Creation of Nanometre-Scale Islands, Wires and Holes on Silicon Surfaces for Microelectronics Ph.D. Thesis V. Palermo University of Bologna March 2003 © All Rights Reserved
Creation of Nanometre-Scale
Islands Wires and Holes on Silicon Surfaces
for Microelectronics
PhD Thesis
V Palermo
University of Bologna
March 2003
copy All Rights Reserved
2 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 3
Universitagrave degli studi di Bologna
Creazione di isole fili e fori nanoscopici su superfici di silicio per microelettronica
Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Tesi di Dottorato di Ricerca Scienze Chimiche
XV ciclo
Presentata da Vincenzo Palermo
RelatoreProf Alberto Ripamonti
Co-Relatore Dott Derek Jones
Coordinatore di Dottorato Prof Goffredo Rosini
Marzo 2003
4 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 5
OMNIA IN MENSURA ET NUMERO ET PONDERE
Sapientiae Salomonis 1120
6 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 7
1 Introduction nanotechnology and the future of computers The motivations of nanotechnology research
In 1965 an electronic engineer named Gordon Moore one of the future founders of Intel
noted that the performance of computers and their complexity doubled every 18 months
and foresaw that computer power would continue to grow exponentially over the following
years
This prevision quite provocative for its time actually came about and gained the name of
ldquoMoorersquos Lawrdquo and continues to hold for the trends of todayrsquos computer industry Since
1965 the number of transistors present in an integrated circuit (IC) has increased from
several hundred to more than ten million and the minimum size of transistor elements has
shrunk from several millimetres to asymp130 nanometres (fig 11)
Devices of such tiny dimensions are actually fabricated using lithographic techniques
where light is passed through an optical mask to react with a photo-sensitive layer (resist)
Fig11 Evolution of the number of transistors present on commercial computers [1]
8 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
on the silicon wafer This resist is then selectively removed and used as a mask for
processing the silicon surface (fig 12 left) The maximum resolution attainable depends
upon the wavelength used and current technology is near to its intrinsic resolution limit
On the other hand there is strong scientific and economic demand for further development
in IC miniaturization to obtain more powerful and complex computers Besides every-day
life applications more powerful computers are fundamental for much scientific research
such as climate change tracking genome sequencing and fluid dynamics Increased
miniaturization is also fundamental for reducing power which has to be dissipated by the
chips which run at progressively higher frequencies Energy consumption by
microelectronic devices is already an issue and represents one of the main obstacles for
the continuing growth in wireless communication (cell phones portable computers CD
and DVD players digital cameras etc)
Thus it is expected that new production methods different from current lithographic ones
will be developed methods which allow modification of a surface well below the 100 nm
limit and even down to single atom manipulation Techniques such as Scanning
Tunnelling Microscopy and Atomic Force Microscopy are already capable of moving
single atoms (see fig 12 right) but unfortunately building a working nanodevice in this
way would take a very long time and these techniques are difficult to apply to large scale
production
Nowadays thousands of researchers are working in the nanotechnology field towards a
new generation of microelectronic devices Several possible solutions are competing for
tomorrowrsquos computer architecture and there is still no clear winner It is likely that the
final solution will be the combined use of different techniques and components (including
molecules nanowires and nanodots) as they become available together with conventional
Fig 12 Left conventional litographic process [2] Right atomic manipulation of iron atoms on copper [3]
VPalermo 9
silicon technology
Below is a brief summary of the most recent developments in nanotechnology and
nanoscience
Actual trends in nanotechnology
Perhaps the most fascinating idea for nanodevice construction is to use one single
molecule working as a complete device The first molecular diodes (ie molecules
conducting current only in one direction) were created in 1997 in 1999 a molecular fuse
and a molecular transistor were demonstrated although there was no possibility of wiring
these devices to external contacts In April 2001 James Heath and his group at UCLA
fabricated an array of overlapping crossbars and placed a small molecule of rotaxane
between each crossbar (fig 13 left) This composite molecule is made up of two
component parts the main rod-like molecular axis and a mobile ring ldquothreadedrdquo on it like a
bead on a necklace and can function as a molecular switch A working 16-bit memory
circuit was constructed using these molecules For a brief review of these works see [4] In
June 2002 a single molecule transistor was built by connecting an organic molecule to two
metal contacts the molecule contained one or two atoms of a transition metal (cobalt or
vanadium) forming the active region of the device supported by an organic backbone [5]
Fig13 Working nanodevices Left schematic representation of rotaxane molecules between crossed nanowires [4] Right SEM image of semiconductor nanowires forming a small circuit [9]
10 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another approach to nanodevice fabrication has become possible through the discovery of
carbon nanotubes which were observed for the first time in 1991 by a Japanese electron
microscopist studying the material deposited during arc-evaporation synthesis of fullerenes
[6] They consist of a graphite-like carbon seamless cylinder with a diameter of several
nanometers and lengths of up to a millimetre Carbon nanotubes are very stable can
behave as metals or semiconductors and can host other molecular or ionic species thus
modifying their electrical behaviour In 2001 Avouris and his group reported the first
circuit made with a single nanotube [7] A few months later Cees Dekker presented a
nanotube-based transistor able to amplify an input signal by a factor of ten and built
several logic circuits using these nanotube transistors [8]
One problem with carbon nanotubes is that it is very difficult to control their electronic
properties ie their metallic or semiconducting behaviour An alternative to carbon
nanotubes are semiconductor nanowires Silicon nanowires can be made using a laser to
vaporize the silicon together with a metal catalyst like iron or gold The vapour condenses
in nanosized drops of silicon and metal from which the wires slowly grow out as more
silicon is adsorbed In 2001 a group at Harvard University [9] created a transistor by
crossing two different nanowires After this the same group arranged four nanowires in a
noughts and crosses grid creating something like a 4-bit memory (fig 13 right) Even
metallic nanowires made of platinum and silver can be used in a crossed configuration to
store information [10]
There are some issues common to all these new technologies though First it is difficult to
imagine these methods applied to large-scale production Up to now the insertion of a
molecule between two electrodes is an occasional lucky event while nanotubes and wires
have to be positioned on the surface creating the appropriate contacts on them manually
The large-scale production of integrated circuits using these building blocks will not be
straightforward and does not seem imminent
Another issue is of an economic and not a scientific or technological nature Since 1965
the cost of IC manufacturing plants has sky-rocketed If the cost of semiconductor
production plants continues to rise exponentially in a few years such plants will cost up to
$20 billion This is a sizeable investment even for large companies such as IBM or Intel
For this reason it is likely that IC companies will resist changing to completely new
technologies closing down their existing plants As it is clear that silicon will remain the
VPalermo 11
fundamental raw material of the IC industry for the foreseeable future nanotechnology
developments for microelectronics will need to be silicon-compatible In these early days
of nanotechnology the most valid approach would seem to be the addition of molecular
functions to existing silicon technology ndash using the latter as a foundation on which to build
Fabrication of self-organised structures on silicon
The possibility of using the phenomenon of atomic or molecular self-organization to create
nanostructures on silicon has already been demonstrated The clean silicon surface shows
in some cases a high degree of order and complex surface reconstruction as will be
described later Several different ordered structures form spontaneously on this surface
such as series of monatomic steps or boundaries between reconstructed areas It has been
demonstrated that it is possible to use these structures to fabricate ordered nanodots and
nanolines on the surface [11] More recently well-defined nanometric patterns have been
obtained with selective etching of silicon using nitric oxide [12]
In this study the possibility of creating different types of nanostructures on the silicon
surface is explored Methods had to be developed which were
- Simple They must not need complex masks or lithographic steps to create the
structure but rather exploit self-organisation phenomena
- High resolution the silicon surface modifications should be on a scalelength of below
100 nm
- Fast billions of nanostructures have to form over the whole surface simultaneously to
be compatible with large-scale production requirements
- Cheap they must not require expensive equipment (such as e-beam lithography) but
exploit simple chemical andor physical treatments to produce nanostructures on the
silicon surface
In Chapter 2 the main characteristics of silicon are described Chapter 3 provides a
summary of the techniques used for this research Chapter 4 examines the chemical etching
of silicon in different liquid environments and the effects of this etching on the surface at a
nanoscopic level with the creation of nanoholes
12 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Chapter 5 describes the growth in ultra-high vacuum (UHV) of nanoscopic voids and
islands on the silicon surface and the effect of surface oxide on this growth
Chapter 6 discusses the modification of silicon surfaces in UHV following the adsorption
of molecules and thermal heating to produce nanoislands and nanolines on silicon
The overall conclusions of our work are summarised in Chapter 7
Finally we will give some conclusions based on the results obtained and discuss possible
applications of the methods developed
Bibliography
[1] From wwwintelcom
[2] From wwwsematechorg
[3] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
[4] Molecules Get Wired Service RF Science 294 (5551) 2442-2443 Dec 21 2001
[5] Coulomb Blockade And The Kondo Effect In Single-Atom Transistors Park J
Pasupathy AN Goldsmith JI Chang C Yaish Y Petta JR Rinkoski M Sethna JP
Abruna HD Mceuen PL Ralph DC Nature 417 (6890) 722-725 Jun 13 2002 Kondo
Resonance In A Single-Molecule Transistor Liang WJ Shores MP Bockrath M Long
JR Park H Nature 417 (6890) 725-729 Jun 13 2002 Nanotechnology - Electronics
And The Single Atom De Franceschi S Kouwenhoven L Nature 417 (6890) 701-702
Jun 13 2002
[6] Smallest Carbon Nanotube Ajayan PM Ijima S Nature 358 (6381) 23-23 Jul 2 1992
[7] Carbon Nanotube Inter- And Intramolecular Logic Gates Derycke V Martel R
Appenzeller J Nano Letters 1 (9) 453-456 Sep 2001
[8] Logic Circuits With Carbon Nanotube Transistors Bachtold A Hadley P Nakanishi T
Dekker C Science 294 (5545) 1317-1320 Nov 2001
VPalermo 13
[9] Logic Gates And Computation From Assembled Nanowire Building Blocks Huang Y
Duan XF Cui Y Lauhon LJ Kim Kh Lieber CM Science 294 (5545) 1313-1317
Nov 9 2001
[10] Formation And Disappearance Of A Nanoscale Silver Cluster Realized By Solid
Electrochemical Reaction Terabe K Nakayama T Hasegawa T Aono M Journal Of
Applied Physics 91 (12) 10110-10114 Jun 15 2002
[11] Fabrication And Integration Of Nanostructures On Si Surfaces Ogino T Hibino H
Homma Y Kobayashi Y Prabhakaran K Sumitomo K Omi H Accounts Of Chemical
Research 32 (5) 447-454 May 1999
[12] Ultrafine And Well-Defined Patterns On Silicon Through Reaction Selectivity
Prabhakaran K Hibino H Ogino T Advanced Materials 14 (19) 1418-1421 Oct 2
2002
14 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 15
2 Silicon surfaces
The name silicon (silicio in Italian) comes from the latin word silex Amorphous silicon
was first isolated by Berzelius in 1824 by reaction of potassium with silicon tetrafluoride
Thirty years later the first crystalline silicon was prepared Silicon makes up 25 of
earthrsquos crust and is the second most abundant element after oxygen Elemental silicon is
not found in nature occurring as silicon oxide (sand quartz amethyst flint etc) or
silicates (asbestos clay mica etc) Perhaps no other element and its compounds has such
a wide range of uses Silicon compounds such as sand and clay are used in the building
industry as refractory materials for high-temperature applications and for enamels and
pottery Silica is the main component of glass silicon carbide is an important abrasive and
silicones are commonly used polymers and lubricants
Here the most interesting use of silicon of course is for the production of
microelectronics devices For this application silicon of high purity (999999) and of
high crystallinity is needed Table 21 lists some of the physical characteristics of silicon
High purity polycrystalline silicon is produced by the reaction of gaseous trichlorosilane
with hydrogen in a furnace Then to prepare a single-crystal of silicon the so-called
Czochralski method is commonly used
Polycrystalline silicon is melted in a quartz furnace at 1415degC in an argon atmosphere
Then a seed of single-crystal silicon is lowered into contact with the melt and slowly
pulled out In this way the crystal grows and a crystalline cylindrical ingot several metres
long is created from the initial seed
After cooling down the ingot is sliced into thin silicon wafers The wafer surfaces are
polished using a counter-rotating lapping machine in an Al2O3 slurry until the surface is
very flat and shiny ready for the lithographic processes
16 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another way to obtain single crystal silicon is the Floating Zone (FZ) method in which a
silicon cylinder is slowly passed through a heating ring The area inside the ring melts and
solidifies smoothly crystallising as it comes out of the ring yielding a single silicon crystal
Microelectronic devices are built on the silicon surface which is the surface of interest
here Unfortunately silicon surfaces are normally quite dirty and uneven at the atomic
scale Atmospheric oxygen and humidity react with silicon surfaces creating a thin layer of
oxide (called ldquonative oxiderdquo) which is usually irregular and full of defects Different kinds
of contaminants also adsorb onto the surface These are usually small organic molecules
and microscopic dust particles A clean surface on exposure to the atmosphere is
completely covered with gas molecules in less than 10-9 seconds If the pressure is
reduced letrsquos say to 10-6 mbar this time increases to 1 second This is the reason why to
study a clean surface we have to work in UHV at pressures below 10-10 mbar
The atoms in the silicon crystal have a diamond-like structure each atom having 4 bonds
in a tetrahedral sp3 arrangement with bond angles of 10947 degrees At the crystal
surface some atoms will have non-bonding orbitals ldquodanglingrdquo in the vacuum ie sp3
orbitals with a lone electron which are highly reactive These orbitals are known as
dangling bonds To minimize surface energy the surface will reorganize by decreasing
the number of dangling bonds
Table 21 Physical data of silicon [1]
Atomic Weight 2809 Lattice constant (A) 543095 Crystal structure Face-centered cubic
(diamond) Melting point 1415 degC
Density (gcm3) 2328 Boiling point 2355degC Atomscm3 50E22 Minority carrier
lifetime (s) 25E-3
Dielectric Constant 119 Specific heat (Jg degC)
07
Breakdown field (Vcm)
~3E5 Thermal conductivity (Wcm degC)
15
Electron affinity x(V)
405 Vapour pressure (Pa) 1 at 1650degC 1E-6 at 900deg C
Energy gap (eV) at 300K
112 Reactivity Inert to acids Attacked by halogens and alkaline
solutions Intrinsic carrier
conc (cm-3) 145E10 Oxidation states +4 -4
Intrinsic Debye Length (microm)
24 Energy of a Si-Si bond (eV)
232
Intrinsic resistivity (Ω-cm)
23E5
VPalermo 17
Dangling bond densities and positions and thus the type of surface reconstruction will
depend upon crystal orientation as well as the temperature and kinetics of the system
Fig 21 shows a drawing of the main faces of a silicon crystal The angle α between any
(11n) face and the (100) face can be calculated from 2cos 2 += nnα The angle
between any (11n) face and the (111) face can be calculated from
)2(3)2(cos 2 ++= nnα
The chemistry and physics of the faces are very different a brief description will be given
for the most important orientations
Si (100)
On the (100) surface each atom has two Si-Si bonds connecting it to the bulk and two
dangling bonds pointing outward Surface energy is reduced by the dimerisation of the
surface atoms through overlap interaction of one dangling bond per atom forming rows of
dimers aligned along the (110) direction This is the well-known ldquo2x1rdquo reconstruction of
this silicon surface The symmetric dimers would make the silicon surface metallic but to
reduce surface stress the dimers tend to buckle and the surface is thus semiconductive It
took several years to understand that the dimers are buckled because at room temperature
Fig 21 Schematic view of the principal orientations of a silicon surface Surface atoms are white bulk atoms are black dangling bonds are gray [2]
18 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
they shift easily from one buckling direction to the other thus appearing symmetric under
STM observation Fig 22 shows an STM image of the 2x1 reconstructed surface
Even almost perfect (100) surfaces have a certain number of monoatomic steps and the
dimer rows on atomic layers are aligned at 90deg to those on adjacent layers Dimer rows are
thus perpendicular or parallel to the step When the dimers on the upper side of the step
are parallel to the step the step is called SA if they are perpendicular the step is called SB
Because of this symmetry SA step edges will be smoother compared to the more broken
fragmented edges of the SB steps
A common defect on the Si(100) surface is the presence of nickel contamination which
appear as missing dimers in STM images This type of contamination is so critical that
even if the silicon sample is only briefly brought into contact with stainless steel tools
(tweezers for example) the 2x1 reconstruction of the surface can be blocked
Silicon atoms can diffuse easily over the silicon surface as monomers and dimers
especially at elevated temperatures The anisotropy due to the 2x1 reconstruction causes a
difference in the diffusion energies of adsorbates over the surface Diffusion of these
silicon species along dimer rows for example will be much easier A list of diffusion
energies for monomers and dimers is given below [2]
Diffusion on Si(100) 2x1 Ed (eV) Monomers along dimer rows 06 Monomers across dimer rows 085 Monomer formation energy 18 Dimer along dimer rows 11 Dimer across dimer rows 15 Dimer formation energy 26 Dimer binding energy 076 Vacancies along dimer rows 17 Vacancies across dimer rows 19
SA
SB
Fig 22 STM image of a 2x1reconstructed silicon surfaceshowing the dimer rows and steps40x35 nm Nickel-induced defectsare visible as dark spots SA and SBsteps are indicated
VPalermo 19
So the diffusion energy for both monomers and dimers is nearly 40 greater if they have
to cross a dimer row This difference reduces to sim10 for vacancy diffusion
Si(113)
The (113) surface can be imaged as a sequence of alternating (100) and (111)-like
structures with two and one dangling bonds on alternate atoms respectively Interest in the
(113) surface is scientific as it has been used to study the energetics of the (100) and (111)
surfaces as well as for surface adsorption experiments
Si(100) surfaces can easily develop (113) facets
Si(111)
This surface besides being the first one imaged with STM with atomic resolution is one of
the most studied because it is the best cleavage face of silicon and because it shows one
of the most complex and elegant reconstructions in surface science
All Si-Si bonds in the silicon crystal are perpendicular to a (111) plane so this face will
have the lowest number of dangling bonds created per unit area In fact each Si atom on a
(111) surface shows a single dangling bond oriented perpendicular to the surface and
bonded to three back atoms These three bonds for each surface atom account for the great
chemical and physical stability of the Si(111) surface Surface energy is 009 eV Aring-2
compared to 015 eV Aring-2 for Si(100)
For energy minimization this surface reconstructs forming a huge 7x7 lattice cell
containing 102 atoms described by the Dimer-Adatom-Stacking fault model (DAS) For a
detailed description of cell structure see fig 23
The cell described by this model is very complex being composed of three kinds of atoms
adatoms rest atoms and corner hole atoms Furthermore a subsurface stacking fault is
present in one half of the cell making the two halves of the unit cell look different under
STM (Fig 24) It took 26 years of research to completely understand the exact structure of
the 7x7 reconstruction
Cleaving a silicon crystal along a (111) plane produces a metastable 2x1 reconstruction
the 7x7 reconstruction is easily obtained by flashing at high temperature in UHV At T gt
20 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
830degC a disordered 1x1 phase covers the surface Cooling down to 800degC leads to the
formation of the 7x7 phase If the cooling process is too rapid small 7x7 domains nucleate
and a disordered 1x1 phase is preserved between domain boundaries
Si(110)
Even though as mentioned before the (111) plane is the favoured cleaving plane of
silicon thin (100) commercial wafers will not break along this plane because the angle
between (100) and (111) is too far from 90deg (see table 22) Instead they will break along
the (110) plane because it is perpendicular to the (100) surface Each surface atom on
Si(110) has a Si-Si bond pointing downward one dangling bond pointing outward and two
Si-Si bonds parallel to the surface in a zig-zag pattern (see fig 21) Cleaved (110)
surfaces are disordered but upon annealing at high temperatures an ordered complex 16x2
Fig23 Scheme of the 7x7 DAS model [2] In each unit cell there are 9 dimers 12 adatoms and a stacking layer fault The force driving this complex reconstruction is the minimization of dangling bonddensity The DAS model shows the lowest number of dangling bonds (19) of all possiblereconstructions 12 dangling bonds are at the adatoms 6 at the rest atoms and 1 at the corner hole atom This surface is metallic
VPalermo 21
reconstruction takes place The surface appears as a series of long ridges and valleys
parallel to each other Eventually tilted facets of orientation (17 15 1) can form on this
surface The adsorption of Ge atoms on this surface leads to the formation of self-
assembled nanowires [3]
Table 22 Angles in degrees between different silicon faces [2]
Orientation
113
110
111
100
100
2524
9000
5474
0
111
2950
3526
0
110
6476
0
113
0
Fig 24 STM image of a Si(111)surface with 7x7 reconstruction A unitcell with its adatoms is highlightedImage size 13x13 nm
22 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] Weast RC Handbook Of Chemistry And Physics (Chemical Rubber Co Cleveland
1972)
[2] Dabrowski J Mussig H Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] The Structure Of Clean And SiGe-Covered Si(110) Surfaces Butz R Luth H Surface
Science 365 (3) 807-816 Oct 1 1996
VPalermo 23
3 STM and other surface analysis
techniques
Scanning Tunneling Microscopy
Since the invention of the optical microscope at the end of the 16th century the possibility
of examining surfaces at higher and higher magnification has fascinated mankind
Development of the technique continued and towards the end of the 19th century optical
microscopes were as good as todays standard instruments The physical limits of the
wavelength of visible light (350-800nm) had been reached
In the 1920s de Broglie showed that electrons can behave like waves and the use of these
particles for imaging with much higher resolution soon followed Atomic resolution using
this technique is only possible in the transmission mode with extremely carefully prepared
samples
In 1982 using the peculiar properties of piezoelectric materials Binnig and Rohrer brought
a metallic tip very very close to a silicon surface and scanned it across an extremly small
area (fig31) The tunneling of electrons from the tip into the sample or vice versa allowed
them to obtain a local density of electronic states (LDOS) map of that surface Although
theory (which treated the extreme point of the tip as a sphere) then excluded the possibility
of atomic resolution following a tip crash into the surface Binnig and Rohrer began to
observe the LDOS with atomic resolution For this discovery and their following work
they were awarded the Nobel Prize in Physics in 1986
24 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The basic principle of STM is very simple A metallic tip is scanned over a surface without
making ohmic contact and a tunneling current passes between the tip and the surface An
electronic circuit keeps this current constant by raising and lowering the tip during the
scan In this way recording the tip height at each point a three-dimensional image of local
density of electronic states (LDOS) of the surface can be obtained To explain the
extremely high resolution attainable by this simple technique quantum theory is needed
According to classical physics the current will flow between sample and tip only if they
are in physical contact If there is a vacuum gap between the two the electrons will simply
remain confined for example within the surface without the possibility of passing into the
tip
In quantum physics however the electrons have a certain probability of passing
(tunneling) across the gap appearing on the other side of the gap in this way reaching the
tip It can be shown that the probability of an electron tunneling through a gap of thickness
z is
kzep 22)0( minusprop ψ h
φmk
2= (1)
where ψ(0) is the electron wavefunction at the surface-gap border m is the electron
mass=91x10-28 g and φ is the work function of the metal (ie the energy required to
remove an electron from that material For silicon it is 48 eV) The tunneling current thus
decays exponentially with z and is extremely sensitive to topographical imperfections
present on the scanned surface A rough formula giving the current as a function of z is [1]
zFS eEVI φρ 0251)( minusprop
Fig 31 Binnig and Rohrer with the first STM Image from IBM [2]
VPalermo 25
where ρs(EF) is the local density of states at the Fermi level on the given surface For
example the formula predicts that for silicon an increase in tip-surface distance of 1 Aring
will give a 95 decrease in tunneling current
This huge dependence of tunneling current upon the distance allows detection even of the
sub-nanometre changes in height given by the single atoms of which the surface is
composed and thus to resolve them in the LDOS images Of course this description of the
tunneling process is oversimplified and for a more accurate one the electronic states of
the tip of the sample and their interaction have to be taken into account Fig 32 shows a
schematic representation of the interaction between tip and sample orbitals
The exponential decay of current with distance also yields high lateral resolution If the tip
is approximated as a sphere of radius R and the current passing at the minimum tip-sample
distance is I0 then the current passing at a lateral distance x from this point will be
Rxk
eII 22
0
2minus
=
Assuming a tip radius of 100 nm the current is concentrated in an area sim15 nm wide at tip
apex
Very sharp tips with even smaller curvature radii can be produced in several ways Simple
Fig32 Schematic view of tip-sampleorbitals interaction a) no interaction b)equilibrium c) sample positive d) tippositive [1]
26 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
tungsten wires mechanically cut are capable of obtaining atomic resolution on graphite in
air but for more disordered and rougher samples sharper and more reproducible tips are
needed
STM tips are mostly made by electrochemically etching a W or Pt-Ir wire The tips we
used were prepared using methods based on the work of Fotino [3]
A tungsten wire 038 nm diameter is immersed in a KOH solution(06M) with a thicker
tungsten wire used as a counterelectrode The cathodic and anodic reactions involved in the
etching are
Cathode 6H2O + 6 e- rarr 3H2(g) + 6 OH-
Anode W(s) + 8 OH- rarr WO42- + 4H2O + 6 e-
A potential of 3V ac is applied to the tungsten and the wire is immersed in the solution
until a constant current of sim100 mA is established The etching rapidly removes metal
shaping the wire end as a sharp tip When the potential reaches 11 V the coarse tip etching
is finished The wire is then removed from the solution carefully inserted into an
insulating plastic tube leaving only the tip exposed and re-immersed in the solution with
the tip pointing upwards A more gentle etching is thus made to reduce tip radius Usually
5 minutes etching at 07 V ac is used In this configuration very small hydrogen bubbles
formed on the tip sides sliding upwards with a ldquohoningrdquo effect on the tip
This procedure yields extremely sharp and reproducible tips at the microscopic level
After the etching the tip is thoroughly rinsed in ultrapure water then dipped into
concentrated HF to remove surface oxides and hydroxides [4] The tip is dried with
nitrogen inserted into the UHV system and degassed overnight at sim150degC
The possibility of measuring sub-nanometric distances would be useless without being able
to control tip movement over such a minute scale To scan the tip over the surface a
piezoelectric scanner is used Piezos are usually made of an alloy of PbZrO3 and PbTiO3 a
material which contracts or expands when a voltage is applied to it The Omicron
instrument used in our laboratory has three such piezo scanners for xy and z tip motion
allowing one to scan the tip over the surface with sub-Aringngstrom precision (fig 33)
To isolate the instrument from ambient vibrations the whole STM stage is suspended upon
four springs which eliminate all frequencies above 1 Hz and surrounded by a crown of
VPalermo 27
copper wings and fixed magnets Parasitic currents generated by the magnets into the
copper wings contrast every movement of the stage and efficiently block stage vibrations
The STM can be used not only to explore surface topography but to measure the IV
characteristics of single atoms or molecules on the surface (Scanning Tunneling
Spectroscopy or STS) Furthermore it can be used to modify the surface with voltage
bursts digging into it or delicately moving single molecules or atoms over a surface [5] It
can be used in vacuum in air and with proper lateral isolation of the tip even in liquids
A major drawback of STM is that it works only on conducting and semiconducting
surfaces and thus cannot be used on many surfaces of biological and chemical interest
Another instrument more suitable for these and other applications is the Atomic Force
Microscope (AFM)
Atomic Force Microscopy
AFM was invented in 1986 by Binnig Quate and Gerber after calculating the possibility of
building a cantilever with a force constant of the same order of magnitude as that of a
chemical bond
In AFM a tip mounted on a microscopic cantilever (usually made of Si3N4 fabricated with
optical lithography) is brought close to a surface When the tip touches the surface the
cantilever is very slightly deflected upwards The movement is measured by observing the
Y-PIEZO
SILICON SAMPLE
Fig 33 A picture of theSTM used for theexperiments The tripodpiezo scanner is shown
Z-PIEZO
X-PIEZO
TIP
28 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
deviation of a laser beam hitting the upper face of the cantilever Fig 34 provides a
schematic view of the principle of AFM
The typical force constant of the cantilever varies from 00006 to 2 Nm the typical
resonance frequency is 3 to 120 kHz The AFM tip can apply a force on the sample of up
to 10-9 N The AFM can be used on conductive or insulating surfaces in vacuum air or
liquids Furthermore the tip can be modified to sense electrostatic potentials (electric force
microscopy) or magnetic fields (magnetic force microscopy) it can even be functionalized
with complex molecules such as proteins to interact with biological surfaces
A drawback of the AFM is that the force it exerts can damage the surface under
observation especially if the sample is soft (as in the case of cell membranes for
example) This problem can be overcome using the instrument in tapping mode (where the
tip does not move laterally during its brief contact with the surface) or in non-contact mode
in which the tip oscillates above the surface during the scan and the changes in its
frequency due to interaction with surface are monitored The shifts in the oscillating
frequency of the cantilever due to tip-sample interaction are then used for imaging the
surface In this mode interaction of the tip with the surface is minimal and soft samples
can be imaged
STM and AFM are the main techniques used for this work A brief description of other
techniques used occasionally is given below
Fig 34 Scheme of an atomic force microscope
VPalermo 29
Low Energy Electron Diffraction (LEED)
Electrons with energies in the 20-500 eV range are diffracted by a crystalline surface the
diffraction peaks are visualized on a fluorescent screen This technique probes the long
range order of the surface up to a depth of several nanometres
X-Ray Photoelectron Spectroscopy (XPS)
XPS allows both qualitative and quantitative chemical analysis of the elements present on
or near the sample surface
An X-ray source is used to photoionize the atoms on a surface and produce photoelectrons
By measuring the kinetic energy of the photoelectrons the binding energy of the electronic
levels can be calculated This energy will depend on the chemical environment of the
surface atoms
Although the soft X-rays used penetrate to a depth of ~2000Aring the sampling depth of the
technique is determined by the mean free path of the photoelectrons which allows their
escape from only the first 10-100Aring
Secondary Ion Mass Spectroscopy (SIMS)
High and low energy ions (primary ions) are used to bombard a sample and remove surface
atoms and ions The ionic fragments removed (secondary ions) are then analysed by a mass
spectrometer The surface can be consumed during the measurement and profiles obtained
giving concentrations of the materials composing the sample at different depths (depth
profiles)
A popular variant of SIMS is TOF-SIMS In this technique the secondary ion masses are
measured by a time-of-flight (TOF) measurement The secondary ions generated by the
bombarding primary ions are accelerated to a constant kinetic energy and then move
through a field-free space before they reach the detector where their intensity is measured
as a function of flight time Since ions with different masses have different velocities at a
30 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
given kinetic energy the measured flight times of the ions can easily be converted to their
masses The static nature of this latter technique allows mass spectroscopy surface analysis
with minimal damage to the surface
Bibliography
[1] Chen CJ Introduction To Scanning Tunneling Microscopy (Oxford University Press
Oxford 1993)
[2] From wwwibmcom
[3] Tip Sharpening By Normal And Reverse Electrochemical Etching Fotino M Review
Of Scientific Instruments 64 (1) 159-167 Jan 1993
[4] A Convenient Method For Removing Surface Oxides From Tungsten STM Tips
Hockett LA Creager SE Review Of Scientific Instruments 64 (1) 263-264 Jan 1993
[5] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
VPalermo 31
4 Surface modification of silicon in liquid
Nano-hole creation
Liquid treatments of silicon wafers are very common in the integrated circuit (IC)
manufacturing industry They are used to clean and improve surface uniformity to create
and etch protective oxide layers and to remove photo-resist layers
Crystalline silicon with its native oxide layer is very stable and is resistant to many acids
It is easily attacked by hydrofluoric acid (HF) and alkaline solutions
The thin (~2 nm) passivating layer of native oxide (SiO2) is formed on exposure to the
atmosphere This surface layer contains many defects and contaminants so it is usually
chemically stripped and substituted with a better chemically-formed protective oxide
The most common silicon cleaning procedure is the RCA method named after the Radio
Corporation of America [1] It consists of two steps in the first one the surface is treated
with a hot alkaline solution (H2OH2O2NH4OH 411) to remove particles from the
surface following this a hot acidic solution (H2OH2O2HCl 411) is used to remove
metal contamination Other well-known cleaning methods are IMEC (a sequence of
cleaning steps in H2OO3 and dilute HF) or the Pirana etch (a hot 41 mixture of H2SO4
H2O2)
32 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The standard RCA clean removes surface contaminants etches the native oxide and
oxidizes the silicon surface leaving a uniform layer of silicon oxide which better protects
the surface from further contamination
Etching with fluorine-based solutions
Hydrofluoric acid is one of the most common reagents used in the treatment of silicon
wafers both in the research field and in industrial processes A rapid dip in dilute HF is the
simplest way to remove the native oxide from Si(100) and leaves the surface passivated by
a layer of Si-H bonds Because of the low polarization of Si-H bonds the Si-H layer is
stable even for several days protecting the surface from contamination It has often been
assumed that this short etch does not significantly change the surface morphology of the
silicon substrate[2] even though a prolonged dip in dilute HF leads to surface roughening
[3]
Although dilute HF roughens the Si(100) surface at the atomic scale [45] immersion in
concentrated HF (49) etches the surface oxide without attacking the Si surface
uncovering in this way the buried SiSiO2 interface The final counter-intuitive result is
that dilute HF etches the silicon while concentrated HF leaves the crystalline silicon
untouched [3]
Etching Si with fluorine-containing solutions at different concentrations and pH can
produce different morphologies from rough surfaces to flat nearly ideal Si-H terminated
surfaces
Hessel et al and Higashi et al demonstrated in 1991 that very flat Si(111) surfaces can be
obtained using 40 NH4F while etching with HF always yields rough surfaces The
surface becomes smoother because the etchant rapidly attacks Si atoms at step borders
thus removing surface kinks and irregularities in a step-flow mechanism [6 7] Later on
even smoother and more perfect surfaces were obtained by removing oxygen from the
solution after it was discovered that oxygen dissolved in 40 NH4F initiates the formation
of triangular etch pits It was not possible to obtain flat surfaces by etching Si(100) with
ammonium fluoride solutions which leads to the formation of small 2x1 dimer-row
reconstructed (100) terraces together with (111) facets [8]
This difference is caused by the different hydride terminations prevailing on the (100) and
(111) faces While the ideal Si(111)-H surface is monohydride terminated the more
VPalermo 33
reactive dihydrides predominate on the Si(100)-H surface making it more vulnerable to
etching The etching reaction is thus strongly anisotropic etching (100) facets faster than
(111) thus producing (111) microfaceting on Si(100) crystals
A more uniform Si(100) surface can however be prepared by etching at low pH with an
HFHCl mixture [9] or by using very dilute HF solutions and ultrapure water with low
dissolved oxygen and carbon contents [10]
Electrochemical etching can also be used applying anodic or cathodic bias to the silicon
to obtain different morphologies [11] by varying the potential isotropic or anisotropic
etching is observed The aforementioned results show that despite the simplicity of the
reactants fluoride etching of silicon is quite a complex reaction
Fig 41 Chemical etching of silicon
HO+H
H HO
H
H
H
FSi
Si
SiSi
+H2O
-OH-
+F-
H2O H H
Si H
F OH
H Si
Si Si
-H2
+H2O
H
H
H
FSi
Si
SiSi
H
H
H
OHSi
Si
SiSi
H
H
H
H Si
Si
Si Si
Etching mechanism of silicon
HF rapidly dissolves the SiO2 passivating layer on silicon leaving the surface almost
completely hydrogenated [12] After this two different types of reactions etch the silicon
simultaneously one chemical and the other electrochemical [13] The overall etching
mechanism can be schematized in two stages (see Figure 41)
i) Si-H bonds are substituted by Si-F or Si-OH bonds creating a partial charge on the
surface silicon atom and polarizing its Si-Si backbonds
ii) these polarised backbonds are then more easily attacked by HF or H2O After
rupture of the Si-Si bond the atom is removed leaving behind new Si-H
terminations and the reaction can start again
34 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
These reactions take place although at different rates on both Si(100) and Si(111)
Stage i) is usually the rate-determining step of the reaction and the stability of the Si-H
bonds depends upon the pH the concentration of nucleophilic species in solution and an
eventual potential applied to the crystal
For pH gt5 as in the case of concentrated NH4F solutions reaction begins with attack by
water to give Si-OH (step ArarrC) The -OH group is rapidly substituted by fluorine with
polarization of the underlying Si-Si bonds These bonds are then easily attacked by water
the silicon atom being released into solution as HSiF(OH)2 The Si-OH rarr Si-F substitution
is not fundamental for the reaction and etching can proceed even for Si-OH terminated
atoms but XPS measurements showed the presence of a certain number of Si-F bonds
remaining Furthermore fluorine seems to have a catalytic effect on Si-H substitution as
indicated by the dependence of the etch rate upon the F- concentration at least for pH
values between 4 and 8
Si-F bonds can be easily removed by a water rinse In the case of strongly alkaline
solutions (pH=14) OH- groups act directly as nucleophiles and no fluorine is needed to
catalyze Si-H bond rupture
At pH lt4 almost no free F- ions are present in solution and the etch rate is very small at
pH lt2 all etching reactions are very slow and this explains the stability of Si crystals in
concentrated (50 ww) HF solutions
Si Si
Si Si
H
H
H
H Si
Si
Si Si
H
H
H Si
Si
SiSi
H
H
H
OH-H+ -e
(F-)
-H+
-e
+H2O hellipas in fig1
Fig 42 Electrochemical etching of silicon by H2O
The electrochemical etching of silicon involves electron transfer from the surface atoms to
the valence or conduction band of the crystal (according to the type of doping of the
silicon) Several studies have been carried out at different electric potentials In the case of
p-silicon or n-silicon at anodic potentials where positive charges (h+) are available in the
crystal a silicon-centred radical can be created by capture of a hole from the bulk (fig
42) The F- ion does not participate in the reaction but is thought to contribute through
VPalermo 35
electrostatic interactions by lowering the energy of the interaction step After the
formation of the Si-OH group the reaction proceeds as shown in the scheme of fig 41
Matsumura et al [4] proposed that not only water but HF2- molecules also play a major role
in electrochemical etching of silicon leaving on the surface Si-F terminated bonds which
can be immediately attacked in an autocatalytic process (fig 43)
In the electrochemical reactions described above an external potential is applied to the
silicon crystal The chemical and electrochemical reactions in any case take place
simultaneously most of the time with the chemical path predominating at high pH Even
when no external potential is applied to the silicon partial electrochemical reactions can
take place at different ldquocathodicrdquo and ldquoanodicrdquo sites on the surface with an internal charge
exchange which ensures neutrality [11] This macroscopic silicon etching and hydrogen
bubble formation can lead to visually observable patterns on the surface when Si(100) is
immersed in ammonium fluoride even without applying a potential
Si Si
Si Si
H
H
H
F Si
Si
Si SiH
F F-H+ -2e
+HF2-
H
HF
SiSi
SiSi + F
F F
FH
H
H
F -H+ -2e
+HF2-
Fig 43 Autocatalytic electrochemical etching of silicon by HF2-
Inhomogeneities on silicon surfaces caused by electrochemical reactions and charge
transfer have been studied extensively because they are of fundamental importance in the
formation of porous silicon
Pore formation on silicon
When Si(100) or Si(111) are etched under anodic bias in fluorine-based solutions
microscopic pores form on their surface Several different morphologies of pores have
been observed with pore diameters ranging from 10 nm to several microns with depths of
several microns [14] Pore shape is very variable too ranging from ordered straight pores
to chaotic networks of branched pores (fig 44) Porous silicon has been known since the
fifties but it was only in 1990 that interest in this material increased following the
discovery that porous silicon layers were able to emit bright red light This led to a large
amount of research and now different classes of micropores can be reproducibly created
36 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Fig 44 Different types of Silicon micropores From ref [14]
mostly for optical and micromachining applications However there is still no unified
theory able to explain the nucleation and growth mechanism of all the different kinds of
pores
We will give a short description of some of these theories for more detail see Parkhutik et
al [15]
One model explains pore nucleation on the basis of physical processes such as hole
positive charge migration ion transport to the surface and small perturbations on the
silicon surface modelled as Fourier components The system is shown to be unstable and
some spatial frequencies that lead to pore nucleation evolve from the etching process
A second model focusses on stationary pore growth without explaining the nucleation
stage According to this model silicon dissolves preferentially at pore edges because h+
charges are attracted by the stronger electric field present at these edges
A third class of models explains pore growth as a Diffusion Limited Aggregation (DLA)
process where the random walk of h+ charge carriers through the depleted layer present at
the silicon-liquid interface controls pore shape
Finally the model by Carstensen Cristophersen and Foll [16] proposes that areas of the
surface of some characteristic size LCO are etched by synchronized ldquocurrent burstsrdquo in the
flow of h+ charges These bursts dissolve silicon through cyclic stages of surface oxidation
oxide removal and hydrogen passivation Areas where a burst has recently taken place are
less passivated and thus more likely to be etched again in this way the pore bottom
continues to dissolve while the pore walls are passivated and are thus less favourable areas
towards current bursts
VPalermo 37
EXPERIMENTAL RESULTS
In the following sections we will show some experimental results obtained from STM and
AFM measurements of fluorine-treated Si(100) surfaces In the first part the results of
mild etching using concentrated and dilute HF solutions at low pH are presented In the
second part the results of etching at high pH using ammonium fluoride are presented and
the mechanism of pore formation discussed
Etching of Si(100) in dilute and concentrated HF
Samples were cut from different areas of an 8-inch diameter p-doped silicon(100) wafer
(10 Ω-cm) supplied by MEMC Electronic Materials Each series of STM measurements
was carried out over at least six different areas on at least two identical samples Low
Electron Energy Diffraction (LEED) was used to check the surface cleanliness of the
samples before STM measurements
Table 41 summarizes the different treatments of each sample After etching with
electronic grade HF each sample underwent a final rinse in Ultra-Pure Water (UPW
resistivity gt18 MΩ-cm) Both the HF and the UPW were allowed to flow continuously
over the sample surface Some samples were not etched with HF at all but just washed with
UPW to observe the morphology of the native oxide layer (~2 nm thick) covering the
surface All of these processes were carried out under nitrogen to limit reoxidation and the
samples were then introduced from the nitrogen atmosphere directly into the vacuum
chamber and degassed overnight at sim150degC before LEED and STM measurements
STM images were obtained from each sample using the same measurement parameters
(sample bias 4 V feedback current 1 nA scan speed 800 nm s-1) These parameters
38 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
provided a satisfactory level of reproducibility for all the samples Measurements were
made over an area of 500x500 nm (image size 500x500 pixel) Slope correction was
carried out by subtracting row-wise and column-wise fitted slopes from the entire image
which gave better results than the simple subtraction of a fitted plane especially for the
rougher samples Following slope correction the rms roughness
sum minus=xy
hyxhN
22 ))((1σ
and the 2-D Fourier transform
)(22
2
)(4
)( vyuxi
xyeyxhvuF +sum∆
= π
π
were calculated for each image where N2 is the number of pixels composing the image
h(xy) is the surface height at each point ∆ is the distance between points h is the mean
height and u v are the spatial frequencies The radial power spectrum PS(f) of the STM image data is obtained from the angular
average of the squared Fourier transform with f 2 = u2 + v2
Fig 45 shows the STM images obtained from the various samples Sample A still covered
with its native oxide layer shows an irregular surface with RMS roughness of ~05 nm
(see Table 41) Observing the sample with LEED gave no diffraction pattern even at
relatively high incident electron energies because of the surface oxide coverage After 1
min etching in dilute HF (sample B) the morphology is similar to the original one though
Table 41 Sample treatments average RMS roughness and integrated area of power spectra
Sample Treatment RMS roughness (nm)
PS area (f lt 01 nm-1)
PS area (f gt 01 nm-1)
A Rapid dip in water 053 plusmn 013 366 032
B 1 min in HF 5 + 10 min in water 051 plusmn 008 272 034
C 30 min in HF 5 + 10 min in water 062 plusmn 008 1064 038
D 5 sec in HF 49 + 10 min in water 042 plusmn 004 183 022
VPalermo 39
Fig 45 STM images of each group of samples showing the topography of the silicon surface A) noetching original oxide surface B) after 1 min etching in dilute HF C) after 30 min etching in diluteHF D) after dipping in concentrated HF Grey scale indicates height of the surface from lower (black) to higher (white) The images are 250x250 nm ie representative portions of the images usedfor the roughness measurement and PSD analysis
some of the larger features have disappeared and the image quality is better maybe due to
improved tunnelling due to the cleaner surface The RMS roughness is comparable to that
of the original surface Clear diffraction patterns are visible using LEED though at quite
high energies (200 eV) After prolonged etching (sample C) the RMS roughness increases
to 062 nm and a long-range corrugation is visible on the surface even if the LEED pattern
is good
The samples dipped in concentrated HF (D) reveal the bare SiSiO2 interface which has a
disordered aspect and protrusions over a wide range of dimensions The quality of the
STM images of sample D is very good probably due to the cleanliness of the surface
40 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 E -0 2
1 E -0 1
1 E + 0 0
1 E + 0 1
1 E + 0 2
1 E + 0 3
0 0 0 0 0 1 0 1 0 1 0 0f (1 nm )
nm^4
A a s re c e iv e d
B e tc h e d 1 min
C e tc h e d 3 0 min
D e tc h e d HF 4 9
Fig 46 Log-Log plot of the averaged power spectra of the STM images for all the samples
which gives a more stable tunnelling junction The LEED pattern is excellent showing
clear diffraction peaks at energies as low as 37 eV comparable to that obtained after high
temperature cleaning in UHV
Fig 46 shows the power spectra of the samples The high frequency and low frequency
areas of the power specturm are considered separately Table 41 shows for each sample
together with the roughness the integrated area of the power spectrum for the high and low
frequency part
We first examine the differences between the samples in the low frequency part of the
spectrum (lt01 nm-1) In this spatial range the short dip in concentrated or dilute HF
(samples BD) seems to lower the roughness of the sample removing some of the larger
features of the native oxide surface visible in Fig 45a and thus decreasing the
corresponding part of the power spectrum Sample C however etched with HF for 30 min
shows an increase in roughness peaking at 0012 nm-1 (~85 nm corresponding to the
typical dimensions of the corrugations visible in Fig 45c) On this scale sample C has a
PS density of 435 nm4 compared to 101 nm4 for sample A
Samples A B and C have the same PS in the high frequency range showing that the fine
structure of the surface is not changed by the HF etching Sample D on the other hand
shows a decrease of roughness for frequencies gt 015 nm-1 This effect could be due to the
VPalermo 41
improved surface cleanliness after etching with concentrated HF which would give a more
stable STM junction thus reducing the high frequency noise in the image
Fig 45 and the analysis of the power spectrum of each sample shows that a rapid dip in
HF removes the native oxide but does not lead to major changes in the morphology of the
Si surface its only effect being the removal of some of the larger features present on the
original surface Prolonged etching on the other hand increases the RMS roughness of the
surface
Etching of Si(100) in concentrated ammonium fluoride and nano-hole creation
Two different types of commercial p-doped Si(100) wafers (2Ω-cm and 10Ω-cm) from
MEMC were used Several different samples of 10x5 mm were immersed for 10 minutes
in 40 electronic grade NH4F solution under agitation Previous works used low
temperatures or anodic potentials applied to the silicon to avoid gaseous hydrogen
production and to obtain a uniform surface but in our experiment we worked at room
temperature to check the influence of hydrogen bubbles on surface morphology During
the etching the stirring was sufficient to provide a uniform concentration of reagents over
the whole sample surface but not to mechanically remove the hydrogen bubbles from the
silicon surface
After the etching each sample was rinsed in ultra-pure water to remove any etching
residues and observed with STM AFM and optical microscopy The AFM measurements
were made in air while for STM measurements the samples were rapidly dried with
nitrogen and inserted into the vacuum system to avoid surface reoxidation After insertion
into the vacuum surface cleanliness was checked with LEED and the surface morphology
observed by STM Parameters for STM measurements were sample bias +4 V 1 nA
current 16 Hz scan rate The images obtained were stable and reproducible over several
days Scan parameters for AFM were 20 nN force and 1 Hz scan rate
Some of the samples were cleaned with an RCA standard clean [1] before NH4F etching to
check the influence of possible surface contaminants on the final results Eight different
samples were prepared and more than sixty STM images of the samples were taken at
different points of the various samples
42 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
(110)
(110)
Fig47 abc) STMtopographic images of differentetching morphologies Eachimage is 500x500 nm Z-ranges are 10 10 and 18 nmrespectively d) STM image of a bridgecreated by etching of the lowerlayers of silicon (black arrow)Image is 250x250x6 nm
After ~2 min of immersion in the solution hydrogen bubbles become visible on the sample
surface The production is slow and the bubbles are quite stable on the sample without
detaching Thus some areas of the surface are masked from the liquid etching action
STM observations (fig 47) show that at the nanometer scale the surface is unevenly
covered with holes of radii ranging from 10 to 200 nm with depths of 2-4 nm These holes
have a wide range of different shapes and distributions In most cases the surface was
covered with a uniform distribution of round-shaped holes (fig47a) indicating isotropic
etching The dimension and the density of the holes changed greatly from sample to
sample and even over the surface of a single sample In some cases the etching was
anisotropic yielding nearly square holes and layered structures as shown in fig 47b
Square holes have been previously observed in cases where the etching speed in the (110)
direction is significantly smaller than in the (100) direction [17]
Over these areas (fig 47b and especially 47c) it is clearly visible how once the surface
had been attacked the reaction continued to preferentially remove atoms at step
irregularities (kink atoms) straightening step edges Eventually the exposed underlying
silicon was also attacked and further holes created inside the previously etched larger
ones It was not possible to detect monatomic steps on this kind of surface The smallest
step height observed was ~15 nm corresponding to several atomic layers In the image
shown in Figure 47a the etching was not strong and created only anisotropic holes on the
surface In fig 47b and c the stronger etch proceeded laterally for several tens of
VPalermo 43
nanometres leaving straight steps several tens of nanometers long and roughly rectangular
holes as expected given the structure of the (100) crystal face In some cases a
significative underetch is observed and the formation of suspended bridges and tunnels
can be deduced from the STM images (fig 47d)
The formation of branched pores and suspended structures has been attributed during pore
formation to diffusion limited aggregation effects where the h+ charge carriers necessary
for silicon etching have a higher probability of reacting at pore bottoms than reaching the
upper part of the silicon surface In the case of very deep pores quantum wire effects have
been invoked to explain the pore growth mechanism [15] In our case though the pores
formed were very shallow the underetch depth being only a few nanometres on pores of
sim100 nm width Thus more than diffusion effects the main contribution to the
underetching process must come from anisotropic etching and some kind of autocatalytic
reaction path analogous to the one described by Matsumura et al [4] with some areas of
the silicon surface hydrogenated and thus less vulnerable to etching
Pre-treatment with RCA cleaning has no effect on the final morphology and this seems to
exclude pore nucleation being caused by presence of metallic or organic surface
contaminants
The morphology and the distribution density of the pits was quite uniform over
microscopic areas of the sample but changes were observed over the millimetre scale This
suggests that etching intensity is influenced by some large-scale parameter
Large-area measurements made with AFM or with an optical microscope (fig 48)
showed that the inhomogeneity of surface etching can be correlated with the masking
action of the bubbles While the fluoride dissolved the silicon hydrogen bubbles formed
by the reaction covered some areas of the surface thus blocking the etching over that area
generating macroscopic steps at the bubble-liquid border As the reaction proceeded more
hydrogen accumulated and the bubble diameter increased producing in this way a circular
pattern of steps The increase in bubble diameter was not continuous with time otherwise a
uniform surface slope gradient would have been obtained The formation of this circular
ldquoetching staircaserdquo indicates that the bubble growth was stepwise the bubble accumulating
more and more hydrogen without enlarging across the surface until it relaxed increasing
its diameter stepwise and covering more silicon The circular structures in fig 48a are not
co-axial and their asymmetry could derive from physical processes due to stirring or
irregularities on the surface
44 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
a b Fig 48 a) optical micrograph of etching patterns on Si(100) created by NH4F 12x09 mm b) AFM image of the circles border xy range is 40x24 microm z-range is 30 nm
The step structure was not destroyed by the etching even after the bubble detached from
the surface but on the contrary the etching process seemed to be influenced by the
presence of the step
Observing in detail a series of steps (fig 48b) a quite deep trench is visible at the base of
each step A close-up image of a step and the corresponding line profile of the trench is
shown in fig 49 The trench is sim5 nm deep with respect to the lower surface compared to
a step height of 22 nm
A similar structure has been recently obtained with electrochemical etching of p-type
Si(100) in 4 HF [18] in which a ldquocurrent burstrdquo etching model previously described
was assumed for silicon dissolution In that experiment the trench was created at the
border of silicon nitride masks and began to grow after a nucleation stage Preferential
trench etching was along the (110) direction and the trench growth was explained as an
effect of mechanical stress induced by the nitride mask and of electric field enhanced
dissolution which depended upon an external applied potential
While it is clear that in our system the gas bubbles have a masking effect similar to a
classical solid nitride mask it is unlikely that hydrogen present on the surface can induce a
significant stress in the silicon lattice as in the case of a nitride mask Furthermore no
external field was applied to drive preferential etching at the trench site
It has been proposed [19] that the cathodic and anodic part of the etching reaction
(hydrogen production and silicon oxidative etching respectively) take place at different
points on the surface with a net charge transfer between the different areas In this case
the highest reaction rates will correspond to the silicon area surrounding the bubble border
where a high number of positive charges will be available for the reaction Furthermore a
sharp trench extending into the silicon crystal will be a preferential electrostatic attractor
VPalermo 45
Fig 49 AFM image of the etched surface showing a step created on the surface by bubblemasking A stronger etching action is visible on the right side of the step as well as theprotected area on the upper side of the step (indicated by the arrows) Image is 10x10 micromz-range is 30 nm The profile on the right is taken from the central area of the image
for the h+ charge carriers coming from other ldquocathodic areasrdquo of the sample either from
other regions on the surface or from the back of the silicon chip [18]
In the areas where the hydrogen bubble had detached and the surface was exposed to the
etching the reaction was not uniform in the neighbourhood of the steps It is possible to
observe (fig 48b and 49) an area on the upper side of the step where less or even no
etching at all seems to have taken place as if the step was able to protect the surface from
etching While etching on the lower side with trench formation can be attributed to the
presence of the bubble the surface on the upper step side can be etched only after bubble
detachment so no masking effect can account for this result However a further
preferential attraction of h+ charge carriers from the already formed trench can be
hypothised electrochemically shielding the surrounding area from further etching If this is
true the shielding effect would be very strong with a relatively shallow 5 nm-deep trench
protecting an area of sim1 microm parallel to the step
To summarise the etching of Si(100) in NH4F is a complex process in which different
reaction paths both chemical and electrochemical co-exist Hydrogen bubbles formed by
the reaction act as a mask on the surface and create etching paths and inhomogeneous
etching of the surface Different kinds of pores are observed on the surface and in some
cases the anisotropy of the process is so strong as to give square-shaped holes and
underetching
46 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The diffusion of h+ charge carriers in the crystal is one of the main rate-determining steps
of the reaction and leads to the formation of a deep trench immediately outside the bubble
perimeter These trenches act as charge collectors and reduce the etching of the upper step
surface in the proximity of the steps
Bibliography
[1] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[2] Spectroscopic Ellipsometry Studies Of HF Treated Si (100) Surfaces Yao H Woollam
Ja Alterovitz SA Applied Physics Letters 62 (25) 3324-3326 Jun 21 1993 Influence
Of HF-H2O2 Treatment On Si(100) And Si(111) Surfaces Graf D Bauermayer S
Schnegg A Journal Of Applied Physics 74 (3) 1679-1683 Aug 1 1993 Kinetics Of
Oxidation On Hydrogen-Terminated Si(100) And (111) Surfaces Stored In Air Miura
T Niwano M Shoji D Miyamoto N Journal Of Applied Physics 79 (8) 4373-4380
Part 1 Apr 15 1996
[3] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995 Structure Of The Stepped SiSiO2 Interface After Thermal-Oxidation -
Investigations With Scanning Tunneling Microscopy And Spot-Profile Analysis Of
Low-Energy Electron-Diffraction Pietsch GJ Kohler U Jusko O Henzler M Hahn
PO Applied Physics Letters 60 (11) 1321-1323 Mar 16 1992
[4] Enhanced Etching Rate Of Silicon In Fluoride Containing Solutions At pH 64
Matsumura M Fukidome H Journal Of The Electrochemical Society 143 (8) 2683-
2686 Aug 1996
[5] A Study Comparing Measurements Of Roughness Of Silicon And SiO2 Surfaces And
Interfaces Using Scanning Probe Microscopy And Neutron Reflectivity Crossley A
Sofield CJ Goff JP Lake ACI Hutchings MT Menelle A Journal Of Non-Crystalline
Solids 187 221-226 Jul 1995
VPalermo 47
[6] Step-Flow Mechanism Versus Pit Corrosion - Scanning-Tunneling Microscopy
Observations On Wet Etching Of Si(111) By Hf Solutions Hessel HE Feltz A Reiter
M Memmert U Behm RJ Chemical Physics Letters 186 (2-3) 275-280 Nov 8 1991
[7] Comparison Of Si(111) Surfaces Prepared Using Aqueous-Solutions Of NH4F Versus
HF Higashi GS Becker RS Chabal YJ Becker AJ Applied Physics Letters 58 (15)
1656-1658 Apr 15 1991
[8] Wet Chemical Etching Of Si(100) Surfaces In Concentrated NH4F Solution -
Formation Of (2x1)H Reconstructed Si(100) Terraces Versus (111) Faceting Neuwald
U Hessel HE Feltz A Memmert U Behm RJ Surface Science 296 (1) L8-L14 Oct
10 1993
[9] Ideal Hydrogen Termination Of Si(001) Surface By Wet-Chemical Preparation Morita
Y Tokumoto H Applied Physics Letters 67 (18) 2654-2656 Oct 30 1995
[10] Atomic Structures Of Hydrogen-Terminated Si(001) Surfaces After Wet Cleaning
By Scanning Tunneling Microscopy Endo K Arima K Kataoka T Oshikane Y Inoue
H Mori Y Applied Physics Letters 73 (13) 1853-1855 Sep 28 1998
[11] On The Potential-Dependent Etching Of Si(111) In Aqueous NH4F Solution
Houbertz R Memmert U Behm RJ Surface Science 396 (1-3) 198-211 Jan 20 1998
[12] Etching Process Of SiO2 By HF Molecules Hoshino T Nishioka Y Journal Of
Chemical Physics 111 (5) 2109-2114 Aug 1 1999
[13] Etching Mechanism And Atomic-Structure Of H-Si(111) Surfaces Prepared In
NH4F Allongue P Kieling V Gerischer H Electrochimica Acta 40 (10) 1353-1360
Jul 1995
[14] Pore Formation Mechanisms For The Si-HF System Carstensen J Christophersen
M Foll H Materials Science And Engineering B-Solid State Materials For Advanced
Technology 69 23-28 Sp Iss Si Jan 19 2000
[15] Porous Silicon - Mechanisms Of Growth And Applications Parkhutik V Solid-
State Electronics 43 (6) 1121-1141 Jun 1999
[16] Formation And Application Of Porous Silicon Foll H Christophersen M
Carstensen J Hasse G Materials Science amp Engineering R-Reports 39 (4) 93-141 Nov
1 2002
[17] Pore Morphology And The Mechanism Of Pore Formation In N-Type Silicon
Searson PC Macaulay JM Ross FM Journal Of Applied Physics 72 (1) 253-258 Jul 1
1992
48 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
[18] Deep Electrochemical Trench Etching With Organic Hydrofluoric Electrolytes
Christophersen M Merz P Quenzer J Carstensen J Foll H Sensors And Actuators A-
Physical 88 (3) 241-246 Mar 5 2001
[19] Morphology Of Anodically Etched Si(111) Surfaces - A Structural Comparison Of
NH4F Versus HF Etching Houbertz R Memmert U Behm RJ Journal Of Vacuum
Science amp Technology B 12 (6) 3145-3148 Nov-Dec 1994
VPalermo 49
5 Surface modification of silicon in vacuum void creation and oxide desorption
The main reason for the huge success of silicon in the microelectronics industry is not due
to its superior properties as a semiconductor Other materials for example germanium
have better qualities such as higher mobility of charge carriers and lower noise levels
which would allow the construction of faster and higher performance devices
The widespread use of silicon however is mainly due to the outstanding characteristics of
its oxide Silicon dioxide (SiO2) is a very good electrical insulator easy to form
chemically and thermally stable and is compatible with lithographic and metal deposition
processes Germanium oxide on the contrary is too reactive to be used
Even the use of Si(100) substrates for nearly all microelectronic devices is dictated by
oxide quality The (111) face of silicon crystal can be easily cleaved and flattened and
almost atomically perfect surfaces can be obtained with simple chemical procedures (as
described above) But the density of interfacial defects is highest for Si(111)-SiO2
interfaces and lowest for Si(100)-SiO2 ones so microchips will continue to be fabricated
on Si(100) wafers
SiO2 (silica) is present in 95 of the earthrsquos minerals in different allotropic forms such as
quartz tridymite and cristobalite In the bulk each silicon atom is bonded to four oxygens
in a Si-O-Si tri-dimensional network Si-O bonds are 016 nm long and form an angle
ranging from 120deg to 150deg
Three typical intrinsic defects are present in SiO2 The so-called Ersquo centres are oxygen
vacancies with a hole localised on a silicon atom with only three Si-O bonds
50 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
O3Si +SiO3 Whereas the PR (peroxy radical) defects are holes trapped by a charged
peroxy moiety with a O3Si-O-O+ SiO3 structure The NBOHC (non-bridging oxygen hole
centres) derive from water or hydrogen contamination and are schematized as O3Si- O- H-
O-SiO3
The atomic structure of the Si-SiO2 interface varies enormously Local domains resembling
the tridymite and the cristobalite structure of silica are present but it seems that only 10
of the interface is ordered [1] Far from the interface the SiO2 bulk is completely
disordered The passage from bulk Si to stoichiometric SiO2 passes through a non-
stoichiometric SiOx layer 07 nm thick
When a clean silicon surface is exposed to atmospheric oxygen a thin sim2 nm thick layer
of native oxide forms spontaneously which is usually removed and substituted with
thicker better quality oxide layers before further processing
Silicon is usually oxidised by thermal annealing at temperatures between 800deg and 1100deg
in an atmosphere of pure O2 with some water eventually added to increase oxidation speed
Thermal oxides made in pure oxygen (dry oxides) grow more slowly than oxides produced
in an oxygen-water atmosphere (wet oxides) but are usually of better quality
According to the Deal-Groove formula the time t needed to grow an oxide of thickness X
is given by [1]
1
212minus
minusminus
+=
ABXBXt α
where the constant B and BA decrease exponentially with temperature as
minus
kTEexp
with activation energies for dry oxidation of EB =123 and EBA =20 eV respectively
EB is related to the diffusion of oxygen in silicon while the value of EBA is interpreted as
the energy required to break a Si-Si bond The exponent α is 1 for wet oxidation and 0 for
oxidation at high temperatures and low oxygen pressures It has intermediate values for dry
oxidation This formula does not work well for low values of X at the initial stages of
oxidation and usually empirical corrections are used
An interesting characteristic of silicon is that at high temperature and in vacuum oxygen
can actually etch the silicon crystal giving gaseous products and the oxidized layer present
on the crystal becomes unstable (fig 51) [2]
VPalermo 51
1E-10
1E-09
1E-08
1E-07
1E-06
1E-05
1E-04
1E-03
1E-02
1E-01
0607080911112
1000T (1K)
P (T
orr)
SiO2 + Si rarr 2SiO(g) Oxide decomposition
Si+O2 rarr SiO2(s) oxide formation Fig51 phase diagram of the
oxygen-silicon system
Silicon oxidation apart from the initial nucleation stages at the monolayer level proceeds
uniformly over the whole surface with a planar reaction front moving from the surface
into the bulk
If heated under low oxygen partial pressure (vacuum or inert atmosphere) SiO2 is known to
decompose following the reaction
SiO2 + Si rarr 2SiOuarr (1)
The reaction begins with nucleation at defect points on the SiSiO2 interface and proceeds
in a spatially inhomogeneous manner with the formation of large voids on the oxide
surface [3]
Several studies have been made on the dynamics of void growth both on thick [4] and thin
[5] layers of SiO2 The process has been used to decorate otherwise unobservable defects at
the SiSiO2 interface [6] or to grow nanoislands of silicon on the void surface [7] It has
been suggested that the defects acting as nucleation centres could be metallic contaminants
present on the native surface which aggregate and catalyze SiO2 decomposition [8]
After oxide desorption the silicon surface is very rough In particular on Si(100) square
islands are observed several nanometres high which act as lsquopinning sitesrsquo on the motion of
monatomic steps on the Si surface It has been proposed that these islands can be composed
of silicon carbide coming from the organic contaminants present on the surface [9] or of
oxide clusters during partial reoxidation [10]
In the following section some experimental results obtained studying voids and nanoisland
growth are presented
52 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nanoisland growth on silicon
We used two different samples the first rougher than the second (i) p-type Czochralski Si
(100) with a RMS roughness of 05 nm shown in fig52a and (ii) p-type epitaxial Si(100)
with a RMS roughness of 019 nm shown in fig 52b The roughness was measured with
STM on the lsquoas receivedrsquo samples on different areas of 500x500 nm Every sample was
covered by a layer of native oxide ~2 nm thick
Some of the samples (AC) were introduced into the UHV system without any cleaning
others (BD) were dipped in concentrated HF (49) to remove the surface oxide without
etching the silicon [11]
After degassing the samples were heated resistively increasing the temperature slowly to
keep the pressure within the 10-10 mbar range during heating Each sample was held at
900˚C for 30 min For some of the samples the heating was stopped at lower temperatures
to obtain incomplete oxide desorption and surface roughening thus enabling observation
of the different steps of the process Table 51 summarizes the different treatments for each
group of samples
Fig52 Original surfaces of Czochralski (a) and epitaxial (b) silicon Every image is 500x500 nm
VPalermo 53
Table 51 Summary of different treatements
Silicon type Oxide removed
Heating Islands density (microm-2)
Av Island Volume (nm3)
A Czochralski No 30rsquo at 900˚C 24x103 353
B Czochralski Yes ldquo 20x104 71
C Epitaxial No ldquo 28x102 2450
D Epitaxial Yes ldquo 23x104 63
Fig53 ab) Surface roughening on group A samples after heating 30 min at 800deg and900deg respectively in presence of an oxide layer Image size 500x500 nm cd) Surface roughening on group B samples after heating 30 min at 700deg and 900degrespectively after removing the oxide layer with HF Image size 250x250 nm
Fig 52a is the native oxide surface of sample A quite disordered and irregular This
surface was stable when heated to 700˚C while at higher temperatures the oxide began to
desorb Fig 53a was taken after heating the sample at 800˚C for 30 min
54 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The oxide began to desorb in correspondence with defect points creating oxide-free
surface voids which enlarge radially uncovering the silicon surface Some surface silicon
atoms are removed through reaction (1) Mobile silicon atoms moving randomly over the
surface aggregate forming a nanocrystal at the initial defect point surrounded by a stable
flat oxide-free surface with some steps The initial SiSiO2 interface as observed after
oxide etching with HF [12] is similar to the one shown in fig 52a and thus the surface
observed in the void area is a completely new one coming from surface reorganization
The reaction continues leading to a coalescence of the voids the final result of the process
is shown in fig 53b with protruding islands aligned along the main crystal axes and flat
areas with some steps LEED measurements confirmed that this surface is crystalline
silicon Atomic resolution was obtained on the flat areas of the surface showing the
typical 2x1 reconstruction of Si(100) Figures 53c and 53d show the evolution of the
roughening on sample B after chemically removing the native oxide layer before
introducing the sample into UHV As mentioned above the initial morphology is similar to
that shown in fig 52a though the oxide has been removed as confirmed by LEED
measurements Fig 53c shows the surface after heating at 700˚ for 30 min Without the
oxide layer the surface began to reorganize at lower temperatures and the small irregular
grains of the original surface begin to grow and become more rectangular After heating to
900˚C (fig 53d) the surface is flat with small rectangular islands The shape of the islands
is similar to that shown in fig 53b but the lateral island dimension is one order of
magnitude smaller
To assess the importance of the initial surface on the process the same treatments were
repeated starting from flat epitaxial silicon (samples CD) The initial epitaxial surface is
smooth with periodic steps due to a small miscut angle (fig52b) On heating the epitaxial
silicon without etching (sample C) relatively large islands are obtained similar to the ones
shown in fig 53b where the ldquostep-pinningrdquo effect of the islands is evident The average
dimension of sample C islands was sim2500 nm3 while only 353 nm3 for sample A Island
density decreased from 24000 to 280 microm-2 This result is maybe due to a lower
concentration of defects in the oxide layer of the epitaxial silicon which allows the
expanding void to grow much larger before coalescence with neighbouring voids the
nanostructures within them thus reaching greater dimensions To verify this hypothesis we
applied the same treatment to clean not-so clean and purposely contaminated samples To
check that the defects inducing void nucleation and nanoisland growth derive from carbon
VPalermo 55
contamination we used electronic-grade methanol to contaminate the silicon surface
Figure 54 shows the results of this comparison
State-of-the-art cleaned samples obtained with multi-step RCA cleaning [13] showed the
development of very few voids the oxide desorption leaving large smooth areas of oxide-
and nanostructure-free silicon as shown in Fig 54a where the presence of residual oxide
can be seen Samples cleaned with a simpler one-stage Pirana cleaning solution showed a
higher number of islands per unit area (Fig 54b) and samples purposely contaminated
with methanol (Fig 54c) had the highest density of island nucleation of all three samples
When the native oxide had been chemically removed from the epitaxial silicon before
heating (sample D) the island size and density is similar to the standard as-received
silicon case (sample B) Thus starting from a flat substrate the process of surface
roughening and island creation was the same even when the oxide layer had been
previously removed
From these results some important indications can be drawn The first is that oxide
desorption temperature depends upon the contamination levels of the oxide layers (fig 54)
with desorption at lower temperatures for contaminated surfaces
Second once the oxide layer is removed the silicon atoms become mobile on the surface
at relatively low temperatures (~700˚C) Their initial configuration is disordered and has a
high surface energy so they crystallize in the small islands shown in fig53d reducing
their surface area and lowering the energy of the system When the oxide is present the
surface cannot reorganize at 700˚C and is stable At 800˚C the oxide begins to desorb in
correspondence with defect points on the surface As the void area enlarges a small
cba Fig 54 The effect of contamination on nano-island production Gradient-filtered STM images of (a)sample cleaned two times with RCA 400x400 nm (b) Sample cleaned with a Pirana solution 500x500nm (c) sample contaminated with MeOH before insertion into the UHV system 500x500 nm Annealingtime is 40 min at 800degC for all the samples
56 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
crystallite grows at the centre of the voids Further oxide desorption uncovers larger areas
freeing more silicon atoms which feed the initial island making it bigger
Fig 55 shows the scaled size distribution of the islands for each group of samples Every
distribution has been obtained using several images of different samples for each group
The size distribution of all the samples follows an exponential decay and the decay seems
the same for all the samples though the average volume of the islands differs by more than
one order of magnitude (see Table 51 for details)
The point defects which catalyze void nucleation can be metallic contaminants present on
the initial oxide surface [8] or organic contaminants which at high temperatures can
form SiC nanocrystals [14] We used Scanning Tunneling Spectroscopy (STS) to look for
differences between the islands and the surrounding flat silicon surface but no difference
was found Furthermore the total island volume per surface unit is very high (more than
104 nm3microm2) and it seems unlikely that such a huge volume could consist of surface
contaminants Islands of pure silicon on silicon have been grown without evidence of
surface contamination[14]
To summarise the formation of nano-sized islands on silicon through surface diffusion
was studied At high temperatures the oxide layer covering the surface decomposes non
uniformly and circular voids of clean silicon are created The presence of the oxide layer
blocks surface silicon atomic motion and surface reorganization except within the voids
causing the growth of islands more than 10 nm high and 30 nm wide on the silicon surface
According to STM and LEED measurements we can say that the islands are mostly
Fig 55 Size distribution (scaled) for the samples described in table 1 The line is a reference for the eye
VPalermo 57
composed of silicon with traces of other substances such as carbon Final island densities
and dimensions depend upon the initial purity of the oxide layer
On the other hand if the oxide is removed chemically before heating the island growth
process is not so localised and takes place simultaneously over the whole surface yielding
smaller and more numerous islands
58 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[2] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[3] High-Temperature SiO2 Decomposition At The Sio2Si Interface Tromp R Rubloff
GW Balk P Legoues FK Physical Review Letters 55 2332-2335 Nov 1985
[4] Defect Microchemistry At The SiO2Si Interface Rubloff GW Hofmann K Liher M
Young DR Physical Review Letters 582379-2382 Jun 1987 Defect Formation In
Thermal SiO2 By High Temperature Annealing Hofmann K Rubloff GW Mccorkle
RA Applied Physics Letters 49 1525-1527 Dec 1986 Kinetics Of High-Temperature
Thermal Decomposition Of SiO2 On Si(100) Liher M Lewis JE Rubloff GW Journal
Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 5 1559-1562 Aug
1987
[5] Thermal-Decomposition Of Very Thin Oxide Layers On Si(111) Kobayashi Y Sugii
K Journal Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 10 (4)
2308-2313 Part 3 Jul-Aug 1992 Scanning Tunneling Microscope Study On Mid-
Desorption Stages Of Native Oxides On Si(111) Kobayashi Y Sugii K Journal Of
Vacuum Science amp Technology B 9 (2) 748-751 Part 2 Mar-Apr 1991 Controlled
Growth Of SiO2 Tunnel Barrier And Crystalline Si Quantum Wells For Si Resonant
Tunneling Diodes Wei Y Wallace RM Seabaugh AC Journal Of Applied Physics 81
(9) 6415-6424 May 1 1997
[6] Defect Formation In SiO2Si(100) By Metal Diffusion And Reaction Liher M
Dallaporta H Lewis Je Appl Phys Lett 53 589-591 Aug 1988 SiO2 Film
Decomposition Reaction Initiated By Carbon Impurities Located At A Si- SiO2
Interface Raider Si Herd Sr Walkup Re Applied Physics Letters 59 (19) 2424-2426
Nov 4 1991
[7] Nanometer-Scale Si Selective Epitaxial Growth On Si(001) Surfaces Using The
Thermal Decomposition Of Ultrathin Oxide Films Fujita K Watanabe H Ichikawa M
Applied Physics Letters 70 (21) 2807-2809 May 26 1997 Pyramidal Si Nanocrystals
VPalermo 59
With A Quasiequilibrium Shape Selectively Grown On Si(001) Windows In Ultrathin
SiO2 Films Shibata M Nitta Y Fujita K Ichikawa M Physical Review B 61 (11)
7499-7504 Mar 15 2000
[8] Stacking-Fault-Induced Defect Creation In SiO2 On Si(100) Liher M Bronner Gb
Lewis Je Appl Phys Lett 52 1982-1985 May 1988
[9] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990 AFM And XPS Characterization Of The
Si(111) Surface After Thermal-Treatment Lamontagne B Guay D Roy D Sporken R
Caudano R Applied Surface Science 90 (4) 481-487 Dec 1995
[10] Scanning-Tunneling-Microscopy Study Of Oxide Nucleation And Oxidation-
Induced Roughening At Elevated-Temperatures On The Si(001)-(2x1) Surface Seiple
JV Pelz JP Physical Review Letters 73 (7) 999-1002 Aug 15 1994 Evolution Of
Atomic-Scale Roughening On Si(001)-(2x1) Surfaces Resulting From High-
Temperature Oxidation Seiple JV Pelz JP Journal Of Vacuum Science amp Technology
A-Vacuum Surfaces And Films 13 (3) 772-776 Part 1 May-Jun 1995
[11] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995
[12] Morphological Changes Of The Si [100] Surface After Treatment With
Concentrated And Diluted HF Palermo V Jones D Materials Science In
Semiconductor Processing 4 (5) 437-441 Oct 2001
[13] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[14] Nanoscale Roughening Of Si(001) By Oxide Desorption In Ultrahigh Vacuum
Gray SM Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology
B 14 (2) 1043-1047 Mar-Apr 1996
60 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 61
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and nano-lines
When the silicon surface is not protected by a native oxide layer or by a hydrogenated
passivating layer it is very reactive The surface chemistry of both Si(100) and Si(111) has
been extensively studied starting from clean surfaces prepared in UHV because of its great
relevance to the microelectronics industry and its technological interest The diffusion and
reaction of molecules and atoms on silicon is an interesting scientific problem on its own
apart from technological issues because silicon reconstruction yields a very complex and
anisotropic surface
We give here a brief summary of the adsorption behaviour of several elements on Si
surfaces The summary is not comprehensive and only the most interesting characteristics
for each substance are given
Table 61 Summary of adsorption behaviour of atoms and molecules on silicon surfaces [12]
Hydrogen
Molecular H2 shows low reactivity towards silicon while atomic hydrogen easily forms
Si-H bonds and can even break Si-Si bonds
Alkali metals
Alkali metals diffuse rapidly into SiO2 and can damage silicon-based transistors
Transition
metals
All transition metals apart from gold and silver react with Si forming metal silicides
62 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nickel A very common contaminant of silicon Nickel from even slight contact with stainless
steel tweezers can contaminate silicon samples forming its silicide and is very difficult
to remove even at high temperatures because it segregates on the silicon surface
Palladium Forms silicides especially Pd2Si and can be used to make contacts possessing a low
Schottky barrier
Titanium Widely used to fabricate contacts for silicon-based microdevices TiSi2 is one of the
more stable and highly conductive silicides
Tungsten The silicide is more stable than TiSi2 on polysilicon and is thus used for polysilicon
gate contacts
Platinum Silicide is used for bipolar transistors PtSi contacts on n-silicon give Schottky diodes
Cobalt Cobalt silicide gives better quality contacts than TiSi2 at the nanometrelevel but is less
used at the moment because it consumes too much silicon during its growth
Iron Silicides have been studied for potential optoelectronic applications because FeSi2 has a
directly accessible bandgap and is optically active
Group 13
Group 13 metals are used in IC technology to dope silicon (giving p-type doping) and
to make III-V type semiconductors When evaporated on Si(100) or Si(111) they react
strongly with the surface giving surface reconstruction and kink formation at steps
Boron Widely used for doping Usually deposited with decomposition of boron hydrides
Sticking coefficient of simple hydrides is very low so decaboranes (B10H14) are used for
deposition
Aluminum Aluminum is widely used for microelectronic contacts but in some cases it reacts with
silicon diffusing into the crystal and with SiO2 stealing oxygen atoms to form Al2O3
Group 14
Carbon Carbon can form a composite semiconductor with silicon (silicon carbide) widely used
to fabricate integrated circuits for use at elevated temperatures or in the presence of
ionising radiation When carbon is deposited on silicon usually polycrystalline films
with properties between diamond and graphite are obtained
CO adsorbs molecularly on silicon without breakage of the C=O bond Other
unsaturated hydrocarbons do not react with silicon Even very reactive strained
molecules like cyclopropane have low sticking coefficient Ethylene acetylene and
propylene adsorb molecularly on Si(100) each molecule sitting on a silicon dimer
interacting with the dangling bonds of the surface Acetylene also adsorbs molecularly
on Si(111) on the 7x7 reconstructed surface preferential adsorption on centre atoms
with respect to corner atoms is observed Benzene and other aromatic ring compounds
adsorb molecularly on the surface They can form σ or π-type bonds with the silicon
laying parallel or tilted respect to the surface according to the bond type
Silicon Silicon atoms can be deposited on the surface with molecular beam epitaxy or
VPalermo 63
decomposition of silanes and chlorosilanes Silanes with single Si-Si bonds have a high
sticking probability and dissociate upon adsorption giving trihydrides and
monohydrides
Chlorosilanes have high sticking coefficients too and decompose on the surface but
they can etch the surface through the reaction SiCl4 + Si rarr 2SiCl2
Germanium Germanium has a lattice constant similar to silicon (Ge lc is 4 larger than Si) so that
various alloys of Si1-xGex can be formed The growth of Germanium on silicon is quite
peculiar with several monolayers adsorbing uniformly on the surface followed by
island formation (Stranski-Krastanov growth)
Group 15
Nitrogen Exposure to ammonia and nitrogen gas at high temperatures leads to the formation of
silicon nitride layers (Si3N4) NH3 dissociates on the surface and reacts with the
dangling bonds giving Si-NH2 and Si-H termination Silicon nitride is stable on the
surface up to 1100degC above this temperature it desorbs as Si2N
Phosphorous Phosphine (PH3) adsorbs dissociatively as Si-PH2 and Si-H and behaviour is similar to
that of ammonia
Arsenic A monolayer of As is often deposited on silicon as a substrate for GaAs growth Arsenic
forms dimers on Si(100) creating dimer rows which are parallel (perpendicular) to the
underlying silicon dimer rows when it is deposited at high (low) temperature On
Si(111) Arsenic breaks the 7x7 reconstruction giving a 1x1 pattern This 1x1 As
monolayer acts as a passivating layer and can resist further As adsorption oxygen and
air
Antimony and
Bismuth
Due to their large covalent radii Sb and Bi form only short dimer rows on silicon and
only at high temperatures
Group 16
Oxygen
Oxygen can oxidize or etch silicon according to the reactions
Si(s) + O2 rarr SiO2 (s)
Si(s) + frac12O2 rarr SiO(g)uarr
Silicon oxidation has already been described in the previous chapter Molecular oxygen
adsorption on silicon is mostly dissociative O atoms break the Si-Si bonds forming an
Si-O-Si bridge or a peroxy bridge Si-O-O-Si Hydrided silicon surfaces do not have
surface dangling bonds and are thus more protected from oxygen attack Etching is
favoured at high temperatures and low oxygen pressures [3] At T gt700degC silicon oxide
on silicon decomposes as SiO(g)
Water
H2O adsorbs dissociatively on silicon easily on Si (100) and with more difficulty on
Si(111) The molecule decomposes giving Si-OH and Si-H on adjacent sites
64 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Group 17
Fluorine
Fluorine adsorbs and reacts easily on silicon dissociating on the surface and forming
Si-F bonds It can then penetrate the surface and attack polarised Si-Si backbonds The
surface is etched with production of SiF4 and SiF2
Chlorine
Chlorine reacts aggressively with silicon etching the surface Cl2 and F2 are often used
commercially to etch silicon usually in a plasma Atomic Cl and Cl2 adsorb on Si(100)
up to saturation The most stable configuration seems to be a Cl atom bonded to each Si
atom of a surface dimer in a symmetric structure but metastable configurations with
two Cl atoms bonded to a buckled dimer and even a central Cl atom bridged across a
dimer have been observed
On Si(111) silicon mono- di- and tri-chlorides are formed In particular heating at
Tgt400degC a uniform Si-Cl monolayer is obtained and the 7x7 reconstruction changes
into a bulk like 1x1 lattice The surface transition induced by Cl adsorption is fully
reversible The 7x7 reconstruction can be restored on desorbing the chlorine by heating
at Tgt1100degC The 1x1 domains nucleate at the lower terrace side of steps the 7x7 at the
upper step edge
Bromine
Like hydrogen bromine maintains the 2x1 structure of Si(100) forming Si-Br bonds
with the dangling bonds of the silicon dimer rows At high exposure some etching of
silicon by formation of volatile SiBr3 species has been observed
Surface diffusion on silicon
In the previous chapter we observed nano-island growth on silicon surfaces with native
oxide and surfaces contaminated with organic impurities While this phenomenon has been
observed in several experiments the formation and growth mechanisms of nanoislands has
still not been satisfactorily described theoretically Various substances have been
hypothesised as nanoisland nucleating agents such as organic or metallic contaminants
present on the surface or oxide clusters (see previous chapter) The island growth process
has been attributed by various workers to lsquosome sort of kinetic instabilityrsquo [4] to the
pinning of step flow by SiC clusters [5] or to a mesoscopic atom flux from areas with low
step density to areas of higher step density [6]
One mechanism proposed to explain surface roughening and island growth is the Ehrlich-
Schwoebel effect ie the presence of an energetic barrier that reflects atoms approaching a
VPalermo 65
Molecule (L
O2
H2O MeOH
CO CO2 CH4
clean surface
downward
weak on si
The aim o
surface an
To obtain
the ones
spontaneou
first obser
surface W
nanoisland
Experime
We used S
removed f
41 mixtur
water and
whole hea
1200degC re
kept below
Table 62 Treatment island density and island volume for each molecule tested
Dose angmuir)
Heating time at 800degC (min)
Island density (microm-2)
Ave Volume (nm3)
Equivalent layer thickness (nm)
280 10 no islands - -
90 10 no islands - - 30 10 2000 19 004
180 10 360 516 019 30 60 no islands - -
180 10 no islands - - - 10 no islands - -
step [7] Recent experiments however indicate that the Schwoebel effect is
licon [48]
f our experiments was to study the reaction of simple molecules with the silicon
d to check the ability of these molecules to nucleate nanoisland growth
a better understanding of the process we chose simple molecules starting from
which are more likely to be present as traces in UHV chambers where
s nanoisland growth is often observed after sample heating Using STM we
ved at the atomic level the possible interactions of these molecules with the
e then heated the surface and checked the density and dimensions of the
s produced
ntal procedures and results
i(111) wafers p-doped 0015 Ω cm Gross contamination and particles were
rom the sample surface with a standard Pirana etch (15 min dip in a H2S4H2O2
e at T=80degC) After this the samples were thoroughly washed in ultra-pure
introduced into the UHV system After degassing for 5 hours at 600degC the
ting stage was allowed to cool down and a sequence of rapid flashes to 1100deg-
moved the native oxide layer and cleaned the surface Pressure during flashes is
1x10-9 mbar
66 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Surface flatness and cleanliness were checked with STM and the contaminant molecules
introduced into the UHV system through a leak valve Doses of several Langmuirs were
used for the experiments (1 Langmuir=10-6 torr sec-1) Pressure was kept constant at 10-7
mbar during adsorption The STM tip was retracted during gas deposition to avoid any tip
shielding effect After the adsorption process the leak valve was closed the base pressure
in the UHV chamber returning to below 10-10 mbar and the presence of adsorbed
molecules checked on the surface with STM Finally the sample was heated at 800degC for
several minutes to allow nanoisland growth Several different areas were examined on each
sample
Table 62 summarizes the dose of exposure and the heating time for each molecule Other
experiments were carried out using smaller dosing times but only the significant results
are reported here
As expected oxygen and water react with the surface oxidising it The adsorption process
can be easily followed with the STM by lowering the tip from time to time and taking an
image No effect of the STM tip is observed over the scanned areas Upon molecule
adsorption the order of the crystalline surface rapidly degenerates and after several
minutes the surface looks completely covered by irregular atomic-sized protrusions
corresponding to Si-OH or Si-O-Si species On heating to 800degC the contaminating species
were easily desorbed through SiO formation the surface became clean and the crystal
surface periodicity re-established No residues or nanoislands were observed
Amongst the molecules tested was methanol CH3OH reacts readily with the surface and
covers it with a disorderd layer already after sim 30 Langmuirs exposure (fig 61)
Fig 61 Si(111) surface during methanol adsorption at 0 1 and 5 minutes respectively (corresponding to 0 6 and 30 L exposure) The dark irregular lines are borders between different 7x7 domains Image size 60x60 nm
VPalermo 67
Synchroton radiation photoemission spectroscopy experiments by Carbone et al [9]
showed that methanol reacts with the rest atoms of the 7x7 silicon cell already after 1
Langmuir exposure but that longer exposures are needed for methanol to react with the
remaining silicon atoms Methanol adsorbs dissociatively forming SiOCH3 and Si-CHx
species on the surface
After methanol adsorption the sample is heated to 800degC and observed again with STM
As in the case of water and oxygen the heating restores the crystalline surface and the 7x7
pattern returns visible But this time quite a dense array of nanoislands is observed on the
surface (fig 62) The islands have an average diameter of 20 nm and a height of sim2 nm
At temperatures above 400degC the Si-OCH3 and Si-CHx species decompose oxygen and
hydrogen are desorbed and according to [9] carbon atoms remain as SiC dispersed
uniformly on the surface Our STM measurements indicate that the carbide does not
randomly cover the surface but that C atoms are concentrated at the nanoislands Rough
calculations based on island volume suggest that the islands must be composed of a SixC1-x
alloy with x varying between 05 and 1
Following these results we expected carbon monoxide to behave in a similar manner
adsorbing onto the surface and with nanoisland growth However no adsorption was
observed with STM and the surface looked perfectly clean even after a dose of more than
100 L of carbon monoxide Some species did adsorb onto the surface though because
after 10 minutes subsequent heating nanoisland growth was observed Island density in
this case was only 18 of the density obtained with methanol and island dimensions are
Fig62 Nanoislands on Si(111) created after 30 L adsorption of methanol and 10 min heating at 800degC Image size 200x200 nm
68 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
much bigger (see table 62 and fig 63) To check whether the islands had been nucleated
by some other contaminant present in the vacuum chamber we heated a blank sample
without introducing any molecule No island growth was observed Exposing the surface to
other simple molecules including carbon dioxide and methane also gave no nanoisland
formation
We can estimate the amount of carbon deposited on the surface during gas dosing The
total flux of molecules hitting the surface is obtained from the Hertz-Knudsen formula
TkmPF
π2=
Fig 63 STM images of Si(111) after exposure to methanol (left) and carbonmonoxide (right) contaminating molecules and subsequent nanoisland growthImage size 400x400 nm The size distribution of nanoislands is shown below eachimage
VPalermo 69
Fig64 Left Si(111) after 180 L contamination of CO and 16 hours heating at 800degC Right Si(111)without CO contamination after 16 hours heating at 800degC Image size 250x250 nm Vertical scale is thesame for both images z-ranges are 35 and 5 nm respectively
where P is the pressure in Pascal k=138x10-23 J K-1 is the Boltzmann constant T is the
temperature and m is the molecular weight (sim32 for methanol and sim28 for CO) in
kilograms
Using this formula we obtain a flux of 027 molecules nm-2 sec-1 for methanol and 029
molecules nm-2 sec-1 for CO After ten minutes exposure at 10-7 mbar pressure more than
150 molecules will have hit each square nanometer of the surface Surface density of
Si(111) 7x7 is sim16 atomsnm2 so each surface atom will be hit by several molecules which
could react or be adsorbed even assuming a sticking coefficient much lower than unity
To explain nanoisland growth we hypothesize that CO molecules adsorb molecularly and
-5
0
5
10
15
20
25
30
0 20 40 60 80 100 120
nm
nm
10 min40 min16 h16 h- no CO
Fig65 STM profiles of nanoislandsgrown for different annealing times The lower curve correspond to the surface heated at 16 hours without CO contamination
70 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Table 63 Island density and volume for the CO-contaminated surface at different heating times For each sample the thickness of a uniform layer having the same volume per unit area of the islands is calculated
Molecule Dose
(Langmuir) Heating time at
800degC (min) Island density
(microm-2) Ave Volume
(nm3) Equivalent layer thickness (nm)
CO 180 10 362 517 019 CO 180 40 550 1072 059 CO 180 16 h 505 3675 186
clean surface - 16 h rough - -
very weakly on Si(111) moving rapidly on the surface Because of their rapid diffusion
the STM is not able to image the CO molecules on the silicon [10]
Increasing the temperature part of the CO molecules will surely desorb but some of them
will dissociatively adsorb onto the surface The oxygen will then desorb as SiO while
carbon will remain fixed forming very stable Si-C bonds
CO decomposition will be favored at surface defects surface steps or in the proximity of
already formed SiC clusters An increased reactivity of the CO molecule on a surface in
correspondence with phase boundaries has already been observed on Pt surfaces [10]
In this case a reduced number of nucleation centres will react with the CO molecules
yielding large and fewer islands with respect to the case of methanol which does not have
the possibility of travelling long distances over the surface
Other factors are likely to influence the process of island nucleation after CO adsorption
the co-adsorption of other molecules which can slow down CO and favour its
decomposition cannot be ruled out as well as the formation of new defects at higher
temperatures Measurements with a variable temperature STM or with some other surface
analysis technique are planned in the future to confirm the proposed mechanism
When the surface was further annealed the nanoislands grew in size After 16 hours
heating very large nanoislands with diameters of sim35 nm and heights of sim20 nm became
visible (fig64 left) As a comparison when the same surface was heated without
nucleation centres deriving from contamination an irregular surface was obtained (fig64
right)
While the nanoisland density seemed to reach an asymptotic limit of 500 islandsmicrom2
(table 63) island size continued to grow with time (fig65) even when no further carbon
was supplied to the surface This indicates that even though the initial nucleating core of
the nanoisland is likely to be an SixC1-x alloy further growth is due to silicon atoms
diffusing from the crystal and being adsorbed by the growing island The final volume
VPalermo 71
occupied by the islands corresponds roughly to a 18 nm thick overall layer of removed
silicon
It is noteworthy that the clean sample after 16 hours annealing even though smoother than
the CO-contaminated one looked much more disordered with hardly any flat area visible
on nanometres scale In the presence of nanoislands however further annealing increases
island size but keeps part of the surface quite flat and ordered with flat areas visible
between the islands Thus the nanoislands can be imagined to act as ldquoimpurity sinksrdquo for
further contaminants approaching the surface yielding a greater but more ordered surface
roughening with respect to the uncontaminated silicon surface
Circular arrays of nanoislands
We used the voids described in the previous chapter formed during oxide layer
decomposition as nanoscopic masks to test the validity of the hypothesis outlined above
The small circular areas uncovered during oxide decomposition can act as ldquoskating rinksrdquo
where contaminants and diffusing silicon atoms can move freely over the surface while
the oxide layer all around and higher than the void area as well as being chemically
bonded to the surface silicon atoms will hinder surface diffusion Furthermore the oxide
passivating layer will at least in some cases prevent contaminating molecules from
adsorbing and decomposing on the surface In this way selective nanoisland growth inside
void areas can take place
The experimental procedure consisted of the following steps
1) An oxide-passivated sample was heated for 10 minutes at 800degC Oxide decomposition
began in correspondence with surface defects and spread laterally creating naked areas
of silicon
2) While the void enlarges the same defect that initiated oxide decomposition behaved as
an attractor for moving silicon atoms and nucleated the growth of a nanoisland at the
void centre
3) The void surface was exposed to contaminating molecules
4) The sample was then heated again at 800degC During the second heating the void
continued to enlarge and new nanoislands were created The freshly uncovered silicon
surface provided mobile silicon atoms to feed nanoisland growth
72 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
5) When the heating continued for long enough all the oxide was desorbed and a clean
silicon surface covered by circular groups of nanoislands was obtained
Figure 66 shows the different stages of the process
We thus heated an oxidized sample at 800degC for 10 minutes in UHV to create multiple
voids in the oxide layer After this we exposed the sample to the contaminating molecules
Then we further annealed the surface at the same temperature to promote island growth
The first time we observed nanoisland growth in the void area was by chance after heating
for a second time samples which had been stored for 2 weeks in vacuum Even at pressures
as low as 10-11 mbar some molecules will adsorb onto the surface slowly creating defects
and thus potential nanoisland nucleation centres It is possible to observe with STM that
the silicon crystal surface looks very clean just after a high temperature flash but even
after only a few days randomly adsorbed species will be observable on the surface
Unfortunately it is not possible to identify these species simply from in situ STM
measurements
Mass spectrometry measurements show that the residual gas contaminants in vacuum are
usually He Ar H2 CH4 CO CO2 and N2 [11] Some of these molecules such as
hydrogen will stick to the silicon surface at room temperature but will simply desorb
during annealing without nucleating nanoislands Others like CH4 or CO2 will not react
with the surface to generate nanoislands as demonstrated previously The best candidate
for contamination of samples stored in UHV seems thus carbon monoxide has the ability
321
4 5
Fig 66 Selective nanoisland growth within oxide voids See text for details
VPalermo 73
as shown before to nucleate nanoislands although co-adsorption mechanisms can not be
ruled out Fig 67 (left) shows the circular groups of nano-islands obtained after heating
void-covered samples stored for 2 weeks in UHV Small islands are observed in the former
void area The void itself has enlarged slightly uncovering a clean oxide-free area of the
silicon surface Apart from the central island created during the first period of heating the
surrounding ones have dimensions decreasing from the void border to the inner area of the
void
This confirms that most of the material needed for island growth comes from the silicon
atoms diffusing from the freshly uncovered areas around the original void The outer
nucleating centres are nearest to the silicon atom source and thus generate larger islands
For a faster process we directly exposed the void to a significant concentration of CO
Fig67 (centre) shows the islands grown after a 180 Langmuir exposure and subsequent
heating The islands are preferentially located around the original void perimeter the rest
of the original oxide-free surface remaining untouched This suggests that although the
whole surface is exposed to CO the gas only interacts chemically with the surface at the
SiSiO2 interface around the oxide-free void perimeter The SiSiO2 interface thus provides
preferential nucleation sites for nanoisland growth in the successive thermal annealing
step This recalls a similar phenomenon observed in the catalytic oxidation of CO on the
Pt(111) surface where CO molecules reacted with adsorbed O species only along the
perimeters of oxygen islands [10]
On repeating the CO exposure and thermal annealing steps a second circle of nano-islands
Fig67 Left circular area of nanoislands obtained from a sample stored 2 weeks in UHV 500x500 nmCentre nanoisland circle obtained after exposing the voids to 180 L of CO and heating for 10 min at800degC 400x400 nm Right two concentric circles obtained with further CO adsorption followed byheating 650x650 nm
74 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
was formed around the newly-expanded surface void perimeter (fig67 right) This
process can in principle be repeated again to obtain a series of concentric groups of
nanoislands the only limit depending on the minimal distance between neighbouring
expanding voids We have thus observed that contaminants adsorbed at low pressure over
long periods of time generate random island growth over the whole void area while CO
molecules adsorbed at relatively high pressures over short periods yield selective
nanoisland growth at the former void border Several theories can explain the different
results CH4 and CO2 as mentioned above did not nucleate islands when when the silicon
surface was exposed to them for short periods However the lower surface mobility of
these contaminants and the greater time available for reaction with the surface could be
the cause of nucleation of SiC clusters over the whole oxide-free void area and not only at
its border The coadsorption of water another UHV residual gas with CO on the naked
silicon areas could perhaps allow a surface reaction similar to that between CO and the
SiSiO2 interface during CO exposure Another explanation for the formation of these
random nucleation sites within the voids could be the coadsorption of residual carbon-
containing species present in the UHV system with hydrogen the major residual gas in
stainless steel UHV systems
The possibility of decorating silicon oxide borders with nanoislands using the high
mobility and selective decomposition of CO molecules suggests interesting applications
for nanofabrication processes that will be discussed below
Silicon nanowire creation on Si(100)
We performed most of the experiments described above on silicon (111) 7x7
reconstructed because this surface is easily prepared in UHV and above all because it is
isotropic and has no preferential directions for atom diffusion The 2x1 reconstruction of
silicon (100) even though it has a much simpler unit cell than Si(111) 7x7 shows a more
complicated morphology with lots of monoatomic and biatomic steps and dimer rows
parallel to each other The orientation of the dimer rows changes by 90deg on alternate
atomic layers Surface diffusion in this case is much more complicated because atoms will
experience the effects of the surface anisotropy and diffuse preferentially along or across
the dimer rows The diffusion energy of a silicon atom as example is 06 eV along a
dimer row and 085 eV across different dimer rows This while being a complication
VPalermo 75
offers interesting possibilities for the creation of ordered structures The monoatomic steps
and the dimer rows can for example act as templates for the formation of elongated
structures of composition similar to the islands described previously
On heating a Si(100) surface oxide decomposition and void growth takes place as
described for Si(111) In some cases the void shape reflects the surface symmetry
especially for very thin oxide layers and the voids often have a slightly squared shape
Nanoisland growth is also influenced by the substrate symmetry and the islands look
rectangular as irregular clusters surrounded by concentric patterns of monoatomic steps
The whole structure reminds vaguely a ldquozigguratrdquo a kind of stepped pyramid found in
Mesopotamia (fig68 inset)
Apart from nanoisland growth in some cases we observed spontaneous formation of
nanowire-like SiC structures on the Si(100) surface The wires were generated after
exposing a Si(100) surface in vacuum to traces of CO followed by sample annealing at
800degC for 15 minutes The wires exhibit lengths ranging from 10 to 100 nm and average
widths of ca 5 nm Although being randomly positioned on the surface they are perfectly
aligned along the crystal axes of the substrate (Fig68)
Several descending monoatomic steps can be observed around each nano-line and nano-
island Very often one of the line extremities coincided with an island
Nanoline
Nanoisland
Fig 68 A Si(100) surface covered by nanoislands and nanowires The monoatomic steps present on the surface are visible The two insets show a typical nanoisland and a nanowire obtained on this surface The typical ldquodimer-row ldquostripes are visible along the sides of the nanowire
76 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Couples of parallel lines starting from the same nano-island were also observed At high
resolution (fig 69) the lines showed a lamellar periodic structure with a periodicity of
around 4 nm and frequent defects Scanning Tunneling Spectroscopy measurements
showed no significant difference between the line and the surrounding silicon surface
The proposed formation mechanism for this kind of structure involves two stages in the
first one organic contaminating molecules adsorb and diffuse on the surface from the
vacuum environment stopping preferentially at surface singularities including steps and
defects
Several substances such as atomic carbon carbon monoxide ethylene and fullerene
decompose when adsorbed on silicon surfaces at temperatures above 500degC and create
localised defects on the surface inducing strain deformations and a metastable surface
reconstruction [12]
If these molecules remain attached to the steps they can diffuse rapidly along step border
until they meet a line and decompose In this way long lines of SiC defects parallel to the
surface steps can be produced In the second phase the clusters of silicon carbide can
-005
0
005
01
015
02
025
03
035
04
-3 -2 -1 0 1 2 3
V
dId
V lt
dId
Vgt
linesilicon
Current- Voltage behaviour measured with STS Fourier analysis of wire periodicity
asymp 4 Aring
High resolution image of a line
Fig 69 Top high resolution image of a nanowire showing the lamellar structure Fourier analysis (bottom left) shows that the lamellae periodicity is sim 04 nm bottom right the IV characteristics measured with the STM tip on the line and the silicon surface
VPalermo 77
25-600degC
Final linear structure
600-800degC
Contaminant Silicon
Fig 610 Schematic representation of nanowire formation mechanism
locally inhibit the spontaneous surface diffusion of silicon atoms acting as templates for
the reorganisation of surface steps (fig 610)
The final morphology shows disordered lines and islands each surrounded by a complex
pattern of silicon monatomic layers
At 800degC silicon atoms are highly mobile on the surface Due to atom diffusion the
surface reorganizes with a step-flow mechanism The presence of a fixed line of
contaminants blocks atom diffusion and step flow generating the monoatomic steps
around each wire (fig 611) SiC clusters act as lsquopinning sitesrsquo on silicon atom surface
diffusion and can be purposely used to direct the localised growth of nano-islands [13]
Although the possibility of using methanol and carbon monoxide as nano-island precursors
has been demonstrated the exact nature of the substances nucleating nano-wire growth is
still unknown
Further experiments will be needed to confirm these hypotheses as well as to identify the
contaminating species nucleating nanowire growth Carbon monoxide is a main candidate
for nanowire growth but the real nucleation process is likely to be a complex one
78 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Step flowblocked
Fig 611 Representation of step flow blocked by the presence of a nanowire 105x105nm
Bibliography
[1] Surface-Chemistry Of Silicon Waltenburg HN Yates JT Chemical Reviews 95 (5)
1589-1673 Jul-Aug 1995
[2] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[4] Growth Of Si On The Si(111) Surface Lanczycki Cj Kotlyar R Fu E Yang Yn
Williams Ed Das Sarma S Physical Review B 57 (20) 13132-13148 May 15 1998
[5] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
VPalermo 79
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990
[6] Nanoscale roughening of Si(001) by oxide desorption in ultrahigh vacuum Gray SM
Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology B 14 (2)
1043-1047 Mar-Apr 1996
[7] Step Motion On Crystal Surfaces Schwoebel Rl Journal Of Applied Physics 40 614-
618 Feb 1969
[8] Scanning Tunneling Microscopy Investigation At High Temperatures Of Islands And
Holes On Si(111)7x7 In Real Time Evidence For Diffusion-Limited Decay
Hildebrandt S Kraus A Kulla R Wilhelmi G Hanbucken M Neddermeyer H Surface
Science 486 (1-2) 24-32 Jul 1 2001
[9] Methanol Adsorption On Si(111)-(7x7) Investigated By Core-Line Photoemission And
Mass Spectrometry Of Photodesorbed Ions Carbone M Piancastelli Mn Zanoni R
Comtet G Dujardin G Hellner L Surface Science 370 (1) L179-L184 Jan 1 1997
[10] Atomic And Macroscopic Reaction Rates Of A Surface-Catalyzed Reaction
Wintterlin J Volkening S Janssens Tvw Zambelli T Ertl G Science 278 (5345)
1931-1934 Dec 12 1997
[11] Redhead PA Hobson JP Kornelsen EV The Physical Basis Of Ultrahigh
Vacuum Chapter 12 (Chapman amp Hall London 1968)
[12] The Si(001) C(4 X 4) Surface Reconstruction A Comprehensive Experimental
Study Norenberg H Briggs Gad Surface Science 430 (1-3) 154-164 Jun 21 1999
[13] Production Of Nanostructures Of Silicon On Silicon By Atomic Self-Organization
Observed By Scanning Tunneling Microscopy Jones D Palermo V Applied Physics
Letters 80 (4) 673-675 Jan 28 2002
80 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 81
7 Conclusions and proposals for future work It is very difficult to foresee the future of nanoscience and nanotechnology even over the
next few years exciting discoveries are continuously made opening up new possibilities
and progress in this sector is now one of the fastest and most varied So it is impossible to
predict whether computers will in the future be made with nanowires nanotubes single
electron devices or some kind of quantum-based transistor What does seem clear
however is that silicon devices can still further extend their levels of miniaturization to
remain competitive for at least the next ten years [1] During this period new emerging
technologies will reach scientific maturity and arrive at the production lines
Even when new technology prototypes do become available for full-scale production the
astronomical cost of changing from silicon to new technologies will be a major concern
Moreover a vast amount of extremely detailed knowledge is already available on all
aspects of silicon technology Thus the possibility of integrating innovative
nanotechnologies with standard CMOS silicon technology already seems an attractive
prospect [2]
For this and other reasons we have focussed our research efforts on surface phenomena on
silicon wafers which could be potentially useful for the development of silicon-compatible
nano-devices
Some important conclusions can be drawn from the work described in this thesis
bull Silicon surfaces of different crystal faces were studied at atomic resolution in ultra-
high-vacuum following transformations induced on the surface by chemical (etching)
and physico-chemical (organic contamination heating) treatments
82 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull The presence of surface contaminants induces at high temperature the growth of
nanometre-sized islands and lines on silicon
bull Final island densities and dimensions strongly depend upon the chemical nature of the
contaminant molecules and the treatments used Molecules interacting only weakly
with silicon will desorb upon heating whereas molecules which decompose easily on
Si will give widespread random nano-island nucleation Between these two extreme
cases complex behaviours of diffusion and subsequent decomposition are possible
bull Nanoscopic naked silicon voids were produced in oxide layers through carefully-
controlled thermal annealing It was found that void density and size depends upon
oxide purity surface contamination levels and annealing conditions
bull The growth of nano-islands and nano-voids was exploited simultaneously and the
voids used as nanoscopic masks to control the positions of growing islands Complex
silicon-on-silicon nano-structures were obtained with multi-step growth processes
bull Finally the use of macroscopic treatments and simple physical processes to produce
large quantities of nanometre-sized structures on silicon surfaces was demonstrated
These processes and the structures produced are of potential interest to the
microelectronics industry
We showed in previous chapters how it is possible to modify silicon surfaces using
chemical and physical methods The use of self-organization for the production of Si-on-Si
20 nm
Fig 71 Left SEM image of field emitter arrays made by conventional lithography [3] Center 3-D STM image of a self-organized void with a central nano-island Right Lateral schematic view of a field emitter device
VPalermo 83
nano-structures has the advantage with respect to other emerging techniques of being
silicon compatible and thus is a potential candidate for its implementation in the
production of new devices Furthermore this technique can create billions of strucures on a
wafer in only a few minutes being thus applicable to large-scale production
Even when ordered structures are obtained these methods still show a poor reproducibility
and control Chemically modified surfaces look very disordered at high magnification
with trenches and holes etched into the surface and progress has recently been made in
producing very small structures electrochemically [4] Oxide decomposition with
nanoisland growth in UHV can also yield ordered structures The void holes with
nanoislands located at their centres although quite variable in dimensions are more similar
to engineered devices than to a spontaneous random surface-roughening process In fact
there is a surprising similarity between the nanovoid-nanoisland structures and well-known
commercially available devices called field emitter arrays (FEA) These are small tips used
as electron guns in various kinds of displays and other devices Both the commercial
devices and our self-organized structures consist of a protruding tip at the centre of
microscopic holes on an insulating layer covering a semiconductor surface (see fig71)
Of course the order and size uniformity of commercial FEAs is much better than that
obtained with our process but it is encouraging to see that a self-organized structure
obtained with simple heating has a very similar surface topography and composition of
commercial working devices obtained after many expensive stages of optical lithography
deposition and etching
Nano-island density and size can to some extent be controlled and different
contaminants can be adsorbed onto the surface to modify the growth process In the future
with a better understanding of diffusion dynamics and growth mechanisms a multi-stage
fabrication process can be envisaged where purposely-designed molecules are adsorbed
onto silicon surfaces to diffuse organize in ordered arrays and favour surface
reorganization at the nanometre scale The same molecules could be used to deliver
dopants to obtain for example highly conducting nanostructures on otherwise undoped
highly resistive silicon The selective formation of nanostructures at the SiSiO2 border
can be used to scale down mesoscopic patterns to the nanometric level A silicon oxide
layer can be easily patterned on the surface with normal optical lithography and then
nanostructures can be built following the pattern edge In this way the production of very
small nanowire based devices may be possible (fig 72)
84 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
D
Gate
S
D
S
SiO2
SiO2
2 Heating nanowire creation
and oxide removal
3 Source-drain deposition by conventional lithography
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
4 Insulating layer (oxide) and gate (metal) deposition
Fig 72 Schematic drawing of the construction of a self-organized nanowire-based transistor
Surface diffusion and decomposition of molecules can be sensitive to even smaller surface
features The nanoscopic lines formed on Si(100) are very regular and neat if their
nucleation and growth mechanism could be controlled ordered networks of nanowires
could be made on silicon
VPalermo 85
The techniques developed here are still experimental ones the careful control and
reproducibility of each step is still an issue and further studies will be needed to clearly
understand the dynamics of formation of these structures Nonetheless the results obtained
so far open up the possibilities of producing novel devices such as for example nano-
transistors (see scheme in fig73 where the mechanism described in fig 72 is applied for
the fabrication of a device array) The possibility of using the border of a lithographic
pattern to draw nanolines can in principle allow the creation of nanodevices using normal
IC manufacturing techniques
The results obtained and their description in this thesis are not the first examples of the use
of self-organization for the production of ordered structures and will certainly not be the
last In our opinion the most important outcome of these findings is that it is possible to
use surface diffusion to overcome the inherent limits of lithographic techniques in
microdevice production processes and to force properly selected molecules to react with
the surface in a spatially differentiated manner at particular sites thereby creating ordered
series of nanostructures The role of the surface dynamics of the substrate atoms
themselves is extremely important in this process
Outstanding results have been and continue to be obtained in nanoscience and
nanotechnology research promising the emergence of new production technologies Sooner
or later those emerging technologies will have to face the issues of cost and compatibility
with the enormous investment made in existing technologies and processing facilities The
commercial production of nanostructures based on surface diffusion processes could in
principle be cheap simple and compatible with existing technology Moreover the
processes of diffusion in bulk silicon is well-known in the microelectronics industry which
has both the know-how and the equipment to better understand and exploit the surface
diffusion processes on silicon surfaces studied here
Extremely important and innovative results are often obtained by exploiting very simple
ideas and in our opinion physical processes such as surface diffusion used to create
complex nanostructures on silicon surfaces is a prime example
86 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
2 Heating nanowire creation and
oxide removal
3 Source-drain deposition by conventional lithography
4 Insulating layer (oxide) and gate (metal) deposition
Fig 73 Schematic macroscopic outline of the construction of a self-organized nanowire-based transistor
VPalermo 87
Bibliography
[1] P Gargini Intel Technology Coordinator ldquoFrom Microelectronics To
Nanotechnologiesrdquo Invited Campus Colloquia Lecture CNR Research Area Bologna
February 6th 2003
[2] Toward A Hybrid Micro-Nanoelectronics Cerofolini Gf Ferla G Journal Of
Nanoparticle Research 4 (3) 185-191 Jun 2002
[3] Recent Progress In Field Emitter Array Development For High Performance
Applications Temple D Materials Science amp Engineering R-Reports 24 (5) 185-239
Jan 25 1999
[4] Silicon Dioxide Micropillars For Sieving Fabricated By Macroporous Silicon-Based
Micromachining Izuo S Ohji H French Pj Tsutsumi K Kimata M Sensors And
Materials 14 (5) 239-251 2002 Electrochemical Etching In HF Solution For Silicon
Micromachining Barillaro G Nannini A Piotto M Sensors And Actuators A-Physical
102 (1-2) 195-201 Dec 1 2002
88 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 89
Acknowledgements
This work would not have been possible without the help and guidance of Dr Derek Jones
he introduced me to the use of STM and other techniques found the funding for this PhD
and helped me with the thousands of small and big problems I had to face during my
research I am grateful to Prof Alberto Ripamonti too for being the supervisor of this
thesis Thanks are also due to Dr Giancarlo Seconi director of ISOF-CNR where most of
this work was carried out Financial support from the Italian National Research Council
(CNR) is also gratefully acknowledged
My passion for surface science began when Fabio Biscarini showed me that it was possible
to really see the atoms and for this I will always thank him
During these years I had the luck to collaborate with many good scientists among those I
would like to cite Enrichetta Susi Massimo Cocchi and Anna Mazzone from CNR
Claudio Zannoni Anna Cavallini Daniela Cavalcoli and Antonio Castaldini from the
University of Bologna Sergio Pizzini and Maurizio Acciarri from the University of Milan
One of the most stimulating periods of my PhD was the one I spent at the Steacie Institute
for Molecular Sciences in Ottawa working with Robert Wolkow Dan Wayner Greg
Lopinski and Peter Kruse I thank them for discussing together many exciting scientific
ideas and trying to put some of them into practice
Special thanks go to my colleagues Paolo Samorigrave and Stefano De Cesari for long and
useful discussions on Science Life and their interactions often with the help of a good
pint
Finally I would like to thank Vassilia Gaetano Simone Claudia Silvia Letizia
Alessandro Angela and Sebastien even though not directly involved in this work they
have been over these three years a continuous source of happiness and support
Bologna March 2003
90 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
SCIENTIFIC PUBLICATIONS bull Lateral diffusion of titanium disilicide as a way to contacting
hybrid Si-organic nanostructures Palermo V Buchanan M Bezinger A Wolkow RA APPLIED PHYSICS LETTERS 2002 v81 p 3636 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v6 issue 20) bull Surface modifications in Si after Rapid Thermal Annealing Castaldini A Cavalcoli D Cavallini A Jones D Palermo V Susi E JOURNAL OF THE ELECTROCHEMICAL SOCIETY 2002 v 149 pG633 bull Production of nanostructures of silicon on silicon by atomic self-
organisation observed by scanning tunnelling microscopy Jones D Palermo V APPLIED PHYSICS LETTERS 2002 v 80 p 673 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v5 issue 5) bull Nucleation of nanostructures from surface defects on silicon Palermo V Jones D SOLID STATE PHENOMENA 2002 v 82-84 p 687 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo MATERIALS SCIENCE AND ENGINEERING B 2002 v88 (2-3) p220 bull Morphological changes of Si[100] surface after treatment with
concentrated and dilute HF Palermo V Jones D MATERIALS SCIENCE IN SEMICONDUCTOR PROCESSING 2001 v 4 p 437 bull Electrical and structural properties of processed silicon surfaces Susi E Cavallini A Castaldini A Cavalcoli D Jones D Palermo V ldquoRECENT RESEARCH DEVELOPMENTS IN VACUUM SCIENCE amp TECHNOLOGYrdquo 2001 v 3 p189 bull Numerical Solutions of the Stochastic Equations of Crystal
Growth Mazzone A M Palermo V INTERNATIONAL JOURNAL OF MODERN PHYSICS C 2000 v 11 Part 1 p195-204 bull Advances in silicon surface characterisation using light beam
injection techniques
VPalermo 91
Acciarri M Pizzini S Simone G Jones D Palermo V MATERIALS SCIENCE AND ENGINEERING B 2000 V73 (1-3) p 235 - 239 bull Abrupt orientational changes for liquid crystals adsorbed on a
graphite surface Palermo V Biscarini F Zannoni C PHYSICAL REVIEW -SERIES E- 1998 V 57 NUMBER 3A p R2519-R2522 Oral presentations bull The use of oxide desorption and surface diffusion for the
creation of silicon on silicon nanostructures Palermo V Jones D 1ST NATIONAL WORKSHOP ON CURRENT TRENDS IN NANOTECHNOLOGIES Catania (Italy) Feb 2002 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo V E-MRS Spring Meeting Strasbourg (France) June 2001 bull STM study of surface transformations on silicon during UHV
annealing Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999 bull Advances in silicon surface characterisation using light beam
injection technique Acciarri M Pizzini S Simone G Jones D Palermo V E_MRS SPRING MEETING June 1999 Posters bull Spontaneous nano-wire growth on silicon Palermo V Jones D SMARTON workshop Leuven (Belgium) October 2002 bull Ordered circles of nano-islands on silicon from CO adsorption Palermo V Jones D TRENDS IN NANOTECHNOLOGY 2002 Santiago de Compostela (Spain) September 2002 bull Production of nanostructures of silicon on silicon by atomic self-
organisation Palermo V Jones D EUROMAT Rimini (Italy) June 2001 bull Etching holes and anisotropic corrosion on silicon [100] Palermo V Jones D Susi E Asoli B SILICON WORKSHOP Genova (Italy) February 2001
92 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull Morphological and electrical characteristics of damaged silicon surfaces
Susi E Castaldini A Cavalcoli D Cavallini A Jones D Palermo V SILICON WORKSHOP Genova (Italy) February 2001 bull Effect of HF etching on the roughness of a silicon surface Palermo V Jones D NATIONAL CONFERENCE ON PHYSICS OF MATTER Genova (Italy) June 2000 bull Effect of Native Oxide Desorption upon the Surface Morphology of Si[100] by STM and LEED Palermo V Jones D SILICON WORKSHOP Genova (Italy) February 2000 bull Characterisation of silicon surfaces for microelectronics through STM measurements Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999
VPalermo 93
Contact Vincenzo Palermo ISOF- Institute for Organic Synthesis and Photoreactivity Via Gobetti 101 40129 Bologna ITALY Tel +39-051-6398336 Fax +39-051-6398349 Mail palermoisofcnrit
94 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Index
1 Introduction nanotechnology and the future of computers
2 Silicon surfaces
3 STM and other surface analysis techniques
4 Surface modification of silicon in liquid Nano-hole creation
5 Surface modification of silicon in vacuum void creation and oxide desorption
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and
nano-lines
7 Conclusions and proposals for future work
2 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 3
Universitagrave degli studi di Bologna
Creazione di isole fili e fori nanoscopici su superfici di silicio per microelettronica
Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Tesi di Dottorato di Ricerca Scienze Chimiche
XV ciclo
Presentata da Vincenzo Palermo
RelatoreProf Alberto Ripamonti
Co-Relatore Dott Derek Jones
Coordinatore di Dottorato Prof Goffredo Rosini
Marzo 2003
4 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 5
OMNIA IN MENSURA ET NUMERO ET PONDERE
Sapientiae Salomonis 1120
6 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 7
1 Introduction nanotechnology and the future of computers The motivations of nanotechnology research
In 1965 an electronic engineer named Gordon Moore one of the future founders of Intel
noted that the performance of computers and their complexity doubled every 18 months
and foresaw that computer power would continue to grow exponentially over the following
years
This prevision quite provocative for its time actually came about and gained the name of
ldquoMoorersquos Lawrdquo and continues to hold for the trends of todayrsquos computer industry Since
1965 the number of transistors present in an integrated circuit (IC) has increased from
several hundred to more than ten million and the minimum size of transistor elements has
shrunk from several millimetres to asymp130 nanometres (fig 11)
Devices of such tiny dimensions are actually fabricated using lithographic techniques
where light is passed through an optical mask to react with a photo-sensitive layer (resist)
Fig11 Evolution of the number of transistors present on commercial computers [1]
8 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
on the silicon wafer This resist is then selectively removed and used as a mask for
processing the silicon surface (fig 12 left) The maximum resolution attainable depends
upon the wavelength used and current technology is near to its intrinsic resolution limit
On the other hand there is strong scientific and economic demand for further development
in IC miniaturization to obtain more powerful and complex computers Besides every-day
life applications more powerful computers are fundamental for much scientific research
such as climate change tracking genome sequencing and fluid dynamics Increased
miniaturization is also fundamental for reducing power which has to be dissipated by the
chips which run at progressively higher frequencies Energy consumption by
microelectronic devices is already an issue and represents one of the main obstacles for
the continuing growth in wireless communication (cell phones portable computers CD
and DVD players digital cameras etc)
Thus it is expected that new production methods different from current lithographic ones
will be developed methods which allow modification of a surface well below the 100 nm
limit and even down to single atom manipulation Techniques such as Scanning
Tunnelling Microscopy and Atomic Force Microscopy are already capable of moving
single atoms (see fig 12 right) but unfortunately building a working nanodevice in this
way would take a very long time and these techniques are difficult to apply to large scale
production
Nowadays thousands of researchers are working in the nanotechnology field towards a
new generation of microelectronic devices Several possible solutions are competing for
tomorrowrsquos computer architecture and there is still no clear winner It is likely that the
final solution will be the combined use of different techniques and components (including
molecules nanowires and nanodots) as they become available together with conventional
Fig 12 Left conventional litographic process [2] Right atomic manipulation of iron atoms on copper [3]
VPalermo 9
silicon technology
Below is a brief summary of the most recent developments in nanotechnology and
nanoscience
Actual trends in nanotechnology
Perhaps the most fascinating idea for nanodevice construction is to use one single
molecule working as a complete device The first molecular diodes (ie molecules
conducting current only in one direction) were created in 1997 in 1999 a molecular fuse
and a molecular transistor were demonstrated although there was no possibility of wiring
these devices to external contacts In April 2001 James Heath and his group at UCLA
fabricated an array of overlapping crossbars and placed a small molecule of rotaxane
between each crossbar (fig 13 left) This composite molecule is made up of two
component parts the main rod-like molecular axis and a mobile ring ldquothreadedrdquo on it like a
bead on a necklace and can function as a molecular switch A working 16-bit memory
circuit was constructed using these molecules For a brief review of these works see [4] In
June 2002 a single molecule transistor was built by connecting an organic molecule to two
metal contacts the molecule contained one or two atoms of a transition metal (cobalt or
vanadium) forming the active region of the device supported by an organic backbone [5]
Fig13 Working nanodevices Left schematic representation of rotaxane molecules between crossed nanowires [4] Right SEM image of semiconductor nanowires forming a small circuit [9]
10 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another approach to nanodevice fabrication has become possible through the discovery of
carbon nanotubes which were observed for the first time in 1991 by a Japanese electron
microscopist studying the material deposited during arc-evaporation synthesis of fullerenes
[6] They consist of a graphite-like carbon seamless cylinder with a diameter of several
nanometers and lengths of up to a millimetre Carbon nanotubes are very stable can
behave as metals or semiconductors and can host other molecular or ionic species thus
modifying their electrical behaviour In 2001 Avouris and his group reported the first
circuit made with a single nanotube [7] A few months later Cees Dekker presented a
nanotube-based transistor able to amplify an input signal by a factor of ten and built
several logic circuits using these nanotube transistors [8]
One problem with carbon nanotubes is that it is very difficult to control their electronic
properties ie their metallic or semiconducting behaviour An alternative to carbon
nanotubes are semiconductor nanowires Silicon nanowires can be made using a laser to
vaporize the silicon together with a metal catalyst like iron or gold The vapour condenses
in nanosized drops of silicon and metal from which the wires slowly grow out as more
silicon is adsorbed In 2001 a group at Harvard University [9] created a transistor by
crossing two different nanowires After this the same group arranged four nanowires in a
noughts and crosses grid creating something like a 4-bit memory (fig 13 right) Even
metallic nanowires made of platinum and silver can be used in a crossed configuration to
store information [10]
There are some issues common to all these new technologies though First it is difficult to
imagine these methods applied to large-scale production Up to now the insertion of a
molecule between two electrodes is an occasional lucky event while nanotubes and wires
have to be positioned on the surface creating the appropriate contacts on them manually
The large-scale production of integrated circuits using these building blocks will not be
straightforward and does not seem imminent
Another issue is of an economic and not a scientific or technological nature Since 1965
the cost of IC manufacturing plants has sky-rocketed If the cost of semiconductor
production plants continues to rise exponentially in a few years such plants will cost up to
$20 billion This is a sizeable investment even for large companies such as IBM or Intel
For this reason it is likely that IC companies will resist changing to completely new
technologies closing down their existing plants As it is clear that silicon will remain the
VPalermo 11
fundamental raw material of the IC industry for the foreseeable future nanotechnology
developments for microelectronics will need to be silicon-compatible In these early days
of nanotechnology the most valid approach would seem to be the addition of molecular
functions to existing silicon technology ndash using the latter as a foundation on which to build
Fabrication of self-organised structures on silicon
The possibility of using the phenomenon of atomic or molecular self-organization to create
nanostructures on silicon has already been demonstrated The clean silicon surface shows
in some cases a high degree of order and complex surface reconstruction as will be
described later Several different ordered structures form spontaneously on this surface
such as series of monatomic steps or boundaries between reconstructed areas It has been
demonstrated that it is possible to use these structures to fabricate ordered nanodots and
nanolines on the surface [11] More recently well-defined nanometric patterns have been
obtained with selective etching of silicon using nitric oxide [12]
In this study the possibility of creating different types of nanostructures on the silicon
surface is explored Methods had to be developed which were
- Simple They must not need complex masks or lithographic steps to create the
structure but rather exploit self-organisation phenomena
- High resolution the silicon surface modifications should be on a scalelength of below
100 nm
- Fast billions of nanostructures have to form over the whole surface simultaneously to
be compatible with large-scale production requirements
- Cheap they must not require expensive equipment (such as e-beam lithography) but
exploit simple chemical andor physical treatments to produce nanostructures on the
silicon surface
In Chapter 2 the main characteristics of silicon are described Chapter 3 provides a
summary of the techniques used for this research Chapter 4 examines the chemical etching
of silicon in different liquid environments and the effects of this etching on the surface at a
nanoscopic level with the creation of nanoholes
12 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Chapter 5 describes the growth in ultra-high vacuum (UHV) of nanoscopic voids and
islands on the silicon surface and the effect of surface oxide on this growth
Chapter 6 discusses the modification of silicon surfaces in UHV following the adsorption
of molecules and thermal heating to produce nanoislands and nanolines on silicon
The overall conclusions of our work are summarised in Chapter 7
Finally we will give some conclusions based on the results obtained and discuss possible
applications of the methods developed
Bibliography
[1] From wwwintelcom
[2] From wwwsematechorg
[3] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
[4] Molecules Get Wired Service RF Science 294 (5551) 2442-2443 Dec 21 2001
[5] Coulomb Blockade And The Kondo Effect In Single-Atom Transistors Park J
Pasupathy AN Goldsmith JI Chang C Yaish Y Petta JR Rinkoski M Sethna JP
Abruna HD Mceuen PL Ralph DC Nature 417 (6890) 722-725 Jun 13 2002 Kondo
Resonance In A Single-Molecule Transistor Liang WJ Shores MP Bockrath M Long
JR Park H Nature 417 (6890) 725-729 Jun 13 2002 Nanotechnology - Electronics
And The Single Atom De Franceschi S Kouwenhoven L Nature 417 (6890) 701-702
Jun 13 2002
[6] Smallest Carbon Nanotube Ajayan PM Ijima S Nature 358 (6381) 23-23 Jul 2 1992
[7] Carbon Nanotube Inter- And Intramolecular Logic Gates Derycke V Martel R
Appenzeller J Nano Letters 1 (9) 453-456 Sep 2001
[8] Logic Circuits With Carbon Nanotube Transistors Bachtold A Hadley P Nakanishi T
Dekker C Science 294 (5545) 1317-1320 Nov 2001
VPalermo 13
[9] Logic Gates And Computation From Assembled Nanowire Building Blocks Huang Y
Duan XF Cui Y Lauhon LJ Kim Kh Lieber CM Science 294 (5545) 1313-1317
Nov 9 2001
[10] Formation And Disappearance Of A Nanoscale Silver Cluster Realized By Solid
Electrochemical Reaction Terabe K Nakayama T Hasegawa T Aono M Journal Of
Applied Physics 91 (12) 10110-10114 Jun 15 2002
[11] Fabrication And Integration Of Nanostructures On Si Surfaces Ogino T Hibino H
Homma Y Kobayashi Y Prabhakaran K Sumitomo K Omi H Accounts Of Chemical
Research 32 (5) 447-454 May 1999
[12] Ultrafine And Well-Defined Patterns On Silicon Through Reaction Selectivity
Prabhakaran K Hibino H Ogino T Advanced Materials 14 (19) 1418-1421 Oct 2
2002
14 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 15
2 Silicon surfaces
The name silicon (silicio in Italian) comes from the latin word silex Amorphous silicon
was first isolated by Berzelius in 1824 by reaction of potassium with silicon tetrafluoride
Thirty years later the first crystalline silicon was prepared Silicon makes up 25 of
earthrsquos crust and is the second most abundant element after oxygen Elemental silicon is
not found in nature occurring as silicon oxide (sand quartz amethyst flint etc) or
silicates (asbestos clay mica etc) Perhaps no other element and its compounds has such
a wide range of uses Silicon compounds such as sand and clay are used in the building
industry as refractory materials for high-temperature applications and for enamels and
pottery Silica is the main component of glass silicon carbide is an important abrasive and
silicones are commonly used polymers and lubricants
Here the most interesting use of silicon of course is for the production of
microelectronics devices For this application silicon of high purity (999999) and of
high crystallinity is needed Table 21 lists some of the physical characteristics of silicon
High purity polycrystalline silicon is produced by the reaction of gaseous trichlorosilane
with hydrogen in a furnace Then to prepare a single-crystal of silicon the so-called
Czochralski method is commonly used
Polycrystalline silicon is melted in a quartz furnace at 1415degC in an argon atmosphere
Then a seed of single-crystal silicon is lowered into contact with the melt and slowly
pulled out In this way the crystal grows and a crystalline cylindrical ingot several metres
long is created from the initial seed
After cooling down the ingot is sliced into thin silicon wafers The wafer surfaces are
polished using a counter-rotating lapping machine in an Al2O3 slurry until the surface is
very flat and shiny ready for the lithographic processes
16 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another way to obtain single crystal silicon is the Floating Zone (FZ) method in which a
silicon cylinder is slowly passed through a heating ring The area inside the ring melts and
solidifies smoothly crystallising as it comes out of the ring yielding a single silicon crystal
Microelectronic devices are built on the silicon surface which is the surface of interest
here Unfortunately silicon surfaces are normally quite dirty and uneven at the atomic
scale Atmospheric oxygen and humidity react with silicon surfaces creating a thin layer of
oxide (called ldquonative oxiderdquo) which is usually irregular and full of defects Different kinds
of contaminants also adsorb onto the surface These are usually small organic molecules
and microscopic dust particles A clean surface on exposure to the atmosphere is
completely covered with gas molecules in less than 10-9 seconds If the pressure is
reduced letrsquos say to 10-6 mbar this time increases to 1 second This is the reason why to
study a clean surface we have to work in UHV at pressures below 10-10 mbar
The atoms in the silicon crystal have a diamond-like structure each atom having 4 bonds
in a tetrahedral sp3 arrangement with bond angles of 10947 degrees At the crystal
surface some atoms will have non-bonding orbitals ldquodanglingrdquo in the vacuum ie sp3
orbitals with a lone electron which are highly reactive These orbitals are known as
dangling bonds To minimize surface energy the surface will reorganize by decreasing
the number of dangling bonds
Table 21 Physical data of silicon [1]
Atomic Weight 2809 Lattice constant (A) 543095 Crystal structure Face-centered cubic
(diamond) Melting point 1415 degC
Density (gcm3) 2328 Boiling point 2355degC Atomscm3 50E22 Minority carrier
lifetime (s) 25E-3
Dielectric Constant 119 Specific heat (Jg degC)
07
Breakdown field (Vcm)
~3E5 Thermal conductivity (Wcm degC)
15
Electron affinity x(V)
405 Vapour pressure (Pa) 1 at 1650degC 1E-6 at 900deg C
Energy gap (eV) at 300K
112 Reactivity Inert to acids Attacked by halogens and alkaline
solutions Intrinsic carrier
conc (cm-3) 145E10 Oxidation states +4 -4
Intrinsic Debye Length (microm)
24 Energy of a Si-Si bond (eV)
232
Intrinsic resistivity (Ω-cm)
23E5
VPalermo 17
Dangling bond densities and positions and thus the type of surface reconstruction will
depend upon crystal orientation as well as the temperature and kinetics of the system
Fig 21 shows a drawing of the main faces of a silicon crystal The angle α between any
(11n) face and the (100) face can be calculated from 2cos 2 += nnα The angle
between any (11n) face and the (111) face can be calculated from
)2(3)2(cos 2 ++= nnα
The chemistry and physics of the faces are very different a brief description will be given
for the most important orientations
Si (100)
On the (100) surface each atom has two Si-Si bonds connecting it to the bulk and two
dangling bonds pointing outward Surface energy is reduced by the dimerisation of the
surface atoms through overlap interaction of one dangling bond per atom forming rows of
dimers aligned along the (110) direction This is the well-known ldquo2x1rdquo reconstruction of
this silicon surface The symmetric dimers would make the silicon surface metallic but to
reduce surface stress the dimers tend to buckle and the surface is thus semiconductive It
took several years to understand that the dimers are buckled because at room temperature
Fig 21 Schematic view of the principal orientations of a silicon surface Surface atoms are white bulk atoms are black dangling bonds are gray [2]
18 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
they shift easily from one buckling direction to the other thus appearing symmetric under
STM observation Fig 22 shows an STM image of the 2x1 reconstructed surface
Even almost perfect (100) surfaces have a certain number of monoatomic steps and the
dimer rows on atomic layers are aligned at 90deg to those on adjacent layers Dimer rows are
thus perpendicular or parallel to the step When the dimers on the upper side of the step
are parallel to the step the step is called SA if they are perpendicular the step is called SB
Because of this symmetry SA step edges will be smoother compared to the more broken
fragmented edges of the SB steps
A common defect on the Si(100) surface is the presence of nickel contamination which
appear as missing dimers in STM images This type of contamination is so critical that
even if the silicon sample is only briefly brought into contact with stainless steel tools
(tweezers for example) the 2x1 reconstruction of the surface can be blocked
Silicon atoms can diffuse easily over the silicon surface as monomers and dimers
especially at elevated temperatures The anisotropy due to the 2x1 reconstruction causes a
difference in the diffusion energies of adsorbates over the surface Diffusion of these
silicon species along dimer rows for example will be much easier A list of diffusion
energies for monomers and dimers is given below [2]
Diffusion on Si(100) 2x1 Ed (eV) Monomers along dimer rows 06 Monomers across dimer rows 085 Monomer formation energy 18 Dimer along dimer rows 11 Dimer across dimer rows 15 Dimer formation energy 26 Dimer binding energy 076 Vacancies along dimer rows 17 Vacancies across dimer rows 19
SA
SB
Fig 22 STM image of a 2x1reconstructed silicon surfaceshowing the dimer rows and steps40x35 nm Nickel-induced defectsare visible as dark spots SA and SBsteps are indicated
VPalermo 19
So the diffusion energy for both monomers and dimers is nearly 40 greater if they have
to cross a dimer row This difference reduces to sim10 for vacancy diffusion
Si(113)
The (113) surface can be imaged as a sequence of alternating (100) and (111)-like
structures with two and one dangling bonds on alternate atoms respectively Interest in the
(113) surface is scientific as it has been used to study the energetics of the (100) and (111)
surfaces as well as for surface adsorption experiments
Si(100) surfaces can easily develop (113) facets
Si(111)
This surface besides being the first one imaged with STM with atomic resolution is one of
the most studied because it is the best cleavage face of silicon and because it shows one
of the most complex and elegant reconstructions in surface science
All Si-Si bonds in the silicon crystal are perpendicular to a (111) plane so this face will
have the lowest number of dangling bonds created per unit area In fact each Si atom on a
(111) surface shows a single dangling bond oriented perpendicular to the surface and
bonded to three back atoms These three bonds for each surface atom account for the great
chemical and physical stability of the Si(111) surface Surface energy is 009 eV Aring-2
compared to 015 eV Aring-2 for Si(100)
For energy minimization this surface reconstructs forming a huge 7x7 lattice cell
containing 102 atoms described by the Dimer-Adatom-Stacking fault model (DAS) For a
detailed description of cell structure see fig 23
The cell described by this model is very complex being composed of three kinds of atoms
adatoms rest atoms and corner hole atoms Furthermore a subsurface stacking fault is
present in one half of the cell making the two halves of the unit cell look different under
STM (Fig 24) It took 26 years of research to completely understand the exact structure of
the 7x7 reconstruction
Cleaving a silicon crystal along a (111) plane produces a metastable 2x1 reconstruction
the 7x7 reconstruction is easily obtained by flashing at high temperature in UHV At T gt
20 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
830degC a disordered 1x1 phase covers the surface Cooling down to 800degC leads to the
formation of the 7x7 phase If the cooling process is too rapid small 7x7 domains nucleate
and a disordered 1x1 phase is preserved between domain boundaries
Si(110)
Even though as mentioned before the (111) plane is the favoured cleaving plane of
silicon thin (100) commercial wafers will not break along this plane because the angle
between (100) and (111) is too far from 90deg (see table 22) Instead they will break along
the (110) plane because it is perpendicular to the (100) surface Each surface atom on
Si(110) has a Si-Si bond pointing downward one dangling bond pointing outward and two
Si-Si bonds parallel to the surface in a zig-zag pattern (see fig 21) Cleaved (110)
surfaces are disordered but upon annealing at high temperatures an ordered complex 16x2
Fig23 Scheme of the 7x7 DAS model [2] In each unit cell there are 9 dimers 12 adatoms and a stacking layer fault The force driving this complex reconstruction is the minimization of dangling bonddensity The DAS model shows the lowest number of dangling bonds (19) of all possiblereconstructions 12 dangling bonds are at the adatoms 6 at the rest atoms and 1 at the corner hole atom This surface is metallic
VPalermo 21
reconstruction takes place The surface appears as a series of long ridges and valleys
parallel to each other Eventually tilted facets of orientation (17 15 1) can form on this
surface The adsorption of Ge atoms on this surface leads to the formation of self-
assembled nanowires [3]
Table 22 Angles in degrees between different silicon faces [2]
Orientation
113
110
111
100
100
2524
9000
5474
0
111
2950
3526
0
110
6476
0
113
0
Fig 24 STM image of a Si(111)surface with 7x7 reconstruction A unitcell with its adatoms is highlightedImage size 13x13 nm
22 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] Weast RC Handbook Of Chemistry And Physics (Chemical Rubber Co Cleveland
1972)
[2] Dabrowski J Mussig H Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] The Structure Of Clean And SiGe-Covered Si(110) Surfaces Butz R Luth H Surface
Science 365 (3) 807-816 Oct 1 1996
VPalermo 23
3 STM and other surface analysis
techniques
Scanning Tunneling Microscopy
Since the invention of the optical microscope at the end of the 16th century the possibility
of examining surfaces at higher and higher magnification has fascinated mankind
Development of the technique continued and towards the end of the 19th century optical
microscopes were as good as todays standard instruments The physical limits of the
wavelength of visible light (350-800nm) had been reached
In the 1920s de Broglie showed that electrons can behave like waves and the use of these
particles for imaging with much higher resolution soon followed Atomic resolution using
this technique is only possible in the transmission mode with extremely carefully prepared
samples
In 1982 using the peculiar properties of piezoelectric materials Binnig and Rohrer brought
a metallic tip very very close to a silicon surface and scanned it across an extremly small
area (fig31) The tunneling of electrons from the tip into the sample or vice versa allowed
them to obtain a local density of electronic states (LDOS) map of that surface Although
theory (which treated the extreme point of the tip as a sphere) then excluded the possibility
of atomic resolution following a tip crash into the surface Binnig and Rohrer began to
observe the LDOS with atomic resolution For this discovery and their following work
they were awarded the Nobel Prize in Physics in 1986
24 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The basic principle of STM is very simple A metallic tip is scanned over a surface without
making ohmic contact and a tunneling current passes between the tip and the surface An
electronic circuit keeps this current constant by raising and lowering the tip during the
scan In this way recording the tip height at each point a three-dimensional image of local
density of electronic states (LDOS) of the surface can be obtained To explain the
extremely high resolution attainable by this simple technique quantum theory is needed
According to classical physics the current will flow between sample and tip only if they
are in physical contact If there is a vacuum gap between the two the electrons will simply
remain confined for example within the surface without the possibility of passing into the
tip
In quantum physics however the electrons have a certain probability of passing
(tunneling) across the gap appearing on the other side of the gap in this way reaching the
tip It can be shown that the probability of an electron tunneling through a gap of thickness
z is
kzep 22)0( minusprop ψ h
φmk
2= (1)
where ψ(0) is the electron wavefunction at the surface-gap border m is the electron
mass=91x10-28 g and φ is the work function of the metal (ie the energy required to
remove an electron from that material For silicon it is 48 eV) The tunneling current thus
decays exponentially with z and is extremely sensitive to topographical imperfections
present on the scanned surface A rough formula giving the current as a function of z is [1]
zFS eEVI φρ 0251)( minusprop
Fig 31 Binnig and Rohrer with the first STM Image from IBM [2]
VPalermo 25
where ρs(EF) is the local density of states at the Fermi level on the given surface For
example the formula predicts that for silicon an increase in tip-surface distance of 1 Aring
will give a 95 decrease in tunneling current
This huge dependence of tunneling current upon the distance allows detection even of the
sub-nanometre changes in height given by the single atoms of which the surface is
composed and thus to resolve them in the LDOS images Of course this description of the
tunneling process is oversimplified and for a more accurate one the electronic states of
the tip of the sample and their interaction have to be taken into account Fig 32 shows a
schematic representation of the interaction between tip and sample orbitals
The exponential decay of current with distance also yields high lateral resolution If the tip
is approximated as a sphere of radius R and the current passing at the minimum tip-sample
distance is I0 then the current passing at a lateral distance x from this point will be
Rxk
eII 22
0
2minus
=
Assuming a tip radius of 100 nm the current is concentrated in an area sim15 nm wide at tip
apex
Very sharp tips with even smaller curvature radii can be produced in several ways Simple
Fig32 Schematic view of tip-sampleorbitals interaction a) no interaction b)equilibrium c) sample positive d) tippositive [1]
26 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
tungsten wires mechanically cut are capable of obtaining atomic resolution on graphite in
air but for more disordered and rougher samples sharper and more reproducible tips are
needed
STM tips are mostly made by electrochemically etching a W or Pt-Ir wire The tips we
used were prepared using methods based on the work of Fotino [3]
A tungsten wire 038 nm diameter is immersed in a KOH solution(06M) with a thicker
tungsten wire used as a counterelectrode The cathodic and anodic reactions involved in the
etching are
Cathode 6H2O + 6 e- rarr 3H2(g) + 6 OH-
Anode W(s) + 8 OH- rarr WO42- + 4H2O + 6 e-
A potential of 3V ac is applied to the tungsten and the wire is immersed in the solution
until a constant current of sim100 mA is established The etching rapidly removes metal
shaping the wire end as a sharp tip When the potential reaches 11 V the coarse tip etching
is finished The wire is then removed from the solution carefully inserted into an
insulating plastic tube leaving only the tip exposed and re-immersed in the solution with
the tip pointing upwards A more gentle etching is thus made to reduce tip radius Usually
5 minutes etching at 07 V ac is used In this configuration very small hydrogen bubbles
formed on the tip sides sliding upwards with a ldquohoningrdquo effect on the tip
This procedure yields extremely sharp and reproducible tips at the microscopic level
After the etching the tip is thoroughly rinsed in ultrapure water then dipped into
concentrated HF to remove surface oxides and hydroxides [4] The tip is dried with
nitrogen inserted into the UHV system and degassed overnight at sim150degC
The possibility of measuring sub-nanometric distances would be useless without being able
to control tip movement over such a minute scale To scan the tip over the surface a
piezoelectric scanner is used Piezos are usually made of an alloy of PbZrO3 and PbTiO3 a
material which contracts or expands when a voltage is applied to it The Omicron
instrument used in our laboratory has three such piezo scanners for xy and z tip motion
allowing one to scan the tip over the surface with sub-Aringngstrom precision (fig 33)
To isolate the instrument from ambient vibrations the whole STM stage is suspended upon
four springs which eliminate all frequencies above 1 Hz and surrounded by a crown of
VPalermo 27
copper wings and fixed magnets Parasitic currents generated by the magnets into the
copper wings contrast every movement of the stage and efficiently block stage vibrations
The STM can be used not only to explore surface topography but to measure the IV
characteristics of single atoms or molecules on the surface (Scanning Tunneling
Spectroscopy or STS) Furthermore it can be used to modify the surface with voltage
bursts digging into it or delicately moving single molecules or atoms over a surface [5] It
can be used in vacuum in air and with proper lateral isolation of the tip even in liquids
A major drawback of STM is that it works only on conducting and semiconducting
surfaces and thus cannot be used on many surfaces of biological and chemical interest
Another instrument more suitable for these and other applications is the Atomic Force
Microscope (AFM)
Atomic Force Microscopy
AFM was invented in 1986 by Binnig Quate and Gerber after calculating the possibility of
building a cantilever with a force constant of the same order of magnitude as that of a
chemical bond
In AFM a tip mounted on a microscopic cantilever (usually made of Si3N4 fabricated with
optical lithography) is brought close to a surface When the tip touches the surface the
cantilever is very slightly deflected upwards The movement is measured by observing the
Y-PIEZO
SILICON SAMPLE
Fig 33 A picture of theSTM used for theexperiments The tripodpiezo scanner is shown
Z-PIEZO
X-PIEZO
TIP
28 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
deviation of a laser beam hitting the upper face of the cantilever Fig 34 provides a
schematic view of the principle of AFM
The typical force constant of the cantilever varies from 00006 to 2 Nm the typical
resonance frequency is 3 to 120 kHz The AFM tip can apply a force on the sample of up
to 10-9 N The AFM can be used on conductive or insulating surfaces in vacuum air or
liquids Furthermore the tip can be modified to sense electrostatic potentials (electric force
microscopy) or magnetic fields (magnetic force microscopy) it can even be functionalized
with complex molecules such as proteins to interact with biological surfaces
A drawback of the AFM is that the force it exerts can damage the surface under
observation especially if the sample is soft (as in the case of cell membranes for
example) This problem can be overcome using the instrument in tapping mode (where the
tip does not move laterally during its brief contact with the surface) or in non-contact mode
in which the tip oscillates above the surface during the scan and the changes in its
frequency due to interaction with surface are monitored The shifts in the oscillating
frequency of the cantilever due to tip-sample interaction are then used for imaging the
surface In this mode interaction of the tip with the surface is minimal and soft samples
can be imaged
STM and AFM are the main techniques used for this work A brief description of other
techniques used occasionally is given below
Fig 34 Scheme of an atomic force microscope
VPalermo 29
Low Energy Electron Diffraction (LEED)
Electrons with energies in the 20-500 eV range are diffracted by a crystalline surface the
diffraction peaks are visualized on a fluorescent screen This technique probes the long
range order of the surface up to a depth of several nanometres
X-Ray Photoelectron Spectroscopy (XPS)
XPS allows both qualitative and quantitative chemical analysis of the elements present on
or near the sample surface
An X-ray source is used to photoionize the atoms on a surface and produce photoelectrons
By measuring the kinetic energy of the photoelectrons the binding energy of the electronic
levels can be calculated This energy will depend on the chemical environment of the
surface atoms
Although the soft X-rays used penetrate to a depth of ~2000Aring the sampling depth of the
technique is determined by the mean free path of the photoelectrons which allows their
escape from only the first 10-100Aring
Secondary Ion Mass Spectroscopy (SIMS)
High and low energy ions (primary ions) are used to bombard a sample and remove surface
atoms and ions The ionic fragments removed (secondary ions) are then analysed by a mass
spectrometer The surface can be consumed during the measurement and profiles obtained
giving concentrations of the materials composing the sample at different depths (depth
profiles)
A popular variant of SIMS is TOF-SIMS In this technique the secondary ion masses are
measured by a time-of-flight (TOF) measurement The secondary ions generated by the
bombarding primary ions are accelerated to a constant kinetic energy and then move
through a field-free space before they reach the detector where their intensity is measured
as a function of flight time Since ions with different masses have different velocities at a
30 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
given kinetic energy the measured flight times of the ions can easily be converted to their
masses The static nature of this latter technique allows mass spectroscopy surface analysis
with minimal damage to the surface
Bibliography
[1] Chen CJ Introduction To Scanning Tunneling Microscopy (Oxford University Press
Oxford 1993)
[2] From wwwibmcom
[3] Tip Sharpening By Normal And Reverse Electrochemical Etching Fotino M Review
Of Scientific Instruments 64 (1) 159-167 Jan 1993
[4] A Convenient Method For Removing Surface Oxides From Tungsten STM Tips
Hockett LA Creager SE Review Of Scientific Instruments 64 (1) 263-264 Jan 1993
[5] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
VPalermo 31
4 Surface modification of silicon in liquid
Nano-hole creation
Liquid treatments of silicon wafers are very common in the integrated circuit (IC)
manufacturing industry They are used to clean and improve surface uniformity to create
and etch protective oxide layers and to remove photo-resist layers
Crystalline silicon with its native oxide layer is very stable and is resistant to many acids
It is easily attacked by hydrofluoric acid (HF) and alkaline solutions
The thin (~2 nm) passivating layer of native oxide (SiO2) is formed on exposure to the
atmosphere This surface layer contains many defects and contaminants so it is usually
chemically stripped and substituted with a better chemically-formed protective oxide
The most common silicon cleaning procedure is the RCA method named after the Radio
Corporation of America [1] It consists of two steps in the first one the surface is treated
with a hot alkaline solution (H2OH2O2NH4OH 411) to remove particles from the
surface following this a hot acidic solution (H2OH2O2HCl 411) is used to remove
metal contamination Other well-known cleaning methods are IMEC (a sequence of
cleaning steps in H2OO3 and dilute HF) or the Pirana etch (a hot 41 mixture of H2SO4
H2O2)
32 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The standard RCA clean removes surface contaminants etches the native oxide and
oxidizes the silicon surface leaving a uniform layer of silicon oxide which better protects
the surface from further contamination
Etching with fluorine-based solutions
Hydrofluoric acid is one of the most common reagents used in the treatment of silicon
wafers both in the research field and in industrial processes A rapid dip in dilute HF is the
simplest way to remove the native oxide from Si(100) and leaves the surface passivated by
a layer of Si-H bonds Because of the low polarization of Si-H bonds the Si-H layer is
stable even for several days protecting the surface from contamination It has often been
assumed that this short etch does not significantly change the surface morphology of the
silicon substrate[2] even though a prolonged dip in dilute HF leads to surface roughening
[3]
Although dilute HF roughens the Si(100) surface at the atomic scale [45] immersion in
concentrated HF (49) etches the surface oxide without attacking the Si surface
uncovering in this way the buried SiSiO2 interface The final counter-intuitive result is
that dilute HF etches the silicon while concentrated HF leaves the crystalline silicon
untouched [3]
Etching Si with fluorine-containing solutions at different concentrations and pH can
produce different morphologies from rough surfaces to flat nearly ideal Si-H terminated
surfaces
Hessel et al and Higashi et al demonstrated in 1991 that very flat Si(111) surfaces can be
obtained using 40 NH4F while etching with HF always yields rough surfaces The
surface becomes smoother because the etchant rapidly attacks Si atoms at step borders
thus removing surface kinks and irregularities in a step-flow mechanism [6 7] Later on
even smoother and more perfect surfaces were obtained by removing oxygen from the
solution after it was discovered that oxygen dissolved in 40 NH4F initiates the formation
of triangular etch pits It was not possible to obtain flat surfaces by etching Si(100) with
ammonium fluoride solutions which leads to the formation of small 2x1 dimer-row
reconstructed (100) terraces together with (111) facets [8]
This difference is caused by the different hydride terminations prevailing on the (100) and
(111) faces While the ideal Si(111)-H surface is monohydride terminated the more
VPalermo 33
reactive dihydrides predominate on the Si(100)-H surface making it more vulnerable to
etching The etching reaction is thus strongly anisotropic etching (100) facets faster than
(111) thus producing (111) microfaceting on Si(100) crystals
A more uniform Si(100) surface can however be prepared by etching at low pH with an
HFHCl mixture [9] or by using very dilute HF solutions and ultrapure water with low
dissolved oxygen and carbon contents [10]
Electrochemical etching can also be used applying anodic or cathodic bias to the silicon
to obtain different morphologies [11] by varying the potential isotropic or anisotropic
etching is observed The aforementioned results show that despite the simplicity of the
reactants fluoride etching of silicon is quite a complex reaction
Fig 41 Chemical etching of silicon
HO+H
H HO
H
H
H
FSi
Si
SiSi
+H2O
-OH-
+F-
H2O H H
Si H
F OH
H Si
Si Si
-H2
+H2O
H
H
H
FSi
Si
SiSi
H
H
H
OHSi
Si
SiSi
H
H
H
H Si
Si
Si Si
Etching mechanism of silicon
HF rapidly dissolves the SiO2 passivating layer on silicon leaving the surface almost
completely hydrogenated [12] After this two different types of reactions etch the silicon
simultaneously one chemical and the other electrochemical [13] The overall etching
mechanism can be schematized in two stages (see Figure 41)
i) Si-H bonds are substituted by Si-F or Si-OH bonds creating a partial charge on the
surface silicon atom and polarizing its Si-Si backbonds
ii) these polarised backbonds are then more easily attacked by HF or H2O After
rupture of the Si-Si bond the atom is removed leaving behind new Si-H
terminations and the reaction can start again
34 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
These reactions take place although at different rates on both Si(100) and Si(111)
Stage i) is usually the rate-determining step of the reaction and the stability of the Si-H
bonds depends upon the pH the concentration of nucleophilic species in solution and an
eventual potential applied to the crystal
For pH gt5 as in the case of concentrated NH4F solutions reaction begins with attack by
water to give Si-OH (step ArarrC) The -OH group is rapidly substituted by fluorine with
polarization of the underlying Si-Si bonds These bonds are then easily attacked by water
the silicon atom being released into solution as HSiF(OH)2 The Si-OH rarr Si-F substitution
is not fundamental for the reaction and etching can proceed even for Si-OH terminated
atoms but XPS measurements showed the presence of a certain number of Si-F bonds
remaining Furthermore fluorine seems to have a catalytic effect on Si-H substitution as
indicated by the dependence of the etch rate upon the F- concentration at least for pH
values between 4 and 8
Si-F bonds can be easily removed by a water rinse In the case of strongly alkaline
solutions (pH=14) OH- groups act directly as nucleophiles and no fluorine is needed to
catalyze Si-H bond rupture
At pH lt4 almost no free F- ions are present in solution and the etch rate is very small at
pH lt2 all etching reactions are very slow and this explains the stability of Si crystals in
concentrated (50 ww) HF solutions
Si Si
Si Si
H
H
H
H Si
Si
Si Si
H
H
H Si
Si
SiSi
H
H
H
OH-H+ -e
(F-)
-H+
-e
+H2O hellipas in fig1
Fig 42 Electrochemical etching of silicon by H2O
The electrochemical etching of silicon involves electron transfer from the surface atoms to
the valence or conduction band of the crystal (according to the type of doping of the
silicon) Several studies have been carried out at different electric potentials In the case of
p-silicon or n-silicon at anodic potentials where positive charges (h+) are available in the
crystal a silicon-centred radical can be created by capture of a hole from the bulk (fig
42) The F- ion does not participate in the reaction but is thought to contribute through
VPalermo 35
electrostatic interactions by lowering the energy of the interaction step After the
formation of the Si-OH group the reaction proceeds as shown in the scheme of fig 41
Matsumura et al [4] proposed that not only water but HF2- molecules also play a major role
in electrochemical etching of silicon leaving on the surface Si-F terminated bonds which
can be immediately attacked in an autocatalytic process (fig 43)
In the electrochemical reactions described above an external potential is applied to the
silicon crystal The chemical and electrochemical reactions in any case take place
simultaneously most of the time with the chemical path predominating at high pH Even
when no external potential is applied to the silicon partial electrochemical reactions can
take place at different ldquocathodicrdquo and ldquoanodicrdquo sites on the surface with an internal charge
exchange which ensures neutrality [11] This macroscopic silicon etching and hydrogen
bubble formation can lead to visually observable patterns on the surface when Si(100) is
immersed in ammonium fluoride even without applying a potential
Si Si
Si Si
H
H
H
F Si
Si
Si SiH
F F-H+ -2e
+HF2-
H
HF
SiSi
SiSi + F
F F
FH
H
H
F -H+ -2e
+HF2-
Fig 43 Autocatalytic electrochemical etching of silicon by HF2-
Inhomogeneities on silicon surfaces caused by electrochemical reactions and charge
transfer have been studied extensively because they are of fundamental importance in the
formation of porous silicon
Pore formation on silicon
When Si(100) or Si(111) are etched under anodic bias in fluorine-based solutions
microscopic pores form on their surface Several different morphologies of pores have
been observed with pore diameters ranging from 10 nm to several microns with depths of
several microns [14] Pore shape is very variable too ranging from ordered straight pores
to chaotic networks of branched pores (fig 44) Porous silicon has been known since the
fifties but it was only in 1990 that interest in this material increased following the
discovery that porous silicon layers were able to emit bright red light This led to a large
amount of research and now different classes of micropores can be reproducibly created
36 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Fig 44 Different types of Silicon micropores From ref [14]
mostly for optical and micromachining applications However there is still no unified
theory able to explain the nucleation and growth mechanism of all the different kinds of
pores
We will give a short description of some of these theories for more detail see Parkhutik et
al [15]
One model explains pore nucleation on the basis of physical processes such as hole
positive charge migration ion transport to the surface and small perturbations on the
silicon surface modelled as Fourier components The system is shown to be unstable and
some spatial frequencies that lead to pore nucleation evolve from the etching process
A second model focusses on stationary pore growth without explaining the nucleation
stage According to this model silicon dissolves preferentially at pore edges because h+
charges are attracted by the stronger electric field present at these edges
A third class of models explains pore growth as a Diffusion Limited Aggregation (DLA)
process where the random walk of h+ charge carriers through the depleted layer present at
the silicon-liquid interface controls pore shape
Finally the model by Carstensen Cristophersen and Foll [16] proposes that areas of the
surface of some characteristic size LCO are etched by synchronized ldquocurrent burstsrdquo in the
flow of h+ charges These bursts dissolve silicon through cyclic stages of surface oxidation
oxide removal and hydrogen passivation Areas where a burst has recently taken place are
less passivated and thus more likely to be etched again in this way the pore bottom
continues to dissolve while the pore walls are passivated and are thus less favourable areas
towards current bursts
VPalermo 37
EXPERIMENTAL RESULTS
In the following sections we will show some experimental results obtained from STM and
AFM measurements of fluorine-treated Si(100) surfaces In the first part the results of
mild etching using concentrated and dilute HF solutions at low pH are presented In the
second part the results of etching at high pH using ammonium fluoride are presented and
the mechanism of pore formation discussed
Etching of Si(100) in dilute and concentrated HF
Samples were cut from different areas of an 8-inch diameter p-doped silicon(100) wafer
(10 Ω-cm) supplied by MEMC Electronic Materials Each series of STM measurements
was carried out over at least six different areas on at least two identical samples Low
Electron Energy Diffraction (LEED) was used to check the surface cleanliness of the
samples before STM measurements
Table 41 summarizes the different treatments of each sample After etching with
electronic grade HF each sample underwent a final rinse in Ultra-Pure Water (UPW
resistivity gt18 MΩ-cm) Both the HF and the UPW were allowed to flow continuously
over the sample surface Some samples were not etched with HF at all but just washed with
UPW to observe the morphology of the native oxide layer (~2 nm thick) covering the
surface All of these processes were carried out under nitrogen to limit reoxidation and the
samples were then introduced from the nitrogen atmosphere directly into the vacuum
chamber and degassed overnight at sim150degC before LEED and STM measurements
STM images were obtained from each sample using the same measurement parameters
(sample bias 4 V feedback current 1 nA scan speed 800 nm s-1) These parameters
38 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
provided a satisfactory level of reproducibility for all the samples Measurements were
made over an area of 500x500 nm (image size 500x500 pixel) Slope correction was
carried out by subtracting row-wise and column-wise fitted slopes from the entire image
which gave better results than the simple subtraction of a fitted plane especially for the
rougher samples Following slope correction the rms roughness
sum minus=xy
hyxhN
22 ))((1σ
and the 2-D Fourier transform
)(22
2
)(4
)( vyuxi
xyeyxhvuF +sum∆
= π
π
were calculated for each image where N2 is the number of pixels composing the image
h(xy) is the surface height at each point ∆ is the distance between points h is the mean
height and u v are the spatial frequencies The radial power spectrum PS(f) of the STM image data is obtained from the angular
average of the squared Fourier transform with f 2 = u2 + v2
Fig 45 shows the STM images obtained from the various samples Sample A still covered
with its native oxide layer shows an irregular surface with RMS roughness of ~05 nm
(see Table 41) Observing the sample with LEED gave no diffraction pattern even at
relatively high incident electron energies because of the surface oxide coverage After 1
min etching in dilute HF (sample B) the morphology is similar to the original one though
Table 41 Sample treatments average RMS roughness and integrated area of power spectra
Sample Treatment RMS roughness (nm)
PS area (f lt 01 nm-1)
PS area (f gt 01 nm-1)
A Rapid dip in water 053 plusmn 013 366 032
B 1 min in HF 5 + 10 min in water 051 plusmn 008 272 034
C 30 min in HF 5 + 10 min in water 062 plusmn 008 1064 038
D 5 sec in HF 49 + 10 min in water 042 plusmn 004 183 022
VPalermo 39
Fig 45 STM images of each group of samples showing the topography of the silicon surface A) noetching original oxide surface B) after 1 min etching in dilute HF C) after 30 min etching in diluteHF D) after dipping in concentrated HF Grey scale indicates height of the surface from lower (black) to higher (white) The images are 250x250 nm ie representative portions of the images usedfor the roughness measurement and PSD analysis
some of the larger features have disappeared and the image quality is better maybe due to
improved tunnelling due to the cleaner surface The RMS roughness is comparable to that
of the original surface Clear diffraction patterns are visible using LEED though at quite
high energies (200 eV) After prolonged etching (sample C) the RMS roughness increases
to 062 nm and a long-range corrugation is visible on the surface even if the LEED pattern
is good
The samples dipped in concentrated HF (D) reveal the bare SiSiO2 interface which has a
disordered aspect and protrusions over a wide range of dimensions The quality of the
STM images of sample D is very good probably due to the cleanliness of the surface
40 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 E -0 2
1 E -0 1
1 E + 0 0
1 E + 0 1
1 E + 0 2
1 E + 0 3
0 0 0 0 0 1 0 1 0 1 0 0f (1 nm )
nm^4
A a s re c e iv e d
B e tc h e d 1 min
C e tc h e d 3 0 min
D e tc h e d HF 4 9
Fig 46 Log-Log plot of the averaged power spectra of the STM images for all the samples
which gives a more stable tunnelling junction The LEED pattern is excellent showing
clear diffraction peaks at energies as low as 37 eV comparable to that obtained after high
temperature cleaning in UHV
Fig 46 shows the power spectra of the samples The high frequency and low frequency
areas of the power specturm are considered separately Table 41 shows for each sample
together with the roughness the integrated area of the power spectrum for the high and low
frequency part
We first examine the differences between the samples in the low frequency part of the
spectrum (lt01 nm-1) In this spatial range the short dip in concentrated or dilute HF
(samples BD) seems to lower the roughness of the sample removing some of the larger
features of the native oxide surface visible in Fig 45a and thus decreasing the
corresponding part of the power spectrum Sample C however etched with HF for 30 min
shows an increase in roughness peaking at 0012 nm-1 (~85 nm corresponding to the
typical dimensions of the corrugations visible in Fig 45c) On this scale sample C has a
PS density of 435 nm4 compared to 101 nm4 for sample A
Samples A B and C have the same PS in the high frequency range showing that the fine
structure of the surface is not changed by the HF etching Sample D on the other hand
shows a decrease of roughness for frequencies gt 015 nm-1 This effect could be due to the
VPalermo 41
improved surface cleanliness after etching with concentrated HF which would give a more
stable STM junction thus reducing the high frequency noise in the image
Fig 45 and the analysis of the power spectrum of each sample shows that a rapid dip in
HF removes the native oxide but does not lead to major changes in the morphology of the
Si surface its only effect being the removal of some of the larger features present on the
original surface Prolonged etching on the other hand increases the RMS roughness of the
surface
Etching of Si(100) in concentrated ammonium fluoride and nano-hole creation
Two different types of commercial p-doped Si(100) wafers (2Ω-cm and 10Ω-cm) from
MEMC were used Several different samples of 10x5 mm were immersed for 10 minutes
in 40 electronic grade NH4F solution under agitation Previous works used low
temperatures or anodic potentials applied to the silicon to avoid gaseous hydrogen
production and to obtain a uniform surface but in our experiment we worked at room
temperature to check the influence of hydrogen bubbles on surface morphology During
the etching the stirring was sufficient to provide a uniform concentration of reagents over
the whole sample surface but not to mechanically remove the hydrogen bubbles from the
silicon surface
After the etching each sample was rinsed in ultra-pure water to remove any etching
residues and observed with STM AFM and optical microscopy The AFM measurements
were made in air while for STM measurements the samples were rapidly dried with
nitrogen and inserted into the vacuum system to avoid surface reoxidation After insertion
into the vacuum surface cleanliness was checked with LEED and the surface morphology
observed by STM Parameters for STM measurements were sample bias +4 V 1 nA
current 16 Hz scan rate The images obtained were stable and reproducible over several
days Scan parameters for AFM were 20 nN force and 1 Hz scan rate
Some of the samples were cleaned with an RCA standard clean [1] before NH4F etching to
check the influence of possible surface contaminants on the final results Eight different
samples were prepared and more than sixty STM images of the samples were taken at
different points of the various samples
42 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
(110)
(110)
Fig47 abc) STMtopographic images of differentetching morphologies Eachimage is 500x500 nm Z-ranges are 10 10 and 18 nmrespectively d) STM image of a bridgecreated by etching of the lowerlayers of silicon (black arrow)Image is 250x250x6 nm
After ~2 min of immersion in the solution hydrogen bubbles become visible on the sample
surface The production is slow and the bubbles are quite stable on the sample without
detaching Thus some areas of the surface are masked from the liquid etching action
STM observations (fig 47) show that at the nanometer scale the surface is unevenly
covered with holes of radii ranging from 10 to 200 nm with depths of 2-4 nm These holes
have a wide range of different shapes and distributions In most cases the surface was
covered with a uniform distribution of round-shaped holes (fig47a) indicating isotropic
etching The dimension and the density of the holes changed greatly from sample to
sample and even over the surface of a single sample In some cases the etching was
anisotropic yielding nearly square holes and layered structures as shown in fig 47b
Square holes have been previously observed in cases where the etching speed in the (110)
direction is significantly smaller than in the (100) direction [17]
Over these areas (fig 47b and especially 47c) it is clearly visible how once the surface
had been attacked the reaction continued to preferentially remove atoms at step
irregularities (kink atoms) straightening step edges Eventually the exposed underlying
silicon was also attacked and further holes created inside the previously etched larger
ones It was not possible to detect monatomic steps on this kind of surface The smallest
step height observed was ~15 nm corresponding to several atomic layers In the image
shown in Figure 47a the etching was not strong and created only anisotropic holes on the
surface In fig 47b and c the stronger etch proceeded laterally for several tens of
VPalermo 43
nanometres leaving straight steps several tens of nanometers long and roughly rectangular
holes as expected given the structure of the (100) crystal face In some cases a
significative underetch is observed and the formation of suspended bridges and tunnels
can be deduced from the STM images (fig 47d)
The formation of branched pores and suspended structures has been attributed during pore
formation to diffusion limited aggregation effects where the h+ charge carriers necessary
for silicon etching have a higher probability of reacting at pore bottoms than reaching the
upper part of the silicon surface In the case of very deep pores quantum wire effects have
been invoked to explain the pore growth mechanism [15] In our case though the pores
formed were very shallow the underetch depth being only a few nanometres on pores of
sim100 nm width Thus more than diffusion effects the main contribution to the
underetching process must come from anisotropic etching and some kind of autocatalytic
reaction path analogous to the one described by Matsumura et al [4] with some areas of
the silicon surface hydrogenated and thus less vulnerable to etching
Pre-treatment with RCA cleaning has no effect on the final morphology and this seems to
exclude pore nucleation being caused by presence of metallic or organic surface
contaminants
The morphology and the distribution density of the pits was quite uniform over
microscopic areas of the sample but changes were observed over the millimetre scale This
suggests that etching intensity is influenced by some large-scale parameter
Large-area measurements made with AFM or with an optical microscope (fig 48)
showed that the inhomogeneity of surface etching can be correlated with the masking
action of the bubbles While the fluoride dissolved the silicon hydrogen bubbles formed
by the reaction covered some areas of the surface thus blocking the etching over that area
generating macroscopic steps at the bubble-liquid border As the reaction proceeded more
hydrogen accumulated and the bubble diameter increased producing in this way a circular
pattern of steps The increase in bubble diameter was not continuous with time otherwise a
uniform surface slope gradient would have been obtained The formation of this circular
ldquoetching staircaserdquo indicates that the bubble growth was stepwise the bubble accumulating
more and more hydrogen without enlarging across the surface until it relaxed increasing
its diameter stepwise and covering more silicon The circular structures in fig 48a are not
co-axial and their asymmetry could derive from physical processes due to stirring or
irregularities on the surface
44 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
a b Fig 48 a) optical micrograph of etching patterns on Si(100) created by NH4F 12x09 mm b) AFM image of the circles border xy range is 40x24 microm z-range is 30 nm
The step structure was not destroyed by the etching even after the bubble detached from
the surface but on the contrary the etching process seemed to be influenced by the
presence of the step
Observing in detail a series of steps (fig 48b) a quite deep trench is visible at the base of
each step A close-up image of a step and the corresponding line profile of the trench is
shown in fig 49 The trench is sim5 nm deep with respect to the lower surface compared to
a step height of 22 nm
A similar structure has been recently obtained with electrochemical etching of p-type
Si(100) in 4 HF [18] in which a ldquocurrent burstrdquo etching model previously described
was assumed for silicon dissolution In that experiment the trench was created at the
border of silicon nitride masks and began to grow after a nucleation stage Preferential
trench etching was along the (110) direction and the trench growth was explained as an
effect of mechanical stress induced by the nitride mask and of electric field enhanced
dissolution which depended upon an external applied potential
While it is clear that in our system the gas bubbles have a masking effect similar to a
classical solid nitride mask it is unlikely that hydrogen present on the surface can induce a
significant stress in the silicon lattice as in the case of a nitride mask Furthermore no
external field was applied to drive preferential etching at the trench site
It has been proposed [19] that the cathodic and anodic part of the etching reaction
(hydrogen production and silicon oxidative etching respectively) take place at different
points on the surface with a net charge transfer between the different areas In this case
the highest reaction rates will correspond to the silicon area surrounding the bubble border
where a high number of positive charges will be available for the reaction Furthermore a
sharp trench extending into the silicon crystal will be a preferential electrostatic attractor
VPalermo 45
Fig 49 AFM image of the etched surface showing a step created on the surface by bubblemasking A stronger etching action is visible on the right side of the step as well as theprotected area on the upper side of the step (indicated by the arrows) Image is 10x10 micromz-range is 30 nm The profile on the right is taken from the central area of the image
for the h+ charge carriers coming from other ldquocathodic areasrdquo of the sample either from
other regions on the surface or from the back of the silicon chip [18]
In the areas where the hydrogen bubble had detached and the surface was exposed to the
etching the reaction was not uniform in the neighbourhood of the steps It is possible to
observe (fig 48b and 49) an area on the upper side of the step where less or even no
etching at all seems to have taken place as if the step was able to protect the surface from
etching While etching on the lower side with trench formation can be attributed to the
presence of the bubble the surface on the upper step side can be etched only after bubble
detachment so no masking effect can account for this result However a further
preferential attraction of h+ charge carriers from the already formed trench can be
hypothised electrochemically shielding the surrounding area from further etching If this is
true the shielding effect would be very strong with a relatively shallow 5 nm-deep trench
protecting an area of sim1 microm parallel to the step
To summarise the etching of Si(100) in NH4F is a complex process in which different
reaction paths both chemical and electrochemical co-exist Hydrogen bubbles formed by
the reaction act as a mask on the surface and create etching paths and inhomogeneous
etching of the surface Different kinds of pores are observed on the surface and in some
cases the anisotropy of the process is so strong as to give square-shaped holes and
underetching
46 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The diffusion of h+ charge carriers in the crystal is one of the main rate-determining steps
of the reaction and leads to the formation of a deep trench immediately outside the bubble
perimeter These trenches act as charge collectors and reduce the etching of the upper step
surface in the proximity of the steps
Bibliography
[1] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[2] Spectroscopic Ellipsometry Studies Of HF Treated Si (100) Surfaces Yao H Woollam
Ja Alterovitz SA Applied Physics Letters 62 (25) 3324-3326 Jun 21 1993 Influence
Of HF-H2O2 Treatment On Si(100) And Si(111) Surfaces Graf D Bauermayer S
Schnegg A Journal Of Applied Physics 74 (3) 1679-1683 Aug 1 1993 Kinetics Of
Oxidation On Hydrogen-Terminated Si(100) And (111) Surfaces Stored In Air Miura
T Niwano M Shoji D Miyamoto N Journal Of Applied Physics 79 (8) 4373-4380
Part 1 Apr 15 1996
[3] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995 Structure Of The Stepped SiSiO2 Interface After Thermal-Oxidation -
Investigations With Scanning Tunneling Microscopy And Spot-Profile Analysis Of
Low-Energy Electron-Diffraction Pietsch GJ Kohler U Jusko O Henzler M Hahn
PO Applied Physics Letters 60 (11) 1321-1323 Mar 16 1992
[4] Enhanced Etching Rate Of Silicon In Fluoride Containing Solutions At pH 64
Matsumura M Fukidome H Journal Of The Electrochemical Society 143 (8) 2683-
2686 Aug 1996
[5] A Study Comparing Measurements Of Roughness Of Silicon And SiO2 Surfaces And
Interfaces Using Scanning Probe Microscopy And Neutron Reflectivity Crossley A
Sofield CJ Goff JP Lake ACI Hutchings MT Menelle A Journal Of Non-Crystalline
Solids 187 221-226 Jul 1995
VPalermo 47
[6] Step-Flow Mechanism Versus Pit Corrosion - Scanning-Tunneling Microscopy
Observations On Wet Etching Of Si(111) By Hf Solutions Hessel HE Feltz A Reiter
M Memmert U Behm RJ Chemical Physics Letters 186 (2-3) 275-280 Nov 8 1991
[7] Comparison Of Si(111) Surfaces Prepared Using Aqueous-Solutions Of NH4F Versus
HF Higashi GS Becker RS Chabal YJ Becker AJ Applied Physics Letters 58 (15)
1656-1658 Apr 15 1991
[8] Wet Chemical Etching Of Si(100) Surfaces In Concentrated NH4F Solution -
Formation Of (2x1)H Reconstructed Si(100) Terraces Versus (111) Faceting Neuwald
U Hessel HE Feltz A Memmert U Behm RJ Surface Science 296 (1) L8-L14 Oct
10 1993
[9] Ideal Hydrogen Termination Of Si(001) Surface By Wet-Chemical Preparation Morita
Y Tokumoto H Applied Physics Letters 67 (18) 2654-2656 Oct 30 1995
[10] Atomic Structures Of Hydrogen-Terminated Si(001) Surfaces After Wet Cleaning
By Scanning Tunneling Microscopy Endo K Arima K Kataoka T Oshikane Y Inoue
H Mori Y Applied Physics Letters 73 (13) 1853-1855 Sep 28 1998
[11] On The Potential-Dependent Etching Of Si(111) In Aqueous NH4F Solution
Houbertz R Memmert U Behm RJ Surface Science 396 (1-3) 198-211 Jan 20 1998
[12] Etching Process Of SiO2 By HF Molecules Hoshino T Nishioka Y Journal Of
Chemical Physics 111 (5) 2109-2114 Aug 1 1999
[13] Etching Mechanism And Atomic-Structure Of H-Si(111) Surfaces Prepared In
NH4F Allongue P Kieling V Gerischer H Electrochimica Acta 40 (10) 1353-1360
Jul 1995
[14] Pore Formation Mechanisms For The Si-HF System Carstensen J Christophersen
M Foll H Materials Science And Engineering B-Solid State Materials For Advanced
Technology 69 23-28 Sp Iss Si Jan 19 2000
[15] Porous Silicon - Mechanisms Of Growth And Applications Parkhutik V Solid-
State Electronics 43 (6) 1121-1141 Jun 1999
[16] Formation And Application Of Porous Silicon Foll H Christophersen M
Carstensen J Hasse G Materials Science amp Engineering R-Reports 39 (4) 93-141 Nov
1 2002
[17] Pore Morphology And The Mechanism Of Pore Formation In N-Type Silicon
Searson PC Macaulay JM Ross FM Journal Of Applied Physics 72 (1) 253-258 Jul 1
1992
48 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
[18] Deep Electrochemical Trench Etching With Organic Hydrofluoric Electrolytes
Christophersen M Merz P Quenzer J Carstensen J Foll H Sensors And Actuators A-
Physical 88 (3) 241-246 Mar 5 2001
[19] Morphology Of Anodically Etched Si(111) Surfaces - A Structural Comparison Of
NH4F Versus HF Etching Houbertz R Memmert U Behm RJ Journal Of Vacuum
Science amp Technology B 12 (6) 3145-3148 Nov-Dec 1994
VPalermo 49
5 Surface modification of silicon in vacuum void creation and oxide desorption
The main reason for the huge success of silicon in the microelectronics industry is not due
to its superior properties as a semiconductor Other materials for example germanium
have better qualities such as higher mobility of charge carriers and lower noise levels
which would allow the construction of faster and higher performance devices
The widespread use of silicon however is mainly due to the outstanding characteristics of
its oxide Silicon dioxide (SiO2) is a very good electrical insulator easy to form
chemically and thermally stable and is compatible with lithographic and metal deposition
processes Germanium oxide on the contrary is too reactive to be used
Even the use of Si(100) substrates for nearly all microelectronic devices is dictated by
oxide quality The (111) face of silicon crystal can be easily cleaved and flattened and
almost atomically perfect surfaces can be obtained with simple chemical procedures (as
described above) But the density of interfacial defects is highest for Si(111)-SiO2
interfaces and lowest for Si(100)-SiO2 ones so microchips will continue to be fabricated
on Si(100) wafers
SiO2 (silica) is present in 95 of the earthrsquos minerals in different allotropic forms such as
quartz tridymite and cristobalite In the bulk each silicon atom is bonded to four oxygens
in a Si-O-Si tri-dimensional network Si-O bonds are 016 nm long and form an angle
ranging from 120deg to 150deg
Three typical intrinsic defects are present in SiO2 The so-called Ersquo centres are oxygen
vacancies with a hole localised on a silicon atom with only three Si-O bonds
50 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
O3Si +SiO3 Whereas the PR (peroxy radical) defects are holes trapped by a charged
peroxy moiety with a O3Si-O-O+ SiO3 structure The NBOHC (non-bridging oxygen hole
centres) derive from water or hydrogen contamination and are schematized as O3Si- O- H-
O-SiO3
The atomic structure of the Si-SiO2 interface varies enormously Local domains resembling
the tridymite and the cristobalite structure of silica are present but it seems that only 10
of the interface is ordered [1] Far from the interface the SiO2 bulk is completely
disordered The passage from bulk Si to stoichiometric SiO2 passes through a non-
stoichiometric SiOx layer 07 nm thick
When a clean silicon surface is exposed to atmospheric oxygen a thin sim2 nm thick layer
of native oxide forms spontaneously which is usually removed and substituted with
thicker better quality oxide layers before further processing
Silicon is usually oxidised by thermal annealing at temperatures between 800deg and 1100deg
in an atmosphere of pure O2 with some water eventually added to increase oxidation speed
Thermal oxides made in pure oxygen (dry oxides) grow more slowly than oxides produced
in an oxygen-water atmosphere (wet oxides) but are usually of better quality
According to the Deal-Groove formula the time t needed to grow an oxide of thickness X
is given by [1]
1
212minus
minusminus
+=
ABXBXt α
where the constant B and BA decrease exponentially with temperature as
minus
kTEexp
with activation energies for dry oxidation of EB =123 and EBA =20 eV respectively
EB is related to the diffusion of oxygen in silicon while the value of EBA is interpreted as
the energy required to break a Si-Si bond The exponent α is 1 for wet oxidation and 0 for
oxidation at high temperatures and low oxygen pressures It has intermediate values for dry
oxidation This formula does not work well for low values of X at the initial stages of
oxidation and usually empirical corrections are used
An interesting characteristic of silicon is that at high temperature and in vacuum oxygen
can actually etch the silicon crystal giving gaseous products and the oxidized layer present
on the crystal becomes unstable (fig 51) [2]
VPalermo 51
1E-10
1E-09
1E-08
1E-07
1E-06
1E-05
1E-04
1E-03
1E-02
1E-01
0607080911112
1000T (1K)
P (T
orr)
SiO2 + Si rarr 2SiO(g) Oxide decomposition
Si+O2 rarr SiO2(s) oxide formation Fig51 phase diagram of the
oxygen-silicon system
Silicon oxidation apart from the initial nucleation stages at the monolayer level proceeds
uniformly over the whole surface with a planar reaction front moving from the surface
into the bulk
If heated under low oxygen partial pressure (vacuum or inert atmosphere) SiO2 is known to
decompose following the reaction
SiO2 + Si rarr 2SiOuarr (1)
The reaction begins with nucleation at defect points on the SiSiO2 interface and proceeds
in a spatially inhomogeneous manner with the formation of large voids on the oxide
surface [3]
Several studies have been made on the dynamics of void growth both on thick [4] and thin
[5] layers of SiO2 The process has been used to decorate otherwise unobservable defects at
the SiSiO2 interface [6] or to grow nanoislands of silicon on the void surface [7] It has
been suggested that the defects acting as nucleation centres could be metallic contaminants
present on the native surface which aggregate and catalyze SiO2 decomposition [8]
After oxide desorption the silicon surface is very rough In particular on Si(100) square
islands are observed several nanometres high which act as lsquopinning sitesrsquo on the motion of
monatomic steps on the Si surface It has been proposed that these islands can be composed
of silicon carbide coming from the organic contaminants present on the surface [9] or of
oxide clusters during partial reoxidation [10]
In the following section some experimental results obtained studying voids and nanoisland
growth are presented
52 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nanoisland growth on silicon
We used two different samples the first rougher than the second (i) p-type Czochralski Si
(100) with a RMS roughness of 05 nm shown in fig52a and (ii) p-type epitaxial Si(100)
with a RMS roughness of 019 nm shown in fig 52b The roughness was measured with
STM on the lsquoas receivedrsquo samples on different areas of 500x500 nm Every sample was
covered by a layer of native oxide ~2 nm thick
Some of the samples (AC) were introduced into the UHV system without any cleaning
others (BD) were dipped in concentrated HF (49) to remove the surface oxide without
etching the silicon [11]
After degassing the samples were heated resistively increasing the temperature slowly to
keep the pressure within the 10-10 mbar range during heating Each sample was held at
900˚C for 30 min For some of the samples the heating was stopped at lower temperatures
to obtain incomplete oxide desorption and surface roughening thus enabling observation
of the different steps of the process Table 51 summarizes the different treatments for each
group of samples
Fig52 Original surfaces of Czochralski (a) and epitaxial (b) silicon Every image is 500x500 nm
VPalermo 53
Table 51 Summary of different treatements
Silicon type Oxide removed
Heating Islands density (microm-2)
Av Island Volume (nm3)
A Czochralski No 30rsquo at 900˚C 24x103 353
B Czochralski Yes ldquo 20x104 71
C Epitaxial No ldquo 28x102 2450
D Epitaxial Yes ldquo 23x104 63
Fig53 ab) Surface roughening on group A samples after heating 30 min at 800deg and900deg respectively in presence of an oxide layer Image size 500x500 nm cd) Surface roughening on group B samples after heating 30 min at 700deg and 900degrespectively after removing the oxide layer with HF Image size 250x250 nm
Fig 52a is the native oxide surface of sample A quite disordered and irregular This
surface was stable when heated to 700˚C while at higher temperatures the oxide began to
desorb Fig 53a was taken after heating the sample at 800˚C for 30 min
54 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The oxide began to desorb in correspondence with defect points creating oxide-free
surface voids which enlarge radially uncovering the silicon surface Some surface silicon
atoms are removed through reaction (1) Mobile silicon atoms moving randomly over the
surface aggregate forming a nanocrystal at the initial defect point surrounded by a stable
flat oxide-free surface with some steps The initial SiSiO2 interface as observed after
oxide etching with HF [12] is similar to the one shown in fig 52a and thus the surface
observed in the void area is a completely new one coming from surface reorganization
The reaction continues leading to a coalescence of the voids the final result of the process
is shown in fig 53b with protruding islands aligned along the main crystal axes and flat
areas with some steps LEED measurements confirmed that this surface is crystalline
silicon Atomic resolution was obtained on the flat areas of the surface showing the
typical 2x1 reconstruction of Si(100) Figures 53c and 53d show the evolution of the
roughening on sample B after chemically removing the native oxide layer before
introducing the sample into UHV As mentioned above the initial morphology is similar to
that shown in fig 52a though the oxide has been removed as confirmed by LEED
measurements Fig 53c shows the surface after heating at 700˚ for 30 min Without the
oxide layer the surface began to reorganize at lower temperatures and the small irregular
grains of the original surface begin to grow and become more rectangular After heating to
900˚C (fig 53d) the surface is flat with small rectangular islands The shape of the islands
is similar to that shown in fig 53b but the lateral island dimension is one order of
magnitude smaller
To assess the importance of the initial surface on the process the same treatments were
repeated starting from flat epitaxial silicon (samples CD) The initial epitaxial surface is
smooth with periodic steps due to a small miscut angle (fig52b) On heating the epitaxial
silicon without etching (sample C) relatively large islands are obtained similar to the ones
shown in fig 53b where the ldquostep-pinningrdquo effect of the islands is evident The average
dimension of sample C islands was sim2500 nm3 while only 353 nm3 for sample A Island
density decreased from 24000 to 280 microm-2 This result is maybe due to a lower
concentration of defects in the oxide layer of the epitaxial silicon which allows the
expanding void to grow much larger before coalescence with neighbouring voids the
nanostructures within them thus reaching greater dimensions To verify this hypothesis we
applied the same treatment to clean not-so clean and purposely contaminated samples To
check that the defects inducing void nucleation and nanoisland growth derive from carbon
VPalermo 55
contamination we used electronic-grade methanol to contaminate the silicon surface
Figure 54 shows the results of this comparison
State-of-the-art cleaned samples obtained with multi-step RCA cleaning [13] showed the
development of very few voids the oxide desorption leaving large smooth areas of oxide-
and nanostructure-free silicon as shown in Fig 54a where the presence of residual oxide
can be seen Samples cleaned with a simpler one-stage Pirana cleaning solution showed a
higher number of islands per unit area (Fig 54b) and samples purposely contaminated
with methanol (Fig 54c) had the highest density of island nucleation of all three samples
When the native oxide had been chemically removed from the epitaxial silicon before
heating (sample D) the island size and density is similar to the standard as-received
silicon case (sample B) Thus starting from a flat substrate the process of surface
roughening and island creation was the same even when the oxide layer had been
previously removed
From these results some important indications can be drawn The first is that oxide
desorption temperature depends upon the contamination levels of the oxide layers (fig 54)
with desorption at lower temperatures for contaminated surfaces
Second once the oxide layer is removed the silicon atoms become mobile on the surface
at relatively low temperatures (~700˚C) Their initial configuration is disordered and has a
high surface energy so they crystallize in the small islands shown in fig53d reducing
their surface area and lowering the energy of the system When the oxide is present the
surface cannot reorganize at 700˚C and is stable At 800˚C the oxide begins to desorb in
correspondence with defect points on the surface As the void area enlarges a small
cba Fig 54 The effect of contamination on nano-island production Gradient-filtered STM images of (a)sample cleaned two times with RCA 400x400 nm (b) Sample cleaned with a Pirana solution 500x500nm (c) sample contaminated with MeOH before insertion into the UHV system 500x500 nm Annealingtime is 40 min at 800degC for all the samples
56 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
crystallite grows at the centre of the voids Further oxide desorption uncovers larger areas
freeing more silicon atoms which feed the initial island making it bigger
Fig 55 shows the scaled size distribution of the islands for each group of samples Every
distribution has been obtained using several images of different samples for each group
The size distribution of all the samples follows an exponential decay and the decay seems
the same for all the samples though the average volume of the islands differs by more than
one order of magnitude (see Table 51 for details)
The point defects which catalyze void nucleation can be metallic contaminants present on
the initial oxide surface [8] or organic contaminants which at high temperatures can
form SiC nanocrystals [14] We used Scanning Tunneling Spectroscopy (STS) to look for
differences between the islands and the surrounding flat silicon surface but no difference
was found Furthermore the total island volume per surface unit is very high (more than
104 nm3microm2) and it seems unlikely that such a huge volume could consist of surface
contaminants Islands of pure silicon on silicon have been grown without evidence of
surface contamination[14]
To summarise the formation of nano-sized islands on silicon through surface diffusion
was studied At high temperatures the oxide layer covering the surface decomposes non
uniformly and circular voids of clean silicon are created The presence of the oxide layer
blocks surface silicon atomic motion and surface reorganization except within the voids
causing the growth of islands more than 10 nm high and 30 nm wide on the silicon surface
According to STM and LEED measurements we can say that the islands are mostly
Fig 55 Size distribution (scaled) for the samples described in table 1 The line is a reference for the eye
VPalermo 57
composed of silicon with traces of other substances such as carbon Final island densities
and dimensions depend upon the initial purity of the oxide layer
On the other hand if the oxide is removed chemically before heating the island growth
process is not so localised and takes place simultaneously over the whole surface yielding
smaller and more numerous islands
58 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[2] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[3] High-Temperature SiO2 Decomposition At The Sio2Si Interface Tromp R Rubloff
GW Balk P Legoues FK Physical Review Letters 55 2332-2335 Nov 1985
[4] Defect Microchemistry At The SiO2Si Interface Rubloff GW Hofmann K Liher M
Young DR Physical Review Letters 582379-2382 Jun 1987 Defect Formation In
Thermal SiO2 By High Temperature Annealing Hofmann K Rubloff GW Mccorkle
RA Applied Physics Letters 49 1525-1527 Dec 1986 Kinetics Of High-Temperature
Thermal Decomposition Of SiO2 On Si(100) Liher M Lewis JE Rubloff GW Journal
Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 5 1559-1562 Aug
1987
[5] Thermal-Decomposition Of Very Thin Oxide Layers On Si(111) Kobayashi Y Sugii
K Journal Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 10 (4)
2308-2313 Part 3 Jul-Aug 1992 Scanning Tunneling Microscope Study On Mid-
Desorption Stages Of Native Oxides On Si(111) Kobayashi Y Sugii K Journal Of
Vacuum Science amp Technology B 9 (2) 748-751 Part 2 Mar-Apr 1991 Controlled
Growth Of SiO2 Tunnel Barrier And Crystalline Si Quantum Wells For Si Resonant
Tunneling Diodes Wei Y Wallace RM Seabaugh AC Journal Of Applied Physics 81
(9) 6415-6424 May 1 1997
[6] Defect Formation In SiO2Si(100) By Metal Diffusion And Reaction Liher M
Dallaporta H Lewis Je Appl Phys Lett 53 589-591 Aug 1988 SiO2 Film
Decomposition Reaction Initiated By Carbon Impurities Located At A Si- SiO2
Interface Raider Si Herd Sr Walkup Re Applied Physics Letters 59 (19) 2424-2426
Nov 4 1991
[7] Nanometer-Scale Si Selective Epitaxial Growth On Si(001) Surfaces Using The
Thermal Decomposition Of Ultrathin Oxide Films Fujita K Watanabe H Ichikawa M
Applied Physics Letters 70 (21) 2807-2809 May 26 1997 Pyramidal Si Nanocrystals
VPalermo 59
With A Quasiequilibrium Shape Selectively Grown On Si(001) Windows In Ultrathin
SiO2 Films Shibata M Nitta Y Fujita K Ichikawa M Physical Review B 61 (11)
7499-7504 Mar 15 2000
[8] Stacking-Fault-Induced Defect Creation In SiO2 On Si(100) Liher M Bronner Gb
Lewis Je Appl Phys Lett 52 1982-1985 May 1988
[9] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990 AFM And XPS Characterization Of The
Si(111) Surface After Thermal-Treatment Lamontagne B Guay D Roy D Sporken R
Caudano R Applied Surface Science 90 (4) 481-487 Dec 1995
[10] Scanning-Tunneling-Microscopy Study Of Oxide Nucleation And Oxidation-
Induced Roughening At Elevated-Temperatures On The Si(001)-(2x1) Surface Seiple
JV Pelz JP Physical Review Letters 73 (7) 999-1002 Aug 15 1994 Evolution Of
Atomic-Scale Roughening On Si(001)-(2x1) Surfaces Resulting From High-
Temperature Oxidation Seiple JV Pelz JP Journal Of Vacuum Science amp Technology
A-Vacuum Surfaces And Films 13 (3) 772-776 Part 1 May-Jun 1995
[11] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995
[12] Morphological Changes Of The Si [100] Surface After Treatment With
Concentrated And Diluted HF Palermo V Jones D Materials Science In
Semiconductor Processing 4 (5) 437-441 Oct 2001
[13] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[14] Nanoscale Roughening Of Si(001) By Oxide Desorption In Ultrahigh Vacuum
Gray SM Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology
B 14 (2) 1043-1047 Mar-Apr 1996
60 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 61
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and nano-lines
When the silicon surface is not protected by a native oxide layer or by a hydrogenated
passivating layer it is very reactive The surface chemistry of both Si(100) and Si(111) has
been extensively studied starting from clean surfaces prepared in UHV because of its great
relevance to the microelectronics industry and its technological interest The diffusion and
reaction of molecules and atoms on silicon is an interesting scientific problem on its own
apart from technological issues because silicon reconstruction yields a very complex and
anisotropic surface
We give here a brief summary of the adsorption behaviour of several elements on Si
surfaces The summary is not comprehensive and only the most interesting characteristics
for each substance are given
Table 61 Summary of adsorption behaviour of atoms and molecules on silicon surfaces [12]
Hydrogen
Molecular H2 shows low reactivity towards silicon while atomic hydrogen easily forms
Si-H bonds and can even break Si-Si bonds
Alkali metals
Alkali metals diffuse rapidly into SiO2 and can damage silicon-based transistors
Transition
metals
All transition metals apart from gold and silver react with Si forming metal silicides
62 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nickel A very common contaminant of silicon Nickel from even slight contact with stainless
steel tweezers can contaminate silicon samples forming its silicide and is very difficult
to remove even at high temperatures because it segregates on the silicon surface
Palladium Forms silicides especially Pd2Si and can be used to make contacts possessing a low
Schottky barrier
Titanium Widely used to fabricate contacts for silicon-based microdevices TiSi2 is one of the
more stable and highly conductive silicides
Tungsten The silicide is more stable than TiSi2 on polysilicon and is thus used for polysilicon
gate contacts
Platinum Silicide is used for bipolar transistors PtSi contacts on n-silicon give Schottky diodes
Cobalt Cobalt silicide gives better quality contacts than TiSi2 at the nanometrelevel but is less
used at the moment because it consumes too much silicon during its growth
Iron Silicides have been studied for potential optoelectronic applications because FeSi2 has a
directly accessible bandgap and is optically active
Group 13
Group 13 metals are used in IC technology to dope silicon (giving p-type doping) and
to make III-V type semiconductors When evaporated on Si(100) or Si(111) they react
strongly with the surface giving surface reconstruction and kink formation at steps
Boron Widely used for doping Usually deposited with decomposition of boron hydrides
Sticking coefficient of simple hydrides is very low so decaboranes (B10H14) are used for
deposition
Aluminum Aluminum is widely used for microelectronic contacts but in some cases it reacts with
silicon diffusing into the crystal and with SiO2 stealing oxygen atoms to form Al2O3
Group 14
Carbon Carbon can form a composite semiconductor with silicon (silicon carbide) widely used
to fabricate integrated circuits for use at elevated temperatures or in the presence of
ionising radiation When carbon is deposited on silicon usually polycrystalline films
with properties between diamond and graphite are obtained
CO adsorbs molecularly on silicon without breakage of the C=O bond Other
unsaturated hydrocarbons do not react with silicon Even very reactive strained
molecules like cyclopropane have low sticking coefficient Ethylene acetylene and
propylene adsorb molecularly on Si(100) each molecule sitting on a silicon dimer
interacting with the dangling bonds of the surface Acetylene also adsorbs molecularly
on Si(111) on the 7x7 reconstructed surface preferential adsorption on centre atoms
with respect to corner atoms is observed Benzene and other aromatic ring compounds
adsorb molecularly on the surface They can form σ or π-type bonds with the silicon
laying parallel or tilted respect to the surface according to the bond type
Silicon Silicon atoms can be deposited on the surface with molecular beam epitaxy or
VPalermo 63
decomposition of silanes and chlorosilanes Silanes with single Si-Si bonds have a high
sticking probability and dissociate upon adsorption giving trihydrides and
monohydrides
Chlorosilanes have high sticking coefficients too and decompose on the surface but
they can etch the surface through the reaction SiCl4 + Si rarr 2SiCl2
Germanium Germanium has a lattice constant similar to silicon (Ge lc is 4 larger than Si) so that
various alloys of Si1-xGex can be formed The growth of Germanium on silicon is quite
peculiar with several monolayers adsorbing uniformly on the surface followed by
island formation (Stranski-Krastanov growth)
Group 15
Nitrogen Exposure to ammonia and nitrogen gas at high temperatures leads to the formation of
silicon nitride layers (Si3N4) NH3 dissociates on the surface and reacts with the
dangling bonds giving Si-NH2 and Si-H termination Silicon nitride is stable on the
surface up to 1100degC above this temperature it desorbs as Si2N
Phosphorous Phosphine (PH3) adsorbs dissociatively as Si-PH2 and Si-H and behaviour is similar to
that of ammonia
Arsenic A monolayer of As is often deposited on silicon as a substrate for GaAs growth Arsenic
forms dimers on Si(100) creating dimer rows which are parallel (perpendicular) to the
underlying silicon dimer rows when it is deposited at high (low) temperature On
Si(111) Arsenic breaks the 7x7 reconstruction giving a 1x1 pattern This 1x1 As
monolayer acts as a passivating layer and can resist further As adsorption oxygen and
air
Antimony and
Bismuth
Due to their large covalent radii Sb and Bi form only short dimer rows on silicon and
only at high temperatures
Group 16
Oxygen
Oxygen can oxidize or etch silicon according to the reactions
Si(s) + O2 rarr SiO2 (s)
Si(s) + frac12O2 rarr SiO(g)uarr
Silicon oxidation has already been described in the previous chapter Molecular oxygen
adsorption on silicon is mostly dissociative O atoms break the Si-Si bonds forming an
Si-O-Si bridge or a peroxy bridge Si-O-O-Si Hydrided silicon surfaces do not have
surface dangling bonds and are thus more protected from oxygen attack Etching is
favoured at high temperatures and low oxygen pressures [3] At T gt700degC silicon oxide
on silicon decomposes as SiO(g)
Water
H2O adsorbs dissociatively on silicon easily on Si (100) and with more difficulty on
Si(111) The molecule decomposes giving Si-OH and Si-H on adjacent sites
64 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Group 17
Fluorine
Fluorine adsorbs and reacts easily on silicon dissociating on the surface and forming
Si-F bonds It can then penetrate the surface and attack polarised Si-Si backbonds The
surface is etched with production of SiF4 and SiF2
Chlorine
Chlorine reacts aggressively with silicon etching the surface Cl2 and F2 are often used
commercially to etch silicon usually in a plasma Atomic Cl and Cl2 adsorb on Si(100)
up to saturation The most stable configuration seems to be a Cl atom bonded to each Si
atom of a surface dimer in a symmetric structure but metastable configurations with
two Cl atoms bonded to a buckled dimer and even a central Cl atom bridged across a
dimer have been observed
On Si(111) silicon mono- di- and tri-chlorides are formed In particular heating at
Tgt400degC a uniform Si-Cl monolayer is obtained and the 7x7 reconstruction changes
into a bulk like 1x1 lattice The surface transition induced by Cl adsorption is fully
reversible The 7x7 reconstruction can be restored on desorbing the chlorine by heating
at Tgt1100degC The 1x1 domains nucleate at the lower terrace side of steps the 7x7 at the
upper step edge
Bromine
Like hydrogen bromine maintains the 2x1 structure of Si(100) forming Si-Br bonds
with the dangling bonds of the silicon dimer rows At high exposure some etching of
silicon by formation of volatile SiBr3 species has been observed
Surface diffusion on silicon
In the previous chapter we observed nano-island growth on silicon surfaces with native
oxide and surfaces contaminated with organic impurities While this phenomenon has been
observed in several experiments the formation and growth mechanisms of nanoislands has
still not been satisfactorily described theoretically Various substances have been
hypothesised as nanoisland nucleating agents such as organic or metallic contaminants
present on the surface or oxide clusters (see previous chapter) The island growth process
has been attributed by various workers to lsquosome sort of kinetic instabilityrsquo [4] to the
pinning of step flow by SiC clusters [5] or to a mesoscopic atom flux from areas with low
step density to areas of higher step density [6]
One mechanism proposed to explain surface roughening and island growth is the Ehrlich-
Schwoebel effect ie the presence of an energetic barrier that reflects atoms approaching a
VPalermo 65
Molecule (L
O2
H2O MeOH
CO CO2 CH4
clean surface
downward
weak on si
The aim o
surface an
To obtain
the ones
spontaneou
first obser
surface W
nanoisland
Experime
We used S
removed f
41 mixtur
water and
whole hea
1200degC re
kept below
Table 62 Treatment island density and island volume for each molecule tested
Dose angmuir)
Heating time at 800degC (min)
Island density (microm-2)
Ave Volume (nm3)
Equivalent layer thickness (nm)
280 10 no islands - -
90 10 no islands - - 30 10 2000 19 004
180 10 360 516 019 30 60 no islands - -
180 10 no islands - - - 10 no islands - -
step [7] Recent experiments however indicate that the Schwoebel effect is
licon [48]
f our experiments was to study the reaction of simple molecules with the silicon
d to check the ability of these molecules to nucleate nanoisland growth
a better understanding of the process we chose simple molecules starting from
which are more likely to be present as traces in UHV chambers where
s nanoisland growth is often observed after sample heating Using STM we
ved at the atomic level the possible interactions of these molecules with the
e then heated the surface and checked the density and dimensions of the
s produced
ntal procedures and results
i(111) wafers p-doped 0015 Ω cm Gross contamination and particles were
rom the sample surface with a standard Pirana etch (15 min dip in a H2S4H2O2
e at T=80degC) After this the samples were thoroughly washed in ultra-pure
introduced into the UHV system After degassing for 5 hours at 600degC the
ting stage was allowed to cool down and a sequence of rapid flashes to 1100deg-
moved the native oxide layer and cleaned the surface Pressure during flashes is
1x10-9 mbar
66 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Surface flatness and cleanliness were checked with STM and the contaminant molecules
introduced into the UHV system through a leak valve Doses of several Langmuirs were
used for the experiments (1 Langmuir=10-6 torr sec-1) Pressure was kept constant at 10-7
mbar during adsorption The STM tip was retracted during gas deposition to avoid any tip
shielding effect After the adsorption process the leak valve was closed the base pressure
in the UHV chamber returning to below 10-10 mbar and the presence of adsorbed
molecules checked on the surface with STM Finally the sample was heated at 800degC for
several minutes to allow nanoisland growth Several different areas were examined on each
sample
Table 62 summarizes the dose of exposure and the heating time for each molecule Other
experiments were carried out using smaller dosing times but only the significant results
are reported here
As expected oxygen and water react with the surface oxidising it The adsorption process
can be easily followed with the STM by lowering the tip from time to time and taking an
image No effect of the STM tip is observed over the scanned areas Upon molecule
adsorption the order of the crystalline surface rapidly degenerates and after several
minutes the surface looks completely covered by irregular atomic-sized protrusions
corresponding to Si-OH or Si-O-Si species On heating to 800degC the contaminating species
were easily desorbed through SiO formation the surface became clean and the crystal
surface periodicity re-established No residues or nanoislands were observed
Amongst the molecules tested was methanol CH3OH reacts readily with the surface and
covers it with a disorderd layer already after sim 30 Langmuirs exposure (fig 61)
Fig 61 Si(111) surface during methanol adsorption at 0 1 and 5 minutes respectively (corresponding to 0 6 and 30 L exposure) The dark irregular lines are borders between different 7x7 domains Image size 60x60 nm
VPalermo 67
Synchroton radiation photoemission spectroscopy experiments by Carbone et al [9]
showed that methanol reacts with the rest atoms of the 7x7 silicon cell already after 1
Langmuir exposure but that longer exposures are needed for methanol to react with the
remaining silicon atoms Methanol adsorbs dissociatively forming SiOCH3 and Si-CHx
species on the surface
After methanol adsorption the sample is heated to 800degC and observed again with STM
As in the case of water and oxygen the heating restores the crystalline surface and the 7x7
pattern returns visible But this time quite a dense array of nanoislands is observed on the
surface (fig 62) The islands have an average diameter of 20 nm and a height of sim2 nm
At temperatures above 400degC the Si-OCH3 and Si-CHx species decompose oxygen and
hydrogen are desorbed and according to [9] carbon atoms remain as SiC dispersed
uniformly on the surface Our STM measurements indicate that the carbide does not
randomly cover the surface but that C atoms are concentrated at the nanoislands Rough
calculations based on island volume suggest that the islands must be composed of a SixC1-x
alloy with x varying between 05 and 1
Following these results we expected carbon monoxide to behave in a similar manner
adsorbing onto the surface and with nanoisland growth However no adsorption was
observed with STM and the surface looked perfectly clean even after a dose of more than
100 L of carbon monoxide Some species did adsorb onto the surface though because
after 10 minutes subsequent heating nanoisland growth was observed Island density in
this case was only 18 of the density obtained with methanol and island dimensions are
Fig62 Nanoislands on Si(111) created after 30 L adsorption of methanol and 10 min heating at 800degC Image size 200x200 nm
68 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
much bigger (see table 62 and fig 63) To check whether the islands had been nucleated
by some other contaminant present in the vacuum chamber we heated a blank sample
without introducing any molecule No island growth was observed Exposing the surface to
other simple molecules including carbon dioxide and methane also gave no nanoisland
formation
We can estimate the amount of carbon deposited on the surface during gas dosing The
total flux of molecules hitting the surface is obtained from the Hertz-Knudsen formula
TkmPF
π2=
Fig 63 STM images of Si(111) after exposure to methanol (left) and carbonmonoxide (right) contaminating molecules and subsequent nanoisland growthImage size 400x400 nm The size distribution of nanoislands is shown below eachimage
VPalermo 69
Fig64 Left Si(111) after 180 L contamination of CO and 16 hours heating at 800degC Right Si(111)without CO contamination after 16 hours heating at 800degC Image size 250x250 nm Vertical scale is thesame for both images z-ranges are 35 and 5 nm respectively
where P is the pressure in Pascal k=138x10-23 J K-1 is the Boltzmann constant T is the
temperature and m is the molecular weight (sim32 for methanol and sim28 for CO) in
kilograms
Using this formula we obtain a flux of 027 molecules nm-2 sec-1 for methanol and 029
molecules nm-2 sec-1 for CO After ten minutes exposure at 10-7 mbar pressure more than
150 molecules will have hit each square nanometer of the surface Surface density of
Si(111) 7x7 is sim16 atomsnm2 so each surface atom will be hit by several molecules which
could react or be adsorbed even assuming a sticking coefficient much lower than unity
To explain nanoisland growth we hypothesize that CO molecules adsorb molecularly and
-5
0
5
10
15
20
25
30
0 20 40 60 80 100 120
nm
nm
10 min40 min16 h16 h- no CO
Fig65 STM profiles of nanoislandsgrown for different annealing times The lower curve correspond to the surface heated at 16 hours without CO contamination
70 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Table 63 Island density and volume for the CO-contaminated surface at different heating times For each sample the thickness of a uniform layer having the same volume per unit area of the islands is calculated
Molecule Dose
(Langmuir) Heating time at
800degC (min) Island density
(microm-2) Ave Volume
(nm3) Equivalent layer thickness (nm)
CO 180 10 362 517 019 CO 180 40 550 1072 059 CO 180 16 h 505 3675 186
clean surface - 16 h rough - -
very weakly on Si(111) moving rapidly on the surface Because of their rapid diffusion
the STM is not able to image the CO molecules on the silicon [10]
Increasing the temperature part of the CO molecules will surely desorb but some of them
will dissociatively adsorb onto the surface The oxygen will then desorb as SiO while
carbon will remain fixed forming very stable Si-C bonds
CO decomposition will be favored at surface defects surface steps or in the proximity of
already formed SiC clusters An increased reactivity of the CO molecule on a surface in
correspondence with phase boundaries has already been observed on Pt surfaces [10]
In this case a reduced number of nucleation centres will react with the CO molecules
yielding large and fewer islands with respect to the case of methanol which does not have
the possibility of travelling long distances over the surface
Other factors are likely to influence the process of island nucleation after CO adsorption
the co-adsorption of other molecules which can slow down CO and favour its
decomposition cannot be ruled out as well as the formation of new defects at higher
temperatures Measurements with a variable temperature STM or with some other surface
analysis technique are planned in the future to confirm the proposed mechanism
When the surface was further annealed the nanoislands grew in size After 16 hours
heating very large nanoislands with diameters of sim35 nm and heights of sim20 nm became
visible (fig64 left) As a comparison when the same surface was heated without
nucleation centres deriving from contamination an irregular surface was obtained (fig64
right)
While the nanoisland density seemed to reach an asymptotic limit of 500 islandsmicrom2
(table 63) island size continued to grow with time (fig65) even when no further carbon
was supplied to the surface This indicates that even though the initial nucleating core of
the nanoisland is likely to be an SixC1-x alloy further growth is due to silicon atoms
diffusing from the crystal and being adsorbed by the growing island The final volume
VPalermo 71
occupied by the islands corresponds roughly to a 18 nm thick overall layer of removed
silicon
It is noteworthy that the clean sample after 16 hours annealing even though smoother than
the CO-contaminated one looked much more disordered with hardly any flat area visible
on nanometres scale In the presence of nanoislands however further annealing increases
island size but keeps part of the surface quite flat and ordered with flat areas visible
between the islands Thus the nanoislands can be imagined to act as ldquoimpurity sinksrdquo for
further contaminants approaching the surface yielding a greater but more ordered surface
roughening with respect to the uncontaminated silicon surface
Circular arrays of nanoislands
We used the voids described in the previous chapter formed during oxide layer
decomposition as nanoscopic masks to test the validity of the hypothesis outlined above
The small circular areas uncovered during oxide decomposition can act as ldquoskating rinksrdquo
where contaminants and diffusing silicon atoms can move freely over the surface while
the oxide layer all around and higher than the void area as well as being chemically
bonded to the surface silicon atoms will hinder surface diffusion Furthermore the oxide
passivating layer will at least in some cases prevent contaminating molecules from
adsorbing and decomposing on the surface In this way selective nanoisland growth inside
void areas can take place
The experimental procedure consisted of the following steps
1) An oxide-passivated sample was heated for 10 minutes at 800degC Oxide decomposition
began in correspondence with surface defects and spread laterally creating naked areas
of silicon
2) While the void enlarges the same defect that initiated oxide decomposition behaved as
an attractor for moving silicon atoms and nucleated the growth of a nanoisland at the
void centre
3) The void surface was exposed to contaminating molecules
4) The sample was then heated again at 800degC During the second heating the void
continued to enlarge and new nanoislands were created The freshly uncovered silicon
surface provided mobile silicon atoms to feed nanoisland growth
72 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
5) When the heating continued for long enough all the oxide was desorbed and a clean
silicon surface covered by circular groups of nanoislands was obtained
Figure 66 shows the different stages of the process
We thus heated an oxidized sample at 800degC for 10 minutes in UHV to create multiple
voids in the oxide layer After this we exposed the sample to the contaminating molecules
Then we further annealed the surface at the same temperature to promote island growth
The first time we observed nanoisland growth in the void area was by chance after heating
for a second time samples which had been stored for 2 weeks in vacuum Even at pressures
as low as 10-11 mbar some molecules will adsorb onto the surface slowly creating defects
and thus potential nanoisland nucleation centres It is possible to observe with STM that
the silicon crystal surface looks very clean just after a high temperature flash but even
after only a few days randomly adsorbed species will be observable on the surface
Unfortunately it is not possible to identify these species simply from in situ STM
measurements
Mass spectrometry measurements show that the residual gas contaminants in vacuum are
usually He Ar H2 CH4 CO CO2 and N2 [11] Some of these molecules such as
hydrogen will stick to the silicon surface at room temperature but will simply desorb
during annealing without nucleating nanoislands Others like CH4 or CO2 will not react
with the surface to generate nanoislands as demonstrated previously The best candidate
for contamination of samples stored in UHV seems thus carbon monoxide has the ability
321
4 5
Fig 66 Selective nanoisland growth within oxide voids See text for details
VPalermo 73
as shown before to nucleate nanoislands although co-adsorption mechanisms can not be
ruled out Fig 67 (left) shows the circular groups of nano-islands obtained after heating
void-covered samples stored for 2 weeks in UHV Small islands are observed in the former
void area The void itself has enlarged slightly uncovering a clean oxide-free area of the
silicon surface Apart from the central island created during the first period of heating the
surrounding ones have dimensions decreasing from the void border to the inner area of the
void
This confirms that most of the material needed for island growth comes from the silicon
atoms diffusing from the freshly uncovered areas around the original void The outer
nucleating centres are nearest to the silicon atom source and thus generate larger islands
For a faster process we directly exposed the void to a significant concentration of CO
Fig67 (centre) shows the islands grown after a 180 Langmuir exposure and subsequent
heating The islands are preferentially located around the original void perimeter the rest
of the original oxide-free surface remaining untouched This suggests that although the
whole surface is exposed to CO the gas only interacts chemically with the surface at the
SiSiO2 interface around the oxide-free void perimeter The SiSiO2 interface thus provides
preferential nucleation sites for nanoisland growth in the successive thermal annealing
step This recalls a similar phenomenon observed in the catalytic oxidation of CO on the
Pt(111) surface where CO molecules reacted with adsorbed O species only along the
perimeters of oxygen islands [10]
On repeating the CO exposure and thermal annealing steps a second circle of nano-islands
Fig67 Left circular area of nanoislands obtained from a sample stored 2 weeks in UHV 500x500 nmCentre nanoisland circle obtained after exposing the voids to 180 L of CO and heating for 10 min at800degC 400x400 nm Right two concentric circles obtained with further CO adsorption followed byheating 650x650 nm
74 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
was formed around the newly-expanded surface void perimeter (fig67 right) This
process can in principle be repeated again to obtain a series of concentric groups of
nanoislands the only limit depending on the minimal distance between neighbouring
expanding voids We have thus observed that contaminants adsorbed at low pressure over
long periods of time generate random island growth over the whole void area while CO
molecules adsorbed at relatively high pressures over short periods yield selective
nanoisland growth at the former void border Several theories can explain the different
results CH4 and CO2 as mentioned above did not nucleate islands when when the silicon
surface was exposed to them for short periods However the lower surface mobility of
these contaminants and the greater time available for reaction with the surface could be
the cause of nucleation of SiC clusters over the whole oxide-free void area and not only at
its border The coadsorption of water another UHV residual gas with CO on the naked
silicon areas could perhaps allow a surface reaction similar to that between CO and the
SiSiO2 interface during CO exposure Another explanation for the formation of these
random nucleation sites within the voids could be the coadsorption of residual carbon-
containing species present in the UHV system with hydrogen the major residual gas in
stainless steel UHV systems
The possibility of decorating silicon oxide borders with nanoislands using the high
mobility and selective decomposition of CO molecules suggests interesting applications
for nanofabrication processes that will be discussed below
Silicon nanowire creation on Si(100)
We performed most of the experiments described above on silicon (111) 7x7
reconstructed because this surface is easily prepared in UHV and above all because it is
isotropic and has no preferential directions for atom diffusion The 2x1 reconstruction of
silicon (100) even though it has a much simpler unit cell than Si(111) 7x7 shows a more
complicated morphology with lots of monoatomic and biatomic steps and dimer rows
parallel to each other The orientation of the dimer rows changes by 90deg on alternate
atomic layers Surface diffusion in this case is much more complicated because atoms will
experience the effects of the surface anisotropy and diffuse preferentially along or across
the dimer rows The diffusion energy of a silicon atom as example is 06 eV along a
dimer row and 085 eV across different dimer rows This while being a complication
VPalermo 75
offers interesting possibilities for the creation of ordered structures The monoatomic steps
and the dimer rows can for example act as templates for the formation of elongated
structures of composition similar to the islands described previously
On heating a Si(100) surface oxide decomposition and void growth takes place as
described for Si(111) In some cases the void shape reflects the surface symmetry
especially for very thin oxide layers and the voids often have a slightly squared shape
Nanoisland growth is also influenced by the substrate symmetry and the islands look
rectangular as irregular clusters surrounded by concentric patterns of monoatomic steps
The whole structure reminds vaguely a ldquozigguratrdquo a kind of stepped pyramid found in
Mesopotamia (fig68 inset)
Apart from nanoisland growth in some cases we observed spontaneous formation of
nanowire-like SiC structures on the Si(100) surface The wires were generated after
exposing a Si(100) surface in vacuum to traces of CO followed by sample annealing at
800degC for 15 minutes The wires exhibit lengths ranging from 10 to 100 nm and average
widths of ca 5 nm Although being randomly positioned on the surface they are perfectly
aligned along the crystal axes of the substrate (Fig68)
Several descending monoatomic steps can be observed around each nano-line and nano-
island Very often one of the line extremities coincided with an island
Nanoline
Nanoisland
Fig 68 A Si(100) surface covered by nanoislands and nanowires The monoatomic steps present on the surface are visible The two insets show a typical nanoisland and a nanowire obtained on this surface The typical ldquodimer-row ldquostripes are visible along the sides of the nanowire
76 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Couples of parallel lines starting from the same nano-island were also observed At high
resolution (fig 69) the lines showed a lamellar periodic structure with a periodicity of
around 4 nm and frequent defects Scanning Tunneling Spectroscopy measurements
showed no significant difference between the line and the surrounding silicon surface
The proposed formation mechanism for this kind of structure involves two stages in the
first one organic contaminating molecules adsorb and diffuse on the surface from the
vacuum environment stopping preferentially at surface singularities including steps and
defects
Several substances such as atomic carbon carbon monoxide ethylene and fullerene
decompose when adsorbed on silicon surfaces at temperatures above 500degC and create
localised defects on the surface inducing strain deformations and a metastable surface
reconstruction [12]
If these molecules remain attached to the steps they can diffuse rapidly along step border
until they meet a line and decompose In this way long lines of SiC defects parallel to the
surface steps can be produced In the second phase the clusters of silicon carbide can
-005
0
005
01
015
02
025
03
035
04
-3 -2 -1 0 1 2 3
V
dId
V lt
dId
Vgt
linesilicon
Current- Voltage behaviour measured with STS Fourier analysis of wire periodicity
asymp 4 Aring
High resolution image of a line
Fig 69 Top high resolution image of a nanowire showing the lamellar structure Fourier analysis (bottom left) shows that the lamellae periodicity is sim 04 nm bottom right the IV characteristics measured with the STM tip on the line and the silicon surface
VPalermo 77
25-600degC
Final linear structure
600-800degC
Contaminant Silicon
Fig 610 Schematic representation of nanowire formation mechanism
locally inhibit the spontaneous surface diffusion of silicon atoms acting as templates for
the reorganisation of surface steps (fig 610)
The final morphology shows disordered lines and islands each surrounded by a complex
pattern of silicon monatomic layers
At 800degC silicon atoms are highly mobile on the surface Due to atom diffusion the
surface reorganizes with a step-flow mechanism The presence of a fixed line of
contaminants blocks atom diffusion and step flow generating the monoatomic steps
around each wire (fig 611) SiC clusters act as lsquopinning sitesrsquo on silicon atom surface
diffusion and can be purposely used to direct the localised growth of nano-islands [13]
Although the possibility of using methanol and carbon monoxide as nano-island precursors
has been demonstrated the exact nature of the substances nucleating nano-wire growth is
still unknown
Further experiments will be needed to confirm these hypotheses as well as to identify the
contaminating species nucleating nanowire growth Carbon monoxide is a main candidate
for nanowire growth but the real nucleation process is likely to be a complex one
78 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Step flowblocked
Fig 611 Representation of step flow blocked by the presence of a nanowire 105x105nm
Bibliography
[1] Surface-Chemistry Of Silicon Waltenburg HN Yates JT Chemical Reviews 95 (5)
1589-1673 Jul-Aug 1995
[2] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[4] Growth Of Si On The Si(111) Surface Lanczycki Cj Kotlyar R Fu E Yang Yn
Williams Ed Das Sarma S Physical Review B 57 (20) 13132-13148 May 15 1998
[5] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
VPalermo 79
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990
[6] Nanoscale roughening of Si(001) by oxide desorption in ultrahigh vacuum Gray SM
Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology B 14 (2)
1043-1047 Mar-Apr 1996
[7] Step Motion On Crystal Surfaces Schwoebel Rl Journal Of Applied Physics 40 614-
618 Feb 1969
[8] Scanning Tunneling Microscopy Investigation At High Temperatures Of Islands And
Holes On Si(111)7x7 In Real Time Evidence For Diffusion-Limited Decay
Hildebrandt S Kraus A Kulla R Wilhelmi G Hanbucken M Neddermeyer H Surface
Science 486 (1-2) 24-32 Jul 1 2001
[9] Methanol Adsorption On Si(111)-(7x7) Investigated By Core-Line Photoemission And
Mass Spectrometry Of Photodesorbed Ions Carbone M Piancastelli Mn Zanoni R
Comtet G Dujardin G Hellner L Surface Science 370 (1) L179-L184 Jan 1 1997
[10] Atomic And Macroscopic Reaction Rates Of A Surface-Catalyzed Reaction
Wintterlin J Volkening S Janssens Tvw Zambelli T Ertl G Science 278 (5345)
1931-1934 Dec 12 1997
[11] Redhead PA Hobson JP Kornelsen EV The Physical Basis Of Ultrahigh
Vacuum Chapter 12 (Chapman amp Hall London 1968)
[12] The Si(001) C(4 X 4) Surface Reconstruction A Comprehensive Experimental
Study Norenberg H Briggs Gad Surface Science 430 (1-3) 154-164 Jun 21 1999
[13] Production Of Nanostructures Of Silicon On Silicon By Atomic Self-Organization
Observed By Scanning Tunneling Microscopy Jones D Palermo V Applied Physics
Letters 80 (4) 673-675 Jan 28 2002
80 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 81
7 Conclusions and proposals for future work It is very difficult to foresee the future of nanoscience and nanotechnology even over the
next few years exciting discoveries are continuously made opening up new possibilities
and progress in this sector is now one of the fastest and most varied So it is impossible to
predict whether computers will in the future be made with nanowires nanotubes single
electron devices or some kind of quantum-based transistor What does seem clear
however is that silicon devices can still further extend their levels of miniaturization to
remain competitive for at least the next ten years [1] During this period new emerging
technologies will reach scientific maturity and arrive at the production lines
Even when new technology prototypes do become available for full-scale production the
astronomical cost of changing from silicon to new technologies will be a major concern
Moreover a vast amount of extremely detailed knowledge is already available on all
aspects of silicon technology Thus the possibility of integrating innovative
nanotechnologies with standard CMOS silicon technology already seems an attractive
prospect [2]
For this and other reasons we have focussed our research efforts on surface phenomena on
silicon wafers which could be potentially useful for the development of silicon-compatible
nano-devices
Some important conclusions can be drawn from the work described in this thesis
bull Silicon surfaces of different crystal faces were studied at atomic resolution in ultra-
high-vacuum following transformations induced on the surface by chemical (etching)
and physico-chemical (organic contamination heating) treatments
82 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull The presence of surface contaminants induces at high temperature the growth of
nanometre-sized islands and lines on silicon
bull Final island densities and dimensions strongly depend upon the chemical nature of the
contaminant molecules and the treatments used Molecules interacting only weakly
with silicon will desorb upon heating whereas molecules which decompose easily on
Si will give widespread random nano-island nucleation Between these two extreme
cases complex behaviours of diffusion and subsequent decomposition are possible
bull Nanoscopic naked silicon voids were produced in oxide layers through carefully-
controlled thermal annealing It was found that void density and size depends upon
oxide purity surface contamination levels and annealing conditions
bull The growth of nano-islands and nano-voids was exploited simultaneously and the
voids used as nanoscopic masks to control the positions of growing islands Complex
silicon-on-silicon nano-structures were obtained with multi-step growth processes
bull Finally the use of macroscopic treatments and simple physical processes to produce
large quantities of nanometre-sized structures on silicon surfaces was demonstrated
These processes and the structures produced are of potential interest to the
microelectronics industry
We showed in previous chapters how it is possible to modify silicon surfaces using
chemical and physical methods The use of self-organization for the production of Si-on-Si
20 nm
Fig 71 Left SEM image of field emitter arrays made by conventional lithography [3] Center 3-D STM image of a self-organized void with a central nano-island Right Lateral schematic view of a field emitter device
VPalermo 83
nano-structures has the advantage with respect to other emerging techniques of being
silicon compatible and thus is a potential candidate for its implementation in the
production of new devices Furthermore this technique can create billions of strucures on a
wafer in only a few minutes being thus applicable to large-scale production
Even when ordered structures are obtained these methods still show a poor reproducibility
and control Chemically modified surfaces look very disordered at high magnification
with trenches and holes etched into the surface and progress has recently been made in
producing very small structures electrochemically [4] Oxide decomposition with
nanoisland growth in UHV can also yield ordered structures The void holes with
nanoislands located at their centres although quite variable in dimensions are more similar
to engineered devices than to a spontaneous random surface-roughening process In fact
there is a surprising similarity between the nanovoid-nanoisland structures and well-known
commercially available devices called field emitter arrays (FEA) These are small tips used
as electron guns in various kinds of displays and other devices Both the commercial
devices and our self-organized structures consist of a protruding tip at the centre of
microscopic holes on an insulating layer covering a semiconductor surface (see fig71)
Of course the order and size uniformity of commercial FEAs is much better than that
obtained with our process but it is encouraging to see that a self-organized structure
obtained with simple heating has a very similar surface topography and composition of
commercial working devices obtained after many expensive stages of optical lithography
deposition and etching
Nano-island density and size can to some extent be controlled and different
contaminants can be adsorbed onto the surface to modify the growth process In the future
with a better understanding of diffusion dynamics and growth mechanisms a multi-stage
fabrication process can be envisaged where purposely-designed molecules are adsorbed
onto silicon surfaces to diffuse organize in ordered arrays and favour surface
reorganization at the nanometre scale The same molecules could be used to deliver
dopants to obtain for example highly conducting nanostructures on otherwise undoped
highly resistive silicon The selective formation of nanostructures at the SiSiO2 border
can be used to scale down mesoscopic patterns to the nanometric level A silicon oxide
layer can be easily patterned on the surface with normal optical lithography and then
nanostructures can be built following the pattern edge In this way the production of very
small nanowire based devices may be possible (fig 72)
84 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
D
Gate
S
D
S
SiO2
SiO2
2 Heating nanowire creation
and oxide removal
3 Source-drain deposition by conventional lithography
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
4 Insulating layer (oxide) and gate (metal) deposition
Fig 72 Schematic drawing of the construction of a self-organized nanowire-based transistor
Surface diffusion and decomposition of molecules can be sensitive to even smaller surface
features The nanoscopic lines formed on Si(100) are very regular and neat if their
nucleation and growth mechanism could be controlled ordered networks of nanowires
could be made on silicon
VPalermo 85
The techniques developed here are still experimental ones the careful control and
reproducibility of each step is still an issue and further studies will be needed to clearly
understand the dynamics of formation of these structures Nonetheless the results obtained
so far open up the possibilities of producing novel devices such as for example nano-
transistors (see scheme in fig73 where the mechanism described in fig 72 is applied for
the fabrication of a device array) The possibility of using the border of a lithographic
pattern to draw nanolines can in principle allow the creation of nanodevices using normal
IC manufacturing techniques
The results obtained and their description in this thesis are not the first examples of the use
of self-organization for the production of ordered structures and will certainly not be the
last In our opinion the most important outcome of these findings is that it is possible to
use surface diffusion to overcome the inherent limits of lithographic techniques in
microdevice production processes and to force properly selected molecules to react with
the surface in a spatially differentiated manner at particular sites thereby creating ordered
series of nanostructures The role of the surface dynamics of the substrate atoms
themselves is extremely important in this process
Outstanding results have been and continue to be obtained in nanoscience and
nanotechnology research promising the emergence of new production technologies Sooner
or later those emerging technologies will have to face the issues of cost and compatibility
with the enormous investment made in existing technologies and processing facilities The
commercial production of nanostructures based on surface diffusion processes could in
principle be cheap simple and compatible with existing technology Moreover the
processes of diffusion in bulk silicon is well-known in the microelectronics industry which
has both the know-how and the equipment to better understand and exploit the surface
diffusion processes on silicon surfaces studied here
Extremely important and innovative results are often obtained by exploiting very simple
ideas and in our opinion physical processes such as surface diffusion used to create
complex nanostructures on silicon surfaces is a prime example
86 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
2 Heating nanowire creation and
oxide removal
3 Source-drain deposition by conventional lithography
4 Insulating layer (oxide) and gate (metal) deposition
Fig 73 Schematic macroscopic outline of the construction of a self-organized nanowire-based transistor
VPalermo 87
Bibliography
[1] P Gargini Intel Technology Coordinator ldquoFrom Microelectronics To
Nanotechnologiesrdquo Invited Campus Colloquia Lecture CNR Research Area Bologna
February 6th 2003
[2] Toward A Hybrid Micro-Nanoelectronics Cerofolini Gf Ferla G Journal Of
Nanoparticle Research 4 (3) 185-191 Jun 2002
[3] Recent Progress In Field Emitter Array Development For High Performance
Applications Temple D Materials Science amp Engineering R-Reports 24 (5) 185-239
Jan 25 1999
[4] Silicon Dioxide Micropillars For Sieving Fabricated By Macroporous Silicon-Based
Micromachining Izuo S Ohji H French Pj Tsutsumi K Kimata M Sensors And
Materials 14 (5) 239-251 2002 Electrochemical Etching In HF Solution For Silicon
Micromachining Barillaro G Nannini A Piotto M Sensors And Actuators A-Physical
102 (1-2) 195-201 Dec 1 2002
88 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 89
Acknowledgements
This work would not have been possible without the help and guidance of Dr Derek Jones
he introduced me to the use of STM and other techniques found the funding for this PhD
and helped me with the thousands of small and big problems I had to face during my
research I am grateful to Prof Alberto Ripamonti too for being the supervisor of this
thesis Thanks are also due to Dr Giancarlo Seconi director of ISOF-CNR where most of
this work was carried out Financial support from the Italian National Research Council
(CNR) is also gratefully acknowledged
My passion for surface science began when Fabio Biscarini showed me that it was possible
to really see the atoms and for this I will always thank him
During these years I had the luck to collaborate with many good scientists among those I
would like to cite Enrichetta Susi Massimo Cocchi and Anna Mazzone from CNR
Claudio Zannoni Anna Cavallini Daniela Cavalcoli and Antonio Castaldini from the
University of Bologna Sergio Pizzini and Maurizio Acciarri from the University of Milan
One of the most stimulating periods of my PhD was the one I spent at the Steacie Institute
for Molecular Sciences in Ottawa working with Robert Wolkow Dan Wayner Greg
Lopinski and Peter Kruse I thank them for discussing together many exciting scientific
ideas and trying to put some of them into practice
Special thanks go to my colleagues Paolo Samorigrave and Stefano De Cesari for long and
useful discussions on Science Life and their interactions often with the help of a good
pint
Finally I would like to thank Vassilia Gaetano Simone Claudia Silvia Letizia
Alessandro Angela and Sebastien even though not directly involved in this work they
have been over these three years a continuous source of happiness and support
Bologna March 2003
90 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
SCIENTIFIC PUBLICATIONS bull Lateral diffusion of titanium disilicide as a way to contacting
hybrid Si-organic nanostructures Palermo V Buchanan M Bezinger A Wolkow RA APPLIED PHYSICS LETTERS 2002 v81 p 3636 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v6 issue 20) bull Surface modifications in Si after Rapid Thermal Annealing Castaldini A Cavalcoli D Cavallini A Jones D Palermo V Susi E JOURNAL OF THE ELECTROCHEMICAL SOCIETY 2002 v 149 pG633 bull Production of nanostructures of silicon on silicon by atomic self-
organisation observed by scanning tunnelling microscopy Jones D Palermo V APPLIED PHYSICS LETTERS 2002 v 80 p 673 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v5 issue 5) bull Nucleation of nanostructures from surface defects on silicon Palermo V Jones D SOLID STATE PHENOMENA 2002 v 82-84 p 687 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo MATERIALS SCIENCE AND ENGINEERING B 2002 v88 (2-3) p220 bull Morphological changes of Si[100] surface after treatment with
concentrated and dilute HF Palermo V Jones D MATERIALS SCIENCE IN SEMICONDUCTOR PROCESSING 2001 v 4 p 437 bull Electrical and structural properties of processed silicon surfaces Susi E Cavallini A Castaldini A Cavalcoli D Jones D Palermo V ldquoRECENT RESEARCH DEVELOPMENTS IN VACUUM SCIENCE amp TECHNOLOGYrdquo 2001 v 3 p189 bull Numerical Solutions of the Stochastic Equations of Crystal
Growth Mazzone A M Palermo V INTERNATIONAL JOURNAL OF MODERN PHYSICS C 2000 v 11 Part 1 p195-204 bull Advances in silicon surface characterisation using light beam
injection techniques
VPalermo 91
Acciarri M Pizzini S Simone G Jones D Palermo V MATERIALS SCIENCE AND ENGINEERING B 2000 V73 (1-3) p 235 - 239 bull Abrupt orientational changes for liquid crystals adsorbed on a
graphite surface Palermo V Biscarini F Zannoni C PHYSICAL REVIEW -SERIES E- 1998 V 57 NUMBER 3A p R2519-R2522 Oral presentations bull The use of oxide desorption and surface diffusion for the
creation of silicon on silicon nanostructures Palermo V Jones D 1ST NATIONAL WORKSHOP ON CURRENT TRENDS IN NANOTECHNOLOGIES Catania (Italy) Feb 2002 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo V E-MRS Spring Meeting Strasbourg (France) June 2001 bull STM study of surface transformations on silicon during UHV
annealing Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999 bull Advances in silicon surface characterisation using light beam
injection technique Acciarri M Pizzini S Simone G Jones D Palermo V E_MRS SPRING MEETING June 1999 Posters bull Spontaneous nano-wire growth on silicon Palermo V Jones D SMARTON workshop Leuven (Belgium) October 2002 bull Ordered circles of nano-islands on silicon from CO adsorption Palermo V Jones D TRENDS IN NANOTECHNOLOGY 2002 Santiago de Compostela (Spain) September 2002 bull Production of nanostructures of silicon on silicon by atomic self-
organisation Palermo V Jones D EUROMAT Rimini (Italy) June 2001 bull Etching holes and anisotropic corrosion on silicon [100] Palermo V Jones D Susi E Asoli B SILICON WORKSHOP Genova (Italy) February 2001
92 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull Morphological and electrical characteristics of damaged silicon surfaces
Susi E Castaldini A Cavalcoli D Cavallini A Jones D Palermo V SILICON WORKSHOP Genova (Italy) February 2001 bull Effect of HF etching on the roughness of a silicon surface Palermo V Jones D NATIONAL CONFERENCE ON PHYSICS OF MATTER Genova (Italy) June 2000 bull Effect of Native Oxide Desorption upon the Surface Morphology of Si[100] by STM and LEED Palermo V Jones D SILICON WORKSHOP Genova (Italy) February 2000 bull Characterisation of silicon surfaces for microelectronics through STM measurements Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999
VPalermo 93
Contact Vincenzo Palermo ISOF- Institute for Organic Synthesis and Photoreactivity Via Gobetti 101 40129 Bologna ITALY Tel +39-051-6398336 Fax +39-051-6398349 Mail palermoisofcnrit
94 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Index
1 Introduction nanotechnology and the future of computers
2 Silicon surfaces
3 STM and other surface analysis techniques
4 Surface modification of silicon in liquid Nano-hole creation
5 Surface modification of silicon in vacuum void creation and oxide desorption
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and
nano-lines
7 Conclusions and proposals for future work
VPalermo 3
Universitagrave degli studi di Bologna
Creazione di isole fili e fori nanoscopici su superfici di silicio per microelettronica
Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Tesi di Dottorato di Ricerca Scienze Chimiche
XV ciclo
Presentata da Vincenzo Palermo
RelatoreProf Alberto Ripamonti
Co-Relatore Dott Derek Jones
Coordinatore di Dottorato Prof Goffredo Rosini
Marzo 2003
4 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 5
OMNIA IN MENSURA ET NUMERO ET PONDERE
Sapientiae Salomonis 1120
6 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 7
1 Introduction nanotechnology and the future of computers The motivations of nanotechnology research
In 1965 an electronic engineer named Gordon Moore one of the future founders of Intel
noted that the performance of computers and their complexity doubled every 18 months
and foresaw that computer power would continue to grow exponentially over the following
years
This prevision quite provocative for its time actually came about and gained the name of
ldquoMoorersquos Lawrdquo and continues to hold for the trends of todayrsquos computer industry Since
1965 the number of transistors present in an integrated circuit (IC) has increased from
several hundred to more than ten million and the minimum size of transistor elements has
shrunk from several millimetres to asymp130 nanometres (fig 11)
Devices of such tiny dimensions are actually fabricated using lithographic techniques
where light is passed through an optical mask to react with a photo-sensitive layer (resist)
Fig11 Evolution of the number of transistors present on commercial computers [1]
8 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
on the silicon wafer This resist is then selectively removed and used as a mask for
processing the silicon surface (fig 12 left) The maximum resolution attainable depends
upon the wavelength used and current technology is near to its intrinsic resolution limit
On the other hand there is strong scientific and economic demand for further development
in IC miniaturization to obtain more powerful and complex computers Besides every-day
life applications more powerful computers are fundamental for much scientific research
such as climate change tracking genome sequencing and fluid dynamics Increased
miniaturization is also fundamental for reducing power which has to be dissipated by the
chips which run at progressively higher frequencies Energy consumption by
microelectronic devices is already an issue and represents one of the main obstacles for
the continuing growth in wireless communication (cell phones portable computers CD
and DVD players digital cameras etc)
Thus it is expected that new production methods different from current lithographic ones
will be developed methods which allow modification of a surface well below the 100 nm
limit and even down to single atom manipulation Techniques such as Scanning
Tunnelling Microscopy and Atomic Force Microscopy are already capable of moving
single atoms (see fig 12 right) but unfortunately building a working nanodevice in this
way would take a very long time and these techniques are difficult to apply to large scale
production
Nowadays thousands of researchers are working in the nanotechnology field towards a
new generation of microelectronic devices Several possible solutions are competing for
tomorrowrsquos computer architecture and there is still no clear winner It is likely that the
final solution will be the combined use of different techniques and components (including
molecules nanowires and nanodots) as they become available together with conventional
Fig 12 Left conventional litographic process [2] Right atomic manipulation of iron atoms on copper [3]
VPalermo 9
silicon technology
Below is a brief summary of the most recent developments in nanotechnology and
nanoscience
Actual trends in nanotechnology
Perhaps the most fascinating idea for nanodevice construction is to use one single
molecule working as a complete device The first molecular diodes (ie molecules
conducting current only in one direction) were created in 1997 in 1999 a molecular fuse
and a molecular transistor were demonstrated although there was no possibility of wiring
these devices to external contacts In April 2001 James Heath and his group at UCLA
fabricated an array of overlapping crossbars and placed a small molecule of rotaxane
between each crossbar (fig 13 left) This composite molecule is made up of two
component parts the main rod-like molecular axis and a mobile ring ldquothreadedrdquo on it like a
bead on a necklace and can function as a molecular switch A working 16-bit memory
circuit was constructed using these molecules For a brief review of these works see [4] In
June 2002 a single molecule transistor was built by connecting an organic molecule to two
metal contacts the molecule contained one or two atoms of a transition metal (cobalt or
vanadium) forming the active region of the device supported by an organic backbone [5]
Fig13 Working nanodevices Left schematic representation of rotaxane molecules between crossed nanowires [4] Right SEM image of semiconductor nanowires forming a small circuit [9]
10 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another approach to nanodevice fabrication has become possible through the discovery of
carbon nanotubes which were observed for the first time in 1991 by a Japanese electron
microscopist studying the material deposited during arc-evaporation synthesis of fullerenes
[6] They consist of a graphite-like carbon seamless cylinder with a diameter of several
nanometers and lengths of up to a millimetre Carbon nanotubes are very stable can
behave as metals or semiconductors and can host other molecular or ionic species thus
modifying their electrical behaviour In 2001 Avouris and his group reported the first
circuit made with a single nanotube [7] A few months later Cees Dekker presented a
nanotube-based transistor able to amplify an input signal by a factor of ten and built
several logic circuits using these nanotube transistors [8]
One problem with carbon nanotubes is that it is very difficult to control their electronic
properties ie their metallic or semiconducting behaviour An alternative to carbon
nanotubes are semiconductor nanowires Silicon nanowires can be made using a laser to
vaporize the silicon together with a metal catalyst like iron or gold The vapour condenses
in nanosized drops of silicon and metal from which the wires slowly grow out as more
silicon is adsorbed In 2001 a group at Harvard University [9] created a transistor by
crossing two different nanowires After this the same group arranged four nanowires in a
noughts and crosses grid creating something like a 4-bit memory (fig 13 right) Even
metallic nanowires made of platinum and silver can be used in a crossed configuration to
store information [10]
There are some issues common to all these new technologies though First it is difficult to
imagine these methods applied to large-scale production Up to now the insertion of a
molecule between two electrodes is an occasional lucky event while nanotubes and wires
have to be positioned on the surface creating the appropriate contacts on them manually
The large-scale production of integrated circuits using these building blocks will not be
straightforward and does not seem imminent
Another issue is of an economic and not a scientific or technological nature Since 1965
the cost of IC manufacturing plants has sky-rocketed If the cost of semiconductor
production plants continues to rise exponentially in a few years such plants will cost up to
$20 billion This is a sizeable investment even for large companies such as IBM or Intel
For this reason it is likely that IC companies will resist changing to completely new
technologies closing down their existing plants As it is clear that silicon will remain the
VPalermo 11
fundamental raw material of the IC industry for the foreseeable future nanotechnology
developments for microelectronics will need to be silicon-compatible In these early days
of nanotechnology the most valid approach would seem to be the addition of molecular
functions to existing silicon technology ndash using the latter as a foundation on which to build
Fabrication of self-organised structures on silicon
The possibility of using the phenomenon of atomic or molecular self-organization to create
nanostructures on silicon has already been demonstrated The clean silicon surface shows
in some cases a high degree of order and complex surface reconstruction as will be
described later Several different ordered structures form spontaneously on this surface
such as series of monatomic steps or boundaries between reconstructed areas It has been
demonstrated that it is possible to use these structures to fabricate ordered nanodots and
nanolines on the surface [11] More recently well-defined nanometric patterns have been
obtained with selective etching of silicon using nitric oxide [12]
In this study the possibility of creating different types of nanostructures on the silicon
surface is explored Methods had to be developed which were
- Simple They must not need complex masks or lithographic steps to create the
structure but rather exploit self-organisation phenomena
- High resolution the silicon surface modifications should be on a scalelength of below
100 nm
- Fast billions of nanostructures have to form over the whole surface simultaneously to
be compatible with large-scale production requirements
- Cheap they must not require expensive equipment (such as e-beam lithography) but
exploit simple chemical andor physical treatments to produce nanostructures on the
silicon surface
In Chapter 2 the main characteristics of silicon are described Chapter 3 provides a
summary of the techniques used for this research Chapter 4 examines the chemical etching
of silicon in different liquid environments and the effects of this etching on the surface at a
nanoscopic level with the creation of nanoholes
12 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Chapter 5 describes the growth in ultra-high vacuum (UHV) of nanoscopic voids and
islands on the silicon surface and the effect of surface oxide on this growth
Chapter 6 discusses the modification of silicon surfaces in UHV following the adsorption
of molecules and thermal heating to produce nanoislands and nanolines on silicon
The overall conclusions of our work are summarised in Chapter 7
Finally we will give some conclusions based on the results obtained and discuss possible
applications of the methods developed
Bibliography
[1] From wwwintelcom
[2] From wwwsematechorg
[3] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
[4] Molecules Get Wired Service RF Science 294 (5551) 2442-2443 Dec 21 2001
[5] Coulomb Blockade And The Kondo Effect In Single-Atom Transistors Park J
Pasupathy AN Goldsmith JI Chang C Yaish Y Petta JR Rinkoski M Sethna JP
Abruna HD Mceuen PL Ralph DC Nature 417 (6890) 722-725 Jun 13 2002 Kondo
Resonance In A Single-Molecule Transistor Liang WJ Shores MP Bockrath M Long
JR Park H Nature 417 (6890) 725-729 Jun 13 2002 Nanotechnology - Electronics
And The Single Atom De Franceschi S Kouwenhoven L Nature 417 (6890) 701-702
Jun 13 2002
[6] Smallest Carbon Nanotube Ajayan PM Ijima S Nature 358 (6381) 23-23 Jul 2 1992
[7] Carbon Nanotube Inter- And Intramolecular Logic Gates Derycke V Martel R
Appenzeller J Nano Letters 1 (9) 453-456 Sep 2001
[8] Logic Circuits With Carbon Nanotube Transistors Bachtold A Hadley P Nakanishi T
Dekker C Science 294 (5545) 1317-1320 Nov 2001
VPalermo 13
[9] Logic Gates And Computation From Assembled Nanowire Building Blocks Huang Y
Duan XF Cui Y Lauhon LJ Kim Kh Lieber CM Science 294 (5545) 1313-1317
Nov 9 2001
[10] Formation And Disappearance Of A Nanoscale Silver Cluster Realized By Solid
Electrochemical Reaction Terabe K Nakayama T Hasegawa T Aono M Journal Of
Applied Physics 91 (12) 10110-10114 Jun 15 2002
[11] Fabrication And Integration Of Nanostructures On Si Surfaces Ogino T Hibino H
Homma Y Kobayashi Y Prabhakaran K Sumitomo K Omi H Accounts Of Chemical
Research 32 (5) 447-454 May 1999
[12] Ultrafine And Well-Defined Patterns On Silicon Through Reaction Selectivity
Prabhakaran K Hibino H Ogino T Advanced Materials 14 (19) 1418-1421 Oct 2
2002
14 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 15
2 Silicon surfaces
The name silicon (silicio in Italian) comes from the latin word silex Amorphous silicon
was first isolated by Berzelius in 1824 by reaction of potassium with silicon tetrafluoride
Thirty years later the first crystalline silicon was prepared Silicon makes up 25 of
earthrsquos crust and is the second most abundant element after oxygen Elemental silicon is
not found in nature occurring as silicon oxide (sand quartz amethyst flint etc) or
silicates (asbestos clay mica etc) Perhaps no other element and its compounds has such
a wide range of uses Silicon compounds such as sand and clay are used in the building
industry as refractory materials for high-temperature applications and for enamels and
pottery Silica is the main component of glass silicon carbide is an important abrasive and
silicones are commonly used polymers and lubricants
Here the most interesting use of silicon of course is for the production of
microelectronics devices For this application silicon of high purity (999999) and of
high crystallinity is needed Table 21 lists some of the physical characteristics of silicon
High purity polycrystalline silicon is produced by the reaction of gaseous trichlorosilane
with hydrogen in a furnace Then to prepare a single-crystal of silicon the so-called
Czochralski method is commonly used
Polycrystalline silicon is melted in a quartz furnace at 1415degC in an argon atmosphere
Then a seed of single-crystal silicon is lowered into contact with the melt and slowly
pulled out In this way the crystal grows and a crystalline cylindrical ingot several metres
long is created from the initial seed
After cooling down the ingot is sliced into thin silicon wafers The wafer surfaces are
polished using a counter-rotating lapping machine in an Al2O3 slurry until the surface is
very flat and shiny ready for the lithographic processes
16 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another way to obtain single crystal silicon is the Floating Zone (FZ) method in which a
silicon cylinder is slowly passed through a heating ring The area inside the ring melts and
solidifies smoothly crystallising as it comes out of the ring yielding a single silicon crystal
Microelectronic devices are built on the silicon surface which is the surface of interest
here Unfortunately silicon surfaces are normally quite dirty and uneven at the atomic
scale Atmospheric oxygen and humidity react with silicon surfaces creating a thin layer of
oxide (called ldquonative oxiderdquo) which is usually irregular and full of defects Different kinds
of contaminants also adsorb onto the surface These are usually small organic molecules
and microscopic dust particles A clean surface on exposure to the atmosphere is
completely covered with gas molecules in less than 10-9 seconds If the pressure is
reduced letrsquos say to 10-6 mbar this time increases to 1 second This is the reason why to
study a clean surface we have to work in UHV at pressures below 10-10 mbar
The atoms in the silicon crystal have a diamond-like structure each atom having 4 bonds
in a tetrahedral sp3 arrangement with bond angles of 10947 degrees At the crystal
surface some atoms will have non-bonding orbitals ldquodanglingrdquo in the vacuum ie sp3
orbitals with a lone electron which are highly reactive These orbitals are known as
dangling bonds To minimize surface energy the surface will reorganize by decreasing
the number of dangling bonds
Table 21 Physical data of silicon [1]
Atomic Weight 2809 Lattice constant (A) 543095 Crystal structure Face-centered cubic
(diamond) Melting point 1415 degC
Density (gcm3) 2328 Boiling point 2355degC Atomscm3 50E22 Minority carrier
lifetime (s) 25E-3
Dielectric Constant 119 Specific heat (Jg degC)
07
Breakdown field (Vcm)
~3E5 Thermal conductivity (Wcm degC)
15
Electron affinity x(V)
405 Vapour pressure (Pa) 1 at 1650degC 1E-6 at 900deg C
Energy gap (eV) at 300K
112 Reactivity Inert to acids Attacked by halogens and alkaline
solutions Intrinsic carrier
conc (cm-3) 145E10 Oxidation states +4 -4
Intrinsic Debye Length (microm)
24 Energy of a Si-Si bond (eV)
232
Intrinsic resistivity (Ω-cm)
23E5
VPalermo 17
Dangling bond densities and positions and thus the type of surface reconstruction will
depend upon crystal orientation as well as the temperature and kinetics of the system
Fig 21 shows a drawing of the main faces of a silicon crystal The angle α between any
(11n) face and the (100) face can be calculated from 2cos 2 += nnα The angle
between any (11n) face and the (111) face can be calculated from
)2(3)2(cos 2 ++= nnα
The chemistry and physics of the faces are very different a brief description will be given
for the most important orientations
Si (100)
On the (100) surface each atom has two Si-Si bonds connecting it to the bulk and two
dangling bonds pointing outward Surface energy is reduced by the dimerisation of the
surface atoms through overlap interaction of one dangling bond per atom forming rows of
dimers aligned along the (110) direction This is the well-known ldquo2x1rdquo reconstruction of
this silicon surface The symmetric dimers would make the silicon surface metallic but to
reduce surface stress the dimers tend to buckle and the surface is thus semiconductive It
took several years to understand that the dimers are buckled because at room temperature
Fig 21 Schematic view of the principal orientations of a silicon surface Surface atoms are white bulk atoms are black dangling bonds are gray [2]
18 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
they shift easily from one buckling direction to the other thus appearing symmetric under
STM observation Fig 22 shows an STM image of the 2x1 reconstructed surface
Even almost perfect (100) surfaces have a certain number of monoatomic steps and the
dimer rows on atomic layers are aligned at 90deg to those on adjacent layers Dimer rows are
thus perpendicular or parallel to the step When the dimers on the upper side of the step
are parallel to the step the step is called SA if they are perpendicular the step is called SB
Because of this symmetry SA step edges will be smoother compared to the more broken
fragmented edges of the SB steps
A common defect on the Si(100) surface is the presence of nickel contamination which
appear as missing dimers in STM images This type of contamination is so critical that
even if the silicon sample is only briefly brought into contact with stainless steel tools
(tweezers for example) the 2x1 reconstruction of the surface can be blocked
Silicon atoms can diffuse easily over the silicon surface as monomers and dimers
especially at elevated temperatures The anisotropy due to the 2x1 reconstruction causes a
difference in the diffusion energies of adsorbates over the surface Diffusion of these
silicon species along dimer rows for example will be much easier A list of diffusion
energies for monomers and dimers is given below [2]
Diffusion on Si(100) 2x1 Ed (eV) Monomers along dimer rows 06 Monomers across dimer rows 085 Monomer formation energy 18 Dimer along dimer rows 11 Dimer across dimer rows 15 Dimer formation energy 26 Dimer binding energy 076 Vacancies along dimer rows 17 Vacancies across dimer rows 19
SA
SB
Fig 22 STM image of a 2x1reconstructed silicon surfaceshowing the dimer rows and steps40x35 nm Nickel-induced defectsare visible as dark spots SA and SBsteps are indicated
VPalermo 19
So the diffusion energy for both monomers and dimers is nearly 40 greater if they have
to cross a dimer row This difference reduces to sim10 for vacancy diffusion
Si(113)
The (113) surface can be imaged as a sequence of alternating (100) and (111)-like
structures with two and one dangling bonds on alternate atoms respectively Interest in the
(113) surface is scientific as it has been used to study the energetics of the (100) and (111)
surfaces as well as for surface adsorption experiments
Si(100) surfaces can easily develop (113) facets
Si(111)
This surface besides being the first one imaged with STM with atomic resolution is one of
the most studied because it is the best cleavage face of silicon and because it shows one
of the most complex and elegant reconstructions in surface science
All Si-Si bonds in the silicon crystal are perpendicular to a (111) plane so this face will
have the lowest number of dangling bonds created per unit area In fact each Si atom on a
(111) surface shows a single dangling bond oriented perpendicular to the surface and
bonded to three back atoms These three bonds for each surface atom account for the great
chemical and physical stability of the Si(111) surface Surface energy is 009 eV Aring-2
compared to 015 eV Aring-2 for Si(100)
For energy minimization this surface reconstructs forming a huge 7x7 lattice cell
containing 102 atoms described by the Dimer-Adatom-Stacking fault model (DAS) For a
detailed description of cell structure see fig 23
The cell described by this model is very complex being composed of three kinds of atoms
adatoms rest atoms and corner hole atoms Furthermore a subsurface stacking fault is
present in one half of the cell making the two halves of the unit cell look different under
STM (Fig 24) It took 26 years of research to completely understand the exact structure of
the 7x7 reconstruction
Cleaving a silicon crystal along a (111) plane produces a metastable 2x1 reconstruction
the 7x7 reconstruction is easily obtained by flashing at high temperature in UHV At T gt
20 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
830degC a disordered 1x1 phase covers the surface Cooling down to 800degC leads to the
formation of the 7x7 phase If the cooling process is too rapid small 7x7 domains nucleate
and a disordered 1x1 phase is preserved between domain boundaries
Si(110)
Even though as mentioned before the (111) plane is the favoured cleaving plane of
silicon thin (100) commercial wafers will not break along this plane because the angle
between (100) and (111) is too far from 90deg (see table 22) Instead they will break along
the (110) plane because it is perpendicular to the (100) surface Each surface atom on
Si(110) has a Si-Si bond pointing downward one dangling bond pointing outward and two
Si-Si bonds parallel to the surface in a zig-zag pattern (see fig 21) Cleaved (110)
surfaces are disordered but upon annealing at high temperatures an ordered complex 16x2
Fig23 Scheme of the 7x7 DAS model [2] In each unit cell there are 9 dimers 12 adatoms and a stacking layer fault The force driving this complex reconstruction is the minimization of dangling bonddensity The DAS model shows the lowest number of dangling bonds (19) of all possiblereconstructions 12 dangling bonds are at the adatoms 6 at the rest atoms and 1 at the corner hole atom This surface is metallic
VPalermo 21
reconstruction takes place The surface appears as a series of long ridges and valleys
parallel to each other Eventually tilted facets of orientation (17 15 1) can form on this
surface The adsorption of Ge atoms on this surface leads to the formation of self-
assembled nanowires [3]
Table 22 Angles in degrees between different silicon faces [2]
Orientation
113
110
111
100
100
2524
9000
5474
0
111
2950
3526
0
110
6476
0
113
0
Fig 24 STM image of a Si(111)surface with 7x7 reconstruction A unitcell with its adatoms is highlightedImage size 13x13 nm
22 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] Weast RC Handbook Of Chemistry And Physics (Chemical Rubber Co Cleveland
1972)
[2] Dabrowski J Mussig H Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] The Structure Of Clean And SiGe-Covered Si(110) Surfaces Butz R Luth H Surface
Science 365 (3) 807-816 Oct 1 1996
VPalermo 23
3 STM and other surface analysis
techniques
Scanning Tunneling Microscopy
Since the invention of the optical microscope at the end of the 16th century the possibility
of examining surfaces at higher and higher magnification has fascinated mankind
Development of the technique continued and towards the end of the 19th century optical
microscopes were as good as todays standard instruments The physical limits of the
wavelength of visible light (350-800nm) had been reached
In the 1920s de Broglie showed that electrons can behave like waves and the use of these
particles for imaging with much higher resolution soon followed Atomic resolution using
this technique is only possible in the transmission mode with extremely carefully prepared
samples
In 1982 using the peculiar properties of piezoelectric materials Binnig and Rohrer brought
a metallic tip very very close to a silicon surface and scanned it across an extremly small
area (fig31) The tunneling of electrons from the tip into the sample or vice versa allowed
them to obtain a local density of electronic states (LDOS) map of that surface Although
theory (which treated the extreme point of the tip as a sphere) then excluded the possibility
of atomic resolution following a tip crash into the surface Binnig and Rohrer began to
observe the LDOS with atomic resolution For this discovery and their following work
they were awarded the Nobel Prize in Physics in 1986
24 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The basic principle of STM is very simple A metallic tip is scanned over a surface without
making ohmic contact and a tunneling current passes between the tip and the surface An
electronic circuit keeps this current constant by raising and lowering the tip during the
scan In this way recording the tip height at each point a three-dimensional image of local
density of electronic states (LDOS) of the surface can be obtained To explain the
extremely high resolution attainable by this simple technique quantum theory is needed
According to classical physics the current will flow between sample and tip only if they
are in physical contact If there is a vacuum gap between the two the electrons will simply
remain confined for example within the surface without the possibility of passing into the
tip
In quantum physics however the electrons have a certain probability of passing
(tunneling) across the gap appearing on the other side of the gap in this way reaching the
tip It can be shown that the probability of an electron tunneling through a gap of thickness
z is
kzep 22)0( minusprop ψ h
φmk
2= (1)
where ψ(0) is the electron wavefunction at the surface-gap border m is the electron
mass=91x10-28 g and φ is the work function of the metal (ie the energy required to
remove an electron from that material For silicon it is 48 eV) The tunneling current thus
decays exponentially with z and is extremely sensitive to topographical imperfections
present on the scanned surface A rough formula giving the current as a function of z is [1]
zFS eEVI φρ 0251)( minusprop
Fig 31 Binnig and Rohrer with the first STM Image from IBM [2]
VPalermo 25
where ρs(EF) is the local density of states at the Fermi level on the given surface For
example the formula predicts that for silicon an increase in tip-surface distance of 1 Aring
will give a 95 decrease in tunneling current
This huge dependence of tunneling current upon the distance allows detection even of the
sub-nanometre changes in height given by the single atoms of which the surface is
composed and thus to resolve them in the LDOS images Of course this description of the
tunneling process is oversimplified and for a more accurate one the electronic states of
the tip of the sample and their interaction have to be taken into account Fig 32 shows a
schematic representation of the interaction between tip and sample orbitals
The exponential decay of current with distance also yields high lateral resolution If the tip
is approximated as a sphere of radius R and the current passing at the minimum tip-sample
distance is I0 then the current passing at a lateral distance x from this point will be
Rxk
eII 22
0
2minus
=
Assuming a tip radius of 100 nm the current is concentrated in an area sim15 nm wide at tip
apex
Very sharp tips with even smaller curvature radii can be produced in several ways Simple
Fig32 Schematic view of tip-sampleorbitals interaction a) no interaction b)equilibrium c) sample positive d) tippositive [1]
26 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
tungsten wires mechanically cut are capable of obtaining atomic resolution on graphite in
air but for more disordered and rougher samples sharper and more reproducible tips are
needed
STM tips are mostly made by electrochemically etching a W or Pt-Ir wire The tips we
used were prepared using methods based on the work of Fotino [3]
A tungsten wire 038 nm diameter is immersed in a KOH solution(06M) with a thicker
tungsten wire used as a counterelectrode The cathodic and anodic reactions involved in the
etching are
Cathode 6H2O + 6 e- rarr 3H2(g) + 6 OH-
Anode W(s) + 8 OH- rarr WO42- + 4H2O + 6 e-
A potential of 3V ac is applied to the tungsten and the wire is immersed in the solution
until a constant current of sim100 mA is established The etching rapidly removes metal
shaping the wire end as a sharp tip When the potential reaches 11 V the coarse tip etching
is finished The wire is then removed from the solution carefully inserted into an
insulating plastic tube leaving only the tip exposed and re-immersed in the solution with
the tip pointing upwards A more gentle etching is thus made to reduce tip radius Usually
5 minutes etching at 07 V ac is used In this configuration very small hydrogen bubbles
formed on the tip sides sliding upwards with a ldquohoningrdquo effect on the tip
This procedure yields extremely sharp and reproducible tips at the microscopic level
After the etching the tip is thoroughly rinsed in ultrapure water then dipped into
concentrated HF to remove surface oxides and hydroxides [4] The tip is dried with
nitrogen inserted into the UHV system and degassed overnight at sim150degC
The possibility of measuring sub-nanometric distances would be useless without being able
to control tip movement over such a minute scale To scan the tip over the surface a
piezoelectric scanner is used Piezos are usually made of an alloy of PbZrO3 and PbTiO3 a
material which contracts or expands when a voltage is applied to it The Omicron
instrument used in our laboratory has three such piezo scanners for xy and z tip motion
allowing one to scan the tip over the surface with sub-Aringngstrom precision (fig 33)
To isolate the instrument from ambient vibrations the whole STM stage is suspended upon
four springs which eliminate all frequencies above 1 Hz and surrounded by a crown of
VPalermo 27
copper wings and fixed magnets Parasitic currents generated by the magnets into the
copper wings contrast every movement of the stage and efficiently block stage vibrations
The STM can be used not only to explore surface topography but to measure the IV
characteristics of single atoms or molecules on the surface (Scanning Tunneling
Spectroscopy or STS) Furthermore it can be used to modify the surface with voltage
bursts digging into it or delicately moving single molecules or atoms over a surface [5] It
can be used in vacuum in air and with proper lateral isolation of the tip even in liquids
A major drawback of STM is that it works only on conducting and semiconducting
surfaces and thus cannot be used on many surfaces of biological and chemical interest
Another instrument more suitable for these and other applications is the Atomic Force
Microscope (AFM)
Atomic Force Microscopy
AFM was invented in 1986 by Binnig Quate and Gerber after calculating the possibility of
building a cantilever with a force constant of the same order of magnitude as that of a
chemical bond
In AFM a tip mounted on a microscopic cantilever (usually made of Si3N4 fabricated with
optical lithography) is brought close to a surface When the tip touches the surface the
cantilever is very slightly deflected upwards The movement is measured by observing the
Y-PIEZO
SILICON SAMPLE
Fig 33 A picture of theSTM used for theexperiments The tripodpiezo scanner is shown
Z-PIEZO
X-PIEZO
TIP
28 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
deviation of a laser beam hitting the upper face of the cantilever Fig 34 provides a
schematic view of the principle of AFM
The typical force constant of the cantilever varies from 00006 to 2 Nm the typical
resonance frequency is 3 to 120 kHz The AFM tip can apply a force on the sample of up
to 10-9 N The AFM can be used on conductive or insulating surfaces in vacuum air or
liquids Furthermore the tip can be modified to sense electrostatic potentials (electric force
microscopy) or magnetic fields (magnetic force microscopy) it can even be functionalized
with complex molecules such as proteins to interact with biological surfaces
A drawback of the AFM is that the force it exerts can damage the surface under
observation especially if the sample is soft (as in the case of cell membranes for
example) This problem can be overcome using the instrument in tapping mode (where the
tip does not move laterally during its brief contact with the surface) or in non-contact mode
in which the tip oscillates above the surface during the scan and the changes in its
frequency due to interaction with surface are monitored The shifts in the oscillating
frequency of the cantilever due to tip-sample interaction are then used for imaging the
surface In this mode interaction of the tip with the surface is minimal and soft samples
can be imaged
STM and AFM are the main techniques used for this work A brief description of other
techniques used occasionally is given below
Fig 34 Scheme of an atomic force microscope
VPalermo 29
Low Energy Electron Diffraction (LEED)
Electrons with energies in the 20-500 eV range are diffracted by a crystalline surface the
diffraction peaks are visualized on a fluorescent screen This technique probes the long
range order of the surface up to a depth of several nanometres
X-Ray Photoelectron Spectroscopy (XPS)
XPS allows both qualitative and quantitative chemical analysis of the elements present on
or near the sample surface
An X-ray source is used to photoionize the atoms on a surface and produce photoelectrons
By measuring the kinetic energy of the photoelectrons the binding energy of the electronic
levels can be calculated This energy will depend on the chemical environment of the
surface atoms
Although the soft X-rays used penetrate to a depth of ~2000Aring the sampling depth of the
technique is determined by the mean free path of the photoelectrons which allows their
escape from only the first 10-100Aring
Secondary Ion Mass Spectroscopy (SIMS)
High and low energy ions (primary ions) are used to bombard a sample and remove surface
atoms and ions The ionic fragments removed (secondary ions) are then analysed by a mass
spectrometer The surface can be consumed during the measurement and profiles obtained
giving concentrations of the materials composing the sample at different depths (depth
profiles)
A popular variant of SIMS is TOF-SIMS In this technique the secondary ion masses are
measured by a time-of-flight (TOF) measurement The secondary ions generated by the
bombarding primary ions are accelerated to a constant kinetic energy and then move
through a field-free space before they reach the detector where their intensity is measured
as a function of flight time Since ions with different masses have different velocities at a
30 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
given kinetic energy the measured flight times of the ions can easily be converted to their
masses The static nature of this latter technique allows mass spectroscopy surface analysis
with minimal damage to the surface
Bibliography
[1] Chen CJ Introduction To Scanning Tunneling Microscopy (Oxford University Press
Oxford 1993)
[2] From wwwibmcom
[3] Tip Sharpening By Normal And Reverse Electrochemical Etching Fotino M Review
Of Scientific Instruments 64 (1) 159-167 Jan 1993
[4] A Convenient Method For Removing Surface Oxides From Tungsten STM Tips
Hockett LA Creager SE Review Of Scientific Instruments 64 (1) 263-264 Jan 1993
[5] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
VPalermo 31
4 Surface modification of silicon in liquid
Nano-hole creation
Liquid treatments of silicon wafers are very common in the integrated circuit (IC)
manufacturing industry They are used to clean and improve surface uniformity to create
and etch protective oxide layers and to remove photo-resist layers
Crystalline silicon with its native oxide layer is very stable and is resistant to many acids
It is easily attacked by hydrofluoric acid (HF) and alkaline solutions
The thin (~2 nm) passivating layer of native oxide (SiO2) is formed on exposure to the
atmosphere This surface layer contains many defects and contaminants so it is usually
chemically stripped and substituted with a better chemically-formed protective oxide
The most common silicon cleaning procedure is the RCA method named after the Radio
Corporation of America [1] It consists of two steps in the first one the surface is treated
with a hot alkaline solution (H2OH2O2NH4OH 411) to remove particles from the
surface following this a hot acidic solution (H2OH2O2HCl 411) is used to remove
metal contamination Other well-known cleaning methods are IMEC (a sequence of
cleaning steps in H2OO3 and dilute HF) or the Pirana etch (a hot 41 mixture of H2SO4
H2O2)
32 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The standard RCA clean removes surface contaminants etches the native oxide and
oxidizes the silicon surface leaving a uniform layer of silicon oxide which better protects
the surface from further contamination
Etching with fluorine-based solutions
Hydrofluoric acid is one of the most common reagents used in the treatment of silicon
wafers both in the research field and in industrial processes A rapid dip in dilute HF is the
simplest way to remove the native oxide from Si(100) and leaves the surface passivated by
a layer of Si-H bonds Because of the low polarization of Si-H bonds the Si-H layer is
stable even for several days protecting the surface from contamination It has often been
assumed that this short etch does not significantly change the surface morphology of the
silicon substrate[2] even though a prolonged dip in dilute HF leads to surface roughening
[3]
Although dilute HF roughens the Si(100) surface at the atomic scale [45] immersion in
concentrated HF (49) etches the surface oxide without attacking the Si surface
uncovering in this way the buried SiSiO2 interface The final counter-intuitive result is
that dilute HF etches the silicon while concentrated HF leaves the crystalline silicon
untouched [3]
Etching Si with fluorine-containing solutions at different concentrations and pH can
produce different morphologies from rough surfaces to flat nearly ideal Si-H terminated
surfaces
Hessel et al and Higashi et al demonstrated in 1991 that very flat Si(111) surfaces can be
obtained using 40 NH4F while etching with HF always yields rough surfaces The
surface becomes smoother because the etchant rapidly attacks Si atoms at step borders
thus removing surface kinks and irregularities in a step-flow mechanism [6 7] Later on
even smoother and more perfect surfaces were obtained by removing oxygen from the
solution after it was discovered that oxygen dissolved in 40 NH4F initiates the formation
of triangular etch pits It was not possible to obtain flat surfaces by etching Si(100) with
ammonium fluoride solutions which leads to the formation of small 2x1 dimer-row
reconstructed (100) terraces together with (111) facets [8]
This difference is caused by the different hydride terminations prevailing on the (100) and
(111) faces While the ideal Si(111)-H surface is monohydride terminated the more
VPalermo 33
reactive dihydrides predominate on the Si(100)-H surface making it more vulnerable to
etching The etching reaction is thus strongly anisotropic etching (100) facets faster than
(111) thus producing (111) microfaceting on Si(100) crystals
A more uniform Si(100) surface can however be prepared by etching at low pH with an
HFHCl mixture [9] or by using very dilute HF solutions and ultrapure water with low
dissolved oxygen and carbon contents [10]
Electrochemical etching can also be used applying anodic or cathodic bias to the silicon
to obtain different morphologies [11] by varying the potential isotropic or anisotropic
etching is observed The aforementioned results show that despite the simplicity of the
reactants fluoride etching of silicon is quite a complex reaction
Fig 41 Chemical etching of silicon
HO+H
H HO
H
H
H
FSi
Si
SiSi
+H2O
-OH-
+F-
H2O H H
Si H
F OH
H Si
Si Si
-H2
+H2O
H
H
H
FSi
Si
SiSi
H
H
H
OHSi
Si
SiSi
H
H
H
H Si
Si
Si Si
Etching mechanism of silicon
HF rapidly dissolves the SiO2 passivating layer on silicon leaving the surface almost
completely hydrogenated [12] After this two different types of reactions etch the silicon
simultaneously one chemical and the other electrochemical [13] The overall etching
mechanism can be schematized in two stages (see Figure 41)
i) Si-H bonds are substituted by Si-F or Si-OH bonds creating a partial charge on the
surface silicon atom and polarizing its Si-Si backbonds
ii) these polarised backbonds are then more easily attacked by HF or H2O After
rupture of the Si-Si bond the atom is removed leaving behind new Si-H
terminations and the reaction can start again
34 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
These reactions take place although at different rates on both Si(100) and Si(111)
Stage i) is usually the rate-determining step of the reaction and the stability of the Si-H
bonds depends upon the pH the concentration of nucleophilic species in solution and an
eventual potential applied to the crystal
For pH gt5 as in the case of concentrated NH4F solutions reaction begins with attack by
water to give Si-OH (step ArarrC) The -OH group is rapidly substituted by fluorine with
polarization of the underlying Si-Si bonds These bonds are then easily attacked by water
the silicon atom being released into solution as HSiF(OH)2 The Si-OH rarr Si-F substitution
is not fundamental for the reaction and etching can proceed even for Si-OH terminated
atoms but XPS measurements showed the presence of a certain number of Si-F bonds
remaining Furthermore fluorine seems to have a catalytic effect on Si-H substitution as
indicated by the dependence of the etch rate upon the F- concentration at least for pH
values between 4 and 8
Si-F bonds can be easily removed by a water rinse In the case of strongly alkaline
solutions (pH=14) OH- groups act directly as nucleophiles and no fluorine is needed to
catalyze Si-H bond rupture
At pH lt4 almost no free F- ions are present in solution and the etch rate is very small at
pH lt2 all etching reactions are very slow and this explains the stability of Si crystals in
concentrated (50 ww) HF solutions
Si Si
Si Si
H
H
H
H Si
Si
Si Si
H
H
H Si
Si
SiSi
H
H
H
OH-H+ -e
(F-)
-H+
-e
+H2O hellipas in fig1
Fig 42 Electrochemical etching of silicon by H2O
The electrochemical etching of silicon involves electron transfer from the surface atoms to
the valence or conduction band of the crystal (according to the type of doping of the
silicon) Several studies have been carried out at different electric potentials In the case of
p-silicon or n-silicon at anodic potentials where positive charges (h+) are available in the
crystal a silicon-centred radical can be created by capture of a hole from the bulk (fig
42) The F- ion does not participate in the reaction but is thought to contribute through
VPalermo 35
electrostatic interactions by lowering the energy of the interaction step After the
formation of the Si-OH group the reaction proceeds as shown in the scheme of fig 41
Matsumura et al [4] proposed that not only water but HF2- molecules also play a major role
in electrochemical etching of silicon leaving on the surface Si-F terminated bonds which
can be immediately attacked in an autocatalytic process (fig 43)
In the electrochemical reactions described above an external potential is applied to the
silicon crystal The chemical and electrochemical reactions in any case take place
simultaneously most of the time with the chemical path predominating at high pH Even
when no external potential is applied to the silicon partial electrochemical reactions can
take place at different ldquocathodicrdquo and ldquoanodicrdquo sites on the surface with an internal charge
exchange which ensures neutrality [11] This macroscopic silicon etching and hydrogen
bubble formation can lead to visually observable patterns on the surface when Si(100) is
immersed in ammonium fluoride even without applying a potential
Si Si
Si Si
H
H
H
F Si
Si
Si SiH
F F-H+ -2e
+HF2-
H
HF
SiSi
SiSi + F
F F
FH
H
H
F -H+ -2e
+HF2-
Fig 43 Autocatalytic electrochemical etching of silicon by HF2-
Inhomogeneities on silicon surfaces caused by electrochemical reactions and charge
transfer have been studied extensively because they are of fundamental importance in the
formation of porous silicon
Pore formation on silicon
When Si(100) or Si(111) are etched under anodic bias in fluorine-based solutions
microscopic pores form on their surface Several different morphologies of pores have
been observed with pore diameters ranging from 10 nm to several microns with depths of
several microns [14] Pore shape is very variable too ranging from ordered straight pores
to chaotic networks of branched pores (fig 44) Porous silicon has been known since the
fifties but it was only in 1990 that interest in this material increased following the
discovery that porous silicon layers were able to emit bright red light This led to a large
amount of research and now different classes of micropores can be reproducibly created
36 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Fig 44 Different types of Silicon micropores From ref [14]
mostly for optical and micromachining applications However there is still no unified
theory able to explain the nucleation and growth mechanism of all the different kinds of
pores
We will give a short description of some of these theories for more detail see Parkhutik et
al [15]
One model explains pore nucleation on the basis of physical processes such as hole
positive charge migration ion transport to the surface and small perturbations on the
silicon surface modelled as Fourier components The system is shown to be unstable and
some spatial frequencies that lead to pore nucleation evolve from the etching process
A second model focusses on stationary pore growth without explaining the nucleation
stage According to this model silicon dissolves preferentially at pore edges because h+
charges are attracted by the stronger electric field present at these edges
A third class of models explains pore growth as a Diffusion Limited Aggregation (DLA)
process where the random walk of h+ charge carriers through the depleted layer present at
the silicon-liquid interface controls pore shape
Finally the model by Carstensen Cristophersen and Foll [16] proposes that areas of the
surface of some characteristic size LCO are etched by synchronized ldquocurrent burstsrdquo in the
flow of h+ charges These bursts dissolve silicon through cyclic stages of surface oxidation
oxide removal and hydrogen passivation Areas where a burst has recently taken place are
less passivated and thus more likely to be etched again in this way the pore bottom
continues to dissolve while the pore walls are passivated and are thus less favourable areas
towards current bursts
VPalermo 37
EXPERIMENTAL RESULTS
In the following sections we will show some experimental results obtained from STM and
AFM measurements of fluorine-treated Si(100) surfaces In the first part the results of
mild etching using concentrated and dilute HF solutions at low pH are presented In the
second part the results of etching at high pH using ammonium fluoride are presented and
the mechanism of pore formation discussed
Etching of Si(100) in dilute and concentrated HF
Samples were cut from different areas of an 8-inch diameter p-doped silicon(100) wafer
(10 Ω-cm) supplied by MEMC Electronic Materials Each series of STM measurements
was carried out over at least six different areas on at least two identical samples Low
Electron Energy Diffraction (LEED) was used to check the surface cleanliness of the
samples before STM measurements
Table 41 summarizes the different treatments of each sample After etching with
electronic grade HF each sample underwent a final rinse in Ultra-Pure Water (UPW
resistivity gt18 MΩ-cm) Both the HF and the UPW were allowed to flow continuously
over the sample surface Some samples were not etched with HF at all but just washed with
UPW to observe the morphology of the native oxide layer (~2 nm thick) covering the
surface All of these processes were carried out under nitrogen to limit reoxidation and the
samples were then introduced from the nitrogen atmosphere directly into the vacuum
chamber and degassed overnight at sim150degC before LEED and STM measurements
STM images were obtained from each sample using the same measurement parameters
(sample bias 4 V feedback current 1 nA scan speed 800 nm s-1) These parameters
38 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
provided a satisfactory level of reproducibility for all the samples Measurements were
made over an area of 500x500 nm (image size 500x500 pixel) Slope correction was
carried out by subtracting row-wise and column-wise fitted slopes from the entire image
which gave better results than the simple subtraction of a fitted plane especially for the
rougher samples Following slope correction the rms roughness
sum minus=xy
hyxhN
22 ))((1σ
and the 2-D Fourier transform
)(22
2
)(4
)( vyuxi
xyeyxhvuF +sum∆
= π
π
were calculated for each image where N2 is the number of pixels composing the image
h(xy) is the surface height at each point ∆ is the distance between points h is the mean
height and u v are the spatial frequencies The radial power spectrum PS(f) of the STM image data is obtained from the angular
average of the squared Fourier transform with f 2 = u2 + v2
Fig 45 shows the STM images obtained from the various samples Sample A still covered
with its native oxide layer shows an irregular surface with RMS roughness of ~05 nm
(see Table 41) Observing the sample with LEED gave no diffraction pattern even at
relatively high incident electron energies because of the surface oxide coverage After 1
min etching in dilute HF (sample B) the morphology is similar to the original one though
Table 41 Sample treatments average RMS roughness and integrated area of power spectra
Sample Treatment RMS roughness (nm)
PS area (f lt 01 nm-1)
PS area (f gt 01 nm-1)
A Rapid dip in water 053 plusmn 013 366 032
B 1 min in HF 5 + 10 min in water 051 plusmn 008 272 034
C 30 min in HF 5 + 10 min in water 062 plusmn 008 1064 038
D 5 sec in HF 49 + 10 min in water 042 plusmn 004 183 022
VPalermo 39
Fig 45 STM images of each group of samples showing the topography of the silicon surface A) noetching original oxide surface B) after 1 min etching in dilute HF C) after 30 min etching in diluteHF D) after dipping in concentrated HF Grey scale indicates height of the surface from lower (black) to higher (white) The images are 250x250 nm ie representative portions of the images usedfor the roughness measurement and PSD analysis
some of the larger features have disappeared and the image quality is better maybe due to
improved tunnelling due to the cleaner surface The RMS roughness is comparable to that
of the original surface Clear diffraction patterns are visible using LEED though at quite
high energies (200 eV) After prolonged etching (sample C) the RMS roughness increases
to 062 nm and a long-range corrugation is visible on the surface even if the LEED pattern
is good
The samples dipped in concentrated HF (D) reveal the bare SiSiO2 interface which has a
disordered aspect and protrusions over a wide range of dimensions The quality of the
STM images of sample D is very good probably due to the cleanliness of the surface
40 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 E -0 2
1 E -0 1
1 E + 0 0
1 E + 0 1
1 E + 0 2
1 E + 0 3
0 0 0 0 0 1 0 1 0 1 0 0f (1 nm )
nm^4
A a s re c e iv e d
B e tc h e d 1 min
C e tc h e d 3 0 min
D e tc h e d HF 4 9
Fig 46 Log-Log plot of the averaged power spectra of the STM images for all the samples
which gives a more stable tunnelling junction The LEED pattern is excellent showing
clear diffraction peaks at energies as low as 37 eV comparable to that obtained after high
temperature cleaning in UHV
Fig 46 shows the power spectra of the samples The high frequency and low frequency
areas of the power specturm are considered separately Table 41 shows for each sample
together with the roughness the integrated area of the power spectrum for the high and low
frequency part
We first examine the differences between the samples in the low frequency part of the
spectrum (lt01 nm-1) In this spatial range the short dip in concentrated or dilute HF
(samples BD) seems to lower the roughness of the sample removing some of the larger
features of the native oxide surface visible in Fig 45a and thus decreasing the
corresponding part of the power spectrum Sample C however etched with HF for 30 min
shows an increase in roughness peaking at 0012 nm-1 (~85 nm corresponding to the
typical dimensions of the corrugations visible in Fig 45c) On this scale sample C has a
PS density of 435 nm4 compared to 101 nm4 for sample A
Samples A B and C have the same PS in the high frequency range showing that the fine
structure of the surface is not changed by the HF etching Sample D on the other hand
shows a decrease of roughness for frequencies gt 015 nm-1 This effect could be due to the
VPalermo 41
improved surface cleanliness after etching with concentrated HF which would give a more
stable STM junction thus reducing the high frequency noise in the image
Fig 45 and the analysis of the power spectrum of each sample shows that a rapid dip in
HF removes the native oxide but does not lead to major changes in the morphology of the
Si surface its only effect being the removal of some of the larger features present on the
original surface Prolonged etching on the other hand increases the RMS roughness of the
surface
Etching of Si(100) in concentrated ammonium fluoride and nano-hole creation
Two different types of commercial p-doped Si(100) wafers (2Ω-cm and 10Ω-cm) from
MEMC were used Several different samples of 10x5 mm were immersed for 10 minutes
in 40 electronic grade NH4F solution under agitation Previous works used low
temperatures or anodic potentials applied to the silicon to avoid gaseous hydrogen
production and to obtain a uniform surface but in our experiment we worked at room
temperature to check the influence of hydrogen bubbles on surface morphology During
the etching the stirring was sufficient to provide a uniform concentration of reagents over
the whole sample surface but not to mechanically remove the hydrogen bubbles from the
silicon surface
After the etching each sample was rinsed in ultra-pure water to remove any etching
residues and observed with STM AFM and optical microscopy The AFM measurements
were made in air while for STM measurements the samples were rapidly dried with
nitrogen and inserted into the vacuum system to avoid surface reoxidation After insertion
into the vacuum surface cleanliness was checked with LEED and the surface morphology
observed by STM Parameters for STM measurements were sample bias +4 V 1 nA
current 16 Hz scan rate The images obtained were stable and reproducible over several
days Scan parameters for AFM were 20 nN force and 1 Hz scan rate
Some of the samples were cleaned with an RCA standard clean [1] before NH4F etching to
check the influence of possible surface contaminants on the final results Eight different
samples were prepared and more than sixty STM images of the samples were taken at
different points of the various samples
42 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
(110)
(110)
Fig47 abc) STMtopographic images of differentetching morphologies Eachimage is 500x500 nm Z-ranges are 10 10 and 18 nmrespectively d) STM image of a bridgecreated by etching of the lowerlayers of silicon (black arrow)Image is 250x250x6 nm
After ~2 min of immersion in the solution hydrogen bubbles become visible on the sample
surface The production is slow and the bubbles are quite stable on the sample without
detaching Thus some areas of the surface are masked from the liquid etching action
STM observations (fig 47) show that at the nanometer scale the surface is unevenly
covered with holes of radii ranging from 10 to 200 nm with depths of 2-4 nm These holes
have a wide range of different shapes and distributions In most cases the surface was
covered with a uniform distribution of round-shaped holes (fig47a) indicating isotropic
etching The dimension and the density of the holes changed greatly from sample to
sample and even over the surface of a single sample In some cases the etching was
anisotropic yielding nearly square holes and layered structures as shown in fig 47b
Square holes have been previously observed in cases where the etching speed in the (110)
direction is significantly smaller than in the (100) direction [17]
Over these areas (fig 47b and especially 47c) it is clearly visible how once the surface
had been attacked the reaction continued to preferentially remove atoms at step
irregularities (kink atoms) straightening step edges Eventually the exposed underlying
silicon was also attacked and further holes created inside the previously etched larger
ones It was not possible to detect monatomic steps on this kind of surface The smallest
step height observed was ~15 nm corresponding to several atomic layers In the image
shown in Figure 47a the etching was not strong and created only anisotropic holes on the
surface In fig 47b and c the stronger etch proceeded laterally for several tens of
VPalermo 43
nanometres leaving straight steps several tens of nanometers long and roughly rectangular
holes as expected given the structure of the (100) crystal face In some cases a
significative underetch is observed and the formation of suspended bridges and tunnels
can be deduced from the STM images (fig 47d)
The formation of branched pores and suspended structures has been attributed during pore
formation to diffusion limited aggregation effects where the h+ charge carriers necessary
for silicon etching have a higher probability of reacting at pore bottoms than reaching the
upper part of the silicon surface In the case of very deep pores quantum wire effects have
been invoked to explain the pore growth mechanism [15] In our case though the pores
formed were very shallow the underetch depth being only a few nanometres on pores of
sim100 nm width Thus more than diffusion effects the main contribution to the
underetching process must come from anisotropic etching and some kind of autocatalytic
reaction path analogous to the one described by Matsumura et al [4] with some areas of
the silicon surface hydrogenated and thus less vulnerable to etching
Pre-treatment with RCA cleaning has no effect on the final morphology and this seems to
exclude pore nucleation being caused by presence of metallic or organic surface
contaminants
The morphology and the distribution density of the pits was quite uniform over
microscopic areas of the sample but changes were observed over the millimetre scale This
suggests that etching intensity is influenced by some large-scale parameter
Large-area measurements made with AFM or with an optical microscope (fig 48)
showed that the inhomogeneity of surface etching can be correlated with the masking
action of the bubbles While the fluoride dissolved the silicon hydrogen bubbles formed
by the reaction covered some areas of the surface thus blocking the etching over that area
generating macroscopic steps at the bubble-liquid border As the reaction proceeded more
hydrogen accumulated and the bubble diameter increased producing in this way a circular
pattern of steps The increase in bubble diameter was not continuous with time otherwise a
uniform surface slope gradient would have been obtained The formation of this circular
ldquoetching staircaserdquo indicates that the bubble growth was stepwise the bubble accumulating
more and more hydrogen without enlarging across the surface until it relaxed increasing
its diameter stepwise and covering more silicon The circular structures in fig 48a are not
co-axial and their asymmetry could derive from physical processes due to stirring or
irregularities on the surface
44 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
a b Fig 48 a) optical micrograph of etching patterns on Si(100) created by NH4F 12x09 mm b) AFM image of the circles border xy range is 40x24 microm z-range is 30 nm
The step structure was not destroyed by the etching even after the bubble detached from
the surface but on the contrary the etching process seemed to be influenced by the
presence of the step
Observing in detail a series of steps (fig 48b) a quite deep trench is visible at the base of
each step A close-up image of a step and the corresponding line profile of the trench is
shown in fig 49 The trench is sim5 nm deep with respect to the lower surface compared to
a step height of 22 nm
A similar structure has been recently obtained with electrochemical etching of p-type
Si(100) in 4 HF [18] in which a ldquocurrent burstrdquo etching model previously described
was assumed for silicon dissolution In that experiment the trench was created at the
border of silicon nitride masks and began to grow after a nucleation stage Preferential
trench etching was along the (110) direction and the trench growth was explained as an
effect of mechanical stress induced by the nitride mask and of electric field enhanced
dissolution which depended upon an external applied potential
While it is clear that in our system the gas bubbles have a masking effect similar to a
classical solid nitride mask it is unlikely that hydrogen present on the surface can induce a
significant stress in the silicon lattice as in the case of a nitride mask Furthermore no
external field was applied to drive preferential etching at the trench site
It has been proposed [19] that the cathodic and anodic part of the etching reaction
(hydrogen production and silicon oxidative etching respectively) take place at different
points on the surface with a net charge transfer between the different areas In this case
the highest reaction rates will correspond to the silicon area surrounding the bubble border
where a high number of positive charges will be available for the reaction Furthermore a
sharp trench extending into the silicon crystal will be a preferential electrostatic attractor
VPalermo 45
Fig 49 AFM image of the etched surface showing a step created on the surface by bubblemasking A stronger etching action is visible on the right side of the step as well as theprotected area on the upper side of the step (indicated by the arrows) Image is 10x10 micromz-range is 30 nm The profile on the right is taken from the central area of the image
for the h+ charge carriers coming from other ldquocathodic areasrdquo of the sample either from
other regions on the surface or from the back of the silicon chip [18]
In the areas where the hydrogen bubble had detached and the surface was exposed to the
etching the reaction was not uniform in the neighbourhood of the steps It is possible to
observe (fig 48b and 49) an area on the upper side of the step where less or even no
etching at all seems to have taken place as if the step was able to protect the surface from
etching While etching on the lower side with trench formation can be attributed to the
presence of the bubble the surface on the upper step side can be etched only after bubble
detachment so no masking effect can account for this result However a further
preferential attraction of h+ charge carriers from the already formed trench can be
hypothised electrochemically shielding the surrounding area from further etching If this is
true the shielding effect would be very strong with a relatively shallow 5 nm-deep trench
protecting an area of sim1 microm parallel to the step
To summarise the etching of Si(100) in NH4F is a complex process in which different
reaction paths both chemical and electrochemical co-exist Hydrogen bubbles formed by
the reaction act as a mask on the surface and create etching paths and inhomogeneous
etching of the surface Different kinds of pores are observed on the surface and in some
cases the anisotropy of the process is so strong as to give square-shaped holes and
underetching
46 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The diffusion of h+ charge carriers in the crystal is one of the main rate-determining steps
of the reaction and leads to the formation of a deep trench immediately outside the bubble
perimeter These trenches act as charge collectors and reduce the etching of the upper step
surface in the proximity of the steps
Bibliography
[1] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[2] Spectroscopic Ellipsometry Studies Of HF Treated Si (100) Surfaces Yao H Woollam
Ja Alterovitz SA Applied Physics Letters 62 (25) 3324-3326 Jun 21 1993 Influence
Of HF-H2O2 Treatment On Si(100) And Si(111) Surfaces Graf D Bauermayer S
Schnegg A Journal Of Applied Physics 74 (3) 1679-1683 Aug 1 1993 Kinetics Of
Oxidation On Hydrogen-Terminated Si(100) And (111) Surfaces Stored In Air Miura
T Niwano M Shoji D Miyamoto N Journal Of Applied Physics 79 (8) 4373-4380
Part 1 Apr 15 1996
[3] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995 Structure Of The Stepped SiSiO2 Interface After Thermal-Oxidation -
Investigations With Scanning Tunneling Microscopy And Spot-Profile Analysis Of
Low-Energy Electron-Diffraction Pietsch GJ Kohler U Jusko O Henzler M Hahn
PO Applied Physics Letters 60 (11) 1321-1323 Mar 16 1992
[4] Enhanced Etching Rate Of Silicon In Fluoride Containing Solutions At pH 64
Matsumura M Fukidome H Journal Of The Electrochemical Society 143 (8) 2683-
2686 Aug 1996
[5] A Study Comparing Measurements Of Roughness Of Silicon And SiO2 Surfaces And
Interfaces Using Scanning Probe Microscopy And Neutron Reflectivity Crossley A
Sofield CJ Goff JP Lake ACI Hutchings MT Menelle A Journal Of Non-Crystalline
Solids 187 221-226 Jul 1995
VPalermo 47
[6] Step-Flow Mechanism Versus Pit Corrosion - Scanning-Tunneling Microscopy
Observations On Wet Etching Of Si(111) By Hf Solutions Hessel HE Feltz A Reiter
M Memmert U Behm RJ Chemical Physics Letters 186 (2-3) 275-280 Nov 8 1991
[7] Comparison Of Si(111) Surfaces Prepared Using Aqueous-Solutions Of NH4F Versus
HF Higashi GS Becker RS Chabal YJ Becker AJ Applied Physics Letters 58 (15)
1656-1658 Apr 15 1991
[8] Wet Chemical Etching Of Si(100) Surfaces In Concentrated NH4F Solution -
Formation Of (2x1)H Reconstructed Si(100) Terraces Versus (111) Faceting Neuwald
U Hessel HE Feltz A Memmert U Behm RJ Surface Science 296 (1) L8-L14 Oct
10 1993
[9] Ideal Hydrogen Termination Of Si(001) Surface By Wet-Chemical Preparation Morita
Y Tokumoto H Applied Physics Letters 67 (18) 2654-2656 Oct 30 1995
[10] Atomic Structures Of Hydrogen-Terminated Si(001) Surfaces After Wet Cleaning
By Scanning Tunneling Microscopy Endo K Arima K Kataoka T Oshikane Y Inoue
H Mori Y Applied Physics Letters 73 (13) 1853-1855 Sep 28 1998
[11] On The Potential-Dependent Etching Of Si(111) In Aqueous NH4F Solution
Houbertz R Memmert U Behm RJ Surface Science 396 (1-3) 198-211 Jan 20 1998
[12] Etching Process Of SiO2 By HF Molecules Hoshino T Nishioka Y Journal Of
Chemical Physics 111 (5) 2109-2114 Aug 1 1999
[13] Etching Mechanism And Atomic-Structure Of H-Si(111) Surfaces Prepared In
NH4F Allongue P Kieling V Gerischer H Electrochimica Acta 40 (10) 1353-1360
Jul 1995
[14] Pore Formation Mechanisms For The Si-HF System Carstensen J Christophersen
M Foll H Materials Science And Engineering B-Solid State Materials For Advanced
Technology 69 23-28 Sp Iss Si Jan 19 2000
[15] Porous Silicon - Mechanisms Of Growth And Applications Parkhutik V Solid-
State Electronics 43 (6) 1121-1141 Jun 1999
[16] Formation And Application Of Porous Silicon Foll H Christophersen M
Carstensen J Hasse G Materials Science amp Engineering R-Reports 39 (4) 93-141 Nov
1 2002
[17] Pore Morphology And The Mechanism Of Pore Formation In N-Type Silicon
Searson PC Macaulay JM Ross FM Journal Of Applied Physics 72 (1) 253-258 Jul 1
1992
48 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
[18] Deep Electrochemical Trench Etching With Organic Hydrofluoric Electrolytes
Christophersen M Merz P Quenzer J Carstensen J Foll H Sensors And Actuators A-
Physical 88 (3) 241-246 Mar 5 2001
[19] Morphology Of Anodically Etched Si(111) Surfaces - A Structural Comparison Of
NH4F Versus HF Etching Houbertz R Memmert U Behm RJ Journal Of Vacuum
Science amp Technology B 12 (6) 3145-3148 Nov-Dec 1994
VPalermo 49
5 Surface modification of silicon in vacuum void creation and oxide desorption
The main reason for the huge success of silicon in the microelectronics industry is not due
to its superior properties as a semiconductor Other materials for example germanium
have better qualities such as higher mobility of charge carriers and lower noise levels
which would allow the construction of faster and higher performance devices
The widespread use of silicon however is mainly due to the outstanding characteristics of
its oxide Silicon dioxide (SiO2) is a very good electrical insulator easy to form
chemically and thermally stable and is compatible with lithographic and metal deposition
processes Germanium oxide on the contrary is too reactive to be used
Even the use of Si(100) substrates for nearly all microelectronic devices is dictated by
oxide quality The (111) face of silicon crystal can be easily cleaved and flattened and
almost atomically perfect surfaces can be obtained with simple chemical procedures (as
described above) But the density of interfacial defects is highest for Si(111)-SiO2
interfaces and lowest for Si(100)-SiO2 ones so microchips will continue to be fabricated
on Si(100) wafers
SiO2 (silica) is present in 95 of the earthrsquos minerals in different allotropic forms such as
quartz tridymite and cristobalite In the bulk each silicon atom is bonded to four oxygens
in a Si-O-Si tri-dimensional network Si-O bonds are 016 nm long and form an angle
ranging from 120deg to 150deg
Three typical intrinsic defects are present in SiO2 The so-called Ersquo centres are oxygen
vacancies with a hole localised on a silicon atom with only three Si-O bonds
50 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
O3Si +SiO3 Whereas the PR (peroxy radical) defects are holes trapped by a charged
peroxy moiety with a O3Si-O-O+ SiO3 structure The NBOHC (non-bridging oxygen hole
centres) derive from water or hydrogen contamination and are schematized as O3Si- O- H-
O-SiO3
The atomic structure of the Si-SiO2 interface varies enormously Local domains resembling
the tridymite and the cristobalite structure of silica are present but it seems that only 10
of the interface is ordered [1] Far from the interface the SiO2 bulk is completely
disordered The passage from bulk Si to stoichiometric SiO2 passes through a non-
stoichiometric SiOx layer 07 nm thick
When a clean silicon surface is exposed to atmospheric oxygen a thin sim2 nm thick layer
of native oxide forms spontaneously which is usually removed and substituted with
thicker better quality oxide layers before further processing
Silicon is usually oxidised by thermal annealing at temperatures between 800deg and 1100deg
in an atmosphere of pure O2 with some water eventually added to increase oxidation speed
Thermal oxides made in pure oxygen (dry oxides) grow more slowly than oxides produced
in an oxygen-water atmosphere (wet oxides) but are usually of better quality
According to the Deal-Groove formula the time t needed to grow an oxide of thickness X
is given by [1]
1
212minus
minusminus
+=
ABXBXt α
where the constant B and BA decrease exponentially with temperature as
minus
kTEexp
with activation energies for dry oxidation of EB =123 and EBA =20 eV respectively
EB is related to the diffusion of oxygen in silicon while the value of EBA is interpreted as
the energy required to break a Si-Si bond The exponent α is 1 for wet oxidation and 0 for
oxidation at high temperatures and low oxygen pressures It has intermediate values for dry
oxidation This formula does not work well for low values of X at the initial stages of
oxidation and usually empirical corrections are used
An interesting characteristic of silicon is that at high temperature and in vacuum oxygen
can actually etch the silicon crystal giving gaseous products and the oxidized layer present
on the crystal becomes unstable (fig 51) [2]
VPalermo 51
1E-10
1E-09
1E-08
1E-07
1E-06
1E-05
1E-04
1E-03
1E-02
1E-01
0607080911112
1000T (1K)
P (T
orr)
SiO2 + Si rarr 2SiO(g) Oxide decomposition
Si+O2 rarr SiO2(s) oxide formation Fig51 phase diagram of the
oxygen-silicon system
Silicon oxidation apart from the initial nucleation stages at the monolayer level proceeds
uniformly over the whole surface with a planar reaction front moving from the surface
into the bulk
If heated under low oxygen partial pressure (vacuum or inert atmosphere) SiO2 is known to
decompose following the reaction
SiO2 + Si rarr 2SiOuarr (1)
The reaction begins with nucleation at defect points on the SiSiO2 interface and proceeds
in a spatially inhomogeneous manner with the formation of large voids on the oxide
surface [3]
Several studies have been made on the dynamics of void growth both on thick [4] and thin
[5] layers of SiO2 The process has been used to decorate otherwise unobservable defects at
the SiSiO2 interface [6] or to grow nanoislands of silicon on the void surface [7] It has
been suggested that the defects acting as nucleation centres could be metallic contaminants
present on the native surface which aggregate and catalyze SiO2 decomposition [8]
After oxide desorption the silicon surface is very rough In particular on Si(100) square
islands are observed several nanometres high which act as lsquopinning sitesrsquo on the motion of
monatomic steps on the Si surface It has been proposed that these islands can be composed
of silicon carbide coming from the organic contaminants present on the surface [9] or of
oxide clusters during partial reoxidation [10]
In the following section some experimental results obtained studying voids and nanoisland
growth are presented
52 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nanoisland growth on silicon
We used two different samples the first rougher than the second (i) p-type Czochralski Si
(100) with a RMS roughness of 05 nm shown in fig52a and (ii) p-type epitaxial Si(100)
with a RMS roughness of 019 nm shown in fig 52b The roughness was measured with
STM on the lsquoas receivedrsquo samples on different areas of 500x500 nm Every sample was
covered by a layer of native oxide ~2 nm thick
Some of the samples (AC) were introduced into the UHV system without any cleaning
others (BD) were dipped in concentrated HF (49) to remove the surface oxide without
etching the silicon [11]
After degassing the samples were heated resistively increasing the temperature slowly to
keep the pressure within the 10-10 mbar range during heating Each sample was held at
900˚C for 30 min For some of the samples the heating was stopped at lower temperatures
to obtain incomplete oxide desorption and surface roughening thus enabling observation
of the different steps of the process Table 51 summarizes the different treatments for each
group of samples
Fig52 Original surfaces of Czochralski (a) and epitaxial (b) silicon Every image is 500x500 nm
VPalermo 53
Table 51 Summary of different treatements
Silicon type Oxide removed
Heating Islands density (microm-2)
Av Island Volume (nm3)
A Czochralski No 30rsquo at 900˚C 24x103 353
B Czochralski Yes ldquo 20x104 71
C Epitaxial No ldquo 28x102 2450
D Epitaxial Yes ldquo 23x104 63
Fig53 ab) Surface roughening on group A samples after heating 30 min at 800deg and900deg respectively in presence of an oxide layer Image size 500x500 nm cd) Surface roughening on group B samples after heating 30 min at 700deg and 900degrespectively after removing the oxide layer with HF Image size 250x250 nm
Fig 52a is the native oxide surface of sample A quite disordered and irregular This
surface was stable when heated to 700˚C while at higher temperatures the oxide began to
desorb Fig 53a was taken after heating the sample at 800˚C for 30 min
54 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The oxide began to desorb in correspondence with defect points creating oxide-free
surface voids which enlarge radially uncovering the silicon surface Some surface silicon
atoms are removed through reaction (1) Mobile silicon atoms moving randomly over the
surface aggregate forming a nanocrystal at the initial defect point surrounded by a stable
flat oxide-free surface with some steps The initial SiSiO2 interface as observed after
oxide etching with HF [12] is similar to the one shown in fig 52a and thus the surface
observed in the void area is a completely new one coming from surface reorganization
The reaction continues leading to a coalescence of the voids the final result of the process
is shown in fig 53b with protruding islands aligned along the main crystal axes and flat
areas with some steps LEED measurements confirmed that this surface is crystalline
silicon Atomic resolution was obtained on the flat areas of the surface showing the
typical 2x1 reconstruction of Si(100) Figures 53c and 53d show the evolution of the
roughening on sample B after chemically removing the native oxide layer before
introducing the sample into UHV As mentioned above the initial morphology is similar to
that shown in fig 52a though the oxide has been removed as confirmed by LEED
measurements Fig 53c shows the surface after heating at 700˚ for 30 min Without the
oxide layer the surface began to reorganize at lower temperatures and the small irregular
grains of the original surface begin to grow and become more rectangular After heating to
900˚C (fig 53d) the surface is flat with small rectangular islands The shape of the islands
is similar to that shown in fig 53b but the lateral island dimension is one order of
magnitude smaller
To assess the importance of the initial surface on the process the same treatments were
repeated starting from flat epitaxial silicon (samples CD) The initial epitaxial surface is
smooth with periodic steps due to a small miscut angle (fig52b) On heating the epitaxial
silicon without etching (sample C) relatively large islands are obtained similar to the ones
shown in fig 53b where the ldquostep-pinningrdquo effect of the islands is evident The average
dimension of sample C islands was sim2500 nm3 while only 353 nm3 for sample A Island
density decreased from 24000 to 280 microm-2 This result is maybe due to a lower
concentration of defects in the oxide layer of the epitaxial silicon which allows the
expanding void to grow much larger before coalescence with neighbouring voids the
nanostructures within them thus reaching greater dimensions To verify this hypothesis we
applied the same treatment to clean not-so clean and purposely contaminated samples To
check that the defects inducing void nucleation and nanoisland growth derive from carbon
VPalermo 55
contamination we used electronic-grade methanol to contaminate the silicon surface
Figure 54 shows the results of this comparison
State-of-the-art cleaned samples obtained with multi-step RCA cleaning [13] showed the
development of very few voids the oxide desorption leaving large smooth areas of oxide-
and nanostructure-free silicon as shown in Fig 54a where the presence of residual oxide
can be seen Samples cleaned with a simpler one-stage Pirana cleaning solution showed a
higher number of islands per unit area (Fig 54b) and samples purposely contaminated
with methanol (Fig 54c) had the highest density of island nucleation of all three samples
When the native oxide had been chemically removed from the epitaxial silicon before
heating (sample D) the island size and density is similar to the standard as-received
silicon case (sample B) Thus starting from a flat substrate the process of surface
roughening and island creation was the same even when the oxide layer had been
previously removed
From these results some important indications can be drawn The first is that oxide
desorption temperature depends upon the contamination levels of the oxide layers (fig 54)
with desorption at lower temperatures for contaminated surfaces
Second once the oxide layer is removed the silicon atoms become mobile on the surface
at relatively low temperatures (~700˚C) Their initial configuration is disordered and has a
high surface energy so they crystallize in the small islands shown in fig53d reducing
their surface area and lowering the energy of the system When the oxide is present the
surface cannot reorganize at 700˚C and is stable At 800˚C the oxide begins to desorb in
correspondence with defect points on the surface As the void area enlarges a small
cba Fig 54 The effect of contamination on nano-island production Gradient-filtered STM images of (a)sample cleaned two times with RCA 400x400 nm (b) Sample cleaned with a Pirana solution 500x500nm (c) sample contaminated with MeOH before insertion into the UHV system 500x500 nm Annealingtime is 40 min at 800degC for all the samples
56 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
crystallite grows at the centre of the voids Further oxide desorption uncovers larger areas
freeing more silicon atoms which feed the initial island making it bigger
Fig 55 shows the scaled size distribution of the islands for each group of samples Every
distribution has been obtained using several images of different samples for each group
The size distribution of all the samples follows an exponential decay and the decay seems
the same for all the samples though the average volume of the islands differs by more than
one order of magnitude (see Table 51 for details)
The point defects which catalyze void nucleation can be metallic contaminants present on
the initial oxide surface [8] or organic contaminants which at high temperatures can
form SiC nanocrystals [14] We used Scanning Tunneling Spectroscopy (STS) to look for
differences between the islands and the surrounding flat silicon surface but no difference
was found Furthermore the total island volume per surface unit is very high (more than
104 nm3microm2) and it seems unlikely that such a huge volume could consist of surface
contaminants Islands of pure silicon on silicon have been grown without evidence of
surface contamination[14]
To summarise the formation of nano-sized islands on silicon through surface diffusion
was studied At high temperatures the oxide layer covering the surface decomposes non
uniformly and circular voids of clean silicon are created The presence of the oxide layer
blocks surface silicon atomic motion and surface reorganization except within the voids
causing the growth of islands more than 10 nm high and 30 nm wide on the silicon surface
According to STM and LEED measurements we can say that the islands are mostly
Fig 55 Size distribution (scaled) for the samples described in table 1 The line is a reference for the eye
VPalermo 57
composed of silicon with traces of other substances such as carbon Final island densities
and dimensions depend upon the initial purity of the oxide layer
On the other hand if the oxide is removed chemically before heating the island growth
process is not so localised and takes place simultaneously over the whole surface yielding
smaller and more numerous islands
58 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[2] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[3] High-Temperature SiO2 Decomposition At The Sio2Si Interface Tromp R Rubloff
GW Balk P Legoues FK Physical Review Letters 55 2332-2335 Nov 1985
[4] Defect Microchemistry At The SiO2Si Interface Rubloff GW Hofmann K Liher M
Young DR Physical Review Letters 582379-2382 Jun 1987 Defect Formation In
Thermal SiO2 By High Temperature Annealing Hofmann K Rubloff GW Mccorkle
RA Applied Physics Letters 49 1525-1527 Dec 1986 Kinetics Of High-Temperature
Thermal Decomposition Of SiO2 On Si(100) Liher M Lewis JE Rubloff GW Journal
Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 5 1559-1562 Aug
1987
[5] Thermal-Decomposition Of Very Thin Oxide Layers On Si(111) Kobayashi Y Sugii
K Journal Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 10 (4)
2308-2313 Part 3 Jul-Aug 1992 Scanning Tunneling Microscope Study On Mid-
Desorption Stages Of Native Oxides On Si(111) Kobayashi Y Sugii K Journal Of
Vacuum Science amp Technology B 9 (2) 748-751 Part 2 Mar-Apr 1991 Controlled
Growth Of SiO2 Tunnel Barrier And Crystalline Si Quantum Wells For Si Resonant
Tunneling Diodes Wei Y Wallace RM Seabaugh AC Journal Of Applied Physics 81
(9) 6415-6424 May 1 1997
[6] Defect Formation In SiO2Si(100) By Metal Diffusion And Reaction Liher M
Dallaporta H Lewis Je Appl Phys Lett 53 589-591 Aug 1988 SiO2 Film
Decomposition Reaction Initiated By Carbon Impurities Located At A Si- SiO2
Interface Raider Si Herd Sr Walkup Re Applied Physics Letters 59 (19) 2424-2426
Nov 4 1991
[7] Nanometer-Scale Si Selective Epitaxial Growth On Si(001) Surfaces Using The
Thermal Decomposition Of Ultrathin Oxide Films Fujita K Watanabe H Ichikawa M
Applied Physics Letters 70 (21) 2807-2809 May 26 1997 Pyramidal Si Nanocrystals
VPalermo 59
With A Quasiequilibrium Shape Selectively Grown On Si(001) Windows In Ultrathin
SiO2 Films Shibata M Nitta Y Fujita K Ichikawa M Physical Review B 61 (11)
7499-7504 Mar 15 2000
[8] Stacking-Fault-Induced Defect Creation In SiO2 On Si(100) Liher M Bronner Gb
Lewis Je Appl Phys Lett 52 1982-1985 May 1988
[9] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990 AFM And XPS Characterization Of The
Si(111) Surface After Thermal-Treatment Lamontagne B Guay D Roy D Sporken R
Caudano R Applied Surface Science 90 (4) 481-487 Dec 1995
[10] Scanning-Tunneling-Microscopy Study Of Oxide Nucleation And Oxidation-
Induced Roughening At Elevated-Temperatures On The Si(001)-(2x1) Surface Seiple
JV Pelz JP Physical Review Letters 73 (7) 999-1002 Aug 15 1994 Evolution Of
Atomic-Scale Roughening On Si(001)-(2x1) Surfaces Resulting From High-
Temperature Oxidation Seiple JV Pelz JP Journal Of Vacuum Science amp Technology
A-Vacuum Surfaces And Films 13 (3) 772-776 Part 1 May-Jun 1995
[11] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995
[12] Morphological Changes Of The Si [100] Surface After Treatment With
Concentrated And Diluted HF Palermo V Jones D Materials Science In
Semiconductor Processing 4 (5) 437-441 Oct 2001
[13] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[14] Nanoscale Roughening Of Si(001) By Oxide Desorption In Ultrahigh Vacuum
Gray SM Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology
B 14 (2) 1043-1047 Mar-Apr 1996
60 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 61
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and nano-lines
When the silicon surface is not protected by a native oxide layer or by a hydrogenated
passivating layer it is very reactive The surface chemistry of both Si(100) and Si(111) has
been extensively studied starting from clean surfaces prepared in UHV because of its great
relevance to the microelectronics industry and its technological interest The diffusion and
reaction of molecules and atoms on silicon is an interesting scientific problem on its own
apart from technological issues because silicon reconstruction yields a very complex and
anisotropic surface
We give here a brief summary of the adsorption behaviour of several elements on Si
surfaces The summary is not comprehensive and only the most interesting characteristics
for each substance are given
Table 61 Summary of adsorption behaviour of atoms and molecules on silicon surfaces [12]
Hydrogen
Molecular H2 shows low reactivity towards silicon while atomic hydrogen easily forms
Si-H bonds and can even break Si-Si bonds
Alkali metals
Alkali metals diffuse rapidly into SiO2 and can damage silicon-based transistors
Transition
metals
All transition metals apart from gold and silver react with Si forming metal silicides
62 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nickel A very common contaminant of silicon Nickel from even slight contact with stainless
steel tweezers can contaminate silicon samples forming its silicide and is very difficult
to remove even at high temperatures because it segregates on the silicon surface
Palladium Forms silicides especially Pd2Si and can be used to make contacts possessing a low
Schottky barrier
Titanium Widely used to fabricate contacts for silicon-based microdevices TiSi2 is one of the
more stable and highly conductive silicides
Tungsten The silicide is more stable than TiSi2 on polysilicon and is thus used for polysilicon
gate contacts
Platinum Silicide is used for bipolar transistors PtSi contacts on n-silicon give Schottky diodes
Cobalt Cobalt silicide gives better quality contacts than TiSi2 at the nanometrelevel but is less
used at the moment because it consumes too much silicon during its growth
Iron Silicides have been studied for potential optoelectronic applications because FeSi2 has a
directly accessible bandgap and is optically active
Group 13
Group 13 metals are used in IC technology to dope silicon (giving p-type doping) and
to make III-V type semiconductors When evaporated on Si(100) or Si(111) they react
strongly with the surface giving surface reconstruction and kink formation at steps
Boron Widely used for doping Usually deposited with decomposition of boron hydrides
Sticking coefficient of simple hydrides is very low so decaboranes (B10H14) are used for
deposition
Aluminum Aluminum is widely used for microelectronic contacts but in some cases it reacts with
silicon diffusing into the crystal and with SiO2 stealing oxygen atoms to form Al2O3
Group 14
Carbon Carbon can form a composite semiconductor with silicon (silicon carbide) widely used
to fabricate integrated circuits for use at elevated temperatures or in the presence of
ionising radiation When carbon is deposited on silicon usually polycrystalline films
with properties between diamond and graphite are obtained
CO adsorbs molecularly on silicon without breakage of the C=O bond Other
unsaturated hydrocarbons do not react with silicon Even very reactive strained
molecules like cyclopropane have low sticking coefficient Ethylene acetylene and
propylene adsorb molecularly on Si(100) each molecule sitting on a silicon dimer
interacting with the dangling bonds of the surface Acetylene also adsorbs molecularly
on Si(111) on the 7x7 reconstructed surface preferential adsorption on centre atoms
with respect to corner atoms is observed Benzene and other aromatic ring compounds
adsorb molecularly on the surface They can form σ or π-type bonds with the silicon
laying parallel or tilted respect to the surface according to the bond type
Silicon Silicon atoms can be deposited on the surface with molecular beam epitaxy or
VPalermo 63
decomposition of silanes and chlorosilanes Silanes with single Si-Si bonds have a high
sticking probability and dissociate upon adsorption giving trihydrides and
monohydrides
Chlorosilanes have high sticking coefficients too and decompose on the surface but
they can etch the surface through the reaction SiCl4 + Si rarr 2SiCl2
Germanium Germanium has a lattice constant similar to silicon (Ge lc is 4 larger than Si) so that
various alloys of Si1-xGex can be formed The growth of Germanium on silicon is quite
peculiar with several monolayers adsorbing uniformly on the surface followed by
island formation (Stranski-Krastanov growth)
Group 15
Nitrogen Exposure to ammonia and nitrogen gas at high temperatures leads to the formation of
silicon nitride layers (Si3N4) NH3 dissociates on the surface and reacts with the
dangling bonds giving Si-NH2 and Si-H termination Silicon nitride is stable on the
surface up to 1100degC above this temperature it desorbs as Si2N
Phosphorous Phosphine (PH3) adsorbs dissociatively as Si-PH2 and Si-H and behaviour is similar to
that of ammonia
Arsenic A monolayer of As is often deposited on silicon as a substrate for GaAs growth Arsenic
forms dimers on Si(100) creating dimer rows which are parallel (perpendicular) to the
underlying silicon dimer rows when it is deposited at high (low) temperature On
Si(111) Arsenic breaks the 7x7 reconstruction giving a 1x1 pattern This 1x1 As
monolayer acts as a passivating layer and can resist further As adsorption oxygen and
air
Antimony and
Bismuth
Due to their large covalent radii Sb and Bi form only short dimer rows on silicon and
only at high temperatures
Group 16
Oxygen
Oxygen can oxidize or etch silicon according to the reactions
Si(s) + O2 rarr SiO2 (s)
Si(s) + frac12O2 rarr SiO(g)uarr
Silicon oxidation has already been described in the previous chapter Molecular oxygen
adsorption on silicon is mostly dissociative O atoms break the Si-Si bonds forming an
Si-O-Si bridge or a peroxy bridge Si-O-O-Si Hydrided silicon surfaces do not have
surface dangling bonds and are thus more protected from oxygen attack Etching is
favoured at high temperatures and low oxygen pressures [3] At T gt700degC silicon oxide
on silicon decomposes as SiO(g)
Water
H2O adsorbs dissociatively on silicon easily on Si (100) and with more difficulty on
Si(111) The molecule decomposes giving Si-OH and Si-H on adjacent sites
64 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Group 17
Fluorine
Fluorine adsorbs and reacts easily on silicon dissociating on the surface and forming
Si-F bonds It can then penetrate the surface and attack polarised Si-Si backbonds The
surface is etched with production of SiF4 and SiF2
Chlorine
Chlorine reacts aggressively with silicon etching the surface Cl2 and F2 are often used
commercially to etch silicon usually in a plasma Atomic Cl and Cl2 adsorb on Si(100)
up to saturation The most stable configuration seems to be a Cl atom bonded to each Si
atom of a surface dimer in a symmetric structure but metastable configurations with
two Cl atoms bonded to a buckled dimer and even a central Cl atom bridged across a
dimer have been observed
On Si(111) silicon mono- di- and tri-chlorides are formed In particular heating at
Tgt400degC a uniform Si-Cl monolayer is obtained and the 7x7 reconstruction changes
into a bulk like 1x1 lattice The surface transition induced by Cl adsorption is fully
reversible The 7x7 reconstruction can be restored on desorbing the chlorine by heating
at Tgt1100degC The 1x1 domains nucleate at the lower terrace side of steps the 7x7 at the
upper step edge
Bromine
Like hydrogen bromine maintains the 2x1 structure of Si(100) forming Si-Br bonds
with the dangling bonds of the silicon dimer rows At high exposure some etching of
silicon by formation of volatile SiBr3 species has been observed
Surface diffusion on silicon
In the previous chapter we observed nano-island growth on silicon surfaces with native
oxide and surfaces contaminated with organic impurities While this phenomenon has been
observed in several experiments the formation and growth mechanisms of nanoislands has
still not been satisfactorily described theoretically Various substances have been
hypothesised as nanoisland nucleating agents such as organic or metallic contaminants
present on the surface or oxide clusters (see previous chapter) The island growth process
has been attributed by various workers to lsquosome sort of kinetic instabilityrsquo [4] to the
pinning of step flow by SiC clusters [5] or to a mesoscopic atom flux from areas with low
step density to areas of higher step density [6]
One mechanism proposed to explain surface roughening and island growth is the Ehrlich-
Schwoebel effect ie the presence of an energetic barrier that reflects atoms approaching a
VPalermo 65
Molecule (L
O2
H2O MeOH
CO CO2 CH4
clean surface
downward
weak on si
The aim o
surface an
To obtain
the ones
spontaneou
first obser
surface W
nanoisland
Experime
We used S
removed f
41 mixtur
water and
whole hea
1200degC re
kept below
Table 62 Treatment island density and island volume for each molecule tested
Dose angmuir)
Heating time at 800degC (min)
Island density (microm-2)
Ave Volume (nm3)
Equivalent layer thickness (nm)
280 10 no islands - -
90 10 no islands - - 30 10 2000 19 004
180 10 360 516 019 30 60 no islands - -
180 10 no islands - - - 10 no islands - -
step [7] Recent experiments however indicate that the Schwoebel effect is
licon [48]
f our experiments was to study the reaction of simple molecules with the silicon
d to check the ability of these molecules to nucleate nanoisland growth
a better understanding of the process we chose simple molecules starting from
which are more likely to be present as traces in UHV chambers where
s nanoisland growth is often observed after sample heating Using STM we
ved at the atomic level the possible interactions of these molecules with the
e then heated the surface and checked the density and dimensions of the
s produced
ntal procedures and results
i(111) wafers p-doped 0015 Ω cm Gross contamination and particles were
rom the sample surface with a standard Pirana etch (15 min dip in a H2S4H2O2
e at T=80degC) After this the samples were thoroughly washed in ultra-pure
introduced into the UHV system After degassing for 5 hours at 600degC the
ting stage was allowed to cool down and a sequence of rapid flashes to 1100deg-
moved the native oxide layer and cleaned the surface Pressure during flashes is
1x10-9 mbar
66 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Surface flatness and cleanliness were checked with STM and the contaminant molecules
introduced into the UHV system through a leak valve Doses of several Langmuirs were
used for the experiments (1 Langmuir=10-6 torr sec-1) Pressure was kept constant at 10-7
mbar during adsorption The STM tip was retracted during gas deposition to avoid any tip
shielding effect After the adsorption process the leak valve was closed the base pressure
in the UHV chamber returning to below 10-10 mbar and the presence of adsorbed
molecules checked on the surface with STM Finally the sample was heated at 800degC for
several minutes to allow nanoisland growth Several different areas were examined on each
sample
Table 62 summarizes the dose of exposure and the heating time for each molecule Other
experiments were carried out using smaller dosing times but only the significant results
are reported here
As expected oxygen and water react with the surface oxidising it The adsorption process
can be easily followed with the STM by lowering the tip from time to time and taking an
image No effect of the STM tip is observed over the scanned areas Upon molecule
adsorption the order of the crystalline surface rapidly degenerates and after several
minutes the surface looks completely covered by irregular atomic-sized protrusions
corresponding to Si-OH or Si-O-Si species On heating to 800degC the contaminating species
were easily desorbed through SiO formation the surface became clean and the crystal
surface periodicity re-established No residues or nanoislands were observed
Amongst the molecules tested was methanol CH3OH reacts readily with the surface and
covers it with a disorderd layer already after sim 30 Langmuirs exposure (fig 61)
Fig 61 Si(111) surface during methanol adsorption at 0 1 and 5 minutes respectively (corresponding to 0 6 and 30 L exposure) The dark irregular lines are borders between different 7x7 domains Image size 60x60 nm
VPalermo 67
Synchroton radiation photoemission spectroscopy experiments by Carbone et al [9]
showed that methanol reacts with the rest atoms of the 7x7 silicon cell already after 1
Langmuir exposure but that longer exposures are needed for methanol to react with the
remaining silicon atoms Methanol adsorbs dissociatively forming SiOCH3 and Si-CHx
species on the surface
After methanol adsorption the sample is heated to 800degC and observed again with STM
As in the case of water and oxygen the heating restores the crystalline surface and the 7x7
pattern returns visible But this time quite a dense array of nanoislands is observed on the
surface (fig 62) The islands have an average diameter of 20 nm and a height of sim2 nm
At temperatures above 400degC the Si-OCH3 and Si-CHx species decompose oxygen and
hydrogen are desorbed and according to [9] carbon atoms remain as SiC dispersed
uniformly on the surface Our STM measurements indicate that the carbide does not
randomly cover the surface but that C atoms are concentrated at the nanoislands Rough
calculations based on island volume suggest that the islands must be composed of a SixC1-x
alloy with x varying between 05 and 1
Following these results we expected carbon monoxide to behave in a similar manner
adsorbing onto the surface and with nanoisland growth However no adsorption was
observed with STM and the surface looked perfectly clean even after a dose of more than
100 L of carbon monoxide Some species did adsorb onto the surface though because
after 10 minutes subsequent heating nanoisland growth was observed Island density in
this case was only 18 of the density obtained with methanol and island dimensions are
Fig62 Nanoislands on Si(111) created after 30 L adsorption of methanol and 10 min heating at 800degC Image size 200x200 nm
68 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
much bigger (see table 62 and fig 63) To check whether the islands had been nucleated
by some other contaminant present in the vacuum chamber we heated a blank sample
without introducing any molecule No island growth was observed Exposing the surface to
other simple molecules including carbon dioxide and methane also gave no nanoisland
formation
We can estimate the amount of carbon deposited on the surface during gas dosing The
total flux of molecules hitting the surface is obtained from the Hertz-Knudsen formula
TkmPF
π2=
Fig 63 STM images of Si(111) after exposure to methanol (left) and carbonmonoxide (right) contaminating molecules and subsequent nanoisland growthImage size 400x400 nm The size distribution of nanoislands is shown below eachimage
VPalermo 69
Fig64 Left Si(111) after 180 L contamination of CO and 16 hours heating at 800degC Right Si(111)without CO contamination after 16 hours heating at 800degC Image size 250x250 nm Vertical scale is thesame for both images z-ranges are 35 and 5 nm respectively
where P is the pressure in Pascal k=138x10-23 J K-1 is the Boltzmann constant T is the
temperature and m is the molecular weight (sim32 for methanol and sim28 for CO) in
kilograms
Using this formula we obtain a flux of 027 molecules nm-2 sec-1 for methanol and 029
molecules nm-2 sec-1 for CO After ten minutes exposure at 10-7 mbar pressure more than
150 molecules will have hit each square nanometer of the surface Surface density of
Si(111) 7x7 is sim16 atomsnm2 so each surface atom will be hit by several molecules which
could react or be adsorbed even assuming a sticking coefficient much lower than unity
To explain nanoisland growth we hypothesize that CO molecules adsorb molecularly and
-5
0
5
10
15
20
25
30
0 20 40 60 80 100 120
nm
nm
10 min40 min16 h16 h- no CO
Fig65 STM profiles of nanoislandsgrown for different annealing times The lower curve correspond to the surface heated at 16 hours without CO contamination
70 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Table 63 Island density and volume for the CO-contaminated surface at different heating times For each sample the thickness of a uniform layer having the same volume per unit area of the islands is calculated
Molecule Dose
(Langmuir) Heating time at
800degC (min) Island density
(microm-2) Ave Volume
(nm3) Equivalent layer thickness (nm)
CO 180 10 362 517 019 CO 180 40 550 1072 059 CO 180 16 h 505 3675 186
clean surface - 16 h rough - -
very weakly on Si(111) moving rapidly on the surface Because of their rapid diffusion
the STM is not able to image the CO molecules on the silicon [10]
Increasing the temperature part of the CO molecules will surely desorb but some of them
will dissociatively adsorb onto the surface The oxygen will then desorb as SiO while
carbon will remain fixed forming very stable Si-C bonds
CO decomposition will be favored at surface defects surface steps or in the proximity of
already formed SiC clusters An increased reactivity of the CO molecule on a surface in
correspondence with phase boundaries has already been observed on Pt surfaces [10]
In this case a reduced number of nucleation centres will react with the CO molecules
yielding large and fewer islands with respect to the case of methanol which does not have
the possibility of travelling long distances over the surface
Other factors are likely to influence the process of island nucleation after CO adsorption
the co-adsorption of other molecules which can slow down CO and favour its
decomposition cannot be ruled out as well as the formation of new defects at higher
temperatures Measurements with a variable temperature STM or with some other surface
analysis technique are planned in the future to confirm the proposed mechanism
When the surface was further annealed the nanoislands grew in size After 16 hours
heating very large nanoislands with diameters of sim35 nm and heights of sim20 nm became
visible (fig64 left) As a comparison when the same surface was heated without
nucleation centres deriving from contamination an irregular surface was obtained (fig64
right)
While the nanoisland density seemed to reach an asymptotic limit of 500 islandsmicrom2
(table 63) island size continued to grow with time (fig65) even when no further carbon
was supplied to the surface This indicates that even though the initial nucleating core of
the nanoisland is likely to be an SixC1-x alloy further growth is due to silicon atoms
diffusing from the crystal and being adsorbed by the growing island The final volume
VPalermo 71
occupied by the islands corresponds roughly to a 18 nm thick overall layer of removed
silicon
It is noteworthy that the clean sample after 16 hours annealing even though smoother than
the CO-contaminated one looked much more disordered with hardly any flat area visible
on nanometres scale In the presence of nanoislands however further annealing increases
island size but keeps part of the surface quite flat and ordered with flat areas visible
between the islands Thus the nanoislands can be imagined to act as ldquoimpurity sinksrdquo for
further contaminants approaching the surface yielding a greater but more ordered surface
roughening with respect to the uncontaminated silicon surface
Circular arrays of nanoislands
We used the voids described in the previous chapter formed during oxide layer
decomposition as nanoscopic masks to test the validity of the hypothesis outlined above
The small circular areas uncovered during oxide decomposition can act as ldquoskating rinksrdquo
where contaminants and diffusing silicon atoms can move freely over the surface while
the oxide layer all around and higher than the void area as well as being chemically
bonded to the surface silicon atoms will hinder surface diffusion Furthermore the oxide
passivating layer will at least in some cases prevent contaminating molecules from
adsorbing and decomposing on the surface In this way selective nanoisland growth inside
void areas can take place
The experimental procedure consisted of the following steps
1) An oxide-passivated sample was heated for 10 minutes at 800degC Oxide decomposition
began in correspondence with surface defects and spread laterally creating naked areas
of silicon
2) While the void enlarges the same defect that initiated oxide decomposition behaved as
an attractor for moving silicon atoms and nucleated the growth of a nanoisland at the
void centre
3) The void surface was exposed to contaminating molecules
4) The sample was then heated again at 800degC During the second heating the void
continued to enlarge and new nanoislands were created The freshly uncovered silicon
surface provided mobile silicon atoms to feed nanoisland growth
72 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
5) When the heating continued for long enough all the oxide was desorbed and a clean
silicon surface covered by circular groups of nanoislands was obtained
Figure 66 shows the different stages of the process
We thus heated an oxidized sample at 800degC for 10 minutes in UHV to create multiple
voids in the oxide layer After this we exposed the sample to the contaminating molecules
Then we further annealed the surface at the same temperature to promote island growth
The first time we observed nanoisland growth in the void area was by chance after heating
for a second time samples which had been stored for 2 weeks in vacuum Even at pressures
as low as 10-11 mbar some molecules will adsorb onto the surface slowly creating defects
and thus potential nanoisland nucleation centres It is possible to observe with STM that
the silicon crystal surface looks very clean just after a high temperature flash but even
after only a few days randomly adsorbed species will be observable on the surface
Unfortunately it is not possible to identify these species simply from in situ STM
measurements
Mass spectrometry measurements show that the residual gas contaminants in vacuum are
usually He Ar H2 CH4 CO CO2 and N2 [11] Some of these molecules such as
hydrogen will stick to the silicon surface at room temperature but will simply desorb
during annealing without nucleating nanoislands Others like CH4 or CO2 will not react
with the surface to generate nanoislands as demonstrated previously The best candidate
for contamination of samples stored in UHV seems thus carbon monoxide has the ability
321
4 5
Fig 66 Selective nanoisland growth within oxide voids See text for details
VPalermo 73
as shown before to nucleate nanoislands although co-adsorption mechanisms can not be
ruled out Fig 67 (left) shows the circular groups of nano-islands obtained after heating
void-covered samples stored for 2 weeks in UHV Small islands are observed in the former
void area The void itself has enlarged slightly uncovering a clean oxide-free area of the
silicon surface Apart from the central island created during the first period of heating the
surrounding ones have dimensions decreasing from the void border to the inner area of the
void
This confirms that most of the material needed for island growth comes from the silicon
atoms diffusing from the freshly uncovered areas around the original void The outer
nucleating centres are nearest to the silicon atom source and thus generate larger islands
For a faster process we directly exposed the void to a significant concentration of CO
Fig67 (centre) shows the islands grown after a 180 Langmuir exposure and subsequent
heating The islands are preferentially located around the original void perimeter the rest
of the original oxide-free surface remaining untouched This suggests that although the
whole surface is exposed to CO the gas only interacts chemically with the surface at the
SiSiO2 interface around the oxide-free void perimeter The SiSiO2 interface thus provides
preferential nucleation sites for nanoisland growth in the successive thermal annealing
step This recalls a similar phenomenon observed in the catalytic oxidation of CO on the
Pt(111) surface where CO molecules reacted with adsorbed O species only along the
perimeters of oxygen islands [10]
On repeating the CO exposure and thermal annealing steps a second circle of nano-islands
Fig67 Left circular area of nanoislands obtained from a sample stored 2 weeks in UHV 500x500 nmCentre nanoisland circle obtained after exposing the voids to 180 L of CO and heating for 10 min at800degC 400x400 nm Right two concentric circles obtained with further CO adsorption followed byheating 650x650 nm
74 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
was formed around the newly-expanded surface void perimeter (fig67 right) This
process can in principle be repeated again to obtain a series of concentric groups of
nanoislands the only limit depending on the minimal distance between neighbouring
expanding voids We have thus observed that contaminants adsorbed at low pressure over
long periods of time generate random island growth over the whole void area while CO
molecules adsorbed at relatively high pressures over short periods yield selective
nanoisland growth at the former void border Several theories can explain the different
results CH4 and CO2 as mentioned above did not nucleate islands when when the silicon
surface was exposed to them for short periods However the lower surface mobility of
these contaminants and the greater time available for reaction with the surface could be
the cause of nucleation of SiC clusters over the whole oxide-free void area and not only at
its border The coadsorption of water another UHV residual gas with CO on the naked
silicon areas could perhaps allow a surface reaction similar to that between CO and the
SiSiO2 interface during CO exposure Another explanation for the formation of these
random nucleation sites within the voids could be the coadsorption of residual carbon-
containing species present in the UHV system with hydrogen the major residual gas in
stainless steel UHV systems
The possibility of decorating silicon oxide borders with nanoislands using the high
mobility and selective decomposition of CO molecules suggests interesting applications
for nanofabrication processes that will be discussed below
Silicon nanowire creation on Si(100)
We performed most of the experiments described above on silicon (111) 7x7
reconstructed because this surface is easily prepared in UHV and above all because it is
isotropic and has no preferential directions for atom diffusion The 2x1 reconstruction of
silicon (100) even though it has a much simpler unit cell than Si(111) 7x7 shows a more
complicated morphology with lots of monoatomic and biatomic steps and dimer rows
parallel to each other The orientation of the dimer rows changes by 90deg on alternate
atomic layers Surface diffusion in this case is much more complicated because atoms will
experience the effects of the surface anisotropy and diffuse preferentially along or across
the dimer rows The diffusion energy of a silicon atom as example is 06 eV along a
dimer row and 085 eV across different dimer rows This while being a complication
VPalermo 75
offers interesting possibilities for the creation of ordered structures The monoatomic steps
and the dimer rows can for example act as templates for the formation of elongated
structures of composition similar to the islands described previously
On heating a Si(100) surface oxide decomposition and void growth takes place as
described for Si(111) In some cases the void shape reflects the surface symmetry
especially for very thin oxide layers and the voids often have a slightly squared shape
Nanoisland growth is also influenced by the substrate symmetry and the islands look
rectangular as irregular clusters surrounded by concentric patterns of monoatomic steps
The whole structure reminds vaguely a ldquozigguratrdquo a kind of stepped pyramid found in
Mesopotamia (fig68 inset)
Apart from nanoisland growth in some cases we observed spontaneous formation of
nanowire-like SiC structures on the Si(100) surface The wires were generated after
exposing a Si(100) surface in vacuum to traces of CO followed by sample annealing at
800degC for 15 minutes The wires exhibit lengths ranging from 10 to 100 nm and average
widths of ca 5 nm Although being randomly positioned on the surface they are perfectly
aligned along the crystal axes of the substrate (Fig68)
Several descending monoatomic steps can be observed around each nano-line and nano-
island Very often one of the line extremities coincided with an island
Nanoline
Nanoisland
Fig 68 A Si(100) surface covered by nanoislands and nanowires The monoatomic steps present on the surface are visible The two insets show a typical nanoisland and a nanowire obtained on this surface The typical ldquodimer-row ldquostripes are visible along the sides of the nanowire
76 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Couples of parallel lines starting from the same nano-island were also observed At high
resolution (fig 69) the lines showed a lamellar periodic structure with a periodicity of
around 4 nm and frequent defects Scanning Tunneling Spectroscopy measurements
showed no significant difference between the line and the surrounding silicon surface
The proposed formation mechanism for this kind of structure involves two stages in the
first one organic contaminating molecules adsorb and diffuse on the surface from the
vacuum environment stopping preferentially at surface singularities including steps and
defects
Several substances such as atomic carbon carbon monoxide ethylene and fullerene
decompose when adsorbed on silicon surfaces at temperatures above 500degC and create
localised defects on the surface inducing strain deformations and a metastable surface
reconstruction [12]
If these molecules remain attached to the steps they can diffuse rapidly along step border
until they meet a line and decompose In this way long lines of SiC defects parallel to the
surface steps can be produced In the second phase the clusters of silicon carbide can
-005
0
005
01
015
02
025
03
035
04
-3 -2 -1 0 1 2 3
V
dId
V lt
dId
Vgt
linesilicon
Current- Voltage behaviour measured with STS Fourier analysis of wire periodicity
asymp 4 Aring
High resolution image of a line
Fig 69 Top high resolution image of a nanowire showing the lamellar structure Fourier analysis (bottom left) shows that the lamellae periodicity is sim 04 nm bottom right the IV characteristics measured with the STM tip on the line and the silicon surface
VPalermo 77
25-600degC
Final linear structure
600-800degC
Contaminant Silicon
Fig 610 Schematic representation of nanowire formation mechanism
locally inhibit the spontaneous surface diffusion of silicon atoms acting as templates for
the reorganisation of surface steps (fig 610)
The final morphology shows disordered lines and islands each surrounded by a complex
pattern of silicon monatomic layers
At 800degC silicon atoms are highly mobile on the surface Due to atom diffusion the
surface reorganizes with a step-flow mechanism The presence of a fixed line of
contaminants blocks atom diffusion and step flow generating the monoatomic steps
around each wire (fig 611) SiC clusters act as lsquopinning sitesrsquo on silicon atom surface
diffusion and can be purposely used to direct the localised growth of nano-islands [13]
Although the possibility of using methanol and carbon monoxide as nano-island precursors
has been demonstrated the exact nature of the substances nucleating nano-wire growth is
still unknown
Further experiments will be needed to confirm these hypotheses as well as to identify the
contaminating species nucleating nanowire growth Carbon monoxide is a main candidate
for nanowire growth but the real nucleation process is likely to be a complex one
78 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Step flowblocked
Fig 611 Representation of step flow blocked by the presence of a nanowire 105x105nm
Bibliography
[1] Surface-Chemistry Of Silicon Waltenburg HN Yates JT Chemical Reviews 95 (5)
1589-1673 Jul-Aug 1995
[2] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[4] Growth Of Si On The Si(111) Surface Lanczycki Cj Kotlyar R Fu E Yang Yn
Williams Ed Das Sarma S Physical Review B 57 (20) 13132-13148 May 15 1998
[5] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
VPalermo 79
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990
[6] Nanoscale roughening of Si(001) by oxide desorption in ultrahigh vacuum Gray SM
Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology B 14 (2)
1043-1047 Mar-Apr 1996
[7] Step Motion On Crystal Surfaces Schwoebel Rl Journal Of Applied Physics 40 614-
618 Feb 1969
[8] Scanning Tunneling Microscopy Investigation At High Temperatures Of Islands And
Holes On Si(111)7x7 In Real Time Evidence For Diffusion-Limited Decay
Hildebrandt S Kraus A Kulla R Wilhelmi G Hanbucken M Neddermeyer H Surface
Science 486 (1-2) 24-32 Jul 1 2001
[9] Methanol Adsorption On Si(111)-(7x7) Investigated By Core-Line Photoemission And
Mass Spectrometry Of Photodesorbed Ions Carbone M Piancastelli Mn Zanoni R
Comtet G Dujardin G Hellner L Surface Science 370 (1) L179-L184 Jan 1 1997
[10] Atomic And Macroscopic Reaction Rates Of A Surface-Catalyzed Reaction
Wintterlin J Volkening S Janssens Tvw Zambelli T Ertl G Science 278 (5345)
1931-1934 Dec 12 1997
[11] Redhead PA Hobson JP Kornelsen EV The Physical Basis Of Ultrahigh
Vacuum Chapter 12 (Chapman amp Hall London 1968)
[12] The Si(001) C(4 X 4) Surface Reconstruction A Comprehensive Experimental
Study Norenberg H Briggs Gad Surface Science 430 (1-3) 154-164 Jun 21 1999
[13] Production Of Nanostructures Of Silicon On Silicon By Atomic Self-Organization
Observed By Scanning Tunneling Microscopy Jones D Palermo V Applied Physics
Letters 80 (4) 673-675 Jan 28 2002
80 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 81
7 Conclusions and proposals for future work It is very difficult to foresee the future of nanoscience and nanotechnology even over the
next few years exciting discoveries are continuously made opening up new possibilities
and progress in this sector is now one of the fastest and most varied So it is impossible to
predict whether computers will in the future be made with nanowires nanotubes single
electron devices or some kind of quantum-based transistor What does seem clear
however is that silicon devices can still further extend their levels of miniaturization to
remain competitive for at least the next ten years [1] During this period new emerging
technologies will reach scientific maturity and arrive at the production lines
Even when new technology prototypes do become available for full-scale production the
astronomical cost of changing from silicon to new technologies will be a major concern
Moreover a vast amount of extremely detailed knowledge is already available on all
aspects of silicon technology Thus the possibility of integrating innovative
nanotechnologies with standard CMOS silicon technology already seems an attractive
prospect [2]
For this and other reasons we have focussed our research efforts on surface phenomena on
silicon wafers which could be potentially useful for the development of silicon-compatible
nano-devices
Some important conclusions can be drawn from the work described in this thesis
bull Silicon surfaces of different crystal faces were studied at atomic resolution in ultra-
high-vacuum following transformations induced on the surface by chemical (etching)
and physico-chemical (organic contamination heating) treatments
82 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull The presence of surface contaminants induces at high temperature the growth of
nanometre-sized islands and lines on silicon
bull Final island densities and dimensions strongly depend upon the chemical nature of the
contaminant molecules and the treatments used Molecules interacting only weakly
with silicon will desorb upon heating whereas molecules which decompose easily on
Si will give widespread random nano-island nucleation Between these two extreme
cases complex behaviours of diffusion and subsequent decomposition are possible
bull Nanoscopic naked silicon voids were produced in oxide layers through carefully-
controlled thermal annealing It was found that void density and size depends upon
oxide purity surface contamination levels and annealing conditions
bull The growth of nano-islands and nano-voids was exploited simultaneously and the
voids used as nanoscopic masks to control the positions of growing islands Complex
silicon-on-silicon nano-structures were obtained with multi-step growth processes
bull Finally the use of macroscopic treatments and simple physical processes to produce
large quantities of nanometre-sized structures on silicon surfaces was demonstrated
These processes and the structures produced are of potential interest to the
microelectronics industry
We showed in previous chapters how it is possible to modify silicon surfaces using
chemical and physical methods The use of self-organization for the production of Si-on-Si
20 nm
Fig 71 Left SEM image of field emitter arrays made by conventional lithography [3] Center 3-D STM image of a self-organized void with a central nano-island Right Lateral schematic view of a field emitter device
VPalermo 83
nano-structures has the advantage with respect to other emerging techniques of being
silicon compatible and thus is a potential candidate for its implementation in the
production of new devices Furthermore this technique can create billions of strucures on a
wafer in only a few minutes being thus applicable to large-scale production
Even when ordered structures are obtained these methods still show a poor reproducibility
and control Chemically modified surfaces look very disordered at high magnification
with trenches and holes etched into the surface and progress has recently been made in
producing very small structures electrochemically [4] Oxide decomposition with
nanoisland growth in UHV can also yield ordered structures The void holes with
nanoislands located at their centres although quite variable in dimensions are more similar
to engineered devices than to a spontaneous random surface-roughening process In fact
there is a surprising similarity between the nanovoid-nanoisland structures and well-known
commercially available devices called field emitter arrays (FEA) These are small tips used
as electron guns in various kinds of displays and other devices Both the commercial
devices and our self-organized structures consist of a protruding tip at the centre of
microscopic holes on an insulating layer covering a semiconductor surface (see fig71)
Of course the order and size uniformity of commercial FEAs is much better than that
obtained with our process but it is encouraging to see that a self-organized structure
obtained with simple heating has a very similar surface topography and composition of
commercial working devices obtained after many expensive stages of optical lithography
deposition and etching
Nano-island density and size can to some extent be controlled and different
contaminants can be adsorbed onto the surface to modify the growth process In the future
with a better understanding of diffusion dynamics and growth mechanisms a multi-stage
fabrication process can be envisaged where purposely-designed molecules are adsorbed
onto silicon surfaces to diffuse organize in ordered arrays and favour surface
reorganization at the nanometre scale The same molecules could be used to deliver
dopants to obtain for example highly conducting nanostructures on otherwise undoped
highly resistive silicon The selective formation of nanostructures at the SiSiO2 border
can be used to scale down mesoscopic patterns to the nanometric level A silicon oxide
layer can be easily patterned on the surface with normal optical lithography and then
nanostructures can be built following the pattern edge In this way the production of very
small nanowire based devices may be possible (fig 72)
84 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
D
Gate
S
D
S
SiO2
SiO2
2 Heating nanowire creation
and oxide removal
3 Source-drain deposition by conventional lithography
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
4 Insulating layer (oxide) and gate (metal) deposition
Fig 72 Schematic drawing of the construction of a self-organized nanowire-based transistor
Surface diffusion and decomposition of molecules can be sensitive to even smaller surface
features The nanoscopic lines formed on Si(100) are very regular and neat if their
nucleation and growth mechanism could be controlled ordered networks of nanowires
could be made on silicon
VPalermo 85
The techniques developed here are still experimental ones the careful control and
reproducibility of each step is still an issue and further studies will be needed to clearly
understand the dynamics of formation of these structures Nonetheless the results obtained
so far open up the possibilities of producing novel devices such as for example nano-
transistors (see scheme in fig73 where the mechanism described in fig 72 is applied for
the fabrication of a device array) The possibility of using the border of a lithographic
pattern to draw nanolines can in principle allow the creation of nanodevices using normal
IC manufacturing techniques
The results obtained and their description in this thesis are not the first examples of the use
of self-organization for the production of ordered structures and will certainly not be the
last In our opinion the most important outcome of these findings is that it is possible to
use surface diffusion to overcome the inherent limits of lithographic techniques in
microdevice production processes and to force properly selected molecules to react with
the surface in a spatially differentiated manner at particular sites thereby creating ordered
series of nanostructures The role of the surface dynamics of the substrate atoms
themselves is extremely important in this process
Outstanding results have been and continue to be obtained in nanoscience and
nanotechnology research promising the emergence of new production technologies Sooner
or later those emerging technologies will have to face the issues of cost and compatibility
with the enormous investment made in existing technologies and processing facilities The
commercial production of nanostructures based on surface diffusion processes could in
principle be cheap simple and compatible with existing technology Moreover the
processes of diffusion in bulk silicon is well-known in the microelectronics industry which
has both the know-how and the equipment to better understand and exploit the surface
diffusion processes on silicon surfaces studied here
Extremely important and innovative results are often obtained by exploiting very simple
ideas and in our opinion physical processes such as surface diffusion used to create
complex nanostructures on silicon surfaces is a prime example
86 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
2 Heating nanowire creation and
oxide removal
3 Source-drain deposition by conventional lithography
4 Insulating layer (oxide) and gate (metal) deposition
Fig 73 Schematic macroscopic outline of the construction of a self-organized nanowire-based transistor
VPalermo 87
Bibliography
[1] P Gargini Intel Technology Coordinator ldquoFrom Microelectronics To
Nanotechnologiesrdquo Invited Campus Colloquia Lecture CNR Research Area Bologna
February 6th 2003
[2] Toward A Hybrid Micro-Nanoelectronics Cerofolini Gf Ferla G Journal Of
Nanoparticle Research 4 (3) 185-191 Jun 2002
[3] Recent Progress In Field Emitter Array Development For High Performance
Applications Temple D Materials Science amp Engineering R-Reports 24 (5) 185-239
Jan 25 1999
[4] Silicon Dioxide Micropillars For Sieving Fabricated By Macroporous Silicon-Based
Micromachining Izuo S Ohji H French Pj Tsutsumi K Kimata M Sensors And
Materials 14 (5) 239-251 2002 Electrochemical Etching In HF Solution For Silicon
Micromachining Barillaro G Nannini A Piotto M Sensors And Actuators A-Physical
102 (1-2) 195-201 Dec 1 2002
88 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 89
Acknowledgements
This work would not have been possible without the help and guidance of Dr Derek Jones
he introduced me to the use of STM and other techniques found the funding for this PhD
and helped me with the thousands of small and big problems I had to face during my
research I am grateful to Prof Alberto Ripamonti too for being the supervisor of this
thesis Thanks are also due to Dr Giancarlo Seconi director of ISOF-CNR where most of
this work was carried out Financial support from the Italian National Research Council
(CNR) is also gratefully acknowledged
My passion for surface science began when Fabio Biscarini showed me that it was possible
to really see the atoms and for this I will always thank him
During these years I had the luck to collaborate with many good scientists among those I
would like to cite Enrichetta Susi Massimo Cocchi and Anna Mazzone from CNR
Claudio Zannoni Anna Cavallini Daniela Cavalcoli and Antonio Castaldini from the
University of Bologna Sergio Pizzini and Maurizio Acciarri from the University of Milan
One of the most stimulating periods of my PhD was the one I spent at the Steacie Institute
for Molecular Sciences in Ottawa working with Robert Wolkow Dan Wayner Greg
Lopinski and Peter Kruse I thank them for discussing together many exciting scientific
ideas and trying to put some of them into practice
Special thanks go to my colleagues Paolo Samorigrave and Stefano De Cesari for long and
useful discussions on Science Life and their interactions often with the help of a good
pint
Finally I would like to thank Vassilia Gaetano Simone Claudia Silvia Letizia
Alessandro Angela and Sebastien even though not directly involved in this work they
have been over these three years a continuous source of happiness and support
Bologna March 2003
90 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
SCIENTIFIC PUBLICATIONS bull Lateral diffusion of titanium disilicide as a way to contacting
hybrid Si-organic nanostructures Palermo V Buchanan M Bezinger A Wolkow RA APPLIED PHYSICS LETTERS 2002 v81 p 3636 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v6 issue 20) bull Surface modifications in Si after Rapid Thermal Annealing Castaldini A Cavalcoli D Cavallini A Jones D Palermo V Susi E JOURNAL OF THE ELECTROCHEMICAL SOCIETY 2002 v 149 pG633 bull Production of nanostructures of silicon on silicon by atomic self-
organisation observed by scanning tunnelling microscopy Jones D Palermo V APPLIED PHYSICS LETTERS 2002 v 80 p 673 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v5 issue 5) bull Nucleation of nanostructures from surface defects on silicon Palermo V Jones D SOLID STATE PHENOMENA 2002 v 82-84 p 687 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo MATERIALS SCIENCE AND ENGINEERING B 2002 v88 (2-3) p220 bull Morphological changes of Si[100] surface after treatment with
concentrated and dilute HF Palermo V Jones D MATERIALS SCIENCE IN SEMICONDUCTOR PROCESSING 2001 v 4 p 437 bull Electrical and structural properties of processed silicon surfaces Susi E Cavallini A Castaldini A Cavalcoli D Jones D Palermo V ldquoRECENT RESEARCH DEVELOPMENTS IN VACUUM SCIENCE amp TECHNOLOGYrdquo 2001 v 3 p189 bull Numerical Solutions of the Stochastic Equations of Crystal
Growth Mazzone A M Palermo V INTERNATIONAL JOURNAL OF MODERN PHYSICS C 2000 v 11 Part 1 p195-204 bull Advances in silicon surface characterisation using light beam
injection techniques
VPalermo 91
Acciarri M Pizzini S Simone G Jones D Palermo V MATERIALS SCIENCE AND ENGINEERING B 2000 V73 (1-3) p 235 - 239 bull Abrupt orientational changes for liquid crystals adsorbed on a
graphite surface Palermo V Biscarini F Zannoni C PHYSICAL REVIEW -SERIES E- 1998 V 57 NUMBER 3A p R2519-R2522 Oral presentations bull The use of oxide desorption and surface diffusion for the
creation of silicon on silicon nanostructures Palermo V Jones D 1ST NATIONAL WORKSHOP ON CURRENT TRENDS IN NANOTECHNOLOGIES Catania (Italy) Feb 2002 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo V E-MRS Spring Meeting Strasbourg (France) June 2001 bull STM study of surface transformations on silicon during UHV
annealing Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999 bull Advances in silicon surface characterisation using light beam
injection technique Acciarri M Pizzini S Simone G Jones D Palermo V E_MRS SPRING MEETING June 1999 Posters bull Spontaneous nano-wire growth on silicon Palermo V Jones D SMARTON workshop Leuven (Belgium) October 2002 bull Ordered circles of nano-islands on silicon from CO adsorption Palermo V Jones D TRENDS IN NANOTECHNOLOGY 2002 Santiago de Compostela (Spain) September 2002 bull Production of nanostructures of silicon on silicon by atomic self-
organisation Palermo V Jones D EUROMAT Rimini (Italy) June 2001 bull Etching holes and anisotropic corrosion on silicon [100] Palermo V Jones D Susi E Asoli B SILICON WORKSHOP Genova (Italy) February 2001
92 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull Morphological and electrical characteristics of damaged silicon surfaces
Susi E Castaldini A Cavalcoli D Cavallini A Jones D Palermo V SILICON WORKSHOP Genova (Italy) February 2001 bull Effect of HF etching on the roughness of a silicon surface Palermo V Jones D NATIONAL CONFERENCE ON PHYSICS OF MATTER Genova (Italy) June 2000 bull Effect of Native Oxide Desorption upon the Surface Morphology of Si[100] by STM and LEED Palermo V Jones D SILICON WORKSHOP Genova (Italy) February 2000 bull Characterisation of silicon surfaces for microelectronics through STM measurements Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999
VPalermo 93
Contact Vincenzo Palermo ISOF- Institute for Organic Synthesis and Photoreactivity Via Gobetti 101 40129 Bologna ITALY Tel +39-051-6398336 Fax +39-051-6398349 Mail palermoisofcnrit
94 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Index
1 Introduction nanotechnology and the future of computers
2 Silicon surfaces
3 STM and other surface analysis techniques
4 Surface modification of silicon in liquid Nano-hole creation
5 Surface modification of silicon in vacuum void creation and oxide desorption
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and
nano-lines
7 Conclusions and proposals for future work
4 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 5
OMNIA IN MENSURA ET NUMERO ET PONDERE
Sapientiae Salomonis 1120
6 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 7
1 Introduction nanotechnology and the future of computers The motivations of nanotechnology research
In 1965 an electronic engineer named Gordon Moore one of the future founders of Intel
noted that the performance of computers and their complexity doubled every 18 months
and foresaw that computer power would continue to grow exponentially over the following
years
This prevision quite provocative for its time actually came about and gained the name of
ldquoMoorersquos Lawrdquo and continues to hold for the trends of todayrsquos computer industry Since
1965 the number of transistors present in an integrated circuit (IC) has increased from
several hundred to more than ten million and the minimum size of transistor elements has
shrunk from several millimetres to asymp130 nanometres (fig 11)
Devices of such tiny dimensions are actually fabricated using lithographic techniques
where light is passed through an optical mask to react with a photo-sensitive layer (resist)
Fig11 Evolution of the number of transistors present on commercial computers [1]
8 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
on the silicon wafer This resist is then selectively removed and used as a mask for
processing the silicon surface (fig 12 left) The maximum resolution attainable depends
upon the wavelength used and current technology is near to its intrinsic resolution limit
On the other hand there is strong scientific and economic demand for further development
in IC miniaturization to obtain more powerful and complex computers Besides every-day
life applications more powerful computers are fundamental for much scientific research
such as climate change tracking genome sequencing and fluid dynamics Increased
miniaturization is also fundamental for reducing power which has to be dissipated by the
chips which run at progressively higher frequencies Energy consumption by
microelectronic devices is already an issue and represents one of the main obstacles for
the continuing growth in wireless communication (cell phones portable computers CD
and DVD players digital cameras etc)
Thus it is expected that new production methods different from current lithographic ones
will be developed methods which allow modification of a surface well below the 100 nm
limit and even down to single atom manipulation Techniques such as Scanning
Tunnelling Microscopy and Atomic Force Microscopy are already capable of moving
single atoms (see fig 12 right) but unfortunately building a working nanodevice in this
way would take a very long time and these techniques are difficult to apply to large scale
production
Nowadays thousands of researchers are working in the nanotechnology field towards a
new generation of microelectronic devices Several possible solutions are competing for
tomorrowrsquos computer architecture and there is still no clear winner It is likely that the
final solution will be the combined use of different techniques and components (including
molecules nanowires and nanodots) as they become available together with conventional
Fig 12 Left conventional litographic process [2] Right atomic manipulation of iron atoms on copper [3]
VPalermo 9
silicon technology
Below is a brief summary of the most recent developments in nanotechnology and
nanoscience
Actual trends in nanotechnology
Perhaps the most fascinating idea for nanodevice construction is to use one single
molecule working as a complete device The first molecular diodes (ie molecules
conducting current only in one direction) were created in 1997 in 1999 a molecular fuse
and a molecular transistor were demonstrated although there was no possibility of wiring
these devices to external contacts In April 2001 James Heath and his group at UCLA
fabricated an array of overlapping crossbars and placed a small molecule of rotaxane
between each crossbar (fig 13 left) This composite molecule is made up of two
component parts the main rod-like molecular axis and a mobile ring ldquothreadedrdquo on it like a
bead on a necklace and can function as a molecular switch A working 16-bit memory
circuit was constructed using these molecules For a brief review of these works see [4] In
June 2002 a single molecule transistor was built by connecting an organic molecule to two
metal contacts the molecule contained one or two atoms of a transition metal (cobalt or
vanadium) forming the active region of the device supported by an organic backbone [5]
Fig13 Working nanodevices Left schematic representation of rotaxane molecules between crossed nanowires [4] Right SEM image of semiconductor nanowires forming a small circuit [9]
10 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another approach to nanodevice fabrication has become possible through the discovery of
carbon nanotubes which were observed for the first time in 1991 by a Japanese electron
microscopist studying the material deposited during arc-evaporation synthesis of fullerenes
[6] They consist of a graphite-like carbon seamless cylinder with a diameter of several
nanometers and lengths of up to a millimetre Carbon nanotubes are very stable can
behave as metals or semiconductors and can host other molecular or ionic species thus
modifying their electrical behaviour In 2001 Avouris and his group reported the first
circuit made with a single nanotube [7] A few months later Cees Dekker presented a
nanotube-based transistor able to amplify an input signal by a factor of ten and built
several logic circuits using these nanotube transistors [8]
One problem with carbon nanotubes is that it is very difficult to control their electronic
properties ie their metallic or semiconducting behaviour An alternative to carbon
nanotubes are semiconductor nanowires Silicon nanowires can be made using a laser to
vaporize the silicon together with a metal catalyst like iron or gold The vapour condenses
in nanosized drops of silicon and metal from which the wires slowly grow out as more
silicon is adsorbed In 2001 a group at Harvard University [9] created a transistor by
crossing two different nanowires After this the same group arranged four nanowires in a
noughts and crosses grid creating something like a 4-bit memory (fig 13 right) Even
metallic nanowires made of platinum and silver can be used in a crossed configuration to
store information [10]
There are some issues common to all these new technologies though First it is difficult to
imagine these methods applied to large-scale production Up to now the insertion of a
molecule between two electrodes is an occasional lucky event while nanotubes and wires
have to be positioned on the surface creating the appropriate contacts on them manually
The large-scale production of integrated circuits using these building blocks will not be
straightforward and does not seem imminent
Another issue is of an economic and not a scientific or technological nature Since 1965
the cost of IC manufacturing plants has sky-rocketed If the cost of semiconductor
production plants continues to rise exponentially in a few years such plants will cost up to
$20 billion This is a sizeable investment even for large companies such as IBM or Intel
For this reason it is likely that IC companies will resist changing to completely new
technologies closing down their existing plants As it is clear that silicon will remain the
VPalermo 11
fundamental raw material of the IC industry for the foreseeable future nanotechnology
developments for microelectronics will need to be silicon-compatible In these early days
of nanotechnology the most valid approach would seem to be the addition of molecular
functions to existing silicon technology ndash using the latter as a foundation on which to build
Fabrication of self-organised structures on silicon
The possibility of using the phenomenon of atomic or molecular self-organization to create
nanostructures on silicon has already been demonstrated The clean silicon surface shows
in some cases a high degree of order and complex surface reconstruction as will be
described later Several different ordered structures form spontaneously on this surface
such as series of monatomic steps or boundaries between reconstructed areas It has been
demonstrated that it is possible to use these structures to fabricate ordered nanodots and
nanolines on the surface [11] More recently well-defined nanometric patterns have been
obtained with selective etching of silicon using nitric oxide [12]
In this study the possibility of creating different types of nanostructures on the silicon
surface is explored Methods had to be developed which were
- Simple They must not need complex masks or lithographic steps to create the
structure but rather exploit self-organisation phenomena
- High resolution the silicon surface modifications should be on a scalelength of below
100 nm
- Fast billions of nanostructures have to form over the whole surface simultaneously to
be compatible with large-scale production requirements
- Cheap they must not require expensive equipment (such as e-beam lithography) but
exploit simple chemical andor physical treatments to produce nanostructures on the
silicon surface
In Chapter 2 the main characteristics of silicon are described Chapter 3 provides a
summary of the techniques used for this research Chapter 4 examines the chemical etching
of silicon in different liquid environments and the effects of this etching on the surface at a
nanoscopic level with the creation of nanoholes
12 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Chapter 5 describes the growth in ultra-high vacuum (UHV) of nanoscopic voids and
islands on the silicon surface and the effect of surface oxide on this growth
Chapter 6 discusses the modification of silicon surfaces in UHV following the adsorption
of molecules and thermal heating to produce nanoislands and nanolines on silicon
The overall conclusions of our work are summarised in Chapter 7
Finally we will give some conclusions based on the results obtained and discuss possible
applications of the methods developed
Bibliography
[1] From wwwintelcom
[2] From wwwsematechorg
[3] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
[4] Molecules Get Wired Service RF Science 294 (5551) 2442-2443 Dec 21 2001
[5] Coulomb Blockade And The Kondo Effect In Single-Atom Transistors Park J
Pasupathy AN Goldsmith JI Chang C Yaish Y Petta JR Rinkoski M Sethna JP
Abruna HD Mceuen PL Ralph DC Nature 417 (6890) 722-725 Jun 13 2002 Kondo
Resonance In A Single-Molecule Transistor Liang WJ Shores MP Bockrath M Long
JR Park H Nature 417 (6890) 725-729 Jun 13 2002 Nanotechnology - Electronics
And The Single Atom De Franceschi S Kouwenhoven L Nature 417 (6890) 701-702
Jun 13 2002
[6] Smallest Carbon Nanotube Ajayan PM Ijima S Nature 358 (6381) 23-23 Jul 2 1992
[7] Carbon Nanotube Inter- And Intramolecular Logic Gates Derycke V Martel R
Appenzeller J Nano Letters 1 (9) 453-456 Sep 2001
[8] Logic Circuits With Carbon Nanotube Transistors Bachtold A Hadley P Nakanishi T
Dekker C Science 294 (5545) 1317-1320 Nov 2001
VPalermo 13
[9] Logic Gates And Computation From Assembled Nanowire Building Blocks Huang Y
Duan XF Cui Y Lauhon LJ Kim Kh Lieber CM Science 294 (5545) 1313-1317
Nov 9 2001
[10] Formation And Disappearance Of A Nanoscale Silver Cluster Realized By Solid
Electrochemical Reaction Terabe K Nakayama T Hasegawa T Aono M Journal Of
Applied Physics 91 (12) 10110-10114 Jun 15 2002
[11] Fabrication And Integration Of Nanostructures On Si Surfaces Ogino T Hibino H
Homma Y Kobayashi Y Prabhakaran K Sumitomo K Omi H Accounts Of Chemical
Research 32 (5) 447-454 May 1999
[12] Ultrafine And Well-Defined Patterns On Silicon Through Reaction Selectivity
Prabhakaran K Hibino H Ogino T Advanced Materials 14 (19) 1418-1421 Oct 2
2002
14 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 15
2 Silicon surfaces
The name silicon (silicio in Italian) comes from the latin word silex Amorphous silicon
was first isolated by Berzelius in 1824 by reaction of potassium with silicon tetrafluoride
Thirty years later the first crystalline silicon was prepared Silicon makes up 25 of
earthrsquos crust and is the second most abundant element after oxygen Elemental silicon is
not found in nature occurring as silicon oxide (sand quartz amethyst flint etc) or
silicates (asbestos clay mica etc) Perhaps no other element and its compounds has such
a wide range of uses Silicon compounds such as sand and clay are used in the building
industry as refractory materials for high-temperature applications and for enamels and
pottery Silica is the main component of glass silicon carbide is an important abrasive and
silicones are commonly used polymers and lubricants
Here the most interesting use of silicon of course is for the production of
microelectronics devices For this application silicon of high purity (999999) and of
high crystallinity is needed Table 21 lists some of the physical characteristics of silicon
High purity polycrystalline silicon is produced by the reaction of gaseous trichlorosilane
with hydrogen in a furnace Then to prepare a single-crystal of silicon the so-called
Czochralski method is commonly used
Polycrystalline silicon is melted in a quartz furnace at 1415degC in an argon atmosphere
Then a seed of single-crystal silicon is lowered into contact with the melt and slowly
pulled out In this way the crystal grows and a crystalline cylindrical ingot several metres
long is created from the initial seed
After cooling down the ingot is sliced into thin silicon wafers The wafer surfaces are
polished using a counter-rotating lapping machine in an Al2O3 slurry until the surface is
very flat and shiny ready for the lithographic processes
16 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another way to obtain single crystal silicon is the Floating Zone (FZ) method in which a
silicon cylinder is slowly passed through a heating ring The area inside the ring melts and
solidifies smoothly crystallising as it comes out of the ring yielding a single silicon crystal
Microelectronic devices are built on the silicon surface which is the surface of interest
here Unfortunately silicon surfaces are normally quite dirty and uneven at the atomic
scale Atmospheric oxygen and humidity react with silicon surfaces creating a thin layer of
oxide (called ldquonative oxiderdquo) which is usually irregular and full of defects Different kinds
of contaminants also adsorb onto the surface These are usually small organic molecules
and microscopic dust particles A clean surface on exposure to the atmosphere is
completely covered with gas molecules in less than 10-9 seconds If the pressure is
reduced letrsquos say to 10-6 mbar this time increases to 1 second This is the reason why to
study a clean surface we have to work in UHV at pressures below 10-10 mbar
The atoms in the silicon crystal have a diamond-like structure each atom having 4 bonds
in a tetrahedral sp3 arrangement with bond angles of 10947 degrees At the crystal
surface some atoms will have non-bonding orbitals ldquodanglingrdquo in the vacuum ie sp3
orbitals with a lone electron which are highly reactive These orbitals are known as
dangling bonds To minimize surface energy the surface will reorganize by decreasing
the number of dangling bonds
Table 21 Physical data of silicon [1]
Atomic Weight 2809 Lattice constant (A) 543095 Crystal structure Face-centered cubic
(diamond) Melting point 1415 degC
Density (gcm3) 2328 Boiling point 2355degC Atomscm3 50E22 Minority carrier
lifetime (s) 25E-3
Dielectric Constant 119 Specific heat (Jg degC)
07
Breakdown field (Vcm)
~3E5 Thermal conductivity (Wcm degC)
15
Electron affinity x(V)
405 Vapour pressure (Pa) 1 at 1650degC 1E-6 at 900deg C
Energy gap (eV) at 300K
112 Reactivity Inert to acids Attacked by halogens and alkaline
solutions Intrinsic carrier
conc (cm-3) 145E10 Oxidation states +4 -4
Intrinsic Debye Length (microm)
24 Energy of a Si-Si bond (eV)
232
Intrinsic resistivity (Ω-cm)
23E5
VPalermo 17
Dangling bond densities and positions and thus the type of surface reconstruction will
depend upon crystal orientation as well as the temperature and kinetics of the system
Fig 21 shows a drawing of the main faces of a silicon crystal The angle α between any
(11n) face and the (100) face can be calculated from 2cos 2 += nnα The angle
between any (11n) face and the (111) face can be calculated from
)2(3)2(cos 2 ++= nnα
The chemistry and physics of the faces are very different a brief description will be given
for the most important orientations
Si (100)
On the (100) surface each atom has two Si-Si bonds connecting it to the bulk and two
dangling bonds pointing outward Surface energy is reduced by the dimerisation of the
surface atoms through overlap interaction of one dangling bond per atom forming rows of
dimers aligned along the (110) direction This is the well-known ldquo2x1rdquo reconstruction of
this silicon surface The symmetric dimers would make the silicon surface metallic but to
reduce surface stress the dimers tend to buckle and the surface is thus semiconductive It
took several years to understand that the dimers are buckled because at room temperature
Fig 21 Schematic view of the principal orientations of a silicon surface Surface atoms are white bulk atoms are black dangling bonds are gray [2]
18 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
they shift easily from one buckling direction to the other thus appearing symmetric under
STM observation Fig 22 shows an STM image of the 2x1 reconstructed surface
Even almost perfect (100) surfaces have a certain number of monoatomic steps and the
dimer rows on atomic layers are aligned at 90deg to those on adjacent layers Dimer rows are
thus perpendicular or parallel to the step When the dimers on the upper side of the step
are parallel to the step the step is called SA if they are perpendicular the step is called SB
Because of this symmetry SA step edges will be smoother compared to the more broken
fragmented edges of the SB steps
A common defect on the Si(100) surface is the presence of nickel contamination which
appear as missing dimers in STM images This type of contamination is so critical that
even if the silicon sample is only briefly brought into contact with stainless steel tools
(tweezers for example) the 2x1 reconstruction of the surface can be blocked
Silicon atoms can diffuse easily over the silicon surface as monomers and dimers
especially at elevated temperatures The anisotropy due to the 2x1 reconstruction causes a
difference in the diffusion energies of adsorbates over the surface Diffusion of these
silicon species along dimer rows for example will be much easier A list of diffusion
energies for monomers and dimers is given below [2]
Diffusion on Si(100) 2x1 Ed (eV) Monomers along dimer rows 06 Monomers across dimer rows 085 Monomer formation energy 18 Dimer along dimer rows 11 Dimer across dimer rows 15 Dimer formation energy 26 Dimer binding energy 076 Vacancies along dimer rows 17 Vacancies across dimer rows 19
SA
SB
Fig 22 STM image of a 2x1reconstructed silicon surfaceshowing the dimer rows and steps40x35 nm Nickel-induced defectsare visible as dark spots SA and SBsteps are indicated
VPalermo 19
So the diffusion energy for both monomers and dimers is nearly 40 greater if they have
to cross a dimer row This difference reduces to sim10 for vacancy diffusion
Si(113)
The (113) surface can be imaged as a sequence of alternating (100) and (111)-like
structures with two and one dangling bonds on alternate atoms respectively Interest in the
(113) surface is scientific as it has been used to study the energetics of the (100) and (111)
surfaces as well as for surface adsorption experiments
Si(100) surfaces can easily develop (113) facets
Si(111)
This surface besides being the first one imaged with STM with atomic resolution is one of
the most studied because it is the best cleavage face of silicon and because it shows one
of the most complex and elegant reconstructions in surface science
All Si-Si bonds in the silicon crystal are perpendicular to a (111) plane so this face will
have the lowest number of dangling bonds created per unit area In fact each Si atom on a
(111) surface shows a single dangling bond oriented perpendicular to the surface and
bonded to three back atoms These three bonds for each surface atom account for the great
chemical and physical stability of the Si(111) surface Surface energy is 009 eV Aring-2
compared to 015 eV Aring-2 for Si(100)
For energy minimization this surface reconstructs forming a huge 7x7 lattice cell
containing 102 atoms described by the Dimer-Adatom-Stacking fault model (DAS) For a
detailed description of cell structure see fig 23
The cell described by this model is very complex being composed of three kinds of atoms
adatoms rest atoms and corner hole atoms Furthermore a subsurface stacking fault is
present in one half of the cell making the two halves of the unit cell look different under
STM (Fig 24) It took 26 years of research to completely understand the exact structure of
the 7x7 reconstruction
Cleaving a silicon crystal along a (111) plane produces a metastable 2x1 reconstruction
the 7x7 reconstruction is easily obtained by flashing at high temperature in UHV At T gt
20 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
830degC a disordered 1x1 phase covers the surface Cooling down to 800degC leads to the
formation of the 7x7 phase If the cooling process is too rapid small 7x7 domains nucleate
and a disordered 1x1 phase is preserved between domain boundaries
Si(110)
Even though as mentioned before the (111) plane is the favoured cleaving plane of
silicon thin (100) commercial wafers will not break along this plane because the angle
between (100) and (111) is too far from 90deg (see table 22) Instead they will break along
the (110) plane because it is perpendicular to the (100) surface Each surface atom on
Si(110) has a Si-Si bond pointing downward one dangling bond pointing outward and two
Si-Si bonds parallel to the surface in a zig-zag pattern (see fig 21) Cleaved (110)
surfaces are disordered but upon annealing at high temperatures an ordered complex 16x2
Fig23 Scheme of the 7x7 DAS model [2] In each unit cell there are 9 dimers 12 adatoms and a stacking layer fault The force driving this complex reconstruction is the minimization of dangling bonddensity The DAS model shows the lowest number of dangling bonds (19) of all possiblereconstructions 12 dangling bonds are at the adatoms 6 at the rest atoms and 1 at the corner hole atom This surface is metallic
VPalermo 21
reconstruction takes place The surface appears as a series of long ridges and valleys
parallel to each other Eventually tilted facets of orientation (17 15 1) can form on this
surface The adsorption of Ge atoms on this surface leads to the formation of self-
assembled nanowires [3]
Table 22 Angles in degrees between different silicon faces [2]
Orientation
113
110
111
100
100
2524
9000
5474
0
111
2950
3526
0
110
6476
0
113
0
Fig 24 STM image of a Si(111)surface with 7x7 reconstruction A unitcell with its adatoms is highlightedImage size 13x13 nm
22 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] Weast RC Handbook Of Chemistry And Physics (Chemical Rubber Co Cleveland
1972)
[2] Dabrowski J Mussig H Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] The Structure Of Clean And SiGe-Covered Si(110) Surfaces Butz R Luth H Surface
Science 365 (3) 807-816 Oct 1 1996
VPalermo 23
3 STM and other surface analysis
techniques
Scanning Tunneling Microscopy
Since the invention of the optical microscope at the end of the 16th century the possibility
of examining surfaces at higher and higher magnification has fascinated mankind
Development of the technique continued and towards the end of the 19th century optical
microscopes were as good as todays standard instruments The physical limits of the
wavelength of visible light (350-800nm) had been reached
In the 1920s de Broglie showed that electrons can behave like waves and the use of these
particles for imaging with much higher resolution soon followed Atomic resolution using
this technique is only possible in the transmission mode with extremely carefully prepared
samples
In 1982 using the peculiar properties of piezoelectric materials Binnig and Rohrer brought
a metallic tip very very close to a silicon surface and scanned it across an extremly small
area (fig31) The tunneling of electrons from the tip into the sample or vice versa allowed
them to obtain a local density of electronic states (LDOS) map of that surface Although
theory (which treated the extreme point of the tip as a sphere) then excluded the possibility
of atomic resolution following a tip crash into the surface Binnig and Rohrer began to
observe the LDOS with atomic resolution For this discovery and their following work
they were awarded the Nobel Prize in Physics in 1986
24 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The basic principle of STM is very simple A metallic tip is scanned over a surface without
making ohmic contact and a tunneling current passes between the tip and the surface An
electronic circuit keeps this current constant by raising and lowering the tip during the
scan In this way recording the tip height at each point a three-dimensional image of local
density of electronic states (LDOS) of the surface can be obtained To explain the
extremely high resolution attainable by this simple technique quantum theory is needed
According to classical physics the current will flow between sample and tip only if they
are in physical contact If there is a vacuum gap between the two the electrons will simply
remain confined for example within the surface without the possibility of passing into the
tip
In quantum physics however the electrons have a certain probability of passing
(tunneling) across the gap appearing on the other side of the gap in this way reaching the
tip It can be shown that the probability of an electron tunneling through a gap of thickness
z is
kzep 22)0( minusprop ψ h
φmk
2= (1)
where ψ(0) is the electron wavefunction at the surface-gap border m is the electron
mass=91x10-28 g and φ is the work function of the metal (ie the energy required to
remove an electron from that material For silicon it is 48 eV) The tunneling current thus
decays exponentially with z and is extremely sensitive to topographical imperfections
present on the scanned surface A rough formula giving the current as a function of z is [1]
zFS eEVI φρ 0251)( minusprop
Fig 31 Binnig and Rohrer with the first STM Image from IBM [2]
VPalermo 25
where ρs(EF) is the local density of states at the Fermi level on the given surface For
example the formula predicts that for silicon an increase in tip-surface distance of 1 Aring
will give a 95 decrease in tunneling current
This huge dependence of tunneling current upon the distance allows detection even of the
sub-nanometre changes in height given by the single atoms of which the surface is
composed and thus to resolve them in the LDOS images Of course this description of the
tunneling process is oversimplified and for a more accurate one the electronic states of
the tip of the sample and their interaction have to be taken into account Fig 32 shows a
schematic representation of the interaction between tip and sample orbitals
The exponential decay of current with distance also yields high lateral resolution If the tip
is approximated as a sphere of radius R and the current passing at the minimum tip-sample
distance is I0 then the current passing at a lateral distance x from this point will be
Rxk
eII 22
0
2minus
=
Assuming a tip radius of 100 nm the current is concentrated in an area sim15 nm wide at tip
apex
Very sharp tips with even smaller curvature radii can be produced in several ways Simple
Fig32 Schematic view of tip-sampleorbitals interaction a) no interaction b)equilibrium c) sample positive d) tippositive [1]
26 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
tungsten wires mechanically cut are capable of obtaining atomic resolution on graphite in
air but for more disordered and rougher samples sharper and more reproducible tips are
needed
STM tips are mostly made by electrochemically etching a W or Pt-Ir wire The tips we
used were prepared using methods based on the work of Fotino [3]
A tungsten wire 038 nm diameter is immersed in a KOH solution(06M) with a thicker
tungsten wire used as a counterelectrode The cathodic and anodic reactions involved in the
etching are
Cathode 6H2O + 6 e- rarr 3H2(g) + 6 OH-
Anode W(s) + 8 OH- rarr WO42- + 4H2O + 6 e-
A potential of 3V ac is applied to the tungsten and the wire is immersed in the solution
until a constant current of sim100 mA is established The etching rapidly removes metal
shaping the wire end as a sharp tip When the potential reaches 11 V the coarse tip etching
is finished The wire is then removed from the solution carefully inserted into an
insulating plastic tube leaving only the tip exposed and re-immersed in the solution with
the tip pointing upwards A more gentle etching is thus made to reduce tip radius Usually
5 minutes etching at 07 V ac is used In this configuration very small hydrogen bubbles
formed on the tip sides sliding upwards with a ldquohoningrdquo effect on the tip
This procedure yields extremely sharp and reproducible tips at the microscopic level
After the etching the tip is thoroughly rinsed in ultrapure water then dipped into
concentrated HF to remove surface oxides and hydroxides [4] The tip is dried with
nitrogen inserted into the UHV system and degassed overnight at sim150degC
The possibility of measuring sub-nanometric distances would be useless without being able
to control tip movement over such a minute scale To scan the tip over the surface a
piezoelectric scanner is used Piezos are usually made of an alloy of PbZrO3 and PbTiO3 a
material which contracts or expands when a voltage is applied to it The Omicron
instrument used in our laboratory has three such piezo scanners for xy and z tip motion
allowing one to scan the tip over the surface with sub-Aringngstrom precision (fig 33)
To isolate the instrument from ambient vibrations the whole STM stage is suspended upon
four springs which eliminate all frequencies above 1 Hz and surrounded by a crown of
VPalermo 27
copper wings and fixed magnets Parasitic currents generated by the magnets into the
copper wings contrast every movement of the stage and efficiently block stage vibrations
The STM can be used not only to explore surface topography but to measure the IV
characteristics of single atoms or molecules on the surface (Scanning Tunneling
Spectroscopy or STS) Furthermore it can be used to modify the surface with voltage
bursts digging into it or delicately moving single molecules or atoms over a surface [5] It
can be used in vacuum in air and with proper lateral isolation of the tip even in liquids
A major drawback of STM is that it works only on conducting and semiconducting
surfaces and thus cannot be used on many surfaces of biological and chemical interest
Another instrument more suitable for these and other applications is the Atomic Force
Microscope (AFM)
Atomic Force Microscopy
AFM was invented in 1986 by Binnig Quate and Gerber after calculating the possibility of
building a cantilever with a force constant of the same order of magnitude as that of a
chemical bond
In AFM a tip mounted on a microscopic cantilever (usually made of Si3N4 fabricated with
optical lithography) is brought close to a surface When the tip touches the surface the
cantilever is very slightly deflected upwards The movement is measured by observing the
Y-PIEZO
SILICON SAMPLE
Fig 33 A picture of theSTM used for theexperiments The tripodpiezo scanner is shown
Z-PIEZO
X-PIEZO
TIP
28 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
deviation of a laser beam hitting the upper face of the cantilever Fig 34 provides a
schematic view of the principle of AFM
The typical force constant of the cantilever varies from 00006 to 2 Nm the typical
resonance frequency is 3 to 120 kHz The AFM tip can apply a force on the sample of up
to 10-9 N The AFM can be used on conductive or insulating surfaces in vacuum air or
liquids Furthermore the tip can be modified to sense electrostatic potentials (electric force
microscopy) or magnetic fields (magnetic force microscopy) it can even be functionalized
with complex molecules such as proteins to interact with biological surfaces
A drawback of the AFM is that the force it exerts can damage the surface under
observation especially if the sample is soft (as in the case of cell membranes for
example) This problem can be overcome using the instrument in tapping mode (where the
tip does not move laterally during its brief contact with the surface) or in non-contact mode
in which the tip oscillates above the surface during the scan and the changes in its
frequency due to interaction with surface are monitored The shifts in the oscillating
frequency of the cantilever due to tip-sample interaction are then used for imaging the
surface In this mode interaction of the tip with the surface is minimal and soft samples
can be imaged
STM and AFM are the main techniques used for this work A brief description of other
techniques used occasionally is given below
Fig 34 Scheme of an atomic force microscope
VPalermo 29
Low Energy Electron Diffraction (LEED)
Electrons with energies in the 20-500 eV range are diffracted by a crystalline surface the
diffraction peaks are visualized on a fluorescent screen This technique probes the long
range order of the surface up to a depth of several nanometres
X-Ray Photoelectron Spectroscopy (XPS)
XPS allows both qualitative and quantitative chemical analysis of the elements present on
or near the sample surface
An X-ray source is used to photoionize the atoms on a surface and produce photoelectrons
By measuring the kinetic energy of the photoelectrons the binding energy of the electronic
levels can be calculated This energy will depend on the chemical environment of the
surface atoms
Although the soft X-rays used penetrate to a depth of ~2000Aring the sampling depth of the
technique is determined by the mean free path of the photoelectrons which allows their
escape from only the first 10-100Aring
Secondary Ion Mass Spectroscopy (SIMS)
High and low energy ions (primary ions) are used to bombard a sample and remove surface
atoms and ions The ionic fragments removed (secondary ions) are then analysed by a mass
spectrometer The surface can be consumed during the measurement and profiles obtained
giving concentrations of the materials composing the sample at different depths (depth
profiles)
A popular variant of SIMS is TOF-SIMS In this technique the secondary ion masses are
measured by a time-of-flight (TOF) measurement The secondary ions generated by the
bombarding primary ions are accelerated to a constant kinetic energy and then move
through a field-free space before they reach the detector where their intensity is measured
as a function of flight time Since ions with different masses have different velocities at a
30 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
given kinetic energy the measured flight times of the ions can easily be converted to their
masses The static nature of this latter technique allows mass spectroscopy surface analysis
with minimal damage to the surface
Bibliography
[1] Chen CJ Introduction To Scanning Tunneling Microscopy (Oxford University Press
Oxford 1993)
[2] From wwwibmcom
[3] Tip Sharpening By Normal And Reverse Electrochemical Etching Fotino M Review
Of Scientific Instruments 64 (1) 159-167 Jan 1993
[4] A Convenient Method For Removing Surface Oxides From Tungsten STM Tips
Hockett LA Creager SE Review Of Scientific Instruments 64 (1) 263-264 Jan 1993
[5] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
VPalermo 31
4 Surface modification of silicon in liquid
Nano-hole creation
Liquid treatments of silicon wafers are very common in the integrated circuit (IC)
manufacturing industry They are used to clean and improve surface uniformity to create
and etch protective oxide layers and to remove photo-resist layers
Crystalline silicon with its native oxide layer is very stable and is resistant to many acids
It is easily attacked by hydrofluoric acid (HF) and alkaline solutions
The thin (~2 nm) passivating layer of native oxide (SiO2) is formed on exposure to the
atmosphere This surface layer contains many defects and contaminants so it is usually
chemically stripped and substituted with a better chemically-formed protective oxide
The most common silicon cleaning procedure is the RCA method named after the Radio
Corporation of America [1] It consists of two steps in the first one the surface is treated
with a hot alkaline solution (H2OH2O2NH4OH 411) to remove particles from the
surface following this a hot acidic solution (H2OH2O2HCl 411) is used to remove
metal contamination Other well-known cleaning methods are IMEC (a sequence of
cleaning steps in H2OO3 and dilute HF) or the Pirana etch (a hot 41 mixture of H2SO4
H2O2)
32 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The standard RCA clean removes surface contaminants etches the native oxide and
oxidizes the silicon surface leaving a uniform layer of silicon oxide which better protects
the surface from further contamination
Etching with fluorine-based solutions
Hydrofluoric acid is one of the most common reagents used in the treatment of silicon
wafers both in the research field and in industrial processes A rapid dip in dilute HF is the
simplest way to remove the native oxide from Si(100) and leaves the surface passivated by
a layer of Si-H bonds Because of the low polarization of Si-H bonds the Si-H layer is
stable even for several days protecting the surface from contamination It has often been
assumed that this short etch does not significantly change the surface morphology of the
silicon substrate[2] even though a prolonged dip in dilute HF leads to surface roughening
[3]
Although dilute HF roughens the Si(100) surface at the atomic scale [45] immersion in
concentrated HF (49) etches the surface oxide without attacking the Si surface
uncovering in this way the buried SiSiO2 interface The final counter-intuitive result is
that dilute HF etches the silicon while concentrated HF leaves the crystalline silicon
untouched [3]
Etching Si with fluorine-containing solutions at different concentrations and pH can
produce different morphologies from rough surfaces to flat nearly ideal Si-H terminated
surfaces
Hessel et al and Higashi et al demonstrated in 1991 that very flat Si(111) surfaces can be
obtained using 40 NH4F while etching with HF always yields rough surfaces The
surface becomes smoother because the etchant rapidly attacks Si atoms at step borders
thus removing surface kinks and irregularities in a step-flow mechanism [6 7] Later on
even smoother and more perfect surfaces were obtained by removing oxygen from the
solution after it was discovered that oxygen dissolved in 40 NH4F initiates the formation
of triangular etch pits It was not possible to obtain flat surfaces by etching Si(100) with
ammonium fluoride solutions which leads to the formation of small 2x1 dimer-row
reconstructed (100) terraces together with (111) facets [8]
This difference is caused by the different hydride terminations prevailing on the (100) and
(111) faces While the ideal Si(111)-H surface is monohydride terminated the more
VPalermo 33
reactive dihydrides predominate on the Si(100)-H surface making it more vulnerable to
etching The etching reaction is thus strongly anisotropic etching (100) facets faster than
(111) thus producing (111) microfaceting on Si(100) crystals
A more uniform Si(100) surface can however be prepared by etching at low pH with an
HFHCl mixture [9] or by using very dilute HF solutions and ultrapure water with low
dissolved oxygen and carbon contents [10]
Electrochemical etching can also be used applying anodic or cathodic bias to the silicon
to obtain different morphologies [11] by varying the potential isotropic or anisotropic
etching is observed The aforementioned results show that despite the simplicity of the
reactants fluoride etching of silicon is quite a complex reaction
Fig 41 Chemical etching of silicon
HO+H
H HO
H
H
H
FSi
Si
SiSi
+H2O
-OH-
+F-
H2O H H
Si H
F OH
H Si
Si Si
-H2
+H2O
H
H
H
FSi
Si
SiSi
H
H
H
OHSi
Si
SiSi
H
H
H
H Si
Si
Si Si
Etching mechanism of silicon
HF rapidly dissolves the SiO2 passivating layer on silicon leaving the surface almost
completely hydrogenated [12] After this two different types of reactions etch the silicon
simultaneously one chemical and the other electrochemical [13] The overall etching
mechanism can be schematized in two stages (see Figure 41)
i) Si-H bonds are substituted by Si-F or Si-OH bonds creating a partial charge on the
surface silicon atom and polarizing its Si-Si backbonds
ii) these polarised backbonds are then more easily attacked by HF or H2O After
rupture of the Si-Si bond the atom is removed leaving behind new Si-H
terminations and the reaction can start again
34 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
These reactions take place although at different rates on both Si(100) and Si(111)
Stage i) is usually the rate-determining step of the reaction and the stability of the Si-H
bonds depends upon the pH the concentration of nucleophilic species in solution and an
eventual potential applied to the crystal
For pH gt5 as in the case of concentrated NH4F solutions reaction begins with attack by
water to give Si-OH (step ArarrC) The -OH group is rapidly substituted by fluorine with
polarization of the underlying Si-Si bonds These bonds are then easily attacked by water
the silicon atom being released into solution as HSiF(OH)2 The Si-OH rarr Si-F substitution
is not fundamental for the reaction and etching can proceed even for Si-OH terminated
atoms but XPS measurements showed the presence of a certain number of Si-F bonds
remaining Furthermore fluorine seems to have a catalytic effect on Si-H substitution as
indicated by the dependence of the etch rate upon the F- concentration at least for pH
values between 4 and 8
Si-F bonds can be easily removed by a water rinse In the case of strongly alkaline
solutions (pH=14) OH- groups act directly as nucleophiles and no fluorine is needed to
catalyze Si-H bond rupture
At pH lt4 almost no free F- ions are present in solution and the etch rate is very small at
pH lt2 all etching reactions are very slow and this explains the stability of Si crystals in
concentrated (50 ww) HF solutions
Si Si
Si Si
H
H
H
H Si
Si
Si Si
H
H
H Si
Si
SiSi
H
H
H
OH-H+ -e
(F-)
-H+
-e
+H2O hellipas in fig1
Fig 42 Electrochemical etching of silicon by H2O
The electrochemical etching of silicon involves electron transfer from the surface atoms to
the valence or conduction band of the crystal (according to the type of doping of the
silicon) Several studies have been carried out at different electric potentials In the case of
p-silicon or n-silicon at anodic potentials where positive charges (h+) are available in the
crystal a silicon-centred radical can be created by capture of a hole from the bulk (fig
42) The F- ion does not participate in the reaction but is thought to contribute through
VPalermo 35
electrostatic interactions by lowering the energy of the interaction step After the
formation of the Si-OH group the reaction proceeds as shown in the scheme of fig 41
Matsumura et al [4] proposed that not only water but HF2- molecules also play a major role
in electrochemical etching of silicon leaving on the surface Si-F terminated bonds which
can be immediately attacked in an autocatalytic process (fig 43)
In the electrochemical reactions described above an external potential is applied to the
silicon crystal The chemical and electrochemical reactions in any case take place
simultaneously most of the time with the chemical path predominating at high pH Even
when no external potential is applied to the silicon partial electrochemical reactions can
take place at different ldquocathodicrdquo and ldquoanodicrdquo sites on the surface with an internal charge
exchange which ensures neutrality [11] This macroscopic silicon etching and hydrogen
bubble formation can lead to visually observable patterns on the surface when Si(100) is
immersed in ammonium fluoride even without applying a potential
Si Si
Si Si
H
H
H
F Si
Si
Si SiH
F F-H+ -2e
+HF2-
H
HF
SiSi
SiSi + F
F F
FH
H
H
F -H+ -2e
+HF2-
Fig 43 Autocatalytic electrochemical etching of silicon by HF2-
Inhomogeneities on silicon surfaces caused by electrochemical reactions and charge
transfer have been studied extensively because they are of fundamental importance in the
formation of porous silicon
Pore formation on silicon
When Si(100) or Si(111) are etched under anodic bias in fluorine-based solutions
microscopic pores form on their surface Several different morphologies of pores have
been observed with pore diameters ranging from 10 nm to several microns with depths of
several microns [14] Pore shape is very variable too ranging from ordered straight pores
to chaotic networks of branched pores (fig 44) Porous silicon has been known since the
fifties but it was only in 1990 that interest in this material increased following the
discovery that porous silicon layers were able to emit bright red light This led to a large
amount of research and now different classes of micropores can be reproducibly created
36 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Fig 44 Different types of Silicon micropores From ref [14]
mostly for optical and micromachining applications However there is still no unified
theory able to explain the nucleation and growth mechanism of all the different kinds of
pores
We will give a short description of some of these theories for more detail see Parkhutik et
al [15]
One model explains pore nucleation on the basis of physical processes such as hole
positive charge migration ion transport to the surface and small perturbations on the
silicon surface modelled as Fourier components The system is shown to be unstable and
some spatial frequencies that lead to pore nucleation evolve from the etching process
A second model focusses on stationary pore growth without explaining the nucleation
stage According to this model silicon dissolves preferentially at pore edges because h+
charges are attracted by the stronger electric field present at these edges
A third class of models explains pore growth as a Diffusion Limited Aggregation (DLA)
process where the random walk of h+ charge carriers through the depleted layer present at
the silicon-liquid interface controls pore shape
Finally the model by Carstensen Cristophersen and Foll [16] proposes that areas of the
surface of some characteristic size LCO are etched by synchronized ldquocurrent burstsrdquo in the
flow of h+ charges These bursts dissolve silicon through cyclic stages of surface oxidation
oxide removal and hydrogen passivation Areas where a burst has recently taken place are
less passivated and thus more likely to be etched again in this way the pore bottom
continues to dissolve while the pore walls are passivated and are thus less favourable areas
towards current bursts
VPalermo 37
EXPERIMENTAL RESULTS
In the following sections we will show some experimental results obtained from STM and
AFM measurements of fluorine-treated Si(100) surfaces In the first part the results of
mild etching using concentrated and dilute HF solutions at low pH are presented In the
second part the results of etching at high pH using ammonium fluoride are presented and
the mechanism of pore formation discussed
Etching of Si(100) in dilute and concentrated HF
Samples were cut from different areas of an 8-inch diameter p-doped silicon(100) wafer
(10 Ω-cm) supplied by MEMC Electronic Materials Each series of STM measurements
was carried out over at least six different areas on at least two identical samples Low
Electron Energy Diffraction (LEED) was used to check the surface cleanliness of the
samples before STM measurements
Table 41 summarizes the different treatments of each sample After etching with
electronic grade HF each sample underwent a final rinse in Ultra-Pure Water (UPW
resistivity gt18 MΩ-cm) Both the HF and the UPW were allowed to flow continuously
over the sample surface Some samples were not etched with HF at all but just washed with
UPW to observe the morphology of the native oxide layer (~2 nm thick) covering the
surface All of these processes were carried out under nitrogen to limit reoxidation and the
samples were then introduced from the nitrogen atmosphere directly into the vacuum
chamber and degassed overnight at sim150degC before LEED and STM measurements
STM images were obtained from each sample using the same measurement parameters
(sample bias 4 V feedback current 1 nA scan speed 800 nm s-1) These parameters
38 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
provided a satisfactory level of reproducibility for all the samples Measurements were
made over an area of 500x500 nm (image size 500x500 pixel) Slope correction was
carried out by subtracting row-wise and column-wise fitted slopes from the entire image
which gave better results than the simple subtraction of a fitted plane especially for the
rougher samples Following slope correction the rms roughness
sum minus=xy
hyxhN
22 ))((1σ
and the 2-D Fourier transform
)(22
2
)(4
)( vyuxi
xyeyxhvuF +sum∆
= π
π
were calculated for each image where N2 is the number of pixels composing the image
h(xy) is the surface height at each point ∆ is the distance between points h is the mean
height and u v are the spatial frequencies The radial power spectrum PS(f) of the STM image data is obtained from the angular
average of the squared Fourier transform with f 2 = u2 + v2
Fig 45 shows the STM images obtained from the various samples Sample A still covered
with its native oxide layer shows an irregular surface with RMS roughness of ~05 nm
(see Table 41) Observing the sample with LEED gave no diffraction pattern even at
relatively high incident electron energies because of the surface oxide coverage After 1
min etching in dilute HF (sample B) the morphology is similar to the original one though
Table 41 Sample treatments average RMS roughness and integrated area of power spectra
Sample Treatment RMS roughness (nm)
PS area (f lt 01 nm-1)
PS area (f gt 01 nm-1)
A Rapid dip in water 053 plusmn 013 366 032
B 1 min in HF 5 + 10 min in water 051 plusmn 008 272 034
C 30 min in HF 5 + 10 min in water 062 plusmn 008 1064 038
D 5 sec in HF 49 + 10 min in water 042 plusmn 004 183 022
VPalermo 39
Fig 45 STM images of each group of samples showing the topography of the silicon surface A) noetching original oxide surface B) after 1 min etching in dilute HF C) after 30 min etching in diluteHF D) after dipping in concentrated HF Grey scale indicates height of the surface from lower (black) to higher (white) The images are 250x250 nm ie representative portions of the images usedfor the roughness measurement and PSD analysis
some of the larger features have disappeared and the image quality is better maybe due to
improved tunnelling due to the cleaner surface The RMS roughness is comparable to that
of the original surface Clear diffraction patterns are visible using LEED though at quite
high energies (200 eV) After prolonged etching (sample C) the RMS roughness increases
to 062 nm and a long-range corrugation is visible on the surface even if the LEED pattern
is good
The samples dipped in concentrated HF (D) reveal the bare SiSiO2 interface which has a
disordered aspect and protrusions over a wide range of dimensions The quality of the
STM images of sample D is very good probably due to the cleanliness of the surface
40 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 E -0 2
1 E -0 1
1 E + 0 0
1 E + 0 1
1 E + 0 2
1 E + 0 3
0 0 0 0 0 1 0 1 0 1 0 0f (1 nm )
nm^4
A a s re c e iv e d
B e tc h e d 1 min
C e tc h e d 3 0 min
D e tc h e d HF 4 9
Fig 46 Log-Log plot of the averaged power spectra of the STM images for all the samples
which gives a more stable tunnelling junction The LEED pattern is excellent showing
clear diffraction peaks at energies as low as 37 eV comparable to that obtained after high
temperature cleaning in UHV
Fig 46 shows the power spectra of the samples The high frequency and low frequency
areas of the power specturm are considered separately Table 41 shows for each sample
together with the roughness the integrated area of the power spectrum for the high and low
frequency part
We first examine the differences between the samples in the low frequency part of the
spectrum (lt01 nm-1) In this spatial range the short dip in concentrated or dilute HF
(samples BD) seems to lower the roughness of the sample removing some of the larger
features of the native oxide surface visible in Fig 45a and thus decreasing the
corresponding part of the power spectrum Sample C however etched with HF for 30 min
shows an increase in roughness peaking at 0012 nm-1 (~85 nm corresponding to the
typical dimensions of the corrugations visible in Fig 45c) On this scale sample C has a
PS density of 435 nm4 compared to 101 nm4 for sample A
Samples A B and C have the same PS in the high frequency range showing that the fine
structure of the surface is not changed by the HF etching Sample D on the other hand
shows a decrease of roughness for frequencies gt 015 nm-1 This effect could be due to the
VPalermo 41
improved surface cleanliness after etching with concentrated HF which would give a more
stable STM junction thus reducing the high frequency noise in the image
Fig 45 and the analysis of the power spectrum of each sample shows that a rapid dip in
HF removes the native oxide but does not lead to major changes in the morphology of the
Si surface its only effect being the removal of some of the larger features present on the
original surface Prolonged etching on the other hand increases the RMS roughness of the
surface
Etching of Si(100) in concentrated ammonium fluoride and nano-hole creation
Two different types of commercial p-doped Si(100) wafers (2Ω-cm and 10Ω-cm) from
MEMC were used Several different samples of 10x5 mm were immersed for 10 minutes
in 40 electronic grade NH4F solution under agitation Previous works used low
temperatures or anodic potentials applied to the silicon to avoid gaseous hydrogen
production and to obtain a uniform surface but in our experiment we worked at room
temperature to check the influence of hydrogen bubbles on surface morphology During
the etching the stirring was sufficient to provide a uniform concentration of reagents over
the whole sample surface but not to mechanically remove the hydrogen bubbles from the
silicon surface
After the etching each sample was rinsed in ultra-pure water to remove any etching
residues and observed with STM AFM and optical microscopy The AFM measurements
were made in air while for STM measurements the samples were rapidly dried with
nitrogen and inserted into the vacuum system to avoid surface reoxidation After insertion
into the vacuum surface cleanliness was checked with LEED and the surface morphology
observed by STM Parameters for STM measurements were sample bias +4 V 1 nA
current 16 Hz scan rate The images obtained were stable and reproducible over several
days Scan parameters for AFM were 20 nN force and 1 Hz scan rate
Some of the samples were cleaned with an RCA standard clean [1] before NH4F etching to
check the influence of possible surface contaminants on the final results Eight different
samples were prepared and more than sixty STM images of the samples were taken at
different points of the various samples
42 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
(110)
(110)
Fig47 abc) STMtopographic images of differentetching morphologies Eachimage is 500x500 nm Z-ranges are 10 10 and 18 nmrespectively d) STM image of a bridgecreated by etching of the lowerlayers of silicon (black arrow)Image is 250x250x6 nm
After ~2 min of immersion in the solution hydrogen bubbles become visible on the sample
surface The production is slow and the bubbles are quite stable on the sample without
detaching Thus some areas of the surface are masked from the liquid etching action
STM observations (fig 47) show that at the nanometer scale the surface is unevenly
covered with holes of radii ranging from 10 to 200 nm with depths of 2-4 nm These holes
have a wide range of different shapes and distributions In most cases the surface was
covered with a uniform distribution of round-shaped holes (fig47a) indicating isotropic
etching The dimension and the density of the holes changed greatly from sample to
sample and even over the surface of a single sample In some cases the etching was
anisotropic yielding nearly square holes and layered structures as shown in fig 47b
Square holes have been previously observed in cases where the etching speed in the (110)
direction is significantly smaller than in the (100) direction [17]
Over these areas (fig 47b and especially 47c) it is clearly visible how once the surface
had been attacked the reaction continued to preferentially remove atoms at step
irregularities (kink atoms) straightening step edges Eventually the exposed underlying
silicon was also attacked and further holes created inside the previously etched larger
ones It was not possible to detect monatomic steps on this kind of surface The smallest
step height observed was ~15 nm corresponding to several atomic layers In the image
shown in Figure 47a the etching was not strong and created only anisotropic holes on the
surface In fig 47b and c the stronger etch proceeded laterally for several tens of
VPalermo 43
nanometres leaving straight steps several tens of nanometers long and roughly rectangular
holes as expected given the structure of the (100) crystal face In some cases a
significative underetch is observed and the formation of suspended bridges and tunnels
can be deduced from the STM images (fig 47d)
The formation of branched pores and suspended structures has been attributed during pore
formation to diffusion limited aggregation effects where the h+ charge carriers necessary
for silicon etching have a higher probability of reacting at pore bottoms than reaching the
upper part of the silicon surface In the case of very deep pores quantum wire effects have
been invoked to explain the pore growth mechanism [15] In our case though the pores
formed were very shallow the underetch depth being only a few nanometres on pores of
sim100 nm width Thus more than diffusion effects the main contribution to the
underetching process must come from anisotropic etching and some kind of autocatalytic
reaction path analogous to the one described by Matsumura et al [4] with some areas of
the silicon surface hydrogenated and thus less vulnerable to etching
Pre-treatment with RCA cleaning has no effect on the final morphology and this seems to
exclude pore nucleation being caused by presence of metallic or organic surface
contaminants
The morphology and the distribution density of the pits was quite uniform over
microscopic areas of the sample but changes were observed over the millimetre scale This
suggests that etching intensity is influenced by some large-scale parameter
Large-area measurements made with AFM or with an optical microscope (fig 48)
showed that the inhomogeneity of surface etching can be correlated with the masking
action of the bubbles While the fluoride dissolved the silicon hydrogen bubbles formed
by the reaction covered some areas of the surface thus blocking the etching over that area
generating macroscopic steps at the bubble-liquid border As the reaction proceeded more
hydrogen accumulated and the bubble diameter increased producing in this way a circular
pattern of steps The increase in bubble diameter was not continuous with time otherwise a
uniform surface slope gradient would have been obtained The formation of this circular
ldquoetching staircaserdquo indicates that the bubble growth was stepwise the bubble accumulating
more and more hydrogen without enlarging across the surface until it relaxed increasing
its diameter stepwise and covering more silicon The circular structures in fig 48a are not
co-axial and their asymmetry could derive from physical processes due to stirring or
irregularities on the surface
44 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
a b Fig 48 a) optical micrograph of etching patterns on Si(100) created by NH4F 12x09 mm b) AFM image of the circles border xy range is 40x24 microm z-range is 30 nm
The step structure was not destroyed by the etching even after the bubble detached from
the surface but on the contrary the etching process seemed to be influenced by the
presence of the step
Observing in detail a series of steps (fig 48b) a quite deep trench is visible at the base of
each step A close-up image of a step and the corresponding line profile of the trench is
shown in fig 49 The trench is sim5 nm deep with respect to the lower surface compared to
a step height of 22 nm
A similar structure has been recently obtained with electrochemical etching of p-type
Si(100) in 4 HF [18] in which a ldquocurrent burstrdquo etching model previously described
was assumed for silicon dissolution In that experiment the trench was created at the
border of silicon nitride masks and began to grow after a nucleation stage Preferential
trench etching was along the (110) direction and the trench growth was explained as an
effect of mechanical stress induced by the nitride mask and of electric field enhanced
dissolution which depended upon an external applied potential
While it is clear that in our system the gas bubbles have a masking effect similar to a
classical solid nitride mask it is unlikely that hydrogen present on the surface can induce a
significant stress in the silicon lattice as in the case of a nitride mask Furthermore no
external field was applied to drive preferential etching at the trench site
It has been proposed [19] that the cathodic and anodic part of the etching reaction
(hydrogen production and silicon oxidative etching respectively) take place at different
points on the surface with a net charge transfer between the different areas In this case
the highest reaction rates will correspond to the silicon area surrounding the bubble border
where a high number of positive charges will be available for the reaction Furthermore a
sharp trench extending into the silicon crystal will be a preferential electrostatic attractor
VPalermo 45
Fig 49 AFM image of the etched surface showing a step created on the surface by bubblemasking A stronger etching action is visible on the right side of the step as well as theprotected area on the upper side of the step (indicated by the arrows) Image is 10x10 micromz-range is 30 nm The profile on the right is taken from the central area of the image
for the h+ charge carriers coming from other ldquocathodic areasrdquo of the sample either from
other regions on the surface or from the back of the silicon chip [18]
In the areas where the hydrogen bubble had detached and the surface was exposed to the
etching the reaction was not uniform in the neighbourhood of the steps It is possible to
observe (fig 48b and 49) an area on the upper side of the step where less or even no
etching at all seems to have taken place as if the step was able to protect the surface from
etching While etching on the lower side with trench formation can be attributed to the
presence of the bubble the surface on the upper step side can be etched only after bubble
detachment so no masking effect can account for this result However a further
preferential attraction of h+ charge carriers from the already formed trench can be
hypothised electrochemically shielding the surrounding area from further etching If this is
true the shielding effect would be very strong with a relatively shallow 5 nm-deep trench
protecting an area of sim1 microm parallel to the step
To summarise the etching of Si(100) in NH4F is a complex process in which different
reaction paths both chemical and electrochemical co-exist Hydrogen bubbles formed by
the reaction act as a mask on the surface and create etching paths and inhomogeneous
etching of the surface Different kinds of pores are observed on the surface and in some
cases the anisotropy of the process is so strong as to give square-shaped holes and
underetching
46 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The diffusion of h+ charge carriers in the crystal is one of the main rate-determining steps
of the reaction and leads to the formation of a deep trench immediately outside the bubble
perimeter These trenches act as charge collectors and reduce the etching of the upper step
surface in the proximity of the steps
Bibliography
[1] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[2] Spectroscopic Ellipsometry Studies Of HF Treated Si (100) Surfaces Yao H Woollam
Ja Alterovitz SA Applied Physics Letters 62 (25) 3324-3326 Jun 21 1993 Influence
Of HF-H2O2 Treatment On Si(100) And Si(111) Surfaces Graf D Bauermayer S
Schnegg A Journal Of Applied Physics 74 (3) 1679-1683 Aug 1 1993 Kinetics Of
Oxidation On Hydrogen-Terminated Si(100) And (111) Surfaces Stored In Air Miura
T Niwano M Shoji D Miyamoto N Journal Of Applied Physics 79 (8) 4373-4380
Part 1 Apr 15 1996
[3] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995 Structure Of The Stepped SiSiO2 Interface After Thermal-Oxidation -
Investigations With Scanning Tunneling Microscopy And Spot-Profile Analysis Of
Low-Energy Electron-Diffraction Pietsch GJ Kohler U Jusko O Henzler M Hahn
PO Applied Physics Letters 60 (11) 1321-1323 Mar 16 1992
[4] Enhanced Etching Rate Of Silicon In Fluoride Containing Solutions At pH 64
Matsumura M Fukidome H Journal Of The Electrochemical Society 143 (8) 2683-
2686 Aug 1996
[5] A Study Comparing Measurements Of Roughness Of Silicon And SiO2 Surfaces And
Interfaces Using Scanning Probe Microscopy And Neutron Reflectivity Crossley A
Sofield CJ Goff JP Lake ACI Hutchings MT Menelle A Journal Of Non-Crystalline
Solids 187 221-226 Jul 1995
VPalermo 47
[6] Step-Flow Mechanism Versus Pit Corrosion - Scanning-Tunneling Microscopy
Observations On Wet Etching Of Si(111) By Hf Solutions Hessel HE Feltz A Reiter
M Memmert U Behm RJ Chemical Physics Letters 186 (2-3) 275-280 Nov 8 1991
[7] Comparison Of Si(111) Surfaces Prepared Using Aqueous-Solutions Of NH4F Versus
HF Higashi GS Becker RS Chabal YJ Becker AJ Applied Physics Letters 58 (15)
1656-1658 Apr 15 1991
[8] Wet Chemical Etching Of Si(100) Surfaces In Concentrated NH4F Solution -
Formation Of (2x1)H Reconstructed Si(100) Terraces Versus (111) Faceting Neuwald
U Hessel HE Feltz A Memmert U Behm RJ Surface Science 296 (1) L8-L14 Oct
10 1993
[9] Ideal Hydrogen Termination Of Si(001) Surface By Wet-Chemical Preparation Morita
Y Tokumoto H Applied Physics Letters 67 (18) 2654-2656 Oct 30 1995
[10] Atomic Structures Of Hydrogen-Terminated Si(001) Surfaces After Wet Cleaning
By Scanning Tunneling Microscopy Endo K Arima K Kataoka T Oshikane Y Inoue
H Mori Y Applied Physics Letters 73 (13) 1853-1855 Sep 28 1998
[11] On The Potential-Dependent Etching Of Si(111) In Aqueous NH4F Solution
Houbertz R Memmert U Behm RJ Surface Science 396 (1-3) 198-211 Jan 20 1998
[12] Etching Process Of SiO2 By HF Molecules Hoshino T Nishioka Y Journal Of
Chemical Physics 111 (5) 2109-2114 Aug 1 1999
[13] Etching Mechanism And Atomic-Structure Of H-Si(111) Surfaces Prepared In
NH4F Allongue P Kieling V Gerischer H Electrochimica Acta 40 (10) 1353-1360
Jul 1995
[14] Pore Formation Mechanisms For The Si-HF System Carstensen J Christophersen
M Foll H Materials Science And Engineering B-Solid State Materials For Advanced
Technology 69 23-28 Sp Iss Si Jan 19 2000
[15] Porous Silicon - Mechanisms Of Growth And Applications Parkhutik V Solid-
State Electronics 43 (6) 1121-1141 Jun 1999
[16] Formation And Application Of Porous Silicon Foll H Christophersen M
Carstensen J Hasse G Materials Science amp Engineering R-Reports 39 (4) 93-141 Nov
1 2002
[17] Pore Morphology And The Mechanism Of Pore Formation In N-Type Silicon
Searson PC Macaulay JM Ross FM Journal Of Applied Physics 72 (1) 253-258 Jul 1
1992
48 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
[18] Deep Electrochemical Trench Etching With Organic Hydrofluoric Electrolytes
Christophersen M Merz P Quenzer J Carstensen J Foll H Sensors And Actuators A-
Physical 88 (3) 241-246 Mar 5 2001
[19] Morphology Of Anodically Etched Si(111) Surfaces - A Structural Comparison Of
NH4F Versus HF Etching Houbertz R Memmert U Behm RJ Journal Of Vacuum
Science amp Technology B 12 (6) 3145-3148 Nov-Dec 1994
VPalermo 49
5 Surface modification of silicon in vacuum void creation and oxide desorption
The main reason for the huge success of silicon in the microelectronics industry is not due
to its superior properties as a semiconductor Other materials for example germanium
have better qualities such as higher mobility of charge carriers and lower noise levels
which would allow the construction of faster and higher performance devices
The widespread use of silicon however is mainly due to the outstanding characteristics of
its oxide Silicon dioxide (SiO2) is a very good electrical insulator easy to form
chemically and thermally stable and is compatible with lithographic and metal deposition
processes Germanium oxide on the contrary is too reactive to be used
Even the use of Si(100) substrates for nearly all microelectronic devices is dictated by
oxide quality The (111) face of silicon crystal can be easily cleaved and flattened and
almost atomically perfect surfaces can be obtained with simple chemical procedures (as
described above) But the density of interfacial defects is highest for Si(111)-SiO2
interfaces and lowest for Si(100)-SiO2 ones so microchips will continue to be fabricated
on Si(100) wafers
SiO2 (silica) is present in 95 of the earthrsquos minerals in different allotropic forms such as
quartz tridymite and cristobalite In the bulk each silicon atom is bonded to four oxygens
in a Si-O-Si tri-dimensional network Si-O bonds are 016 nm long and form an angle
ranging from 120deg to 150deg
Three typical intrinsic defects are present in SiO2 The so-called Ersquo centres are oxygen
vacancies with a hole localised on a silicon atom with only three Si-O bonds
50 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
O3Si +SiO3 Whereas the PR (peroxy radical) defects are holes trapped by a charged
peroxy moiety with a O3Si-O-O+ SiO3 structure The NBOHC (non-bridging oxygen hole
centres) derive from water or hydrogen contamination and are schematized as O3Si- O- H-
O-SiO3
The atomic structure of the Si-SiO2 interface varies enormously Local domains resembling
the tridymite and the cristobalite structure of silica are present but it seems that only 10
of the interface is ordered [1] Far from the interface the SiO2 bulk is completely
disordered The passage from bulk Si to stoichiometric SiO2 passes through a non-
stoichiometric SiOx layer 07 nm thick
When a clean silicon surface is exposed to atmospheric oxygen a thin sim2 nm thick layer
of native oxide forms spontaneously which is usually removed and substituted with
thicker better quality oxide layers before further processing
Silicon is usually oxidised by thermal annealing at temperatures between 800deg and 1100deg
in an atmosphere of pure O2 with some water eventually added to increase oxidation speed
Thermal oxides made in pure oxygen (dry oxides) grow more slowly than oxides produced
in an oxygen-water atmosphere (wet oxides) but are usually of better quality
According to the Deal-Groove formula the time t needed to grow an oxide of thickness X
is given by [1]
1
212minus
minusminus
+=
ABXBXt α
where the constant B and BA decrease exponentially with temperature as
minus
kTEexp
with activation energies for dry oxidation of EB =123 and EBA =20 eV respectively
EB is related to the diffusion of oxygen in silicon while the value of EBA is interpreted as
the energy required to break a Si-Si bond The exponent α is 1 for wet oxidation and 0 for
oxidation at high temperatures and low oxygen pressures It has intermediate values for dry
oxidation This formula does not work well for low values of X at the initial stages of
oxidation and usually empirical corrections are used
An interesting characteristic of silicon is that at high temperature and in vacuum oxygen
can actually etch the silicon crystal giving gaseous products and the oxidized layer present
on the crystal becomes unstable (fig 51) [2]
VPalermo 51
1E-10
1E-09
1E-08
1E-07
1E-06
1E-05
1E-04
1E-03
1E-02
1E-01
0607080911112
1000T (1K)
P (T
orr)
SiO2 + Si rarr 2SiO(g) Oxide decomposition
Si+O2 rarr SiO2(s) oxide formation Fig51 phase diagram of the
oxygen-silicon system
Silicon oxidation apart from the initial nucleation stages at the monolayer level proceeds
uniformly over the whole surface with a planar reaction front moving from the surface
into the bulk
If heated under low oxygen partial pressure (vacuum or inert atmosphere) SiO2 is known to
decompose following the reaction
SiO2 + Si rarr 2SiOuarr (1)
The reaction begins with nucleation at defect points on the SiSiO2 interface and proceeds
in a spatially inhomogeneous manner with the formation of large voids on the oxide
surface [3]
Several studies have been made on the dynamics of void growth both on thick [4] and thin
[5] layers of SiO2 The process has been used to decorate otherwise unobservable defects at
the SiSiO2 interface [6] or to grow nanoislands of silicon on the void surface [7] It has
been suggested that the defects acting as nucleation centres could be metallic contaminants
present on the native surface which aggregate and catalyze SiO2 decomposition [8]
After oxide desorption the silicon surface is very rough In particular on Si(100) square
islands are observed several nanometres high which act as lsquopinning sitesrsquo on the motion of
monatomic steps on the Si surface It has been proposed that these islands can be composed
of silicon carbide coming from the organic contaminants present on the surface [9] or of
oxide clusters during partial reoxidation [10]
In the following section some experimental results obtained studying voids and nanoisland
growth are presented
52 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nanoisland growth on silicon
We used two different samples the first rougher than the second (i) p-type Czochralski Si
(100) with a RMS roughness of 05 nm shown in fig52a and (ii) p-type epitaxial Si(100)
with a RMS roughness of 019 nm shown in fig 52b The roughness was measured with
STM on the lsquoas receivedrsquo samples on different areas of 500x500 nm Every sample was
covered by a layer of native oxide ~2 nm thick
Some of the samples (AC) were introduced into the UHV system without any cleaning
others (BD) were dipped in concentrated HF (49) to remove the surface oxide without
etching the silicon [11]
After degassing the samples were heated resistively increasing the temperature slowly to
keep the pressure within the 10-10 mbar range during heating Each sample was held at
900˚C for 30 min For some of the samples the heating was stopped at lower temperatures
to obtain incomplete oxide desorption and surface roughening thus enabling observation
of the different steps of the process Table 51 summarizes the different treatments for each
group of samples
Fig52 Original surfaces of Czochralski (a) and epitaxial (b) silicon Every image is 500x500 nm
VPalermo 53
Table 51 Summary of different treatements
Silicon type Oxide removed
Heating Islands density (microm-2)
Av Island Volume (nm3)
A Czochralski No 30rsquo at 900˚C 24x103 353
B Czochralski Yes ldquo 20x104 71
C Epitaxial No ldquo 28x102 2450
D Epitaxial Yes ldquo 23x104 63
Fig53 ab) Surface roughening on group A samples after heating 30 min at 800deg and900deg respectively in presence of an oxide layer Image size 500x500 nm cd) Surface roughening on group B samples after heating 30 min at 700deg and 900degrespectively after removing the oxide layer with HF Image size 250x250 nm
Fig 52a is the native oxide surface of sample A quite disordered and irregular This
surface was stable when heated to 700˚C while at higher temperatures the oxide began to
desorb Fig 53a was taken after heating the sample at 800˚C for 30 min
54 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The oxide began to desorb in correspondence with defect points creating oxide-free
surface voids which enlarge radially uncovering the silicon surface Some surface silicon
atoms are removed through reaction (1) Mobile silicon atoms moving randomly over the
surface aggregate forming a nanocrystal at the initial defect point surrounded by a stable
flat oxide-free surface with some steps The initial SiSiO2 interface as observed after
oxide etching with HF [12] is similar to the one shown in fig 52a and thus the surface
observed in the void area is a completely new one coming from surface reorganization
The reaction continues leading to a coalescence of the voids the final result of the process
is shown in fig 53b with protruding islands aligned along the main crystal axes and flat
areas with some steps LEED measurements confirmed that this surface is crystalline
silicon Atomic resolution was obtained on the flat areas of the surface showing the
typical 2x1 reconstruction of Si(100) Figures 53c and 53d show the evolution of the
roughening on sample B after chemically removing the native oxide layer before
introducing the sample into UHV As mentioned above the initial morphology is similar to
that shown in fig 52a though the oxide has been removed as confirmed by LEED
measurements Fig 53c shows the surface after heating at 700˚ for 30 min Without the
oxide layer the surface began to reorganize at lower temperatures and the small irregular
grains of the original surface begin to grow and become more rectangular After heating to
900˚C (fig 53d) the surface is flat with small rectangular islands The shape of the islands
is similar to that shown in fig 53b but the lateral island dimension is one order of
magnitude smaller
To assess the importance of the initial surface on the process the same treatments were
repeated starting from flat epitaxial silicon (samples CD) The initial epitaxial surface is
smooth with periodic steps due to a small miscut angle (fig52b) On heating the epitaxial
silicon without etching (sample C) relatively large islands are obtained similar to the ones
shown in fig 53b where the ldquostep-pinningrdquo effect of the islands is evident The average
dimension of sample C islands was sim2500 nm3 while only 353 nm3 for sample A Island
density decreased from 24000 to 280 microm-2 This result is maybe due to a lower
concentration of defects in the oxide layer of the epitaxial silicon which allows the
expanding void to grow much larger before coalescence with neighbouring voids the
nanostructures within them thus reaching greater dimensions To verify this hypothesis we
applied the same treatment to clean not-so clean and purposely contaminated samples To
check that the defects inducing void nucleation and nanoisland growth derive from carbon
VPalermo 55
contamination we used electronic-grade methanol to contaminate the silicon surface
Figure 54 shows the results of this comparison
State-of-the-art cleaned samples obtained with multi-step RCA cleaning [13] showed the
development of very few voids the oxide desorption leaving large smooth areas of oxide-
and nanostructure-free silicon as shown in Fig 54a where the presence of residual oxide
can be seen Samples cleaned with a simpler one-stage Pirana cleaning solution showed a
higher number of islands per unit area (Fig 54b) and samples purposely contaminated
with methanol (Fig 54c) had the highest density of island nucleation of all three samples
When the native oxide had been chemically removed from the epitaxial silicon before
heating (sample D) the island size and density is similar to the standard as-received
silicon case (sample B) Thus starting from a flat substrate the process of surface
roughening and island creation was the same even when the oxide layer had been
previously removed
From these results some important indications can be drawn The first is that oxide
desorption temperature depends upon the contamination levels of the oxide layers (fig 54)
with desorption at lower temperatures for contaminated surfaces
Second once the oxide layer is removed the silicon atoms become mobile on the surface
at relatively low temperatures (~700˚C) Their initial configuration is disordered and has a
high surface energy so they crystallize in the small islands shown in fig53d reducing
their surface area and lowering the energy of the system When the oxide is present the
surface cannot reorganize at 700˚C and is stable At 800˚C the oxide begins to desorb in
correspondence with defect points on the surface As the void area enlarges a small
cba Fig 54 The effect of contamination on nano-island production Gradient-filtered STM images of (a)sample cleaned two times with RCA 400x400 nm (b) Sample cleaned with a Pirana solution 500x500nm (c) sample contaminated with MeOH before insertion into the UHV system 500x500 nm Annealingtime is 40 min at 800degC for all the samples
56 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
crystallite grows at the centre of the voids Further oxide desorption uncovers larger areas
freeing more silicon atoms which feed the initial island making it bigger
Fig 55 shows the scaled size distribution of the islands for each group of samples Every
distribution has been obtained using several images of different samples for each group
The size distribution of all the samples follows an exponential decay and the decay seems
the same for all the samples though the average volume of the islands differs by more than
one order of magnitude (see Table 51 for details)
The point defects which catalyze void nucleation can be metallic contaminants present on
the initial oxide surface [8] or organic contaminants which at high temperatures can
form SiC nanocrystals [14] We used Scanning Tunneling Spectroscopy (STS) to look for
differences between the islands and the surrounding flat silicon surface but no difference
was found Furthermore the total island volume per surface unit is very high (more than
104 nm3microm2) and it seems unlikely that such a huge volume could consist of surface
contaminants Islands of pure silicon on silicon have been grown without evidence of
surface contamination[14]
To summarise the formation of nano-sized islands on silicon through surface diffusion
was studied At high temperatures the oxide layer covering the surface decomposes non
uniformly and circular voids of clean silicon are created The presence of the oxide layer
blocks surface silicon atomic motion and surface reorganization except within the voids
causing the growth of islands more than 10 nm high and 30 nm wide on the silicon surface
According to STM and LEED measurements we can say that the islands are mostly
Fig 55 Size distribution (scaled) for the samples described in table 1 The line is a reference for the eye
VPalermo 57
composed of silicon with traces of other substances such as carbon Final island densities
and dimensions depend upon the initial purity of the oxide layer
On the other hand if the oxide is removed chemically before heating the island growth
process is not so localised and takes place simultaneously over the whole surface yielding
smaller and more numerous islands
58 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[2] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[3] High-Temperature SiO2 Decomposition At The Sio2Si Interface Tromp R Rubloff
GW Balk P Legoues FK Physical Review Letters 55 2332-2335 Nov 1985
[4] Defect Microchemistry At The SiO2Si Interface Rubloff GW Hofmann K Liher M
Young DR Physical Review Letters 582379-2382 Jun 1987 Defect Formation In
Thermal SiO2 By High Temperature Annealing Hofmann K Rubloff GW Mccorkle
RA Applied Physics Letters 49 1525-1527 Dec 1986 Kinetics Of High-Temperature
Thermal Decomposition Of SiO2 On Si(100) Liher M Lewis JE Rubloff GW Journal
Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 5 1559-1562 Aug
1987
[5] Thermal-Decomposition Of Very Thin Oxide Layers On Si(111) Kobayashi Y Sugii
K Journal Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 10 (4)
2308-2313 Part 3 Jul-Aug 1992 Scanning Tunneling Microscope Study On Mid-
Desorption Stages Of Native Oxides On Si(111) Kobayashi Y Sugii K Journal Of
Vacuum Science amp Technology B 9 (2) 748-751 Part 2 Mar-Apr 1991 Controlled
Growth Of SiO2 Tunnel Barrier And Crystalline Si Quantum Wells For Si Resonant
Tunneling Diodes Wei Y Wallace RM Seabaugh AC Journal Of Applied Physics 81
(9) 6415-6424 May 1 1997
[6] Defect Formation In SiO2Si(100) By Metal Diffusion And Reaction Liher M
Dallaporta H Lewis Je Appl Phys Lett 53 589-591 Aug 1988 SiO2 Film
Decomposition Reaction Initiated By Carbon Impurities Located At A Si- SiO2
Interface Raider Si Herd Sr Walkup Re Applied Physics Letters 59 (19) 2424-2426
Nov 4 1991
[7] Nanometer-Scale Si Selective Epitaxial Growth On Si(001) Surfaces Using The
Thermal Decomposition Of Ultrathin Oxide Films Fujita K Watanabe H Ichikawa M
Applied Physics Letters 70 (21) 2807-2809 May 26 1997 Pyramidal Si Nanocrystals
VPalermo 59
With A Quasiequilibrium Shape Selectively Grown On Si(001) Windows In Ultrathin
SiO2 Films Shibata M Nitta Y Fujita K Ichikawa M Physical Review B 61 (11)
7499-7504 Mar 15 2000
[8] Stacking-Fault-Induced Defect Creation In SiO2 On Si(100) Liher M Bronner Gb
Lewis Je Appl Phys Lett 52 1982-1985 May 1988
[9] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990 AFM And XPS Characterization Of The
Si(111) Surface After Thermal-Treatment Lamontagne B Guay D Roy D Sporken R
Caudano R Applied Surface Science 90 (4) 481-487 Dec 1995
[10] Scanning-Tunneling-Microscopy Study Of Oxide Nucleation And Oxidation-
Induced Roughening At Elevated-Temperatures On The Si(001)-(2x1) Surface Seiple
JV Pelz JP Physical Review Letters 73 (7) 999-1002 Aug 15 1994 Evolution Of
Atomic-Scale Roughening On Si(001)-(2x1) Surfaces Resulting From High-
Temperature Oxidation Seiple JV Pelz JP Journal Of Vacuum Science amp Technology
A-Vacuum Surfaces And Films 13 (3) 772-776 Part 1 May-Jun 1995
[11] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995
[12] Morphological Changes Of The Si [100] Surface After Treatment With
Concentrated And Diluted HF Palermo V Jones D Materials Science In
Semiconductor Processing 4 (5) 437-441 Oct 2001
[13] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[14] Nanoscale Roughening Of Si(001) By Oxide Desorption In Ultrahigh Vacuum
Gray SM Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology
B 14 (2) 1043-1047 Mar-Apr 1996
60 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 61
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and nano-lines
When the silicon surface is not protected by a native oxide layer or by a hydrogenated
passivating layer it is very reactive The surface chemistry of both Si(100) and Si(111) has
been extensively studied starting from clean surfaces prepared in UHV because of its great
relevance to the microelectronics industry and its technological interest The diffusion and
reaction of molecules and atoms on silicon is an interesting scientific problem on its own
apart from technological issues because silicon reconstruction yields a very complex and
anisotropic surface
We give here a brief summary of the adsorption behaviour of several elements on Si
surfaces The summary is not comprehensive and only the most interesting characteristics
for each substance are given
Table 61 Summary of adsorption behaviour of atoms and molecules on silicon surfaces [12]
Hydrogen
Molecular H2 shows low reactivity towards silicon while atomic hydrogen easily forms
Si-H bonds and can even break Si-Si bonds
Alkali metals
Alkali metals diffuse rapidly into SiO2 and can damage silicon-based transistors
Transition
metals
All transition metals apart from gold and silver react with Si forming metal silicides
62 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nickel A very common contaminant of silicon Nickel from even slight contact with stainless
steel tweezers can contaminate silicon samples forming its silicide and is very difficult
to remove even at high temperatures because it segregates on the silicon surface
Palladium Forms silicides especially Pd2Si and can be used to make contacts possessing a low
Schottky barrier
Titanium Widely used to fabricate contacts for silicon-based microdevices TiSi2 is one of the
more stable and highly conductive silicides
Tungsten The silicide is more stable than TiSi2 on polysilicon and is thus used for polysilicon
gate contacts
Platinum Silicide is used for bipolar transistors PtSi contacts on n-silicon give Schottky diodes
Cobalt Cobalt silicide gives better quality contacts than TiSi2 at the nanometrelevel but is less
used at the moment because it consumes too much silicon during its growth
Iron Silicides have been studied for potential optoelectronic applications because FeSi2 has a
directly accessible bandgap and is optically active
Group 13
Group 13 metals are used in IC technology to dope silicon (giving p-type doping) and
to make III-V type semiconductors When evaporated on Si(100) or Si(111) they react
strongly with the surface giving surface reconstruction and kink formation at steps
Boron Widely used for doping Usually deposited with decomposition of boron hydrides
Sticking coefficient of simple hydrides is very low so decaboranes (B10H14) are used for
deposition
Aluminum Aluminum is widely used for microelectronic contacts but in some cases it reacts with
silicon diffusing into the crystal and with SiO2 stealing oxygen atoms to form Al2O3
Group 14
Carbon Carbon can form a composite semiconductor with silicon (silicon carbide) widely used
to fabricate integrated circuits for use at elevated temperatures or in the presence of
ionising radiation When carbon is deposited on silicon usually polycrystalline films
with properties between diamond and graphite are obtained
CO adsorbs molecularly on silicon without breakage of the C=O bond Other
unsaturated hydrocarbons do not react with silicon Even very reactive strained
molecules like cyclopropane have low sticking coefficient Ethylene acetylene and
propylene adsorb molecularly on Si(100) each molecule sitting on a silicon dimer
interacting with the dangling bonds of the surface Acetylene also adsorbs molecularly
on Si(111) on the 7x7 reconstructed surface preferential adsorption on centre atoms
with respect to corner atoms is observed Benzene and other aromatic ring compounds
adsorb molecularly on the surface They can form σ or π-type bonds with the silicon
laying parallel or tilted respect to the surface according to the bond type
Silicon Silicon atoms can be deposited on the surface with molecular beam epitaxy or
VPalermo 63
decomposition of silanes and chlorosilanes Silanes with single Si-Si bonds have a high
sticking probability and dissociate upon adsorption giving trihydrides and
monohydrides
Chlorosilanes have high sticking coefficients too and decompose on the surface but
they can etch the surface through the reaction SiCl4 + Si rarr 2SiCl2
Germanium Germanium has a lattice constant similar to silicon (Ge lc is 4 larger than Si) so that
various alloys of Si1-xGex can be formed The growth of Germanium on silicon is quite
peculiar with several monolayers adsorbing uniformly on the surface followed by
island formation (Stranski-Krastanov growth)
Group 15
Nitrogen Exposure to ammonia and nitrogen gas at high temperatures leads to the formation of
silicon nitride layers (Si3N4) NH3 dissociates on the surface and reacts with the
dangling bonds giving Si-NH2 and Si-H termination Silicon nitride is stable on the
surface up to 1100degC above this temperature it desorbs as Si2N
Phosphorous Phosphine (PH3) adsorbs dissociatively as Si-PH2 and Si-H and behaviour is similar to
that of ammonia
Arsenic A monolayer of As is often deposited on silicon as a substrate for GaAs growth Arsenic
forms dimers on Si(100) creating dimer rows which are parallel (perpendicular) to the
underlying silicon dimer rows when it is deposited at high (low) temperature On
Si(111) Arsenic breaks the 7x7 reconstruction giving a 1x1 pattern This 1x1 As
monolayer acts as a passivating layer and can resist further As adsorption oxygen and
air
Antimony and
Bismuth
Due to their large covalent radii Sb and Bi form only short dimer rows on silicon and
only at high temperatures
Group 16
Oxygen
Oxygen can oxidize or etch silicon according to the reactions
Si(s) + O2 rarr SiO2 (s)
Si(s) + frac12O2 rarr SiO(g)uarr
Silicon oxidation has already been described in the previous chapter Molecular oxygen
adsorption on silicon is mostly dissociative O atoms break the Si-Si bonds forming an
Si-O-Si bridge or a peroxy bridge Si-O-O-Si Hydrided silicon surfaces do not have
surface dangling bonds and are thus more protected from oxygen attack Etching is
favoured at high temperatures and low oxygen pressures [3] At T gt700degC silicon oxide
on silicon decomposes as SiO(g)
Water
H2O adsorbs dissociatively on silicon easily on Si (100) and with more difficulty on
Si(111) The molecule decomposes giving Si-OH and Si-H on adjacent sites
64 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Group 17
Fluorine
Fluorine adsorbs and reacts easily on silicon dissociating on the surface and forming
Si-F bonds It can then penetrate the surface and attack polarised Si-Si backbonds The
surface is etched with production of SiF4 and SiF2
Chlorine
Chlorine reacts aggressively with silicon etching the surface Cl2 and F2 are often used
commercially to etch silicon usually in a plasma Atomic Cl and Cl2 adsorb on Si(100)
up to saturation The most stable configuration seems to be a Cl atom bonded to each Si
atom of a surface dimer in a symmetric structure but metastable configurations with
two Cl atoms bonded to a buckled dimer and even a central Cl atom bridged across a
dimer have been observed
On Si(111) silicon mono- di- and tri-chlorides are formed In particular heating at
Tgt400degC a uniform Si-Cl monolayer is obtained and the 7x7 reconstruction changes
into a bulk like 1x1 lattice The surface transition induced by Cl adsorption is fully
reversible The 7x7 reconstruction can be restored on desorbing the chlorine by heating
at Tgt1100degC The 1x1 domains nucleate at the lower terrace side of steps the 7x7 at the
upper step edge
Bromine
Like hydrogen bromine maintains the 2x1 structure of Si(100) forming Si-Br bonds
with the dangling bonds of the silicon dimer rows At high exposure some etching of
silicon by formation of volatile SiBr3 species has been observed
Surface diffusion on silicon
In the previous chapter we observed nano-island growth on silicon surfaces with native
oxide and surfaces contaminated with organic impurities While this phenomenon has been
observed in several experiments the formation and growth mechanisms of nanoislands has
still not been satisfactorily described theoretically Various substances have been
hypothesised as nanoisland nucleating agents such as organic or metallic contaminants
present on the surface or oxide clusters (see previous chapter) The island growth process
has been attributed by various workers to lsquosome sort of kinetic instabilityrsquo [4] to the
pinning of step flow by SiC clusters [5] or to a mesoscopic atom flux from areas with low
step density to areas of higher step density [6]
One mechanism proposed to explain surface roughening and island growth is the Ehrlich-
Schwoebel effect ie the presence of an energetic barrier that reflects atoms approaching a
VPalermo 65
Molecule (L
O2
H2O MeOH
CO CO2 CH4
clean surface
downward
weak on si
The aim o
surface an
To obtain
the ones
spontaneou
first obser
surface W
nanoisland
Experime
We used S
removed f
41 mixtur
water and
whole hea
1200degC re
kept below
Table 62 Treatment island density and island volume for each molecule tested
Dose angmuir)
Heating time at 800degC (min)
Island density (microm-2)
Ave Volume (nm3)
Equivalent layer thickness (nm)
280 10 no islands - -
90 10 no islands - - 30 10 2000 19 004
180 10 360 516 019 30 60 no islands - -
180 10 no islands - - - 10 no islands - -
step [7] Recent experiments however indicate that the Schwoebel effect is
licon [48]
f our experiments was to study the reaction of simple molecules with the silicon
d to check the ability of these molecules to nucleate nanoisland growth
a better understanding of the process we chose simple molecules starting from
which are more likely to be present as traces in UHV chambers where
s nanoisland growth is often observed after sample heating Using STM we
ved at the atomic level the possible interactions of these molecules with the
e then heated the surface and checked the density and dimensions of the
s produced
ntal procedures and results
i(111) wafers p-doped 0015 Ω cm Gross contamination and particles were
rom the sample surface with a standard Pirana etch (15 min dip in a H2S4H2O2
e at T=80degC) After this the samples were thoroughly washed in ultra-pure
introduced into the UHV system After degassing for 5 hours at 600degC the
ting stage was allowed to cool down and a sequence of rapid flashes to 1100deg-
moved the native oxide layer and cleaned the surface Pressure during flashes is
1x10-9 mbar
66 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Surface flatness and cleanliness were checked with STM and the contaminant molecules
introduced into the UHV system through a leak valve Doses of several Langmuirs were
used for the experiments (1 Langmuir=10-6 torr sec-1) Pressure was kept constant at 10-7
mbar during adsorption The STM tip was retracted during gas deposition to avoid any tip
shielding effect After the adsorption process the leak valve was closed the base pressure
in the UHV chamber returning to below 10-10 mbar and the presence of adsorbed
molecules checked on the surface with STM Finally the sample was heated at 800degC for
several minutes to allow nanoisland growth Several different areas were examined on each
sample
Table 62 summarizes the dose of exposure and the heating time for each molecule Other
experiments were carried out using smaller dosing times but only the significant results
are reported here
As expected oxygen and water react with the surface oxidising it The adsorption process
can be easily followed with the STM by lowering the tip from time to time and taking an
image No effect of the STM tip is observed over the scanned areas Upon molecule
adsorption the order of the crystalline surface rapidly degenerates and after several
minutes the surface looks completely covered by irregular atomic-sized protrusions
corresponding to Si-OH or Si-O-Si species On heating to 800degC the contaminating species
were easily desorbed through SiO formation the surface became clean and the crystal
surface periodicity re-established No residues or nanoislands were observed
Amongst the molecules tested was methanol CH3OH reacts readily with the surface and
covers it with a disorderd layer already after sim 30 Langmuirs exposure (fig 61)
Fig 61 Si(111) surface during methanol adsorption at 0 1 and 5 minutes respectively (corresponding to 0 6 and 30 L exposure) The dark irregular lines are borders between different 7x7 domains Image size 60x60 nm
VPalermo 67
Synchroton radiation photoemission spectroscopy experiments by Carbone et al [9]
showed that methanol reacts with the rest atoms of the 7x7 silicon cell already after 1
Langmuir exposure but that longer exposures are needed for methanol to react with the
remaining silicon atoms Methanol adsorbs dissociatively forming SiOCH3 and Si-CHx
species on the surface
After methanol adsorption the sample is heated to 800degC and observed again with STM
As in the case of water and oxygen the heating restores the crystalline surface and the 7x7
pattern returns visible But this time quite a dense array of nanoislands is observed on the
surface (fig 62) The islands have an average diameter of 20 nm and a height of sim2 nm
At temperatures above 400degC the Si-OCH3 and Si-CHx species decompose oxygen and
hydrogen are desorbed and according to [9] carbon atoms remain as SiC dispersed
uniformly on the surface Our STM measurements indicate that the carbide does not
randomly cover the surface but that C atoms are concentrated at the nanoislands Rough
calculations based on island volume suggest that the islands must be composed of a SixC1-x
alloy with x varying between 05 and 1
Following these results we expected carbon monoxide to behave in a similar manner
adsorbing onto the surface and with nanoisland growth However no adsorption was
observed with STM and the surface looked perfectly clean even after a dose of more than
100 L of carbon monoxide Some species did adsorb onto the surface though because
after 10 minutes subsequent heating nanoisland growth was observed Island density in
this case was only 18 of the density obtained with methanol and island dimensions are
Fig62 Nanoislands on Si(111) created after 30 L adsorption of methanol and 10 min heating at 800degC Image size 200x200 nm
68 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
much bigger (see table 62 and fig 63) To check whether the islands had been nucleated
by some other contaminant present in the vacuum chamber we heated a blank sample
without introducing any molecule No island growth was observed Exposing the surface to
other simple molecules including carbon dioxide and methane also gave no nanoisland
formation
We can estimate the amount of carbon deposited on the surface during gas dosing The
total flux of molecules hitting the surface is obtained from the Hertz-Knudsen formula
TkmPF
π2=
Fig 63 STM images of Si(111) after exposure to methanol (left) and carbonmonoxide (right) contaminating molecules and subsequent nanoisland growthImage size 400x400 nm The size distribution of nanoislands is shown below eachimage
VPalermo 69
Fig64 Left Si(111) after 180 L contamination of CO and 16 hours heating at 800degC Right Si(111)without CO contamination after 16 hours heating at 800degC Image size 250x250 nm Vertical scale is thesame for both images z-ranges are 35 and 5 nm respectively
where P is the pressure in Pascal k=138x10-23 J K-1 is the Boltzmann constant T is the
temperature and m is the molecular weight (sim32 for methanol and sim28 for CO) in
kilograms
Using this formula we obtain a flux of 027 molecules nm-2 sec-1 for methanol and 029
molecules nm-2 sec-1 for CO After ten minutes exposure at 10-7 mbar pressure more than
150 molecules will have hit each square nanometer of the surface Surface density of
Si(111) 7x7 is sim16 atomsnm2 so each surface atom will be hit by several molecules which
could react or be adsorbed even assuming a sticking coefficient much lower than unity
To explain nanoisland growth we hypothesize that CO molecules adsorb molecularly and
-5
0
5
10
15
20
25
30
0 20 40 60 80 100 120
nm
nm
10 min40 min16 h16 h- no CO
Fig65 STM profiles of nanoislandsgrown for different annealing times The lower curve correspond to the surface heated at 16 hours without CO contamination
70 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Table 63 Island density and volume for the CO-contaminated surface at different heating times For each sample the thickness of a uniform layer having the same volume per unit area of the islands is calculated
Molecule Dose
(Langmuir) Heating time at
800degC (min) Island density
(microm-2) Ave Volume
(nm3) Equivalent layer thickness (nm)
CO 180 10 362 517 019 CO 180 40 550 1072 059 CO 180 16 h 505 3675 186
clean surface - 16 h rough - -
very weakly on Si(111) moving rapidly on the surface Because of their rapid diffusion
the STM is not able to image the CO molecules on the silicon [10]
Increasing the temperature part of the CO molecules will surely desorb but some of them
will dissociatively adsorb onto the surface The oxygen will then desorb as SiO while
carbon will remain fixed forming very stable Si-C bonds
CO decomposition will be favored at surface defects surface steps or in the proximity of
already formed SiC clusters An increased reactivity of the CO molecule on a surface in
correspondence with phase boundaries has already been observed on Pt surfaces [10]
In this case a reduced number of nucleation centres will react with the CO molecules
yielding large and fewer islands with respect to the case of methanol which does not have
the possibility of travelling long distances over the surface
Other factors are likely to influence the process of island nucleation after CO adsorption
the co-adsorption of other molecules which can slow down CO and favour its
decomposition cannot be ruled out as well as the formation of new defects at higher
temperatures Measurements with a variable temperature STM or with some other surface
analysis technique are planned in the future to confirm the proposed mechanism
When the surface was further annealed the nanoislands grew in size After 16 hours
heating very large nanoislands with diameters of sim35 nm and heights of sim20 nm became
visible (fig64 left) As a comparison when the same surface was heated without
nucleation centres deriving from contamination an irregular surface was obtained (fig64
right)
While the nanoisland density seemed to reach an asymptotic limit of 500 islandsmicrom2
(table 63) island size continued to grow with time (fig65) even when no further carbon
was supplied to the surface This indicates that even though the initial nucleating core of
the nanoisland is likely to be an SixC1-x alloy further growth is due to silicon atoms
diffusing from the crystal and being adsorbed by the growing island The final volume
VPalermo 71
occupied by the islands corresponds roughly to a 18 nm thick overall layer of removed
silicon
It is noteworthy that the clean sample after 16 hours annealing even though smoother than
the CO-contaminated one looked much more disordered with hardly any flat area visible
on nanometres scale In the presence of nanoislands however further annealing increases
island size but keeps part of the surface quite flat and ordered with flat areas visible
between the islands Thus the nanoislands can be imagined to act as ldquoimpurity sinksrdquo for
further contaminants approaching the surface yielding a greater but more ordered surface
roughening with respect to the uncontaminated silicon surface
Circular arrays of nanoislands
We used the voids described in the previous chapter formed during oxide layer
decomposition as nanoscopic masks to test the validity of the hypothesis outlined above
The small circular areas uncovered during oxide decomposition can act as ldquoskating rinksrdquo
where contaminants and diffusing silicon atoms can move freely over the surface while
the oxide layer all around and higher than the void area as well as being chemically
bonded to the surface silicon atoms will hinder surface diffusion Furthermore the oxide
passivating layer will at least in some cases prevent contaminating molecules from
adsorbing and decomposing on the surface In this way selective nanoisland growth inside
void areas can take place
The experimental procedure consisted of the following steps
1) An oxide-passivated sample was heated for 10 minutes at 800degC Oxide decomposition
began in correspondence with surface defects and spread laterally creating naked areas
of silicon
2) While the void enlarges the same defect that initiated oxide decomposition behaved as
an attractor for moving silicon atoms and nucleated the growth of a nanoisland at the
void centre
3) The void surface was exposed to contaminating molecules
4) The sample was then heated again at 800degC During the second heating the void
continued to enlarge and new nanoislands were created The freshly uncovered silicon
surface provided mobile silicon atoms to feed nanoisland growth
72 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
5) When the heating continued for long enough all the oxide was desorbed and a clean
silicon surface covered by circular groups of nanoislands was obtained
Figure 66 shows the different stages of the process
We thus heated an oxidized sample at 800degC for 10 minutes in UHV to create multiple
voids in the oxide layer After this we exposed the sample to the contaminating molecules
Then we further annealed the surface at the same temperature to promote island growth
The first time we observed nanoisland growth in the void area was by chance after heating
for a second time samples which had been stored for 2 weeks in vacuum Even at pressures
as low as 10-11 mbar some molecules will adsorb onto the surface slowly creating defects
and thus potential nanoisland nucleation centres It is possible to observe with STM that
the silicon crystal surface looks very clean just after a high temperature flash but even
after only a few days randomly adsorbed species will be observable on the surface
Unfortunately it is not possible to identify these species simply from in situ STM
measurements
Mass spectrometry measurements show that the residual gas contaminants in vacuum are
usually He Ar H2 CH4 CO CO2 and N2 [11] Some of these molecules such as
hydrogen will stick to the silicon surface at room temperature but will simply desorb
during annealing without nucleating nanoislands Others like CH4 or CO2 will not react
with the surface to generate nanoislands as demonstrated previously The best candidate
for contamination of samples stored in UHV seems thus carbon monoxide has the ability
321
4 5
Fig 66 Selective nanoisland growth within oxide voids See text for details
VPalermo 73
as shown before to nucleate nanoislands although co-adsorption mechanisms can not be
ruled out Fig 67 (left) shows the circular groups of nano-islands obtained after heating
void-covered samples stored for 2 weeks in UHV Small islands are observed in the former
void area The void itself has enlarged slightly uncovering a clean oxide-free area of the
silicon surface Apart from the central island created during the first period of heating the
surrounding ones have dimensions decreasing from the void border to the inner area of the
void
This confirms that most of the material needed for island growth comes from the silicon
atoms diffusing from the freshly uncovered areas around the original void The outer
nucleating centres are nearest to the silicon atom source and thus generate larger islands
For a faster process we directly exposed the void to a significant concentration of CO
Fig67 (centre) shows the islands grown after a 180 Langmuir exposure and subsequent
heating The islands are preferentially located around the original void perimeter the rest
of the original oxide-free surface remaining untouched This suggests that although the
whole surface is exposed to CO the gas only interacts chemically with the surface at the
SiSiO2 interface around the oxide-free void perimeter The SiSiO2 interface thus provides
preferential nucleation sites for nanoisland growth in the successive thermal annealing
step This recalls a similar phenomenon observed in the catalytic oxidation of CO on the
Pt(111) surface where CO molecules reacted with adsorbed O species only along the
perimeters of oxygen islands [10]
On repeating the CO exposure and thermal annealing steps a second circle of nano-islands
Fig67 Left circular area of nanoislands obtained from a sample stored 2 weeks in UHV 500x500 nmCentre nanoisland circle obtained after exposing the voids to 180 L of CO and heating for 10 min at800degC 400x400 nm Right two concentric circles obtained with further CO adsorption followed byheating 650x650 nm
74 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
was formed around the newly-expanded surface void perimeter (fig67 right) This
process can in principle be repeated again to obtain a series of concentric groups of
nanoislands the only limit depending on the minimal distance between neighbouring
expanding voids We have thus observed that contaminants adsorbed at low pressure over
long periods of time generate random island growth over the whole void area while CO
molecules adsorbed at relatively high pressures over short periods yield selective
nanoisland growth at the former void border Several theories can explain the different
results CH4 and CO2 as mentioned above did not nucleate islands when when the silicon
surface was exposed to them for short periods However the lower surface mobility of
these contaminants and the greater time available for reaction with the surface could be
the cause of nucleation of SiC clusters over the whole oxide-free void area and not only at
its border The coadsorption of water another UHV residual gas with CO on the naked
silicon areas could perhaps allow a surface reaction similar to that between CO and the
SiSiO2 interface during CO exposure Another explanation for the formation of these
random nucleation sites within the voids could be the coadsorption of residual carbon-
containing species present in the UHV system with hydrogen the major residual gas in
stainless steel UHV systems
The possibility of decorating silicon oxide borders with nanoislands using the high
mobility and selective decomposition of CO molecules suggests interesting applications
for nanofabrication processes that will be discussed below
Silicon nanowire creation on Si(100)
We performed most of the experiments described above on silicon (111) 7x7
reconstructed because this surface is easily prepared in UHV and above all because it is
isotropic and has no preferential directions for atom diffusion The 2x1 reconstruction of
silicon (100) even though it has a much simpler unit cell than Si(111) 7x7 shows a more
complicated morphology with lots of monoatomic and biatomic steps and dimer rows
parallel to each other The orientation of the dimer rows changes by 90deg on alternate
atomic layers Surface diffusion in this case is much more complicated because atoms will
experience the effects of the surface anisotropy and diffuse preferentially along or across
the dimer rows The diffusion energy of a silicon atom as example is 06 eV along a
dimer row and 085 eV across different dimer rows This while being a complication
VPalermo 75
offers interesting possibilities for the creation of ordered structures The monoatomic steps
and the dimer rows can for example act as templates for the formation of elongated
structures of composition similar to the islands described previously
On heating a Si(100) surface oxide decomposition and void growth takes place as
described for Si(111) In some cases the void shape reflects the surface symmetry
especially for very thin oxide layers and the voids often have a slightly squared shape
Nanoisland growth is also influenced by the substrate symmetry and the islands look
rectangular as irregular clusters surrounded by concentric patterns of monoatomic steps
The whole structure reminds vaguely a ldquozigguratrdquo a kind of stepped pyramid found in
Mesopotamia (fig68 inset)
Apart from nanoisland growth in some cases we observed spontaneous formation of
nanowire-like SiC structures on the Si(100) surface The wires were generated after
exposing a Si(100) surface in vacuum to traces of CO followed by sample annealing at
800degC for 15 minutes The wires exhibit lengths ranging from 10 to 100 nm and average
widths of ca 5 nm Although being randomly positioned on the surface they are perfectly
aligned along the crystal axes of the substrate (Fig68)
Several descending monoatomic steps can be observed around each nano-line and nano-
island Very often one of the line extremities coincided with an island
Nanoline
Nanoisland
Fig 68 A Si(100) surface covered by nanoislands and nanowires The monoatomic steps present on the surface are visible The two insets show a typical nanoisland and a nanowire obtained on this surface The typical ldquodimer-row ldquostripes are visible along the sides of the nanowire
76 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Couples of parallel lines starting from the same nano-island were also observed At high
resolution (fig 69) the lines showed a lamellar periodic structure with a periodicity of
around 4 nm and frequent defects Scanning Tunneling Spectroscopy measurements
showed no significant difference between the line and the surrounding silicon surface
The proposed formation mechanism for this kind of structure involves two stages in the
first one organic contaminating molecules adsorb and diffuse on the surface from the
vacuum environment stopping preferentially at surface singularities including steps and
defects
Several substances such as atomic carbon carbon monoxide ethylene and fullerene
decompose when adsorbed on silicon surfaces at temperatures above 500degC and create
localised defects on the surface inducing strain deformations and a metastable surface
reconstruction [12]
If these molecules remain attached to the steps they can diffuse rapidly along step border
until they meet a line and decompose In this way long lines of SiC defects parallel to the
surface steps can be produced In the second phase the clusters of silicon carbide can
-005
0
005
01
015
02
025
03
035
04
-3 -2 -1 0 1 2 3
V
dId
V lt
dId
Vgt
linesilicon
Current- Voltage behaviour measured with STS Fourier analysis of wire periodicity
asymp 4 Aring
High resolution image of a line
Fig 69 Top high resolution image of a nanowire showing the lamellar structure Fourier analysis (bottom left) shows that the lamellae periodicity is sim 04 nm bottom right the IV characteristics measured with the STM tip on the line and the silicon surface
VPalermo 77
25-600degC
Final linear structure
600-800degC
Contaminant Silicon
Fig 610 Schematic representation of nanowire formation mechanism
locally inhibit the spontaneous surface diffusion of silicon atoms acting as templates for
the reorganisation of surface steps (fig 610)
The final morphology shows disordered lines and islands each surrounded by a complex
pattern of silicon monatomic layers
At 800degC silicon atoms are highly mobile on the surface Due to atom diffusion the
surface reorganizes with a step-flow mechanism The presence of a fixed line of
contaminants blocks atom diffusion and step flow generating the monoatomic steps
around each wire (fig 611) SiC clusters act as lsquopinning sitesrsquo on silicon atom surface
diffusion and can be purposely used to direct the localised growth of nano-islands [13]
Although the possibility of using methanol and carbon monoxide as nano-island precursors
has been demonstrated the exact nature of the substances nucleating nano-wire growth is
still unknown
Further experiments will be needed to confirm these hypotheses as well as to identify the
contaminating species nucleating nanowire growth Carbon monoxide is a main candidate
for nanowire growth but the real nucleation process is likely to be a complex one
78 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Step flowblocked
Fig 611 Representation of step flow blocked by the presence of a nanowire 105x105nm
Bibliography
[1] Surface-Chemistry Of Silicon Waltenburg HN Yates JT Chemical Reviews 95 (5)
1589-1673 Jul-Aug 1995
[2] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[4] Growth Of Si On The Si(111) Surface Lanczycki Cj Kotlyar R Fu E Yang Yn
Williams Ed Das Sarma S Physical Review B 57 (20) 13132-13148 May 15 1998
[5] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
VPalermo 79
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990
[6] Nanoscale roughening of Si(001) by oxide desorption in ultrahigh vacuum Gray SM
Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology B 14 (2)
1043-1047 Mar-Apr 1996
[7] Step Motion On Crystal Surfaces Schwoebel Rl Journal Of Applied Physics 40 614-
618 Feb 1969
[8] Scanning Tunneling Microscopy Investigation At High Temperatures Of Islands And
Holes On Si(111)7x7 In Real Time Evidence For Diffusion-Limited Decay
Hildebrandt S Kraus A Kulla R Wilhelmi G Hanbucken M Neddermeyer H Surface
Science 486 (1-2) 24-32 Jul 1 2001
[9] Methanol Adsorption On Si(111)-(7x7) Investigated By Core-Line Photoemission And
Mass Spectrometry Of Photodesorbed Ions Carbone M Piancastelli Mn Zanoni R
Comtet G Dujardin G Hellner L Surface Science 370 (1) L179-L184 Jan 1 1997
[10] Atomic And Macroscopic Reaction Rates Of A Surface-Catalyzed Reaction
Wintterlin J Volkening S Janssens Tvw Zambelli T Ertl G Science 278 (5345)
1931-1934 Dec 12 1997
[11] Redhead PA Hobson JP Kornelsen EV The Physical Basis Of Ultrahigh
Vacuum Chapter 12 (Chapman amp Hall London 1968)
[12] The Si(001) C(4 X 4) Surface Reconstruction A Comprehensive Experimental
Study Norenberg H Briggs Gad Surface Science 430 (1-3) 154-164 Jun 21 1999
[13] Production Of Nanostructures Of Silicon On Silicon By Atomic Self-Organization
Observed By Scanning Tunneling Microscopy Jones D Palermo V Applied Physics
Letters 80 (4) 673-675 Jan 28 2002
80 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 81
7 Conclusions and proposals for future work It is very difficult to foresee the future of nanoscience and nanotechnology even over the
next few years exciting discoveries are continuously made opening up new possibilities
and progress in this sector is now one of the fastest and most varied So it is impossible to
predict whether computers will in the future be made with nanowires nanotubes single
electron devices or some kind of quantum-based transistor What does seem clear
however is that silicon devices can still further extend their levels of miniaturization to
remain competitive for at least the next ten years [1] During this period new emerging
technologies will reach scientific maturity and arrive at the production lines
Even when new technology prototypes do become available for full-scale production the
astronomical cost of changing from silicon to new technologies will be a major concern
Moreover a vast amount of extremely detailed knowledge is already available on all
aspects of silicon technology Thus the possibility of integrating innovative
nanotechnologies with standard CMOS silicon technology already seems an attractive
prospect [2]
For this and other reasons we have focussed our research efforts on surface phenomena on
silicon wafers which could be potentially useful for the development of silicon-compatible
nano-devices
Some important conclusions can be drawn from the work described in this thesis
bull Silicon surfaces of different crystal faces were studied at atomic resolution in ultra-
high-vacuum following transformations induced on the surface by chemical (etching)
and physico-chemical (organic contamination heating) treatments
82 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull The presence of surface contaminants induces at high temperature the growth of
nanometre-sized islands and lines on silicon
bull Final island densities and dimensions strongly depend upon the chemical nature of the
contaminant molecules and the treatments used Molecules interacting only weakly
with silicon will desorb upon heating whereas molecules which decompose easily on
Si will give widespread random nano-island nucleation Between these two extreme
cases complex behaviours of diffusion and subsequent decomposition are possible
bull Nanoscopic naked silicon voids were produced in oxide layers through carefully-
controlled thermal annealing It was found that void density and size depends upon
oxide purity surface contamination levels and annealing conditions
bull The growth of nano-islands and nano-voids was exploited simultaneously and the
voids used as nanoscopic masks to control the positions of growing islands Complex
silicon-on-silicon nano-structures were obtained with multi-step growth processes
bull Finally the use of macroscopic treatments and simple physical processes to produce
large quantities of nanometre-sized structures on silicon surfaces was demonstrated
These processes and the structures produced are of potential interest to the
microelectronics industry
We showed in previous chapters how it is possible to modify silicon surfaces using
chemical and physical methods The use of self-organization for the production of Si-on-Si
20 nm
Fig 71 Left SEM image of field emitter arrays made by conventional lithography [3] Center 3-D STM image of a self-organized void with a central nano-island Right Lateral schematic view of a field emitter device
VPalermo 83
nano-structures has the advantage with respect to other emerging techniques of being
silicon compatible and thus is a potential candidate for its implementation in the
production of new devices Furthermore this technique can create billions of strucures on a
wafer in only a few minutes being thus applicable to large-scale production
Even when ordered structures are obtained these methods still show a poor reproducibility
and control Chemically modified surfaces look very disordered at high magnification
with trenches and holes etched into the surface and progress has recently been made in
producing very small structures electrochemically [4] Oxide decomposition with
nanoisland growth in UHV can also yield ordered structures The void holes with
nanoislands located at their centres although quite variable in dimensions are more similar
to engineered devices than to a spontaneous random surface-roughening process In fact
there is a surprising similarity between the nanovoid-nanoisland structures and well-known
commercially available devices called field emitter arrays (FEA) These are small tips used
as electron guns in various kinds of displays and other devices Both the commercial
devices and our self-organized structures consist of a protruding tip at the centre of
microscopic holes on an insulating layer covering a semiconductor surface (see fig71)
Of course the order and size uniformity of commercial FEAs is much better than that
obtained with our process but it is encouraging to see that a self-organized structure
obtained with simple heating has a very similar surface topography and composition of
commercial working devices obtained after many expensive stages of optical lithography
deposition and etching
Nano-island density and size can to some extent be controlled and different
contaminants can be adsorbed onto the surface to modify the growth process In the future
with a better understanding of diffusion dynamics and growth mechanisms a multi-stage
fabrication process can be envisaged where purposely-designed molecules are adsorbed
onto silicon surfaces to diffuse organize in ordered arrays and favour surface
reorganization at the nanometre scale The same molecules could be used to deliver
dopants to obtain for example highly conducting nanostructures on otherwise undoped
highly resistive silicon The selective formation of nanostructures at the SiSiO2 border
can be used to scale down mesoscopic patterns to the nanometric level A silicon oxide
layer can be easily patterned on the surface with normal optical lithography and then
nanostructures can be built following the pattern edge In this way the production of very
small nanowire based devices may be possible (fig 72)
84 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
D
Gate
S
D
S
SiO2
SiO2
2 Heating nanowire creation
and oxide removal
3 Source-drain deposition by conventional lithography
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
4 Insulating layer (oxide) and gate (metal) deposition
Fig 72 Schematic drawing of the construction of a self-organized nanowire-based transistor
Surface diffusion and decomposition of molecules can be sensitive to even smaller surface
features The nanoscopic lines formed on Si(100) are very regular and neat if their
nucleation and growth mechanism could be controlled ordered networks of nanowires
could be made on silicon
VPalermo 85
The techniques developed here are still experimental ones the careful control and
reproducibility of each step is still an issue and further studies will be needed to clearly
understand the dynamics of formation of these structures Nonetheless the results obtained
so far open up the possibilities of producing novel devices such as for example nano-
transistors (see scheme in fig73 where the mechanism described in fig 72 is applied for
the fabrication of a device array) The possibility of using the border of a lithographic
pattern to draw nanolines can in principle allow the creation of nanodevices using normal
IC manufacturing techniques
The results obtained and their description in this thesis are not the first examples of the use
of self-organization for the production of ordered structures and will certainly not be the
last In our opinion the most important outcome of these findings is that it is possible to
use surface diffusion to overcome the inherent limits of lithographic techniques in
microdevice production processes and to force properly selected molecules to react with
the surface in a spatially differentiated manner at particular sites thereby creating ordered
series of nanostructures The role of the surface dynamics of the substrate atoms
themselves is extremely important in this process
Outstanding results have been and continue to be obtained in nanoscience and
nanotechnology research promising the emergence of new production technologies Sooner
or later those emerging technologies will have to face the issues of cost and compatibility
with the enormous investment made in existing technologies and processing facilities The
commercial production of nanostructures based on surface diffusion processes could in
principle be cheap simple and compatible with existing technology Moreover the
processes of diffusion in bulk silicon is well-known in the microelectronics industry which
has both the know-how and the equipment to better understand and exploit the surface
diffusion processes on silicon surfaces studied here
Extremely important and innovative results are often obtained by exploiting very simple
ideas and in our opinion physical processes such as surface diffusion used to create
complex nanostructures on silicon surfaces is a prime example
86 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
2 Heating nanowire creation and
oxide removal
3 Source-drain deposition by conventional lithography
4 Insulating layer (oxide) and gate (metal) deposition
Fig 73 Schematic macroscopic outline of the construction of a self-organized nanowire-based transistor
VPalermo 87
Bibliography
[1] P Gargini Intel Technology Coordinator ldquoFrom Microelectronics To
Nanotechnologiesrdquo Invited Campus Colloquia Lecture CNR Research Area Bologna
February 6th 2003
[2] Toward A Hybrid Micro-Nanoelectronics Cerofolini Gf Ferla G Journal Of
Nanoparticle Research 4 (3) 185-191 Jun 2002
[3] Recent Progress In Field Emitter Array Development For High Performance
Applications Temple D Materials Science amp Engineering R-Reports 24 (5) 185-239
Jan 25 1999
[4] Silicon Dioxide Micropillars For Sieving Fabricated By Macroporous Silicon-Based
Micromachining Izuo S Ohji H French Pj Tsutsumi K Kimata M Sensors And
Materials 14 (5) 239-251 2002 Electrochemical Etching In HF Solution For Silicon
Micromachining Barillaro G Nannini A Piotto M Sensors And Actuators A-Physical
102 (1-2) 195-201 Dec 1 2002
88 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 89
Acknowledgements
This work would not have been possible without the help and guidance of Dr Derek Jones
he introduced me to the use of STM and other techniques found the funding for this PhD
and helped me with the thousands of small and big problems I had to face during my
research I am grateful to Prof Alberto Ripamonti too for being the supervisor of this
thesis Thanks are also due to Dr Giancarlo Seconi director of ISOF-CNR where most of
this work was carried out Financial support from the Italian National Research Council
(CNR) is also gratefully acknowledged
My passion for surface science began when Fabio Biscarini showed me that it was possible
to really see the atoms and for this I will always thank him
During these years I had the luck to collaborate with many good scientists among those I
would like to cite Enrichetta Susi Massimo Cocchi and Anna Mazzone from CNR
Claudio Zannoni Anna Cavallini Daniela Cavalcoli and Antonio Castaldini from the
University of Bologna Sergio Pizzini and Maurizio Acciarri from the University of Milan
One of the most stimulating periods of my PhD was the one I spent at the Steacie Institute
for Molecular Sciences in Ottawa working with Robert Wolkow Dan Wayner Greg
Lopinski and Peter Kruse I thank them for discussing together many exciting scientific
ideas and trying to put some of them into practice
Special thanks go to my colleagues Paolo Samorigrave and Stefano De Cesari for long and
useful discussions on Science Life and their interactions often with the help of a good
pint
Finally I would like to thank Vassilia Gaetano Simone Claudia Silvia Letizia
Alessandro Angela and Sebastien even though not directly involved in this work they
have been over these three years a continuous source of happiness and support
Bologna March 2003
90 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
SCIENTIFIC PUBLICATIONS bull Lateral diffusion of titanium disilicide as a way to contacting
hybrid Si-organic nanostructures Palermo V Buchanan M Bezinger A Wolkow RA APPLIED PHYSICS LETTERS 2002 v81 p 3636 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v6 issue 20) bull Surface modifications in Si after Rapid Thermal Annealing Castaldini A Cavalcoli D Cavallini A Jones D Palermo V Susi E JOURNAL OF THE ELECTROCHEMICAL SOCIETY 2002 v 149 pG633 bull Production of nanostructures of silicon on silicon by atomic self-
organisation observed by scanning tunnelling microscopy Jones D Palermo V APPLIED PHYSICS LETTERS 2002 v 80 p 673 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v5 issue 5) bull Nucleation of nanostructures from surface defects on silicon Palermo V Jones D SOLID STATE PHENOMENA 2002 v 82-84 p 687 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo MATERIALS SCIENCE AND ENGINEERING B 2002 v88 (2-3) p220 bull Morphological changes of Si[100] surface after treatment with
concentrated and dilute HF Palermo V Jones D MATERIALS SCIENCE IN SEMICONDUCTOR PROCESSING 2001 v 4 p 437 bull Electrical and structural properties of processed silicon surfaces Susi E Cavallini A Castaldini A Cavalcoli D Jones D Palermo V ldquoRECENT RESEARCH DEVELOPMENTS IN VACUUM SCIENCE amp TECHNOLOGYrdquo 2001 v 3 p189 bull Numerical Solutions of the Stochastic Equations of Crystal
Growth Mazzone A M Palermo V INTERNATIONAL JOURNAL OF MODERN PHYSICS C 2000 v 11 Part 1 p195-204 bull Advances in silicon surface characterisation using light beam
injection techniques
VPalermo 91
Acciarri M Pizzini S Simone G Jones D Palermo V MATERIALS SCIENCE AND ENGINEERING B 2000 V73 (1-3) p 235 - 239 bull Abrupt orientational changes for liquid crystals adsorbed on a
graphite surface Palermo V Biscarini F Zannoni C PHYSICAL REVIEW -SERIES E- 1998 V 57 NUMBER 3A p R2519-R2522 Oral presentations bull The use of oxide desorption and surface diffusion for the
creation of silicon on silicon nanostructures Palermo V Jones D 1ST NATIONAL WORKSHOP ON CURRENT TRENDS IN NANOTECHNOLOGIES Catania (Italy) Feb 2002 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo V E-MRS Spring Meeting Strasbourg (France) June 2001 bull STM study of surface transformations on silicon during UHV
annealing Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999 bull Advances in silicon surface characterisation using light beam
injection technique Acciarri M Pizzini S Simone G Jones D Palermo V E_MRS SPRING MEETING June 1999 Posters bull Spontaneous nano-wire growth on silicon Palermo V Jones D SMARTON workshop Leuven (Belgium) October 2002 bull Ordered circles of nano-islands on silicon from CO adsorption Palermo V Jones D TRENDS IN NANOTECHNOLOGY 2002 Santiago de Compostela (Spain) September 2002 bull Production of nanostructures of silicon on silicon by atomic self-
organisation Palermo V Jones D EUROMAT Rimini (Italy) June 2001 bull Etching holes and anisotropic corrosion on silicon [100] Palermo V Jones D Susi E Asoli B SILICON WORKSHOP Genova (Italy) February 2001
92 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull Morphological and electrical characteristics of damaged silicon surfaces
Susi E Castaldini A Cavalcoli D Cavallini A Jones D Palermo V SILICON WORKSHOP Genova (Italy) February 2001 bull Effect of HF etching on the roughness of a silicon surface Palermo V Jones D NATIONAL CONFERENCE ON PHYSICS OF MATTER Genova (Italy) June 2000 bull Effect of Native Oxide Desorption upon the Surface Morphology of Si[100] by STM and LEED Palermo V Jones D SILICON WORKSHOP Genova (Italy) February 2000 bull Characterisation of silicon surfaces for microelectronics through STM measurements Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999
VPalermo 93
Contact Vincenzo Palermo ISOF- Institute for Organic Synthesis and Photoreactivity Via Gobetti 101 40129 Bologna ITALY Tel +39-051-6398336 Fax +39-051-6398349 Mail palermoisofcnrit
94 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Index
1 Introduction nanotechnology and the future of computers
2 Silicon surfaces
3 STM and other surface analysis techniques
4 Surface modification of silicon in liquid Nano-hole creation
5 Surface modification of silicon in vacuum void creation and oxide desorption
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and
nano-lines
7 Conclusions and proposals for future work
VPalermo 5
OMNIA IN MENSURA ET NUMERO ET PONDERE
Sapientiae Salomonis 1120
6 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 7
1 Introduction nanotechnology and the future of computers The motivations of nanotechnology research
In 1965 an electronic engineer named Gordon Moore one of the future founders of Intel
noted that the performance of computers and their complexity doubled every 18 months
and foresaw that computer power would continue to grow exponentially over the following
years
This prevision quite provocative for its time actually came about and gained the name of
ldquoMoorersquos Lawrdquo and continues to hold for the trends of todayrsquos computer industry Since
1965 the number of transistors present in an integrated circuit (IC) has increased from
several hundred to more than ten million and the minimum size of transistor elements has
shrunk from several millimetres to asymp130 nanometres (fig 11)
Devices of such tiny dimensions are actually fabricated using lithographic techniques
where light is passed through an optical mask to react with a photo-sensitive layer (resist)
Fig11 Evolution of the number of transistors present on commercial computers [1]
8 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
on the silicon wafer This resist is then selectively removed and used as a mask for
processing the silicon surface (fig 12 left) The maximum resolution attainable depends
upon the wavelength used and current technology is near to its intrinsic resolution limit
On the other hand there is strong scientific and economic demand for further development
in IC miniaturization to obtain more powerful and complex computers Besides every-day
life applications more powerful computers are fundamental for much scientific research
such as climate change tracking genome sequencing and fluid dynamics Increased
miniaturization is also fundamental for reducing power which has to be dissipated by the
chips which run at progressively higher frequencies Energy consumption by
microelectronic devices is already an issue and represents one of the main obstacles for
the continuing growth in wireless communication (cell phones portable computers CD
and DVD players digital cameras etc)
Thus it is expected that new production methods different from current lithographic ones
will be developed methods which allow modification of a surface well below the 100 nm
limit and even down to single atom manipulation Techniques such as Scanning
Tunnelling Microscopy and Atomic Force Microscopy are already capable of moving
single atoms (see fig 12 right) but unfortunately building a working nanodevice in this
way would take a very long time and these techniques are difficult to apply to large scale
production
Nowadays thousands of researchers are working in the nanotechnology field towards a
new generation of microelectronic devices Several possible solutions are competing for
tomorrowrsquos computer architecture and there is still no clear winner It is likely that the
final solution will be the combined use of different techniques and components (including
molecules nanowires and nanodots) as they become available together with conventional
Fig 12 Left conventional litographic process [2] Right atomic manipulation of iron atoms on copper [3]
VPalermo 9
silicon technology
Below is a brief summary of the most recent developments in nanotechnology and
nanoscience
Actual trends in nanotechnology
Perhaps the most fascinating idea for nanodevice construction is to use one single
molecule working as a complete device The first molecular diodes (ie molecules
conducting current only in one direction) were created in 1997 in 1999 a molecular fuse
and a molecular transistor were demonstrated although there was no possibility of wiring
these devices to external contacts In April 2001 James Heath and his group at UCLA
fabricated an array of overlapping crossbars and placed a small molecule of rotaxane
between each crossbar (fig 13 left) This composite molecule is made up of two
component parts the main rod-like molecular axis and a mobile ring ldquothreadedrdquo on it like a
bead on a necklace and can function as a molecular switch A working 16-bit memory
circuit was constructed using these molecules For a brief review of these works see [4] In
June 2002 a single molecule transistor was built by connecting an organic molecule to two
metal contacts the molecule contained one or two atoms of a transition metal (cobalt or
vanadium) forming the active region of the device supported by an organic backbone [5]
Fig13 Working nanodevices Left schematic representation of rotaxane molecules between crossed nanowires [4] Right SEM image of semiconductor nanowires forming a small circuit [9]
10 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another approach to nanodevice fabrication has become possible through the discovery of
carbon nanotubes which were observed for the first time in 1991 by a Japanese electron
microscopist studying the material deposited during arc-evaporation synthesis of fullerenes
[6] They consist of a graphite-like carbon seamless cylinder with a diameter of several
nanometers and lengths of up to a millimetre Carbon nanotubes are very stable can
behave as metals or semiconductors and can host other molecular or ionic species thus
modifying their electrical behaviour In 2001 Avouris and his group reported the first
circuit made with a single nanotube [7] A few months later Cees Dekker presented a
nanotube-based transistor able to amplify an input signal by a factor of ten and built
several logic circuits using these nanotube transistors [8]
One problem with carbon nanotubes is that it is very difficult to control their electronic
properties ie their metallic or semiconducting behaviour An alternative to carbon
nanotubes are semiconductor nanowires Silicon nanowires can be made using a laser to
vaporize the silicon together with a metal catalyst like iron or gold The vapour condenses
in nanosized drops of silicon and metal from which the wires slowly grow out as more
silicon is adsorbed In 2001 a group at Harvard University [9] created a transistor by
crossing two different nanowires After this the same group arranged four nanowires in a
noughts and crosses grid creating something like a 4-bit memory (fig 13 right) Even
metallic nanowires made of platinum and silver can be used in a crossed configuration to
store information [10]
There are some issues common to all these new technologies though First it is difficult to
imagine these methods applied to large-scale production Up to now the insertion of a
molecule between two electrodes is an occasional lucky event while nanotubes and wires
have to be positioned on the surface creating the appropriate contacts on them manually
The large-scale production of integrated circuits using these building blocks will not be
straightforward and does not seem imminent
Another issue is of an economic and not a scientific or technological nature Since 1965
the cost of IC manufacturing plants has sky-rocketed If the cost of semiconductor
production plants continues to rise exponentially in a few years such plants will cost up to
$20 billion This is a sizeable investment even for large companies such as IBM or Intel
For this reason it is likely that IC companies will resist changing to completely new
technologies closing down their existing plants As it is clear that silicon will remain the
VPalermo 11
fundamental raw material of the IC industry for the foreseeable future nanotechnology
developments for microelectronics will need to be silicon-compatible In these early days
of nanotechnology the most valid approach would seem to be the addition of molecular
functions to existing silicon technology ndash using the latter as a foundation on which to build
Fabrication of self-organised structures on silicon
The possibility of using the phenomenon of atomic or molecular self-organization to create
nanostructures on silicon has already been demonstrated The clean silicon surface shows
in some cases a high degree of order and complex surface reconstruction as will be
described later Several different ordered structures form spontaneously on this surface
such as series of monatomic steps or boundaries between reconstructed areas It has been
demonstrated that it is possible to use these structures to fabricate ordered nanodots and
nanolines on the surface [11] More recently well-defined nanometric patterns have been
obtained with selective etching of silicon using nitric oxide [12]
In this study the possibility of creating different types of nanostructures on the silicon
surface is explored Methods had to be developed which were
- Simple They must not need complex masks or lithographic steps to create the
structure but rather exploit self-organisation phenomena
- High resolution the silicon surface modifications should be on a scalelength of below
100 nm
- Fast billions of nanostructures have to form over the whole surface simultaneously to
be compatible with large-scale production requirements
- Cheap they must not require expensive equipment (such as e-beam lithography) but
exploit simple chemical andor physical treatments to produce nanostructures on the
silicon surface
In Chapter 2 the main characteristics of silicon are described Chapter 3 provides a
summary of the techniques used for this research Chapter 4 examines the chemical etching
of silicon in different liquid environments and the effects of this etching on the surface at a
nanoscopic level with the creation of nanoholes
12 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Chapter 5 describes the growth in ultra-high vacuum (UHV) of nanoscopic voids and
islands on the silicon surface and the effect of surface oxide on this growth
Chapter 6 discusses the modification of silicon surfaces in UHV following the adsorption
of molecules and thermal heating to produce nanoislands and nanolines on silicon
The overall conclusions of our work are summarised in Chapter 7
Finally we will give some conclusions based on the results obtained and discuss possible
applications of the methods developed
Bibliography
[1] From wwwintelcom
[2] From wwwsematechorg
[3] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
[4] Molecules Get Wired Service RF Science 294 (5551) 2442-2443 Dec 21 2001
[5] Coulomb Blockade And The Kondo Effect In Single-Atom Transistors Park J
Pasupathy AN Goldsmith JI Chang C Yaish Y Petta JR Rinkoski M Sethna JP
Abruna HD Mceuen PL Ralph DC Nature 417 (6890) 722-725 Jun 13 2002 Kondo
Resonance In A Single-Molecule Transistor Liang WJ Shores MP Bockrath M Long
JR Park H Nature 417 (6890) 725-729 Jun 13 2002 Nanotechnology - Electronics
And The Single Atom De Franceschi S Kouwenhoven L Nature 417 (6890) 701-702
Jun 13 2002
[6] Smallest Carbon Nanotube Ajayan PM Ijima S Nature 358 (6381) 23-23 Jul 2 1992
[7] Carbon Nanotube Inter- And Intramolecular Logic Gates Derycke V Martel R
Appenzeller J Nano Letters 1 (9) 453-456 Sep 2001
[8] Logic Circuits With Carbon Nanotube Transistors Bachtold A Hadley P Nakanishi T
Dekker C Science 294 (5545) 1317-1320 Nov 2001
VPalermo 13
[9] Logic Gates And Computation From Assembled Nanowire Building Blocks Huang Y
Duan XF Cui Y Lauhon LJ Kim Kh Lieber CM Science 294 (5545) 1313-1317
Nov 9 2001
[10] Formation And Disappearance Of A Nanoscale Silver Cluster Realized By Solid
Electrochemical Reaction Terabe K Nakayama T Hasegawa T Aono M Journal Of
Applied Physics 91 (12) 10110-10114 Jun 15 2002
[11] Fabrication And Integration Of Nanostructures On Si Surfaces Ogino T Hibino H
Homma Y Kobayashi Y Prabhakaran K Sumitomo K Omi H Accounts Of Chemical
Research 32 (5) 447-454 May 1999
[12] Ultrafine And Well-Defined Patterns On Silicon Through Reaction Selectivity
Prabhakaran K Hibino H Ogino T Advanced Materials 14 (19) 1418-1421 Oct 2
2002
14 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 15
2 Silicon surfaces
The name silicon (silicio in Italian) comes from the latin word silex Amorphous silicon
was first isolated by Berzelius in 1824 by reaction of potassium with silicon tetrafluoride
Thirty years later the first crystalline silicon was prepared Silicon makes up 25 of
earthrsquos crust and is the second most abundant element after oxygen Elemental silicon is
not found in nature occurring as silicon oxide (sand quartz amethyst flint etc) or
silicates (asbestos clay mica etc) Perhaps no other element and its compounds has such
a wide range of uses Silicon compounds such as sand and clay are used in the building
industry as refractory materials for high-temperature applications and for enamels and
pottery Silica is the main component of glass silicon carbide is an important abrasive and
silicones are commonly used polymers and lubricants
Here the most interesting use of silicon of course is for the production of
microelectronics devices For this application silicon of high purity (999999) and of
high crystallinity is needed Table 21 lists some of the physical characteristics of silicon
High purity polycrystalline silicon is produced by the reaction of gaseous trichlorosilane
with hydrogen in a furnace Then to prepare a single-crystal of silicon the so-called
Czochralski method is commonly used
Polycrystalline silicon is melted in a quartz furnace at 1415degC in an argon atmosphere
Then a seed of single-crystal silicon is lowered into contact with the melt and slowly
pulled out In this way the crystal grows and a crystalline cylindrical ingot several metres
long is created from the initial seed
After cooling down the ingot is sliced into thin silicon wafers The wafer surfaces are
polished using a counter-rotating lapping machine in an Al2O3 slurry until the surface is
very flat and shiny ready for the lithographic processes
16 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another way to obtain single crystal silicon is the Floating Zone (FZ) method in which a
silicon cylinder is slowly passed through a heating ring The area inside the ring melts and
solidifies smoothly crystallising as it comes out of the ring yielding a single silicon crystal
Microelectronic devices are built on the silicon surface which is the surface of interest
here Unfortunately silicon surfaces are normally quite dirty and uneven at the atomic
scale Atmospheric oxygen and humidity react with silicon surfaces creating a thin layer of
oxide (called ldquonative oxiderdquo) which is usually irregular and full of defects Different kinds
of contaminants also adsorb onto the surface These are usually small organic molecules
and microscopic dust particles A clean surface on exposure to the atmosphere is
completely covered with gas molecules in less than 10-9 seconds If the pressure is
reduced letrsquos say to 10-6 mbar this time increases to 1 second This is the reason why to
study a clean surface we have to work in UHV at pressures below 10-10 mbar
The atoms in the silicon crystal have a diamond-like structure each atom having 4 bonds
in a tetrahedral sp3 arrangement with bond angles of 10947 degrees At the crystal
surface some atoms will have non-bonding orbitals ldquodanglingrdquo in the vacuum ie sp3
orbitals with a lone electron which are highly reactive These orbitals are known as
dangling bonds To minimize surface energy the surface will reorganize by decreasing
the number of dangling bonds
Table 21 Physical data of silicon [1]
Atomic Weight 2809 Lattice constant (A) 543095 Crystal structure Face-centered cubic
(diamond) Melting point 1415 degC
Density (gcm3) 2328 Boiling point 2355degC Atomscm3 50E22 Minority carrier
lifetime (s) 25E-3
Dielectric Constant 119 Specific heat (Jg degC)
07
Breakdown field (Vcm)
~3E5 Thermal conductivity (Wcm degC)
15
Electron affinity x(V)
405 Vapour pressure (Pa) 1 at 1650degC 1E-6 at 900deg C
Energy gap (eV) at 300K
112 Reactivity Inert to acids Attacked by halogens and alkaline
solutions Intrinsic carrier
conc (cm-3) 145E10 Oxidation states +4 -4
Intrinsic Debye Length (microm)
24 Energy of a Si-Si bond (eV)
232
Intrinsic resistivity (Ω-cm)
23E5
VPalermo 17
Dangling bond densities and positions and thus the type of surface reconstruction will
depend upon crystal orientation as well as the temperature and kinetics of the system
Fig 21 shows a drawing of the main faces of a silicon crystal The angle α between any
(11n) face and the (100) face can be calculated from 2cos 2 += nnα The angle
between any (11n) face and the (111) face can be calculated from
)2(3)2(cos 2 ++= nnα
The chemistry and physics of the faces are very different a brief description will be given
for the most important orientations
Si (100)
On the (100) surface each atom has two Si-Si bonds connecting it to the bulk and two
dangling bonds pointing outward Surface energy is reduced by the dimerisation of the
surface atoms through overlap interaction of one dangling bond per atom forming rows of
dimers aligned along the (110) direction This is the well-known ldquo2x1rdquo reconstruction of
this silicon surface The symmetric dimers would make the silicon surface metallic but to
reduce surface stress the dimers tend to buckle and the surface is thus semiconductive It
took several years to understand that the dimers are buckled because at room temperature
Fig 21 Schematic view of the principal orientations of a silicon surface Surface atoms are white bulk atoms are black dangling bonds are gray [2]
18 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
they shift easily from one buckling direction to the other thus appearing symmetric under
STM observation Fig 22 shows an STM image of the 2x1 reconstructed surface
Even almost perfect (100) surfaces have a certain number of monoatomic steps and the
dimer rows on atomic layers are aligned at 90deg to those on adjacent layers Dimer rows are
thus perpendicular or parallel to the step When the dimers on the upper side of the step
are parallel to the step the step is called SA if they are perpendicular the step is called SB
Because of this symmetry SA step edges will be smoother compared to the more broken
fragmented edges of the SB steps
A common defect on the Si(100) surface is the presence of nickel contamination which
appear as missing dimers in STM images This type of contamination is so critical that
even if the silicon sample is only briefly brought into contact with stainless steel tools
(tweezers for example) the 2x1 reconstruction of the surface can be blocked
Silicon atoms can diffuse easily over the silicon surface as monomers and dimers
especially at elevated temperatures The anisotropy due to the 2x1 reconstruction causes a
difference in the diffusion energies of adsorbates over the surface Diffusion of these
silicon species along dimer rows for example will be much easier A list of diffusion
energies for monomers and dimers is given below [2]
Diffusion on Si(100) 2x1 Ed (eV) Monomers along dimer rows 06 Monomers across dimer rows 085 Monomer formation energy 18 Dimer along dimer rows 11 Dimer across dimer rows 15 Dimer formation energy 26 Dimer binding energy 076 Vacancies along dimer rows 17 Vacancies across dimer rows 19
SA
SB
Fig 22 STM image of a 2x1reconstructed silicon surfaceshowing the dimer rows and steps40x35 nm Nickel-induced defectsare visible as dark spots SA and SBsteps are indicated
VPalermo 19
So the diffusion energy for both monomers and dimers is nearly 40 greater if they have
to cross a dimer row This difference reduces to sim10 for vacancy diffusion
Si(113)
The (113) surface can be imaged as a sequence of alternating (100) and (111)-like
structures with two and one dangling bonds on alternate atoms respectively Interest in the
(113) surface is scientific as it has been used to study the energetics of the (100) and (111)
surfaces as well as for surface adsorption experiments
Si(100) surfaces can easily develop (113) facets
Si(111)
This surface besides being the first one imaged with STM with atomic resolution is one of
the most studied because it is the best cleavage face of silicon and because it shows one
of the most complex and elegant reconstructions in surface science
All Si-Si bonds in the silicon crystal are perpendicular to a (111) plane so this face will
have the lowest number of dangling bonds created per unit area In fact each Si atom on a
(111) surface shows a single dangling bond oriented perpendicular to the surface and
bonded to three back atoms These three bonds for each surface atom account for the great
chemical and physical stability of the Si(111) surface Surface energy is 009 eV Aring-2
compared to 015 eV Aring-2 for Si(100)
For energy minimization this surface reconstructs forming a huge 7x7 lattice cell
containing 102 atoms described by the Dimer-Adatom-Stacking fault model (DAS) For a
detailed description of cell structure see fig 23
The cell described by this model is very complex being composed of three kinds of atoms
adatoms rest atoms and corner hole atoms Furthermore a subsurface stacking fault is
present in one half of the cell making the two halves of the unit cell look different under
STM (Fig 24) It took 26 years of research to completely understand the exact structure of
the 7x7 reconstruction
Cleaving a silicon crystal along a (111) plane produces a metastable 2x1 reconstruction
the 7x7 reconstruction is easily obtained by flashing at high temperature in UHV At T gt
20 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
830degC a disordered 1x1 phase covers the surface Cooling down to 800degC leads to the
formation of the 7x7 phase If the cooling process is too rapid small 7x7 domains nucleate
and a disordered 1x1 phase is preserved between domain boundaries
Si(110)
Even though as mentioned before the (111) plane is the favoured cleaving plane of
silicon thin (100) commercial wafers will not break along this plane because the angle
between (100) and (111) is too far from 90deg (see table 22) Instead they will break along
the (110) plane because it is perpendicular to the (100) surface Each surface atom on
Si(110) has a Si-Si bond pointing downward one dangling bond pointing outward and two
Si-Si bonds parallel to the surface in a zig-zag pattern (see fig 21) Cleaved (110)
surfaces are disordered but upon annealing at high temperatures an ordered complex 16x2
Fig23 Scheme of the 7x7 DAS model [2] In each unit cell there are 9 dimers 12 adatoms and a stacking layer fault The force driving this complex reconstruction is the minimization of dangling bonddensity The DAS model shows the lowest number of dangling bonds (19) of all possiblereconstructions 12 dangling bonds are at the adatoms 6 at the rest atoms and 1 at the corner hole atom This surface is metallic
VPalermo 21
reconstruction takes place The surface appears as a series of long ridges and valleys
parallel to each other Eventually tilted facets of orientation (17 15 1) can form on this
surface The adsorption of Ge atoms on this surface leads to the formation of self-
assembled nanowires [3]
Table 22 Angles in degrees between different silicon faces [2]
Orientation
113
110
111
100
100
2524
9000
5474
0
111
2950
3526
0
110
6476
0
113
0
Fig 24 STM image of a Si(111)surface with 7x7 reconstruction A unitcell with its adatoms is highlightedImage size 13x13 nm
22 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] Weast RC Handbook Of Chemistry And Physics (Chemical Rubber Co Cleveland
1972)
[2] Dabrowski J Mussig H Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] The Structure Of Clean And SiGe-Covered Si(110) Surfaces Butz R Luth H Surface
Science 365 (3) 807-816 Oct 1 1996
VPalermo 23
3 STM and other surface analysis
techniques
Scanning Tunneling Microscopy
Since the invention of the optical microscope at the end of the 16th century the possibility
of examining surfaces at higher and higher magnification has fascinated mankind
Development of the technique continued and towards the end of the 19th century optical
microscopes were as good as todays standard instruments The physical limits of the
wavelength of visible light (350-800nm) had been reached
In the 1920s de Broglie showed that electrons can behave like waves and the use of these
particles for imaging with much higher resolution soon followed Atomic resolution using
this technique is only possible in the transmission mode with extremely carefully prepared
samples
In 1982 using the peculiar properties of piezoelectric materials Binnig and Rohrer brought
a metallic tip very very close to a silicon surface and scanned it across an extremly small
area (fig31) The tunneling of electrons from the tip into the sample or vice versa allowed
them to obtain a local density of electronic states (LDOS) map of that surface Although
theory (which treated the extreme point of the tip as a sphere) then excluded the possibility
of atomic resolution following a tip crash into the surface Binnig and Rohrer began to
observe the LDOS with atomic resolution For this discovery and their following work
they were awarded the Nobel Prize in Physics in 1986
24 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The basic principle of STM is very simple A metallic tip is scanned over a surface without
making ohmic contact and a tunneling current passes between the tip and the surface An
electronic circuit keeps this current constant by raising and lowering the tip during the
scan In this way recording the tip height at each point a three-dimensional image of local
density of electronic states (LDOS) of the surface can be obtained To explain the
extremely high resolution attainable by this simple technique quantum theory is needed
According to classical physics the current will flow between sample and tip only if they
are in physical contact If there is a vacuum gap between the two the electrons will simply
remain confined for example within the surface without the possibility of passing into the
tip
In quantum physics however the electrons have a certain probability of passing
(tunneling) across the gap appearing on the other side of the gap in this way reaching the
tip It can be shown that the probability of an electron tunneling through a gap of thickness
z is
kzep 22)0( minusprop ψ h
φmk
2= (1)
where ψ(0) is the electron wavefunction at the surface-gap border m is the electron
mass=91x10-28 g and φ is the work function of the metal (ie the energy required to
remove an electron from that material For silicon it is 48 eV) The tunneling current thus
decays exponentially with z and is extremely sensitive to topographical imperfections
present on the scanned surface A rough formula giving the current as a function of z is [1]
zFS eEVI φρ 0251)( minusprop
Fig 31 Binnig and Rohrer with the first STM Image from IBM [2]
VPalermo 25
where ρs(EF) is the local density of states at the Fermi level on the given surface For
example the formula predicts that for silicon an increase in tip-surface distance of 1 Aring
will give a 95 decrease in tunneling current
This huge dependence of tunneling current upon the distance allows detection even of the
sub-nanometre changes in height given by the single atoms of which the surface is
composed and thus to resolve them in the LDOS images Of course this description of the
tunneling process is oversimplified and for a more accurate one the electronic states of
the tip of the sample and their interaction have to be taken into account Fig 32 shows a
schematic representation of the interaction between tip and sample orbitals
The exponential decay of current with distance also yields high lateral resolution If the tip
is approximated as a sphere of radius R and the current passing at the minimum tip-sample
distance is I0 then the current passing at a lateral distance x from this point will be
Rxk
eII 22
0
2minus
=
Assuming a tip radius of 100 nm the current is concentrated in an area sim15 nm wide at tip
apex
Very sharp tips with even smaller curvature radii can be produced in several ways Simple
Fig32 Schematic view of tip-sampleorbitals interaction a) no interaction b)equilibrium c) sample positive d) tippositive [1]
26 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
tungsten wires mechanically cut are capable of obtaining atomic resolution on graphite in
air but for more disordered and rougher samples sharper and more reproducible tips are
needed
STM tips are mostly made by electrochemically etching a W or Pt-Ir wire The tips we
used were prepared using methods based on the work of Fotino [3]
A tungsten wire 038 nm diameter is immersed in a KOH solution(06M) with a thicker
tungsten wire used as a counterelectrode The cathodic and anodic reactions involved in the
etching are
Cathode 6H2O + 6 e- rarr 3H2(g) + 6 OH-
Anode W(s) + 8 OH- rarr WO42- + 4H2O + 6 e-
A potential of 3V ac is applied to the tungsten and the wire is immersed in the solution
until a constant current of sim100 mA is established The etching rapidly removes metal
shaping the wire end as a sharp tip When the potential reaches 11 V the coarse tip etching
is finished The wire is then removed from the solution carefully inserted into an
insulating plastic tube leaving only the tip exposed and re-immersed in the solution with
the tip pointing upwards A more gentle etching is thus made to reduce tip radius Usually
5 minutes etching at 07 V ac is used In this configuration very small hydrogen bubbles
formed on the tip sides sliding upwards with a ldquohoningrdquo effect on the tip
This procedure yields extremely sharp and reproducible tips at the microscopic level
After the etching the tip is thoroughly rinsed in ultrapure water then dipped into
concentrated HF to remove surface oxides and hydroxides [4] The tip is dried with
nitrogen inserted into the UHV system and degassed overnight at sim150degC
The possibility of measuring sub-nanometric distances would be useless without being able
to control tip movement over such a minute scale To scan the tip over the surface a
piezoelectric scanner is used Piezos are usually made of an alloy of PbZrO3 and PbTiO3 a
material which contracts or expands when a voltage is applied to it The Omicron
instrument used in our laboratory has three such piezo scanners for xy and z tip motion
allowing one to scan the tip over the surface with sub-Aringngstrom precision (fig 33)
To isolate the instrument from ambient vibrations the whole STM stage is suspended upon
four springs which eliminate all frequencies above 1 Hz and surrounded by a crown of
VPalermo 27
copper wings and fixed magnets Parasitic currents generated by the magnets into the
copper wings contrast every movement of the stage and efficiently block stage vibrations
The STM can be used not only to explore surface topography but to measure the IV
characteristics of single atoms or molecules on the surface (Scanning Tunneling
Spectroscopy or STS) Furthermore it can be used to modify the surface with voltage
bursts digging into it or delicately moving single molecules or atoms over a surface [5] It
can be used in vacuum in air and with proper lateral isolation of the tip even in liquids
A major drawback of STM is that it works only on conducting and semiconducting
surfaces and thus cannot be used on many surfaces of biological and chemical interest
Another instrument more suitable for these and other applications is the Atomic Force
Microscope (AFM)
Atomic Force Microscopy
AFM was invented in 1986 by Binnig Quate and Gerber after calculating the possibility of
building a cantilever with a force constant of the same order of magnitude as that of a
chemical bond
In AFM a tip mounted on a microscopic cantilever (usually made of Si3N4 fabricated with
optical lithography) is brought close to a surface When the tip touches the surface the
cantilever is very slightly deflected upwards The movement is measured by observing the
Y-PIEZO
SILICON SAMPLE
Fig 33 A picture of theSTM used for theexperiments The tripodpiezo scanner is shown
Z-PIEZO
X-PIEZO
TIP
28 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
deviation of a laser beam hitting the upper face of the cantilever Fig 34 provides a
schematic view of the principle of AFM
The typical force constant of the cantilever varies from 00006 to 2 Nm the typical
resonance frequency is 3 to 120 kHz The AFM tip can apply a force on the sample of up
to 10-9 N The AFM can be used on conductive or insulating surfaces in vacuum air or
liquids Furthermore the tip can be modified to sense electrostatic potentials (electric force
microscopy) or magnetic fields (magnetic force microscopy) it can even be functionalized
with complex molecules such as proteins to interact with biological surfaces
A drawback of the AFM is that the force it exerts can damage the surface under
observation especially if the sample is soft (as in the case of cell membranes for
example) This problem can be overcome using the instrument in tapping mode (where the
tip does not move laterally during its brief contact with the surface) or in non-contact mode
in which the tip oscillates above the surface during the scan and the changes in its
frequency due to interaction with surface are monitored The shifts in the oscillating
frequency of the cantilever due to tip-sample interaction are then used for imaging the
surface In this mode interaction of the tip with the surface is minimal and soft samples
can be imaged
STM and AFM are the main techniques used for this work A brief description of other
techniques used occasionally is given below
Fig 34 Scheme of an atomic force microscope
VPalermo 29
Low Energy Electron Diffraction (LEED)
Electrons with energies in the 20-500 eV range are diffracted by a crystalline surface the
diffraction peaks are visualized on a fluorescent screen This technique probes the long
range order of the surface up to a depth of several nanometres
X-Ray Photoelectron Spectroscopy (XPS)
XPS allows both qualitative and quantitative chemical analysis of the elements present on
or near the sample surface
An X-ray source is used to photoionize the atoms on a surface and produce photoelectrons
By measuring the kinetic energy of the photoelectrons the binding energy of the electronic
levels can be calculated This energy will depend on the chemical environment of the
surface atoms
Although the soft X-rays used penetrate to a depth of ~2000Aring the sampling depth of the
technique is determined by the mean free path of the photoelectrons which allows their
escape from only the first 10-100Aring
Secondary Ion Mass Spectroscopy (SIMS)
High and low energy ions (primary ions) are used to bombard a sample and remove surface
atoms and ions The ionic fragments removed (secondary ions) are then analysed by a mass
spectrometer The surface can be consumed during the measurement and profiles obtained
giving concentrations of the materials composing the sample at different depths (depth
profiles)
A popular variant of SIMS is TOF-SIMS In this technique the secondary ion masses are
measured by a time-of-flight (TOF) measurement The secondary ions generated by the
bombarding primary ions are accelerated to a constant kinetic energy and then move
through a field-free space before they reach the detector where their intensity is measured
as a function of flight time Since ions with different masses have different velocities at a
30 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
given kinetic energy the measured flight times of the ions can easily be converted to their
masses The static nature of this latter technique allows mass spectroscopy surface analysis
with minimal damage to the surface
Bibliography
[1] Chen CJ Introduction To Scanning Tunneling Microscopy (Oxford University Press
Oxford 1993)
[2] From wwwibmcom
[3] Tip Sharpening By Normal And Reverse Electrochemical Etching Fotino M Review
Of Scientific Instruments 64 (1) 159-167 Jan 1993
[4] A Convenient Method For Removing Surface Oxides From Tungsten STM Tips
Hockett LA Creager SE Review Of Scientific Instruments 64 (1) 263-264 Jan 1993
[5] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
VPalermo 31
4 Surface modification of silicon in liquid
Nano-hole creation
Liquid treatments of silicon wafers are very common in the integrated circuit (IC)
manufacturing industry They are used to clean and improve surface uniformity to create
and etch protective oxide layers and to remove photo-resist layers
Crystalline silicon with its native oxide layer is very stable and is resistant to many acids
It is easily attacked by hydrofluoric acid (HF) and alkaline solutions
The thin (~2 nm) passivating layer of native oxide (SiO2) is formed on exposure to the
atmosphere This surface layer contains many defects and contaminants so it is usually
chemically stripped and substituted with a better chemically-formed protective oxide
The most common silicon cleaning procedure is the RCA method named after the Radio
Corporation of America [1] It consists of two steps in the first one the surface is treated
with a hot alkaline solution (H2OH2O2NH4OH 411) to remove particles from the
surface following this a hot acidic solution (H2OH2O2HCl 411) is used to remove
metal contamination Other well-known cleaning methods are IMEC (a sequence of
cleaning steps in H2OO3 and dilute HF) or the Pirana etch (a hot 41 mixture of H2SO4
H2O2)
32 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The standard RCA clean removes surface contaminants etches the native oxide and
oxidizes the silicon surface leaving a uniform layer of silicon oxide which better protects
the surface from further contamination
Etching with fluorine-based solutions
Hydrofluoric acid is one of the most common reagents used in the treatment of silicon
wafers both in the research field and in industrial processes A rapid dip in dilute HF is the
simplest way to remove the native oxide from Si(100) and leaves the surface passivated by
a layer of Si-H bonds Because of the low polarization of Si-H bonds the Si-H layer is
stable even for several days protecting the surface from contamination It has often been
assumed that this short etch does not significantly change the surface morphology of the
silicon substrate[2] even though a prolonged dip in dilute HF leads to surface roughening
[3]
Although dilute HF roughens the Si(100) surface at the atomic scale [45] immersion in
concentrated HF (49) etches the surface oxide without attacking the Si surface
uncovering in this way the buried SiSiO2 interface The final counter-intuitive result is
that dilute HF etches the silicon while concentrated HF leaves the crystalline silicon
untouched [3]
Etching Si with fluorine-containing solutions at different concentrations and pH can
produce different morphologies from rough surfaces to flat nearly ideal Si-H terminated
surfaces
Hessel et al and Higashi et al demonstrated in 1991 that very flat Si(111) surfaces can be
obtained using 40 NH4F while etching with HF always yields rough surfaces The
surface becomes smoother because the etchant rapidly attacks Si atoms at step borders
thus removing surface kinks and irregularities in a step-flow mechanism [6 7] Later on
even smoother and more perfect surfaces were obtained by removing oxygen from the
solution after it was discovered that oxygen dissolved in 40 NH4F initiates the formation
of triangular etch pits It was not possible to obtain flat surfaces by etching Si(100) with
ammonium fluoride solutions which leads to the formation of small 2x1 dimer-row
reconstructed (100) terraces together with (111) facets [8]
This difference is caused by the different hydride terminations prevailing on the (100) and
(111) faces While the ideal Si(111)-H surface is monohydride terminated the more
VPalermo 33
reactive dihydrides predominate on the Si(100)-H surface making it more vulnerable to
etching The etching reaction is thus strongly anisotropic etching (100) facets faster than
(111) thus producing (111) microfaceting on Si(100) crystals
A more uniform Si(100) surface can however be prepared by etching at low pH with an
HFHCl mixture [9] or by using very dilute HF solutions and ultrapure water with low
dissolved oxygen and carbon contents [10]
Electrochemical etching can also be used applying anodic or cathodic bias to the silicon
to obtain different morphologies [11] by varying the potential isotropic or anisotropic
etching is observed The aforementioned results show that despite the simplicity of the
reactants fluoride etching of silicon is quite a complex reaction
Fig 41 Chemical etching of silicon
HO+H
H HO
H
H
H
FSi
Si
SiSi
+H2O
-OH-
+F-
H2O H H
Si H
F OH
H Si
Si Si
-H2
+H2O
H
H
H
FSi
Si
SiSi
H
H
H
OHSi
Si
SiSi
H
H
H
H Si
Si
Si Si
Etching mechanism of silicon
HF rapidly dissolves the SiO2 passivating layer on silicon leaving the surface almost
completely hydrogenated [12] After this two different types of reactions etch the silicon
simultaneously one chemical and the other electrochemical [13] The overall etching
mechanism can be schematized in two stages (see Figure 41)
i) Si-H bonds are substituted by Si-F or Si-OH bonds creating a partial charge on the
surface silicon atom and polarizing its Si-Si backbonds
ii) these polarised backbonds are then more easily attacked by HF or H2O After
rupture of the Si-Si bond the atom is removed leaving behind new Si-H
terminations and the reaction can start again
34 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
These reactions take place although at different rates on both Si(100) and Si(111)
Stage i) is usually the rate-determining step of the reaction and the stability of the Si-H
bonds depends upon the pH the concentration of nucleophilic species in solution and an
eventual potential applied to the crystal
For pH gt5 as in the case of concentrated NH4F solutions reaction begins with attack by
water to give Si-OH (step ArarrC) The -OH group is rapidly substituted by fluorine with
polarization of the underlying Si-Si bonds These bonds are then easily attacked by water
the silicon atom being released into solution as HSiF(OH)2 The Si-OH rarr Si-F substitution
is not fundamental for the reaction and etching can proceed even for Si-OH terminated
atoms but XPS measurements showed the presence of a certain number of Si-F bonds
remaining Furthermore fluorine seems to have a catalytic effect on Si-H substitution as
indicated by the dependence of the etch rate upon the F- concentration at least for pH
values between 4 and 8
Si-F bonds can be easily removed by a water rinse In the case of strongly alkaline
solutions (pH=14) OH- groups act directly as nucleophiles and no fluorine is needed to
catalyze Si-H bond rupture
At pH lt4 almost no free F- ions are present in solution and the etch rate is very small at
pH lt2 all etching reactions are very slow and this explains the stability of Si crystals in
concentrated (50 ww) HF solutions
Si Si
Si Si
H
H
H
H Si
Si
Si Si
H
H
H Si
Si
SiSi
H
H
H
OH-H+ -e
(F-)
-H+
-e
+H2O hellipas in fig1
Fig 42 Electrochemical etching of silicon by H2O
The electrochemical etching of silicon involves electron transfer from the surface atoms to
the valence or conduction band of the crystal (according to the type of doping of the
silicon) Several studies have been carried out at different electric potentials In the case of
p-silicon or n-silicon at anodic potentials where positive charges (h+) are available in the
crystal a silicon-centred radical can be created by capture of a hole from the bulk (fig
42) The F- ion does not participate in the reaction but is thought to contribute through
VPalermo 35
electrostatic interactions by lowering the energy of the interaction step After the
formation of the Si-OH group the reaction proceeds as shown in the scheme of fig 41
Matsumura et al [4] proposed that not only water but HF2- molecules also play a major role
in electrochemical etching of silicon leaving on the surface Si-F terminated bonds which
can be immediately attacked in an autocatalytic process (fig 43)
In the electrochemical reactions described above an external potential is applied to the
silicon crystal The chemical and electrochemical reactions in any case take place
simultaneously most of the time with the chemical path predominating at high pH Even
when no external potential is applied to the silicon partial electrochemical reactions can
take place at different ldquocathodicrdquo and ldquoanodicrdquo sites on the surface with an internal charge
exchange which ensures neutrality [11] This macroscopic silicon etching and hydrogen
bubble formation can lead to visually observable patterns on the surface when Si(100) is
immersed in ammonium fluoride even without applying a potential
Si Si
Si Si
H
H
H
F Si
Si
Si SiH
F F-H+ -2e
+HF2-
H
HF
SiSi
SiSi + F
F F
FH
H
H
F -H+ -2e
+HF2-
Fig 43 Autocatalytic electrochemical etching of silicon by HF2-
Inhomogeneities on silicon surfaces caused by electrochemical reactions and charge
transfer have been studied extensively because they are of fundamental importance in the
formation of porous silicon
Pore formation on silicon
When Si(100) or Si(111) are etched under anodic bias in fluorine-based solutions
microscopic pores form on their surface Several different morphologies of pores have
been observed with pore diameters ranging from 10 nm to several microns with depths of
several microns [14] Pore shape is very variable too ranging from ordered straight pores
to chaotic networks of branched pores (fig 44) Porous silicon has been known since the
fifties but it was only in 1990 that interest in this material increased following the
discovery that porous silicon layers were able to emit bright red light This led to a large
amount of research and now different classes of micropores can be reproducibly created
36 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Fig 44 Different types of Silicon micropores From ref [14]
mostly for optical and micromachining applications However there is still no unified
theory able to explain the nucleation and growth mechanism of all the different kinds of
pores
We will give a short description of some of these theories for more detail see Parkhutik et
al [15]
One model explains pore nucleation on the basis of physical processes such as hole
positive charge migration ion transport to the surface and small perturbations on the
silicon surface modelled as Fourier components The system is shown to be unstable and
some spatial frequencies that lead to pore nucleation evolve from the etching process
A second model focusses on stationary pore growth without explaining the nucleation
stage According to this model silicon dissolves preferentially at pore edges because h+
charges are attracted by the stronger electric field present at these edges
A third class of models explains pore growth as a Diffusion Limited Aggregation (DLA)
process where the random walk of h+ charge carriers through the depleted layer present at
the silicon-liquid interface controls pore shape
Finally the model by Carstensen Cristophersen and Foll [16] proposes that areas of the
surface of some characteristic size LCO are etched by synchronized ldquocurrent burstsrdquo in the
flow of h+ charges These bursts dissolve silicon through cyclic stages of surface oxidation
oxide removal and hydrogen passivation Areas where a burst has recently taken place are
less passivated and thus more likely to be etched again in this way the pore bottom
continues to dissolve while the pore walls are passivated and are thus less favourable areas
towards current bursts
VPalermo 37
EXPERIMENTAL RESULTS
In the following sections we will show some experimental results obtained from STM and
AFM measurements of fluorine-treated Si(100) surfaces In the first part the results of
mild etching using concentrated and dilute HF solutions at low pH are presented In the
second part the results of etching at high pH using ammonium fluoride are presented and
the mechanism of pore formation discussed
Etching of Si(100) in dilute and concentrated HF
Samples were cut from different areas of an 8-inch diameter p-doped silicon(100) wafer
(10 Ω-cm) supplied by MEMC Electronic Materials Each series of STM measurements
was carried out over at least six different areas on at least two identical samples Low
Electron Energy Diffraction (LEED) was used to check the surface cleanliness of the
samples before STM measurements
Table 41 summarizes the different treatments of each sample After etching with
electronic grade HF each sample underwent a final rinse in Ultra-Pure Water (UPW
resistivity gt18 MΩ-cm) Both the HF and the UPW were allowed to flow continuously
over the sample surface Some samples were not etched with HF at all but just washed with
UPW to observe the morphology of the native oxide layer (~2 nm thick) covering the
surface All of these processes were carried out under nitrogen to limit reoxidation and the
samples were then introduced from the nitrogen atmosphere directly into the vacuum
chamber and degassed overnight at sim150degC before LEED and STM measurements
STM images were obtained from each sample using the same measurement parameters
(sample bias 4 V feedback current 1 nA scan speed 800 nm s-1) These parameters
38 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
provided a satisfactory level of reproducibility for all the samples Measurements were
made over an area of 500x500 nm (image size 500x500 pixel) Slope correction was
carried out by subtracting row-wise and column-wise fitted slopes from the entire image
which gave better results than the simple subtraction of a fitted plane especially for the
rougher samples Following slope correction the rms roughness
sum minus=xy
hyxhN
22 ))((1σ
and the 2-D Fourier transform
)(22
2
)(4
)( vyuxi
xyeyxhvuF +sum∆
= π
π
were calculated for each image where N2 is the number of pixels composing the image
h(xy) is the surface height at each point ∆ is the distance between points h is the mean
height and u v are the spatial frequencies The radial power spectrum PS(f) of the STM image data is obtained from the angular
average of the squared Fourier transform with f 2 = u2 + v2
Fig 45 shows the STM images obtained from the various samples Sample A still covered
with its native oxide layer shows an irregular surface with RMS roughness of ~05 nm
(see Table 41) Observing the sample with LEED gave no diffraction pattern even at
relatively high incident electron energies because of the surface oxide coverage After 1
min etching in dilute HF (sample B) the morphology is similar to the original one though
Table 41 Sample treatments average RMS roughness and integrated area of power spectra
Sample Treatment RMS roughness (nm)
PS area (f lt 01 nm-1)
PS area (f gt 01 nm-1)
A Rapid dip in water 053 plusmn 013 366 032
B 1 min in HF 5 + 10 min in water 051 plusmn 008 272 034
C 30 min in HF 5 + 10 min in water 062 plusmn 008 1064 038
D 5 sec in HF 49 + 10 min in water 042 plusmn 004 183 022
VPalermo 39
Fig 45 STM images of each group of samples showing the topography of the silicon surface A) noetching original oxide surface B) after 1 min etching in dilute HF C) after 30 min etching in diluteHF D) after dipping in concentrated HF Grey scale indicates height of the surface from lower (black) to higher (white) The images are 250x250 nm ie representative portions of the images usedfor the roughness measurement and PSD analysis
some of the larger features have disappeared and the image quality is better maybe due to
improved tunnelling due to the cleaner surface The RMS roughness is comparable to that
of the original surface Clear diffraction patterns are visible using LEED though at quite
high energies (200 eV) After prolonged etching (sample C) the RMS roughness increases
to 062 nm and a long-range corrugation is visible on the surface even if the LEED pattern
is good
The samples dipped in concentrated HF (D) reveal the bare SiSiO2 interface which has a
disordered aspect and protrusions over a wide range of dimensions The quality of the
STM images of sample D is very good probably due to the cleanliness of the surface
40 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 E -0 2
1 E -0 1
1 E + 0 0
1 E + 0 1
1 E + 0 2
1 E + 0 3
0 0 0 0 0 1 0 1 0 1 0 0f (1 nm )
nm^4
A a s re c e iv e d
B e tc h e d 1 min
C e tc h e d 3 0 min
D e tc h e d HF 4 9
Fig 46 Log-Log plot of the averaged power spectra of the STM images for all the samples
which gives a more stable tunnelling junction The LEED pattern is excellent showing
clear diffraction peaks at energies as low as 37 eV comparable to that obtained after high
temperature cleaning in UHV
Fig 46 shows the power spectra of the samples The high frequency and low frequency
areas of the power specturm are considered separately Table 41 shows for each sample
together with the roughness the integrated area of the power spectrum for the high and low
frequency part
We first examine the differences between the samples in the low frequency part of the
spectrum (lt01 nm-1) In this spatial range the short dip in concentrated or dilute HF
(samples BD) seems to lower the roughness of the sample removing some of the larger
features of the native oxide surface visible in Fig 45a and thus decreasing the
corresponding part of the power spectrum Sample C however etched with HF for 30 min
shows an increase in roughness peaking at 0012 nm-1 (~85 nm corresponding to the
typical dimensions of the corrugations visible in Fig 45c) On this scale sample C has a
PS density of 435 nm4 compared to 101 nm4 for sample A
Samples A B and C have the same PS in the high frequency range showing that the fine
structure of the surface is not changed by the HF etching Sample D on the other hand
shows a decrease of roughness for frequencies gt 015 nm-1 This effect could be due to the
VPalermo 41
improved surface cleanliness after etching with concentrated HF which would give a more
stable STM junction thus reducing the high frequency noise in the image
Fig 45 and the analysis of the power spectrum of each sample shows that a rapid dip in
HF removes the native oxide but does not lead to major changes in the morphology of the
Si surface its only effect being the removal of some of the larger features present on the
original surface Prolonged etching on the other hand increases the RMS roughness of the
surface
Etching of Si(100) in concentrated ammonium fluoride and nano-hole creation
Two different types of commercial p-doped Si(100) wafers (2Ω-cm and 10Ω-cm) from
MEMC were used Several different samples of 10x5 mm were immersed for 10 minutes
in 40 electronic grade NH4F solution under agitation Previous works used low
temperatures or anodic potentials applied to the silicon to avoid gaseous hydrogen
production and to obtain a uniform surface but in our experiment we worked at room
temperature to check the influence of hydrogen bubbles on surface morphology During
the etching the stirring was sufficient to provide a uniform concentration of reagents over
the whole sample surface but not to mechanically remove the hydrogen bubbles from the
silicon surface
After the etching each sample was rinsed in ultra-pure water to remove any etching
residues and observed with STM AFM and optical microscopy The AFM measurements
were made in air while for STM measurements the samples were rapidly dried with
nitrogen and inserted into the vacuum system to avoid surface reoxidation After insertion
into the vacuum surface cleanliness was checked with LEED and the surface morphology
observed by STM Parameters for STM measurements were sample bias +4 V 1 nA
current 16 Hz scan rate The images obtained were stable and reproducible over several
days Scan parameters for AFM were 20 nN force and 1 Hz scan rate
Some of the samples were cleaned with an RCA standard clean [1] before NH4F etching to
check the influence of possible surface contaminants on the final results Eight different
samples were prepared and more than sixty STM images of the samples were taken at
different points of the various samples
42 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
(110)
(110)
Fig47 abc) STMtopographic images of differentetching morphologies Eachimage is 500x500 nm Z-ranges are 10 10 and 18 nmrespectively d) STM image of a bridgecreated by etching of the lowerlayers of silicon (black arrow)Image is 250x250x6 nm
After ~2 min of immersion in the solution hydrogen bubbles become visible on the sample
surface The production is slow and the bubbles are quite stable on the sample without
detaching Thus some areas of the surface are masked from the liquid etching action
STM observations (fig 47) show that at the nanometer scale the surface is unevenly
covered with holes of radii ranging from 10 to 200 nm with depths of 2-4 nm These holes
have a wide range of different shapes and distributions In most cases the surface was
covered with a uniform distribution of round-shaped holes (fig47a) indicating isotropic
etching The dimension and the density of the holes changed greatly from sample to
sample and even over the surface of a single sample In some cases the etching was
anisotropic yielding nearly square holes and layered structures as shown in fig 47b
Square holes have been previously observed in cases where the etching speed in the (110)
direction is significantly smaller than in the (100) direction [17]
Over these areas (fig 47b and especially 47c) it is clearly visible how once the surface
had been attacked the reaction continued to preferentially remove atoms at step
irregularities (kink atoms) straightening step edges Eventually the exposed underlying
silicon was also attacked and further holes created inside the previously etched larger
ones It was not possible to detect monatomic steps on this kind of surface The smallest
step height observed was ~15 nm corresponding to several atomic layers In the image
shown in Figure 47a the etching was not strong and created only anisotropic holes on the
surface In fig 47b and c the stronger etch proceeded laterally for several tens of
VPalermo 43
nanometres leaving straight steps several tens of nanometers long and roughly rectangular
holes as expected given the structure of the (100) crystal face In some cases a
significative underetch is observed and the formation of suspended bridges and tunnels
can be deduced from the STM images (fig 47d)
The formation of branched pores and suspended structures has been attributed during pore
formation to diffusion limited aggregation effects where the h+ charge carriers necessary
for silicon etching have a higher probability of reacting at pore bottoms than reaching the
upper part of the silicon surface In the case of very deep pores quantum wire effects have
been invoked to explain the pore growth mechanism [15] In our case though the pores
formed were very shallow the underetch depth being only a few nanometres on pores of
sim100 nm width Thus more than diffusion effects the main contribution to the
underetching process must come from anisotropic etching and some kind of autocatalytic
reaction path analogous to the one described by Matsumura et al [4] with some areas of
the silicon surface hydrogenated and thus less vulnerable to etching
Pre-treatment with RCA cleaning has no effect on the final morphology and this seems to
exclude pore nucleation being caused by presence of metallic or organic surface
contaminants
The morphology and the distribution density of the pits was quite uniform over
microscopic areas of the sample but changes were observed over the millimetre scale This
suggests that etching intensity is influenced by some large-scale parameter
Large-area measurements made with AFM or with an optical microscope (fig 48)
showed that the inhomogeneity of surface etching can be correlated with the masking
action of the bubbles While the fluoride dissolved the silicon hydrogen bubbles formed
by the reaction covered some areas of the surface thus blocking the etching over that area
generating macroscopic steps at the bubble-liquid border As the reaction proceeded more
hydrogen accumulated and the bubble diameter increased producing in this way a circular
pattern of steps The increase in bubble diameter was not continuous with time otherwise a
uniform surface slope gradient would have been obtained The formation of this circular
ldquoetching staircaserdquo indicates that the bubble growth was stepwise the bubble accumulating
more and more hydrogen without enlarging across the surface until it relaxed increasing
its diameter stepwise and covering more silicon The circular structures in fig 48a are not
co-axial and their asymmetry could derive from physical processes due to stirring or
irregularities on the surface
44 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
a b Fig 48 a) optical micrograph of etching patterns on Si(100) created by NH4F 12x09 mm b) AFM image of the circles border xy range is 40x24 microm z-range is 30 nm
The step structure was not destroyed by the etching even after the bubble detached from
the surface but on the contrary the etching process seemed to be influenced by the
presence of the step
Observing in detail a series of steps (fig 48b) a quite deep trench is visible at the base of
each step A close-up image of a step and the corresponding line profile of the trench is
shown in fig 49 The trench is sim5 nm deep with respect to the lower surface compared to
a step height of 22 nm
A similar structure has been recently obtained with electrochemical etching of p-type
Si(100) in 4 HF [18] in which a ldquocurrent burstrdquo etching model previously described
was assumed for silicon dissolution In that experiment the trench was created at the
border of silicon nitride masks and began to grow after a nucleation stage Preferential
trench etching was along the (110) direction and the trench growth was explained as an
effect of mechanical stress induced by the nitride mask and of electric field enhanced
dissolution which depended upon an external applied potential
While it is clear that in our system the gas bubbles have a masking effect similar to a
classical solid nitride mask it is unlikely that hydrogen present on the surface can induce a
significant stress in the silicon lattice as in the case of a nitride mask Furthermore no
external field was applied to drive preferential etching at the trench site
It has been proposed [19] that the cathodic and anodic part of the etching reaction
(hydrogen production and silicon oxidative etching respectively) take place at different
points on the surface with a net charge transfer between the different areas In this case
the highest reaction rates will correspond to the silicon area surrounding the bubble border
where a high number of positive charges will be available for the reaction Furthermore a
sharp trench extending into the silicon crystal will be a preferential electrostatic attractor
VPalermo 45
Fig 49 AFM image of the etched surface showing a step created on the surface by bubblemasking A stronger etching action is visible on the right side of the step as well as theprotected area on the upper side of the step (indicated by the arrows) Image is 10x10 micromz-range is 30 nm The profile on the right is taken from the central area of the image
for the h+ charge carriers coming from other ldquocathodic areasrdquo of the sample either from
other regions on the surface or from the back of the silicon chip [18]
In the areas where the hydrogen bubble had detached and the surface was exposed to the
etching the reaction was not uniform in the neighbourhood of the steps It is possible to
observe (fig 48b and 49) an area on the upper side of the step where less or even no
etching at all seems to have taken place as if the step was able to protect the surface from
etching While etching on the lower side with trench formation can be attributed to the
presence of the bubble the surface on the upper step side can be etched only after bubble
detachment so no masking effect can account for this result However a further
preferential attraction of h+ charge carriers from the already formed trench can be
hypothised electrochemically shielding the surrounding area from further etching If this is
true the shielding effect would be very strong with a relatively shallow 5 nm-deep trench
protecting an area of sim1 microm parallel to the step
To summarise the etching of Si(100) in NH4F is a complex process in which different
reaction paths both chemical and electrochemical co-exist Hydrogen bubbles formed by
the reaction act as a mask on the surface and create etching paths and inhomogeneous
etching of the surface Different kinds of pores are observed on the surface and in some
cases the anisotropy of the process is so strong as to give square-shaped holes and
underetching
46 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The diffusion of h+ charge carriers in the crystal is one of the main rate-determining steps
of the reaction and leads to the formation of a deep trench immediately outside the bubble
perimeter These trenches act as charge collectors and reduce the etching of the upper step
surface in the proximity of the steps
Bibliography
[1] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[2] Spectroscopic Ellipsometry Studies Of HF Treated Si (100) Surfaces Yao H Woollam
Ja Alterovitz SA Applied Physics Letters 62 (25) 3324-3326 Jun 21 1993 Influence
Of HF-H2O2 Treatment On Si(100) And Si(111) Surfaces Graf D Bauermayer S
Schnegg A Journal Of Applied Physics 74 (3) 1679-1683 Aug 1 1993 Kinetics Of
Oxidation On Hydrogen-Terminated Si(100) And (111) Surfaces Stored In Air Miura
T Niwano M Shoji D Miyamoto N Journal Of Applied Physics 79 (8) 4373-4380
Part 1 Apr 15 1996
[3] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995 Structure Of The Stepped SiSiO2 Interface After Thermal-Oxidation -
Investigations With Scanning Tunneling Microscopy And Spot-Profile Analysis Of
Low-Energy Electron-Diffraction Pietsch GJ Kohler U Jusko O Henzler M Hahn
PO Applied Physics Letters 60 (11) 1321-1323 Mar 16 1992
[4] Enhanced Etching Rate Of Silicon In Fluoride Containing Solutions At pH 64
Matsumura M Fukidome H Journal Of The Electrochemical Society 143 (8) 2683-
2686 Aug 1996
[5] A Study Comparing Measurements Of Roughness Of Silicon And SiO2 Surfaces And
Interfaces Using Scanning Probe Microscopy And Neutron Reflectivity Crossley A
Sofield CJ Goff JP Lake ACI Hutchings MT Menelle A Journal Of Non-Crystalline
Solids 187 221-226 Jul 1995
VPalermo 47
[6] Step-Flow Mechanism Versus Pit Corrosion - Scanning-Tunneling Microscopy
Observations On Wet Etching Of Si(111) By Hf Solutions Hessel HE Feltz A Reiter
M Memmert U Behm RJ Chemical Physics Letters 186 (2-3) 275-280 Nov 8 1991
[7] Comparison Of Si(111) Surfaces Prepared Using Aqueous-Solutions Of NH4F Versus
HF Higashi GS Becker RS Chabal YJ Becker AJ Applied Physics Letters 58 (15)
1656-1658 Apr 15 1991
[8] Wet Chemical Etching Of Si(100) Surfaces In Concentrated NH4F Solution -
Formation Of (2x1)H Reconstructed Si(100) Terraces Versus (111) Faceting Neuwald
U Hessel HE Feltz A Memmert U Behm RJ Surface Science 296 (1) L8-L14 Oct
10 1993
[9] Ideal Hydrogen Termination Of Si(001) Surface By Wet-Chemical Preparation Morita
Y Tokumoto H Applied Physics Letters 67 (18) 2654-2656 Oct 30 1995
[10] Atomic Structures Of Hydrogen-Terminated Si(001) Surfaces After Wet Cleaning
By Scanning Tunneling Microscopy Endo K Arima K Kataoka T Oshikane Y Inoue
H Mori Y Applied Physics Letters 73 (13) 1853-1855 Sep 28 1998
[11] On The Potential-Dependent Etching Of Si(111) In Aqueous NH4F Solution
Houbertz R Memmert U Behm RJ Surface Science 396 (1-3) 198-211 Jan 20 1998
[12] Etching Process Of SiO2 By HF Molecules Hoshino T Nishioka Y Journal Of
Chemical Physics 111 (5) 2109-2114 Aug 1 1999
[13] Etching Mechanism And Atomic-Structure Of H-Si(111) Surfaces Prepared In
NH4F Allongue P Kieling V Gerischer H Electrochimica Acta 40 (10) 1353-1360
Jul 1995
[14] Pore Formation Mechanisms For The Si-HF System Carstensen J Christophersen
M Foll H Materials Science And Engineering B-Solid State Materials For Advanced
Technology 69 23-28 Sp Iss Si Jan 19 2000
[15] Porous Silicon - Mechanisms Of Growth And Applications Parkhutik V Solid-
State Electronics 43 (6) 1121-1141 Jun 1999
[16] Formation And Application Of Porous Silicon Foll H Christophersen M
Carstensen J Hasse G Materials Science amp Engineering R-Reports 39 (4) 93-141 Nov
1 2002
[17] Pore Morphology And The Mechanism Of Pore Formation In N-Type Silicon
Searson PC Macaulay JM Ross FM Journal Of Applied Physics 72 (1) 253-258 Jul 1
1992
48 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
[18] Deep Electrochemical Trench Etching With Organic Hydrofluoric Electrolytes
Christophersen M Merz P Quenzer J Carstensen J Foll H Sensors And Actuators A-
Physical 88 (3) 241-246 Mar 5 2001
[19] Morphology Of Anodically Etched Si(111) Surfaces - A Structural Comparison Of
NH4F Versus HF Etching Houbertz R Memmert U Behm RJ Journal Of Vacuum
Science amp Technology B 12 (6) 3145-3148 Nov-Dec 1994
VPalermo 49
5 Surface modification of silicon in vacuum void creation and oxide desorption
The main reason for the huge success of silicon in the microelectronics industry is not due
to its superior properties as a semiconductor Other materials for example germanium
have better qualities such as higher mobility of charge carriers and lower noise levels
which would allow the construction of faster and higher performance devices
The widespread use of silicon however is mainly due to the outstanding characteristics of
its oxide Silicon dioxide (SiO2) is a very good electrical insulator easy to form
chemically and thermally stable and is compatible with lithographic and metal deposition
processes Germanium oxide on the contrary is too reactive to be used
Even the use of Si(100) substrates for nearly all microelectronic devices is dictated by
oxide quality The (111) face of silicon crystal can be easily cleaved and flattened and
almost atomically perfect surfaces can be obtained with simple chemical procedures (as
described above) But the density of interfacial defects is highest for Si(111)-SiO2
interfaces and lowest for Si(100)-SiO2 ones so microchips will continue to be fabricated
on Si(100) wafers
SiO2 (silica) is present in 95 of the earthrsquos minerals in different allotropic forms such as
quartz tridymite and cristobalite In the bulk each silicon atom is bonded to four oxygens
in a Si-O-Si tri-dimensional network Si-O bonds are 016 nm long and form an angle
ranging from 120deg to 150deg
Three typical intrinsic defects are present in SiO2 The so-called Ersquo centres are oxygen
vacancies with a hole localised on a silicon atom with only three Si-O bonds
50 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
O3Si +SiO3 Whereas the PR (peroxy radical) defects are holes trapped by a charged
peroxy moiety with a O3Si-O-O+ SiO3 structure The NBOHC (non-bridging oxygen hole
centres) derive from water or hydrogen contamination and are schematized as O3Si- O- H-
O-SiO3
The atomic structure of the Si-SiO2 interface varies enormously Local domains resembling
the tridymite and the cristobalite structure of silica are present but it seems that only 10
of the interface is ordered [1] Far from the interface the SiO2 bulk is completely
disordered The passage from bulk Si to stoichiometric SiO2 passes through a non-
stoichiometric SiOx layer 07 nm thick
When a clean silicon surface is exposed to atmospheric oxygen a thin sim2 nm thick layer
of native oxide forms spontaneously which is usually removed and substituted with
thicker better quality oxide layers before further processing
Silicon is usually oxidised by thermal annealing at temperatures between 800deg and 1100deg
in an atmosphere of pure O2 with some water eventually added to increase oxidation speed
Thermal oxides made in pure oxygen (dry oxides) grow more slowly than oxides produced
in an oxygen-water atmosphere (wet oxides) but are usually of better quality
According to the Deal-Groove formula the time t needed to grow an oxide of thickness X
is given by [1]
1
212minus
minusminus
+=
ABXBXt α
where the constant B and BA decrease exponentially with temperature as
minus
kTEexp
with activation energies for dry oxidation of EB =123 and EBA =20 eV respectively
EB is related to the diffusion of oxygen in silicon while the value of EBA is interpreted as
the energy required to break a Si-Si bond The exponent α is 1 for wet oxidation and 0 for
oxidation at high temperatures and low oxygen pressures It has intermediate values for dry
oxidation This formula does not work well for low values of X at the initial stages of
oxidation and usually empirical corrections are used
An interesting characteristic of silicon is that at high temperature and in vacuum oxygen
can actually etch the silicon crystal giving gaseous products and the oxidized layer present
on the crystal becomes unstable (fig 51) [2]
VPalermo 51
1E-10
1E-09
1E-08
1E-07
1E-06
1E-05
1E-04
1E-03
1E-02
1E-01
0607080911112
1000T (1K)
P (T
orr)
SiO2 + Si rarr 2SiO(g) Oxide decomposition
Si+O2 rarr SiO2(s) oxide formation Fig51 phase diagram of the
oxygen-silicon system
Silicon oxidation apart from the initial nucleation stages at the monolayer level proceeds
uniformly over the whole surface with a planar reaction front moving from the surface
into the bulk
If heated under low oxygen partial pressure (vacuum or inert atmosphere) SiO2 is known to
decompose following the reaction
SiO2 + Si rarr 2SiOuarr (1)
The reaction begins with nucleation at defect points on the SiSiO2 interface and proceeds
in a spatially inhomogeneous manner with the formation of large voids on the oxide
surface [3]
Several studies have been made on the dynamics of void growth both on thick [4] and thin
[5] layers of SiO2 The process has been used to decorate otherwise unobservable defects at
the SiSiO2 interface [6] or to grow nanoislands of silicon on the void surface [7] It has
been suggested that the defects acting as nucleation centres could be metallic contaminants
present on the native surface which aggregate and catalyze SiO2 decomposition [8]
After oxide desorption the silicon surface is very rough In particular on Si(100) square
islands are observed several nanometres high which act as lsquopinning sitesrsquo on the motion of
monatomic steps on the Si surface It has been proposed that these islands can be composed
of silicon carbide coming from the organic contaminants present on the surface [9] or of
oxide clusters during partial reoxidation [10]
In the following section some experimental results obtained studying voids and nanoisland
growth are presented
52 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nanoisland growth on silicon
We used two different samples the first rougher than the second (i) p-type Czochralski Si
(100) with a RMS roughness of 05 nm shown in fig52a and (ii) p-type epitaxial Si(100)
with a RMS roughness of 019 nm shown in fig 52b The roughness was measured with
STM on the lsquoas receivedrsquo samples on different areas of 500x500 nm Every sample was
covered by a layer of native oxide ~2 nm thick
Some of the samples (AC) were introduced into the UHV system without any cleaning
others (BD) were dipped in concentrated HF (49) to remove the surface oxide without
etching the silicon [11]
After degassing the samples were heated resistively increasing the temperature slowly to
keep the pressure within the 10-10 mbar range during heating Each sample was held at
900˚C for 30 min For some of the samples the heating was stopped at lower temperatures
to obtain incomplete oxide desorption and surface roughening thus enabling observation
of the different steps of the process Table 51 summarizes the different treatments for each
group of samples
Fig52 Original surfaces of Czochralski (a) and epitaxial (b) silicon Every image is 500x500 nm
VPalermo 53
Table 51 Summary of different treatements
Silicon type Oxide removed
Heating Islands density (microm-2)
Av Island Volume (nm3)
A Czochralski No 30rsquo at 900˚C 24x103 353
B Czochralski Yes ldquo 20x104 71
C Epitaxial No ldquo 28x102 2450
D Epitaxial Yes ldquo 23x104 63
Fig53 ab) Surface roughening on group A samples after heating 30 min at 800deg and900deg respectively in presence of an oxide layer Image size 500x500 nm cd) Surface roughening on group B samples after heating 30 min at 700deg and 900degrespectively after removing the oxide layer with HF Image size 250x250 nm
Fig 52a is the native oxide surface of sample A quite disordered and irregular This
surface was stable when heated to 700˚C while at higher temperatures the oxide began to
desorb Fig 53a was taken after heating the sample at 800˚C for 30 min
54 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The oxide began to desorb in correspondence with defect points creating oxide-free
surface voids which enlarge radially uncovering the silicon surface Some surface silicon
atoms are removed through reaction (1) Mobile silicon atoms moving randomly over the
surface aggregate forming a nanocrystal at the initial defect point surrounded by a stable
flat oxide-free surface with some steps The initial SiSiO2 interface as observed after
oxide etching with HF [12] is similar to the one shown in fig 52a and thus the surface
observed in the void area is a completely new one coming from surface reorganization
The reaction continues leading to a coalescence of the voids the final result of the process
is shown in fig 53b with protruding islands aligned along the main crystal axes and flat
areas with some steps LEED measurements confirmed that this surface is crystalline
silicon Atomic resolution was obtained on the flat areas of the surface showing the
typical 2x1 reconstruction of Si(100) Figures 53c and 53d show the evolution of the
roughening on sample B after chemically removing the native oxide layer before
introducing the sample into UHV As mentioned above the initial morphology is similar to
that shown in fig 52a though the oxide has been removed as confirmed by LEED
measurements Fig 53c shows the surface after heating at 700˚ for 30 min Without the
oxide layer the surface began to reorganize at lower temperatures and the small irregular
grains of the original surface begin to grow and become more rectangular After heating to
900˚C (fig 53d) the surface is flat with small rectangular islands The shape of the islands
is similar to that shown in fig 53b but the lateral island dimension is one order of
magnitude smaller
To assess the importance of the initial surface on the process the same treatments were
repeated starting from flat epitaxial silicon (samples CD) The initial epitaxial surface is
smooth with periodic steps due to a small miscut angle (fig52b) On heating the epitaxial
silicon without etching (sample C) relatively large islands are obtained similar to the ones
shown in fig 53b where the ldquostep-pinningrdquo effect of the islands is evident The average
dimension of sample C islands was sim2500 nm3 while only 353 nm3 for sample A Island
density decreased from 24000 to 280 microm-2 This result is maybe due to a lower
concentration of defects in the oxide layer of the epitaxial silicon which allows the
expanding void to grow much larger before coalescence with neighbouring voids the
nanostructures within them thus reaching greater dimensions To verify this hypothesis we
applied the same treatment to clean not-so clean and purposely contaminated samples To
check that the defects inducing void nucleation and nanoisland growth derive from carbon
VPalermo 55
contamination we used electronic-grade methanol to contaminate the silicon surface
Figure 54 shows the results of this comparison
State-of-the-art cleaned samples obtained with multi-step RCA cleaning [13] showed the
development of very few voids the oxide desorption leaving large smooth areas of oxide-
and nanostructure-free silicon as shown in Fig 54a where the presence of residual oxide
can be seen Samples cleaned with a simpler one-stage Pirana cleaning solution showed a
higher number of islands per unit area (Fig 54b) and samples purposely contaminated
with methanol (Fig 54c) had the highest density of island nucleation of all three samples
When the native oxide had been chemically removed from the epitaxial silicon before
heating (sample D) the island size and density is similar to the standard as-received
silicon case (sample B) Thus starting from a flat substrate the process of surface
roughening and island creation was the same even when the oxide layer had been
previously removed
From these results some important indications can be drawn The first is that oxide
desorption temperature depends upon the contamination levels of the oxide layers (fig 54)
with desorption at lower temperatures for contaminated surfaces
Second once the oxide layer is removed the silicon atoms become mobile on the surface
at relatively low temperatures (~700˚C) Their initial configuration is disordered and has a
high surface energy so they crystallize in the small islands shown in fig53d reducing
their surface area and lowering the energy of the system When the oxide is present the
surface cannot reorganize at 700˚C and is stable At 800˚C the oxide begins to desorb in
correspondence with defect points on the surface As the void area enlarges a small
cba Fig 54 The effect of contamination on nano-island production Gradient-filtered STM images of (a)sample cleaned two times with RCA 400x400 nm (b) Sample cleaned with a Pirana solution 500x500nm (c) sample contaminated with MeOH before insertion into the UHV system 500x500 nm Annealingtime is 40 min at 800degC for all the samples
56 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
crystallite grows at the centre of the voids Further oxide desorption uncovers larger areas
freeing more silicon atoms which feed the initial island making it bigger
Fig 55 shows the scaled size distribution of the islands for each group of samples Every
distribution has been obtained using several images of different samples for each group
The size distribution of all the samples follows an exponential decay and the decay seems
the same for all the samples though the average volume of the islands differs by more than
one order of magnitude (see Table 51 for details)
The point defects which catalyze void nucleation can be metallic contaminants present on
the initial oxide surface [8] or organic contaminants which at high temperatures can
form SiC nanocrystals [14] We used Scanning Tunneling Spectroscopy (STS) to look for
differences between the islands and the surrounding flat silicon surface but no difference
was found Furthermore the total island volume per surface unit is very high (more than
104 nm3microm2) and it seems unlikely that such a huge volume could consist of surface
contaminants Islands of pure silicon on silicon have been grown without evidence of
surface contamination[14]
To summarise the formation of nano-sized islands on silicon through surface diffusion
was studied At high temperatures the oxide layer covering the surface decomposes non
uniformly and circular voids of clean silicon are created The presence of the oxide layer
blocks surface silicon atomic motion and surface reorganization except within the voids
causing the growth of islands more than 10 nm high and 30 nm wide on the silicon surface
According to STM and LEED measurements we can say that the islands are mostly
Fig 55 Size distribution (scaled) for the samples described in table 1 The line is a reference for the eye
VPalermo 57
composed of silicon with traces of other substances such as carbon Final island densities
and dimensions depend upon the initial purity of the oxide layer
On the other hand if the oxide is removed chemically before heating the island growth
process is not so localised and takes place simultaneously over the whole surface yielding
smaller and more numerous islands
58 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[2] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[3] High-Temperature SiO2 Decomposition At The Sio2Si Interface Tromp R Rubloff
GW Balk P Legoues FK Physical Review Letters 55 2332-2335 Nov 1985
[4] Defect Microchemistry At The SiO2Si Interface Rubloff GW Hofmann K Liher M
Young DR Physical Review Letters 582379-2382 Jun 1987 Defect Formation In
Thermal SiO2 By High Temperature Annealing Hofmann K Rubloff GW Mccorkle
RA Applied Physics Letters 49 1525-1527 Dec 1986 Kinetics Of High-Temperature
Thermal Decomposition Of SiO2 On Si(100) Liher M Lewis JE Rubloff GW Journal
Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 5 1559-1562 Aug
1987
[5] Thermal-Decomposition Of Very Thin Oxide Layers On Si(111) Kobayashi Y Sugii
K Journal Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 10 (4)
2308-2313 Part 3 Jul-Aug 1992 Scanning Tunneling Microscope Study On Mid-
Desorption Stages Of Native Oxides On Si(111) Kobayashi Y Sugii K Journal Of
Vacuum Science amp Technology B 9 (2) 748-751 Part 2 Mar-Apr 1991 Controlled
Growth Of SiO2 Tunnel Barrier And Crystalline Si Quantum Wells For Si Resonant
Tunneling Diodes Wei Y Wallace RM Seabaugh AC Journal Of Applied Physics 81
(9) 6415-6424 May 1 1997
[6] Defect Formation In SiO2Si(100) By Metal Diffusion And Reaction Liher M
Dallaporta H Lewis Je Appl Phys Lett 53 589-591 Aug 1988 SiO2 Film
Decomposition Reaction Initiated By Carbon Impurities Located At A Si- SiO2
Interface Raider Si Herd Sr Walkup Re Applied Physics Letters 59 (19) 2424-2426
Nov 4 1991
[7] Nanometer-Scale Si Selective Epitaxial Growth On Si(001) Surfaces Using The
Thermal Decomposition Of Ultrathin Oxide Films Fujita K Watanabe H Ichikawa M
Applied Physics Letters 70 (21) 2807-2809 May 26 1997 Pyramidal Si Nanocrystals
VPalermo 59
With A Quasiequilibrium Shape Selectively Grown On Si(001) Windows In Ultrathin
SiO2 Films Shibata M Nitta Y Fujita K Ichikawa M Physical Review B 61 (11)
7499-7504 Mar 15 2000
[8] Stacking-Fault-Induced Defect Creation In SiO2 On Si(100) Liher M Bronner Gb
Lewis Je Appl Phys Lett 52 1982-1985 May 1988
[9] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990 AFM And XPS Characterization Of The
Si(111) Surface After Thermal-Treatment Lamontagne B Guay D Roy D Sporken R
Caudano R Applied Surface Science 90 (4) 481-487 Dec 1995
[10] Scanning-Tunneling-Microscopy Study Of Oxide Nucleation And Oxidation-
Induced Roughening At Elevated-Temperatures On The Si(001)-(2x1) Surface Seiple
JV Pelz JP Physical Review Letters 73 (7) 999-1002 Aug 15 1994 Evolution Of
Atomic-Scale Roughening On Si(001)-(2x1) Surfaces Resulting From High-
Temperature Oxidation Seiple JV Pelz JP Journal Of Vacuum Science amp Technology
A-Vacuum Surfaces And Films 13 (3) 772-776 Part 1 May-Jun 1995
[11] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995
[12] Morphological Changes Of The Si [100] Surface After Treatment With
Concentrated And Diluted HF Palermo V Jones D Materials Science In
Semiconductor Processing 4 (5) 437-441 Oct 2001
[13] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[14] Nanoscale Roughening Of Si(001) By Oxide Desorption In Ultrahigh Vacuum
Gray SM Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology
B 14 (2) 1043-1047 Mar-Apr 1996
60 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 61
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and nano-lines
When the silicon surface is not protected by a native oxide layer or by a hydrogenated
passivating layer it is very reactive The surface chemistry of both Si(100) and Si(111) has
been extensively studied starting from clean surfaces prepared in UHV because of its great
relevance to the microelectronics industry and its technological interest The diffusion and
reaction of molecules and atoms on silicon is an interesting scientific problem on its own
apart from technological issues because silicon reconstruction yields a very complex and
anisotropic surface
We give here a brief summary of the adsorption behaviour of several elements on Si
surfaces The summary is not comprehensive and only the most interesting characteristics
for each substance are given
Table 61 Summary of adsorption behaviour of atoms and molecules on silicon surfaces [12]
Hydrogen
Molecular H2 shows low reactivity towards silicon while atomic hydrogen easily forms
Si-H bonds and can even break Si-Si bonds
Alkali metals
Alkali metals diffuse rapidly into SiO2 and can damage silicon-based transistors
Transition
metals
All transition metals apart from gold and silver react with Si forming metal silicides
62 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nickel A very common contaminant of silicon Nickel from even slight contact with stainless
steel tweezers can contaminate silicon samples forming its silicide and is very difficult
to remove even at high temperatures because it segregates on the silicon surface
Palladium Forms silicides especially Pd2Si and can be used to make contacts possessing a low
Schottky barrier
Titanium Widely used to fabricate contacts for silicon-based microdevices TiSi2 is one of the
more stable and highly conductive silicides
Tungsten The silicide is more stable than TiSi2 on polysilicon and is thus used for polysilicon
gate contacts
Platinum Silicide is used for bipolar transistors PtSi contacts on n-silicon give Schottky diodes
Cobalt Cobalt silicide gives better quality contacts than TiSi2 at the nanometrelevel but is less
used at the moment because it consumes too much silicon during its growth
Iron Silicides have been studied for potential optoelectronic applications because FeSi2 has a
directly accessible bandgap and is optically active
Group 13
Group 13 metals are used in IC technology to dope silicon (giving p-type doping) and
to make III-V type semiconductors When evaporated on Si(100) or Si(111) they react
strongly with the surface giving surface reconstruction and kink formation at steps
Boron Widely used for doping Usually deposited with decomposition of boron hydrides
Sticking coefficient of simple hydrides is very low so decaboranes (B10H14) are used for
deposition
Aluminum Aluminum is widely used for microelectronic contacts but in some cases it reacts with
silicon diffusing into the crystal and with SiO2 stealing oxygen atoms to form Al2O3
Group 14
Carbon Carbon can form a composite semiconductor with silicon (silicon carbide) widely used
to fabricate integrated circuits for use at elevated temperatures or in the presence of
ionising radiation When carbon is deposited on silicon usually polycrystalline films
with properties between diamond and graphite are obtained
CO adsorbs molecularly on silicon without breakage of the C=O bond Other
unsaturated hydrocarbons do not react with silicon Even very reactive strained
molecules like cyclopropane have low sticking coefficient Ethylene acetylene and
propylene adsorb molecularly on Si(100) each molecule sitting on a silicon dimer
interacting with the dangling bonds of the surface Acetylene also adsorbs molecularly
on Si(111) on the 7x7 reconstructed surface preferential adsorption on centre atoms
with respect to corner atoms is observed Benzene and other aromatic ring compounds
adsorb molecularly on the surface They can form σ or π-type bonds with the silicon
laying parallel or tilted respect to the surface according to the bond type
Silicon Silicon atoms can be deposited on the surface with molecular beam epitaxy or
VPalermo 63
decomposition of silanes and chlorosilanes Silanes with single Si-Si bonds have a high
sticking probability and dissociate upon adsorption giving trihydrides and
monohydrides
Chlorosilanes have high sticking coefficients too and decompose on the surface but
they can etch the surface through the reaction SiCl4 + Si rarr 2SiCl2
Germanium Germanium has a lattice constant similar to silicon (Ge lc is 4 larger than Si) so that
various alloys of Si1-xGex can be formed The growth of Germanium on silicon is quite
peculiar with several monolayers adsorbing uniformly on the surface followed by
island formation (Stranski-Krastanov growth)
Group 15
Nitrogen Exposure to ammonia and nitrogen gas at high temperatures leads to the formation of
silicon nitride layers (Si3N4) NH3 dissociates on the surface and reacts with the
dangling bonds giving Si-NH2 and Si-H termination Silicon nitride is stable on the
surface up to 1100degC above this temperature it desorbs as Si2N
Phosphorous Phosphine (PH3) adsorbs dissociatively as Si-PH2 and Si-H and behaviour is similar to
that of ammonia
Arsenic A monolayer of As is often deposited on silicon as a substrate for GaAs growth Arsenic
forms dimers on Si(100) creating dimer rows which are parallel (perpendicular) to the
underlying silicon dimer rows when it is deposited at high (low) temperature On
Si(111) Arsenic breaks the 7x7 reconstruction giving a 1x1 pattern This 1x1 As
monolayer acts as a passivating layer and can resist further As adsorption oxygen and
air
Antimony and
Bismuth
Due to their large covalent radii Sb and Bi form only short dimer rows on silicon and
only at high temperatures
Group 16
Oxygen
Oxygen can oxidize or etch silicon according to the reactions
Si(s) + O2 rarr SiO2 (s)
Si(s) + frac12O2 rarr SiO(g)uarr
Silicon oxidation has already been described in the previous chapter Molecular oxygen
adsorption on silicon is mostly dissociative O atoms break the Si-Si bonds forming an
Si-O-Si bridge or a peroxy bridge Si-O-O-Si Hydrided silicon surfaces do not have
surface dangling bonds and are thus more protected from oxygen attack Etching is
favoured at high temperatures and low oxygen pressures [3] At T gt700degC silicon oxide
on silicon decomposes as SiO(g)
Water
H2O adsorbs dissociatively on silicon easily on Si (100) and with more difficulty on
Si(111) The molecule decomposes giving Si-OH and Si-H on adjacent sites
64 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Group 17
Fluorine
Fluorine adsorbs and reacts easily on silicon dissociating on the surface and forming
Si-F bonds It can then penetrate the surface and attack polarised Si-Si backbonds The
surface is etched with production of SiF4 and SiF2
Chlorine
Chlorine reacts aggressively with silicon etching the surface Cl2 and F2 are often used
commercially to etch silicon usually in a plasma Atomic Cl and Cl2 adsorb on Si(100)
up to saturation The most stable configuration seems to be a Cl atom bonded to each Si
atom of a surface dimer in a symmetric structure but metastable configurations with
two Cl atoms bonded to a buckled dimer and even a central Cl atom bridged across a
dimer have been observed
On Si(111) silicon mono- di- and tri-chlorides are formed In particular heating at
Tgt400degC a uniform Si-Cl monolayer is obtained and the 7x7 reconstruction changes
into a bulk like 1x1 lattice The surface transition induced by Cl adsorption is fully
reversible The 7x7 reconstruction can be restored on desorbing the chlorine by heating
at Tgt1100degC The 1x1 domains nucleate at the lower terrace side of steps the 7x7 at the
upper step edge
Bromine
Like hydrogen bromine maintains the 2x1 structure of Si(100) forming Si-Br bonds
with the dangling bonds of the silicon dimer rows At high exposure some etching of
silicon by formation of volatile SiBr3 species has been observed
Surface diffusion on silicon
In the previous chapter we observed nano-island growth on silicon surfaces with native
oxide and surfaces contaminated with organic impurities While this phenomenon has been
observed in several experiments the formation and growth mechanisms of nanoislands has
still not been satisfactorily described theoretically Various substances have been
hypothesised as nanoisland nucleating agents such as organic or metallic contaminants
present on the surface or oxide clusters (see previous chapter) The island growth process
has been attributed by various workers to lsquosome sort of kinetic instabilityrsquo [4] to the
pinning of step flow by SiC clusters [5] or to a mesoscopic atom flux from areas with low
step density to areas of higher step density [6]
One mechanism proposed to explain surface roughening and island growth is the Ehrlich-
Schwoebel effect ie the presence of an energetic barrier that reflects atoms approaching a
VPalermo 65
Molecule (L
O2
H2O MeOH
CO CO2 CH4
clean surface
downward
weak on si
The aim o
surface an
To obtain
the ones
spontaneou
first obser
surface W
nanoisland
Experime
We used S
removed f
41 mixtur
water and
whole hea
1200degC re
kept below
Table 62 Treatment island density and island volume for each molecule tested
Dose angmuir)
Heating time at 800degC (min)
Island density (microm-2)
Ave Volume (nm3)
Equivalent layer thickness (nm)
280 10 no islands - -
90 10 no islands - - 30 10 2000 19 004
180 10 360 516 019 30 60 no islands - -
180 10 no islands - - - 10 no islands - -
step [7] Recent experiments however indicate that the Schwoebel effect is
licon [48]
f our experiments was to study the reaction of simple molecules with the silicon
d to check the ability of these molecules to nucleate nanoisland growth
a better understanding of the process we chose simple molecules starting from
which are more likely to be present as traces in UHV chambers where
s nanoisland growth is often observed after sample heating Using STM we
ved at the atomic level the possible interactions of these molecules with the
e then heated the surface and checked the density and dimensions of the
s produced
ntal procedures and results
i(111) wafers p-doped 0015 Ω cm Gross contamination and particles were
rom the sample surface with a standard Pirana etch (15 min dip in a H2S4H2O2
e at T=80degC) After this the samples were thoroughly washed in ultra-pure
introduced into the UHV system After degassing for 5 hours at 600degC the
ting stage was allowed to cool down and a sequence of rapid flashes to 1100deg-
moved the native oxide layer and cleaned the surface Pressure during flashes is
1x10-9 mbar
66 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Surface flatness and cleanliness were checked with STM and the contaminant molecules
introduced into the UHV system through a leak valve Doses of several Langmuirs were
used for the experiments (1 Langmuir=10-6 torr sec-1) Pressure was kept constant at 10-7
mbar during adsorption The STM tip was retracted during gas deposition to avoid any tip
shielding effect After the adsorption process the leak valve was closed the base pressure
in the UHV chamber returning to below 10-10 mbar and the presence of adsorbed
molecules checked on the surface with STM Finally the sample was heated at 800degC for
several minutes to allow nanoisland growth Several different areas were examined on each
sample
Table 62 summarizes the dose of exposure and the heating time for each molecule Other
experiments were carried out using smaller dosing times but only the significant results
are reported here
As expected oxygen and water react with the surface oxidising it The adsorption process
can be easily followed with the STM by lowering the tip from time to time and taking an
image No effect of the STM tip is observed over the scanned areas Upon molecule
adsorption the order of the crystalline surface rapidly degenerates and after several
minutes the surface looks completely covered by irregular atomic-sized protrusions
corresponding to Si-OH or Si-O-Si species On heating to 800degC the contaminating species
were easily desorbed through SiO formation the surface became clean and the crystal
surface periodicity re-established No residues or nanoislands were observed
Amongst the molecules tested was methanol CH3OH reacts readily with the surface and
covers it with a disorderd layer already after sim 30 Langmuirs exposure (fig 61)
Fig 61 Si(111) surface during methanol adsorption at 0 1 and 5 minutes respectively (corresponding to 0 6 and 30 L exposure) The dark irregular lines are borders between different 7x7 domains Image size 60x60 nm
VPalermo 67
Synchroton radiation photoemission spectroscopy experiments by Carbone et al [9]
showed that methanol reacts with the rest atoms of the 7x7 silicon cell already after 1
Langmuir exposure but that longer exposures are needed for methanol to react with the
remaining silicon atoms Methanol adsorbs dissociatively forming SiOCH3 and Si-CHx
species on the surface
After methanol adsorption the sample is heated to 800degC and observed again with STM
As in the case of water and oxygen the heating restores the crystalline surface and the 7x7
pattern returns visible But this time quite a dense array of nanoislands is observed on the
surface (fig 62) The islands have an average diameter of 20 nm and a height of sim2 nm
At temperatures above 400degC the Si-OCH3 and Si-CHx species decompose oxygen and
hydrogen are desorbed and according to [9] carbon atoms remain as SiC dispersed
uniformly on the surface Our STM measurements indicate that the carbide does not
randomly cover the surface but that C atoms are concentrated at the nanoislands Rough
calculations based on island volume suggest that the islands must be composed of a SixC1-x
alloy with x varying between 05 and 1
Following these results we expected carbon monoxide to behave in a similar manner
adsorbing onto the surface and with nanoisland growth However no adsorption was
observed with STM and the surface looked perfectly clean even after a dose of more than
100 L of carbon monoxide Some species did adsorb onto the surface though because
after 10 minutes subsequent heating nanoisland growth was observed Island density in
this case was only 18 of the density obtained with methanol and island dimensions are
Fig62 Nanoislands on Si(111) created after 30 L adsorption of methanol and 10 min heating at 800degC Image size 200x200 nm
68 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
much bigger (see table 62 and fig 63) To check whether the islands had been nucleated
by some other contaminant present in the vacuum chamber we heated a blank sample
without introducing any molecule No island growth was observed Exposing the surface to
other simple molecules including carbon dioxide and methane also gave no nanoisland
formation
We can estimate the amount of carbon deposited on the surface during gas dosing The
total flux of molecules hitting the surface is obtained from the Hertz-Knudsen formula
TkmPF
π2=
Fig 63 STM images of Si(111) after exposure to methanol (left) and carbonmonoxide (right) contaminating molecules and subsequent nanoisland growthImage size 400x400 nm The size distribution of nanoislands is shown below eachimage
VPalermo 69
Fig64 Left Si(111) after 180 L contamination of CO and 16 hours heating at 800degC Right Si(111)without CO contamination after 16 hours heating at 800degC Image size 250x250 nm Vertical scale is thesame for both images z-ranges are 35 and 5 nm respectively
where P is the pressure in Pascal k=138x10-23 J K-1 is the Boltzmann constant T is the
temperature and m is the molecular weight (sim32 for methanol and sim28 for CO) in
kilograms
Using this formula we obtain a flux of 027 molecules nm-2 sec-1 for methanol and 029
molecules nm-2 sec-1 for CO After ten minutes exposure at 10-7 mbar pressure more than
150 molecules will have hit each square nanometer of the surface Surface density of
Si(111) 7x7 is sim16 atomsnm2 so each surface atom will be hit by several molecules which
could react or be adsorbed even assuming a sticking coefficient much lower than unity
To explain nanoisland growth we hypothesize that CO molecules adsorb molecularly and
-5
0
5
10
15
20
25
30
0 20 40 60 80 100 120
nm
nm
10 min40 min16 h16 h- no CO
Fig65 STM profiles of nanoislandsgrown for different annealing times The lower curve correspond to the surface heated at 16 hours without CO contamination
70 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Table 63 Island density and volume for the CO-contaminated surface at different heating times For each sample the thickness of a uniform layer having the same volume per unit area of the islands is calculated
Molecule Dose
(Langmuir) Heating time at
800degC (min) Island density
(microm-2) Ave Volume
(nm3) Equivalent layer thickness (nm)
CO 180 10 362 517 019 CO 180 40 550 1072 059 CO 180 16 h 505 3675 186
clean surface - 16 h rough - -
very weakly on Si(111) moving rapidly on the surface Because of their rapid diffusion
the STM is not able to image the CO molecules on the silicon [10]
Increasing the temperature part of the CO molecules will surely desorb but some of them
will dissociatively adsorb onto the surface The oxygen will then desorb as SiO while
carbon will remain fixed forming very stable Si-C bonds
CO decomposition will be favored at surface defects surface steps or in the proximity of
already formed SiC clusters An increased reactivity of the CO molecule on a surface in
correspondence with phase boundaries has already been observed on Pt surfaces [10]
In this case a reduced number of nucleation centres will react with the CO molecules
yielding large and fewer islands with respect to the case of methanol which does not have
the possibility of travelling long distances over the surface
Other factors are likely to influence the process of island nucleation after CO adsorption
the co-adsorption of other molecules which can slow down CO and favour its
decomposition cannot be ruled out as well as the formation of new defects at higher
temperatures Measurements with a variable temperature STM or with some other surface
analysis technique are planned in the future to confirm the proposed mechanism
When the surface was further annealed the nanoislands grew in size After 16 hours
heating very large nanoislands with diameters of sim35 nm and heights of sim20 nm became
visible (fig64 left) As a comparison when the same surface was heated without
nucleation centres deriving from contamination an irregular surface was obtained (fig64
right)
While the nanoisland density seemed to reach an asymptotic limit of 500 islandsmicrom2
(table 63) island size continued to grow with time (fig65) even when no further carbon
was supplied to the surface This indicates that even though the initial nucleating core of
the nanoisland is likely to be an SixC1-x alloy further growth is due to silicon atoms
diffusing from the crystal and being adsorbed by the growing island The final volume
VPalermo 71
occupied by the islands corresponds roughly to a 18 nm thick overall layer of removed
silicon
It is noteworthy that the clean sample after 16 hours annealing even though smoother than
the CO-contaminated one looked much more disordered with hardly any flat area visible
on nanometres scale In the presence of nanoislands however further annealing increases
island size but keeps part of the surface quite flat and ordered with flat areas visible
between the islands Thus the nanoislands can be imagined to act as ldquoimpurity sinksrdquo for
further contaminants approaching the surface yielding a greater but more ordered surface
roughening with respect to the uncontaminated silicon surface
Circular arrays of nanoislands
We used the voids described in the previous chapter formed during oxide layer
decomposition as nanoscopic masks to test the validity of the hypothesis outlined above
The small circular areas uncovered during oxide decomposition can act as ldquoskating rinksrdquo
where contaminants and diffusing silicon atoms can move freely over the surface while
the oxide layer all around and higher than the void area as well as being chemically
bonded to the surface silicon atoms will hinder surface diffusion Furthermore the oxide
passivating layer will at least in some cases prevent contaminating molecules from
adsorbing and decomposing on the surface In this way selective nanoisland growth inside
void areas can take place
The experimental procedure consisted of the following steps
1) An oxide-passivated sample was heated for 10 minutes at 800degC Oxide decomposition
began in correspondence with surface defects and spread laterally creating naked areas
of silicon
2) While the void enlarges the same defect that initiated oxide decomposition behaved as
an attractor for moving silicon atoms and nucleated the growth of a nanoisland at the
void centre
3) The void surface was exposed to contaminating molecules
4) The sample was then heated again at 800degC During the second heating the void
continued to enlarge and new nanoislands were created The freshly uncovered silicon
surface provided mobile silicon atoms to feed nanoisland growth
72 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
5) When the heating continued for long enough all the oxide was desorbed and a clean
silicon surface covered by circular groups of nanoislands was obtained
Figure 66 shows the different stages of the process
We thus heated an oxidized sample at 800degC for 10 minutes in UHV to create multiple
voids in the oxide layer After this we exposed the sample to the contaminating molecules
Then we further annealed the surface at the same temperature to promote island growth
The first time we observed nanoisland growth in the void area was by chance after heating
for a second time samples which had been stored for 2 weeks in vacuum Even at pressures
as low as 10-11 mbar some molecules will adsorb onto the surface slowly creating defects
and thus potential nanoisland nucleation centres It is possible to observe with STM that
the silicon crystal surface looks very clean just after a high temperature flash but even
after only a few days randomly adsorbed species will be observable on the surface
Unfortunately it is not possible to identify these species simply from in situ STM
measurements
Mass spectrometry measurements show that the residual gas contaminants in vacuum are
usually He Ar H2 CH4 CO CO2 and N2 [11] Some of these molecules such as
hydrogen will stick to the silicon surface at room temperature but will simply desorb
during annealing without nucleating nanoislands Others like CH4 or CO2 will not react
with the surface to generate nanoislands as demonstrated previously The best candidate
for contamination of samples stored in UHV seems thus carbon monoxide has the ability
321
4 5
Fig 66 Selective nanoisland growth within oxide voids See text for details
VPalermo 73
as shown before to nucleate nanoislands although co-adsorption mechanisms can not be
ruled out Fig 67 (left) shows the circular groups of nano-islands obtained after heating
void-covered samples stored for 2 weeks in UHV Small islands are observed in the former
void area The void itself has enlarged slightly uncovering a clean oxide-free area of the
silicon surface Apart from the central island created during the first period of heating the
surrounding ones have dimensions decreasing from the void border to the inner area of the
void
This confirms that most of the material needed for island growth comes from the silicon
atoms diffusing from the freshly uncovered areas around the original void The outer
nucleating centres are nearest to the silicon atom source and thus generate larger islands
For a faster process we directly exposed the void to a significant concentration of CO
Fig67 (centre) shows the islands grown after a 180 Langmuir exposure and subsequent
heating The islands are preferentially located around the original void perimeter the rest
of the original oxide-free surface remaining untouched This suggests that although the
whole surface is exposed to CO the gas only interacts chemically with the surface at the
SiSiO2 interface around the oxide-free void perimeter The SiSiO2 interface thus provides
preferential nucleation sites for nanoisland growth in the successive thermal annealing
step This recalls a similar phenomenon observed in the catalytic oxidation of CO on the
Pt(111) surface where CO molecules reacted with adsorbed O species only along the
perimeters of oxygen islands [10]
On repeating the CO exposure and thermal annealing steps a second circle of nano-islands
Fig67 Left circular area of nanoislands obtained from a sample stored 2 weeks in UHV 500x500 nmCentre nanoisland circle obtained after exposing the voids to 180 L of CO and heating for 10 min at800degC 400x400 nm Right two concentric circles obtained with further CO adsorption followed byheating 650x650 nm
74 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
was formed around the newly-expanded surface void perimeter (fig67 right) This
process can in principle be repeated again to obtain a series of concentric groups of
nanoislands the only limit depending on the minimal distance between neighbouring
expanding voids We have thus observed that contaminants adsorbed at low pressure over
long periods of time generate random island growth over the whole void area while CO
molecules adsorbed at relatively high pressures over short periods yield selective
nanoisland growth at the former void border Several theories can explain the different
results CH4 and CO2 as mentioned above did not nucleate islands when when the silicon
surface was exposed to them for short periods However the lower surface mobility of
these contaminants and the greater time available for reaction with the surface could be
the cause of nucleation of SiC clusters over the whole oxide-free void area and not only at
its border The coadsorption of water another UHV residual gas with CO on the naked
silicon areas could perhaps allow a surface reaction similar to that between CO and the
SiSiO2 interface during CO exposure Another explanation for the formation of these
random nucleation sites within the voids could be the coadsorption of residual carbon-
containing species present in the UHV system with hydrogen the major residual gas in
stainless steel UHV systems
The possibility of decorating silicon oxide borders with nanoislands using the high
mobility and selective decomposition of CO molecules suggests interesting applications
for nanofabrication processes that will be discussed below
Silicon nanowire creation on Si(100)
We performed most of the experiments described above on silicon (111) 7x7
reconstructed because this surface is easily prepared in UHV and above all because it is
isotropic and has no preferential directions for atom diffusion The 2x1 reconstruction of
silicon (100) even though it has a much simpler unit cell than Si(111) 7x7 shows a more
complicated morphology with lots of monoatomic and biatomic steps and dimer rows
parallel to each other The orientation of the dimer rows changes by 90deg on alternate
atomic layers Surface diffusion in this case is much more complicated because atoms will
experience the effects of the surface anisotropy and diffuse preferentially along or across
the dimer rows The diffusion energy of a silicon atom as example is 06 eV along a
dimer row and 085 eV across different dimer rows This while being a complication
VPalermo 75
offers interesting possibilities for the creation of ordered structures The monoatomic steps
and the dimer rows can for example act as templates for the formation of elongated
structures of composition similar to the islands described previously
On heating a Si(100) surface oxide decomposition and void growth takes place as
described for Si(111) In some cases the void shape reflects the surface symmetry
especially for very thin oxide layers and the voids often have a slightly squared shape
Nanoisland growth is also influenced by the substrate symmetry and the islands look
rectangular as irregular clusters surrounded by concentric patterns of monoatomic steps
The whole structure reminds vaguely a ldquozigguratrdquo a kind of stepped pyramid found in
Mesopotamia (fig68 inset)
Apart from nanoisland growth in some cases we observed spontaneous formation of
nanowire-like SiC structures on the Si(100) surface The wires were generated after
exposing a Si(100) surface in vacuum to traces of CO followed by sample annealing at
800degC for 15 minutes The wires exhibit lengths ranging from 10 to 100 nm and average
widths of ca 5 nm Although being randomly positioned on the surface they are perfectly
aligned along the crystal axes of the substrate (Fig68)
Several descending monoatomic steps can be observed around each nano-line and nano-
island Very often one of the line extremities coincided with an island
Nanoline
Nanoisland
Fig 68 A Si(100) surface covered by nanoislands and nanowires The monoatomic steps present on the surface are visible The two insets show a typical nanoisland and a nanowire obtained on this surface The typical ldquodimer-row ldquostripes are visible along the sides of the nanowire
76 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Couples of parallel lines starting from the same nano-island were also observed At high
resolution (fig 69) the lines showed a lamellar periodic structure with a periodicity of
around 4 nm and frequent defects Scanning Tunneling Spectroscopy measurements
showed no significant difference between the line and the surrounding silicon surface
The proposed formation mechanism for this kind of structure involves two stages in the
first one organic contaminating molecules adsorb and diffuse on the surface from the
vacuum environment stopping preferentially at surface singularities including steps and
defects
Several substances such as atomic carbon carbon monoxide ethylene and fullerene
decompose when adsorbed on silicon surfaces at temperatures above 500degC and create
localised defects on the surface inducing strain deformations and a metastable surface
reconstruction [12]
If these molecules remain attached to the steps they can diffuse rapidly along step border
until they meet a line and decompose In this way long lines of SiC defects parallel to the
surface steps can be produced In the second phase the clusters of silicon carbide can
-005
0
005
01
015
02
025
03
035
04
-3 -2 -1 0 1 2 3
V
dId
V lt
dId
Vgt
linesilicon
Current- Voltage behaviour measured with STS Fourier analysis of wire periodicity
asymp 4 Aring
High resolution image of a line
Fig 69 Top high resolution image of a nanowire showing the lamellar structure Fourier analysis (bottom left) shows that the lamellae periodicity is sim 04 nm bottom right the IV characteristics measured with the STM tip on the line and the silicon surface
VPalermo 77
25-600degC
Final linear structure
600-800degC
Contaminant Silicon
Fig 610 Schematic representation of nanowire formation mechanism
locally inhibit the spontaneous surface diffusion of silicon atoms acting as templates for
the reorganisation of surface steps (fig 610)
The final morphology shows disordered lines and islands each surrounded by a complex
pattern of silicon monatomic layers
At 800degC silicon atoms are highly mobile on the surface Due to atom diffusion the
surface reorganizes with a step-flow mechanism The presence of a fixed line of
contaminants blocks atom diffusion and step flow generating the monoatomic steps
around each wire (fig 611) SiC clusters act as lsquopinning sitesrsquo on silicon atom surface
diffusion and can be purposely used to direct the localised growth of nano-islands [13]
Although the possibility of using methanol and carbon monoxide as nano-island precursors
has been demonstrated the exact nature of the substances nucleating nano-wire growth is
still unknown
Further experiments will be needed to confirm these hypotheses as well as to identify the
contaminating species nucleating nanowire growth Carbon monoxide is a main candidate
for nanowire growth but the real nucleation process is likely to be a complex one
78 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Step flowblocked
Fig 611 Representation of step flow blocked by the presence of a nanowire 105x105nm
Bibliography
[1] Surface-Chemistry Of Silicon Waltenburg HN Yates JT Chemical Reviews 95 (5)
1589-1673 Jul-Aug 1995
[2] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[4] Growth Of Si On The Si(111) Surface Lanczycki Cj Kotlyar R Fu E Yang Yn
Williams Ed Das Sarma S Physical Review B 57 (20) 13132-13148 May 15 1998
[5] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
VPalermo 79
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990
[6] Nanoscale roughening of Si(001) by oxide desorption in ultrahigh vacuum Gray SM
Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology B 14 (2)
1043-1047 Mar-Apr 1996
[7] Step Motion On Crystal Surfaces Schwoebel Rl Journal Of Applied Physics 40 614-
618 Feb 1969
[8] Scanning Tunneling Microscopy Investigation At High Temperatures Of Islands And
Holes On Si(111)7x7 In Real Time Evidence For Diffusion-Limited Decay
Hildebrandt S Kraus A Kulla R Wilhelmi G Hanbucken M Neddermeyer H Surface
Science 486 (1-2) 24-32 Jul 1 2001
[9] Methanol Adsorption On Si(111)-(7x7) Investigated By Core-Line Photoemission And
Mass Spectrometry Of Photodesorbed Ions Carbone M Piancastelli Mn Zanoni R
Comtet G Dujardin G Hellner L Surface Science 370 (1) L179-L184 Jan 1 1997
[10] Atomic And Macroscopic Reaction Rates Of A Surface-Catalyzed Reaction
Wintterlin J Volkening S Janssens Tvw Zambelli T Ertl G Science 278 (5345)
1931-1934 Dec 12 1997
[11] Redhead PA Hobson JP Kornelsen EV The Physical Basis Of Ultrahigh
Vacuum Chapter 12 (Chapman amp Hall London 1968)
[12] The Si(001) C(4 X 4) Surface Reconstruction A Comprehensive Experimental
Study Norenberg H Briggs Gad Surface Science 430 (1-3) 154-164 Jun 21 1999
[13] Production Of Nanostructures Of Silicon On Silicon By Atomic Self-Organization
Observed By Scanning Tunneling Microscopy Jones D Palermo V Applied Physics
Letters 80 (4) 673-675 Jan 28 2002
80 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 81
7 Conclusions and proposals for future work It is very difficult to foresee the future of nanoscience and nanotechnology even over the
next few years exciting discoveries are continuously made opening up new possibilities
and progress in this sector is now one of the fastest and most varied So it is impossible to
predict whether computers will in the future be made with nanowires nanotubes single
electron devices or some kind of quantum-based transistor What does seem clear
however is that silicon devices can still further extend their levels of miniaturization to
remain competitive for at least the next ten years [1] During this period new emerging
technologies will reach scientific maturity and arrive at the production lines
Even when new technology prototypes do become available for full-scale production the
astronomical cost of changing from silicon to new technologies will be a major concern
Moreover a vast amount of extremely detailed knowledge is already available on all
aspects of silicon technology Thus the possibility of integrating innovative
nanotechnologies with standard CMOS silicon technology already seems an attractive
prospect [2]
For this and other reasons we have focussed our research efforts on surface phenomena on
silicon wafers which could be potentially useful for the development of silicon-compatible
nano-devices
Some important conclusions can be drawn from the work described in this thesis
bull Silicon surfaces of different crystal faces were studied at atomic resolution in ultra-
high-vacuum following transformations induced on the surface by chemical (etching)
and physico-chemical (organic contamination heating) treatments
82 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull The presence of surface contaminants induces at high temperature the growth of
nanometre-sized islands and lines on silicon
bull Final island densities and dimensions strongly depend upon the chemical nature of the
contaminant molecules and the treatments used Molecules interacting only weakly
with silicon will desorb upon heating whereas molecules which decompose easily on
Si will give widespread random nano-island nucleation Between these two extreme
cases complex behaviours of diffusion and subsequent decomposition are possible
bull Nanoscopic naked silicon voids were produced in oxide layers through carefully-
controlled thermal annealing It was found that void density and size depends upon
oxide purity surface contamination levels and annealing conditions
bull The growth of nano-islands and nano-voids was exploited simultaneously and the
voids used as nanoscopic masks to control the positions of growing islands Complex
silicon-on-silicon nano-structures were obtained with multi-step growth processes
bull Finally the use of macroscopic treatments and simple physical processes to produce
large quantities of nanometre-sized structures on silicon surfaces was demonstrated
These processes and the structures produced are of potential interest to the
microelectronics industry
We showed in previous chapters how it is possible to modify silicon surfaces using
chemical and physical methods The use of self-organization for the production of Si-on-Si
20 nm
Fig 71 Left SEM image of field emitter arrays made by conventional lithography [3] Center 3-D STM image of a self-organized void with a central nano-island Right Lateral schematic view of a field emitter device
VPalermo 83
nano-structures has the advantage with respect to other emerging techniques of being
silicon compatible and thus is a potential candidate for its implementation in the
production of new devices Furthermore this technique can create billions of strucures on a
wafer in only a few minutes being thus applicable to large-scale production
Even when ordered structures are obtained these methods still show a poor reproducibility
and control Chemically modified surfaces look very disordered at high magnification
with trenches and holes etched into the surface and progress has recently been made in
producing very small structures electrochemically [4] Oxide decomposition with
nanoisland growth in UHV can also yield ordered structures The void holes with
nanoislands located at their centres although quite variable in dimensions are more similar
to engineered devices than to a spontaneous random surface-roughening process In fact
there is a surprising similarity between the nanovoid-nanoisland structures and well-known
commercially available devices called field emitter arrays (FEA) These are small tips used
as electron guns in various kinds of displays and other devices Both the commercial
devices and our self-organized structures consist of a protruding tip at the centre of
microscopic holes on an insulating layer covering a semiconductor surface (see fig71)
Of course the order and size uniformity of commercial FEAs is much better than that
obtained with our process but it is encouraging to see that a self-organized structure
obtained with simple heating has a very similar surface topography and composition of
commercial working devices obtained after many expensive stages of optical lithography
deposition and etching
Nano-island density and size can to some extent be controlled and different
contaminants can be adsorbed onto the surface to modify the growth process In the future
with a better understanding of diffusion dynamics and growth mechanisms a multi-stage
fabrication process can be envisaged where purposely-designed molecules are adsorbed
onto silicon surfaces to diffuse organize in ordered arrays and favour surface
reorganization at the nanometre scale The same molecules could be used to deliver
dopants to obtain for example highly conducting nanostructures on otherwise undoped
highly resistive silicon The selective formation of nanostructures at the SiSiO2 border
can be used to scale down mesoscopic patterns to the nanometric level A silicon oxide
layer can be easily patterned on the surface with normal optical lithography and then
nanostructures can be built following the pattern edge In this way the production of very
small nanowire based devices may be possible (fig 72)
84 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
D
Gate
S
D
S
SiO2
SiO2
2 Heating nanowire creation
and oxide removal
3 Source-drain deposition by conventional lithography
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
4 Insulating layer (oxide) and gate (metal) deposition
Fig 72 Schematic drawing of the construction of a self-organized nanowire-based transistor
Surface diffusion and decomposition of molecules can be sensitive to even smaller surface
features The nanoscopic lines formed on Si(100) are very regular and neat if their
nucleation and growth mechanism could be controlled ordered networks of nanowires
could be made on silicon
VPalermo 85
The techniques developed here are still experimental ones the careful control and
reproducibility of each step is still an issue and further studies will be needed to clearly
understand the dynamics of formation of these structures Nonetheless the results obtained
so far open up the possibilities of producing novel devices such as for example nano-
transistors (see scheme in fig73 where the mechanism described in fig 72 is applied for
the fabrication of a device array) The possibility of using the border of a lithographic
pattern to draw nanolines can in principle allow the creation of nanodevices using normal
IC manufacturing techniques
The results obtained and their description in this thesis are not the first examples of the use
of self-organization for the production of ordered structures and will certainly not be the
last In our opinion the most important outcome of these findings is that it is possible to
use surface diffusion to overcome the inherent limits of lithographic techniques in
microdevice production processes and to force properly selected molecules to react with
the surface in a spatially differentiated manner at particular sites thereby creating ordered
series of nanostructures The role of the surface dynamics of the substrate atoms
themselves is extremely important in this process
Outstanding results have been and continue to be obtained in nanoscience and
nanotechnology research promising the emergence of new production technologies Sooner
or later those emerging technologies will have to face the issues of cost and compatibility
with the enormous investment made in existing technologies and processing facilities The
commercial production of nanostructures based on surface diffusion processes could in
principle be cheap simple and compatible with existing technology Moreover the
processes of diffusion in bulk silicon is well-known in the microelectronics industry which
has both the know-how and the equipment to better understand and exploit the surface
diffusion processes on silicon surfaces studied here
Extremely important and innovative results are often obtained by exploiting very simple
ideas and in our opinion physical processes such as surface diffusion used to create
complex nanostructures on silicon surfaces is a prime example
86 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
2 Heating nanowire creation and
oxide removal
3 Source-drain deposition by conventional lithography
4 Insulating layer (oxide) and gate (metal) deposition
Fig 73 Schematic macroscopic outline of the construction of a self-organized nanowire-based transistor
VPalermo 87
Bibliography
[1] P Gargini Intel Technology Coordinator ldquoFrom Microelectronics To
Nanotechnologiesrdquo Invited Campus Colloquia Lecture CNR Research Area Bologna
February 6th 2003
[2] Toward A Hybrid Micro-Nanoelectronics Cerofolini Gf Ferla G Journal Of
Nanoparticle Research 4 (3) 185-191 Jun 2002
[3] Recent Progress In Field Emitter Array Development For High Performance
Applications Temple D Materials Science amp Engineering R-Reports 24 (5) 185-239
Jan 25 1999
[4] Silicon Dioxide Micropillars For Sieving Fabricated By Macroporous Silicon-Based
Micromachining Izuo S Ohji H French Pj Tsutsumi K Kimata M Sensors And
Materials 14 (5) 239-251 2002 Electrochemical Etching In HF Solution For Silicon
Micromachining Barillaro G Nannini A Piotto M Sensors And Actuators A-Physical
102 (1-2) 195-201 Dec 1 2002
88 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 89
Acknowledgements
This work would not have been possible without the help and guidance of Dr Derek Jones
he introduced me to the use of STM and other techniques found the funding for this PhD
and helped me with the thousands of small and big problems I had to face during my
research I am grateful to Prof Alberto Ripamonti too for being the supervisor of this
thesis Thanks are also due to Dr Giancarlo Seconi director of ISOF-CNR where most of
this work was carried out Financial support from the Italian National Research Council
(CNR) is also gratefully acknowledged
My passion for surface science began when Fabio Biscarini showed me that it was possible
to really see the atoms and for this I will always thank him
During these years I had the luck to collaborate with many good scientists among those I
would like to cite Enrichetta Susi Massimo Cocchi and Anna Mazzone from CNR
Claudio Zannoni Anna Cavallini Daniela Cavalcoli and Antonio Castaldini from the
University of Bologna Sergio Pizzini and Maurizio Acciarri from the University of Milan
One of the most stimulating periods of my PhD was the one I spent at the Steacie Institute
for Molecular Sciences in Ottawa working with Robert Wolkow Dan Wayner Greg
Lopinski and Peter Kruse I thank them for discussing together many exciting scientific
ideas and trying to put some of them into practice
Special thanks go to my colleagues Paolo Samorigrave and Stefano De Cesari for long and
useful discussions on Science Life and their interactions often with the help of a good
pint
Finally I would like to thank Vassilia Gaetano Simone Claudia Silvia Letizia
Alessandro Angela and Sebastien even though not directly involved in this work they
have been over these three years a continuous source of happiness and support
Bologna March 2003
90 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
SCIENTIFIC PUBLICATIONS bull Lateral diffusion of titanium disilicide as a way to contacting
hybrid Si-organic nanostructures Palermo V Buchanan M Bezinger A Wolkow RA APPLIED PHYSICS LETTERS 2002 v81 p 3636 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v6 issue 20) bull Surface modifications in Si after Rapid Thermal Annealing Castaldini A Cavalcoli D Cavallini A Jones D Palermo V Susi E JOURNAL OF THE ELECTROCHEMICAL SOCIETY 2002 v 149 pG633 bull Production of nanostructures of silicon on silicon by atomic self-
organisation observed by scanning tunnelling microscopy Jones D Palermo V APPLIED PHYSICS LETTERS 2002 v 80 p 673 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v5 issue 5) bull Nucleation of nanostructures from surface defects on silicon Palermo V Jones D SOLID STATE PHENOMENA 2002 v 82-84 p 687 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo MATERIALS SCIENCE AND ENGINEERING B 2002 v88 (2-3) p220 bull Morphological changes of Si[100] surface after treatment with
concentrated and dilute HF Palermo V Jones D MATERIALS SCIENCE IN SEMICONDUCTOR PROCESSING 2001 v 4 p 437 bull Electrical and structural properties of processed silicon surfaces Susi E Cavallini A Castaldini A Cavalcoli D Jones D Palermo V ldquoRECENT RESEARCH DEVELOPMENTS IN VACUUM SCIENCE amp TECHNOLOGYrdquo 2001 v 3 p189 bull Numerical Solutions of the Stochastic Equations of Crystal
Growth Mazzone A M Palermo V INTERNATIONAL JOURNAL OF MODERN PHYSICS C 2000 v 11 Part 1 p195-204 bull Advances in silicon surface characterisation using light beam
injection techniques
VPalermo 91
Acciarri M Pizzini S Simone G Jones D Palermo V MATERIALS SCIENCE AND ENGINEERING B 2000 V73 (1-3) p 235 - 239 bull Abrupt orientational changes for liquid crystals adsorbed on a
graphite surface Palermo V Biscarini F Zannoni C PHYSICAL REVIEW -SERIES E- 1998 V 57 NUMBER 3A p R2519-R2522 Oral presentations bull The use of oxide desorption and surface diffusion for the
creation of silicon on silicon nanostructures Palermo V Jones D 1ST NATIONAL WORKSHOP ON CURRENT TRENDS IN NANOTECHNOLOGIES Catania (Italy) Feb 2002 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo V E-MRS Spring Meeting Strasbourg (France) June 2001 bull STM study of surface transformations on silicon during UHV
annealing Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999 bull Advances in silicon surface characterisation using light beam
injection technique Acciarri M Pizzini S Simone G Jones D Palermo V E_MRS SPRING MEETING June 1999 Posters bull Spontaneous nano-wire growth on silicon Palermo V Jones D SMARTON workshop Leuven (Belgium) October 2002 bull Ordered circles of nano-islands on silicon from CO adsorption Palermo V Jones D TRENDS IN NANOTECHNOLOGY 2002 Santiago de Compostela (Spain) September 2002 bull Production of nanostructures of silicon on silicon by atomic self-
organisation Palermo V Jones D EUROMAT Rimini (Italy) June 2001 bull Etching holes and anisotropic corrosion on silicon [100] Palermo V Jones D Susi E Asoli B SILICON WORKSHOP Genova (Italy) February 2001
92 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull Morphological and electrical characteristics of damaged silicon surfaces
Susi E Castaldini A Cavalcoli D Cavallini A Jones D Palermo V SILICON WORKSHOP Genova (Italy) February 2001 bull Effect of HF etching on the roughness of a silicon surface Palermo V Jones D NATIONAL CONFERENCE ON PHYSICS OF MATTER Genova (Italy) June 2000 bull Effect of Native Oxide Desorption upon the Surface Morphology of Si[100] by STM and LEED Palermo V Jones D SILICON WORKSHOP Genova (Italy) February 2000 bull Characterisation of silicon surfaces for microelectronics through STM measurements Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999
VPalermo 93
Contact Vincenzo Palermo ISOF- Institute for Organic Synthesis and Photoreactivity Via Gobetti 101 40129 Bologna ITALY Tel +39-051-6398336 Fax +39-051-6398349 Mail palermoisofcnrit
94 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Index
1 Introduction nanotechnology and the future of computers
2 Silicon surfaces
3 STM and other surface analysis techniques
4 Surface modification of silicon in liquid Nano-hole creation
5 Surface modification of silicon in vacuum void creation and oxide desorption
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and
nano-lines
7 Conclusions and proposals for future work
6 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 7
1 Introduction nanotechnology and the future of computers The motivations of nanotechnology research
In 1965 an electronic engineer named Gordon Moore one of the future founders of Intel
noted that the performance of computers and their complexity doubled every 18 months
and foresaw that computer power would continue to grow exponentially over the following
years
This prevision quite provocative for its time actually came about and gained the name of
ldquoMoorersquos Lawrdquo and continues to hold for the trends of todayrsquos computer industry Since
1965 the number of transistors present in an integrated circuit (IC) has increased from
several hundred to more than ten million and the minimum size of transistor elements has
shrunk from several millimetres to asymp130 nanometres (fig 11)
Devices of such tiny dimensions are actually fabricated using lithographic techniques
where light is passed through an optical mask to react with a photo-sensitive layer (resist)
Fig11 Evolution of the number of transistors present on commercial computers [1]
8 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
on the silicon wafer This resist is then selectively removed and used as a mask for
processing the silicon surface (fig 12 left) The maximum resolution attainable depends
upon the wavelength used and current technology is near to its intrinsic resolution limit
On the other hand there is strong scientific and economic demand for further development
in IC miniaturization to obtain more powerful and complex computers Besides every-day
life applications more powerful computers are fundamental for much scientific research
such as climate change tracking genome sequencing and fluid dynamics Increased
miniaturization is also fundamental for reducing power which has to be dissipated by the
chips which run at progressively higher frequencies Energy consumption by
microelectronic devices is already an issue and represents one of the main obstacles for
the continuing growth in wireless communication (cell phones portable computers CD
and DVD players digital cameras etc)
Thus it is expected that new production methods different from current lithographic ones
will be developed methods which allow modification of a surface well below the 100 nm
limit and even down to single atom manipulation Techniques such as Scanning
Tunnelling Microscopy and Atomic Force Microscopy are already capable of moving
single atoms (see fig 12 right) but unfortunately building a working nanodevice in this
way would take a very long time and these techniques are difficult to apply to large scale
production
Nowadays thousands of researchers are working in the nanotechnology field towards a
new generation of microelectronic devices Several possible solutions are competing for
tomorrowrsquos computer architecture and there is still no clear winner It is likely that the
final solution will be the combined use of different techniques and components (including
molecules nanowires and nanodots) as they become available together with conventional
Fig 12 Left conventional litographic process [2] Right atomic manipulation of iron atoms on copper [3]
VPalermo 9
silicon technology
Below is a brief summary of the most recent developments in nanotechnology and
nanoscience
Actual trends in nanotechnology
Perhaps the most fascinating idea for nanodevice construction is to use one single
molecule working as a complete device The first molecular diodes (ie molecules
conducting current only in one direction) were created in 1997 in 1999 a molecular fuse
and a molecular transistor were demonstrated although there was no possibility of wiring
these devices to external contacts In April 2001 James Heath and his group at UCLA
fabricated an array of overlapping crossbars and placed a small molecule of rotaxane
between each crossbar (fig 13 left) This composite molecule is made up of two
component parts the main rod-like molecular axis and a mobile ring ldquothreadedrdquo on it like a
bead on a necklace and can function as a molecular switch A working 16-bit memory
circuit was constructed using these molecules For a brief review of these works see [4] In
June 2002 a single molecule transistor was built by connecting an organic molecule to two
metal contacts the molecule contained one or two atoms of a transition metal (cobalt or
vanadium) forming the active region of the device supported by an organic backbone [5]
Fig13 Working nanodevices Left schematic representation of rotaxane molecules between crossed nanowires [4] Right SEM image of semiconductor nanowires forming a small circuit [9]
10 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another approach to nanodevice fabrication has become possible through the discovery of
carbon nanotubes which were observed for the first time in 1991 by a Japanese electron
microscopist studying the material deposited during arc-evaporation synthesis of fullerenes
[6] They consist of a graphite-like carbon seamless cylinder with a diameter of several
nanometers and lengths of up to a millimetre Carbon nanotubes are very stable can
behave as metals or semiconductors and can host other molecular or ionic species thus
modifying their electrical behaviour In 2001 Avouris and his group reported the first
circuit made with a single nanotube [7] A few months later Cees Dekker presented a
nanotube-based transistor able to amplify an input signal by a factor of ten and built
several logic circuits using these nanotube transistors [8]
One problem with carbon nanotubes is that it is very difficult to control their electronic
properties ie their metallic or semiconducting behaviour An alternative to carbon
nanotubes are semiconductor nanowires Silicon nanowires can be made using a laser to
vaporize the silicon together with a metal catalyst like iron or gold The vapour condenses
in nanosized drops of silicon and metal from which the wires slowly grow out as more
silicon is adsorbed In 2001 a group at Harvard University [9] created a transistor by
crossing two different nanowires After this the same group arranged four nanowires in a
noughts and crosses grid creating something like a 4-bit memory (fig 13 right) Even
metallic nanowires made of platinum and silver can be used in a crossed configuration to
store information [10]
There are some issues common to all these new technologies though First it is difficult to
imagine these methods applied to large-scale production Up to now the insertion of a
molecule between two electrodes is an occasional lucky event while nanotubes and wires
have to be positioned on the surface creating the appropriate contacts on them manually
The large-scale production of integrated circuits using these building blocks will not be
straightforward and does not seem imminent
Another issue is of an economic and not a scientific or technological nature Since 1965
the cost of IC manufacturing plants has sky-rocketed If the cost of semiconductor
production plants continues to rise exponentially in a few years such plants will cost up to
$20 billion This is a sizeable investment even for large companies such as IBM or Intel
For this reason it is likely that IC companies will resist changing to completely new
technologies closing down their existing plants As it is clear that silicon will remain the
VPalermo 11
fundamental raw material of the IC industry for the foreseeable future nanotechnology
developments for microelectronics will need to be silicon-compatible In these early days
of nanotechnology the most valid approach would seem to be the addition of molecular
functions to existing silicon technology ndash using the latter as a foundation on which to build
Fabrication of self-organised structures on silicon
The possibility of using the phenomenon of atomic or molecular self-organization to create
nanostructures on silicon has already been demonstrated The clean silicon surface shows
in some cases a high degree of order and complex surface reconstruction as will be
described later Several different ordered structures form spontaneously on this surface
such as series of monatomic steps or boundaries between reconstructed areas It has been
demonstrated that it is possible to use these structures to fabricate ordered nanodots and
nanolines on the surface [11] More recently well-defined nanometric patterns have been
obtained with selective etching of silicon using nitric oxide [12]
In this study the possibility of creating different types of nanostructures on the silicon
surface is explored Methods had to be developed which were
- Simple They must not need complex masks or lithographic steps to create the
structure but rather exploit self-organisation phenomena
- High resolution the silicon surface modifications should be on a scalelength of below
100 nm
- Fast billions of nanostructures have to form over the whole surface simultaneously to
be compatible with large-scale production requirements
- Cheap they must not require expensive equipment (such as e-beam lithography) but
exploit simple chemical andor physical treatments to produce nanostructures on the
silicon surface
In Chapter 2 the main characteristics of silicon are described Chapter 3 provides a
summary of the techniques used for this research Chapter 4 examines the chemical etching
of silicon in different liquid environments and the effects of this etching on the surface at a
nanoscopic level with the creation of nanoholes
12 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Chapter 5 describes the growth in ultra-high vacuum (UHV) of nanoscopic voids and
islands on the silicon surface and the effect of surface oxide on this growth
Chapter 6 discusses the modification of silicon surfaces in UHV following the adsorption
of molecules and thermal heating to produce nanoislands and nanolines on silicon
The overall conclusions of our work are summarised in Chapter 7
Finally we will give some conclusions based on the results obtained and discuss possible
applications of the methods developed
Bibliography
[1] From wwwintelcom
[2] From wwwsematechorg
[3] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
[4] Molecules Get Wired Service RF Science 294 (5551) 2442-2443 Dec 21 2001
[5] Coulomb Blockade And The Kondo Effect In Single-Atom Transistors Park J
Pasupathy AN Goldsmith JI Chang C Yaish Y Petta JR Rinkoski M Sethna JP
Abruna HD Mceuen PL Ralph DC Nature 417 (6890) 722-725 Jun 13 2002 Kondo
Resonance In A Single-Molecule Transistor Liang WJ Shores MP Bockrath M Long
JR Park H Nature 417 (6890) 725-729 Jun 13 2002 Nanotechnology - Electronics
And The Single Atom De Franceschi S Kouwenhoven L Nature 417 (6890) 701-702
Jun 13 2002
[6] Smallest Carbon Nanotube Ajayan PM Ijima S Nature 358 (6381) 23-23 Jul 2 1992
[7] Carbon Nanotube Inter- And Intramolecular Logic Gates Derycke V Martel R
Appenzeller J Nano Letters 1 (9) 453-456 Sep 2001
[8] Logic Circuits With Carbon Nanotube Transistors Bachtold A Hadley P Nakanishi T
Dekker C Science 294 (5545) 1317-1320 Nov 2001
VPalermo 13
[9] Logic Gates And Computation From Assembled Nanowire Building Blocks Huang Y
Duan XF Cui Y Lauhon LJ Kim Kh Lieber CM Science 294 (5545) 1313-1317
Nov 9 2001
[10] Formation And Disappearance Of A Nanoscale Silver Cluster Realized By Solid
Electrochemical Reaction Terabe K Nakayama T Hasegawa T Aono M Journal Of
Applied Physics 91 (12) 10110-10114 Jun 15 2002
[11] Fabrication And Integration Of Nanostructures On Si Surfaces Ogino T Hibino H
Homma Y Kobayashi Y Prabhakaran K Sumitomo K Omi H Accounts Of Chemical
Research 32 (5) 447-454 May 1999
[12] Ultrafine And Well-Defined Patterns On Silicon Through Reaction Selectivity
Prabhakaran K Hibino H Ogino T Advanced Materials 14 (19) 1418-1421 Oct 2
2002
14 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 15
2 Silicon surfaces
The name silicon (silicio in Italian) comes from the latin word silex Amorphous silicon
was first isolated by Berzelius in 1824 by reaction of potassium with silicon tetrafluoride
Thirty years later the first crystalline silicon was prepared Silicon makes up 25 of
earthrsquos crust and is the second most abundant element after oxygen Elemental silicon is
not found in nature occurring as silicon oxide (sand quartz amethyst flint etc) or
silicates (asbestos clay mica etc) Perhaps no other element and its compounds has such
a wide range of uses Silicon compounds such as sand and clay are used in the building
industry as refractory materials for high-temperature applications and for enamels and
pottery Silica is the main component of glass silicon carbide is an important abrasive and
silicones are commonly used polymers and lubricants
Here the most interesting use of silicon of course is for the production of
microelectronics devices For this application silicon of high purity (999999) and of
high crystallinity is needed Table 21 lists some of the physical characteristics of silicon
High purity polycrystalline silicon is produced by the reaction of gaseous trichlorosilane
with hydrogen in a furnace Then to prepare a single-crystal of silicon the so-called
Czochralski method is commonly used
Polycrystalline silicon is melted in a quartz furnace at 1415degC in an argon atmosphere
Then a seed of single-crystal silicon is lowered into contact with the melt and slowly
pulled out In this way the crystal grows and a crystalline cylindrical ingot several metres
long is created from the initial seed
After cooling down the ingot is sliced into thin silicon wafers The wafer surfaces are
polished using a counter-rotating lapping machine in an Al2O3 slurry until the surface is
very flat and shiny ready for the lithographic processes
16 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another way to obtain single crystal silicon is the Floating Zone (FZ) method in which a
silicon cylinder is slowly passed through a heating ring The area inside the ring melts and
solidifies smoothly crystallising as it comes out of the ring yielding a single silicon crystal
Microelectronic devices are built on the silicon surface which is the surface of interest
here Unfortunately silicon surfaces are normally quite dirty and uneven at the atomic
scale Atmospheric oxygen and humidity react with silicon surfaces creating a thin layer of
oxide (called ldquonative oxiderdquo) which is usually irregular and full of defects Different kinds
of contaminants also adsorb onto the surface These are usually small organic molecules
and microscopic dust particles A clean surface on exposure to the atmosphere is
completely covered with gas molecules in less than 10-9 seconds If the pressure is
reduced letrsquos say to 10-6 mbar this time increases to 1 second This is the reason why to
study a clean surface we have to work in UHV at pressures below 10-10 mbar
The atoms in the silicon crystal have a diamond-like structure each atom having 4 bonds
in a tetrahedral sp3 arrangement with bond angles of 10947 degrees At the crystal
surface some atoms will have non-bonding orbitals ldquodanglingrdquo in the vacuum ie sp3
orbitals with a lone electron which are highly reactive These orbitals are known as
dangling bonds To minimize surface energy the surface will reorganize by decreasing
the number of dangling bonds
Table 21 Physical data of silicon [1]
Atomic Weight 2809 Lattice constant (A) 543095 Crystal structure Face-centered cubic
(diamond) Melting point 1415 degC
Density (gcm3) 2328 Boiling point 2355degC Atomscm3 50E22 Minority carrier
lifetime (s) 25E-3
Dielectric Constant 119 Specific heat (Jg degC)
07
Breakdown field (Vcm)
~3E5 Thermal conductivity (Wcm degC)
15
Electron affinity x(V)
405 Vapour pressure (Pa) 1 at 1650degC 1E-6 at 900deg C
Energy gap (eV) at 300K
112 Reactivity Inert to acids Attacked by halogens and alkaline
solutions Intrinsic carrier
conc (cm-3) 145E10 Oxidation states +4 -4
Intrinsic Debye Length (microm)
24 Energy of a Si-Si bond (eV)
232
Intrinsic resistivity (Ω-cm)
23E5
VPalermo 17
Dangling bond densities and positions and thus the type of surface reconstruction will
depend upon crystal orientation as well as the temperature and kinetics of the system
Fig 21 shows a drawing of the main faces of a silicon crystal The angle α between any
(11n) face and the (100) face can be calculated from 2cos 2 += nnα The angle
between any (11n) face and the (111) face can be calculated from
)2(3)2(cos 2 ++= nnα
The chemistry and physics of the faces are very different a brief description will be given
for the most important orientations
Si (100)
On the (100) surface each atom has two Si-Si bonds connecting it to the bulk and two
dangling bonds pointing outward Surface energy is reduced by the dimerisation of the
surface atoms through overlap interaction of one dangling bond per atom forming rows of
dimers aligned along the (110) direction This is the well-known ldquo2x1rdquo reconstruction of
this silicon surface The symmetric dimers would make the silicon surface metallic but to
reduce surface stress the dimers tend to buckle and the surface is thus semiconductive It
took several years to understand that the dimers are buckled because at room temperature
Fig 21 Schematic view of the principal orientations of a silicon surface Surface atoms are white bulk atoms are black dangling bonds are gray [2]
18 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
they shift easily from one buckling direction to the other thus appearing symmetric under
STM observation Fig 22 shows an STM image of the 2x1 reconstructed surface
Even almost perfect (100) surfaces have a certain number of monoatomic steps and the
dimer rows on atomic layers are aligned at 90deg to those on adjacent layers Dimer rows are
thus perpendicular or parallel to the step When the dimers on the upper side of the step
are parallel to the step the step is called SA if they are perpendicular the step is called SB
Because of this symmetry SA step edges will be smoother compared to the more broken
fragmented edges of the SB steps
A common defect on the Si(100) surface is the presence of nickel contamination which
appear as missing dimers in STM images This type of contamination is so critical that
even if the silicon sample is only briefly brought into contact with stainless steel tools
(tweezers for example) the 2x1 reconstruction of the surface can be blocked
Silicon atoms can diffuse easily over the silicon surface as monomers and dimers
especially at elevated temperatures The anisotropy due to the 2x1 reconstruction causes a
difference in the diffusion energies of adsorbates over the surface Diffusion of these
silicon species along dimer rows for example will be much easier A list of diffusion
energies for monomers and dimers is given below [2]
Diffusion on Si(100) 2x1 Ed (eV) Monomers along dimer rows 06 Monomers across dimer rows 085 Monomer formation energy 18 Dimer along dimer rows 11 Dimer across dimer rows 15 Dimer formation energy 26 Dimer binding energy 076 Vacancies along dimer rows 17 Vacancies across dimer rows 19
SA
SB
Fig 22 STM image of a 2x1reconstructed silicon surfaceshowing the dimer rows and steps40x35 nm Nickel-induced defectsare visible as dark spots SA and SBsteps are indicated
VPalermo 19
So the diffusion energy for both monomers and dimers is nearly 40 greater if they have
to cross a dimer row This difference reduces to sim10 for vacancy diffusion
Si(113)
The (113) surface can be imaged as a sequence of alternating (100) and (111)-like
structures with two and one dangling bonds on alternate atoms respectively Interest in the
(113) surface is scientific as it has been used to study the energetics of the (100) and (111)
surfaces as well as for surface adsorption experiments
Si(100) surfaces can easily develop (113) facets
Si(111)
This surface besides being the first one imaged with STM with atomic resolution is one of
the most studied because it is the best cleavage face of silicon and because it shows one
of the most complex and elegant reconstructions in surface science
All Si-Si bonds in the silicon crystal are perpendicular to a (111) plane so this face will
have the lowest number of dangling bonds created per unit area In fact each Si atom on a
(111) surface shows a single dangling bond oriented perpendicular to the surface and
bonded to three back atoms These three bonds for each surface atom account for the great
chemical and physical stability of the Si(111) surface Surface energy is 009 eV Aring-2
compared to 015 eV Aring-2 for Si(100)
For energy minimization this surface reconstructs forming a huge 7x7 lattice cell
containing 102 atoms described by the Dimer-Adatom-Stacking fault model (DAS) For a
detailed description of cell structure see fig 23
The cell described by this model is very complex being composed of three kinds of atoms
adatoms rest atoms and corner hole atoms Furthermore a subsurface stacking fault is
present in one half of the cell making the two halves of the unit cell look different under
STM (Fig 24) It took 26 years of research to completely understand the exact structure of
the 7x7 reconstruction
Cleaving a silicon crystal along a (111) plane produces a metastable 2x1 reconstruction
the 7x7 reconstruction is easily obtained by flashing at high temperature in UHV At T gt
20 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
830degC a disordered 1x1 phase covers the surface Cooling down to 800degC leads to the
formation of the 7x7 phase If the cooling process is too rapid small 7x7 domains nucleate
and a disordered 1x1 phase is preserved between domain boundaries
Si(110)
Even though as mentioned before the (111) plane is the favoured cleaving plane of
silicon thin (100) commercial wafers will not break along this plane because the angle
between (100) and (111) is too far from 90deg (see table 22) Instead they will break along
the (110) plane because it is perpendicular to the (100) surface Each surface atom on
Si(110) has a Si-Si bond pointing downward one dangling bond pointing outward and two
Si-Si bonds parallel to the surface in a zig-zag pattern (see fig 21) Cleaved (110)
surfaces are disordered but upon annealing at high temperatures an ordered complex 16x2
Fig23 Scheme of the 7x7 DAS model [2] In each unit cell there are 9 dimers 12 adatoms and a stacking layer fault The force driving this complex reconstruction is the minimization of dangling bonddensity The DAS model shows the lowest number of dangling bonds (19) of all possiblereconstructions 12 dangling bonds are at the adatoms 6 at the rest atoms and 1 at the corner hole atom This surface is metallic
VPalermo 21
reconstruction takes place The surface appears as a series of long ridges and valleys
parallel to each other Eventually tilted facets of orientation (17 15 1) can form on this
surface The adsorption of Ge atoms on this surface leads to the formation of self-
assembled nanowires [3]
Table 22 Angles in degrees between different silicon faces [2]
Orientation
113
110
111
100
100
2524
9000
5474
0
111
2950
3526
0
110
6476
0
113
0
Fig 24 STM image of a Si(111)surface with 7x7 reconstruction A unitcell with its adatoms is highlightedImage size 13x13 nm
22 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] Weast RC Handbook Of Chemistry And Physics (Chemical Rubber Co Cleveland
1972)
[2] Dabrowski J Mussig H Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] The Structure Of Clean And SiGe-Covered Si(110) Surfaces Butz R Luth H Surface
Science 365 (3) 807-816 Oct 1 1996
VPalermo 23
3 STM and other surface analysis
techniques
Scanning Tunneling Microscopy
Since the invention of the optical microscope at the end of the 16th century the possibility
of examining surfaces at higher and higher magnification has fascinated mankind
Development of the technique continued and towards the end of the 19th century optical
microscopes were as good as todays standard instruments The physical limits of the
wavelength of visible light (350-800nm) had been reached
In the 1920s de Broglie showed that electrons can behave like waves and the use of these
particles for imaging with much higher resolution soon followed Atomic resolution using
this technique is only possible in the transmission mode with extremely carefully prepared
samples
In 1982 using the peculiar properties of piezoelectric materials Binnig and Rohrer brought
a metallic tip very very close to a silicon surface and scanned it across an extremly small
area (fig31) The tunneling of electrons from the tip into the sample or vice versa allowed
them to obtain a local density of electronic states (LDOS) map of that surface Although
theory (which treated the extreme point of the tip as a sphere) then excluded the possibility
of atomic resolution following a tip crash into the surface Binnig and Rohrer began to
observe the LDOS with atomic resolution For this discovery and their following work
they were awarded the Nobel Prize in Physics in 1986
24 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The basic principle of STM is very simple A metallic tip is scanned over a surface without
making ohmic contact and a tunneling current passes between the tip and the surface An
electronic circuit keeps this current constant by raising and lowering the tip during the
scan In this way recording the tip height at each point a three-dimensional image of local
density of electronic states (LDOS) of the surface can be obtained To explain the
extremely high resolution attainable by this simple technique quantum theory is needed
According to classical physics the current will flow between sample and tip only if they
are in physical contact If there is a vacuum gap between the two the electrons will simply
remain confined for example within the surface without the possibility of passing into the
tip
In quantum physics however the electrons have a certain probability of passing
(tunneling) across the gap appearing on the other side of the gap in this way reaching the
tip It can be shown that the probability of an electron tunneling through a gap of thickness
z is
kzep 22)0( minusprop ψ h
φmk
2= (1)
where ψ(0) is the electron wavefunction at the surface-gap border m is the electron
mass=91x10-28 g and φ is the work function of the metal (ie the energy required to
remove an electron from that material For silicon it is 48 eV) The tunneling current thus
decays exponentially with z and is extremely sensitive to topographical imperfections
present on the scanned surface A rough formula giving the current as a function of z is [1]
zFS eEVI φρ 0251)( minusprop
Fig 31 Binnig and Rohrer with the first STM Image from IBM [2]
VPalermo 25
where ρs(EF) is the local density of states at the Fermi level on the given surface For
example the formula predicts that for silicon an increase in tip-surface distance of 1 Aring
will give a 95 decrease in tunneling current
This huge dependence of tunneling current upon the distance allows detection even of the
sub-nanometre changes in height given by the single atoms of which the surface is
composed and thus to resolve them in the LDOS images Of course this description of the
tunneling process is oversimplified and for a more accurate one the electronic states of
the tip of the sample and their interaction have to be taken into account Fig 32 shows a
schematic representation of the interaction between tip and sample orbitals
The exponential decay of current with distance also yields high lateral resolution If the tip
is approximated as a sphere of radius R and the current passing at the minimum tip-sample
distance is I0 then the current passing at a lateral distance x from this point will be
Rxk
eII 22
0
2minus
=
Assuming a tip radius of 100 nm the current is concentrated in an area sim15 nm wide at tip
apex
Very sharp tips with even smaller curvature radii can be produced in several ways Simple
Fig32 Schematic view of tip-sampleorbitals interaction a) no interaction b)equilibrium c) sample positive d) tippositive [1]
26 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
tungsten wires mechanically cut are capable of obtaining atomic resolution on graphite in
air but for more disordered and rougher samples sharper and more reproducible tips are
needed
STM tips are mostly made by electrochemically etching a W or Pt-Ir wire The tips we
used were prepared using methods based on the work of Fotino [3]
A tungsten wire 038 nm diameter is immersed in a KOH solution(06M) with a thicker
tungsten wire used as a counterelectrode The cathodic and anodic reactions involved in the
etching are
Cathode 6H2O + 6 e- rarr 3H2(g) + 6 OH-
Anode W(s) + 8 OH- rarr WO42- + 4H2O + 6 e-
A potential of 3V ac is applied to the tungsten and the wire is immersed in the solution
until a constant current of sim100 mA is established The etching rapidly removes metal
shaping the wire end as a sharp tip When the potential reaches 11 V the coarse tip etching
is finished The wire is then removed from the solution carefully inserted into an
insulating plastic tube leaving only the tip exposed and re-immersed in the solution with
the tip pointing upwards A more gentle etching is thus made to reduce tip radius Usually
5 minutes etching at 07 V ac is used In this configuration very small hydrogen bubbles
formed on the tip sides sliding upwards with a ldquohoningrdquo effect on the tip
This procedure yields extremely sharp and reproducible tips at the microscopic level
After the etching the tip is thoroughly rinsed in ultrapure water then dipped into
concentrated HF to remove surface oxides and hydroxides [4] The tip is dried with
nitrogen inserted into the UHV system and degassed overnight at sim150degC
The possibility of measuring sub-nanometric distances would be useless without being able
to control tip movement over such a minute scale To scan the tip over the surface a
piezoelectric scanner is used Piezos are usually made of an alloy of PbZrO3 and PbTiO3 a
material which contracts or expands when a voltage is applied to it The Omicron
instrument used in our laboratory has three such piezo scanners for xy and z tip motion
allowing one to scan the tip over the surface with sub-Aringngstrom precision (fig 33)
To isolate the instrument from ambient vibrations the whole STM stage is suspended upon
four springs which eliminate all frequencies above 1 Hz and surrounded by a crown of
VPalermo 27
copper wings and fixed magnets Parasitic currents generated by the magnets into the
copper wings contrast every movement of the stage and efficiently block stage vibrations
The STM can be used not only to explore surface topography but to measure the IV
characteristics of single atoms or molecules on the surface (Scanning Tunneling
Spectroscopy or STS) Furthermore it can be used to modify the surface with voltage
bursts digging into it or delicately moving single molecules or atoms over a surface [5] It
can be used in vacuum in air and with proper lateral isolation of the tip even in liquids
A major drawback of STM is that it works only on conducting and semiconducting
surfaces and thus cannot be used on many surfaces of biological and chemical interest
Another instrument more suitable for these and other applications is the Atomic Force
Microscope (AFM)
Atomic Force Microscopy
AFM was invented in 1986 by Binnig Quate and Gerber after calculating the possibility of
building a cantilever with a force constant of the same order of magnitude as that of a
chemical bond
In AFM a tip mounted on a microscopic cantilever (usually made of Si3N4 fabricated with
optical lithography) is brought close to a surface When the tip touches the surface the
cantilever is very slightly deflected upwards The movement is measured by observing the
Y-PIEZO
SILICON SAMPLE
Fig 33 A picture of theSTM used for theexperiments The tripodpiezo scanner is shown
Z-PIEZO
X-PIEZO
TIP
28 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
deviation of a laser beam hitting the upper face of the cantilever Fig 34 provides a
schematic view of the principle of AFM
The typical force constant of the cantilever varies from 00006 to 2 Nm the typical
resonance frequency is 3 to 120 kHz The AFM tip can apply a force on the sample of up
to 10-9 N The AFM can be used on conductive or insulating surfaces in vacuum air or
liquids Furthermore the tip can be modified to sense electrostatic potentials (electric force
microscopy) or magnetic fields (magnetic force microscopy) it can even be functionalized
with complex molecules such as proteins to interact with biological surfaces
A drawback of the AFM is that the force it exerts can damage the surface under
observation especially if the sample is soft (as in the case of cell membranes for
example) This problem can be overcome using the instrument in tapping mode (where the
tip does not move laterally during its brief contact with the surface) or in non-contact mode
in which the tip oscillates above the surface during the scan and the changes in its
frequency due to interaction with surface are monitored The shifts in the oscillating
frequency of the cantilever due to tip-sample interaction are then used for imaging the
surface In this mode interaction of the tip with the surface is minimal and soft samples
can be imaged
STM and AFM are the main techniques used for this work A brief description of other
techniques used occasionally is given below
Fig 34 Scheme of an atomic force microscope
VPalermo 29
Low Energy Electron Diffraction (LEED)
Electrons with energies in the 20-500 eV range are diffracted by a crystalline surface the
diffraction peaks are visualized on a fluorescent screen This technique probes the long
range order of the surface up to a depth of several nanometres
X-Ray Photoelectron Spectroscopy (XPS)
XPS allows both qualitative and quantitative chemical analysis of the elements present on
or near the sample surface
An X-ray source is used to photoionize the atoms on a surface and produce photoelectrons
By measuring the kinetic energy of the photoelectrons the binding energy of the electronic
levels can be calculated This energy will depend on the chemical environment of the
surface atoms
Although the soft X-rays used penetrate to a depth of ~2000Aring the sampling depth of the
technique is determined by the mean free path of the photoelectrons which allows their
escape from only the first 10-100Aring
Secondary Ion Mass Spectroscopy (SIMS)
High and low energy ions (primary ions) are used to bombard a sample and remove surface
atoms and ions The ionic fragments removed (secondary ions) are then analysed by a mass
spectrometer The surface can be consumed during the measurement and profiles obtained
giving concentrations of the materials composing the sample at different depths (depth
profiles)
A popular variant of SIMS is TOF-SIMS In this technique the secondary ion masses are
measured by a time-of-flight (TOF) measurement The secondary ions generated by the
bombarding primary ions are accelerated to a constant kinetic energy and then move
through a field-free space before they reach the detector where their intensity is measured
as a function of flight time Since ions with different masses have different velocities at a
30 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
given kinetic energy the measured flight times of the ions can easily be converted to their
masses The static nature of this latter technique allows mass spectroscopy surface analysis
with minimal damage to the surface
Bibliography
[1] Chen CJ Introduction To Scanning Tunneling Microscopy (Oxford University Press
Oxford 1993)
[2] From wwwibmcom
[3] Tip Sharpening By Normal And Reverse Electrochemical Etching Fotino M Review
Of Scientific Instruments 64 (1) 159-167 Jan 1993
[4] A Convenient Method For Removing Surface Oxides From Tungsten STM Tips
Hockett LA Creager SE Review Of Scientific Instruments 64 (1) 263-264 Jan 1993
[5] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
VPalermo 31
4 Surface modification of silicon in liquid
Nano-hole creation
Liquid treatments of silicon wafers are very common in the integrated circuit (IC)
manufacturing industry They are used to clean and improve surface uniformity to create
and etch protective oxide layers and to remove photo-resist layers
Crystalline silicon with its native oxide layer is very stable and is resistant to many acids
It is easily attacked by hydrofluoric acid (HF) and alkaline solutions
The thin (~2 nm) passivating layer of native oxide (SiO2) is formed on exposure to the
atmosphere This surface layer contains many defects and contaminants so it is usually
chemically stripped and substituted with a better chemically-formed protective oxide
The most common silicon cleaning procedure is the RCA method named after the Radio
Corporation of America [1] It consists of two steps in the first one the surface is treated
with a hot alkaline solution (H2OH2O2NH4OH 411) to remove particles from the
surface following this a hot acidic solution (H2OH2O2HCl 411) is used to remove
metal contamination Other well-known cleaning methods are IMEC (a sequence of
cleaning steps in H2OO3 and dilute HF) or the Pirana etch (a hot 41 mixture of H2SO4
H2O2)
32 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The standard RCA clean removes surface contaminants etches the native oxide and
oxidizes the silicon surface leaving a uniform layer of silicon oxide which better protects
the surface from further contamination
Etching with fluorine-based solutions
Hydrofluoric acid is one of the most common reagents used in the treatment of silicon
wafers both in the research field and in industrial processes A rapid dip in dilute HF is the
simplest way to remove the native oxide from Si(100) and leaves the surface passivated by
a layer of Si-H bonds Because of the low polarization of Si-H bonds the Si-H layer is
stable even for several days protecting the surface from contamination It has often been
assumed that this short etch does not significantly change the surface morphology of the
silicon substrate[2] even though a prolonged dip in dilute HF leads to surface roughening
[3]
Although dilute HF roughens the Si(100) surface at the atomic scale [45] immersion in
concentrated HF (49) etches the surface oxide without attacking the Si surface
uncovering in this way the buried SiSiO2 interface The final counter-intuitive result is
that dilute HF etches the silicon while concentrated HF leaves the crystalline silicon
untouched [3]
Etching Si with fluorine-containing solutions at different concentrations and pH can
produce different morphologies from rough surfaces to flat nearly ideal Si-H terminated
surfaces
Hessel et al and Higashi et al demonstrated in 1991 that very flat Si(111) surfaces can be
obtained using 40 NH4F while etching with HF always yields rough surfaces The
surface becomes smoother because the etchant rapidly attacks Si atoms at step borders
thus removing surface kinks and irregularities in a step-flow mechanism [6 7] Later on
even smoother and more perfect surfaces were obtained by removing oxygen from the
solution after it was discovered that oxygen dissolved in 40 NH4F initiates the formation
of triangular etch pits It was not possible to obtain flat surfaces by etching Si(100) with
ammonium fluoride solutions which leads to the formation of small 2x1 dimer-row
reconstructed (100) terraces together with (111) facets [8]
This difference is caused by the different hydride terminations prevailing on the (100) and
(111) faces While the ideal Si(111)-H surface is monohydride terminated the more
VPalermo 33
reactive dihydrides predominate on the Si(100)-H surface making it more vulnerable to
etching The etching reaction is thus strongly anisotropic etching (100) facets faster than
(111) thus producing (111) microfaceting on Si(100) crystals
A more uniform Si(100) surface can however be prepared by etching at low pH with an
HFHCl mixture [9] or by using very dilute HF solutions and ultrapure water with low
dissolved oxygen and carbon contents [10]
Electrochemical etching can also be used applying anodic or cathodic bias to the silicon
to obtain different morphologies [11] by varying the potential isotropic or anisotropic
etching is observed The aforementioned results show that despite the simplicity of the
reactants fluoride etching of silicon is quite a complex reaction
Fig 41 Chemical etching of silicon
HO+H
H HO
H
H
H
FSi
Si
SiSi
+H2O
-OH-
+F-
H2O H H
Si H
F OH
H Si
Si Si
-H2
+H2O
H
H
H
FSi
Si
SiSi
H
H
H
OHSi
Si
SiSi
H
H
H
H Si
Si
Si Si
Etching mechanism of silicon
HF rapidly dissolves the SiO2 passivating layer on silicon leaving the surface almost
completely hydrogenated [12] After this two different types of reactions etch the silicon
simultaneously one chemical and the other electrochemical [13] The overall etching
mechanism can be schematized in two stages (see Figure 41)
i) Si-H bonds are substituted by Si-F or Si-OH bonds creating a partial charge on the
surface silicon atom and polarizing its Si-Si backbonds
ii) these polarised backbonds are then more easily attacked by HF or H2O After
rupture of the Si-Si bond the atom is removed leaving behind new Si-H
terminations and the reaction can start again
34 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
These reactions take place although at different rates on both Si(100) and Si(111)
Stage i) is usually the rate-determining step of the reaction and the stability of the Si-H
bonds depends upon the pH the concentration of nucleophilic species in solution and an
eventual potential applied to the crystal
For pH gt5 as in the case of concentrated NH4F solutions reaction begins with attack by
water to give Si-OH (step ArarrC) The -OH group is rapidly substituted by fluorine with
polarization of the underlying Si-Si bonds These bonds are then easily attacked by water
the silicon atom being released into solution as HSiF(OH)2 The Si-OH rarr Si-F substitution
is not fundamental for the reaction and etching can proceed even for Si-OH terminated
atoms but XPS measurements showed the presence of a certain number of Si-F bonds
remaining Furthermore fluorine seems to have a catalytic effect on Si-H substitution as
indicated by the dependence of the etch rate upon the F- concentration at least for pH
values between 4 and 8
Si-F bonds can be easily removed by a water rinse In the case of strongly alkaline
solutions (pH=14) OH- groups act directly as nucleophiles and no fluorine is needed to
catalyze Si-H bond rupture
At pH lt4 almost no free F- ions are present in solution and the etch rate is very small at
pH lt2 all etching reactions are very slow and this explains the stability of Si crystals in
concentrated (50 ww) HF solutions
Si Si
Si Si
H
H
H
H Si
Si
Si Si
H
H
H Si
Si
SiSi
H
H
H
OH-H+ -e
(F-)
-H+
-e
+H2O hellipas in fig1
Fig 42 Electrochemical etching of silicon by H2O
The electrochemical etching of silicon involves electron transfer from the surface atoms to
the valence or conduction band of the crystal (according to the type of doping of the
silicon) Several studies have been carried out at different electric potentials In the case of
p-silicon or n-silicon at anodic potentials where positive charges (h+) are available in the
crystal a silicon-centred radical can be created by capture of a hole from the bulk (fig
42) The F- ion does not participate in the reaction but is thought to contribute through
VPalermo 35
electrostatic interactions by lowering the energy of the interaction step After the
formation of the Si-OH group the reaction proceeds as shown in the scheme of fig 41
Matsumura et al [4] proposed that not only water but HF2- molecules also play a major role
in electrochemical etching of silicon leaving on the surface Si-F terminated bonds which
can be immediately attacked in an autocatalytic process (fig 43)
In the electrochemical reactions described above an external potential is applied to the
silicon crystal The chemical and electrochemical reactions in any case take place
simultaneously most of the time with the chemical path predominating at high pH Even
when no external potential is applied to the silicon partial electrochemical reactions can
take place at different ldquocathodicrdquo and ldquoanodicrdquo sites on the surface with an internal charge
exchange which ensures neutrality [11] This macroscopic silicon etching and hydrogen
bubble formation can lead to visually observable patterns on the surface when Si(100) is
immersed in ammonium fluoride even without applying a potential
Si Si
Si Si
H
H
H
F Si
Si
Si SiH
F F-H+ -2e
+HF2-
H
HF
SiSi
SiSi + F
F F
FH
H
H
F -H+ -2e
+HF2-
Fig 43 Autocatalytic electrochemical etching of silicon by HF2-
Inhomogeneities on silicon surfaces caused by electrochemical reactions and charge
transfer have been studied extensively because they are of fundamental importance in the
formation of porous silicon
Pore formation on silicon
When Si(100) or Si(111) are etched under anodic bias in fluorine-based solutions
microscopic pores form on their surface Several different morphologies of pores have
been observed with pore diameters ranging from 10 nm to several microns with depths of
several microns [14] Pore shape is very variable too ranging from ordered straight pores
to chaotic networks of branched pores (fig 44) Porous silicon has been known since the
fifties but it was only in 1990 that interest in this material increased following the
discovery that porous silicon layers were able to emit bright red light This led to a large
amount of research and now different classes of micropores can be reproducibly created
36 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Fig 44 Different types of Silicon micropores From ref [14]
mostly for optical and micromachining applications However there is still no unified
theory able to explain the nucleation and growth mechanism of all the different kinds of
pores
We will give a short description of some of these theories for more detail see Parkhutik et
al [15]
One model explains pore nucleation on the basis of physical processes such as hole
positive charge migration ion transport to the surface and small perturbations on the
silicon surface modelled as Fourier components The system is shown to be unstable and
some spatial frequencies that lead to pore nucleation evolve from the etching process
A second model focusses on stationary pore growth without explaining the nucleation
stage According to this model silicon dissolves preferentially at pore edges because h+
charges are attracted by the stronger electric field present at these edges
A third class of models explains pore growth as a Diffusion Limited Aggregation (DLA)
process where the random walk of h+ charge carriers through the depleted layer present at
the silicon-liquid interface controls pore shape
Finally the model by Carstensen Cristophersen and Foll [16] proposes that areas of the
surface of some characteristic size LCO are etched by synchronized ldquocurrent burstsrdquo in the
flow of h+ charges These bursts dissolve silicon through cyclic stages of surface oxidation
oxide removal and hydrogen passivation Areas where a burst has recently taken place are
less passivated and thus more likely to be etched again in this way the pore bottom
continues to dissolve while the pore walls are passivated and are thus less favourable areas
towards current bursts
VPalermo 37
EXPERIMENTAL RESULTS
In the following sections we will show some experimental results obtained from STM and
AFM measurements of fluorine-treated Si(100) surfaces In the first part the results of
mild etching using concentrated and dilute HF solutions at low pH are presented In the
second part the results of etching at high pH using ammonium fluoride are presented and
the mechanism of pore formation discussed
Etching of Si(100) in dilute and concentrated HF
Samples were cut from different areas of an 8-inch diameter p-doped silicon(100) wafer
(10 Ω-cm) supplied by MEMC Electronic Materials Each series of STM measurements
was carried out over at least six different areas on at least two identical samples Low
Electron Energy Diffraction (LEED) was used to check the surface cleanliness of the
samples before STM measurements
Table 41 summarizes the different treatments of each sample After etching with
electronic grade HF each sample underwent a final rinse in Ultra-Pure Water (UPW
resistivity gt18 MΩ-cm) Both the HF and the UPW were allowed to flow continuously
over the sample surface Some samples were not etched with HF at all but just washed with
UPW to observe the morphology of the native oxide layer (~2 nm thick) covering the
surface All of these processes were carried out under nitrogen to limit reoxidation and the
samples were then introduced from the nitrogen atmosphere directly into the vacuum
chamber and degassed overnight at sim150degC before LEED and STM measurements
STM images were obtained from each sample using the same measurement parameters
(sample bias 4 V feedback current 1 nA scan speed 800 nm s-1) These parameters
38 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
provided a satisfactory level of reproducibility for all the samples Measurements were
made over an area of 500x500 nm (image size 500x500 pixel) Slope correction was
carried out by subtracting row-wise and column-wise fitted slopes from the entire image
which gave better results than the simple subtraction of a fitted plane especially for the
rougher samples Following slope correction the rms roughness
sum minus=xy
hyxhN
22 ))((1σ
and the 2-D Fourier transform
)(22
2
)(4
)( vyuxi
xyeyxhvuF +sum∆
= π
π
were calculated for each image where N2 is the number of pixels composing the image
h(xy) is the surface height at each point ∆ is the distance between points h is the mean
height and u v are the spatial frequencies The radial power spectrum PS(f) of the STM image data is obtained from the angular
average of the squared Fourier transform with f 2 = u2 + v2
Fig 45 shows the STM images obtained from the various samples Sample A still covered
with its native oxide layer shows an irregular surface with RMS roughness of ~05 nm
(see Table 41) Observing the sample with LEED gave no diffraction pattern even at
relatively high incident electron energies because of the surface oxide coverage After 1
min etching in dilute HF (sample B) the morphology is similar to the original one though
Table 41 Sample treatments average RMS roughness and integrated area of power spectra
Sample Treatment RMS roughness (nm)
PS area (f lt 01 nm-1)
PS area (f gt 01 nm-1)
A Rapid dip in water 053 plusmn 013 366 032
B 1 min in HF 5 + 10 min in water 051 plusmn 008 272 034
C 30 min in HF 5 + 10 min in water 062 plusmn 008 1064 038
D 5 sec in HF 49 + 10 min in water 042 plusmn 004 183 022
VPalermo 39
Fig 45 STM images of each group of samples showing the topography of the silicon surface A) noetching original oxide surface B) after 1 min etching in dilute HF C) after 30 min etching in diluteHF D) after dipping in concentrated HF Grey scale indicates height of the surface from lower (black) to higher (white) The images are 250x250 nm ie representative portions of the images usedfor the roughness measurement and PSD analysis
some of the larger features have disappeared and the image quality is better maybe due to
improved tunnelling due to the cleaner surface The RMS roughness is comparable to that
of the original surface Clear diffraction patterns are visible using LEED though at quite
high energies (200 eV) After prolonged etching (sample C) the RMS roughness increases
to 062 nm and a long-range corrugation is visible on the surface even if the LEED pattern
is good
The samples dipped in concentrated HF (D) reveal the bare SiSiO2 interface which has a
disordered aspect and protrusions over a wide range of dimensions The quality of the
STM images of sample D is very good probably due to the cleanliness of the surface
40 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 E -0 2
1 E -0 1
1 E + 0 0
1 E + 0 1
1 E + 0 2
1 E + 0 3
0 0 0 0 0 1 0 1 0 1 0 0f (1 nm )
nm^4
A a s re c e iv e d
B e tc h e d 1 min
C e tc h e d 3 0 min
D e tc h e d HF 4 9
Fig 46 Log-Log plot of the averaged power spectra of the STM images for all the samples
which gives a more stable tunnelling junction The LEED pattern is excellent showing
clear diffraction peaks at energies as low as 37 eV comparable to that obtained after high
temperature cleaning in UHV
Fig 46 shows the power spectra of the samples The high frequency and low frequency
areas of the power specturm are considered separately Table 41 shows for each sample
together with the roughness the integrated area of the power spectrum for the high and low
frequency part
We first examine the differences between the samples in the low frequency part of the
spectrum (lt01 nm-1) In this spatial range the short dip in concentrated or dilute HF
(samples BD) seems to lower the roughness of the sample removing some of the larger
features of the native oxide surface visible in Fig 45a and thus decreasing the
corresponding part of the power spectrum Sample C however etched with HF for 30 min
shows an increase in roughness peaking at 0012 nm-1 (~85 nm corresponding to the
typical dimensions of the corrugations visible in Fig 45c) On this scale sample C has a
PS density of 435 nm4 compared to 101 nm4 for sample A
Samples A B and C have the same PS in the high frequency range showing that the fine
structure of the surface is not changed by the HF etching Sample D on the other hand
shows a decrease of roughness for frequencies gt 015 nm-1 This effect could be due to the
VPalermo 41
improved surface cleanliness after etching with concentrated HF which would give a more
stable STM junction thus reducing the high frequency noise in the image
Fig 45 and the analysis of the power spectrum of each sample shows that a rapid dip in
HF removes the native oxide but does not lead to major changes in the morphology of the
Si surface its only effect being the removal of some of the larger features present on the
original surface Prolonged etching on the other hand increases the RMS roughness of the
surface
Etching of Si(100) in concentrated ammonium fluoride and nano-hole creation
Two different types of commercial p-doped Si(100) wafers (2Ω-cm and 10Ω-cm) from
MEMC were used Several different samples of 10x5 mm were immersed for 10 minutes
in 40 electronic grade NH4F solution under agitation Previous works used low
temperatures or anodic potentials applied to the silicon to avoid gaseous hydrogen
production and to obtain a uniform surface but in our experiment we worked at room
temperature to check the influence of hydrogen bubbles on surface morphology During
the etching the stirring was sufficient to provide a uniform concentration of reagents over
the whole sample surface but not to mechanically remove the hydrogen bubbles from the
silicon surface
After the etching each sample was rinsed in ultra-pure water to remove any etching
residues and observed with STM AFM and optical microscopy The AFM measurements
were made in air while for STM measurements the samples were rapidly dried with
nitrogen and inserted into the vacuum system to avoid surface reoxidation After insertion
into the vacuum surface cleanliness was checked with LEED and the surface morphology
observed by STM Parameters for STM measurements were sample bias +4 V 1 nA
current 16 Hz scan rate The images obtained were stable and reproducible over several
days Scan parameters for AFM were 20 nN force and 1 Hz scan rate
Some of the samples were cleaned with an RCA standard clean [1] before NH4F etching to
check the influence of possible surface contaminants on the final results Eight different
samples were prepared and more than sixty STM images of the samples were taken at
different points of the various samples
42 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
(110)
(110)
Fig47 abc) STMtopographic images of differentetching morphologies Eachimage is 500x500 nm Z-ranges are 10 10 and 18 nmrespectively d) STM image of a bridgecreated by etching of the lowerlayers of silicon (black arrow)Image is 250x250x6 nm
After ~2 min of immersion in the solution hydrogen bubbles become visible on the sample
surface The production is slow and the bubbles are quite stable on the sample without
detaching Thus some areas of the surface are masked from the liquid etching action
STM observations (fig 47) show that at the nanometer scale the surface is unevenly
covered with holes of radii ranging from 10 to 200 nm with depths of 2-4 nm These holes
have a wide range of different shapes and distributions In most cases the surface was
covered with a uniform distribution of round-shaped holes (fig47a) indicating isotropic
etching The dimension and the density of the holes changed greatly from sample to
sample and even over the surface of a single sample In some cases the etching was
anisotropic yielding nearly square holes and layered structures as shown in fig 47b
Square holes have been previously observed in cases where the etching speed in the (110)
direction is significantly smaller than in the (100) direction [17]
Over these areas (fig 47b and especially 47c) it is clearly visible how once the surface
had been attacked the reaction continued to preferentially remove atoms at step
irregularities (kink atoms) straightening step edges Eventually the exposed underlying
silicon was also attacked and further holes created inside the previously etched larger
ones It was not possible to detect monatomic steps on this kind of surface The smallest
step height observed was ~15 nm corresponding to several atomic layers In the image
shown in Figure 47a the etching was not strong and created only anisotropic holes on the
surface In fig 47b and c the stronger etch proceeded laterally for several tens of
VPalermo 43
nanometres leaving straight steps several tens of nanometers long and roughly rectangular
holes as expected given the structure of the (100) crystal face In some cases a
significative underetch is observed and the formation of suspended bridges and tunnels
can be deduced from the STM images (fig 47d)
The formation of branched pores and suspended structures has been attributed during pore
formation to diffusion limited aggregation effects where the h+ charge carriers necessary
for silicon etching have a higher probability of reacting at pore bottoms than reaching the
upper part of the silicon surface In the case of very deep pores quantum wire effects have
been invoked to explain the pore growth mechanism [15] In our case though the pores
formed were very shallow the underetch depth being only a few nanometres on pores of
sim100 nm width Thus more than diffusion effects the main contribution to the
underetching process must come from anisotropic etching and some kind of autocatalytic
reaction path analogous to the one described by Matsumura et al [4] with some areas of
the silicon surface hydrogenated and thus less vulnerable to etching
Pre-treatment with RCA cleaning has no effect on the final morphology and this seems to
exclude pore nucleation being caused by presence of metallic or organic surface
contaminants
The morphology and the distribution density of the pits was quite uniform over
microscopic areas of the sample but changes were observed over the millimetre scale This
suggests that etching intensity is influenced by some large-scale parameter
Large-area measurements made with AFM or with an optical microscope (fig 48)
showed that the inhomogeneity of surface etching can be correlated with the masking
action of the bubbles While the fluoride dissolved the silicon hydrogen bubbles formed
by the reaction covered some areas of the surface thus blocking the etching over that area
generating macroscopic steps at the bubble-liquid border As the reaction proceeded more
hydrogen accumulated and the bubble diameter increased producing in this way a circular
pattern of steps The increase in bubble diameter was not continuous with time otherwise a
uniform surface slope gradient would have been obtained The formation of this circular
ldquoetching staircaserdquo indicates that the bubble growth was stepwise the bubble accumulating
more and more hydrogen without enlarging across the surface until it relaxed increasing
its diameter stepwise and covering more silicon The circular structures in fig 48a are not
co-axial and their asymmetry could derive from physical processes due to stirring or
irregularities on the surface
44 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
a b Fig 48 a) optical micrograph of etching patterns on Si(100) created by NH4F 12x09 mm b) AFM image of the circles border xy range is 40x24 microm z-range is 30 nm
The step structure was not destroyed by the etching even after the bubble detached from
the surface but on the contrary the etching process seemed to be influenced by the
presence of the step
Observing in detail a series of steps (fig 48b) a quite deep trench is visible at the base of
each step A close-up image of a step and the corresponding line profile of the trench is
shown in fig 49 The trench is sim5 nm deep with respect to the lower surface compared to
a step height of 22 nm
A similar structure has been recently obtained with electrochemical etching of p-type
Si(100) in 4 HF [18] in which a ldquocurrent burstrdquo etching model previously described
was assumed for silicon dissolution In that experiment the trench was created at the
border of silicon nitride masks and began to grow after a nucleation stage Preferential
trench etching was along the (110) direction and the trench growth was explained as an
effect of mechanical stress induced by the nitride mask and of electric field enhanced
dissolution which depended upon an external applied potential
While it is clear that in our system the gas bubbles have a masking effect similar to a
classical solid nitride mask it is unlikely that hydrogen present on the surface can induce a
significant stress in the silicon lattice as in the case of a nitride mask Furthermore no
external field was applied to drive preferential etching at the trench site
It has been proposed [19] that the cathodic and anodic part of the etching reaction
(hydrogen production and silicon oxidative etching respectively) take place at different
points on the surface with a net charge transfer between the different areas In this case
the highest reaction rates will correspond to the silicon area surrounding the bubble border
where a high number of positive charges will be available for the reaction Furthermore a
sharp trench extending into the silicon crystal will be a preferential electrostatic attractor
VPalermo 45
Fig 49 AFM image of the etched surface showing a step created on the surface by bubblemasking A stronger etching action is visible on the right side of the step as well as theprotected area on the upper side of the step (indicated by the arrows) Image is 10x10 micromz-range is 30 nm The profile on the right is taken from the central area of the image
for the h+ charge carriers coming from other ldquocathodic areasrdquo of the sample either from
other regions on the surface or from the back of the silicon chip [18]
In the areas where the hydrogen bubble had detached and the surface was exposed to the
etching the reaction was not uniform in the neighbourhood of the steps It is possible to
observe (fig 48b and 49) an area on the upper side of the step where less or even no
etching at all seems to have taken place as if the step was able to protect the surface from
etching While etching on the lower side with trench formation can be attributed to the
presence of the bubble the surface on the upper step side can be etched only after bubble
detachment so no masking effect can account for this result However a further
preferential attraction of h+ charge carriers from the already formed trench can be
hypothised electrochemically shielding the surrounding area from further etching If this is
true the shielding effect would be very strong with a relatively shallow 5 nm-deep trench
protecting an area of sim1 microm parallel to the step
To summarise the etching of Si(100) in NH4F is a complex process in which different
reaction paths both chemical and electrochemical co-exist Hydrogen bubbles formed by
the reaction act as a mask on the surface and create etching paths and inhomogeneous
etching of the surface Different kinds of pores are observed on the surface and in some
cases the anisotropy of the process is so strong as to give square-shaped holes and
underetching
46 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The diffusion of h+ charge carriers in the crystal is one of the main rate-determining steps
of the reaction and leads to the formation of a deep trench immediately outside the bubble
perimeter These trenches act as charge collectors and reduce the etching of the upper step
surface in the proximity of the steps
Bibliography
[1] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[2] Spectroscopic Ellipsometry Studies Of HF Treated Si (100) Surfaces Yao H Woollam
Ja Alterovitz SA Applied Physics Letters 62 (25) 3324-3326 Jun 21 1993 Influence
Of HF-H2O2 Treatment On Si(100) And Si(111) Surfaces Graf D Bauermayer S
Schnegg A Journal Of Applied Physics 74 (3) 1679-1683 Aug 1 1993 Kinetics Of
Oxidation On Hydrogen-Terminated Si(100) And (111) Surfaces Stored In Air Miura
T Niwano M Shoji D Miyamoto N Journal Of Applied Physics 79 (8) 4373-4380
Part 1 Apr 15 1996
[3] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995 Structure Of The Stepped SiSiO2 Interface After Thermal-Oxidation -
Investigations With Scanning Tunneling Microscopy And Spot-Profile Analysis Of
Low-Energy Electron-Diffraction Pietsch GJ Kohler U Jusko O Henzler M Hahn
PO Applied Physics Letters 60 (11) 1321-1323 Mar 16 1992
[4] Enhanced Etching Rate Of Silicon In Fluoride Containing Solutions At pH 64
Matsumura M Fukidome H Journal Of The Electrochemical Society 143 (8) 2683-
2686 Aug 1996
[5] A Study Comparing Measurements Of Roughness Of Silicon And SiO2 Surfaces And
Interfaces Using Scanning Probe Microscopy And Neutron Reflectivity Crossley A
Sofield CJ Goff JP Lake ACI Hutchings MT Menelle A Journal Of Non-Crystalline
Solids 187 221-226 Jul 1995
VPalermo 47
[6] Step-Flow Mechanism Versus Pit Corrosion - Scanning-Tunneling Microscopy
Observations On Wet Etching Of Si(111) By Hf Solutions Hessel HE Feltz A Reiter
M Memmert U Behm RJ Chemical Physics Letters 186 (2-3) 275-280 Nov 8 1991
[7] Comparison Of Si(111) Surfaces Prepared Using Aqueous-Solutions Of NH4F Versus
HF Higashi GS Becker RS Chabal YJ Becker AJ Applied Physics Letters 58 (15)
1656-1658 Apr 15 1991
[8] Wet Chemical Etching Of Si(100) Surfaces In Concentrated NH4F Solution -
Formation Of (2x1)H Reconstructed Si(100) Terraces Versus (111) Faceting Neuwald
U Hessel HE Feltz A Memmert U Behm RJ Surface Science 296 (1) L8-L14 Oct
10 1993
[9] Ideal Hydrogen Termination Of Si(001) Surface By Wet-Chemical Preparation Morita
Y Tokumoto H Applied Physics Letters 67 (18) 2654-2656 Oct 30 1995
[10] Atomic Structures Of Hydrogen-Terminated Si(001) Surfaces After Wet Cleaning
By Scanning Tunneling Microscopy Endo K Arima K Kataoka T Oshikane Y Inoue
H Mori Y Applied Physics Letters 73 (13) 1853-1855 Sep 28 1998
[11] On The Potential-Dependent Etching Of Si(111) In Aqueous NH4F Solution
Houbertz R Memmert U Behm RJ Surface Science 396 (1-3) 198-211 Jan 20 1998
[12] Etching Process Of SiO2 By HF Molecules Hoshino T Nishioka Y Journal Of
Chemical Physics 111 (5) 2109-2114 Aug 1 1999
[13] Etching Mechanism And Atomic-Structure Of H-Si(111) Surfaces Prepared In
NH4F Allongue P Kieling V Gerischer H Electrochimica Acta 40 (10) 1353-1360
Jul 1995
[14] Pore Formation Mechanisms For The Si-HF System Carstensen J Christophersen
M Foll H Materials Science And Engineering B-Solid State Materials For Advanced
Technology 69 23-28 Sp Iss Si Jan 19 2000
[15] Porous Silicon - Mechanisms Of Growth And Applications Parkhutik V Solid-
State Electronics 43 (6) 1121-1141 Jun 1999
[16] Formation And Application Of Porous Silicon Foll H Christophersen M
Carstensen J Hasse G Materials Science amp Engineering R-Reports 39 (4) 93-141 Nov
1 2002
[17] Pore Morphology And The Mechanism Of Pore Formation In N-Type Silicon
Searson PC Macaulay JM Ross FM Journal Of Applied Physics 72 (1) 253-258 Jul 1
1992
48 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
[18] Deep Electrochemical Trench Etching With Organic Hydrofluoric Electrolytes
Christophersen M Merz P Quenzer J Carstensen J Foll H Sensors And Actuators A-
Physical 88 (3) 241-246 Mar 5 2001
[19] Morphology Of Anodically Etched Si(111) Surfaces - A Structural Comparison Of
NH4F Versus HF Etching Houbertz R Memmert U Behm RJ Journal Of Vacuum
Science amp Technology B 12 (6) 3145-3148 Nov-Dec 1994
VPalermo 49
5 Surface modification of silicon in vacuum void creation and oxide desorption
The main reason for the huge success of silicon in the microelectronics industry is not due
to its superior properties as a semiconductor Other materials for example germanium
have better qualities such as higher mobility of charge carriers and lower noise levels
which would allow the construction of faster and higher performance devices
The widespread use of silicon however is mainly due to the outstanding characteristics of
its oxide Silicon dioxide (SiO2) is a very good electrical insulator easy to form
chemically and thermally stable and is compatible with lithographic and metal deposition
processes Germanium oxide on the contrary is too reactive to be used
Even the use of Si(100) substrates for nearly all microelectronic devices is dictated by
oxide quality The (111) face of silicon crystal can be easily cleaved and flattened and
almost atomically perfect surfaces can be obtained with simple chemical procedures (as
described above) But the density of interfacial defects is highest for Si(111)-SiO2
interfaces and lowest for Si(100)-SiO2 ones so microchips will continue to be fabricated
on Si(100) wafers
SiO2 (silica) is present in 95 of the earthrsquos minerals in different allotropic forms such as
quartz tridymite and cristobalite In the bulk each silicon atom is bonded to four oxygens
in a Si-O-Si tri-dimensional network Si-O bonds are 016 nm long and form an angle
ranging from 120deg to 150deg
Three typical intrinsic defects are present in SiO2 The so-called Ersquo centres are oxygen
vacancies with a hole localised on a silicon atom with only three Si-O bonds
50 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
O3Si +SiO3 Whereas the PR (peroxy radical) defects are holes trapped by a charged
peroxy moiety with a O3Si-O-O+ SiO3 structure The NBOHC (non-bridging oxygen hole
centres) derive from water or hydrogen contamination and are schematized as O3Si- O- H-
O-SiO3
The atomic structure of the Si-SiO2 interface varies enormously Local domains resembling
the tridymite and the cristobalite structure of silica are present but it seems that only 10
of the interface is ordered [1] Far from the interface the SiO2 bulk is completely
disordered The passage from bulk Si to stoichiometric SiO2 passes through a non-
stoichiometric SiOx layer 07 nm thick
When a clean silicon surface is exposed to atmospheric oxygen a thin sim2 nm thick layer
of native oxide forms spontaneously which is usually removed and substituted with
thicker better quality oxide layers before further processing
Silicon is usually oxidised by thermal annealing at temperatures between 800deg and 1100deg
in an atmosphere of pure O2 with some water eventually added to increase oxidation speed
Thermal oxides made in pure oxygen (dry oxides) grow more slowly than oxides produced
in an oxygen-water atmosphere (wet oxides) but are usually of better quality
According to the Deal-Groove formula the time t needed to grow an oxide of thickness X
is given by [1]
1
212minus
minusminus
+=
ABXBXt α
where the constant B and BA decrease exponentially with temperature as
minus
kTEexp
with activation energies for dry oxidation of EB =123 and EBA =20 eV respectively
EB is related to the diffusion of oxygen in silicon while the value of EBA is interpreted as
the energy required to break a Si-Si bond The exponent α is 1 for wet oxidation and 0 for
oxidation at high temperatures and low oxygen pressures It has intermediate values for dry
oxidation This formula does not work well for low values of X at the initial stages of
oxidation and usually empirical corrections are used
An interesting characteristic of silicon is that at high temperature and in vacuum oxygen
can actually etch the silicon crystal giving gaseous products and the oxidized layer present
on the crystal becomes unstable (fig 51) [2]
VPalermo 51
1E-10
1E-09
1E-08
1E-07
1E-06
1E-05
1E-04
1E-03
1E-02
1E-01
0607080911112
1000T (1K)
P (T
orr)
SiO2 + Si rarr 2SiO(g) Oxide decomposition
Si+O2 rarr SiO2(s) oxide formation Fig51 phase diagram of the
oxygen-silicon system
Silicon oxidation apart from the initial nucleation stages at the monolayer level proceeds
uniformly over the whole surface with a planar reaction front moving from the surface
into the bulk
If heated under low oxygen partial pressure (vacuum or inert atmosphere) SiO2 is known to
decompose following the reaction
SiO2 + Si rarr 2SiOuarr (1)
The reaction begins with nucleation at defect points on the SiSiO2 interface and proceeds
in a spatially inhomogeneous manner with the formation of large voids on the oxide
surface [3]
Several studies have been made on the dynamics of void growth both on thick [4] and thin
[5] layers of SiO2 The process has been used to decorate otherwise unobservable defects at
the SiSiO2 interface [6] or to grow nanoislands of silicon on the void surface [7] It has
been suggested that the defects acting as nucleation centres could be metallic contaminants
present on the native surface which aggregate and catalyze SiO2 decomposition [8]
After oxide desorption the silicon surface is very rough In particular on Si(100) square
islands are observed several nanometres high which act as lsquopinning sitesrsquo on the motion of
monatomic steps on the Si surface It has been proposed that these islands can be composed
of silicon carbide coming from the organic contaminants present on the surface [9] or of
oxide clusters during partial reoxidation [10]
In the following section some experimental results obtained studying voids and nanoisland
growth are presented
52 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nanoisland growth on silicon
We used two different samples the first rougher than the second (i) p-type Czochralski Si
(100) with a RMS roughness of 05 nm shown in fig52a and (ii) p-type epitaxial Si(100)
with a RMS roughness of 019 nm shown in fig 52b The roughness was measured with
STM on the lsquoas receivedrsquo samples on different areas of 500x500 nm Every sample was
covered by a layer of native oxide ~2 nm thick
Some of the samples (AC) were introduced into the UHV system without any cleaning
others (BD) were dipped in concentrated HF (49) to remove the surface oxide without
etching the silicon [11]
After degassing the samples were heated resistively increasing the temperature slowly to
keep the pressure within the 10-10 mbar range during heating Each sample was held at
900˚C for 30 min For some of the samples the heating was stopped at lower temperatures
to obtain incomplete oxide desorption and surface roughening thus enabling observation
of the different steps of the process Table 51 summarizes the different treatments for each
group of samples
Fig52 Original surfaces of Czochralski (a) and epitaxial (b) silicon Every image is 500x500 nm
VPalermo 53
Table 51 Summary of different treatements
Silicon type Oxide removed
Heating Islands density (microm-2)
Av Island Volume (nm3)
A Czochralski No 30rsquo at 900˚C 24x103 353
B Czochralski Yes ldquo 20x104 71
C Epitaxial No ldquo 28x102 2450
D Epitaxial Yes ldquo 23x104 63
Fig53 ab) Surface roughening on group A samples after heating 30 min at 800deg and900deg respectively in presence of an oxide layer Image size 500x500 nm cd) Surface roughening on group B samples after heating 30 min at 700deg and 900degrespectively after removing the oxide layer with HF Image size 250x250 nm
Fig 52a is the native oxide surface of sample A quite disordered and irregular This
surface was stable when heated to 700˚C while at higher temperatures the oxide began to
desorb Fig 53a was taken after heating the sample at 800˚C for 30 min
54 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The oxide began to desorb in correspondence with defect points creating oxide-free
surface voids which enlarge radially uncovering the silicon surface Some surface silicon
atoms are removed through reaction (1) Mobile silicon atoms moving randomly over the
surface aggregate forming a nanocrystal at the initial defect point surrounded by a stable
flat oxide-free surface with some steps The initial SiSiO2 interface as observed after
oxide etching with HF [12] is similar to the one shown in fig 52a and thus the surface
observed in the void area is a completely new one coming from surface reorganization
The reaction continues leading to a coalescence of the voids the final result of the process
is shown in fig 53b with protruding islands aligned along the main crystal axes and flat
areas with some steps LEED measurements confirmed that this surface is crystalline
silicon Atomic resolution was obtained on the flat areas of the surface showing the
typical 2x1 reconstruction of Si(100) Figures 53c and 53d show the evolution of the
roughening on sample B after chemically removing the native oxide layer before
introducing the sample into UHV As mentioned above the initial morphology is similar to
that shown in fig 52a though the oxide has been removed as confirmed by LEED
measurements Fig 53c shows the surface after heating at 700˚ for 30 min Without the
oxide layer the surface began to reorganize at lower temperatures and the small irregular
grains of the original surface begin to grow and become more rectangular After heating to
900˚C (fig 53d) the surface is flat with small rectangular islands The shape of the islands
is similar to that shown in fig 53b but the lateral island dimension is one order of
magnitude smaller
To assess the importance of the initial surface on the process the same treatments were
repeated starting from flat epitaxial silicon (samples CD) The initial epitaxial surface is
smooth with periodic steps due to a small miscut angle (fig52b) On heating the epitaxial
silicon without etching (sample C) relatively large islands are obtained similar to the ones
shown in fig 53b where the ldquostep-pinningrdquo effect of the islands is evident The average
dimension of sample C islands was sim2500 nm3 while only 353 nm3 for sample A Island
density decreased from 24000 to 280 microm-2 This result is maybe due to a lower
concentration of defects in the oxide layer of the epitaxial silicon which allows the
expanding void to grow much larger before coalescence with neighbouring voids the
nanostructures within them thus reaching greater dimensions To verify this hypothesis we
applied the same treatment to clean not-so clean and purposely contaminated samples To
check that the defects inducing void nucleation and nanoisland growth derive from carbon
VPalermo 55
contamination we used electronic-grade methanol to contaminate the silicon surface
Figure 54 shows the results of this comparison
State-of-the-art cleaned samples obtained with multi-step RCA cleaning [13] showed the
development of very few voids the oxide desorption leaving large smooth areas of oxide-
and nanostructure-free silicon as shown in Fig 54a where the presence of residual oxide
can be seen Samples cleaned with a simpler one-stage Pirana cleaning solution showed a
higher number of islands per unit area (Fig 54b) and samples purposely contaminated
with methanol (Fig 54c) had the highest density of island nucleation of all three samples
When the native oxide had been chemically removed from the epitaxial silicon before
heating (sample D) the island size and density is similar to the standard as-received
silicon case (sample B) Thus starting from a flat substrate the process of surface
roughening and island creation was the same even when the oxide layer had been
previously removed
From these results some important indications can be drawn The first is that oxide
desorption temperature depends upon the contamination levels of the oxide layers (fig 54)
with desorption at lower temperatures for contaminated surfaces
Second once the oxide layer is removed the silicon atoms become mobile on the surface
at relatively low temperatures (~700˚C) Their initial configuration is disordered and has a
high surface energy so they crystallize in the small islands shown in fig53d reducing
their surface area and lowering the energy of the system When the oxide is present the
surface cannot reorganize at 700˚C and is stable At 800˚C the oxide begins to desorb in
correspondence with defect points on the surface As the void area enlarges a small
cba Fig 54 The effect of contamination on nano-island production Gradient-filtered STM images of (a)sample cleaned two times with RCA 400x400 nm (b) Sample cleaned with a Pirana solution 500x500nm (c) sample contaminated with MeOH before insertion into the UHV system 500x500 nm Annealingtime is 40 min at 800degC for all the samples
56 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
crystallite grows at the centre of the voids Further oxide desorption uncovers larger areas
freeing more silicon atoms which feed the initial island making it bigger
Fig 55 shows the scaled size distribution of the islands for each group of samples Every
distribution has been obtained using several images of different samples for each group
The size distribution of all the samples follows an exponential decay and the decay seems
the same for all the samples though the average volume of the islands differs by more than
one order of magnitude (see Table 51 for details)
The point defects which catalyze void nucleation can be metallic contaminants present on
the initial oxide surface [8] or organic contaminants which at high temperatures can
form SiC nanocrystals [14] We used Scanning Tunneling Spectroscopy (STS) to look for
differences between the islands and the surrounding flat silicon surface but no difference
was found Furthermore the total island volume per surface unit is very high (more than
104 nm3microm2) and it seems unlikely that such a huge volume could consist of surface
contaminants Islands of pure silicon on silicon have been grown without evidence of
surface contamination[14]
To summarise the formation of nano-sized islands on silicon through surface diffusion
was studied At high temperatures the oxide layer covering the surface decomposes non
uniformly and circular voids of clean silicon are created The presence of the oxide layer
blocks surface silicon atomic motion and surface reorganization except within the voids
causing the growth of islands more than 10 nm high and 30 nm wide on the silicon surface
According to STM and LEED measurements we can say that the islands are mostly
Fig 55 Size distribution (scaled) for the samples described in table 1 The line is a reference for the eye
VPalermo 57
composed of silicon with traces of other substances such as carbon Final island densities
and dimensions depend upon the initial purity of the oxide layer
On the other hand if the oxide is removed chemically before heating the island growth
process is not so localised and takes place simultaneously over the whole surface yielding
smaller and more numerous islands
58 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[2] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[3] High-Temperature SiO2 Decomposition At The Sio2Si Interface Tromp R Rubloff
GW Balk P Legoues FK Physical Review Letters 55 2332-2335 Nov 1985
[4] Defect Microchemistry At The SiO2Si Interface Rubloff GW Hofmann K Liher M
Young DR Physical Review Letters 582379-2382 Jun 1987 Defect Formation In
Thermal SiO2 By High Temperature Annealing Hofmann K Rubloff GW Mccorkle
RA Applied Physics Letters 49 1525-1527 Dec 1986 Kinetics Of High-Temperature
Thermal Decomposition Of SiO2 On Si(100) Liher M Lewis JE Rubloff GW Journal
Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 5 1559-1562 Aug
1987
[5] Thermal-Decomposition Of Very Thin Oxide Layers On Si(111) Kobayashi Y Sugii
K Journal Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 10 (4)
2308-2313 Part 3 Jul-Aug 1992 Scanning Tunneling Microscope Study On Mid-
Desorption Stages Of Native Oxides On Si(111) Kobayashi Y Sugii K Journal Of
Vacuum Science amp Technology B 9 (2) 748-751 Part 2 Mar-Apr 1991 Controlled
Growth Of SiO2 Tunnel Barrier And Crystalline Si Quantum Wells For Si Resonant
Tunneling Diodes Wei Y Wallace RM Seabaugh AC Journal Of Applied Physics 81
(9) 6415-6424 May 1 1997
[6] Defect Formation In SiO2Si(100) By Metal Diffusion And Reaction Liher M
Dallaporta H Lewis Je Appl Phys Lett 53 589-591 Aug 1988 SiO2 Film
Decomposition Reaction Initiated By Carbon Impurities Located At A Si- SiO2
Interface Raider Si Herd Sr Walkup Re Applied Physics Letters 59 (19) 2424-2426
Nov 4 1991
[7] Nanometer-Scale Si Selective Epitaxial Growth On Si(001) Surfaces Using The
Thermal Decomposition Of Ultrathin Oxide Films Fujita K Watanabe H Ichikawa M
Applied Physics Letters 70 (21) 2807-2809 May 26 1997 Pyramidal Si Nanocrystals
VPalermo 59
With A Quasiequilibrium Shape Selectively Grown On Si(001) Windows In Ultrathin
SiO2 Films Shibata M Nitta Y Fujita K Ichikawa M Physical Review B 61 (11)
7499-7504 Mar 15 2000
[8] Stacking-Fault-Induced Defect Creation In SiO2 On Si(100) Liher M Bronner Gb
Lewis Je Appl Phys Lett 52 1982-1985 May 1988
[9] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990 AFM And XPS Characterization Of The
Si(111) Surface After Thermal-Treatment Lamontagne B Guay D Roy D Sporken R
Caudano R Applied Surface Science 90 (4) 481-487 Dec 1995
[10] Scanning-Tunneling-Microscopy Study Of Oxide Nucleation And Oxidation-
Induced Roughening At Elevated-Temperatures On The Si(001)-(2x1) Surface Seiple
JV Pelz JP Physical Review Letters 73 (7) 999-1002 Aug 15 1994 Evolution Of
Atomic-Scale Roughening On Si(001)-(2x1) Surfaces Resulting From High-
Temperature Oxidation Seiple JV Pelz JP Journal Of Vacuum Science amp Technology
A-Vacuum Surfaces And Films 13 (3) 772-776 Part 1 May-Jun 1995
[11] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995
[12] Morphological Changes Of The Si [100] Surface After Treatment With
Concentrated And Diluted HF Palermo V Jones D Materials Science In
Semiconductor Processing 4 (5) 437-441 Oct 2001
[13] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[14] Nanoscale Roughening Of Si(001) By Oxide Desorption In Ultrahigh Vacuum
Gray SM Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology
B 14 (2) 1043-1047 Mar-Apr 1996
60 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 61
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and nano-lines
When the silicon surface is not protected by a native oxide layer or by a hydrogenated
passivating layer it is very reactive The surface chemistry of both Si(100) and Si(111) has
been extensively studied starting from clean surfaces prepared in UHV because of its great
relevance to the microelectronics industry and its technological interest The diffusion and
reaction of molecules and atoms on silicon is an interesting scientific problem on its own
apart from technological issues because silicon reconstruction yields a very complex and
anisotropic surface
We give here a brief summary of the adsorption behaviour of several elements on Si
surfaces The summary is not comprehensive and only the most interesting characteristics
for each substance are given
Table 61 Summary of adsorption behaviour of atoms and molecules on silicon surfaces [12]
Hydrogen
Molecular H2 shows low reactivity towards silicon while atomic hydrogen easily forms
Si-H bonds and can even break Si-Si bonds
Alkali metals
Alkali metals diffuse rapidly into SiO2 and can damage silicon-based transistors
Transition
metals
All transition metals apart from gold and silver react with Si forming metal silicides
62 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nickel A very common contaminant of silicon Nickel from even slight contact with stainless
steel tweezers can contaminate silicon samples forming its silicide and is very difficult
to remove even at high temperatures because it segregates on the silicon surface
Palladium Forms silicides especially Pd2Si and can be used to make contacts possessing a low
Schottky barrier
Titanium Widely used to fabricate contacts for silicon-based microdevices TiSi2 is one of the
more stable and highly conductive silicides
Tungsten The silicide is more stable than TiSi2 on polysilicon and is thus used for polysilicon
gate contacts
Platinum Silicide is used for bipolar transistors PtSi contacts on n-silicon give Schottky diodes
Cobalt Cobalt silicide gives better quality contacts than TiSi2 at the nanometrelevel but is less
used at the moment because it consumes too much silicon during its growth
Iron Silicides have been studied for potential optoelectronic applications because FeSi2 has a
directly accessible bandgap and is optically active
Group 13
Group 13 metals are used in IC technology to dope silicon (giving p-type doping) and
to make III-V type semiconductors When evaporated on Si(100) or Si(111) they react
strongly with the surface giving surface reconstruction and kink formation at steps
Boron Widely used for doping Usually deposited with decomposition of boron hydrides
Sticking coefficient of simple hydrides is very low so decaboranes (B10H14) are used for
deposition
Aluminum Aluminum is widely used for microelectronic contacts but in some cases it reacts with
silicon diffusing into the crystal and with SiO2 stealing oxygen atoms to form Al2O3
Group 14
Carbon Carbon can form a composite semiconductor with silicon (silicon carbide) widely used
to fabricate integrated circuits for use at elevated temperatures or in the presence of
ionising radiation When carbon is deposited on silicon usually polycrystalline films
with properties between diamond and graphite are obtained
CO adsorbs molecularly on silicon without breakage of the C=O bond Other
unsaturated hydrocarbons do not react with silicon Even very reactive strained
molecules like cyclopropane have low sticking coefficient Ethylene acetylene and
propylene adsorb molecularly on Si(100) each molecule sitting on a silicon dimer
interacting with the dangling bonds of the surface Acetylene also adsorbs molecularly
on Si(111) on the 7x7 reconstructed surface preferential adsorption on centre atoms
with respect to corner atoms is observed Benzene and other aromatic ring compounds
adsorb molecularly on the surface They can form σ or π-type bonds with the silicon
laying parallel or tilted respect to the surface according to the bond type
Silicon Silicon atoms can be deposited on the surface with molecular beam epitaxy or
VPalermo 63
decomposition of silanes and chlorosilanes Silanes with single Si-Si bonds have a high
sticking probability and dissociate upon adsorption giving trihydrides and
monohydrides
Chlorosilanes have high sticking coefficients too and decompose on the surface but
they can etch the surface through the reaction SiCl4 + Si rarr 2SiCl2
Germanium Germanium has a lattice constant similar to silicon (Ge lc is 4 larger than Si) so that
various alloys of Si1-xGex can be formed The growth of Germanium on silicon is quite
peculiar with several monolayers adsorbing uniformly on the surface followed by
island formation (Stranski-Krastanov growth)
Group 15
Nitrogen Exposure to ammonia and nitrogen gas at high temperatures leads to the formation of
silicon nitride layers (Si3N4) NH3 dissociates on the surface and reacts with the
dangling bonds giving Si-NH2 and Si-H termination Silicon nitride is stable on the
surface up to 1100degC above this temperature it desorbs as Si2N
Phosphorous Phosphine (PH3) adsorbs dissociatively as Si-PH2 and Si-H and behaviour is similar to
that of ammonia
Arsenic A monolayer of As is often deposited on silicon as a substrate for GaAs growth Arsenic
forms dimers on Si(100) creating dimer rows which are parallel (perpendicular) to the
underlying silicon dimer rows when it is deposited at high (low) temperature On
Si(111) Arsenic breaks the 7x7 reconstruction giving a 1x1 pattern This 1x1 As
monolayer acts as a passivating layer and can resist further As adsorption oxygen and
air
Antimony and
Bismuth
Due to their large covalent radii Sb and Bi form only short dimer rows on silicon and
only at high temperatures
Group 16
Oxygen
Oxygen can oxidize or etch silicon according to the reactions
Si(s) + O2 rarr SiO2 (s)
Si(s) + frac12O2 rarr SiO(g)uarr
Silicon oxidation has already been described in the previous chapter Molecular oxygen
adsorption on silicon is mostly dissociative O atoms break the Si-Si bonds forming an
Si-O-Si bridge or a peroxy bridge Si-O-O-Si Hydrided silicon surfaces do not have
surface dangling bonds and are thus more protected from oxygen attack Etching is
favoured at high temperatures and low oxygen pressures [3] At T gt700degC silicon oxide
on silicon decomposes as SiO(g)
Water
H2O adsorbs dissociatively on silicon easily on Si (100) and with more difficulty on
Si(111) The molecule decomposes giving Si-OH and Si-H on adjacent sites
64 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Group 17
Fluorine
Fluorine adsorbs and reacts easily on silicon dissociating on the surface and forming
Si-F bonds It can then penetrate the surface and attack polarised Si-Si backbonds The
surface is etched with production of SiF4 and SiF2
Chlorine
Chlorine reacts aggressively with silicon etching the surface Cl2 and F2 are often used
commercially to etch silicon usually in a plasma Atomic Cl and Cl2 adsorb on Si(100)
up to saturation The most stable configuration seems to be a Cl atom bonded to each Si
atom of a surface dimer in a symmetric structure but metastable configurations with
two Cl atoms bonded to a buckled dimer and even a central Cl atom bridged across a
dimer have been observed
On Si(111) silicon mono- di- and tri-chlorides are formed In particular heating at
Tgt400degC a uniform Si-Cl monolayer is obtained and the 7x7 reconstruction changes
into a bulk like 1x1 lattice The surface transition induced by Cl adsorption is fully
reversible The 7x7 reconstruction can be restored on desorbing the chlorine by heating
at Tgt1100degC The 1x1 domains nucleate at the lower terrace side of steps the 7x7 at the
upper step edge
Bromine
Like hydrogen bromine maintains the 2x1 structure of Si(100) forming Si-Br bonds
with the dangling bonds of the silicon dimer rows At high exposure some etching of
silicon by formation of volatile SiBr3 species has been observed
Surface diffusion on silicon
In the previous chapter we observed nano-island growth on silicon surfaces with native
oxide and surfaces contaminated with organic impurities While this phenomenon has been
observed in several experiments the formation and growth mechanisms of nanoislands has
still not been satisfactorily described theoretically Various substances have been
hypothesised as nanoisland nucleating agents such as organic or metallic contaminants
present on the surface or oxide clusters (see previous chapter) The island growth process
has been attributed by various workers to lsquosome sort of kinetic instabilityrsquo [4] to the
pinning of step flow by SiC clusters [5] or to a mesoscopic atom flux from areas with low
step density to areas of higher step density [6]
One mechanism proposed to explain surface roughening and island growth is the Ehrlich-
Schwoebel effect ie the presence of an energetic barrier that reflects atoms approaching a
VPalermo 65
Molecule (L
O2
H2O MeOH
CO CO2 CH4
clean surface
downward
weak on si
The aim o
surface an
To obtain
the ones
spontaneou
first obser
surface W
nanoisland
Experime
We used S
removed f
41 mixtur
water and
whole hea
1200degC re
kept below
Table 62 Treatment island density and island volume for each molecule tested
Dose angmuir)
Heating time at 800degC (min)
Island density (microm-2)
Ave Volume (nm3)
Equivalent layer thickness (nm)
280 10 no islands - -
90 10 no islands - - 30 10 2000 19 004
180 10 360 516 019 30 60 no islands - -
180 10 no islands - - - 10 no islands - -
step [7] Recent experiments however indicate that the Schwoebel effect is
licon [48]
f our experiments was to study the reaction of simple molecules with the silicon
d to check the ability of these molecules to nucleate nanoisland growth
a better understanding of the process we chose simple molecules starting from
which are more likely to be present as traces in UHV chambers where
s nanoisland growth is often observed after sample heating Using STM we
ved at the atomic level the possible interactions of these molecules with the
e then heated the surface and checked the density and dimensions of the
s produced
ntal procedures and results
i(111) wafers p-doped 0015 Ω cm Gross contamination and particles were
rom the sample surface with a standard Pirana etch (15 min dip in a H2S4H2O2
e at T=80degC) After this the samples were thoroughly washed in ultra-pure
introduced into the UHV system After degassing for 5 hours at 600degC the
ting stage was allowed to cool down and a sequence of rapid flashes to 1100deg-
moved the native oxide layer and cleaned the surface Pressure during flashes is
1x10-9 mbar
66 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Surface flatness and cleanliness were checked with STM and the contaminant molecules
introduced into the UHV system through a leak valve Doses of several Langmuirs were
used for the experiments (1 Langmuir=10-6 torr sec-1) Pressure was kept constant at 10-7
mbar during adsorption The STM tip was retracted during gas deposition to avoid any tip
shielding effect After the adsorption process the leak valve was closed the base pressure
in the UHV chamber returning to below 10-10 mbar and the presence of adsorbed
molecules checked on the surface with STM Finally the sample was heated at 800degC for
several minutes to allow nanoisland growth Several different areas were examined on each
sample
Table 62 summarizes the dose of exposure and the heating time for each molecule Other
experiments were carried out using smaller dosing times but only the significant results
are reported here
As expected oxygen and water react with the surface oxidising it The adsorption process
can be easily followed with the STM by lowering the tip from time to time and taking an
image No effect of the STM tip is observed over the scanned areas Upon molecule
adsorption the order of the crystalline surface rapidly degenerates and after several
minutes the surface looks completely covered by irregular atomic-sized protrusions
corresponding to Si-OH or Si-O-Si species On heating to 800degC the contaminating species
were easily desorbed through SiO formation the surface became clean and the crystal
surface periodicity re-established No residues or nanoislands were observed
Amongst the molecules tested was methanol CH3OH reacts readily with the surface and
covers it with a disorderd layer already after sim 30 Langmuirs exposure (fig 61)
Fig 61 Si(111) surface during methanol adsorption at 0 1 and 5 minutes respectively (corresponding to 0 6 and 30 L exposure) The dark irregular lines are borders between different 7x7 domains Image size 60x60 nm
VPalermo 67
Synchroton radiation photoemission spectroscopy experiments by Carbone et al [9]
showed that methanol reacts with the rest atoms of the 7x7 silicon cell already after 1
Langmuir exposure but that longer exposures are needed for methanol to react with the
remaining silicon atoms Methanol adsorbs dissociatively forming SiOCH3 and Si-CHx
species on the surface
After methanol adsorption the sample is heated to 800degC and observed again with STM
As in the case of water and oxygen the heating restores the crystalline surface and the 7x7
pattern returns visible But this time quite a dense array of nanoislands is observed on the
surface (fig 62) The islands have an average diameter of 20 nm and a height of sim2 nm
At temperatures above 400degC the Si-OCH3 and Si-CHx species decompose oxygen and
hydrogen are desorbed and according to [9] carbon atoms remain as SiC dispersed
uniformly on the surface Our STM measurements indicate that the carbide does not
randomly cover the surface but that C atoms are concentrated at the nanoislands Rough
calculations based on island volume suggest that the islands must be composed of a SixC1-x
alloy with x varying between 05 and 1
Following these results we expected carbon monoxide to behave in a similar manner
adsorbing onto the surface and with nanoisland growth However no adsorption was
observed with STM and the surface looked perfectly clean even after a dose of more than
100 L of carbon monoxide Some species did adsorb onto the surface though because
after 10 minutes subsequent heating nanoisland growth was observed Island density in
this case was only 18 of the density obtained with methanol and island dimensions are
Fig62 Nanoislands on Si(111) created after 30 L adsorption of methanol and 10 min heating at 800degC Image size 200x200 nm
68 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
much bigger (see table 62 and fig 63) To check whether the islands had been nucleated
by some other contaminant present in the vacuum chamber we heated a blank sample
without introducing any molecule No island growth was observed Exposing the surface to
other simple molecules including carbon dioxide and methane also gave no nanoisland
formation
We can estimate the amount of carbon deposited on the surface during gas dosing The
total flux of molecules hitting the surface is obtained from the Hertz-Knudsen formula
TkmPF
π2=
Fig 63 STM images of Si(111) after exposure to methanol (left) and carbonmonoxide (right) contaminating molecules and subsequent nanoisland growthImage size 400x400 nm The size distribution of nanoislands is shown below eachimage
VPalermo 69
Fig64 Left Si(111) after 180 L contamination of CO and 16 hours heating at 800degC Right Si(111)without CO contamination after 16 hours heating at 800degC Image size 250x250 nm Vertical scale is thesame for both images z-ranges are 35 and 5 nm respectively
where P is the pressure in Pascal k=138x10-23 J K-1 is the Boltzmann constant T is the
temperature and m is the molecular weight (sim32 for methanol and sim28 for CO) in
kilograms
Using this formula we obtain a flux of 027 molecules nm-2 sec-1 for methanol and 029
molecules nm-2 sec-1 for CO After ten minutes exposure at 10-7 mbar pressure more than
150 molecules will have hit each square nanometer of the surface Surface density of
Si(111) 7x7 is sim16 atomsnm2 so each surface atom will be hit by several molecules which
could react or be adsorbed even assuming a sticking coefficient much lower than unity
To explain nanoisland growth we hypothesize that CO molecules adsorb molecularly and
-5
0
5
10
15
20
25
30
0 20 40 60 80 100 120
nm
nm
10 min40 min16 h16 h- no CO
Fig65 STM profiles of nanoislandsgrown for different annealing times The lower curve correspond to the surface heated at 16 hours without CO contamination
70 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Table 63 Island density and volume for the CO-contaminated surface at different heating times For each sample the thickness of a uniform layer having the same volume per unit area of the islands is calculated
Molecule Dose
(Langmuir) Heating time at
800degC (min) Island density
(microm-2) Ave Volume
(nm3) Equivalent layer thickness (nm)
CO 180 10 362 517 019 CO 180 40 550 1072 059 CO 180 16 h 505 3675 186
clean surface - 16 h rough - -
very weakly on Si(111) moving rapidly on the surface Because of their rapid diffusion
the STM is not able to image the CO molecules on the silicon [10]
Increasing the temperature part of the CO molecules will surely desorb but some of them
will dissociatively adsorb onto the surface The oxygen will then desorb as SiO while
carbon will remain fixed forming very stable Si-C bonds
CO decomposition will be favored at surface defects surface steps or in the proximity of
already formed SiC clusters An increased reactivity of the CO molecule on a surface in
correspondence with phase boundaries has already been observed on Pt surfaces [10]
In this case a reduced number of nucleation centres will react with the CO molecules
yielding large and fewer islands with respect to the case of methanol which does not have
the possibility of travelling long distances over the surface
Other factors are likely to influence the process of island nucleation after CO adsorption
the co-adsorption of other molecules which can slow down CO and favour its
decomposition cannot be ruled out as well as the formation of new defects at higher
temperatures Measurements with a variable temperature STM or with some other surface
analysis technique are planned in the future to confirm the proposed mechanism
When the surface was further annealed the nanoislands grew in size After 16 hours
heating very large nanoislands with diameters of sim35 nm and heights of sim20 nm became
visible (fig64 left) As a comparison when the same surface was heated without
nucleation centres deriving from contamination an irregular surface was obtained (fig64
right)
While the nanoisland density seemed to reach an asymptotic limit of 500 islandsmicrom2
(table 63) island size continued to grow with time (fig65) even when no further carbon
was supplied to the surface This indicates that even though the initial nucleating core of
the nanoisland is likely to be an SixC1-x alloy further growth is due to silicon atoms
diffusing from the crystal and being adsorbed by the growing island The final volume
VPalermo 71
occupied by the islands corresponds roughly to a 18 nm thick overall layer of removed
silicon
It is noteworthy that the clean sample after 16 hours annealing even though smoother than
the CO-contaminated one looked much more disordered with hardly any flat area visible
on nanometres scale In the presence of nanoislands however further annealing increases
island size but keeps part of the surface quite flat and ordered with flat areas visible
between the islands Thus the nanoislands can be imagined to act as ldquoimpurity sinksrdquo for
further contaminants approaching the surface yielding a greater but more ordered surface
roughening with respect to the uncontaminated silicon surface
Circular arrays of nanoislands
We used the voids described in the previous chapter formed during oxide layer
decomposition as nanoscopic masks to test the validity of the hypothesis outlined above
The small circular areas uncovered during oxide decomposition can act as ldquoskating rinksrdquo
where contaminants and diffusing silicon atoms can move freely over the surface while
the oxide layer all around and higher than the void area as well as being chemically
bonded to the surface silicon atoms will hinder surface diffusion Furthermore the oxide
passivating layer will at least in some cases prevent contaminating molecules from
adsorbing and decomposing on the surface In this way selective nanoisland growth inside
void areas can take place
The experimental procedure consisted of the following steps
1) An oxide-passivated sample was heated for 10 minutes at 800degC Oxide decomposition
began in correspondence with surface defects and spread laterally creating naked areas
of silicon
2) While the void enlarges the same defect that initiated oxide decomposition behaved as
an attractor for moving silicon atoms and nucleated the growth of a nanoisland at the
void centre
3) The void surface was exposed to contaminating molecules
4) The sample was then heated again at 800degC During the second heating the void
continued to enlarge and new nanoislands were created The freshly uncovered silicon
surface provided mobile silicon atoms to feed nanoisland growth
72 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
5) When the heating continued for long enough all the oxide was desorbed and a clean
silicon surface covered by circular groups of nanoislands was obtained
Figure 66 shows the different stages of the process
We thus heated an oxidized sample at 800degC for 10 minutes in UHV to create multiple
voids in the oxide layer After this we exposed the sample to the contaminating molecules
Then we further annealed the surface at the same temperature to promote island growth
The first time we observed nanoisland growth in the void area was by chance after heating
for a second time samples which had been stored for 2 weeks in vacuum Even at pressures
as low as 10-11 mbar some molecules will adsorb onto the surface slowly creating defects
and thus potential nanoisland nucleation centres It is possible to observe with STM that
the silicon crystal surface looks very clean just after a high temperature flash but even
after only a few days randomly adsorbed species will be observable on the surface
Unfortunately it is not possible to identify these species simply from in situ STM
measurements
Mass spectrometry measurements show that the residual gas contaminants in vacuum are
usually He Ar H2 CH4 CO CO2 and N2 [11] Some of these molecules such as
hydrogen will stick to the silicon surface at room temperature but will simply desorb
during annealing without nucleating nanoislands Others like CH4 or CO2 will not react
with the surface to generate nanoislands as demonstrated previously The best candidate
for contamination of samples stored in UHV seems thus carbon monoxide has the ability
321
4 5
Fig 66 Selective nanoisland growth within oxide voids See text for details
VPalermo 73
as shown before to nucleate nanoislands although co-adsorption mechanisms can not be
ruled out Fig 67 (left) shows the circular groups of nano-islands obtained after heating
void-covered samples stored for 2 weeks in UHV Small islands are observed in the former
void area The void itself has enlarged slightly uncovering a clean oxide-free area of the
silicon surface Apart from the central island created during the first period of heating the
surrounding ones have dimensions decreasing from the void border to the inner area of the
void
This confirms that most of the material needed for island growth comes from the silicon
atoms diffusing from the freshly uncovered areas around the original void The outer
nucleating centres are nearest to the silicon atom source and thus generate larger islands
For a faster process we directly exposed the void to a significant concentration of CO
Fig67 (centre) shows the islands grown after a 180 Langmuir exposure and subsequent
heating The islands are preferentially located around the original void perimeter the rest
of the original oxide-free surface remaining untouched This suggests that although the
whole surface is exposed to CO the gas only interacts chemically with the surface at the
SiSiO2 interface around the oxide-free void perimeter The SiSiO2 interface thus provides
preferential nucleation sites for nanoisland growth in the successive thermal annealing
step This recalls a similar phenomenon observed in the catalytic oxidation of CO on the
Pt(111) surface where CO molecules reacted with adsorbed O species only along the
perimeters of oxygen islands [10]
On repeating the CO exposure and thermal annealing steps a second circle of nano-islands
Fig67 Left circular area of nanoislands obtained from a sample stored 2 weeks in UHV 500x500 nmCentre nanoisland circle obtained after exposing the voids to 180 L of CO and heating for 10 min at800degC 400x400 nm Right two concentric circles obtained with further CO adsorption followed byheating 650x650 nm
74 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
was formed around the newly-expanded surface void perimeter (fig67 right) This
process can in principle be repeated again to obtain a series of concentric groups of
nanoislands the only limit depending on the minimal distance between neighbouring
expanding voids We have thus observed that contaminants adsorbed at low pressure over
long periods of time generate random island growth over the whole void area while CO
molecules adsorbed at relatively high pressures over short periods yield selective
nanoisland growth at the former void border Several theories can explain the different
results CH4 and CO2 as mentioned above did not nucleate islands when when the silicon
surface was exposed to them for short periods However the lower surface mobility of
these contaminants and the greater time available for reaction with the surface could be
the cause of nucleation of SiC clusters over the whole oxide-free void area and not only at
its border The coadsorption of water another UHV residual gas with CO on the naked
silicon areas could perhaps allow a surface reaction similar to that between CO and the
SiSiO2 interface during CO exposure Another explanation for the formation of these
random nucleation sites within the voids could be the coadsorption of residual carbon-
containing species present in the UHV system with hydrogen the major residual gas in
stainless steel UHV systems
The possibility of decorating silicon oxide borders with nanoislands using the high
mobility and selective decomposition of CO molecules suggests interesting applications
for nanofabrication processes that will be discussed below
Silicon nanowire creation on Si(100)
We performed most of the experiments described above on silicon (111) 7x7
reconstructed because this surface is easily prepared in UHV and above all because it is
isotropic and has no preferential directions for atom diffusion The 2x1 reconstruction of
silicon (100) even though it has a much simpler unit cell than Si(111) 7x7 shows a more
complicated morphology with lots of monoatomic and biatomic steps and dimer rows
parallel to each other The orientation of the dimer rows changes by 90deg on alternate
atomic layers Surface diffusion in this case is much more complicated because atoms will
experience the effects of the surface anisotropy and diffuse preferentially along or across
the dimer rows The diffusion energy of a silicon atom as example is 06 eV along a
dimer row and 085 eV across different dimer rows This while being a complication
VPalermo 75
offers interesting possibilities for the creation of ordered structures The monoatomic steps
and the dimer rows can for example act as templates for the formation of elongated
structures of composition similar to the islands described previously
On heating a Si(100) surface oxide decomposition and void growth takes place as
described for Si(111) In some cases the void shape reflects the surface symmetry
especially for very thin oxide layers and the voids often have a slightly squared shape
Nanoisland growth is also influenced by the substrate symmetry and the islands look
rectangular as irregular clusters surrounded by concentric patterns of monoatomic steps
The whole structure reminds vaguely a ldquozigguratrdquo a kind of stepped pyramid found in
Mesopotamia (fig68 inset)
Apart from nanoisland growth in some cases we observed spontaneous formation of
nanowire-like SiC structures on the Si(100) surface The wires were generated after
exposing a Si(100) surface in vacuum to traces of CO followed by sample annealing at
800degC for 15 minutes The wires exhibit lengths ranging from 10 to 100 nm and average
widths of ca 5 nm Although being randomly positioned on the surface they are perfectly
aligned along the crystal axes of the substrate (Fig68)
Several descending monoatomic steps can be observed around each nano-line and nano-
island Very often one of the line extremities coincided with an island
Nanoline
Nanoisland
Fig 68 A Si(100) surface covered by nanoislands and nanowires The monoatomic steps present on the surface are visible The two insets show a typical nanoisland and a nanowire obtained on this surface The typical ldquodimer-row ldquostripes are visible along the sides of the nanowire
76 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Couples of parallel lines starting from the same nano-island were also observed At high
resolution (fig 69) the lines showed a lamellar periodic structure with a periodicity of
around 4 nm and frequent defects Scanning Tunneling Spectroscopy measurements
showed no significant difference between the line and the surrounding silicon surface
The proposed formation mechanism for this kind of structure involves two stages in the
first one organic contaminating molecules adsorb and diffuse on the surface from the
vacuum environment stopping preferentially at surface singularities including steps and
defects
Several substances such as atomic carbon carbon monoxide ethylene and fullerene
decompose when adsorbed on silicon surfaces at temperatures above 500degC and create
localised defects on the surface inducing strain deformations and a metastable surface
reconstruction [12]
If these molecules remain attached to the steps they can diffuse rapidly along step border
until they meet a line and decompose In this way long lines of SiC defects parallel to the
surface steps can be produced In the second phase the clusters of silicon carbide can
-005
0
005
01
015
02
025
03
035
04
-3 -2 -1 0 1 2 3
V
dId
V lt
dId
Vgt
linesilicon
Current- Voltage behaviour measured with STS Fourier analysis of wire periodicity
asymp 4 Aring
High resolution image of a line
Fig 69 Top high resolution image of a nanowire showing the lamellar structure Fourier analysis (bottom left) shows that the lamellae periodicity is sim 04 nm bottom right the IV characteristics measured with the STM tip on the line and the silicon surface
VPalermo 77
25-600degC
Final linear structure
600-800degC
Contaminant Silicon
Fig 610 Schematic representation of nanowire formation mechanism
locally inhibit the spontaneous surface diffusion of silicon atoms acting as templates for
the reorganisation of surface steps (fig 610)
The final morphology shows disordered lines and islands each surrounded by a complex
pattern of silicon monatomic layers
At 800degC silicon atoms are highly mobile on the surface Due to atom diffusion the
surface reorganizes with a step-flow mechanism The presence of a fixed line of
contaminants blocks atom diffusion and step flow generating the monoatomic steps
around each wire (fig 611) SiC clusters act as lsquopinning sitesrsquo on silicon atom surface
diffusion and can be purposely used to direct the localised growth of nano-islands [13]
Although the possibility of using methanol and carbon monoxide as nano-island precursors
has been demonstrated the exact nature of the substances nucleating nano-wire growth is
still unknown
Further experiments will be needed to confirm these hypotheses as well as to identify the
contaminating species nucleating nanowire growth Carbon monoxide is a main candidate
for nanowire growth but the real nucleation process is likely to be a complex one
78 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Step flowblocked
Fig 611 Representation of step flow blocked by the presence of a nanowire 105x105nm
Bibliography
[1] Surface-Chemistry Of Silicon Waltenburg HN Yates JT Chemical Reviews 95 (5)
1589-1673 Jul-Aug 1995
[2] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[4] Growth Of Si On The Si(111) Surface Lanczycki Cj Kotlyar R Fu E Yang Yn
Williams Ed Das Sarma S Physical Review B 57 (20) 13132-13148 May 15 1998
[5] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
VPalermo 79
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990
[6] Nanoscale roughening of Si(001) by oxide desorption in ultrahigh vacuum Gray SM
Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology B 14 (2)
1043-1047 Mar-Apr 1996
[7] Step Motion On Crystal Surfaces Schwoebel Rl Journal Of Applied Physics 40 614-
618 Feb 1969
[8] Scanning Tunneling Microscopy Investigation At High Temperatures Of Islands And
Holes On Si(111)7x7 In Real Time Evidence For Diffusion-Limited Decay
Hildebrandt S Kraus A Kulla R Wilhelmi G Hanbucken M Neddermeyer H Surface
Science 486 (1-2) 24-32 Jul 1 2001
[9] Methanol Adsorption On Si(111)-(7x7) Investigated By Core-Line Photoemission And
Mass Spectrometry Of Photodesorbed Ions Carbone M Piancastelli Mn Zanoni R
Comtet G Dujardin G Hellner L Surface Science 370 (1) L179-L184 Jan 1 1997
[10] Atomic And Macroscopic Reaction Rates Of A Surface-Catalyzed Reaction
Wintterlin J Volkening S Janssens Tvw Zambelli T Ertl G Science 278 (5345)
1931-1934 Dec 12 1997
[11] Redhead PA Hobson JP Kornelsen EV The Physical Basis Of Ultrahigh
Vacuum Chapter 12 (Chapman amp Hall London 1968)
[12] The Si(001) C(4 X 4) Surface Reconstruction A Comprehensive Experimental
Study Norenberg H Briggs Gad Surface Science 430 (1-3) 154-164 Jun 21 1999
[13] Production Of Nanostructures Of Silicon On Silicon By Atomic Self-Organization
Observed By Scanning Tunneling Microscopy Jones D Palermo V Applied Physics
Letters 80 (4) 673-675 Jan 28 2002
80 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 81
7 Conclusions and proposals for future work It is very difficult to foresee the future of nanoscience and nanotechnology even over the
next few years exciting discoveries are continuously made opening up new possibilities
and progress in this sector is now one of the fastest and most varied So it is impossible to
predict whether computers will in the future be made with nanowires nanotubes single
electron devices or some kind of quantum-based transistor What does seem clear
however is that silicon devices can still further extend their levels of miniaturization to
remain competitive for at least the next ten years [1] During this period new emerging
technologies will reach scientific maturity and arrive at the production lines
Even when new technology prototypes do become available for full-scale production the
astronomical cost of changing from silicon to new technologies will be a major concern
Moreover a vast amount of extremely detailed knowledge is already available on all
aspects of silicon technology Thus the possibility of integrating innovative
nanotechnologies with standard CMOS silicon technology already seems an attractive
prospect [2]
For this and other reasons we have focussed our research efforts on surface phenomena on
silicon wafers which could be potentially useful for the development of silicon-compatible
nano-devices
Some important conclusions can be drawn from the work described in this thesis
bull Silicon surfaces of different crystal faces were studied at atomic resolution in ultra-
high-vacuum following transformations induced on the surface by chemical (etching)
and physico-chemical (organic contamination heating) treatments
82 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull The presence of surface contaminants induces at high temperature the growth of
nanometre-sized islands and lines on silicon
bull Final island densities and dimensions strongly depend upon the chemical nature of the
contaminant molecules and the treatments used Molecules interacting only weakly
with silicon will desorb upon heating whereas molecules which decompose easily on
Si will give widespread random nano-island nucleation Between these two extreme
cases complex behaviours of diffusion and subsequent decomposition are possible
bull Nanoscopic naked silicon voids were produced in oxide layers through carefully-
controlled thermal annealing It was found that void density and size depends upon
oxide purity surface contamination levels and annealing conditions
bull The growth of nano-islands and nano-voids was exploited simultaneously and the
voids used as nanoscopic masks to control the positions of growing islands Complex
silicon-on-silicon nano-structures were obtained with multi-step growth processes
bull Finally the use of macroscopic treatments and simple physical processes to produce
large quantities of nanometre-sized structures on silicon surfaces was demonstrated
These processes and the structures produced are of potential interest to the
microelectronics industry
We showed in previous chapters how it is possible to modify silicon surfaces using
chemical and physical methods The use of self-organization for the production of Si-on-Si
20 nm
Fig 71 Left SEM image of field emitter arrays made by conventional lithography [3] Center 3-D STM image of a self-organized void with a central nano-island Right Lateral schematic view of a field emitter device
VPalermo 83
nano-structures has the advantage with respect to other emerging techniques of being
silicon compatible and thus is a potential candidate for its implementation in the
production of new devices Furthermore this technique can create billions of strucures on a
wafer in only a few minutes being thus applicable to large-scale production
Even when ordered structures are obtained these methods still show a poor reproducibility
and control Chemically modified surfaces look very disordered at high magnification
with trenches and holes etched into the surface and progress has recently been made in
producing very small structures electrochemically [4] Oxide decomposition with
nanoisland growth in UHV can also yield ordered structures The void holes with
nanoislands located at their centres although quite variable in dimensions are more similar
to engineered devices than to a spontaneous random surface-roughening process In fact
there is a surprising similarity between the nanovoid-nanoisland structures and well-known
commercially available devices called field emitter arrays (FEA) These are small tips used
as electron guns in various kinds of displays and other devices Both the commercial
devices and our self-organized structures consist of a protruding tip at the centre of
microscopic holes on an insulating layer covering a semiconductor surface (see fig71)
Of course the order and size uniformity of commercial FEAs is much better than that
obtained with our process but it is encouraging to see that a self-organized structure
obtained with simple heating has a very similar surface topography and composition of
commercial working devices obtained after many expensive stages of optical lithography
deposition and etching
Nano-island density and size can to some extent be controlled and different
contaminants can be adsorbed onto the surface to modify the growth process In the future
with a better understanding of diffusion dynamics and growth mechanisms a multi-stage
fabrication process can be envisaged where purposely-designed molecules are adsorbed
onto silicon surfaces to diffuse organize in ordered arrays and favour surface
reorganization at the nanometre scale The same molecules could be used to deliver
dopants to obtain for example highly conducting nanostructures on otherwise undoped
highly resistive silicon The selective formation of nanostructures at the SiSiO2 border
can be used to scale down mesoscopic patterns to the nanometric level A silicon oxide
layer can be easily patterned on the surface with normal optical lithography and then
nanostructures can be built following the pattern edge In this way the production of very
small nanowire based devices may be possible (fig 72)
84 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
D
Gate
S
D
S
SiO2
SiO2
2 Heating nanowire creation
and oxide removal
3 Source-drain deposition by conventional lithography
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
4 Insulating layer (oxide) and gate (metal) deposition
Fig 72 Schematic drawing of the construction of a self-organized nanowire-based transistor
Surface diffusion and decomposition of molecules can be sensitive to even smaller surface
features The nanoscopic lines formed on Si(100) are very regular and neat if their
nucleation and growth mechanism could be controlled ordered networks of nanowires
could be made on silicon
VPalermo 85
The techniques developed here are still experimental ones the careful control and
reproducibility of each step is still an issue and further studies will be needed to clearly
understand the dynamics of formation of these structures Nonetheless the results obtained
so far open up the possibilities of producing novel devices such as for example nano-
transistors (see scheme in fig73 where the mechanism described in fig 72 is applied for
the fabrication of a device array) The possibility of using the border of a lithographic
pattern to draw nanolines can in principle allow the creation of nanodevices using normal
IC manufacturing techniques
The results obtained and their description in this thesis are not the first examples of the use
of self-organization for the production of ordered structures and will certainly not be the
last In our opinion the most important outcome of these findings is that it is possible to
use surface diffusion to overcome the inherent limits of lithographic techniques in
microdevice production processes and to force properly selected molecules to react with
the surface in a spatially differentiated manner at particular sites thereby creating ordered
series of nanostructures The role of the surface dynamics of the substrate atoms
themselves is extremely important in this process
Outstanding results have been and continue to be obtained in nanoscience and
nanotechnology research promising the emergence of new production technologies Sooner
or later those emerging technologies will have to face the issues of cost and compatibility
with the enormous investment made in existing technologies and processing facilities The
commercial production of nanostructures based on surface diffusion processes could in
principle be cheap simple and compatible with existing technology Moreover the
processes of diffusion in bulk silicon is well-known in the microelectronics industry which
has both the know-how and the equipment to better understand and exploit the surface
diffusion processes on silicon surfaces studied here
Extremely important and innovative results are often obtained by exploiting very simple
ideas and in our opinion physical processes such as surface diffusion used to create
complex nanostructures on silicon surfaces is a prime example
86 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
2 Heating nanowire creation and
oxide removal
3 Source-drain deposition by conventional lithography
4 Insulating layer (oxide) and gate (metal) deposition
Fig 73 Schematic macroscopic outline of the construction of a self-organized nanowire-based transistor
VPalermo 87
Bibliography
[1] P Gargini Intel Technology Coordinator ldquoFrom Microelectronics To
Nanotechnologiesrdquo Invited Campus Colloquia Lecture CNR Research Area Bologna
February 6th 2003
[2] Toward A Hybrid Micro-Nanoelectronics Cerofolini Gf Ferla G Journal Of
Nanoparticle Research 4 (3) 185-191 Jun 2002
[3] Recent Progress In Field Emitter Array Development For High Performance
Applications Temple D Materials Science amp Engineering R-Reports 24 (5) 185-239
Jan 25 1999
[4] Silicon Dioxide Micropillars For Sieving Fabricated By Macroporous Silicon-Based
Micromachining Izuo S Ohji H French Pj Tsutsumi K Kimata M Sensors And
Materials 14 (5) 239-251 2002 Electrochemical Etching In HF Solution For Silicon
Micromachining Barillaro G Nannini A Piotto M Sensors And Actuators A-Physical
102 (1-2) 195-201 Dec 1 2002
88 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 89
Acknowledgements
This work would not have been possible without the help and guidance of Dr Derek Jones
he introduced me to the use of STM and other techniques found the funding for this PhD
and helped me with the thousands of small and big problems I had to face during my
research I am grateful to Prof Alberto Ripamonti too for being the supervisor of this
thesis Thanks are also due to Dr Giancarlo Seconi director of ISOF-CNR where most of
this work was carried out Financial support from the Italian National Research Council
(CNR) is also gratefully acknowledged
My passion for surface science began when Fabio Biscarini showed me that it was possible
to really see the atoms and for this I will always thank him
During these years I had the luck to collaborate with many good scientists among those I
would like to cite Enrichetta Susi Massimo Cocchi and Anna Mazzone from CNR
Claudio Zannoni Anna Cavallini Daniela Cavalcoli and Antonio Castaldini from the
University of Bologna Sergio Pizzini and Maurizio Acciarri from the University of Milan
One of the most stimulating periods of my PhD was the one I spent at the Steacie Institute
for Molecular Sciences in Ottawa working with Robert Wolkow Dan Wayner Greg
Lopinski and Peter Kruse I thank them for discussing together many exciting scientific
ideas and trying to put some of them into practice
Special thanks go to my colleagues Paolo Samorigrave and Stefano De Cesari for long and
useful discussions on Science Life and their interactions often with the help of a good
pint
Finally I would like to thank Vassilia Gaetano Simone Claudia Silvia Letizia
Alessandro Angela and Sebastien even though not directly involved in this work they
have been over these three years a continuous source of happiness and support
Bologna March 2003
90 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
SCIENTIFIC PUBLICATIONS bull Lateral diffusion of titanium disilicide as a way to contacting
hybrid Si-organic nanostructures Palermo V Buchanan M Bezinger A Wolkow RA APPLIED PHYSICS LETTERS 2002 v81 p 3636 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v6 issue 20) bull Surface modifications in Si after Rapid Thermal Annealing Castaldini A Cavalcoli D Cavallini A Jones D Palermo V Susi E JOURNAL OF THE ELECTROCHEMICAL SOCIETY 2002 v 149 pG633 bull Production of nanostructures of silicon on silicon by atomic self-
organisation observed by scanning tunnelling microscopy Jones D Palermo V APPLIED PHYSICS LETTERS 2002 v 80 p 673 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v5 issue 5) bull Nucleation of nanostructures from surface defects on silicon Palermo V Jones D SOLID STATE PHENOMENA 2002 v 82-84 p 687 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo MATERIALS SCIENCE AND ENGINEERING B 2002 v88 (2-3) p220 bull Morphological changes of Si[100] surface after treatment with
concentrated and dilute HF Palermo V Jones D MATERIALS SCIENCE IN SEMICONDUCTOR PROCESSING 2001 v 4 p 437 bull Electrical and structural properties of processed silicon surfaces Susi E Cavallini A Castaldini A Cavalcoli D Jones D Palermo V ldquoRECENT RESEARCH DEVELOPMENTS IN VACUUM SCIENCE amp TECHNOLOGYrdquo 2001 v 3 p189 bull Numerical Solutions of the Stochastic Equations of Crystal
Growth Mazzone A M Palermo V INTERNATIONAL JOURNAL OF MODERN PHYSICS C 2000 v 11 Part 1 p195-204 bull Advances in silicon surface characterisation using light beam
injection techniques
VPalermo 91
Acciarri M Pizzini S Simone G Jones D Palermo V MATERIALS SCIENCE AND ENGINEERING B 2000 V73 (1-3) p 235 - 239 bull Abrupt orientational changes for liquid crystals adsorbed on a
graphite surface Palermo V Biscarini F Zannoni C PHYSICAL REVIEW -SERIES E- 1998 V 57 NUMBER 3A p R2519-R2522 Oral presentations bull The use of oxide desorption and surface diffusion for the
creation of silicon on silicon nanostructures Palermo V Jones D 1ST NATIONAL WORKSHOP ON CURRENT TRENDS IN NANOTECHNOLOGIES Catania (Italy) Feb 2002 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo V E-MRS Spring Meeting Strasbourg (France) June 2001 bull STM study of surface transformations on silicon during UHV
annealing Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999 bull Advances in silicon surface characterisation using light beam
injection technique Acciarri M Pizzini S Simone G Jones D Palermo V E_MRS SPRING MEETING June 1999 Posters bull Spontaneous nano-wire growth on silicon Palermo V Jones D SMARTON workshop Leuven (Belgium) October 2002 bull Ordered circles of nano-islands on silicon from CO adsorption Palermo V Jones D TRENDS IN NANOTECHNOLOGY 2002 Santiago de Compostela (Spain) September 2002 bull Production of nanostructures of silicon on silicon by atomic self-
organisation Palermo V Jones D EUROMAT Rimini (Italy) June 2001 bull Etching holes and anisotropic corrosion on silicon [100] Palermo V Jones D Susi E Asoli B SILICON WORKSHOP Genova (Italy) February 2001
92 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull Morphological and electrical characteristics of damaged silicon surfaces
Susi E Castaldini A Cavalcoli D Cavallini A Jones D Palermo V SILICON WORKSHOP Genova (Italy) February 2001 bull Effect of HF etching on the roughness of a silicon surface Palermo V Jones D NATIONAL CONFERENCE ON PHYSICS OF MATTER Genova (Italy) June 2000 bull Effect of Native Oxide Desorption upon the Surface Morphology of Si[100] by STM and LEED Palermo V Jones D SILICON WORKSHOP Genova (Italy) February 2000 bull Characterisation of silicon surfaces for microelectronics through STM measurements Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999
VPalermo 93
Contact Vincenzo Palermo ISOF- Institute for Organic Synthesis and Photoreactivity Via Gobetti 101 40129 Bologna ITALY Tel +39-051-6398336 Fax +39-051-6398349 Mail palermoisofcnrit
94 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Index
1 Introduction nanotechnology and the future of computers
2 Silicon surfaces
3 STM and other surface analysis techniques
4 Surface modification of silicon in liquid Nano-hole creation
5 Surface modification of silicon in vacuum void creation and oxide desorption
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and
nano-lines
7 Conclusions and proposals for future work
VPalermo 7
1 Introduction nanotechnology and the future of computers The motivations of nanotechnology research
In 1965 an electronic engineer named Gordon Moore one of the future founders of Intel
noted that the performance of computers and their complexity doubled every 18 months
and foresaw that computer power would continue to grow exponentially over the following
years
This prevision quite provocative for its time actually came about and gained the name of
ldquoMoorersquos Lawrdquo and continues to hold for the trends of todayrsquos computer industry Since
1965 the number of transistors present in an integrated circuit (IC) has increased from
several hundred to more than ten million and the minimum size of transistor elements has
shrunk from several millimetres to asymp130 nanometres (fig 11)
Devices of such tiny dimensions are actually fabricated using lithographic techniques
where light is passed through an optical mask to react with a photo-sensitive layer (resist)
Fig11 Evolution of the number of transistors present on commercial computers [1]
8 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
on the silicon wafer This resist is then selectively removed and used as a mask for
processing the silicon surface (fig 12 left) The maximum resolution attainable depends
upon the wavelength used and current technology is near to its intrinsic resolution limit
On the other hand there is strong scientific and economic demand for further development
in IC miniaturization to obtain more powerful and complex computers Besides every-day
life applications more powerful computers are fundamental for much scientific research
such as climate change tracking genome sequencing and fluid dynamics Increased
miniaturization is also fundamental for reducing power which has to be dissipated by the
chips which run at progressively higher frequencies Energy consumption by
microelectronic devices is already an issue and represents one of the main obstacles for
the continuing growth in wireless communication (cell phones portable computers CD
and DVD players digital cameras etc)
Thus it is expected that new production methods different from current lithographic ones
will be developed methods which allow modification of a surface well below the 100 nm
limit and even down to single atom manipulation Techniques such as Scanning
Tunnelling Microscopy and Atomic Force Microscopy are already capable of moving
single atoms (see fig 12 right) but unfortunately building a working nanodevice in this
way would take a very long time and these techniques are difficult to apply to large scale
production
Nowadays thousands of researchers are working in the nanotechnology field towards a
new generation of microelectronic devices Several possible solutions are competing for
tomorrowrsquos computer architecture and there is still no clear winner It is likely that the
final solution will be the combined use of different techniques and components (including
molecules nanowires and nanodots) as they become available together with conventional
Fig 12 Left conventional litographic process [2] Right atomic manipulation of iron atoms on copper [3]
VPalermo 9
silicon technology
Below is a brief summary of the most recent developments in nanotechnology and
nanoscience
Actual trends in nanotechnology
Perhaps the most fascinating idea for nanodevice construction is to use one single
molecule working as a complete device The first molecular diodes (ie molecules
conducting current only in one direction) were created in 1997 in 1999 a molecular fuse
and a molecular transistor were demonstrated although there was no possibility of wiring
these devices to external contacts In April 2001 James Heath and his group at UCLA
fabricated an array of overlapping crossbars and placed a small molecule of rotaxane
between each crossbar (fig 13 left) This composite molecule is made up of two
component parts the main rod-like molecular axis and a mobile ring ldquothreadedrdquo on it like a
bead on a necklace and can function as a molecular switch A working 16-bit memory
circuit was constructed using these molecules For a brief review of these works see [4] In
June 2002 a single molecule transistor was built by connecting an organic molecule to two
metal contacts the molecule contained one or two atoms of a transition metal (cobalt or
vanadium) forming the active region of the device supported by an organic backbone [5]
Fig13 Working nanodevices Left schematic representation of rotaxane molecules between crossed nanowires [4] Right SEM image of semiconductor nanowires forming a small circuit [9]
10 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another approach to nanodevice fabrication has become possible through the discovery of
carbon nanotubes which were observed for the first time in 1991 by a Japanese electron
microscopist studying the material deposited during arc-evaporation synthesis of fullerenes
[6] They consist of a graphite-like carbon seamless cylinder with a diameter of several
nanometers and lengths of up to a millimetre Carbon nanotubes are very stable can
behave as metals or semiconductors and can host other molecular or ionic species thus
modifying their electrical behaviour In 2001 Avouris and his group reported the first
circuit made with a single nanotube [7] A few months later Cees Dekker presented a
nanotube-based transistor able to amplify an input signal by a factor of ten and built
several logic circuits using these nanotube transistors [8]
One problem with carbon nanotubes is that it is very difficult to control their electronic
properties ie their metallic or semiconducting behaviour An alternative to carbon
nanotubes are semiconductor nanowires Silicon nanowires can be made using a laser to
vaporize the silicon together with a metal catalyst like iron or gold The vapour condenses
in nanosized drops of silicon and metal from which the wires slowly grow out as more
silicon is adsorbed In 2001 a group at Harvard University [9] created a transistor by
crossing two different nanowires After this the same group arranged four nanowires in a
noughts and crosses grid creating something like a 4-bit memory (fig 13 right) Even
metallic nanowires made of platinum and silver can be used in a crossed configuration to
store information [10]
There are some issues common to all these new technologies though First it is difficult to
imagine these methods applied to large-scale production Up to now the insertion of a
molecule between two electrodes is an occasional lucky event while nanotubes and wires
have to be positioned on the surface creating the appropriate contacts on them manually
The large-scale production of integrated circuits using these building blocks will not be
straightforward and does not seem imminent
Another issue is of an economic and not a scientific or technological nature Since 1965
the cost of IC manufacturing plants has sky-rocketed If the cost of semiconductor
production plants continues to rise exponentially in a few years such plants will cost up to
$20 billion This is a sizeable investment even for large companies such as IBM or Intel
For this reason it is likely that IC companies will resist changing to completely new
technologies closing down their existing plants As it is clear that silicon will remain the
VPalermo 11
fundamental raw material of the IC industry for the foreseeable future nanotechnology
developments for microelectronics will need to be silicon-compatible In these early days
of nanotechnology the most valid approach would seem to be the addition of molecular
functions to existing silicon technology ndash using the latter as a foundation on which to build
Fabrication of self-organised structures on silicon
The possibility of using the phenomenon of atomic or molecular self-organization to create
nanostructures on silicon has already been demonstrated The clean silicon surface shows
in some cases a high degree of order and complex surface reconstruction as will be
described later Several different ordered structures form spontaneously on this surface
such as series of monatomic steps or boundaries between reconstructed areas It has been
demonstrated that it is possible to use these structures to fabricate ordered nanodots and
nanolines on the surface [11] More recently well-defined nanometric patterns have been
obtained with selective etching of silicon using nitric oxide [12]
In this study the possibility of creating different types of nanostructures on the silicon
surface is explored Methods had to be developed which were
- Simple They must not need complex masks or lithographic steps to create the
structure but rather exploit self-organisation phenomena
- High resolution the silicon surface modifications should be on a scalelength of below
100 nm
- Fast billions of nanostructures have to form over the whole surface simultaneously to
be compatible with large-scale production requirements
- Cheap they must not require expensive equipment (such as e-beam lithography) but
exploit simple chemical andor physical treatments to produce nanostructures on the
silicon surface
In Chapter 2 the main characteristics of silicon are described Chapter 3 provides a
summary of the techniques used for this research Chapter 4 examines the chemical etching
of silicon in different liquid environments and the effects of this etching on the surface at a
nanoscopic level with the creation of nanoholes
12 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Chapter 5 describes the growth in ultra-high vacuum (UHV) of nanoscopic voids and
islands on the silicon surface and the effect of surface oxide on this growth
Chapter 6 discusses the modification of silicon surfaces in UHV following the adsorption
of molecules and thermal heating to produce nanoislands and nanolines on silicon
The overall conclusions of our work are summarised in Chapter 7
Finally we will give some conclusions based on the results obtained and discuss possible
applications of the methods developed
Bibliography
[1] From wwwintelcom
[2] From wwwsematechorg
[3] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
[4] Molecules Get Wired Service RF Science 294 (5551) 2442-2443 Dec 21 2001
[5] Coulomb Blockade And The Kondo Effect In Single-Atom Transistors Park J
Pasupathy AN Goldsmith JI Chang C Yaish Y Petta JR Rinkoski M Sethna JP
Abruna HD Mceuen PL Ralph DC Nature 417 (6890) 722-725 Jun 13 2002 Kondo
Resonance In A Single-Molecule Transistor Liang WJ Shores MP Bockrath M Long
JR Park H Nature 417 (6890) 725-729 Jun 13 2002 Nanotechnology - Electronics
And The Single Atom De Franceschi S Kouwenhoven L Nature 417 (6890) 701-702
Jun 13 2002
[6] Smallest Carbon Nanotube Ajayan PM Ijima S Nature 358 (6381) 23-23 Jul 2 1992
[7] Carbon Nanotube Inter- And Intramolecular Logic Gates Derycke V Martel R
Appenzeller J Nano Letters 1 (9) 453-456 Sep 2001
[8] Logic Circuits With Carbon Nanotube Transistors Bachtold A Hadley P Nakanishi T
Dekker C Science 294 (5545) 1317-1320 Nov 2001
VPalermo 13
[9] Logic Gates And Computation From Assembled Nanowire Building Blocks Huang Y
Duan XF Cui Y Lauhon LJ Kim Kh Lieber CM Science 294 (5545) 1313-1317
Nov 9 2001
[10] Formation And Disappearance Of A Nanoscale Silver Cluster Realized By Solid
Electrochemical Reaction Terabe K Nakayama T Hasegawa T Aono M Journal Of
Applied Physics 91 (12) 10110-10114 Jun 15 2002
[11] Fabrication And Integration Of Nanostructures On Si Surfaces Ogino T Hibino H
Homma Y Kobayashi Y Prabhakaran K Sumitomo K Omi H Accounts Of Chemical
Research 32 (5) 447-454 May 1999
[12] Ultrafine And Well-Defined Patterns On Silicon Through Reaction Selectivity
Prabhakaran K Hibino H Ogino T Advanced Materials 14 (19) 1418-1421 Oct 2
2002
14 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 15
2 Silicon surfaces
The name silicon (silicio in Italian) comes from the latin word silex Amorphous silicon
was first isolated by Berzelius in 1824 by reaction of potassium with silicon tetrafluoride
Thirty years later the first crystalline silicon was prepared Silicon makes up 25 of
earthrsquos crust and is the second most abundant element after oxygen Elemental silicon is
not found in nature occurring as silicon oxide (sand quartz amethyst flint etc) or
silicates (asbestos clay mica etc) Perhaps no other element and its compounds has such
a wide range of uses Silicon compounds such as sand and clay are used in the building
industry as refractory materials for high-temperature applications and for enamels and
pottery Silica is the main component of glass silicon carbide is an important abrasive and
silicones are commonly used polymers and lubricants
Here the most interesting use of silicon of course is for the production of
microelectronics devices For this application silicon of high purity (999999) and of
high crystallinity is needed Table 21 lists some of the physical characteristics of silicon
High purity polycrystalline silicon is produced by the reaction of gaseous trichlorosilane
with hydrogen in a furnace Then to prepare a single-crystal of silicon the so-called
Czochralski method is commonly used
Polycrystalline silicon is melted in a quartz furnace at 1415degC in an argon atmosphere
Then a seed of single-crystal silicon is lowered into contact with the melt and slowly
pulled out In this way the crystal grows and a crystalline cylindrical ingot several metres
long is created from the initial seed
After cooling down the ingot is sliced into thin silicon wafers The wafer surfaces are
polished using a counter-rotating lapping machine in an Al2O3 slurry until the surface is
very flat and shiny ready for the lithographic processes
16 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another way to obtain single crystal silicon is the Floating Zone (FZ) method in which a
silicon cylinder is slowly passed through a heating ring The area inside the ring melts and
solidifies smoothly crystallising as it comes out of the ring yielding a single silicon crystal
Microelectronic devices are built on the silicon surface which is the surface of interest
here Unfortunately silicon surfaces are normally quite dirty and uneven at the atomic
scale Atmospheric oxygen and humidity react with silicon surfaces creating a thin layer of
oxide (called ldquonative oxiderdquo) which is usually irregular and full of defects Different kinds
of contaminants also adsorb onto the surface These are usually small organic molecules
and microscopic dust particles A clean surface on exposure to the atmosphere is
completely covered with gas molecules in less than 10-9 seconds If the pressure is
reduced letrsquos say to 10-6 mbar this time increases to 1 second This is the reason why to
study a clean surface we have to work in UHV at pressures below 10-10 mbar
The atoms in the silicon crystal have a diamond-like structure each atom having 4 bonds
in a tetrahedral sp3 arrangement with bond angles of 10947 degrees At the crystal
surface some atoms will have non-bonding orbitals ldquodanglingrdquo in the vacuum ie sp3
orbitals with a lone electron which are highly reactive These orbitals are known as
dangling bonds To minimize surface energy the surface will reorganize by decreasing
the number of dangling bonds
Table 21 Physical data of silicon [1]
Atomic Weight 2809 Lattice constant (A) 543095 Crystal structure Face-centered cubic
(diamond) Melting point 1415 degC
Density (gcm3) 2328 Boiling point 2355degC Atomscm3 50E22 Minority carrier
lifetime (s) 25E-3
Dielectric Constant 119 Specific heat (Jg degC)
07
Breakdown field (Vcm)
~3E5 Thermal conductivity (Wcm degC)
15
Electron affinity x(V)
405 Vapour pressure (Pa) 1 at 1650degC 1E-6 at 900deg C
Energy gap (eV) at 300K
112 Reactivity Inert to acids Attacked by halogens and alkaline
solutions Intrinsic carrier
conc (cm-3) 145E10 Oxidation states +4 -4
Intrinsic Debye Length (microm)
24 Energy of a Si-Si bond (eV)
232
Intrinsic resistivity (Ω-cm)
23E5
VPalermo 17
Dangling bond densities and positions and thus the type of surface reconstruction will
depend upon crystal orientation as well as the temperature and kinetics of the system
Fig 21 shows a drawing of the main faces of a silicon crystal The angle α between any
(11n) face and the (100) face can be calculated from 2cos 2 += nnα The angle
between any (11n) face and the (111) face can be calculated from
)2(3)2(cos 2 ++= nnα
The chemistry and physics of the faces are very different a brief description will be given
for the most important orientations
Si (100)
On the (100) surface each atom has two Si-Si bonds connecting it to the bulk and two
dangling bonds pointing outward Surface energy is reduced by the dimerisation of the
surface atoms through overlap interaction of one dangling bond per atom forming rows of
dimers aligned along the (110) direction This is the well-known ldquo2x1rdquo reconstruction of
this silicon surface The symmetric dimers would make the silicon surface metallic but to
reduce surface stress the dimers tend to buckle and the surface is thus semiconductive It
took several years to understand that the dimers are buckled because at room temperature
Fig 21 Schematic view of the principal orientations of a silicon surface Surface atoms are white bulk atoms are black dangling bonds are gray [2]
18 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
they shift easily from one buckling direction to the other thus appearing symmetric under
STM observation Fig 22 shows an STM image of the 2x1 reconstructed surface
Even almost perfect (100) surfaces have a certain number of monoatomic steps and the
dimer rows on atomic layers are aligned at 90deg to those on adjacent layers Dimer rows are
thus perpendicular or parallel to the step When the dimers on the upper side of the step
are parallel to the step the step is called SA if they are perpendicular the step is called SB
Because of this symmetry SA step edges will be smoother compared to the more broken
fragmented edges of the SB steps
A common defect on the Si(100) surface is the presence of nickel contamination which
appear as missing dimers in STM images This type of contamination is so critical that
even if the silicon sample is only briefly brought into contact with stainless steel tools
(tweezers for example) the 2x1 reconstruction of the surface can be blocked
Silicon atoms can diffuse easily over the silicon surface as monomers and dimers
especially at elevated temperatures The anisotropy due to the 2x1 reconstruction causes a
difference in the diffusion energies of adsorbates over the surface Diffusion of these
silicon species along dimer rows for example will be much easier A list of diffusion
energies for monomers and dimers is given below [2]
Diffusion on Si(100) 2x1 Ed (eV) Monomers along dimer rows 06 Monomers across dimer rows 085 Monomer formation energy 18 Dimer along dimer rows 11 Dimer across dimer rows 15 Dimer formation energy 26 Dimer binding energy 076 Vacancies along dimer rows 17 Vacancies across dimer rows 19
SA
SB
Fig 22 STM image of a 2x1reconstructed silicon surfaceshowing the dimer rows and steps40x35 nm Nickel-induced defectsare visible as dark spots SA and SBsteps are indicated
VPalermo 19
So the diffusion energy for both monomers and dimers is nearly 40 greater if they have
to cross a dimer row This difference reduces to sim10 for vacancy diffusion
Si(113)
The (113) surface can be imaged as a sequence of alternating (100) and (111)-like
structures with two and one dangling bonds on alternate atoms respectively Interest in the
(113) surface is scientific as it has been used to study the energetics of the (100) and (111)
surfaces as well as for surface adsorption experiments
Si(100) surfaces can easily develop (113) facets
Si(111)
This surface besides being the first one imaged with STM with atomic resolution is one of
the most studied because it is the best cleavage face of silicon and because it shows one
of the most complex and elegant reconstructions in surface science
All Si-Si bonds in the silicon crystal are perpendicular to a (111) plane so this face will
have the lowest number of dangling bonds created per unit area In fact each Si atom on a
(111) surface shows a single dangling bond oriented perpendicular to the surface and
bonded to three back atoms These three bonds for each surface atom account for the great
chemical and physical stability of the Si(111) surface Surface energy is 009 eV Aring-2
compared to 015 eV Aring-2 for Si(100)
For energy minimization this surface reconstructs forming a huge 7x7 lattice cell
containing 102 atoms described by the Dimer-Adatom-Stacking fault model (DAS) For a
detailed description of cell structure see fig 23
The cell described by this model is very complex being composed of three kinds of atoms
adatoms rest atoms and corner hole atoms Furthermore a subsurface stacking fault is
present in one half of the cell making the two halves of the unit cell look different under
STM (Fig 24) It took 26 years of research to completely understand the exact structure of
the 7x7 reconstruction
Cleaving a silicon crystal along a (111) plane produces a metastable 2x1 reconstruction
the 7x7 reconstruction is easily obtained by flashing at high temperature in UHV At T gt
20 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
830degC a disordered 1x1 phase covers the surface Cooling down to 800degC leads to the
formation of the 7x7 phase If the cooling process is too rapid small 7x7 domains nucleate
and a disordered 1x1 phase is preserved between domain boundaries
Si(110)
Even though as mentioned before the (111) plane is the favoured cleaving plane of
silicon thin (100) commercial wafers will not break along this plane because the angle
between (100) and (111) is too far from 90deg (see table 22) Instead they will break along
the (110) plane because it is perpendicular to the (100) surface Each surface atom on
Si(110) has a Si-Si bond pointing downward one dangling bond pointing outward and two
Si-Si bonds parallel to the surface in a zig-zag pattern (see fig 21) Cleaved (110)
surfaces are disordered but upon annealing at high temperatures an ordered complex 16x2
Fig23 Scheme of the 7x7 DAS model [2] In each unit cell there are 9 dimers 12 adatoms and a stacking layer fault The force driving this complex reconstruction is the minimization of dangling bonddensity The DAS model shows the lowest number of dangling bonds (19) of all possiblereconstructions 12 dangling bonds are at the adatoms 6 at the rest atoms and 1 at the corner hole atom This surface is metallic
VPalermo 21
reconstruction takes place The surface appears as a series of long ridges and valleys
parallel to each other Eventually tilted facets of orientation (17 15 1) can form on this
surface The adsorption of Ge atoms on this surface leads to the formation of self-
assembled nanowires [3]
Table 22 Angles in degrees between different silicon faces [2]
Orientation
113
110
111
100
100
2524
9000
5474
0
111
2950
3526
0
110
6476
0
113
0
Fig 24 STM image of a Si(111)surface with 7x7 reconstruction A unitcell with its adatoms is highlightedImage size 13x13 nm
22 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] Weast RC Handbook Of Chemistry And Physics (Chemical Rubber Co Cleveland
1972)
[2] Dabrowski J Mussig H Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] The Structure Of Clean And SiGe-Covered Si(110) Surfaces Butz R Luth H Surface
Science 365 (3) 807-816 Oct 1 1996
VPalermo 23
3 STM and other surface analysis
techniques
Scanning Tunneling Microscopy
Since the invention of the optical microscope at the end of the 16th century the possibility
of examining surfaces at higher and higher magnification has fascinated mankind
Development of the technique continued and towards the end of the 19th century optical
microscopes were as good as todays standard instruments The physical limits of the
wavelength of visible light (350-800nm) had been reached
In the 1920s de Broglie showed that electrons can behave like waves and the use of these
particles for imaging with much higher resolution soon followed Atomic resolution using
this technique is only possible in the transmission mode with extremely carefully prepared
samples
In 1982 using the peculiar properties of piezoelectric materials Binnig and Rohrer brought
a metallic tip very very close to a silicon surface and scanned it across an extremly small
area (fig31) The tunneling of electrons from the tip into the sample or vice versa allowed
them to obtain a local density of electronic states (LDOS) map of that surface Although
theory (which treated the extreme point of the tip as a sphere) then excluded the possibility
of atomic resolution following a tip crash into the surface Binnig and Rohrer began to
observe the LDOS with atomic resolution For this discovery and their following work
they were awarded the Nobel Prize in Physics in 1986
24 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The basic principle of STM is very simple A metallic tip is scanned over a surface without
making ohmic contact and a tunneling current passes between the tip and the surface An
electronic circuit keeps this current constant by raising and lowering the tip during the
scan In this way recording the tip height at each point a three-dimensional image of local
density of electronic states (LDOS) of the surface can be obtained To explain the
extremely high resolution attainable by this simple technique quantum theory is needed
According to classical physics the current will flow between sample and tip only if they
are in physical contact If there is a vacuum gap between the two the electrons will simply
remain confined for example within the surface without the possibility of passing into the
tip
In quantum physics however the electrons have a certain probability of passing
(tunneling) across the gap appearing on the other side of the gap in this way reaching the
tip It can be shown that the probability of an electron tunneling through a gap of thickness
z is
kzep 22)0( minusprop ψ h
φmk
2= (1)
where ψ(0) is the electron wavefunction at the surface-gap border m is the electron
mass=91x10-28 g and φ is the work function of the metal (ie the energy required to
remove an electron from that material For silicon it is 48 eV) The tunneling current thus
decays exponentially with z and is extremely sensitive to topographical imperfections
present on the scanned surface A rough formula giving the current as a function of z is [1]
zFS eEVI φρ 0251)( minusprop
Fig 31 Binnig and Rohrer with the first STM Image from IBM [2]
VPalermo 25
where ρs(EF) is the local density of states at the Fermi level on the given surface For
example the formula predicts that for silicon an increase in tip-surface distance of 1 Aring
will give a 95 decrease in tunneling current
This huge dependence of tunneling current upon the distance allows detection even of the
sub-nanometre changes in height given by the single atoms of which the surface is
composed and thus to resolve them in the LDOS images Of course this description of the
tunneling process is oversimplified and for a more accurate one the electronic states of
the tip of the sample and their interaction have to be taken into account Fig 32 shows a
schematic representation of the interaction between tip and sample orbitals
The exponential decay of current with distance also yields high lateral resolution If the tip
is approximated as a sphere of radius R and the current passing at the minimum tip-sample
distance is I0 then the current passing at a lateral distance x from this point will be
Rxk
eII 22
0
2minus
=
Assuming a tip radius of 100 nm the current is concentrated in an area sim15 nm wide at tip
apex
Very sharp tips with even smaller curvature radii can be produced in several ways Simple
Fig32 Schematic view of tip-sampleorbitals interaction a) no interaction b)equilibrium c) sample positive d) tippositive [1]
26 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
tungsten wires mechanically cut are capable of obtaining atomic resolution on graphite in
air but for more disordered and rougher samples sharper and more reproducible tips are
needed
STM tips are mostly made by electrochemically etching a W or Pt-Ir wire The tips we
used were prepared using methods based on the work of Fotino [3]
A tungsten wire 038 nm diameter is immersed in a KOH solution(06M) with a thicker
tungsten wire used as a counterelectrode The cathodic and anodic reactions involved in the
etching are
Cathode 6H2O + 6 e- rarr 3H2(g) + 6 OH-
Anode W(s) + 8 OH- rarr WO42- + 4H2O + 6 e-
A potential of 3V ac is applied to the tungsten and the wire is immersed in the solution
until a constant current of sim100 mA is established The etching rapidly removes metal
shaping the wire end as a sharp tip When the potential reaches 11 V the coarse tip etching
is finished The wire is then removed from the solution carefully inserted into an
insulating plastic tube leaving only the tip exposed and re-immersed in the solution with
the tip pointing upwards A more gentle etching is thus made to reduce tip radius Usually
5 minutes etching at 07 V ac is used In this configuration very small hydrogen bubbles
formed on the tip sides sliding upwards with a ldquohoningrdquo effect on the tip
This procedure yields extremely sharp and reproducible tips at the microscopic level
After the etching the tip is thoroughly rinsed in ultrapure water then dipped into
concentrated HF to remove surface oxides and hydroxides [4] The tip is dried with
nitrogen inserted into the UHV system and degassed overnight at sim150degC
The possibility of measuring sub-nanometric distances would be useless without being able
to control tip movement over such a minute scale To scan the tip over the surface a
piezoelectric scanner is used Piezos are usually made of an alloy of PbZrO3 and PbTiO3 a
material which contracts or expands when a voltage is applied to it The Omicron
instrument used in our laboratory has three such piezo scanners for xy and z tip motion
allowing one to scan the tip over the surface with sub-Aringngstrom precision (fig 33)
To isolate the instrument from ambient vibrations the whole STM stage is suspended upon
four springs which eliminate all frequencies above 1 Hz and surrounded by a crown of
VPalermo 27
copper wings and fixed magnets Parasitic currents generated by the magnets into the
copper wings contrast every movement of the stage and efficiently block stage vibrations
The STM can be used not only to explore surface topography but to measure the IV
characteristics of single atoms or molecules on the surface (Scanning Tunneling
Spectroscopy or STS) Furthermore it can be used to modify the surface with voltage
bursts digging into it or delicately moving single molecules or atoms over a surface [5] It
can be used in vacuum in air and with proper lateral isolation of the tip even in liquids
A major drawback of STM is that it works only on conducting and semiconducting
surfaces and thus cannot be used on many surfaces of biological and chemical interest
Another instrument more suitable for these and other applications is the Atomic Force
Microscope (AFM)
Atomic Force Microscopy
AFM was invented in 1986 by Binnig Quate and Gerber after calculating the possibility of
building a cantilever with a force constant of the same order of magnitude as that of a
chemical bond
In AFM a tip mounted on a microscopic cantilever (usually made of Si3N4 fabricated with
optical lithography) is brought close to a surface When the tip touches the surface the
cantilever is very slightly deflected upwards The movement is measured by observing the
Y-PIEZO
SILICON SAMPLE
Fig 33 A picture of theSTM used for theexperiments The tripodpiezo scanner is shown
Z-PIEZO
X-PIEZO
TIP
28 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
deviation of a laser beam hitting the upper face of the cantilever Fig 34 provides a
schematic view of the principle of AFM
The typical force constant of the cantilever varies from 00006 to 2 Nm the typical
resonance frequency is 3 to 120 kHz The AFM tip can apply a force on the sample of up
to 10-9 N The AFM can be used on conductive or insulating surfaces in vacuum air or
liquids Furthermore the tip can be modified to sense electrostatic potentials (electric force
microscopy) or magnetic fields (magnetic force microscopy) it can even be functionalized
with complex molecules such as proteins to interact with biological surfaces
A drawback of the AFM is that the force it exerts can damage the surface under
observation especially if the sample is soft (as in the case of cell membranes for
example) This problem can be overcome using the instrument in tapping mode (where the
tip does not move laterally during its brief contact with the surface) or in non-contact mode
in which the tip oscillates above the surface during the scan and the changes in its
frequency due to interaction with surface are monitored The shifts in the oscillating
frequency of the cantilever due to tip-sample interaction are then used for imaging the
surface In this mode interaction of the tip with the surface is minimal and soft samples
can be imaged
STM and AFM are the main techniques used for this work A brief description of other
techniques used occasionally is given below
Fig 34 Scheme of an atomic force microscope
VPalermo 29
Low Energy Electron Diffraction (LEED)
Electrons with energies in the 20-500 eV range are diffracted by a crystalline surface the
diffraction peaks are visualized on a fluorescent screen This technique probes the long
range order of the surface up to a depth of several nanometres
X-Ray Photoelectron Spectroscopy (XPS)
XPS allows both qualitative and quantitative chemical analysis of the elements present on
or near the sample surface
An X-ray source is used to photoionize the atoms on a surface and produce photoelectrons
By measuring the kinetic energy of the photoelectrons the binding energy of the electronic
levels can be calculated This energy will depend on the chemical environment of the
surface atoms
Although the soft X-rays used penetrate to a depth of ~2000Aring the sampling depth of the
technique is determined by the mean free path of the photoelectrons which allows their
escape from only the first 10-100Aring
Secondary Ion Mass Spectroscopy (SIMS)
High and low energy ions (primary ions) are used to bombard a sample and remove surface
atoms and ions The ionic fragments removed (secondary ions) are then analysed by a mass
spectrometer The surface can be consumed during the measurement and profiles obtained
giving concentrations of the materials composing the sample at different depths (depth
profiles)
A popular variant of SIMS is TOF-SIMS In this technique the secondary ion masses are
measured by a time-of-flight (TOF) measurement The secondary ions generated by the
bombarding primary ions are accelerated to a constant kinetic energy and then move
through a field-free space before they reach the detector where their intensity is measured
as a function of flight time Since ions with different masses have different velocities at a
30 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
given kinetic energy the measured flight times of the ions can easily be converted to their
masses The static nature of this latter technique allows mass spectroscopy surface analysis
with minimal damage to the surface
Bibliography
[1] Chen CJ Introduction To Scanning Tunneling Microscopy (Oxford University Press
Oxford 1993)
[2] From wwwibmcom
[3] Tip Sharpening By Normal And Reverse Electrochemical Etching Fotino M Review
Of Scientific Instruments 64 (1) 159-167 Jan 1993
[4] A Convenient Method For Removing Surface Oxides From Tungsten STM Tips
Hockett LA Creager SE Review Of Scientific Instruments 64 (1) 263-264 Jan 1993
[5] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
VPalermo 31
4 Surface modification of silicon in liquid
Nano-hole creation
Liquid treatments of silicon wafers are very common in the integrated circuit (IC)
manufacturing industry They are used to clean and improve surface uniformity to create
and etch protective oxide layers and to remove photo-resist layers
Crystalline silicon with its native oxide layer is very stable and is resistant to many acids
It is easily attacked by hydrofluoric acid (HF) and alkaline solutions
The thin (~2 nm) passivating layer of native oxide (SiO2) is formed on exposure to the
atmosphere This surface layer contains many defects and contaminants so it is usually
chemically stripped and substituted with a better chemically-formed protective oxide
The most common silicon cleaning procedure is the RCA method named after the Radio
Corporation of America [1] It consists of two steps in the first one the surface is treated
with a hot alkaline solution (H2OH2O2NH4OH 411) to remove particles from the
surface following this a hot acidic solution (H2OH2O2HCl 411) is used to remove
metal contamination Other well-known cleaning methods are IMEC (a sequence of
cleaning steps in H2OO3 and dilute HF) or the Pirana etch (a hot 41 mixture of H2SO4
H2O2)
32 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The standard RCA clean removes surface contaminants etches the native oxide and
oxidizes the silicon surface leaving a uniform layer of silicon oxide which better protects
the surface from further contamination
Etching with fluorine-based solutions
Hydrofluoric acid is one of the most common reagents used in the treatment of silicon
wafers both in the research field and in industrial processes A rapid dip in dilute HF is the
simplest way to remove the native oxide from Si(100) and leaves the surface passivated by
a layer of Si-H bonds Because of the low polarization of Si-H bonds the Si-H layer is
stable even for several days protecting the surface from contamination It has often been
assumed that this short etch does not significantly change the surface morphology of the
silicon substrate[2] even though a prolonged dip in dilute HF leads to surface roughening
[3]
Although dilute HF roughens the Si(100) surface at the atomic scale [45] immersion in
concentrated HF (49) etches the surface oxide without attacking the Si surface
uncovering in this way the buried SiSiO2 interface The final counter-intuitive result is
that dilute HF etches the silicon while concentrated HF leaves the crystalline silicon
untouched [3]
Etching Si with fluorine-containing solutions at different concentrations and pH can
produce different morphologies from rough surfaces to flat nearly ideal Si-H terminated
surfaces
Hessel et al and Higashi et al demonstrated in 1991 that very flat Si(111) surfaces can be
obtained using 40 NH4F while etching with HF always yields rough surfaces The
surface becomes smoother because the etchant rapidly attacks Si atoms at step borders
thus removing surface kinks and irregularities in a step-flow mechanism [6 7] Later on
even smoother and more perfect surfaces were obtained by removing oxygen from the
solution after it was discovered that oxygen dissolved in 40 NH4F initiates the formation
of triangular etch pits It was not possible to obtain flat surfaces by etching Si(100) with
ammonium fluoride solutions which leads to the formation of small 2x1 dimer-row
reconstructed (100) terraces together with (111) facets [8]
This difference is caused by the different hydride terminations prevailing on the (100) and
(111) faces While the ideal Si(111)-H surface is monohydride terminated the more
VPalermo 33
reactive dihydrides predominate on the Si(100)-H surface making it more vulnerable to
etching The etching reaction is thus strongly anisotropic etching (100) facets faster than
(111) thus producing (111) microfaceting on Si(100) crystals
A more uniform Si(100) surface can however be prepared by etching at low pH with an
HFHCl mixture [9] or by using very dilute HF solutions and ultrapure water with low
dissolved oxygen and carbon contents [10]
Electrochemical etching can also be used applying anodic or cathodic bias to the silicon
to obtain different morphologies [11] by varying the potential isotropic or anisotropic
etching is observed The aforementioned results show that despite the simplicity of the
reactants fluoride etching of silicon is quite a complex reaction
Fig 41 Chemical etching of silicon
HO+H
H HO
H
H
H
FSi
Si
SiSi
+H2O
-OH-
+F-
H2O H H
Si H
F OH
H Si
Si Si
-H2
+H2O
H
H
H
FSi
Si
SiSi
H
H
H
OHSi
Si
SiSi
H
H
H
H Si
Si
Si Si
Etching mechanism of silicon
HF rapidly dissolves the SiO2 passivating layer on silicon leaving the surface almost
completely hydrogenated [12] After this two different types of reactions etch the silicon
simultaneously one chemical and the other electrochemical [13] The overall etching
mechanism can be schematized in two stages (see Figure 41)
i) Si-H bonds are substituted by Si-F or Si-OH bonds creating a partial charge on the
surface silicon atom and polarizing its Si-Si backbonds
ii) these polarised backbonds are then more easily attacked by HF or H2O After
rupture of the Si-Si bond the atom is removed leaving behind new Si-H
terminations and the reaction can start again
34 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
These reactions take place although at different rates on both Si(100) and Si(111)
Stage i) is usually the rate-determining step of the reaction and the stability of the Si-H
bonds depends upon the pH the concentration of nucleophilic species in solution and an
eventual potential applied to the crystal
For pH gt5 as in the case of concentrated NH4F solutions reaction begins with attack by
water to give Si-OH (step ArarrC) The -OH group is rapidly substituted by fluorine with
polarization of the underlying Si-Si bonds These bonds are then easily attacked by water
the silicon atom being released into solution as HSiF(OH)2 The Si-OH rarr Si-F substitution
is not fundamental for the reaction and etching can proceed even for Si-OH terminated
atoms but XPS measurements showed the presence of a certain number of Si-F bonds
remaining Furthermore fluorine seems to have a catalytic effect on Si-H substitution as
indicated by the dependence of the etch rate upon the F- concentration at least for pH
values between 4 and 8
Si-F bonds can be easily removed by a water rinse In the case of strongly alkaline
solutions (pH=14) OH- groups act directly as nucleophiles and no fluorine is needed to
catalyze Si-H bond rupture
At pH lt4 almost no free F- ions are present in solution and the etch rate is very small at
pH lt2 all etching reactions are very slow and this explains the stability of Si crystals in
concentrated (50 ww) HF solutions
Si Si
Si Si
H
H
H
H Si
Si
Si Si
H
H
H Si
Si
SiSi
H
H
H
OH-H+ -e
(F-)
-H+
-e
+H2O hellipas in fig1
Fig 42 Electrochemical etching of silicon by H2O
The electrochemical etching of silicon involves electron transfer from the surface atoms to
the valence or conduction band of the crystal (according to the type of doping of the
silicon) Several studies have been carried out at different electric potentials In the case of
p-silicon or n-silicon at anodic potentials where positive charges (h+) are available in the
crystal a silicon-centred radical can be created by capture of a hole from the bulk (fig
42) The F- ion does not participate in the reaction but is thought to contribute through
VPalermo 35
electrostatic interactions by lowering the energy of the interaction step After the
formation of the Si-OH group the reaction proceeds as shown in the scheme of fig 41
Matsumura et al [4] proposed that not only water but HF2- molecules also play a major role
in electrochemical etching of silicon leaving on the surface Si-F terminated bonds which
can be immediately attacked in an autocatalytic process (fig 43)
In the electrochemical reactions described above an external potential is applied to the
silicon crystal The chemical and electrochemical reactions in any case take place
simultaneously most of the time with the chemical path predominating at high pH Even
when no external potential is applied to the silicon partial electrochemical reactions can
take place at different ldquocathodicrdquo and ldquoanodicrdquo sites on the surface with an internal charge
exchange which ensures neutrality [11] This macroscopic silicon etching and hydrogen
bubble formation can lead to visually observable patterns on the surface when Si(100) is
immersed in ammonium fluoride even without applying a potential
Si Si
Si Si
H
H
H
F Si
Si
Si SiH
F F-H+ -2e
+HF2-
H
HF
SiSi
SiSi + F
F F
FH
H
H
F -H+ -2e
+HF2-
Fig 43 Autocatalytic electrochemical etching of silicon by HF2-
Inhomogeneities on silicon surfaces caused by electrochemical reactions and charge
transfer have been studied extensively because they are of fundamental importance in the
formation of porous silicon
Pore formation on silicon
When Si(100) or Si(111) are etched under anodic bias in fluorine-based solutions
microscopic pores form on their surface Several different morphologies of pores have
been observed with pore diameters ranging from 10 nm to several microns with depths of
several microns [14] Pore shape is very variable too ranging from ordered straight pores
to chaotic networks of branched pores (fig 44) Porous silicon has been known since the
fifties but it was only in 1990 that interest in this material increased following the
discovery that porous silicon layers were able to emit bright red light This led to a large
amount of research and now different classes of micropores can be reproducibly created
36 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Fig 44 Different types of Silicon micropores From ref [14]
mostly for optical and micromachining applications However there is still no unified
theory able to explain the nucleation and growth mechanism of all the different kinds of
pores
We will give a short description of some of these theories for more detail see Parkhutik et
al [15]
One model explains pore nucleation on the basis of physical processes such as hole
positive charge migration ion transport to the surface and small perturbations on the
silicon surface modelled as Fourier components The system is shown to be unstable and
some spatial frequencies that lead to pore nucleation evolve from the etching process
A second model focusses on stationary pore growth without explaining the nucleation
stage According to this model silicon dissolves preferentially at pore edges because h+
charges are attracted by the stronger electric field present at these edges
A third class of models explains pore growth as a Diffusion Limited Aggregation (DLA)
process where the random walk of h+ charge carriers through the depleted layer present at
the silicon-liquid interface controls pore shape
Finally the model by Carstensen Cristophersen and Foll [16] proposes that areas of the
surface of some characteristic size LCO are etched by synchronized ldquocurrent burstsrdquo in the
flow of h+ charges These bursts dissolve silicon through cyclic stages of surface oxidation
oxide removal and hydrogen passivation Areas where a burst has recently taken place are
less passivated and thus more likely to be etched again in this way the pore bottom
continues to dissolve while the pore walls are passivated and are thus less favourable areas
towards current bursts
VPalermo 37
EXPERIMENTAL RESULTS
In the following sections we will show some experimental results obtained from STM and
AFM measurements of fluorine-treated Si(100) surfaces In the first part the results of
mild etching using concentrated and dilute HF solutions at low pH are presented In the
second part the results of etching at high pH using ammonium fluoride are presented and
the mechanism of pore formation discussed
Etching of Si(100) in dilute and concentrated HF
Samples were cut from different areas of an 8-inch diameter p-doped silicon(100) wafer
(10 Ω-cm) supplied by MEMC Electronic Materials Each series of STM measurements
was carried out over at least six different areas on at least two identical samples Low
Electron Energy Diffraction (LEED) was used to check the surface cleanliness of the
samples before STM measurements
Table 41 summarizes the different treatments of each sample After etching with
electronic grade HF each sample underwent a final rinse in Ultra-Pure Water (UPW
resistivity gt18 MΩ-cm) Both the HF and the UPW were allowed to flow continuously
over the sample surface Some samples were not etched with HF at all but just washed with
UPW to observe the morphology of the native oxide layer (~2 nm thick) covering the
surface All of these processes were carried out under nitrogen to limit reoxidation and the
samples were then introduced from the nitrogen atmosphere directly into the vacuum
chamber and degassed overnight at sim150degC before LEED and STM measurements
STM images were obtained from each sample using the same measurement parameters
(sample bias 4 V feedback current 1 nA scan speed 800 nm s-1) These parameters
38 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
provided a satisfactory level of reproducibility for all the samples Measurements were
made over an area of 500x500 nm (image size 500x500 pixel) Slope correction was
carried out by subtracting row-wise and column-wise fitted slopes from the entire image
which gave better results than the simple subtraction of a fitted plane especially for the
rougher samples Following slope correction the rms roughness
sum minus=xy
hyxhN
22 ))((1σ
and the 2-D Fourier transform
)(22
2
)(4
)( vyuxi
xyeyxhvuF +sum∆
= π
π
were calculated for each image where N2 is the number of pixels composing the image
h(xy) is the surface height at each point ∆ is the distance between points h is the mean
height and u v are the spatial frequencies The radial power spectrum PS(f) of the STM image data is obtained from the angular
average of the squared Fourier transform with f 2 = u2 + v2
Fig 45 shows the STM images obtained from the various samples Sample A still covered
with its native oxide layer shows an irregular surface with RMS roughness of ~05 nm
(see Table 41) Observing the sample with LEED gave no diffraction pattern even at
relatively high incident electron energies because of the surface oxide coverage After 1
min etching in dilute HF (sample B) the morphology is similar to the original one though
Table 41 Sample treatments average RMS roughness and integrated area of power spectra
Sample Treatment RMS roughness (nm)
PS area (f lt 01 nm-1)
PS area (f gt 01 nm-1)
A Rapid dip in water 053 plusmn 013 366 032
B 1 min in HF 5 + 10 min in water 051 plusmn 008 272 034
C 30 min in HF 5 + 10 min in water 062 plusmn 008 1064 038
D 5 sec in HF 49 + 10 min in water 042 plusmn 004 183 022
VPalermo 39
Fig 45 STM images of each group of samples showing the topography of the silicon surface A) noetching original oxide surface B) after 1 min etching in dilute HF C) after 30 min etching in diluteHF D) after dipping in concentrated HF Grey scale indicates height of the surface from lower (black) to higher (white) The images are 250x250 nm ie representative portions of the images usedfor the roughness measurement and PSD analysis
some of the larger features have disappeared and the image quality is better maybe due to
improved tunnelling due to the cleaner surface The RMS roughness is comparable to that
of the original surface Clear diffraction patterns are visible using LEED though at quite
high energies (200 eV) After prolonged etching (sample C) the RMS roughness increases
to 062 nm and a long-range corrugation is visible on the surface even if the LEED pattern
is good
The samples dipped in concentrated HF (D) reveal the bare SiSiO2 interface which has a
disordered aspect and protrusions over a wide range of dimensions The quality of the
STM images of sample D is very good probably due to the cleanliness of the surface
40 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 E -0 2
1 E -0 1
1 E + 0 0
1 E + 0 1
1 E + 0 2
1 E + 0 3
0 0 0 0 0 1 0 1 0 1 0 0f (1 nm )
nm^4
A a s re c e iv e d
B e tc h e d 1 min
C e tc h e d 3 0 min
D e tc h e d HF 4 9
Fig 46 Log-Log plot of the averaged power spectra of the STM images for all the samples
which gives a more stable tunnelling junction The LEED pattern is excellent showing
clear diffraction peaks at energies as low as 37 eV comparable to that obtained after high
temperature cleaning in UHV
Fig 46 shows the power spectra of the samples The high frequency and low frequency
areas of the power specturm are considered separately Table 41 shows for each sample
together with the roughness the integrated area of the power spectrum for the high and low
frequency part
We first examine the differences between the samples in the low frequency part of the
spectrum (lt01 nm-1) In this spatial range the short dip in concentrated or dilute HF
(samples BD) seems to lower the roughness of the sample removing some of the larger
features of the native oxide surface visible in Fig 45a and thus decreasing the
corresponding part of the power spectrum Sample C however etched with HF for 30 min
shows an increase in roughness peaking at 0012 nm-1 (~85 nm corresponding to the
typical dimensions of the corrugations visible in Fig 45c) On this scale sample C has a
PS density of 435 nm4 compared to 101 nm4 for sample A
Samples A B and C have the same PS in the high frequency range showing that the fine
structure of the surface is not changed by the HF etching Sample D on the other hand
shows a decrease of roughness for frequencies gt 015 nm-1 This effect could be due to the
VPalermo 41
improved surface cleanliness after etching with concentrated HF which would give a more
stable STM junction thus reducing the high frequency noise in the image
Fig 45 and the analysis of the power spectrum of each sample shows that a rapid dip in
HF removes the native oxide but does not lead to major changes in the morphology of the
Si surface its only effect being the removal of some of the larger features present on the
original surface Prolonged etching on the other hand increases the RMS roughness of the
surface
Etching of Si(100) in concentrated ammonium fluoride and nano-hole creation
Two different types of commercial p-doped Si(100) wafers (2Ω-cm and 10Ω-cm) from
MEMC were used Several different samples of 10x5 mm were immersed for 10 minutes
in 40 electronic grade NH4F solution under agitation Previous works used low
temperatures or anodic potentials applied to the silicon to avoid gaseous hydrogen
production and to obtain a uniform surface but in our experiment we worked at room
temperature to check the influence of hydrogen bubbles on surface morphology During
the etching the stirring was sufficient to provide a uniform concentration of reagents over
the whole sample surface but not to mechanically remove the hydrogen bubbles from the
silicon surface
After the etching each sample was rinsed in ultra-pure water to remove any etching
residues and observed with STM AFM and optical microscopy The AFM measurements
were made in air while for STM measurements the samples were rapidly dried with
nitrogen and inserted into the vacuum system to avoid surface reoxidation After insertion
into the vacuum surface cleanliness was checked with LEED and the surface morphology
observed by STM Parameters for STM measurements were sample bias +4 V 1 nA
current 16 Hz scan rate The images obtained were stable and reproducible over several
days Scan parameters for AFM were 20 nN force and 1 Hz scan rate
Some of the samples were cleaned with an RCA standard clean [1] before NH4F etching to
check the influence of possible surface contaminants on the final results Eight different
samples were prepared and more than sixty STM images of the samples were taken at
different points of the various samples
42 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
(110)
(110)
Fig47 abc) STMtopographic images of differentetching morphologies Eachimage is 500x500 nm Z-ranges are 10 10 and 18 nmrespectively d) STM image of a bridgecreated by etching of the lowerlayers of silicon (black arrow)Image is 250x250x6 nm
After ~2 min of immersion in the solution hydrogen bubbles become visible on the sample
surface The production is slow and the bubbles are quite stable on the sample without
detaching Thus some areas of the surface are masked from the liquid etching action
STM observations (fig 47) show that at the nanometer scale the surface is unevenly
covered with holes of radii ranging from 10 to 200 nm with depths of 2-4 nm These holes
have a wide range of different shapes and distributions In most cases the surface was
covered with a uniform distribution of round-shaped holes (fig47a) indicating isotropic
etching The dimension and the density of the holes changed greatly from sample to
sample and even over the surface of a single sample In some cases the etching was
anisotropic yielding nearly square holes and layered structures as shown in fig 47b
Square holes have been previously observed in cases where the etching speed in the (110)
direction is significantly smaller than in the (100) direction [17]
Over these areas (fig 47b and especially 47c) it is clearly visible how once the surface
had been attacked the reaction continued to preferentially remove atoms at step
irregularities (kink atoms) straightening step edges Eventually the exposed underlying
silicon was also attacked and further holes created inside the previously etched larger
ones It was not possible to detect monatomic steps on this kind of surface The smallest
step height observed was ~15 nm corresponding to several atomic layers In the image
shown in Figure 47a the etching was not strong and created only anisotropic holes on the
surface In fig 47b and c the stronger etch proceeded laterally for several tens of
VPalermo 43
nanometres leaving straight steps several tens of nanometers long and roughly rectangular
holes as expected given the structure of the (100) crystal face In some cases a
significative underetch is observed and the formation of suspended bridges and tunnels
can be deduced from the STM images (fig 47d)
The formation of branched pores and suspended structures has been attributed during pore
formation to diffusion limited aggregation effects where the h+ charge carriers necessary
for silicon etching have a higher probability of reacting at pore bottoms than reaching the
upper part of the silicon surface In the case of very deep pores quantum wire effects have
been invoked to explain the pore growth mechanism [15] In our case though the pores
formed were very shallow the underetch depth being only a few nanometres on pores of
sim100 nm width Thus more than diffusion effects the main contribution to the
underetching process must come from anisotropic etching and some kind of autocatalytic
reaction path analogous to the one described by Matsumura et al [4] with some areas of
the silicon surface hydrogenated and thus less vulnerable to etching
Pre-treatment with RCA cleaning has no effect on the final morphology and this seems to
exclude pore nucleation being caused by presence of metallic or organic surface
contaminants
The morphology and the distribution density of the pits was quite uniform over
microscopic areas of the sample but changes were observed over the millimetre scale This
suggests that etching intensity is influenced by some large-scale parameter
Large-area measurements made with AFM or with an optical microscope (fig 48)
showed that the inhomogeneity of surface etching can be correlated with the masking
action of the bubbles While the fluoride dissolved the silicon hydrogen bubbles formed
by the reaction covered some areas of the surface thus blocking the etching over that area
generating macroscopic steps at the bubble-liquid border As the reaction proceeded more
hydrogen accumulated and the bubble diameter increased producing in this way a circular
pattern of steps The increase in bubble diameter was not continuous with time otherwise a
uniform surface slope gradient would have been obtained The formation of this circular
ldquoetching staircaserdquo indicates that the bubble growth was stepwise the bubble accumulating
more and more hydrogen without enlarging across the surface until it relaxed increasing
its diameter stepwise and covering more silicon The circular structures in fig 48a are not
co-axial and their asymmetry could derive from physical processes due to stirring or
irregularities on the surface
44 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
a b Fig 48 a) optical micrograph of etching patterns on Si(100) created by NH4F 12x09 mm b) AFM image of the circles border xy range is 40x24 microm z-range is 30 nm
The step structure was not destroyed by the etching even after the bubble detached from
the surface but on the contrary the etching process seemed to be influenced by the
presence of the step
Observing in detail a series of steps (fig 48b) a quite deep trench is visible at the base of
each step A close-up image of a step and the corresponding line profile of the trench is
shown in fig 49 The trench is sim5 nm deep with respect to the lower surface compared to
a step height of 22 nm
A similar structure has been recently obtained with electrochemical etching of p-type
Si(100) in 4 HF [18] in which a ldquocurrent burstrdquo etching model previously described
was assumed for silicon dissolution In that experiment the trench was created at the
border of silicon nitride masks and began to grow after a nucleation stage Preferential
trench etching was along the (110) direction and the trench growth was explained as an
effect of mechanical stress induced by the nitride mask and of electric field enhanced
dissolution which depended upon an external applied potential
While it is clear that in our system the gas bubbles have a masking effect similar to a
classical solid nitride mask it is unlikely that hydrogen present on the surface can induce a
significant stress in the silicon lattice as in the case of a nitride mask Furthermore no
external field was applied to drive preferential etching at the trench site
It has been proposed [19] that the cathodic and anodic part of the etching reaction
(hydrogen production and silicon oxidative etching respectively) take place at different
points on the surface with a net charge transfer between the different areas In this case
the highest reaction rates will correspond to the silicon area surrounding the bubble border
where a high number of positive charges will be available for the reaction Furthermore a
sharp trench extending into the silicon crystal will be a preferential electrostatic attractor
VPalermo 45
Fig 49 AFM image of the etched surface showing a step created on the surface by bubblemasking A stronger etching action is visible on the right side of the step as well as theprotected area on the upper side of the step (indicated by the arrows) Image is 10x10 micromz-range is 30 nm The profile on the right is taken from the central area of the image
for the h+ charge carriers coming from other ldquocathodic areasrdquo of the sample either from
other regions on the surface or from the back of the silicon chip [18]
In the areas where the hydrogen bubble had detached and the surface was exposed to the
etching the reaction was not uniform in the neighbourhood of the steps It is possible to
observe (fig 48b and 49) an area on the upper side of the step where less or even no
etching at all seems to have taken place as if the step was able to protect the surface from
etching While etching on the lower side with trench formation can be attributed to the
presence of the bubble the surface on the upper step side can be etched only after bubble
detachment so no masking effect can account for this result However a further
preferential attraction of h+ charge carriers from the already formed trench can be
hypothised electrochemically shielding the surrounding area from further etching If this is
true the shielding effect would be very strong with a relatively shallow 5 nm-deep trench
protecting an area of sim1 microm parallel to the step
To summarise the etching of Si(100) in NH4F is a complex process in which different
reaction paths both chemical and electrochemical co-exist Hydrogen bubbles formed by
the reaction act as a mask on the surface and create etching paths and inhomogeneous
etching of the surface Different kinds of pores are observed on the surface and in some
cases the anisotropy of the process is so strong as to give square-shaped holes and
underetching
46 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The diffusion of h+ charge carriers in the crystal is one of the main rate-determining steps
of the reaction and leads to the formation of a deep trench immediately outside the bubble
perimeter These trenches act as charge collectors and reduce the etching of the upper step
surface in the proximity of the steps
Bibliography
[1] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[2] Spectroscopic Ellipsometry Studies Of HF Treated Si (100) Surfaces Yao H Woollam
Ja Alterovitz SA Applied Physics Letters 62 (25) 3324-3326 Jun 21 1993 Influence
Of HF-H2O2 Treatment On Si(100) And Si(111) Surfaces Graf D Bauermayer S
Schnegg A Journal Of Applied Physics 74 (3) 1679-1683 Aug 1 1993 Kinetics Of
Oxidation On Hydrogen-Terminated Si(100) And (111) Surfaces Stored In Air Miura
T Niwano M Shoji D Miyamoto N Journal Of Applied Physics 79 (8) 4373-4380
Part 1 Apr 15 1996
[3] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995 Structure Of The Stepped SiSiO2 Interface After Thermal-Oxidation -
Investigations With Scanning Tunneling Microscopy And Spot-Profile Analysis Of
Low-Energy Electron-Diffraction Pietsch GJ Kohler U Jusko O Henzler M Hahn
PO Applied Physics Letters 60 (11) 1321-1323 Mar 16 1992
[4] Enhanced Etching Rate Of Silicon In Fluoride Containing Solutions At pH 64
Matsumura M Fukidome H Journal Of The Electrochemical Society 143 (8) 2683-
2686 Aug 1996
[5] A Study Comparing Measurements Of Roughness Of Silicon And SiO2 Surfaces And
Interfaces Using Scanning Probe Microscopy And Neutron Reflectivity Crossley A
Sofield CJ Goff JP Lake ACI Hutchings MT Menelle A Journal Of Non-Crystalline
Solids 187 221-226 Jul 1995
VPalermo 47
[6] Step-Flow Mechanism Versus Pit Corrosion - Scanning-Tunneling Microscopy
Observations On Wet Etching Of Si(111) By Hf Solutions Hessel HE Feltz A Reiter
M Memmert U Behm RJ Chemical Physics Letters 186 (2-3) 275-280 Nov 8 1991
[7] Comparison Of Si(111) Surfaces Prepared Using Aqueous-Solutions Of NH4F Versus
HF Higashi GS Becker RS Chabal YJ Becker AJ Applied Physics Letters 58 (15)
1656-1658 Apr 15 1991
[8] Wet Chemical Etching Of Si(100) Surfaces In Concentrated NH4F Solution -
Formation Of (2x1)H Reconstructed Si(100) Terraces Versus (111) Faceting Neuwald
U Hessel HE Feltz A Memmert U Behm RJ Surface Science 296 (1) L8-L14 Oct
10 1993
[9] Ideal Hydrogen Termination Of Si(001) Surface By Wet-Chemical Preparation Morita
Y Tokumoto H Applied Physics Letters 67 (18) 2654-2656 Oct 30 1995
[10] Atomic Structures Of Hydrogen-Terminated Si(001) Surfaces After Wet Cleaning
By Scanning Tunneling Microscopy Endo K Arima K Kataoka T Oshikane Y Inoue
H Mori Y Applied Physics Letters 73 (13) 1853-1855 Sep 28 1998
[11] On The Potential-Dependent Etching Of Si(111) In Aqueous NH4F Solution
Houbertz R Memmert U Behm RJ Surface Science 396 (1-3) 198-211 Jan 20 1998
[12] Etching Process Of SiO2 By HF Molecules Hoshino T Nishioka Y Journal Of
Chemical Physics 111 (5) 2109-2114 Aug 1 1999
[13] Etching Mechanism And Atomic-Structure Of H-Si(111) Surfaces Prepared In
NH4F Allongue P Kieling V Gerischer H Electrochimica Acta 40 (10) 1353-1360
Jul 1995
[14] Pore Formation Mechanisms For The Si-HF System Carstensen J Christophersen
M Foll H Materials Science And Engineering B-Solid State Materials For Advanced
Technology 69 23-28 Sp Iss Si Jan 19 2000
[15] Porous Silicon - Mechanisms Of Growth And Applications Parkhutik V Solid-
State Electronics 43 (6) 1121-1141 Jun 1999
[16] Formation And Application Of Porous Silicon Foll H Christophersen M
Carstensen J Hasse G Materials Science amp Engineering R-Reports 39 (4) 93-141 Nov
1 2002
[17] Pore Morphology And The Mechanism Of Pore Formation In N-Type Silicon
Searson PC Macaulay JM Ross FM Journal Of Applied Physics 72 (1) 253-258 Jul 1
1992
48 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
[18] Deep Electrochemical Trench Etching With Organic Hydrofluoric Electrolytes
Christophersen M Merz P Quenzer J Carstensen J Foll H Sensors And Actuators A-
Physical 88 (3) 241-246 Mar 5 2001
[19] Morphology Of Anodically Etched Si(111) Surfaces - A Structural Comparison Of
NH4F Versus HF Etching Houbertz R Memmert U Behm RJ Journal Of Vacuum
Science amp Technology B 12 (6) 3145-3148 Nov-Dec 1994
VPalermo 49
5 Surface modification of silicon in vacuum void creation and oxide desorption
The main reason for the huge success of silicon in the microelectronics industry is not due
to its superior properties as a semiconductor Other materials for example germanium
have better qualities such as higher mobility of charge carriers and lower noise levels
which would allow the construction of faster and higher performance devices
The widespread use of silicon however is mainly due to the outstanding characteristics of
its oxide Silicon dioxide (SiO2) is a very good electrical insulator easy to form
chemically and thermally stable and is compatible with lithographic and metal deposition
processes Germanium oxide on the contrary is too reactive to be used
Even the use of Si(100) substrates for nearly all microelectronic devices is dictated by
oxide quality The (111) face of silicon crystal can be easily cleaved and flattened and
almost atomically perfect surfaces can be obtained with simple chemical procedures (as
described above) But the density of interfacial defects is highest for Si(111)-SiO2
interfaces and lowest for Si(100)-SiO2 ones so microchips will continue to be fabricated
on Si(100) wafers
SiO2 (silica) is present in 95 of the earthrsquos minerals in different allotropic forms such as
quartz tridymite and cristobalite In the bulk each silicon atom is bonded to four oxygens
in a Si-O-Si tri-dimensional network Si-O bonds are 016 nm long and form an angle
ranging from 120deg to 150deg
Three typical intrinsic defects are present in SiO2 The so-called Ersquo centres are oxygen
vacancies with a hole localised on a silicon atom with only three Si-O bonds
50 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
O3Si +SiO3 Whereas the PR (peroxy radical) defects are holes trapped by a charged
peroxy moiety with a O3Si-O-O+ SiO3 structure The NBOHC (non-bridging oxygen hole
centres) derive from water or hydrogen contamination and are schematized as O3Si- O- H-
O-SiO3
The atomic structure of the Si-SiO2 interface varies enormously Local domains resembling
the tridymite and the cristobalite structure of silica are present but it seems that only 10
of the interface is ordered [1] Far from the interface the SiO2 bulk is completely
disordered The passage from bulk Si to stoichiometric SiO2 passes through a non-
stoichiometric SiOx layer 07 nm thick
When a clean silicon surface is exposed to atmospheric oxygen a thin sim2 nm thick layer
of native oxide forms spontaneously which is usually removed and substituted with
thicker better quality oxide layers before further processing
Silicon is usually oxidised by thermal annealing at temperatures between 800deg and 1100deg
in an atmosphere of pure O2 with some water eventually added to increase oxidation speed
Thermal oxides made in pure oxygen (dry oxides) grow more slowly than oxides produced
in an oxygen-water atmosphere (wet oxides) but are usually of better quality
According to the Deal-Groove formula the time t needed to grow an oxide of thickness X
is given by [1]
1
212minus
minusminus
+=
ABXBXt α
where the constant B and BA decrease exponentially with temperature as
minus
kTEexp
with activation energies for dry oxidation of EB =123 and EBA =20 eV respectively
EB is related to the diffusion of oxygen in silicon while the value of EBA is interpreted as
the energy required to break a Si-Si bond The exponent α is 1 for wet oxidation and 0 for
oxidation at high temperatures and low oxygen pressures It has intermediate values for dry
oxidation This formula does not work well for low values of X at the initial stages of
oxidation and usually empirical corrections are used
An interesting characteristic of silicon is that at high temperature and in vacuum oxygen
can actually etch the silicon crystal giving gaseous products and the oxidized layer present
on the crystal becomes unstable (fig 51) [2]
VPalermo 51
1E-10
1E-09
1E-08
1E-07
1E-06
1E-05
1E-04
1E-03
1E-02
1E-01
0607080911112
1000T (1K)
P (T
orr)
SiO2 + Si rarr 2SiO(g) Oxide decomposition
Si+O2 rarr SiO2(s) oxide formation Fig51 phase diagram of the
oxygen-silicon system
Silicon oxidation apart from the initial nucleation stages at the monolayer level proceeds
uniformly over the whole surface with a planar reaction front moving from the surface
into the bulk
If heated under low oxygen partial pressure (vacuum or inert atmosphere) SiO2 is known to
decompose following the reaction
SiO2 + Si rarr 2SiOuarr (1)
The reaction begins with nucleation at defect points on the SiSiO2 interface and proceeds
in a spatially inhomogeneous manner with the formation of large voids on the oxide
surface [3]
Several studies have been made on the dynamics of void growth both on thick [4] and thin
[5] layers of SiO2 The process has been used to decorate otherwise unobservable defects at
the SiSiO2 interface [6] or to grow nanoislands of silicon on the void surface [7] It has
been suggested that the defects acting as nucleation centres could be metallic contaminants
present on the native surface which aggregate and catalyze SiO2 decomposition [8]
After oxide desorption the silicon surface is very rough In particular on Si(100) square
islands are observed several nanometres high which act as lsquopinning sitesrsquo on the motion of
monatomic steps on the Si surface It has been proposed that these islands can be composed
of silicon carbide coming from the organic contaminants present on the surface [9] or of
oxide clusters during partial reoxidation [10]
In the following section some experimental results obtained studying voids and nanoisland
growth are presented
52 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nanoisland growth on silicon
We used two different samples the first rougher than the second (i) p-type Czochralski Si
(100) with a RMS roughness of 05 nm shown in fig52a and (ii) p-type epitaxial Si(100)
with a RMS roughness of 019 nm shown in fig 52b The roughness was measured with
STM on the lsquoas receivedrsquo samples on different areas of 500x500 nm Every sample was
covered by a layer of native oxide ~2 nm thick
Some of the samples (AC) were introduced into the UHV system without any cleaning
others (BD) were dipped in concentrated HF (49) to remove the surface oxide without
etching the silicon [11]
After degassing the samples were heated resistively increasing the temperature slowly to
keep the pressure within the 10-10 mbar range during heating Each sample was held at
900˚C for 30 min For some of the samples the heating was stopped at lower temperatures
to obtain incomplete oxide desorption and surface roughening thus enabling observation
of the different steps of the process Table 51 summarizes the different treatments for each
group of samples
Fig52 Original surfaces of Czochralski (a) and epitaxial (b) silicon Every image is 500x500 nm
VPalermo 53
Table 51 Summary of different treatements
Silicon type Oxide removed
Heating Islands density (microm-2)
Av Island Volume (nm3)
A Czochralski No 30rsquo at 900˚C 24x103 353
B Czochralski Yes ldquo 20x104 71
C Epitaxial No ldquo 28x102 2450
D Epitaxial Yes ldquo 23x104 63
Fig53 ab) Surface roughening on group A samples after heating 30 min at 800deg and900deg respectively in presence of an oxide layer Image size 500x500 nm cd) Surface roughening on group B samples after heating 30 min at 700deg and 900degrespectively after removing the oxide layer with HF Image size 250x250 nm
Fig 52a is the native oxide surface of sample A quite disordered and irregular This
surface was stable when heated to 700˚C while at higher temperatures the oxide began to
desorb Fig 53a was taken after heating the sample at 800˚C for 30 min
54 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The oxide began to desorb in correspondence with defect points creating oxide-free
surface voids which enlarge radially uncovering the silicon surface Some surface silicon
atoms are removed through reaction (1) Mobile silicon atoms moving randomly over the
surface aggregate forming a nanocrystal at the initial defect point surrounded by a stable
flat oxide-free surface with some steps The initial SiSiO2 interface as observed after
oxide etching with HF [12] is similar to the one shown in fig 52a and thus the surface
observed in the void area is a completely new one coming from surface reorganization
The reaction continues leading to a coalescence of the voids the final result of the process
is shown in fig 53b with protruding islands aligned along the main crystal axes and flat
areas with some steps LEED measurements confirmed that this surface is crystalline
silicon Atomic resolution was obtained on the flat areas of the surface showing the
typical 2x1 reconstruction of Si(100) Figures 53c and 53d show the evolution of the
roughening on sample B after chemically removing the native oxide layer before
introducing the sample into UHV As mentioned above the initial morphology is similar to
that shown in fig 52a though the oxide has been removed as confirmed by LEED
measurements Fig 53c shows the surface after heating at 700˚ for 30 min Without the
oxide layer the surface began to reorganize at lower temperatures and the small irregular
grains of the original surface begin to grow and become more rectangular After heating to
900˚C (fig 53d) the surface is flat with small rectangular islands The shape of the islands
is similar to that shown in fig 53b but the lateral island dimension is one order of
magnitude smaller
To assess the importance of the initial surface on the process the same treatments were
repeated starting from flat epitaxial silicon (samples CD) The initial epitaxial surface is
smooth with periodic steps due to a small miscut angle (fig52b) On heating the epitaxial
silicon without etching (sample C) relatively large islands are obtained similar to the ones
shown in fig 53b where the ldquostep-pinningrdquo effect of the islands is evident The average
dimension of sample C islands was sim2500 nm3 while only 353 nm3 for sample A Island
density decreased from 24000 to 280 microm-2 This result is maybe due to a lower
concentration of defects in the oxide layer of the epitaxial silicon which allows the
expanding void to grow much larger before coalescence with neighbouring voids the
nanostructures within them thus reaching greater dimensions To verify this hypothesis we
applied the same treatment to clean not-so clean and purposely contaminated samples To
check that the defects inducing void nucleation and nanoisland growth derive from carbon
VPalermo 55
contamination we used electronic-grade methanol to contaminate the silicon surface
Figure 54 shows the results of this comparison
State-of-the-art cleaned samples obtained with multi-step RCA cleaning [13] showed the
development of very few voids the oxide desorption leaving large smooth areas of oxide-
and nanostructure-free silicon as shown in Fig 54a where the presence of residual oxide
can be seen Samples cleaned with a simpler one-stage Pirana cleaning solution showed a
higher number of islands per unit area (Fig 54b) and samples purposely contaminated
with methanol (Fig 54c) had the highest density of island nucleation of all three samples
When the native oxide had been chemically removed from the epitaxial silicon before
heating (sample D) the island size and density is similar to the standard as-received
silicon case (sample B) Thus starting from a flat substrate the process of surface
roughening and island creation was the same even when the oxide layer had been
previously removed
From these results some important indications can be drawn The first is that oxide
desorption temperature depends upon the contamination levels of the oxide layers (fig 54)
with desorption at lower temperatures for contaminated surfaces
Second once the oxide layer is removed the silicon atoms become mobile on the surface
at relatively low temperatures (~700˚C) Their initial configuration is disordered and has a
high surface energy so they crystallize in the small islands shown in fig53d reducing
their surface area and lowering the energy of the system When the oxide is present the
surface cannot reorganize at 700˚C and is stable At 800˚C the oxide begins to desorb in
correspondence with defect points on the surface As the void area enlarges a small
cba Fig 54 The effect of contamination on nano-island production Gradient-filtered STM images of (a)sample cleaned two times with RCA 400x400 nm (b) Sample cleaned with a Pirana solution 500x500nm (c) sample contaminated with MeOH before insertion into the UHV system 500x500 nm Annealingtime is 40 min at 800degC for all the samples
56 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
crystallite grows at the centre of the voids Further oxide desorption uncovers larger areas
freeing more silicon atoms which feed the initial island making it bigger
Fig 55 shows the scaled size distribution of the islands for each group of samples Every
distribution has been obtained using several images of different samples for each group
The size distribution of all the samples follows an exponential decay and the decay seems
the same for all the samples though the average volume of the islands differs by more than
one order of magnitude (see Table 51 for details)
The point defects which catalyze void nucleation can be metallic contaminants present on
the initial oxide surface [8] or organic contaminants which at high temperatures can
form SiC nanocrystals [14] We used Scanning Tunneling Spectroscopy (STS) to look for
differences between the islands and the surrounding flat silicon surface but no difference
was found Furthermore the total island volume per surface unit is very high (more than
104 nm3microm2) and it seems unlikely that such a huge volume could consist of surface
contaminants Islands of pure silicon on silicon have been grown without evidence of
surface contamination[14]
To summarise the formation of nano-sized islands on silicon through surface diffusion
was studied At high temperatures the oxide layer covering the surface decomposes non
uniformly and circular voids of clean silicon are created The presence of the oxide layer
blocks surface silicon atomic motion and surface reorganization except within the voids
causing the growth of islands more than 10 nm high and 30 nm wide on the silicon surface
According to STM and LEED measurements we can say that the islands are mostly
Fig 55 Size distribution (scaled) for the samples described in table 1 The line is a reference for the eye
VPalermo 57
composed of silicon with traces of other substances such as carbon Final island densities
and dimensions depend upon the initial purity of the oxide layer
On the other hand if the oxide is removed chemically before heating the island growth
process is not so localised and takes place simultaneously over the whole surface yielding
smaller and more numerous islands
58 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[2] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[3] High-Temperature SiO2 Decomposition At The Sio2Si Interface Tromp R Rubloff
GW Balk P Legoues FK Physical Review Letters 55 2332-2335 Nov 1985
[4] Defect Microchemistry At The SiO2Si Interface Rubloff GW Hofmann K Liher M
Young DR Physical Review Letters 582379-2382 Jun 1987 Defect Formation In
Thermal SiO2 By High Temperature Annealing Hofmann K Rubloff GW Mccorkle
RA Applied Physics Letters 49 1525-1527 Dec 1986 Kinetics Of High-Temperature
Thermal Decomposition Of SiO2 On Si(100) Liher M Lewis JE Rubloff GW Journal
Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 5 1559-1562 Aug
1987
[5] Thermal-Decomposition Of Very Thin Oxide Layers On Si(111) Kobayashi Y Sugii
K Journal Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 10 (4)
2308-2313 Part 3 Jul-Aug 1992 Scanning Tunneling Microscope Study On Mid-
Desorption Stages Of Native Oxides On Si(111) Kobayashi Y Sugii K Journal Of
Vacuum Science amp Technology B 9 (2) 748-751 Part 2 Mar-Apr 1991 Controlled
Growth Of SiO2 Tunnel Barrier And Crystalline Si Quantum Wells For Si Resonant
Tunneling Diodes Wei Y Wallace RM Seabaugh AC Journal Of Applied Physics 81
(9) 6415-6424 May 1 1997
[6] Defect Formation In SiO2Si(100) By Metal Diffusion And Reaction Liher M
Dallaporta H Lewis Je Appl Phys Lett 53 589-591 Aug 1988 SiO2 Film
Decomposition Reaction Initiated By Carbon Impurities Located At A Si- SiO2
Interface Raider Si Herd Sr Walkup Re Applied Physics Letters 59 (19) 2424-2426
Nov 4 1991
[7] Nanometer-Scale Si Selective Epitaxial Growth On Si(001) Surfaces Using The
Thermal Decomposition Of Ultrathin Oxide Films Fujita K Watanabe H Ichikawa M
Applied Physics Letters 70 (21) 2807-2809 May 26 1997 Pyramidal Si Nanocrystals
VPalermo 59
With A Quasiequilibrium Shape Selectively Grown On Si(001) Windows In Ultrathin
SiO2 Films Shibata M Nitta Y Fujita K Ichikawa M Physical Review B 61 (11)
7499-7504 Mar 15 2000
[8] Stacking-Fault-Induced Defect Creation In SiO2 On Si(100) Liher M Bronner Gb
Lewis Je Appl Phys Lett 52 1982-1985 May 1988
[9] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990 AFM And XPS Characterization Of The
Si(111) Surface After Thermal-Treatment Lamontagne B Guay D Roy D Sporken R
Caudano R Applied Surface Science 90 (4) 481-487 Dec 1995
[10] Scanning-Tunneling-Microscopy Study Of Oxide Nucleation And Oxidation-
Induced Roughening At Elevated-Temperatures On The Si(001)-(2x1) Surface Seiple
JV Pelz JP Physical Review Letters 73 (7) 999-1002 Aug 15 1994 Evolution Of
Atomic-Scale Roughening On Si(001)-(2x1) Surfaces Resulting From High-
Temperature Oxidation Seiple JV Pelz JP Journal Of Vacuum Science amp Technology
A-Vacuum Surfaces And Films 13 (3) 772-776 Part 1 May-Jun 1995
[11] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995
[12] Morphological Changes Of The Si [100] Surface After Treatment With
Concentrated And Diluted HF Palermo V Jones D Materials Science In
Semiconductor Processing 4 (5) 437-441 Oct 2001
[13] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[14] Nanoscale Roughening Of Si(001) By Oxide Desorption In Ultrahigh Vacuum
Gray SM Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology
B 14 (2) 1043-1047 Mar-Apr 1996
60 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 61
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and nano-lines
When the silicon surface is not protected by a native oxide layer or by a hydrogenated
passivating layer it is very reactive The surface chemistry of both Si(100) and Si(111) has
been extensively studied starting from clean surfaces prepared in UHV because of its great
relevance to the microelectronics industry and its technological interest The diffusion and
reaction of molecules and atoms on silicon is an interesting scientific problem on its own
apart from technological issues because silicon reconstruction yields a very complex and
anisotropic surface
We give here a brief summary of the adsorption behaviour of several elements on Si
surfaces The summary is not comprehensive and only the most interesting characteristics
for each substance are given
Table 61 Summary of adsorption behaviour of atoms and molecules on silicon surfaces [12]
Hydrogen
Molecular H2 shows low reactivity towards silicon while atomic hydrogen easily forms
Si-H bonds and can even break Si-Si bonds
Alkali metals
Alkali metals diffuse rapidly into SiO2 and can damage silicon-based transistors
Transition
metals
All transition metals apart from gold and silver react with Si forming metal silicides
62 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nickel A very common contaminant of silicon Nickel from even slight contact with stainless
steel tweezers can contaminate silicon samples forming its silicide and is very difficult
to remove even at high temperatures because it segregates on the silicon surface
Palladium Forms silicides especially Pd2Si and can be used to make contacts possessing a low
Schottky barrier
Titanium Widely used to fabricate contacts for silicon-based microdevices TiSi2 is one of the
more stable and highly conductive silicides
Tungsten The silicide is more stable than TiSi2 on polysilicon and is thus used for polysilicon
gate contacts
Platinum Silicide is used for bipolar transistors PtSi contacts on n-silicon give Schottky diodes
Cobalt Cobalt silicide gives better quality contacts than TiSi2 at the nanometrelevel but is less
used at the moment because it consumes too much silicon during its growth
Iron Silicides have been studied for potential optoelectronic applications because FeSi2 has a
directly accessible bandgap and is optically active
Group 13
Group 13 metals are used in IC technology to dope silicon (giving p-type doping) and
to make III-V type semiconductors When evaporated on Si(100) or Si(111) they react
strongly with the surface giving surface reconstruction and kink formation at steps
Boron Widely used for doping Usually deposited with decomposition of boron hydrides
Sticking coefficient of simple hydrides is very low so decaboranes (B10H14) are used for
deposition
Aluminum Aluminum is widely used for microelectronic contacts but in some cases it reacts with
silicon diffusing into the crystal and with SiO2 stealing oxygen atoms to form Al2O3
Group 14
Carbon Carbon can form a composite semiconductor with silicon (silicon carbide) widely used
to fabricate integrated circuits for use at elevated temperatures or in the presence of
ionising radiation When carbon is deposited on silicon usually polycrystalline films
with properties between diamond and graphite are obtained
CO adsorbs molecularly on silicon without breakage of the C=O bond Other
unsaturated hydrocarbons do not react with silicon Even very reactive strained
molecules like cyclopropane have low sticking coefficient Ethylene acetylene and
propylene adsorb molecularly on Si(100) each molecule sitting on a silicon dimer
interacting with the dangling bonds of the surface Acetylene also adsorbs molecularly
on Si(111) on the 7x7 reconstructed surface preferential adsorption on centre atoms
with respect to corner atoms is observed Benzene and other aromatic ring compounds
adsorb molecularly on the surface They can form σ or π-type bonds with the silicon
laying parallel or tilted respect to the surface according to the bond type
Silicon Silicon atoms can be deposited on the surface with molecular beam epitaxy or
VPalermo 63
decomposition of silanes and chlorosilanes Silanes with single Si-Si bonds have a high
sticking probability and dissociate upon adsorption giving trihydrides and
monohydrides
Chlorosilanes have high sticking coefficients too and decompose on the surface but
they can etch the surface through the reaction SiCl4 + Si rarr 2SiCl2
Germanium Germanium has a lattice constant similar to silicon (Ge lc is 4 larger than Si) so that
various alloys of Si1-xGex can be formed The growth of Germanium on silicon is quite
peculiar with several monolayers adsorbing uniformly on the surface followed by
island formation (Stranski-Krastanov growth)
Group 15
Nitrogen Exposure to ammonia and nitrogen gas at high temperatures leads to the formation of
silicon nitride layers (Si3N4) NH3 dissociates on the surface and reacts with the
dangling bonds giving Si-NH2 and Si-H termination Silicon nitride is stable on the
surface up to 1100degC above this temperature it desorbs as Si2N
Phosphorous Phosphine (PH3) adsorbs dissociatively as Si-PH2 and Si-H and behaviour is similar to
that of ammonia
Arsenic A monolayer of As is often deposited on silicon as a substrate for GaAs growth Arsenic
forms dimers on Si(100) creating dimer rows which are parallel (perpendicular) to the
underlying silicon dimer rows when it is deposited at high (low) temperature On
Si(111) Arsenic breaks the 7x7 reconstruction giving a 1x1 pattern This 1x1 As
monolayer acts as a passivating layer and can resist further As adsorption oxygen and
air
Antimony and
Bismuth
Due to their large covalent radii Sb and Bi form only short dimer rows on silicon and
only at high temperatures
Group 16
Oxygen
Oxygen can oxidize or etch silicon according to the reactions
Si(s) + O2 rarr SiO2 (s)
Si(s) + frac12O2 rarr SiO(g)uarr
Silicon oxidation has already been described in the previous chapter Molecular oxygen
adsorption on silicon is mostly dissociative O atoms break the Si-Si bonds forming an
Si-O-Si bridge or a peroxy bridge Si-O-O-Si Hydrided silicon surfaces do not have
surface dangling bonds and are thus more protected from oxygen attack Etching is
favoured at high temperatures and low oxygen pressures [3] At T gt700degC silicon oxide
on silicon decomposes as SiO(g)
Water
H2O adsorbs dissociatively on silicon easily on Si (100) and with more difficulty on
Si(111) The molecule decomposes giving Si-OH and Si-H on adjacent sites
64 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Group 17
Fluorine
Fluorine adsorbs and reacts easily on silicon dissociating on the surface and forming
Si-F bonds It can then penetrate the surface and attack polarised Si-Si backbonds The
surface is etched with production of SiF4 and SiF2
Chlorine
Chlorine reacts aggressively with silicon etching the surface Cl2 and F2 are often used
commercially to etch silicon usually in a plasma Atomic Cl and Cl2 adsorb on Si(100)
up to saturation The most stable configuration seems to be a Cl atom bonded to each Si
atom of a surface dimer in a symmetric structure but metastable configurations with
two Cl atoms bonded to a buckled dimer and even a central Cl atom bridged across a
dimer have been observed
On Si(111) silicon mono- di- and tri-chlorides are formed In particular heating at
Tgt400degC a uniform Si-Cl monolayer is obtained and the 7x7 reconstruction changes
into a bulk like 1x1 lattice The surface transition induced by Cl adsorption is fully
reversible The 7x7 reconstruction can be restored on desorbing the chlorine by heating
at Tgt1100degC The 1x1 domains nucleate at the lower terrace side of steps the 7x7 at the
upper step edge
Bromine
Like hydrogen bromine maintains the 2x1 structure of Si(100) forming Si-Br bonds
with the dangling bonds of the silicon dimer rows At high exposure some etching of
silicon by formation of volatile SiBr3 species has been observed
Surface diffusion on silicon
In the previous chapter we observed nano-island growth on silicon surfaces with native
oxide and surfaces contaminated with organic impurities While this phenomenon has been
observed in several experiments the formation and growth mechanisms of nanoislands has
still not been satisfactorily described theoretically Various substances have been
hypothesised as nanoisland nucleating agents such as organic or metallic contaminants
present on the surface or oxide clusters (see previous chapter) The island growth process
has been attributed by various workers to lsquosome sort of kinetic instabilityrsquo [4] to the
pinning of step flow by SiC clusters [5] or to a mesoscopic atom flux from areas with low
step density to areas of higher step density [6]
One mechanism proposed to explain surface roughening and island growth is the Ehrlich-
Schwoebel effect ie the presence of an energetic barrier that reflects atoms approaching a
VPalermo 65
Molecule (L
O2
H2O MeOH
CO CO2 CH4
clean surface
downward
weak on si
The aim o
surface an
To obtain
the ones
spontaneou
first obser
surface W
nanoisland
Experime
We used S
removed f
41 mixtur
water and
whole hea
1200degC re
kept below
Table 62 Treatment island density and island volume for each molecule tested
Dose angmuir)
Heating time at 800degC (min)
Island density (microm-2)
Ave Volume (nm3)
Equivalent layer thickness (nm)
280 10 no islands - -
90 10 no islands - - 30 10 2000 19 004
180 10 360 516 019 30 60 no islands - -
180 10 no islands - - - 10 no islands - -
step [7] Recent experiments however indicate that the Schwoebel effect is
licon [48]
f our experiments was to study the reaction of simple molecules with the silicon
d to check the ability of these molecules to nucleate nanoisland growth
a better understanding of the process we chose simple molecules starting from
which are more likely to be present as traces in UHV chambers where
s nanoisland growth is often observed after sample heating Using STM we
ved at the atomic level the possible interactions of these molecules with the
e then heated the surface and checked the density and dimensions of the
s produced
ntal procedures and results
i(111) wafers p-doped 0015 Ω cm Gross contamination and particles were
rom the sample surface with a standard Pirana etch (15 min dip in a H2S4H2O2
e at T=80degC) After this the samples were thoroughly washed in ultra-pure
introduced into the UHV system After degassing for 5 hours at 600degC the
ting stage was allowed to cool down and a sequence of rapid flashes to 1100deg-
moved the native oxide layer and cleaned the surface Pressure during flashes is
1x10-9 mbar
66 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Surface flatness and cleanliness were checked with STM and the contaminant molecules
introduced into the UHV system through a leak valve Doses of several Langmuirs were
used for the experiments (1 Langmuir=10-6 torr sec-1) Pressure was kept constant at 10-7
mbar during adsorption The STM tip was retracted during gas deposition to avoid any tip
shielding effect After the adsorption process the leak valve was closed the base pressure
in the UHV chamber returning to below 10-10 mbar and the presence of adsorbed
molecules checked on the surface with STM Finally the sample was heated at 800degC for
several minutes to allow nanoisland growth Several different areas were examined on each
sample
Table 62 summarizes the dose of exposure and the heating time for each molecule Other
experiments were carried out using smaller dosing times but only the significant results
are reported here
As expected oxygen and water react with the surface oxidising it The adsorption process
can be easily followed with the STM by lowering the tip from time to time and taking an
image No effect of the STM tip is observed over the scanned areas Upon molecule
adsorption the order of the crystalline surface rapidly degenerates and after several
minutes the surface looks completely covered by irregular atomic-sized protrusions
corresponding to Si-OH or Si-O-Si species On heating to 800degC the contaminating species
were easily desorbed through SiO formation the surface became clean and the crystal
surface periodicity re-established No residues or nanoislands were observed
Amongst the molecules tested was methanol CH3OH reacts readily with the surface and
covers it with a disorderd layer already after sim 30 Langmuirs exposure (fig 61)
Fig 61 Si(111) surface during methanol adsorption at 0 1 and 5 minutes respectively (corresponding to 0 6 and 30 L exposure) The dark irregular lines are borders between different 7x7 domains Image size 60x60 nm
VPalermo 67
Synchroton radiation photoemission spectroscopy experiments by Carbone et al [9]
showed that methanol reacts with the rest atoms of the 7x7 silicon cell already after 1
Langmuir exposure but that longer exposures are needed for methanol to react with the
remaining silicon atoms Methanol adsorbs dissociatively forming SiOCH3 and Si-CHx
species on the surface
After methanol adsorption the sample is heated to 800degC and observed again with STM
As in the case of water and oxygen the heating restores the crystalline surface and the 7x7
pattern returns visible But this time quite a dense array of nanoislands is observed on the
surface (fig 62) The islands have an average diameter of 20 nm and a height of sim2 nm
At temperatures above 400degC the Si-OCH3 and Si-CHx species decompose oxygen and
hydrogen are desorbed and according to [9] carbon atoms remain as SiC dispersed
uniformly on the surface Our STM measurements indicate that the carbide does not
randomly cover the surface but that C atoms are concentrated at the nanoislands Rough
calculations based on island volume suggest that the islands must be composed of a SixC1-x
alloy with x varying between 05 and 1
Following these results we expected carbon monoxide to behave in a similar manner
adsorbing onto the surface and with nanoisland growth However no adsorption was
observed with STM and the surface looked perfectly clean even after a dose of more than
100 L of carbon monoxide Some species did adsorb onto the surface though because
after 10 minutes subsequent heating nanoisland growth was observed Island density in
this case was only 18 of the density obtained with methanol and island dimensions are
Fig62 Nanoislands on Si(111) created after 30 L adsorption of methanol and 10 min heating at 800degC Image size 200x200 nm
68 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
much bigger (see table 62 and fig 63) To check whether the islands had been nucleated
by some other contaminant present in the vacuum chamber we heated a blank sample
without introducing any molecule No island growth was observed Exposing the surface to
other simple molecules including carbon dioxide and methane also gave no nanoisland
formation
We can estimate the amount of carbon deposited on the surface during gas dosing The
total flux of molecules hitting the surface is obtained from the Hertz-Knudsen formula
TkmPF
π2=
Fig 63 STM images of Si(111) after exposure to methanol (left) and carbonmonoxide (right) contaminating molecules and subsequent nanoisland growthImage size 400x400 nm The size distribution of nanoislands is shown below eachimage
VPalermo 69
Fig64 Left Si(111) after 180 L contamination of CO and 16 hours heating at 800degC Right Si(111)without CO contamination after 16 hours heating at 800degC Image size 250x250 nm Vertical scale is thesame for both images z-ranges are 35 and 5 nm respectively
where P is the pressure in Pascal k=138x10-23 J K-1 is the Boltzmann constant T is the
temperature and m is the molecular weight (sim32 for methanol and sim28 for CO) in
kilograms
Using this formula we obtain a flux of 027 molecules nm-2 sec-1 for methanol and 029
molecules nm-2 sec-1 for CO After ten minutes exposure at 10-7 mbar pressure more than
150 molecules will have hit each square nanometer of the surface Surface density of
Si(111) 7x7 is sim16 atomsnm2 so each surface atom will be hit by several molecules which
could react or be adsorbed even assuming a sticking coefficient much lower than unity
To explain nanoisland growth we hypothesize that CO molecules adsorb molecularly and
-5
0
5
10
15
20
25
30
0 20 40 60 80 100 120
nm
nm
10 min40 min16 h16 h- no CO
Fig65 STM profiles of nanoislandsgrown for different annealing times The lower curve correspond to the surface heated at 16 hours without CO contamination
70 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Table 63 Island density and volume for the CO-contaminated surface at different heating times For each sample the thickness of a uniform layer having the same volume per unit area of the islands is calculated
Molecule Dose
(Langmuir) Heating time at
800degC (min) Island density
(microm-2) Ave Volume
(nm3) Equivalent layer thickness (nm)
CO 180 10 362 517 019 CO 180 40 550 1072 059 CO 180 16 h 505 3675 186
clean surface - 16 h rough - -
very weakly on Si(111) moving rapidly on the surface Because of their rapid diffusion
the STM is not able to image the CO molecules on the silicon [10]
Increasing the temperature part of the CO molecules will surely desorb but some of them
will dissociatively adsorb onto the surface The oxygen will then desorb as SiO while
carbon will remain fixed forming very stable Si-C bonds
CO decomposition will be favored at surface defects surface steps or in the proximity of
already formed SiC clusters An increased reactivity of the CO molecule on a surface in
correspondence with phase boundaries has already been observed on Pt surfaces [10]
In this case a reduced number of nucleation centres will react with the CO molecules
yielding large and fewer islands with respect to the case of methanol which does not have
the possibility of travelling long distances over the surface
Other factors are likely to influence the process of island nucleation after CO adsorption
the co-adsorption of other molecules which can slow down CO and favour its
decomposition cannot be ruled out as well as the formation of new defects at higher
temperatures Measurements with a variable temperature STM or with some other surface
analysis technique are planned in the future to confirm the proposed mechanism
When the surface was further annealed the nanoislands grew in size After 16 hours
heating very large nanoislands with diameters of sim35 nm and heights of sim20 nm became
visible (fig64 left) As a comparison when the same surface was heated without
nucleation centres deriving from contamination an irregular surface was obtained (fig64
right)
While the nanoisland density seemed to reach an asymptotic limit of 500 islandsmicrom2
(table 63) island size continued to grow with time (fig65) even when no further carbon
was supplied to the surface This indicates that even though the initial nucleating core of
the nanoisland is likely to be an SixC1-x alloy further growth is due to silicon atoms
diffusing from the crystal and being adsorbed by the growing island The final volume
VPalermo 71
occupied by the islands corresponds roughly to a 18 nm thick overall layer of removed
silicon
It is noteworthy that the clean sample after 16 hours annealing even though smoother than
the CO-contaminated one looked much more disordered with hardly any flat area visible
on nanometres scale In the presence of nanoislands however further annealing increases
island size but keeps part of the surface quite flat and ordered with flat areas visible
between the islands Thus the nanoislands can be imagined to act as ldquoimpurity sinksrdquo for
further contaminants approaching the surface yielding a greater but more ordered surface
roughening with respect to the uncontaminated silicon surface
Circular arrays of nanoislands
We used the voids described in the previous chapter formed during oxide layer
decomposition as nanoscopic masks to test the validity of the hypothesis outlined above
The small circular areas uncovered during oxide decomposition can act as ldquoskating rinksrdquo
where contaminants and diffusing silicon atoms can move freely over the surface while
the oxide layer all around and higher than the void area as well as being chemically
bonded to the surface silicon atoms will hinder surface diffusion Furthermore the oxide
passivating layer will at least in some cases prevent contaminating molecules from
adsorbing and decomposing on the surface In this way selective nanoisland growth inside
void areas can take place
The experimental procedure consisted of the following steps
1) An oxide-passivated sample was heated for 10 minutes at 800degC Oxide decomposition
began in correspondence with surface defects and spread laterally creating naked areas
of silicon
2) While the void enlarges the same defect that initiated oxide decomposition behaved as
an attractor for moving silicon atoms and nucleated the growth of a nanoisland at the
void centre
3) The void surface was exposed to contaminating molecules
4) The sample was then heated again at 800degC During the second heating the void
continued to enlarge and new nanoislands were created The freshly uncovered silicon
surface provided mobile silicon atoms to feed nanoisland growth
72 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
5) When the heating continued for long enough all the oxide was desorbed and a clean
silicon surface covered by circular groups of nanoislands was obtained
Figure 66 shows the different stages of the process
We thus heated an oxidized sample at 800degC for 10 minutes in UHV to create multiple
voids in the oxide layer After this we exposed the sample to the contaminating molecules
Then we further annealed the surface at the same temperature to promote island growth
The first time we observed nanoisland growth in the void area was by chance after heating
for a second time samples which had been stored for 2 weeks in vacuum Even at pressures
as low as 10-11 mbar some molecules will adsorb onto the surface slowly creating defects
and thus potential nanoisland nucleation centres It is possible to observe with STM that
the silicon crystal surface looks very clean just after a high temperature flash but even
after only a few days randomly adsorbed species will be observable on the surface
Unfortunately it is not possible to identify these species simply from in situ STM
measurements
Mass spectrometry measurements show that the residual gas contaminants in vacuum are
usually He Ar H2 CH4 CO CO2 and N2 [11] Some of these molecules such as
hydrogen will stick to the silicon surface at room temperature but will simply desorb
during annealing without nucleating nanoislands Others like CH4 or CO2 will not react
with the surface to generate nanoislands as demonstrated previously The best candidate
for contamination of samples stored in UHV seems thus carbon monoxide has the ability
321
4 5
Fig 66 Selective nanoisland growth within oxide voids See text for details
VPalermo 73
as shown before to nucleate nanoislands although co-adsorption mechanisms can not be
ruled out Fig 67 (left) shows the circular groups of nano-islands obtained after heating
void-covered samples stored for 2 weeks in UHV Small islands are observed in the former
void area The void itself has enlarged slightly uncovering a clean oxide-free area of the
silicon surface Apart from the central island created during the first period of heating the
surrounding ones have dimensions decreasing from the void border to the inner area of the
void
This confirms that most of the material needed for island growth comes from the silicon
atoms diffusing from the freshly uncovered areas around the original void The outer
nucleating centres are nearest to the silicon atom source and thus generate larger islands
For a faster process we directly exposed the void to a significant concentration of CO
Fig67 (centre) shows the islands grown after a 180 Langmuir exposure and subsequent
heating The islands are preferentially located around the original void perimeter the rest
of the original oxide-free surface remaining untouched This suggests that although the
whole surface is exposed to CO the gas only interacts chemically with the surface at the
SiSiO2 interface around the oxide-free void perimeter The SiSiO2 interface thus provides
preferential nucleation sites for nanoisland growth in the successive thermal annealing
step This recalls a similar phenomenon observed in the catalytic oxidation of CO on the
Pt(111) surface where CO molecules reacted with adsorbed O species only along the
perimeters of oxygen islands [10]
On repeating the CO exposure and thermal annealing steps a second circle of nano-islands
Fig67 Left circular area of nanoislands obtained from a sample stored 2 weeks in UHV 500x500 nmCentre nanoisland circle obtained after exposing the voids to 180 L of CO and heating for 10 min at800degC 400x400 nm Right two concentric circles obtained with further CO adsorption followed byheating 650x650 nm
74 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
was formed around the newly-expanded surface void perimeter (fig67 right) This
process can in principle be repeated again to obtain a series of concentric groups of
nanoislands the only limit depending on the minimal distance between neighbouring
expanding voids We have thus observed that contaminants adsorbed at low pressure over
long periods of time generate random island growth over the whole void area while CO
molecules adsorbed at relatively high pressures over short periods yield selective
nanoisland growth at the former void border Several theories can explain the different
results CH4 and CO2 as mentioned above did not nucleate islands when when the silicon
surface was exposed to them for short periods However the lower surface mobility of
these contaminants and the greater time available for reaction with the surface could be
the cause of nucleation of SiC clusters over the whole oxide-free void area and not only at
its border The coadsorption of water another UHV residual gas with CO on the naked
silicon areas could perhaps allow a surface reaction similar to that between CO and the
SiSiO2 interface during CO exposure Another explanation for the formation of these
random nucleation sites within the voids could be the coadsorption of residual carbon-
containing species present in the UHV system with hydrogen the major residual gas in
stainless steel UHV systems
The possibility of decorating silicon oxide borders with nanoislands using the high
mobility and selective decomposition of CO molecules suggests interesting applications
for nanofabrication processes that will be discussed below
Silicon nanowire creation on Si(100)
We performed most of the experiments described above on silicon (111) 7x7
reconstructed because this surface is easily prepared in UHV and above all because it is
isotropic and has no preferential directions for atom diffusion The 2x1 reconstruction of
silicon (100) even though it has a much simpler unit cell than Si(111) 7x7 shows a more
complicated morphology with lots of monoatomic and biatomic steps and dimer rows
parallel to each other The orientation of the dimer rows changes by 90deg on alternate
atomic layers Surface diffusion in this case is much more complicated because atoms will
experience the effects of the surface anisotropy and diffuse preferentially along or across
the dimer rows The diffusion energy of a silicon atom as example is 06 eV along a
dimer row and 085 eV across different dimer rows This while being a complication
VPalermo 75
offers interesting possibilities for the creation of ordered structures The monoatomic steps
and the dimer rows can for example act as templates for the formation of elongated
structures of composition similar to the islands described previously
On heating a Si(100) surface oxide decomposition and void growth takes place as
described for Si(111) In some cases the void shape reflects the surface symmetry
especially for very thin oxide layers and the voids often have a slightly squared shape
Nanoisland growth is also influenced by the substrate symmetry and the islands look
rectangular as irregular clusters surrounded by concentric patterns of monoatomic steps
The whole structure reminds vaguely a ldquozigguratrdquo a kind of stepped pyramid found in
Mesopotamia (fig68 inset)
Apart from nanoisland growth in some cases we observed spontaneous formation of
nanowire-like SiC structures on the Si(100) surface The wires were generated after
exposing a Si(100) surface in vacuum to traces of CO followed by sample annealing at
800degC for 15 minutes The wires exhibit lengths ranging from 10 to 100 nm and average
widths of ca 5 nm Although being randomly positioned on the surface they are perfectly
aligned along the crystal axes of the substrate (Fig68)
Several descending monoatomic steps can be observed around each nano-line and nano-
island Very often one of the line extremities coincided with an island
Nanoline
Nanoisland
Fig 68 A Si(100) surface covered by nanoislands and nanowires The monoatomic steps present on the surface are visible The two insets show a typical nanoisland and a nanowire obtained on this surface The typical ldquodimer-row ldquostripes are visible along the sides of the nanowire
76 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Couples of parallel lines starting from the same nano-island were also observed At high
resolution (fig 69) the lines showed a lamellar periodic structure with a periodicity of
around 4 nm and frequent defects Scanning Tunneling Spectroscopy measurements
showed no significant difference between the line and the surrounding silicon surface
The proposed formation mechanism for this kind of structure involves two stages in the
first one organic contaminating molecules adsorb and diffuse on the surface from the
vacuum environment stopping preferentially at surface singularities including steps and
defects
Several substances such as atomic carbon carbon monoxide ethylene and fullerene
decompose when adsorbed on silicon surfaces at temperatures above 500degC and create
localised defects on the surface inducing strain deformations and a metastable surface
reconstruction [12]
If these molecules remain attached to the steps they can diffuse rapidly along step border
until they meet a line and decompose In this way long lines of SiC defects parallel to the
surface steps can be produced In the second phase the clusters of silicon carbide can
-005
0
005
01
015
02
025
03
035
04
-3 -2 -1 0 1 2 3
V
dId
V lt
dId
Vgt
linesilicon
Current- Voltage behaviour measured with STS Fourier analysis of wire periodicity
asymp 4 Aring
High resolution image of a line
Fig 69 Top high resolution image of a nanowire showing the lamellar structure Fourier analysis (bottom left) shows that the lamellae periodicity is sim 04 nm bottom right the IV characteristics measured with the STM tip on the line and the silicon surface
VPalermo 77
25-600degC
Final linear structure
600-800degC
Contaminant Silicon
Fig 610 Schematic representation of nanowire formation mechanism
locally inhibit the spontaneous surface diffusion of silicon atoms acting as templates for
the reorganisation of surface steps (fig 610)
The final morphology shows disordered lines and islands each surrounded by a complex
pattern of silicon monatomic layers
At 800degC silicon atoms are highly mobile on the surface Due to atom diffusion the
surface reorganizes with a step-flow mechanism The presence of a fixed line of
contaminants blocks atom diffusion and step flow generating the monoatomic steps
around each wire (fig 611) SiC clusters act as lsquopinning sitesrsquo on silicon atom surface
diffusion and can be purposely used to direct the localised growth of nano-islands [13]
Although the possibility of using methanol and carbon monoxide as nano-island precursors
has been demonstrated the exact nature of the substances nucleating nano-wire growth is
still unknown
Further experiments will be needed to confirm these hypotheses as well as to identify the
contaminating species nucleating nanowire growth Carbon monoxide is a main candidate
for nanowire growth but the real nucleation process is likely to be a complex one
78 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Step flowblocked
Fig 611 Representation of step flow blocked by the presence of a nanowire 105x105nm
Bibliography
[1] Surface-Chemistry Of Silicon Waltenburg HN Yates JT Chemical Reviews 95 (5)
1589-1673 Jul-Aug 1995
[2] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[4] Growth Of Si On The Si(111) Surface Lanczycki Cj Kotlyar R Fu E Yang Yn
Williams Ed Das Sarma S Physical Review B 57 (20) 13132-13148 May 15 1998
[5] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
VPalermo 79
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990
[6] Nanoscale roughening of Si(001) by oxide desorption in ultrahigh vacuum Gray SM
Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology B 14 (2)
1043-1047 Mar-Apr 1996
[7] Step Motion On Crystal Surfaces Schwoebel Rl Journal Of Applied Physics 40 614-
618 Feb 1969
[8] Scanning Tunneling Microscopy Investigation At High Temperatures Of Islands And
Holes On Si(111)7x7 In Real Time Evidence For Diffusion-Limited Decay
Hildebrandt S Kraus A Kulla R Wilhelmi G Hanbucken M Neddermeyer H Surface
Science 486 (1-2) 24-32 Jul 1 2001
[9] Methanol Adsorption On Si(111)-(7x7) Investigated By Core-Line Photoemission And
Mass Spectrometry Of Photodesorbed Ions Carbone M Piancastelli Mn Zanoni R
Comtet G Dujardin G Hellner L Surface Science 370 (1) L179-L184 Jan 1 1997
[10] Atomic And Macroscopic Reaction Rates Of A Surface-Catalyzed Reaction
Wintterlin J Volkening S Janssens Tvw Zambelli T Ertl G Science 278 (5345)
1931-1934 Dec 12 1997
[11] Redhead PA Hobson JP Kornelsen EV The Physical Basis Of Ultrahigh
Vacuum Chapter 12 (Chapman amp Hall London 1968)
[12] The Si(001) C(4 X 4) Surface Reconstruction A Comprehensive Experimental
Study Norenberg H Briggs Gad Surface Science 430 (1-3) 154-164 Jun 21 1999
[13] Production Of Nanostructures Of Silicon On Silicon By Atomic Self-Organization
Observed By Scanning Tunneling Microscopy Jones D Palermo V Applied Physics
Letters 80 (4) 673-675 Jan 28 2002
80 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 81
7 Conclusions and proposals for future work It is very difficult to foresee the future of nanoscience and nanotechnology even over the
next few years exciting discoveries are continuously made opening up new possibilities
and progress in this sector is now one of the fastest and most varied So it is impossible to
predict whether computers will in the future be made with nanowires nanotubes single
electron devices or some kind of quantum-based transistor What does seem clear
however is that silicon devices can still further extend their levels of miniaturization to
remain competitive for at least the next ten years [1] During this period new emerging
technologies will reach scientific maturity and arrive at the production lines
Even when new technology prototypes do become available for full-scale production the
astronomical cost of changing from silicon to new technologies will be a major concern
Moreover a vast amount of extremely detailed knowledge is already available on all
aspects of silicon technology Thus the possibility of integrating innovative
nanotechnologies with standard CMOS silicon technology already seems an attractive
prospect [2]
For this and other reasons we have focussed our research efforts on surface phenomena on
silicon wafers which could be potentially useful for the development of silicon-compatible
nano-devices
Some important conclusions can be drawn from the work described in this thesis
bull Silicon surfaces of different crystal faces were studied at atomic resolution in ultra-
high-vacuum following transformations induced on the surface by chemical (etching)
and physico-chemical (organic contamination heating) treatments
82 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull The presence of surface contaminants induces at high temperature the growth of
nanometre-sized islands and lines on silicon
bull Final island densities and dimensions strongly depend upon the chemical nature of the
contaminant molecules and the treatments used Molecules interacting only weakly
with silicon will desorb upon heating whereas molecules which decompose easily on
Si will give widespread random nano-island nucleation Between these two extreme
cases complex behaviours of diffusion and subsequent decomposition are possible
bull Nanoscopic naked silicon voids were produced in oxide layers through carefully-
controlled thermal annealing It was found that void density and size depends upon
oxide purity surface contamination levels and annealing conditions
bull The growth of nano-islands and nano-voids was exploited simultaneously and the
voids used as nanoscopic masks to control the positions of growing islands Complex
silicon-on-silicon nano-structures were obtained with multi-step growth processes
bull Finally the use of macroscopic treatments and simple physical processes to produce
large quantities of nanometre-sized structures on silicon surfaces was demonstrated
These processes and the structures produced are of potential interest to the
microelectronics industry
We showed in previous chapters how it is possible to modify silicon surfaces using
chemical and physical methods The use of self-organization for the production of Si-on-Si
20 nm
Fig 71 Left SEM image of field emitter arrays made by conventional lithography [3] Center 3-D STM image of a self-organized void with a central nano-island Right Lateral schematic view of a field emitter device
VPalermo 83
nano-structures has the advantage with respect to other emerging techniques of being
silicon compatible and thus is a potential candidate for its implementation in the
production of new devices Furthermore this technique can create billions of strucures on a
wafer in only a few minutes being thus applicable to large-scale production
Even when ordered structures are obtained these methods still show a poor reproducibility
and control Chemically modified surfaces look very disordered at high magnification
with trenches and holes etched into the surface and progress has recently been made in
producing very small structures electrochemically [4] Oxide decomposition with
nanoisland growth in UHV can also yield ordered structures The void holes with
nanoislands located at their centres although quite variable in dimensions are more similar
to engineered devices than to a spontaneous random surface-roughening process In fact
there is a surprising similarity between the nanovoid-nanoisland structures and well-known
commercially available devices called field emitter arrays (FEA) These are small tips used
as electron guns in various kinds of displays and other devices Both the commercial
devices and our self-organized structures consist of a protruding tip at the centre of
microscopic holes on an insulating layer covering a semiconductor surface (see fig71)
Of course the order and size uniformity of commercial FEAs is much better than that
obtained with our process but it is encouraging to see that a self-organized structure
obtained with simple heating has a very similar surface topography and composition of
commercial working devices obtained after many expensive stages of optical lithography
deposition and etching
Nano-island density and size can to some extent be controlled and different
contaminants can be adsorbed onto the surface to modify the growth process In the future
with a better understanding of diffusion dynamics and growth mechanisms a multi-stage
fabrication process can be envisaged where purposely-designed molecules are adsorbed
onto silicon surfaces to diffuse organize in ordered arrays and favour surface
reorganization at the nanometre scale The same molecules could be used to deliver
dopants to obtain for example highly conducting nanostructures on otherwise undoped
highly resistive silicon The selective formation of nanostructures at the SiSiO2 border
can be used to scale down mesoscopic patterns to the nanometric level A silicon oxide
layer can be easily patterned on the surface with normal optical lithography and then
nanostructures can be built following the pattern edge In this way the production of very
small nanowire based devices may be possible (fig 72)
84 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
D
Gate
S
D
S
SiO2
SiO2
2 Heating nanowire creation
and oxide removal
3 Source-drain deposition by conventional lithography
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
4 Insulating layer (oxide) and gate (metal) deposition
Fig 72 Schematic drawing of the construction of a self-organized nanowire-based transistor
Surface diffusion and decomposition of molecules can be sensitive to even smaller surface
features The nanoscopic lines formed on Si(100) are very regular and neat if their
nucleation and growth mechanism could be controlled ordered networks of nanowires
could be made on silicon
VPalermo 85
The techniques developed here are still experimental ones the careful control and
reproducibility of each step is still an issue and further studies will be needed to clearly
understand the dynamics of formation of these structures Nonetheless the results obtained
so far open up the possibilities of producing novel devices such as for example nano-
transistors (see scheme in fig73 where the mechanism described in fig 72 is applied for
the fabrication of a device array) The possibility of using the border of a lithographic
pattern to draw nanolines can in principle allow the creation of nanodevices using normal
IC manufacturing techniques
The results obtained and their description in this thesis are not the first examples of the use
of self-organization for the production of ordered structures and will certainly not be the
last In our opinion the most important outcome of these findings is that it is possible to
use surface diffusion to overcome the inherent limits of lithographic techniques in
microdevice production processes and to force properly selected molecules to react with
the surface in a spatially differentiated manner at particular sites thereby creating ordered
series of nanostructures The role of the surface dynamics of the substrate atoms
themselves is extremely important in this process
Outstanding results have been and continue to be obtained in nanoscience and
nanotechnology research promising the emergence of new production technologies Sooner
or later those emerging technologies will have to face the issues of cost and compatibility
with the enormous investment made in existing technologies and processing facilities The
commercial production of nanostructures based on surface diffusion processes could in
principle be cheap simple and compatible with existing technology Moreover the
processes of diffusion in bulk silicon is well-known in the microelectronics industry which
has both the know-how and the equipment to better understand and exploit the surface
diffusion processes on silicon surfaces studied here
Extremely important and innovative results are often obtained by exploiting very simple
ideas and in our opinion physical processes such as surface diffusion used to create
complex nanostructures on silicon surfaces is a prime example
86 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
2 Heating nanowire creation and
oxide removal
3 Source-drain deposition by conventional lithography
4 Insulating layer (oxide) and gate (metal) deposition
Fig 73 Schematic macroscopic outline of the construction of a self-organized nanowire-based transistor
VPalermo 87
Bibliography
[1] P Gargini Intel Technology Coordinator ldquoFrom Microelectronics To
Nanotechnologiesrdquo Invited Campus Colloquia Lecture CNR Research Area Bologna
February 6th 2003
[2] Toward A Hybrid Micro-Nanoelectronics Cerofolini Gf Ferla G Journal Of
Nanoparticle Research 4 (3) 185-191 Jun 2002
[3] Recent Progress In Field Emitter Array Development For High Performance
Applications Temple D Materials Science amp Engineering R-Reports 24 (5) 185-239
Jan 25 1999
[4] Silicon Dioxide Micropillars For Sieving Fabricated By Macroporous Silicon-Based
Micromachining Izuo S Ohji H French Pj Tsutsumi K Kimata M Sensors And
Materials 14 (5) 239-251 2002 Electrochemical Etching In HF Solution For Silicon
Micromachining Barillaro G Nannini A Piotto M Sensors And Actuators A-Physical
102 (1-2) 195-201 Dec 1 2002
88 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 89
Acknowledgements
This work would not have been possible without the help and guidance of Dr Derek Jones
he introduced me to the use of STM and other techniques found the funding for this PhD
and helped me with the thousands of small and big problems I had to face during my
research I am grateful to Prof Alberto Ripamonti too for being the supervisor of this
thesis Thanks are also due to Dr Giancarlo Seconi director of ISOF-CNR where most of
this work was carried out Financial support from the Italian National Research Council
(CNR) is also gratefully acknowledged
My passion for surface science began when Fabio Biscarini showed me that it was possible
to really see the atoms and for this I will always thank him
During these years I had the luck to collaborate with many good scientists among those I
would like to cite Enrichetta Susi Massimo Cocchi and Anna Mazzone from CNR
Claudio Zannoni Anna Cavallini Daniela Cavalcoli and Antonio Castaldini from the
University of Bologna Sergio Pizzini and Maurizio Acciarri from the University of Milan
One of the most stimulating periods of my PhD was the one I spent at the Steacie Institute
for Molecular Sciences in Ottawa working with Robert Wolkow Dan Wayner Greg
Lopinski and Peter Kruse I thank them for discussing together many exciting scientific
ideas and trying to put some of them into practice
Special thanks go to my colleagues Paolo Samorigrave and Stefano De Cesari for long and
useful discussions on Science Life and their interactions often with the help of a good
pint
Finally I would like to thank Vassilia Gaetano Simone Claudia Silvia Letizia
Alessandro Angela and Sebastien even though not directly involved in this work they
have been over these three years a continuous source of happiness and support
Bologna March 2003
90 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
SCIENTIFIC PUBLICATIONS bull Lateral diffusion of titanium disilicide as a way to contacting
hybrid Si-organic nanostructures Palermo V Buchanan M Bezinger A Wolkow RA APPLIED PHYSICS LETTERS 2002 v81 p 3636 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v6 issue 20) bull Surface modifications in Si after Rapid Thermal Annealing Castaldini A Cavalcoli D Cavallini A Jones D Palermo V Susi E JOURNAL OF THE ELECTROCHEMICAL SOCIETY 2002 v 149 pG633 bull Production of nanostructures of silicon on silicon by atomic self-
organisation observed by scanning tunnelling microscopy Jones D Palermo V APPLIED PHYSICS LETTERS 2002 v 80 p 673 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v5 issue 5) bull Nucleation of nanostructures from surface defects on silicon Palermo V Jones D SOLID STATE PHENOMENA 2002 v 82-84 p 687 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo MATERIALS SCIENCE AND ENGINEERING B 2002 v88 (2-3) p220 bull Morphological changes of Si[100] surface after treatment with
concentrated and dilute HF Palermo V Jones D MATERIALS SCIENCE IN SEMICONDUCTOR PROCESSING 2001 v 4 p 437 bull Electrical and structural properties of processed silicon surfaces Susi E Cavallini A Castaldini A Cavalcoli D Jones D Palermo V ldquoRECENT RESEARCH DEVELOPMENTS IN VACUUM SCIENCE amp TECHNOLOGYrdquo 2001 v 3 p189 bull Numerical Solutions of the Stochastic Equations of Crystal
Growth Mazzone A M Palermo V INTERNATIONAL JOURNAL OF MODERN PHYSICS C 2000 v 11 Part 1 p195-204 bull Advances in silicon surface characterisation using light beam
injection techniques
VPalermo 91
Acciarri M Pizzini S Simone G Jones D Palermo V MATERIALS SCIENCE AND ENGINEERING B 2000 V73 (1-3) p 235 - 239 bull Abrupt orientational changes for liquid crystals adsorbed on a
graphite surface Palermo V Biscarini F Zannoni C PHYSICAL REVIEW -SERIES E- 1998 V 57 NUMBER 3A p R2519-R2522 Oral presentations bull The use of oxide desorption and surface diffusion for the
creation of silicon on silicon nanostructures Palermo V Jones D 1ST NATIONAL WORKSHOP ON CURRENT TRENDS IN NANOTECHNOLOGIES Catania (Italy) Feb 2002 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo V E-MRS Spring Meeting Strasbourg (France) June 2001 bull STM study of surface transformations on silicon during UHV
annealing Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999 bull Advances in silicon surface characterisation using light beam
injection technique Acciarri M Pizzini S Simone G Jones D Palermo V E_MRS SPRING MEETING June 1999 Posters bull Spontaneous nano-wire growth on silicon Palermo V Jones D SMARTON workshop Leuven (Belgium) October 2002 bull Ordered circles of nano-islands on silicon from CO adsorption Palermo V Jones D TRENDS IN NANOTECHNOLOGY 2002 Santiago de Compostela (Spain) September 2002 bull Production of nanostructures of silicon on silicon by atomic self-
organisation Palermo V Jones D EUROMAT Rimini (Italy) June 2001 bull Etching holes and anisotropic corrosion on silicon [100] Palermo V Jones D Susi E Asoli B SILICON WORKSHOP Genova (Italy) February 2001
92 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull Morphological and electrical characteristics of damaged silicon surfaces
Susi E Castaldini A Cavalcoli D Cavallini A Jones D Palermo V SILICON WORKSHOP Genova (Italy) February 2001 bull Effect of HF etching on the roughness of a silicon surface Palermo V Jones D NATIONAL CONFERENCE ON PHYSICS OF MATTER Genova (Italy) June 2000 bull Effect of Native Oxide Desorption upon the Surface Morphology of Si[100] by STM and LEED Palermo V Jones D SILICON WORKSHOP Genova (Italy) February 2000 bull Characterisation of silicon surfaces for microelectronics through STM measurements Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999
VPalermo 93
Contact Vincenzo Palermo ISOF- Institute for Organic Synthesis and Photoreactivity Via Gobetti 101 40129 Bologna ITALY Tel +39-051-6398336 Fax +39-051-6398349 Mail palermoisofcnrit
94 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Index
1 Introduction nanotechnology and the future of computers
2 Silicon surfaces
3 STM and other surface analysis techniques
4 Surface modification of silicon in liquid Nano-hole creation
5 Surface modification of silicon in vacuum void creation and oxide desorption
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and
nano-lines
7 Conclusions and proposals for future work
8 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
on the silicon wafer This resist is then selectively removed and used as a mask for
processing the silicon surface (fig 12 left) The maximum resolution attainable depends
upon the wavelength used and current technology is near to its intrinsic resolution limit
On the other hand there is strong scientific and economic demand for further development
in IC miniaturization to obtain more powerful and complex computers Besides every-day
life applications more powerful computers are fundamental for much scientific research
such as climate change tracking genome sequencing and fluid dynamics Increased
miniaturization is also fundamental for reducing power which has to be dissipated by the
chips which run at progressively higher frequencies Energy consumption by
microelectronic devices is already an issue and represents one of the main obstacles for
the continuing growth in wireless communication (cell phones portable computers CD
and DVD players digital cameras etc)
Thus it is expected that new production methods different from current lithographic ones
will be developed methods which allow modification of a surface well below the 100 nm
limit and even down to single atom manipulation Techniques such as Scanning
Tunnelling Microscopy and Atomic Force Microscopy are already capable of moving
single atoms (see fig 12 right) but unfortunately building a working nanodevice in this
way would take a very long time and these techniques are difficult to apply to large scale
production
Nowadays thousands of researchers are working in the nanotechnology field towards a
new generation of microelectronic devices Several possible solutions are competing for
tomorrowrsquos computer architecture and there is still no clear winner It is likely that the
final solution will be the combined use of different techniques and components (including
molecules nanowires and nanodots) as they become available together with conventional
Fig 12 Left conventional litographic process [2] Right atomic manipulation of iron atoms on copper [3]
VPalermo 9
silicon technology
Below is a brief summary of the most recent developments in nanotechnology and
nanoscience
Actual trends in nanotechnology
Perhaps the most fascinating idea for nanodevice construction is to use one single
molecule working as a complete device The first molecular diodes (ie molecules
conducting current only in one direction) were created in 1997 in 1999 a molecular fuse
and a molecular transistor were demonstrated although there was no possibility of wiring
these devices to external contacts In April 2001 James Heath and his group at UCLA
fabricated an array of overlapping crossbars and placed a small molecule of rotaxane
between each crossbar (fig 13 left) This composite molecule is made up of two
component parts the main rod-like molecular axis and a mobile ring ldquothreadedrdquo on it like a
bead on a necklace and can function as a molecular switch A working 16-bit memory
circuit was constructed using these molecules For a brief review of these works see [4] In
June 2002 a single molecule transistor was built by connecting an organic molecule to two
metal contacts the molecule contained one or two atoms of a transition metal (cobalt or
vanadium) forming the active region of the device supported by an organic backbone [5]
Fig13 Working nanodevices Left schematic representation of rotaxane molecules between crossed nanowires [4] Right SEM image of semiconductor nanowires forming a small circuit [9]
10 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another approach to nanodevice fabrication has become possible through the discovery of
carbon nanotubes which were observed for the first time in 1991 by a Japanese electron
microscopist studying the material deposited during arc-evaporation synthesis of fullerenes
[6] They consist of a graphite-like carbon seamless cylinder with a diameter of several
nanometers and lengths of up to a millimetre Carbon nanotubes are very stable can
behave as metals or semiconductors and can host other molecular or ionic species thus
modifying their electrical behaviour In 2001 Avouris and his group reported the first
circuit made with a single nanotube [7] A few months later Cees Dekker presented a
nanotube-based transistor able to amplify an input signal by a factor of ten and built
several logic circuits using these nanotube transistors [8]
One problem with carbon nanotubes is that it is very difficult to control their electronic
properties ie their metallic or semiconducting behaviour An alternative to carbon
nanotubes are semiconductor nanowires Silicon nanowires can be made using a laser to
vaporize the silicon together with a metal catalyst like iron or gold The vapour condenses
in nanosized drops of silicon and metal from which the wires slowly grow out as more
silicon is adsorbed In 2001 a group at Harvard University [9] created a transistor by
crossing two different nanowires After this the same group arranged four nanowires in a
noughts and crosses grid creating something like a 4-bit memory (fig 13 right) Even
metallic nanowires made of platinum and silver can be used in a crossed configuration to
store information [10]
There are some issues common to all these new technologies though First it is difficult to
imagine these methods applied to large-scale production Up to now the insertion of a
molecule between two electrodes is an occasional lucky event while nanotubes and wires
have to be positioned on the surface creating the appropriate contacts on them manually
The large-scale production of integrated circuits using these building blocks will not be
straightforward and does not seem imminent
Another issue is of an economic and not a scientific or technological nature Since 1965
the cost of IC manufacturing plants has sky-rocketed If the cost of semiconductor
production plants continues to rise exponentially in a few years such plants will cost up to
$20 billion This is a sizeable investment even for large companies such as IBM or Intel
For this reason it is likely that IC companies will resist changing to completely new
technologies closing down their existing plants As it is clear that silicon will remain the
VPalermo 11
fundamental raw material of the IC industry for the foreseeable future nanotechnology
developments for microelectronics will need to be silicon-compatible In these early days
of nanotechnology the most valid approach would seem to be the addition of molecular
functions to existing silicon technology ndash using the latter as a foundation on which to build
Fabrication of self-organised structures on silicon
The possibility of using the phenomenon of atomic or molecular self-organization to create
nanostructures on silicon has already been demonstrated The clean silicon surface shows
in some cases a high degree of order and complex surface reconstruction as will be
described later Several different ordered structures form spontaneously on this surface
such as series of monatomic steps or boundaries between reconstructed areas It has been
demonstrated that it is possible to use these structures to fabricate ordered nanodots and
nanolines on the surface [11] More recently well-defined nanometric patterns have been
obtained with selective etching of silicon using nitric oxide [12]
In this study the possibility of creating different types of nanostructures on the silicon
surface is explored Methods had to be developed which were
- Simple They must not need complex masks or lithographic steps to create the
structure but rather exploit self-organisation phenomena
- High resolution the silicon surface modifications should be on a scalelength of below
100 nm
- Fast billions of nanostructures have to form over the whole surface simultaneously to
be compatible with large-scale production requirements
- Cheap they must not require expensive equipment (such as e-beam lithography) but
exploit simple chemical andor physical treatments to produce nanostructures on the
silicon surface
In Chapter 2 the main characteristics of silicon are described Chapter 3 provides a
summary of the techniques used for this research Chapter 4 examines the chemical etching
of silicon in different liquid environments and the effects of this etching on the surface at a
nanoscopic level with the creation of nanoholes
12 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Chapter 5 describes the growth in ultra-high vacuum (UHV) of nanoscopic voids and
islands on the silicon surface and the effect of surface oxide on this growth
Chapter 6 discusses the modification of silicon surfaces in UHV following the adsorption
of molecules and thermal heating to produce nanoislands and nanolines on silicon
The overall conclusions of our work are summarised in Chapter 7
Finally we will give some conclusions based on the results obtained and discuss possible
applications of the methods developed
Bibliography
[1] From wwwintelcom
[2] From wwwsematechorg
[3] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
[4] Molecules Get Wired Service RF Science 294 (5551) 2442-2443 Dec 21 2001
[5] Coulomb Blockade And The Kondo Effect In Single-Atom Transistors Park J
Pasupathy AN Goldsmith JI Chang C Yaish Y Petta JR Rinkoski M Sethna JP
Abruna HD Mceuen PL Ralph DC Nature 417 (6890) 722-725 Jun 13 2002 Kondo
Resonance In A Single-Molecule Transistor Liang WJ Shores MP Bockrath M Long
JR Park H Nature 417 (6890) 725-729 Jun 13 2002 Nanotechnology - Electronics
And The Single Atom De Franceschi S Kouwenhoven L Nature 417 (6890) 701-702
Jun 13 2002
[6] Smallest Carbon Nanotube Ajayan PM Ijima S Nature 358 (6381) 23-23 Jul 2 1992
[7] Carbon Nanotube Inter- And Intramolecular Logic Gates Derycke V Martel R
Appenzeller J Nano Letters 1 (9) 453-456 Sep 2001
[8] Logic Circuits With Carbon Nanotube Transistors Bachtold A Hadley P Nakanishi T
Dekker C Science 294 (5545) 1317-1320 Nov 2001
VPalermo 13
[9] Logic Gates And Computation From Assembled Nanowire Building Blocks Huang Y
Duan XF Cui Y Lauhon LJ Kim Kh Lieber CM Science 294 (5545) 1313-1317
Nov 9 2001
[10] Formation And Disappearance Of A Nanoscale Silver Cluster Realized By Solid
Electrochemical Reaction Terabe K Nakayama T Hasegawa T Aono M Journal Of
Applied Physics 91 (12) 10110-10114 Jun 15 2002
[11] Fabrication And Integration Of Nanostructures On Si Surfaces Ogino T Hibino H
Homma Y Kobayashi Y Prabhakaran K Sumitomo K Omi H Accounts Of Chemical
Research 32 (5) 447-454 May 1999
[12] Ultrafine And Well-Defined Patterns On Silicon Through Reaction Selectivity
Prabhakaran K Hibino H Ogino T Advanced Materials 14 (19) 1418-1421 Oct 2
2002
14 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 15
2 Silicon surfaces
The name silicon (silicio in Italian) comes from the latin word silex Amorphous silicon
was first isolated by Berzelius in 1824 by reaction of potassium with silicon tetrafluoride
Thirty years later the first crystalline silicon was prepared Silicon makes up 25 of
earthrsquos crust and is the second most abundant element after oxygen Elemental silicon is
not found in nature occurring as silicon oxide (sand quartz amethyst flint etc) or
silicates (asbestos clay mica etc) Perhaps no other element and its compounds has such
a wide range of uses Silicon compounds such as sand and clay are used in the building
industry as refractory materials for high-temperature applications and for enamels and
pottery Silica is the main component of glass silicon carbide is an important abrasive and
silicones are commonly used polymers and lubricants
Here the most interesting use of silicon of course is for the production of
microelectronics devices For this application silicon of high purity (999999) and of
high crystallinity is needed Table 21 lists some of the physical characteristics of silicon
High purity polycrystalline silicon is produced by the reaction of gaseous trichlorosilane
with hydrogen in a furnace Then to prepare a single-crystal of silicon the so-called
Czochralski method is commonly used
Polycrystalline silicon is melted in a quartz furnace at 1415degC in an argon atmosphere
Then a seed of single-crystal silicon is lowered into contact with the melt and slowly
pulled out In this way the crystal grows and a crystalline cylindrical ingot several metres
long is created from the initial seed
After cooling down the ingot is sliced into thin silicon wafers The wafer surfaces are
polished using a counter-rotating lapping machine in an Al2O3 slurry until the surface is
very flat and shiny ready for the lithographic processes
16 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another way to obtain single crystal silicon is the Floating Zone (FZ) method in which a
silicon cylinder is slowly passed through a heating ring The area inside the ring melts and
solidifies smoothly crystallising as it comes out of the ring yielding a single silicon crystal
Microelectronic devices are built on the silicon surface which is the surface of interest
here Unfortunately silicon surfaces are normally quite dirty and uneven at the atomic
scale Atmospheric oxygen and humidity react with silicon surfaces creating a thin layer of
oxide (called ldquonative oxiderdquo) which is usually irregular and full of defects Different kinds
of contaminants also adsorb onto the surface These are usually small organic molecules
and microscopic dust particles A clean surface on exposure to the atmosphere is
completely covered with gas molecules in less than 10-9 seconds If the pressure is
reduced letrsquos say to 10-6 mbar this time increases to 1 second This is the reason why to
study a clean surface we have to work in UHV at pressures below 10-10 mbar
The atoms in the silicon crystal have a diamond-like structure each atom having 4 bonds
in a tetrahedral sp3 arrangement with bond angles of 10947 degrees At the crystal
surface some atoms will have non-bonding orbitals ldquodanglingrdquo in the vacuum ie sp3
orbitals with a lone electron which are highly reactive These orbitals are known as
dangling bonds To minimize surface energy the surface will reorganize by decreasing
the number of dangling bonds
Table 21 Physical data of silicon [1]
Atomic Weight 2809 Lattice constant (A) 543095 Crystal structure Face-centered cubic
(diamond) Melting point 1415 degC
Density (gcm3) 2328 Boiling point 2355degC Atomscm3 50E22 Minority carrier
lifetime (s) 25E-3
Dielectric Constant 119 Specific heat (Jg degC)
07
Breakdown field (Vcm)
~3E5 Thermal conductivity (Wcm degC)
15
Electron affinity x(V)
405 Vapour pressure (Pa) 1 at 1650degC 1E-6 at 900deg C
Energy gap (eV) at 300K
112 Reactivity Inert to acids Attacked by halogens and alkaline
solutions Intrinsic carrier
conc (cm-3) 145E10 Oxidation states +4 -4
Intrinsic Debye Length (microm)
24 Energy of a Si-Si bond (eV)
232
Intrinsic resistivity (Ω-cm)
23E5
VPalermo 17
Dangling bond densities and positions and thus the type of surface reconstruction will
depend upon crystal orientation as well as the temperature and kinetics of the system
Fig 21 shows a drawing of the main faces of a silicon crystal The angle α between any
(11n) face and the (100) face can be calculated from 2cos 2 += nnα The angle
between any (11n) face and the (111) face can be calculated from
)2(3)2(cos 2 ++= nnα
The chemistry and physics of the faces are very different a brief description will be given
for the most important orientations
Si (100)
On the (100) surface each atom has two Si-Si bonds connecting it to the bulk and two
dangling bonds pointing outward Surface energy is reduced by the dimerisation of the
surface atoms through overlap interaction of one dangling bond per atom forming rows of
dimers aligned along the (110) direction This is the well-known ldquo2x1rdquo reconstruction of
this silicon surface The symmetric dimers would make the silicon surface metallic but to
reduce surface stress the dimers tend to buckle and the surface is thus semiconductive It
took several years to understand that the dimers are buckled because at room temperature
Fig 21 Schematic view of the principal orientations of a silicon surface Surface atoms are white bulk atoms are black dangling bonds are gray [2]
18 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
they shift easily from one buckling direction to the other thus appearing symmetric under
STM observation Fig 22 shows an STM image of the 2x1 reconstructed surface
Even almost perfect (100) surfaces have a certain number of monoatomic steps and the
dimer rows on atomic layers are aligned at 90deg to those on adjacent layers Dimer rows are
thus perpendicular or parallel to the step When the dimers on the upper side of the step
are parallel to the step the step is called SA if they are perpendicular the step is called SB
Because of this symmetry SA step edges will be smoother compared to the more broken
fragmented edges of the SB steps
A common defect on the Si(100) surface is the presence of nickel contamination which
appear as missing dimers in STM images This type of contamination is so critical that
even if the silicon sample is only briefly brought into contact with stainless steel tools
(tweezers for example) the 2x1 reconstruction of the surface can be blocked
Silicon atoms can diffuse easily over the silicon surface as monomers and dimers
especially at elevated temperatures The anisotropy due to the 2x1 reconstruction causes a
difference in the diffusion energies of adsorbates over the surface Diffusion of these
silicon species along dimer rows for example will be much easier A list of diffusion
energies for monomers and dimers is given below [2]
Diffusion on Si(100) 2x1 Ed (eV) Monomers along dimer rows 06 Monomers across dimer rows 085 Monomer formation energy 18 Dimer along dimer rows 11 Dimer across dimer rows 15 Dimer formation energy 26 Dimer binding energy 076 Vacancies along dimer rows 17 Vacancies across dimer rows 19
SA
SB
Fig 22 STM image of a 2x1reconstructed silicon surfaceshowing the dimer rows and steps40x35 nm Nickel-induced defectsare visible as dark spots SA and SBsteps are indicated
VPalermo 19
So the diffusion energy for both monomers and dimers is nearly 40 greater if they have
to cross a dimer row This difference reduces to sim10 for vacancy diffusion
Si(113)
The (113) surface can be imaged as a sequence of alternating (100) and (111)-like
structures with two and one dangling bonds on alternate atoms respectively Interest in the
(113) surface is scientific as it has been used to study the energetics of the (100) and (111)
surfaces as well as for surface adsorption experiments
Si(100) surfaces can easily develop (113) facets
Si(111)
This surface besides being the first one imaged with STM with atomic resolution is one of
the most studied because it is the best cleavage face of silicon and because it shows one
of the most complex and elegant reconstructions in surface science
All Si-Si bonds in the silicon crystal are perpendicular to a (111) plane so this face will
have the lowest number of dangling bonds created per unit area In fact each Si atom on a
(111) surface shows a single dangling bond oriented perpendicular to the surface and
bonded to three back atoms These three bonds for each surface atom account for the great
chemical and physical stability of the Si(111) surface Surface energy is 009 eV Aring-2
compared to 015 eV Aring-2 for Si(100)
For energy minimization this surface reconstructs forming a huge 7x7 lattice cell
containing 102 atoms described by the Dimer-Adatom-Stacking fault model (DAS) For a
detailed description of cell structure see fig 23
The cell described by this model is very complex being composed of three kinds of atoms
adatoms rest atoms and corner hole atoms Furthermore a subsurface stacking fault is
present in one half of the cell making the two halves of the unit cell look different under
STM (Fig 24) It took 26 years of research to completely understand the exact structure of
the 7x7 reconstruction
Cleaving a silicon crystal along a (111) plane produces a metastable 2x1 reconstruction
the 7x7 reconstruction is easily obtained by flashing at high temperature in UHV At T gt
20 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
830degC a disordered 1x1 phase covers the surface Cooling down to 800degC leads to the
formation of the 7x7 phase If the cooling process is too rapid small 7x7 domains nucleate
and a disordered 1x1 phase is preserved between domain boundaries
Si(110)
Even though as mentioned before the (111) plane is the favoured cleaving plane of
silicon thin (100) commercial wafers will not break along this plane because the angle
between (100) and (111) is too far from 90deg (see table 22) Instead they will break along
the (110) plane because it is perpendicular to the (100) surface Each surface atom on
Si(110) has a Si-Si bond pointing downward one dangling bond pointing outward and two
Si-Si bonds parallel to the surface in a zig-zag pattern (see fig 21) Cleaved (110)
surfaces are disordered but upon annealing at high temperatures an ordered complex 16x2
Fig23 Scheme of the 7x7 DAS model [2] In each unit cell there are 9 dimers 12 adatoms and a stacking layer fault The force driving this complex reconstruction is the minimization of dangling bonddensity The DAS model shows the lowest number of dangling bonds (19) of all possiblereconstructions 12 dangling bonds are at the adatoms 6 at the rest atoms and 1 at the corner hole atom This surface is metallic
VPalermo 21
reconstruction takes place The surface appears as a series of long ridges and valleys
parallel to each other Eventually tilted facets of orientation (17 15 1) can form on this
surface The adsorption of Ge atoms on this surface leads to the formation of self-
assembled nanowires [3]
Table 22 Angles in degrees between different silicon faces [2]
Orientation
113
110
111
100
100
2524
9000
5474
0
111
2950
3526
0
110
6476
0
113
0
Fig 24 STM image of a Si(111)surface with 7x7 reconstruction A unitcell with its adatoms is highlightedImage size 13x13 nm
22 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] Weast RC Handbook Of Chemistry And Physics (Chemical Rubber Co Cleveland
1972)
[2] Dabrowski J Mussig H Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] The Structure Of Clean And SiGe-Covered Si(110) Surfaces Butz R Luth H Surface
Science 365 (3) 807-816 Oct 1 1996
VPalermo 23
3 STM and other surface analysis
techniques
Scanning Tunneling Microscopy
Since the invention of the optical microscope at the end of the 16th century the possibility
of examining surfaces at higher and higher magnification has fascinated mankind
Development of the technique continued and towards the end of the 19th century optical
microscopes were as good as todays standard instruments The physical limits of the
wavelength of visible light (350-800nm) had been reached
In the 1920s de Broglie showed that electrons can behave like waves and the use of these
particles for imaging with much higher resolution soon followed Atomic resolution using
this technique is only possible in the transmission mode with extremely carefully prepared
samples
In 1982 using the peculiar properties of piezoelectric materials Binnig and Rohrer brought
a metallic tip very very close to a silicon surface and scanned it across an extremly small
area (fig31) The tunneling of electrons from the tip into the sample or vice versa allowed
them to obtain a local density of electronic states (LDOS) map of that surface Although
theory (which treated the extreme point of the tip as a sphere) then excluded the possibility
of atomic resolution following a tip crash into the surface Binnig and Rohrer began to
observe the LDOS with atomic resolution For this discovery and their following work
they were awarded the Nobel Prize in Physics in 1986
24 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The basic principle of STM is very simple A metallic tip is scanned over a surface without
making ohmic contact and a tunneling current passes between the tip and the surface An
electronic circuit keeps this current constant by raising and lowering the tip during the
scan In this way recording the tip height at each point a three-dimensional image of local
density of electronic states (LDOS) of the surface can be obtained To explain the
extremely high resolution attainable by this simple technique quantum theory is needed
According to classical physics the current will flow between sample and tip only if they
are in physical contact If there is a vacuum gap between the two the electrons will simply
remain confined for example within the surface without the possibility of passing into the
tip
In quantum physics however the electrons have a certain probability of passing
(tunneling) across the gap appearing on the other side of the gap in this way reaching the
tip It can be shown that the probability of an electron tunneling through a gap of thickness
z is
kzep 22)0( minusprop ψ h
φmk
2= (1)
where ψ(0) is the electron wavefunction at the surface-gap border m is the electron
mass=91x10-28 g and φ is the work function of the metal (ie the energy required to
remove an electron from that material For silicon it is 48 eV) The tunneling current thus
decays exponentially with z and is extremely sensitive to topographical imperfections
present on the scanned surface A rough formula giving the current as a function of z is [1]
zFS eEVI φρ 0251)( minusprop
Fig 31 Binnig and Rohrer with the first STM Image from IBM [2]
VPalermo 25
where ρs(EF) is the local density of states at the Fermi level on the given surface For
example the formula predicts that for silicon an increase in tip-surface distance of 1 Aring
will give a 95 decrease in tunneling current
This huge dependence of tunneling current upon the distance allows detection even of the
sub-nanometre changes in height given by the single atoms of which the surface is
composed and thus to resolve them in the LDOS images Of course this description of the
tunneling process is oversimplified and for a more accurate one the electronic states of
the tip of the sample and their interaction have to be taken into account Fig 32 shows a
schematic representation of the interaction between tip and sample orbitals
The exponential decay of current with distance also yields high lateral resolution If the tip
is approximated as a sphere of radius R and the current passing at the minimum tip-sample
distance is I0 then the current passing at a lateral distance x from this point will be
Rxk
eII 22
0
2minus
=
Assuming a tip radius of 100 nm the current is concentrated in an area sim15 nm wide at tip
apex
Very sharp tips with even smaller curvature radii can be produced in several ways Simple
Fig32 Schematic view of tip-sampleorbitals interaction a) no interaction b)equilibrium c) sample positive d) tippositive [1]
26 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
tungsten wires mechanically cut are capable of obtaining atomic resolution on graphite in
air but for more disordered and rougher samples sharper and more reproducible tips are
needed
STM tips are mostly made by electrochemically etching a W or Pt-Ir wire The tips we
used were prepared using methods based on the work of Fotino [3]
A tungsten wire 038 nm diameter is immersed in a KOH solution(06M) with a thicker
tungsten wire used as a counterelectrode The cathodic and anodic reactions involved in the
etching are
Cathode 6H2O + 6 e- rarr 3H2(g) + 6 OH-
Anode W(s) + 8 OH- rarr WO42- + 4H2O + 6 e-
A potential of 3V ac is applied to the tungsten and the wire is immersed in the solution
until a constant current of sim100 mA is established The etching rapidly removes metal
shaping the wire end as a sharp tip When the potential reaches 11 V the coarse tip etching
is finished The wire is then removed from the solution carefully inserted into an
insulating plastic tube leaving only the tip exposed and re-immersed in the solution with
the tip pointing upwards A more gentle etching is thus made to reduce tip radius Usually
5 minutes etching at 07 V ac is used In this configuration very small hydrogen bubbles
formed on the tip sides sliding upwards with a ldquohoningrdquo effect on the tip
This procedure yields extremely sharp and reproducible tips at the microscopic level
After the etching the tip is thoroughly rinsed in ultrapure water then dipped into
concentrated HF to remove surface oxides and hydroxides [4] The tip is dried with
nitrogen inserted into the UHV system and degassed overnight at sim150degC
The possibility of measuring sub-nanometric distances would be useless without being able
to control tip movement over such a minute scale To scan the tip over the surface a
piezoelectric scanner is used Piezos are usually made of an alloy of PbZrO3 and PbTiO3 a
material which contracts or expands when a voltage is applied to it The Omicron
instrument used in our laboratory has three such piezo scanners for xy and z tip motion
allowing one to scan the tip over the surface with sub-Aringngstrom precision (fig 33)
To isolate the instrument from ambient vibrations the whole STM stage is suspended upon
four springs which eliminate all frequencies above 1 Hz and surrounded by a crown of
VPalermo 27
copper wings and fixed magnets Parasitic currents generated by the magnets into the
copper wings contrast every movement of the stage and efficiently block stage vibrations
The STM can be used not only to explore surface topography but to measure the IV
characteristics of single atoms or molecules on the surface (Scanning Tunneling
Spectroscopy or STS) Furthermore it can be used to modify the surface with voltage
bursts digging into it or delicately moving single molecules or atoms over a surface [5] It
can be used in vacuum in air and with proper lateral isolation of the tip even in liquids
A major drawback of STM is that it works only on conducting and semiconducting
surfaces and thus cannot be used on many surfaces of biological and chemical interest
Another instrument more suitable for these and other applications is the Atomic Force
Microscope (AFM)
Atomic Force Microscopy
AFM was invented in 1986 by Binnig Quate and Gerber after calculating the possibility of
building a cantilever with a force constant of the same order of magnitude as that of a
chemical bond
In AFM a tip mounted on a microscopic cantilever (usually made of Si3N4 fabricated with
optical lithography) is brought close to a surface When the tip touches the surface the
cantilever is very slightly deflected upwards The movement is measured by observing the
Y-PIEZO
SILICON SAMPLE
Fig 33 A picture of theSTM used for theexperiments The tripodpiezo scanner is shown
Z-PIEZO
X-PIEZO
TIP
28 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
deviation of a laser beam hitting the upper face of the cantilever Fig 34 provides a
schematic view of the principle of AFM
The typical force constant of the cantilever varies from 00006 to 2 Nm the typical
resonance frequency is 3 to 120 kHz The AFM tip can apply a force on the sample of up
to 10-9 N The AFM can be used on conductive or insulating surfaces in vacuum air or
liquids Furthermore the tip can be modified to sense electrostatic potentials (electric force
microscopy) or magnetic fields (magnetic force microscopy) it can even be functionalized
with complex molecules such as proteins to interact with biological surfaces
A drawback of the AFM is that the force it exerts can damage the surface under
observation especially if the sample is soft (as in the case of cell membranes for
example) This problem can be overcome using the instrument in tapping mode (where the
tip does not move laterally during its brief contact with the surface) or in non-contact mode
in which the tip oscillates above the surface during the scan and the changes in its
frequency due to interaction with surface are monitored The shifts in the oscillating
frequency of the cantilever due to tip-sample interaction are then used for imaging the
surface In this mode interaction of the tip with the surface is minimal and soft samples
can be imaged
STM and AFM are the main techniques used for this work A brief description of other
techniques used occasionally is given below
Fig 34 Scheme of an atomic force microscope
VPalermo 29
Low Energy Electron Diffraction (LEED)
Electrons with energies in the 20-500 eV range are diffracted by a crystalline surface the
diffraction peaks are visualized on a fluorescent screen This technique probes the long
range order of the surface up to a depth of several nanometres
X-Ray Photoelectron Spectroscopy (XPS)
XPS allows both qualitative and quantitative chemical analysis of the elements present on
or near the sample surface
An X-ray source is used to photoionize the atoms on a surface and produce photoelectrons
By measuring the kinetic energy of the photoelectrons the binding energy of the electronic
levels can be calculated This energy will depend on the chemical environment of the
surface atoms
Although the soft X-rays used penetrate to a depth of ~2000Aring the sampling depth of the
technique is determined by the mean free path of the photoelectrons which allows their
escape from only the first 10-100Aring
Secondary Ion Mass Spectroscopy (SIMS)
High and low energy ions (primary ions) are used to bombard a sample and remove surface
atoms and ions The ionic fragments removed (secondary ions) are then analysed by a mass
spectrometer The surface can be consumed during the measurement and profiles obtained
giving concentrations of the materials composing the sample at different depths (depth
profiles)
A popular variant of SIMS is TOF-SIMS In this technique the secondary ion masses are
measured by a time-of-flight (TOF) measurement The secondary ions generated by the
bombarding primary ions are accelerated to a constant kinetic energy and then move
through a field-free space before they reach the detector where their intensity is measured
as a function of flight time Since ions with different masses have different velocities at a
30 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
given kinetic energy the measured flight times of the ions can easily be converted to their
masses The static nature of this latter technique allows mass spectroscopy surface analysis
with minimal damage to the surface
Bibliography
[1] Chen CJ Introduction To Scanning Tunneling Microscopy (Oxford University Press
Oxford 1993)
[2] From wwwibmcom
[3] Tip Sharpening By Normal And Reverse Electrochemical Etching Fotino M Review
Of Scientific Instruments 64 (1) 159-167 Jan 1993
[4] A Convenient Method For Removing Surface Oxides From Tungsten STM Tips
Hockett LA Creager SE Review Of Scientific Instruments 64 (1) 263-264 Jan 1993
[5] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
VPalermo 31
4 Surface modification of silicon in liquid
Nano-hole creation
Liquid treatments of silicon wafers are very common in the integrated circuit (IC)
manufacturing industry They are used to clean and improve surface uniformity to create
and etch protective oxide layers and to remove photo-resist layers
Crystalline silicon with its native oxide layer is very stable and is resistant to many acids
It is easily attacked by hydrofluoric acid (HF) and alkaline solutions
The thin (~2 nm) passivating layer of native oxide (SiO2) is formed on exposure to the
atmosphere This surface layer contains many defects and contaminants so it is usually
chemically stripped and substituted with a better chemically-formed protective oxide
The most common silicon cleaning procedure is the RCA method named after the Radio
Corporation of America [1] It consists of two steps in the first one the surface is treated
with a hot alkaline solution (H2OH2O2NH4OH 411) to remove particles from the
surface following this a hot acidic solution (H2OH2O2HCl 411) is used to remove
metal contamination Other well-known cleaning methods are IMEC (a sequence of
cleaning steps in H2OO3 and dilute HF) or the Pirana etch (a hot 41 mixture of H2SO4
H2O2)
32 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The standard RCA clean removes surface contaminants etches the native oxide and
oxidizes the silicon surface leaving a uniform layer of silicon oxide which better protects
the surface from further contamination
Etching with fluorine-based solutions
Hydrofluoric acid is one of the most common reagents used in the treatment of silicon
wafers both in the research field and in industrial processes A rapid dip in dilute HF is the
simplest way to remove the native oxide from Si(100) and leaves the surface passivated by
a layer of Si-H bonds Because of the low polarization of Si-H bonds the Si-H layer is
stable even for several days protecting the surface from contamination It has often been
assumed that this short etch does not significantly change the surface morphology of the
silicon substrate[2] even though a prolonged dip in dilute HF leads to surface roughening
[3]
Although dilute HF roughens the Si(100) surface at the atomic scale [45] immersion in
concentrated HF (49) etches the surface oxide without attacking the Si surface
uncovering in this way the buried SiSiO2 interface The final counter-intuitive result is
that dilute HF etches the silicon while concentrated HF leaves the crystalline silicon
untouched [3]
Etching Si with fluorine-containing solutions at different concentrations and pH can
produce different morphologies from rough surfaces to flat nearly ideal Si-H terminated
surfaces
Hessel et al and Higashi et al demonstrated in 1991 that very flat Si(111) surfaces can be
obtained using 40 NH4F while etching with HF always yields rough surfaces The
surface becomes smoother because the etchant rapidly attacks Si atoms at step borders
thus removing surface kinks and irregularities in a step-flow mechanism [6 7] Later on
even smoother and more perfect surfaces were obtained by removing oxygen from the
solution after it was discovered that oxygen dissolved in 40 NH4F initiates the formation
of triangular etch pits It was not possible to obtain flat surfaces by etching Si(100) with
ammonium fluoride solutions which leads to the formation of small 2x1 dimer-row
reconstructed (100) terraces together with (111) facets [8]
This difference is caused by the different hydride terminations prevailing on the (100) and
(111) faces While the ideal Si(111)-H surface is monohydride terminated the more
VPalermo 33
reactive dihydrides predominate on the Si(100)-H surface making it more vulnerable to
etching The etching reaction is thus strongly anisotropic etching (100) facets faster than
(111) thus producing (111) microfaceting on Si(100) crystals
A more uniform Si(100) surface can however be prepared by etching at low pH with an
HFHCl mixture [9] or by using very dilute HF solutions and ultrapure water with low
dissolved oxygen and carbon contents [10]
Electrochemical etching can also be used applying anodic or cathodic bias to the silicon
to obtain different morphologies [11] by varying the potential isotropic or anisotropic
etching is observed The aforementioned results show that despite the simplicity of the
reactants fluoride etching of silicon is quite a complex reaction
Fig 41 Chemical etching of silicon
HO+H
H HO
H
H
H
FSi
Si
SiSi
+H2O
-OH-
+F-
H2O H H
Si H
F OH
H Si
Si Si
-H2
+H2O
H
H
H
FSi
Si
SiSi
H
H
H
OHSi
Si
SiSi
H
H
H
H Si
Si
Si Si
Etching mechanism of silicon
HF rapidly dissolves the SiO2 passivating layer on silicon leaving the surface almost
completely hydrogenated [12] After this two different types of reactions etch the silicon
simultaneously one chemical and the other electrochemical [13] The overall etching
mechanism can be schematized in two stages (see Figure 41)
i) Si-H bonds are substituted by Si-F or Si-OH bonds creating a partial charge on the
surface silicon atom and polarizing its Si-Si backbonds
ii) these polarised backbonds are then more easily attacked by HF or H2O After
rupture of the Si-Si bond the atom is removed leaving behind new Si-H
terminations and the reaction can start again
34 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
These reactions take place although at different rates on both Si(100) and Si(111)
Stage i) is usually the rate-determining step of the reaction and the stability of the Si-H
bonds depends upon the pH the concentration of nucleophilic species in solution and an
eventual potential applied to the crystal
For pH gt5 as in the case of concentrated NH4F solutions reaction begins with attack by
water to give Si-OH (step ArarrC) The -OH group is rapidly substituted by fluorine with
polarization of the underlying Si-Si bonds These bonds are then easily attacked by water
the silicon atom being released into solution as HSiF(OH)2 The Si-OH rarr Si-F substitution
is not fundamental for the reaction and etching can proceed even for Si-OH terminated
atoms but XPS measurements showed the presence of a certain number of Si-F bonds
remaining Furthermore fluorine seems to have a catalytic effect on Si-H substitution as
indicated by the dependence of the etch rate upon the F- concentration at least for pH
values between 4 and 8
Si-F bonds can be easily removed by a water rinse In the case of strongly alkaline
solutions (pH=14) OH- groups act directly as nucleophiles and no fluorine is needed to
catalyze Si-H bond rupture
At pH lt4 almost no free F- ions are present in solution and the etch rate is very small at
pH lt2 all etching reactions are very slow and this explains the stability of Si crystals in
concentrated (50 ww) HF solutions
Si Si
Si Si
H
H
H
H Si
Si
Si Si
H
H
H Si
Si
SiSi
H
H
H
OH-H+ -e
(F-)
-H+
-e
+H2O hellipas in fig1
Fig 42 Electrochemical etching of silicon by H2O
The electrochemical etching of silicon involves electron transfer from the surface atoms to
the valence or conduction band of the crystal (according to the type of doping of the
silicon) Several studies have been carried out at different electric potentials In the case of
p-silicon or n-silicon at anodic potentials where positive charges (h+) are available in the
crystal a silicon-centred radical can be created by capture of a hole from the bulk (fig
42) The F- ion does not participate in the reaction but is thought to contribute through
VPalermo 35
electrostatic interactions by lowering the energy of the interaction step After the
formation of the Si-OH group the reaction proceeds as shown in the scheme of fig 41
Matsumura et al [4] proposed that not only water but HF2- molecules also play a major role
in electrochemical etching of silicon leaving on the surface Si-F terminated bonds which
can be immediately attacked in an autocatalytic process (fig 43)
In the electrochemical reactions described above an external potential is applied to the
silicon crystal The chemical and electrochemical reactions in any case take place
simultaneously most of the time with the chemical path predominating at high pH Even
when no external potential is applied to the silicon partial electrochemical reactions can
take place at different ldquocathodicrdquo and ldquoanodicrdquo sites on the surface with an internal charge
exchange which ensures neutrality [11] This macroscopic silicon etching and hydrogen
bubble formation can lead to visually observable patterns on the surface when Si(100) is
immersed in ammonium fluoride even without applying a potential
Si Si
Si Si
H
H
H
F Si
Si
Si SiH
F F-H+ -2e
+HF2-
H
HF
SiSi
SiSi + F
F F
FH
H
H
F -H+ -2e
+HF2-
Fig 43 Autocatalytic electrochemical etching of silicon by HF2-
Inhomogeneities on silicon surfaces caused by electrochemical reactions and charge
transfer have been studied extensively because they are of fundamental importance in the
formation of porous silicon
Pore formation on silicon
When Si(100) or Si(111) are etched under anodic bias in fluorine-based solutions
microscopic pores form on their surface Several different morphologies of pores have
been observed with pore diameters ranging from 10 nm to several microns with depths of
several microns [14] Pore shape is very variable too ranging from ordered straight pores
to chaotic networks of branched pores (fig 44) Porous silicon has been known since the
fifties but it was only in 1990 that interest in this material increased following the
discovery that porous silicon layers were able to emit bright red light This led to a large
amount of research and now different classes of micropores can be reproducibly created
36 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Fig 44 Different types of Silicon micropores From ref [14]
mostly for optical and micromachining applications However there is still no unified
theory able to explain the nucleation and growth mechanism of all the different kinds of
pores
We will give a short description of some of these theories for more detail see Parkhutik et
al [15]
One model explains pore nucleation on the basis of physical processes such as hole
positive charge migration ion transport to the surface and small perturbations on the
silicon surface modelled as Fourier components The system is shown to be unstable and
some spatial frequencies that lead to pore nucleation evolve from the etching process
A second model focusses on stationary pore growth without explaining the nucleation
stage According to this model silicon dissolves preferentially at pore edges because h+
charges are attracted by the stronger electric field present at these edges
A third class of models explains pore growth as a Diffusion Limited Aggregation (DLA)
process where the random walk of h+ charge carriers through the depleted layer present at
the silicon-liquid interface controls pore shape
Finally the model by Carstensen Cristophersen and Foll [16] proposes that areas of the
surface of some characteristic size LCO are etched by synchronized ldquocurrent burstsrdquo in the
flow of h+ charges These bursts dissolve silicon through cyclic stages of surface oxidation
oxide removal and hydrogen passivation Areas where a burst has recently taken place are
less passivated and thus more likely to be etched again in this way the pore bottom
continues to dissolve while the pore walls are passivated and are thus less favourable areas
towards current bursts
VPalermo 37
EXPERIMENTAL RESULTS
In the following sections we will show some experimental results obtained from STM and
AFM measurements of fluorine-treated Si(100) surfaces In the first part the results of
mild etching using concentrated and dilute HF solutions at low pH are presented In the
second part the results of etching at high pH using ammonium fluoride are presented and
the mechanism of pore formation discussed
Etching of Si(100) in dilute and concentrated HF
Samples were cut from different areas of an 8-inch diameter p-doped silicon(100) wafer
(10 Ω-cm) supplied by MEMC Electronic Materials Each series of STM measurements
was carried out over at least six different areas on at least two identical samples Low
Electron Energy Diffraction (LEED) was used to check the surface cleanliness of the
samples before STM measurements
Table 41 summarizes the different treatments of each sample After etching with
electronic grade HF each sample underwent a final rinse in Ultra-Pure Water (UPW
resistivity gt18 MΩ-cm) Both the HF and the UPW were allowed to flow continuously
over the sample surface Some samples were not etched with HF at all but just washed with
UPW to observe the morphology of the native oxide layer (~2 nm thick) covering the
surface All of these processes were carried out under nitrogen to limit reoxidation and the
samples were then introduced from the nitrogen atmosphere directly into the vacuum
chamber and degassed overnight at sim150degC before LEED and STM measurements
STM images were obtained from each sample using the same measurement parameters
(sample bias 4 V feedback current 1 nA scan speed 800 nm s-1) These parameters
38 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
provided a satisfactory level of reproducibility for all the samples Measurements were
made over an area of 500x500 nm (image size 500x500 pixel) Slope correction was
carried out by subtracting row-wise and column-wise fitted slopes from the entire image
which gave better results than the simple subtraction of a fitted plane especially for the
rougher samples Following slope correction the rms roughness
sum minus=xy
hyxhN
22 ))((1σ
and the 2-D Fourier transform
)(22
2
)(4
)( vyuxi
xyeyxhvuF +sum∆
= π
π
were calculated for each image where N2 is the number of pixels composing the image
h(xy) is the surface height at each point ∆ is the distance between points h is the mean
height and u v are the spatial frequencies The radial power spectrum PS(f) of the STM image data is obtained from the angular
average of the squared Fourier transform with f 2 = u2 + v2
Fig 45 shows the STM images obtained from the various samples Sample A still covered
with its native oxide layer shows an irregular surface with RMS roughness of ~05 nm
(see Table 41) Observing the sample with LEED gave no diffraction pattern even at
relatively high incident electron energies because of the surface oxide coverage After 1
min etching in dilute HF (sample B) the morphology is similar to the original one though
Table 41 Sample treatments average RMS roughness and integrated area of power spectra
Sample Treatment RMS roughness (nm)
PS area (f lt 01 nm-1)
PS area (f gt 01 nm-1)
A Rapid dip in water 053 plusmn 013 366 032
B 1 min in HF 5 + 10 min in water 051 plusmn 008 272 034
C 30 min in HF 5 + 10 min in water 062 plusmn 008 1064 038
D 5 sec in HF 49 + 10 min in water 042 plusmn 004 183 022
VPalermo 39
Fig 45 STM images of each group of samples showing the topography of the silicon surface A) noetching original oxide surface B) after 1 min etching in dilute HF C) after 30 min etching in diluteHF D) after dipping in concentrated HF Grey scale indicates height of the surface from lower (black) to higher (white) The images are 250x250 nm ie representative portions of the images usedfor the roughness measurement and PSD analysis
some of the larger features have disappeared and the image quality is better maybe due to
improved tunnelling due to the cleaner surface The RMS roughness is comparable to that
of the original surface Clear diffraction patterns are visible using LEED though at quite
high energies (200 eV) After prolonged etching (sample C) the RMS roughness increases
to 062 nm and a long-range corrugation is visible on the surface even if the LEED pattern
is good
The samples dipped in concentrated HF (D) reveal the bare SiSiO2 interface which has a
disordered aspect and protrusions over a wide range of dimensions The quality of the
STM images of sample D is very good probably due to the cleanliness of the surface
40 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 E -0 2
1 E -0 1
1 E + 0 0
1 E + 0 1
1 E + 0 2
1 E + 0 3
0 0 0 0 0 1 0 1 0 1 0 0f (1 nm )
nm^4
A a s re c e iv e d
B e tc h e d 1 min
C e tc h e d 3 0 min
D e tc h e d HF 4 9
Fig 46 Log-Log plot of the averaged power spectra of the STM images for all the samples
which gives a more stable tunnelling junction The LEED pattern is excellent showing
clear diffraction peaks at energies as low as 37 eV comparable to that obtained after high
temperature cleaning in UHV
Fig 46 shows the power spectra of the samples The high frequency and low frequency
areas of the power specturm are considered separately Table 41 shows for each sample
together with the roughness the integrated area of the power spectrum for the high and low
frequency part
We first examine the differences between the samples in the low frequency part of the
spectrum (lt01 nm-1) In this spatial range the short dip in concentrated or dilute HF
(samples BD) seems to lower the roughness of the sample removing some of the larger
features of the native oxide surface visible in Fig 45a and thus decreasing the
corresponding part of the power spectrum Sample C however etched with HF for 30 min
shows an increase in roughness peaking at 0012 nm-1 (~85 nm corresponding to the
typical dimensions of the corrugations visible in Fig 45c) On this scale sample C has a
PS density of 435 nm4 compared to 101 nm4 for sample A
Samples A B and C have the same PS in the high frequency range showing that the fine
structure of the surface is not changed by the HF etching Sample D on the other hand
shows a decrease of roughness for frequencies gt 015 nm-1 This effect could be due to the
VPalermo 41
improved surface cleanliness after etching with concentrated HF which would give a more
stable STM junction thus reducing the high frequency noise in the image
Fig 45 and the analysis of the power spectrum of each sample shows that a rapid dip in
HF removes the native oxide but does not lead to major changes in the morphology of the
Si surface its only effect being the removal of some of the larger features present on the
original surface Prolonged etching on the other hand increases the RMS roughness of the
surface
Etching of Si(100) in concentrated ammonium fluoride and nano-hole creation
Two different types of commercial p-doped Si(100) wafers (2Ω-cm and 10Ω-cm) from
MEMC were used Several different samples of 10x5 mm were immersed for 10 minutes
in 40 electronic grade NH4F solution under agitation Previous works used low
temperatures or anodic potentials applied to the silicon to avoid gaseous hydrogen
production and to obtain a uniform surface but in our experiment we worked at room
temperature to check the influence of hydrogen bubbles on surface morphology During
the etching the stirring was sufficient to provide a uniform concentration of reagents over
the whole sample surface but not to mechanically remove the hydrogen bubbles from the
silicon surface
After the etching each sample was rinsed in ultra-pure water to remove any etching
residues and observed with STM AFM and optical microscopy The AFM measurements
were made in air while for STM measurements the samples were rapidly dried with
nitrogen and inserted into the vacuum system to avoid surface reoxidation After insertion
into the vacuum surface cleanliness was checked with LEED and the surface morphology
observed by STM Parameters for STM measurements were sample bias +4 V 1 nA
current 16 Hz scan rate The images obtained were stable and reproducible over several
days Scan parameters for AFM were 20 nN force and 1 Hz scan rate
Some of the samples were cleaned with an RCA standard clean [1] before NH4F etching to
check the influence of possible surface contaminants on the final results Eight different
samples were prepared and more than sixty STM images of the samples were taken at
different points of the various samples
42 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
(110)
(110)
Fig47 abc) STMtopographic images of differentetching morphologies Eachimage is 500x500 nm Z-ranges are 10 10 and 18 nmrespectively d) STM image of a bridgecreated by etching of the lowerlayers of silicon (black arrow)Image is 250x250x6 nm
After ~2 min of immersion in the solution hydrogen bubbles become visible on the sample
surface The production is slow and the bubbles are quite stable on the sample without
detaching Thus some areas of the surface are masked from the liquid etching action
STM observations (fig 47) show that at the nanometer scale the surface is unevenly
covered with holes of radii ranging from 10 to 200 nm with depths of 2-4 nm These holes
have a wide range of different shapes and distributions In most cases the surface was
covered with a uniform distribution of round-shaped holes (fig47a) indicating isotropic
etching The dimension and the density of the holes changed greatly from sample to
sample and even over the surface of a single sample In some cases the etching was
anisotropic yielding nearly square holes and layered structures as shown in fig 47b
Square holes have been previously observed in cases where the etching speed in the (110)
direction is significantly smaller than in the (100) direction [17]
Over these areas (fig 47b and especially 47c) it is clearly visible how once the surface
had been attacked the reaction continued to preferentially remove atoms at step
irregularities (kink atoms) straightening step edges Eventually the exposed underlying
silicon was also attacked and further holes created inside the previously etched larger
ones It was not possible to detect monatomic steps on this kind of surface The smallest
step height observed was ~15 nm corresponding to several atomic layers In the image
shown in Figure 47a the etching was not strong and created only anisotropic holes on the
surface In fig 47b and c the stronger etch proceeded laterally for several tens of
VPalermo 43
nanometres leaving straight steps several tens of nanometers long and roughly rectangular
holes as expected given the structure of the (100) crystal face In some cases a
significative underetch is observed and the formation of suspended bridges and tunnels
can be deduced from the STM images (fig 47d)
The formation of branched pores and suspended structures has been attributed during pore
formation to diffusion limited aggregation effects where the h+ charge carriers necessary
for silicon etching have a higher probability of reacting at pore bottoms than reaching the
upper part of the silicon surface In the case of very deep pores quantum wire effects have
been invoked to explain the pore growth mechanism [15] In our case though the pores
formed were very shallow the underetch depth being only a few nanometres on pores of
sim100 nm width Thus more than diffusion effects the main contribution to the
underetching process must come from anisotropic etching and some kind of autocatalytic
reaction path analogous to the one described by Matsumura et al [4] with some areas of
the silicon surface hydrogenated and thus less vulnerable to etching
Pre-treatment with RCA cleaning has no effect on the final morphology and this seems to
exclude pore nucleation being caused by presence of metallic or organic surface
contaminants
The morphology and the distribution density of the pits was quite uniform over
microscopic areas of the sample but changes were observed over the millimetre scale This
suggests that etching intensity is influenced by some large-scale parameter
Large-area measurements made with AFM or with an optical microscope (fig 48)
showed that the inhomogeneity of surface etching can be correlated with the masking
action of the bubbles While the fluoride dissolved the silicon hydrogen bubbles formed
by the reaction covered some areas of the surface thus blocking the etching over that area
generating macroscopic steps at the bubble-liquid border As the reaction proceeded more
hydrogen accumulated and the bubble diameter increased producing in this way a circular
pattern of steps The increase in bubble diameter was not continuous with time otherwise a
uniform surface slope gradient would have been obtained The formation of this circular
ldquoetching staircaserdquo indicates that the bubble growth was stepwise the bubble accumulating
more and more hydrogen without enlarging across the surface until it relaxed increasing
its diameter stepwise and covering more silicon The circular structures in fig 48a are not
co-axial and their asymmetry could derive from physical processes due to stirring or
irregularities on the surface
44 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
a b Fig 48 a) optical micrograph of etching patterns on Si(100) created by NH4F 12x09 mm b) AFM image of the circles border xy range is 40x24 microm z-range is 30 nm
The step structure was not destroyed by the etching even after the bubble detached from
the surface but on the contrary the etching process seemed to be influenced by the
presence of the step
Observing in detail a series of steps (fig 48b) a quite deep trench is visible at the base of
each step A close-up image of a step and the corresponding line profile of the trench is
shown in fig 49 The trench is sim5 nm deep with respect to the lower surface compared to
a step height of 22 nm
A similar structure has been recently obtained with electrochemical etching of p-type
Si(100) in 4 HF [18] in which a ldquocurrent burstrdquo etching model previously described
was assumed for silicon dissolution In that experiment the trench was created at the
border of silicon nitride masks and began to grow after a nucleation stage Preferential
trench etching was along the (110) direction and the trench growth was explained as an
effect of mechanical stress induced by the nitride mask and of electric field enhanced
dissolution which depended upon an external applied potential
While it is clear that in our system the gas bubbles have a masking effect similar to a
classical solid nitride mask it is unlikely that hydrogen present on the surface can induce a
significant stress in the silicon lattice as in the case of a nitride mask Furthermore no
external field was applied to drive preferential etching at the trench site
It has been proposed [19] that the cathodic and anodic part of the etching reaction
(hydrogen production and silicon oxidative etching respectively) take place at different
points on the surface with a net charge transfer between the different areas In this case
the highest reaction rates will correspond to the silicon area surrounding the bubble border
where a high number of positive charges will be available for the reaction Furthermore a
sharp trench extending into the silicon crystal will be a preferential electrostatic attractor
VPalermo 45
Fig 49 AFM image of the etched surface showing a step created on the surface by bubblemasking A stronger etching action is visible on the right side of the step as well as theprotected area on the upper side of the step (indicated by the arrows) Image is 10x10 micromz-range is 30 nm The profile on the right is taken from the central area of the image
for the h+ charge carriers coming from other ldquocathodic areasrdquo of the sample either from
other regions on the surface or from the back of the silicon chip [18]
In the areas where the hydrogen bubble had detached and the surface was exposed to the
etching the reaction was not uniform in the neighbourhood of the steps It is possible to
observe (fig 48b and 49) an area on the upper side of the step where less or even no
etching at all seems to have taken place as if the step was able to protect the surface from
etching While etching on the lower side with trench formation can be attributed to the
presence of the bubble the surface on the upper step side can be etched only after bubble
detachment so no masking effect can account for this result However a further
preferential attraction of h+ charge carriers from the already formed trench can be
hypothised electrochemically shielding the surrounding area from further etching If this is
true the shielding effect would be very strong with a relatively shallow 5 nm-deep trench
protecting an area of sim1 microm parallel to the step
To summarise the etching of Si(100) in NH4F is a complex process in which different
reaction paths both chemical and electrochemical co-exist Hydrogen bubbles formed by
the reaction act as a mask on the surface and create etching paths and inhomogeneous
etching of the surface Different kinds of pores are observed on the surface and in some
cases the anisotropy of the process is so strong as to give square-shaped holes and
underetching
46 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The diffusion of h+ charge carriers in the crystal is one of the main rate-determining steps
of the reaction and leads to the formation of a deep trench immediately outside the bubble
perimeter These trenches act as charge collectors and reduce the etching of the upper step
surface in the proximity of the steps
Bibliography
[1] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[2] Spectroscopic Ellipsometry Studies Of HF Treated Si (100) Surfaces Yao H Woollam
Ja Alterovitz SA Applied Physics Letters 62 (25) 3324-3326 Jun 21 1993 Influence
Of HF-H2O2 Treatment On Si(100) And Si(111) Surfaces Graf D Bauermayer S
Schnegg A Journal Of Applied Physics 74 (3) 1679-1683 Aug 1 1993 Kinetics Of
Oxidation On Hydrogen-Terminated Si(100) And (111) Surfaces Stored In Air Miura
T Niwano M Shoji D Miyamoto N Journal Of Applied Physics 79 (8) 4373-4380
Part 1 Apr 15 1996
[3] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995 Structure Of The Stepped SiSiO2 Interface After Thermal-Oxidation -
Investigations With Scanning Tunneling Microscopy And Spot-Profile Analysis Of
Low-Energy Electron-Diffraction Pietsch GJ Kohler U Jusko O Henzler M Hahn
PO Applied Physics Letters 60 (11) 1321-1323 Mar 16 1992
[4] Enhanced Etching Rate Of Silicon In Fluoride Containing Solutions At pH 64
Matsumura M Fukidome H Journal Of The Electrochemical Society 143 (8) 2683-
2686 Aug 1996
[5] A Study Comparing Measurements Of Roughness Of Silicon And SiO2 Surfaces And
Interfaces Using Scanning Probe Microscopy And Neutron Reflectivity Crossley A
Sofield CJ Goff JP Lake ACI Hutchings MT Menelle A Journal Of Non-Crystalline
Solids 187 221-226 Jul 1995
VPalermo 47
[6] Step-Flow Mechanism Versus Pit Corrosion - Scanning-Tunneling Microscopy
Observations On Wet Etching Of Si(111) By Hf Solutions Hessel HE Feltz A Reiter
M Memmert U Behm RJ Chemical Physics Letters 186 (2-3) 275-280 Nov 8 1991
[7] Comparison Of Si(111) Surfaces Prepared Using Aqueous-Solutions Of NH4F Versus
HF Higashi GS Becker RS Chabal YJ Becker AJ Applied Physics Letters 58 (15)
1656-1658 Apr 15 1991
[8] Wet Chemical Etching Of Si(100) Surfaces In Concentrated NH4F Solution -
Formation Of (2x1)H Reconstructed Si(100) Terraces Versus (111) Faceting Neuwald
U Hessel HE Feltz A Memmert U Behm RJ Surface Science 296 (1) L8-L14 Oct
10 1993
[9] Ideal Hydrogen Termination Of Si(001) Surface By Wet-Chemical Preparation Morita
Y Tokumoto H Applied Physics Letters 67 (18) 2654-2656 Oct 30 1995
[10] Atomic Structures Of Hydrogen-Terminated Si(001) Surfaces After Wet Cleaning
By Scanning Tunneling Microscopy Endo K Arima K Kataoka T Oshikane Y Inoue
H Mori Y Applied Physics Letters 73 (13) 1853-1855 Sep 28 1998
[11] On The Potential-Dependent Etching Of Si(111) In Aqueous NH4F Solution
Houbertz R Memmert U Behm RJ Surface Science 396 (1-3) 198-211 Jan 20 1998
[12] Etching Process Of SiO2 By HF Molecules Hoshino T Nishioka Y Journal Of
Chemical Physics 111 (5) 2109-2114 Aug 1 1999
[13] Etching Mechanism And Atomic-Structure Of H-Si(111) Surfaces Prepared In
NH4F Allongue P Kieling V Gerischer H Electrochimica Acta 40 (10) 1353-1360
Jul 1995
[14] Pore Formation Mechanisms For The Si-HF System Carstensen J Christophersen
M Foll H Materials Science And Engineering B-Solid State Materials For Advanced
Technology 69 23-28 Sp Iss Si Jan 19 2000
[15] Porous Silicon - Mechanisms Of Growth And Applications Parkhutik V Solid-
State Electronics 43 (6) 1121-1141 Jun 1999
[16] Formation And Application Of Porous Silicon Foll H Christophersen M
Carstensen J Hasse G Materials Science amp Engineering R-Reports 39 (4) 93-141 Nov
1 2002
[17] Pore Morphology And The Mechanism Of Pore Formation In N-Type Silicon
Searson PC Macaulay JM Ross FM Journal Of Applied Physics 72 (1) 253-258 Jul 1
1992
48 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
[18] Deep Electrochemical Trench Etching With Organic Hydrofluoric Electrolytes
Christophersen M Merz P Quenzer J Carstensen J Foll H Sensors And Actuators A-
Physical 88 (3) 241-246 Mar 5 2001
[19] Morphology Of Anodically Etched Si(111) Surfaces - A Structural Comparison Of
NH4F Versus HF Etching Houbertz R Memmert U Behm RJ Journal Of Vacuum
Science amp Technology B 12 (6) 3145-3148 Nov-Dec 1994
VPalermo 49
5 Surface modification of silicon in vacuum void creation and oxide desorption
The main reason for the huge success of silicon in the microelectronics industry is not due
to its superior properties as a semiconductor Other materials for example germanium
have better qualities such as higher mobility of charge carriers and lower noise levels
which would allow the construction of faster and higher performance devices
The widespread use of silicon however is mainly due to the outstanding characteristics of
its oxide Silicon dioxide (SiO2) is a very good electrical insulator easy to form
chemically and thermally stable and is compatible with lithographic and metal deposition
processes Germanium oxide on the contrary is too reactive to be used
Even the use of Si(100) substrates for nearly all microelectronic devices is dictated by
oxide quality The (111) face of silicon crystal can be easily cleaved and flattened and
almost atomically perfect surfaces can be obtained with simple chemical procedures (as
described above) But the density of interfacial defects is highest for Si(111)-SiO2
interfaces and lowest for Si(100)-SiO2 ones so microchips will continue to be fabricated
on Si(100) wafers
SiO2 (silica) is present in 95 of the earthrsquos minerals in different allotropic forms such as
quartz tridymite and cristobalite In the bulk each silicon atom is bonded to four oxygens
in a Si-O-Si tri-dimensional network Si-O bonds are 016 nm long and form an angle
ranging from 120deg to 150deg
Three typical intrinsic defects are present in SiO2 The so-called Ersquo centres are oxygen
vacancies with a hole localised on a silicon atom with only three Si-O bonds
50 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
O3Si +SiO3 Whereas the PR (peroxy radical) defects are holes trapped by a charged
peroxy moiety with a O3Si-O-O+ SiO3 structure The NBOHC (non-bridging oxygen hole
centres) derive from water or hydrogen contamination and are schematized as O3Si- O- H-
O-SiO3
The atomic structure of the Si-SiO2 interface varies enormously Local domains resembling
the tridymite and the cristobalite structure of silica are present but it seems that only 10
of the interface is ordered [1] Far from the interface the SiO2 bulk is completely
disordered The passage from bulk Si to stoichiometric SiO2 passes through a non-
stoichiometric SiOx layer 07 nm thick
When a clean silicon surface is exposed to atmospheric oxygen a thin sim2 nm thick layer
of native oxide forms spontaneously which is usually removed and substituted with
thicker better quality oxide layers before further processing
Silicon is usually oxidised by thermal annealing at temperatures between 800deg and 1100deg
in an atmosphere of pure O2 with some water eventually added to increase oxidation speed
Thermal oxides made in pure oxygen (dry oxides) grow more slowly than oxides produced
in an oxygen-water atmosphere (wet oxides) but are usually of better quality
According to the Deal-Groove formula the time t needed to grow an oxide of thickness X
is given by [1]
1
212minus
minusminus
+=
ABXBXt α
where the constant B and BA decrease exponentially with temperature as
minus
kTEexp
with activation energies for dry oxidation of EB =123 and EBA =20 eV respectively
EB is related to the diffusion of oxygen in silicon while the value of EBA is interpreted as
the energy required to break a Si-Si bond The exponent α is 1 for wet oxidation and 0 for
oxidation at high temperatures and low oxygen pressures It has intermediate values for dry
oxidation This formula does not work well for low values of X at the initial stages of
oxidation and usually empirical corrections are used
An interesting characteristic of silicon is that at high temperature and in vacuum oxygen
can actually etch the silicon crystal giving gaseous products and the oxidized layer present
on the crystal becomes unstable (fig 51) [2]
VPalermo 51
1E-10
1E-09
1E-08
1E-07
1E-06
1E-05
1E-04
1E-03
1E-02
1E-01
0607080911112
1000T (1K)
P (T
orr)
SiO2 + Si rarr 2SiO(g) Oxide decomposition
Si+O2 rarr SiO2(s) oxide formation Fig51 phase diagram of the
oxygen-silicon system
Silicon oxidation apart from the initial nucleation stages at the monolayer level proceeds
uniformly over the whole surface with a planar reaction front moving from the surface
into the bulk
If heated under low oxygen partial pressure (vacuum or inert atmosphere) SiO2 is known to
decompose following the reaction
SiO2 + Si rarr 2SiOuarr (1)
The reaction begins with nucleation at defect points on the SiSiO2 interface and proceeds
in a spatially inhomogeneous manner with the formation of large voids on the oxide
surface [3]
Several studies have been made on the dynamics of void growth both on thick [4] and thin
[5] layers of SiO2 The process has been used to decorate otherwise unobservable defects at
the SiSiO2 interface [6] or to grow nanoislands of silicon on the void surface [7] It has
been suggested that the defects acting as nucleation centres could be metallic contaminants
present on the native surface which aggregate and catalyze SiO2 decomposition [8]
After oxide desorption the silicon surface is very rough In particular on Si(100) square
islands are observed several nanometres high which act as lsquopinning sitesrsquo on the motion of
monatomic steps on the Si surface It has been proposed that these islands can be composed
of silicon carbide coming from the organic contaminants present on the surface [9] or of
oxide clusters during partial reoxidation [10]
In the following section some experimental results obtained studying voids and nanoisland
growth are presented
52 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nanoisland growth on silicon
We used two different samples the first rougher than the second (i) p-type Czochralski Si
(100) with a RMS roughness of 05 nm shown in fig52a and (ii) p-type epitaxial Si(100)
with a RMS roughness of 019 nm shown in fig 52b The roughness was measured with
STM on the lsquoas receivedrsquo samples on different areas of 500x500 nm Every sample was
covered by a layer of native oxide ~2 nm thick
Some of the samples (AC) were introduced into the UHV system without any cleaning
others (BD) were dipped in concentrated HF (49) to remove the surface oxide without
etching the silicon [11]
After degassing the samples were heated resistively increasing the temperature slowly to
keep the pressure within the 10-10 mbar range during heating Each sample was held at
900˚C for 30 min For some of the samples the heating was stopped at lower temperatures
to obtain incomplete oxide desorption and surface roughening thus enabling observation
of the different steps of the process Table 51 summarizes the different treatments for each
group of samples
Fig52 Original surfaces of Czochralski (a) and epitaxial (b) silicon Every image is 500x500 nm
VPalermo 53
Table 51 Summary of different treatements
Silicon type Oxide removed
Heating Islands density (microm-2)
Av Island Volume (nm3)
A Czochralski No 30rsquo at 900˚C 24x103 353
B Czochralski Yes ldquo 20x104 71
C Epitaxial No ldquo 28x102 2450
D Epitaxial Yes ldquo 23x104 63
Fig53 ab) Surface roughening on group A samples after heating 30 min at 800deg and900deg respectively in presence of an oxide layer Image size 500x500 nm cd) Surface roughening on group B samples after heating 30 min at 700deg and 900degrespectively after removing the oxide layer with HF Image size 250x250 nm
Fig 52a is the native oxide surface of sample A quite disordered and irregular This
surface was stable when heated to 700˚C while at higher temperatures the oxide began to
desorb Fig 53a was taken after heating the sample at 800˚C for 30 min
54 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The oxide began to desorb in correspondence with defect points creating oxide-free
surface voids which enlarge radially uncovering the silicon surface Some surface silicon
atoms are removed through reaction (1) Mobile silicon atoms moving randomly over the
surface aggregate forming a nanocrystal at the initial defect point surrounded by a stable
flat oxide-free surface with some steps The initial SiSiO2 interface as observed after
oxide etching with HF [12] is similar to the one shown in fig 52a and thus the surface
observed in the void area is a completely new one coming from surface reorganization
The reaction continues leading to a coalescence of the voids the final result of the process
is shown in fig 53b with protruding islands aligned along the main crystal axes and flat
areas with some steps LEED measurements confirmed that this surface is crystalline
silicon Atomic resolution was obtained on the flat areas of the surface showing the
typical 2x1 reconstruction of Si(100) Figures 53c and 53d show the evolution of the
roughening on sample B after chemically removing the native oxide layer before
introducing the sample into UHV As mentioned above the initial morphology is similar to
that shown in fig 52a though the oxide has been removed as confirmed by LEED
measurements Fig 53c shows the surface after heating at 700˚ for 30 min Without the
oxide layer the surface began to reorganize at lower temperatures and the small irregular
grains of the original surface begin to grow and become more rectangular After heating to
900˚C (fig 53d) the surface is flat with small rectangular islands The shape of the islands
is similar to that shown in fig 53b but the lateral island dimension is one order of
magnitude smaller
To assess the importance of the initial surface on the process the same treatments were
repeated starting from flat epitaxial silicon (samples CD) The initial epitaxial surface is
smooth with periodic steps due to a small miscut angle (fig52b) On heating the epitaxial
silicon without etching (sample C) relatively large islands are obtained similar to the ones
shown in fig 53b where the ldquostep-pinningrdquo effect of the islands is evident The average
dimension of sample C islands was sim2500 nm3 while only 353 nm3 for sample A Island
density decreased from 24000 to 280 microm-2 This result is maybe due to a lower
concentration of defects in the oxide layer of the epitaxial silicon which allows the
expanding void to grow much larger before coalescence with neighbouring voids the
nanostructures within them thus reaching greater dimensions To verify this hypothesis we
applied the same treatment to clean not-so clean and purposely contaminated samples To
check that the defects inducing void nucleation and nanoisland growth derive from carbon
VPalermo 55
contamination we used electronic-grade methanol to contaminate the silicon surface
Figure 54 shows the results of this comparison
State-of-the-art cleaned samples obtained with multi-step RCA cleaning [13] showed the
development of very few voids the oxide desorption leaving large smooth areas of oxide-
and nanostructure-free silicon as shown in Fig 54a where the presence of residual oxide
can be seen Samples cleaned with a simpler one-stage Pirana cleaning solution showed a
higher number of islands per unit area (Fig 54b) and samples purposely contaminated
with methanol (Fig 54c) had the highest density of island nucleation of all three samples
When the native oxide had been chemically removed from the epitaxial silicon before
heating (sample D) the island size and density is similar to the standard as-received
silicon case (sample B) Thus starting from a flat substrate the process of surface
roughening and island creation was the same even when the oxide layer had been
previously removed
From these results some important indications can be drawn The first is that oxide
desorption temperature depends upon the contamination levels of the oxide layers (fig 54)
with desorption at lower temperatures for contaminated surfaces
Second once the oxide layer is removed the silicon atoms become mobile on the surface
at relatively low temperatures (~700˚C) Their initial configuration is disordered and has a
high surface energy so they crystallize in the small islands shown in fig53d reducing
their surface area and lowering the energy of the system When the oxide is present the
surface cannot reorganize at 700˚C and is stable At 800˚C the oxide begins to desorb in
correspondence with defect points on the surface As the void area enlarges a small
cba Fig 54 The effect of contamination on nano-island production Gradient-filtered STM images of (a)sample cleaned two times with RCA 400x400 nm (b) Sample cleaned with a Pirana solution 500x500nm (c) sample contaminated with MeOH before insertion into the UHV system 500x500 nm Annealingtime is 40 min at 800degC for all the samples
56 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
crystallite grows at the centre of the voids Further oxide desorption uncovers larger areas
freeing more silicon atoms which feed the initial island making it bigger
Fig 55 shows the scaled size distribution of the islands for each group of samples Every
distribution has been obtained using several images of different samples for each group
The size distribution of all the samples follows an exponential decay and the decay seems
the same for all the samples though the average volume of the islands differs by more than
one order of magnitude (see Table 51 for details)
The point defects which catalyze void nucleation can be metallic contaminants present on
the initial oxide surface [8] or organic contaminants which at high temperatures can
form SiC nanocrystals [14] We used Scanning Tunneling Spectroscopy (STS) to look for
differences between the islands and the surrounding flat silicon surface but no difference
was found Furthermore the total island volume per surface unit is very high (more than
104 nm3microm2) and it seems unlikely that such a huge volume could consist of surface
contaminants Islands of pure silicon on silicon have been grown without evidence of
surface contamination[14]
To summarise the formation of nano-sized islands on silicon through surface diffusion
was studied At high temperatures the oxide layer covering the surface decomposes non
uniformly and circular voids of clean silicon are created The presence of the oxide layer
blocks surface silicon atomic motion and surface reorganization except within the voids
causing the growth of islands more than 10 nm high and 30 nm wide on the silicon surface
According to STM and LEED measurements we can say that the islands are mostly
Fig 55 Size distribution (scaled) for the samples described in table 1 The line is a reference for the eye
VPalermo 57
composed of silicon with traces of other substances such as carbon Final island densities
and dimensions depend upon the initial purity of the oxide layer
On the other hand if the oxide is removed chemically before heating the island growth
process is not so localised and takes place simultaneously over the whole surface yielding
smaller and more numerous islands
58 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[2] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[3] High-Temperature SiO2 Decomposition At The Sio2Si Interface Tromp R Rubloff
GW Balk P Legoues FK Physical Review Letters 55 2332-2335 Nov 1985
[4] Defect Microchemistry At The SiO2Si Interface Rubloff GW Hofmann K Liher M
Young DR Physical Review Letters 582379-2382 Jun 1987 Defect Formation In
Thermal SiO2 By High Temperature Annealing Hofmann K Rubloff GW Mccorkle
RA Applied Physics Letters 49 1525-1527 Dec 1986 Kinetics Of High-Temperature
Thermal Decomposition Of SiO2 On Si(100) Liher M Lewis JE Rubloff GW Journal
Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 5 1559-1562 Aug
1987
[5] Thermal-Decomposition Of Very Thin Oxide Layers On Si(111) Kobayashi Y Sugii
K Journal Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 10 (4)
2308-2313 Part 3 Jul-Aug 1992 Scanning Tunneling Microscope Study On Mid-
Desorption Stages Of Native Oxides On Si(111) Kobayashi Y Sugii K Journal Of
Vacuum Science amp Technology B 9 (2) 748-751 Part 2 Mar-Apr 1991 Controlled
Growth Of SiO2 Tunnel Barrier And Crystalline Si Quantum Wells For Si Resonant
Tunneling Diodes Wei Y Wallace RM Seabaugh AC Journal Of Applied Physics 81
(9) 6415-6424 May 1 1997
[6] Defect Formation In SiO2Si(100) By Metal Diffusion And Reaction Liher M
Dallaporta H Lewis Je Appl Phys Lett 53 589-591 Aug 1988 SiO2 Film
Decomposition Reaction Initiated By Carbon Impurities Located At A Si- SiO2
Interface Raider Si Herd Sr Walkup Re Applied Physics Letters 59 (19) 2424-2426
Nov 4 1991
[7] Nanometer-Scale Si Selective Epitaxial Growth On Si(001) Surfaces Using The
Thermal Decomposition Of Ultrathin Oxide Films Fujita K Watanabe H Ichikawa M
Applied Physics Letters 70 (21) 2807-2809 May 26 1997 Pyramidal Si Nanocrystals
VPalermo 59
With A Quasiequilibrium Shape Selectively Grown On Si(001) Windows In Ultrathin
SiO2 Films Shibata M Nitta Y Fujita K Ichikawa M Physical Review B 61 (11)
7499-7504 Mar 15 2000
[8] Stacking-Fault-Induced Defect Creation In SiO2 On Si(100) Liher M Bronner Gb
Lewis Je Appl Phys Lett 52 1982-1985 May 1988
[9] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990 AFM And XPS Characterization Of The
Si(111) Surface After Thermal-Treatment Lamontagne B Guay D Roy D Sporken R
Caudano R Applied Surface Science 90 (4) 481-487 Dec 1995
[10] Scanning-Tunneling-Microscopy Study Of Oxide Nucleation And Oxidation-
Induced Roughening At Elevated-Temperatures On The Si(001)-(2x1) Surface Seiple
JV Pelz JP Physical Review Letters 73 (7) 999-1002 Aug 15 1994 Evolution Of
Atomic-Scale Roughening On Si(001)-(2x1) Surfaces Resulting From High-
Temperature Oxidation Seiple JV Pelz JP Journal Of Vacuum Science amp Technology
A-Vacuum Surfaces And Films 13 (3) 772-776 Part 1 May-Jun 1995
[11] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995
[12] Morphological Changes Of The Si [100] Surface After Treatment With
Concentrated And Diluted HF Palermo V Jones D Materials Science In
Semiconductor Processing 4 (5) 437-441 Oct 2001
[13] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[14] Nanoscale Roughening Of Si(001) By Oxide Desorption In Ultrahigh Vacuum
Gray SM Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology
B 14 (2) 1043-1047 Mar-Apr 1996
60 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 61
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and nano-lines
When the silicon surface is not protected by a native oxide layer or by a hydrogenated
passivating layer it is very reactive The surface chemistry of both Si(100) and Si(111) has
been extensively studied starting from clean surfaces prepared in UHV because of its great
relevance to the microelectronics industry and its technological interest The diffusion and
reaction of molecules and atoms on silicon is an interesting scientific problem on its own
apart from technological issues because silicon reconstruction yields a very complex and
anisotropic surface
We give here a brief summary of the adsorption behaviour of several elements on Si
surfaces The summary is not comprehensive and only the most interesting characteristics
for each substance are given
Table 61 Summary of adsorption behaviour of atoms and molecules on silicon surfaces [12]
Hydrogen
Molecular H2 shows low reactivity towards silicon while atomic hydrogen easily forms
Si-H bonds and can even break Si-Si bonds
Alkali metals
Alkali metals diffuse rapidly into SiO2 and can damage silicon-based transistors
Transition
metals
All transition metals apart from gold and silver react with Si forming metal silicides
62 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nickel A very common contaminant of silicon Nickel from even slight contact with stainless
steel tweezers can contaminate silicon samples forming its silicide and is very difficult
to remove even at high temperatures because it segregates on the silicon surface
Palladium Forms silicides especially Pd2Si and can be used to make contacts possessing a low
Schottky barrier
Titanium Widely used to fabricate contacts for silicon-based microdevices TiSi2 is one of the
more stable and highly conductive silicides
Tungsten The silicide is more stable than TiSi2 on polysilicon and is thus used for polysilicon
gate contacts
Platinum Silicide is used for bipolar transistors PtSi contacts on n-silicon give Schottky diodes
Cobalt Cobalt silicide gives better quality contacts than TiSi2 at the nanometrelevel but is less
used at the moment because it consumes too much silicon during its growth
Iron Silicides have been studied for potential optoelectronic applications because FeSi2 has a
directly accessible bandgap and is optically active
Group 13
Group 13 metals are used in IC technology to dope silicon (giving p-type doping) and
to make III-V type semiconductors When evaporated on Si(100) or Si(111) they react
strongly with the surface giving surface reconstruction and kink formation at steps
Boron Widely used for doping Usually deposited with decomposition of boron hydrides
Sticking coefficient of simple hydrides is very low so decaboranes (B10H14) are used for
deposition
Aluminum Aluminum is widely used for microelectronic contacts but in some cases it reacts with
silicon diffusing into the crystal and with SiO2 stealing oxygen atoms to form Al2O3
Group 14
Carbon Carbon can form a composite semiconductor with silicon (silicon carbide) widely used
to fabricate integrated circuits for use at elevated temperatures or in the presence of
ionising radiation When carbon is deposited on silicon usually polycrystalline films
with properties between diamond and graphite are obtained
CO adsorbs molecularly on silicon without breakage of the C=O bond Other
unsaturated hydrocarbons do not react with silicon Even very reactive strained
molecules like cyclopropane have low sticking coefficient Ethylene acetylene and
propylene adsorb molecularly on Si(100) each molecule sitting on a silicon dimer
interacting with the dangling bonds of the surface Acetylene also adsorbs molecularly
on Si(111) on the 7x7 reconstructed surface preferential adsorption on centre atoms
with respect to corner atoms is observed Benzene and other aromatic ring compounds
adsorb molecularly on the surface They can form σ or π-type bonds with the silicon
laying parallel or tilted respect to the surface according to the bond type
Silicon Silicon atoms can be deposited on the surface with molecular beam epitaxy or
VPalermo 63
decomposition of silanes and chlorosilanes Silanes with single Si-Si bonds have a high
sticking probability and dissociate upon adsorption giving trihydrides and
monohydrides
Chlorosilanes have high sticking coefficients too and decompose on the surface but
they can etch the surface through the reaction SiCl4 + Si rarr 2SiCl2
Germanium Germanium has a lattice constant similar to silicon (Ge lc is 4 larger than Si) so that
various alloys of Si1-xGex can be formed The growth of Germanium on silicon is quite
peculiar with several monolayers adsorbing uniformly on the surface followed by
island formation (Stranski-Krastanov growth)
Group 15
Nitrogen Exposure to ammonia and nitrogen gas at high temperatures leads to the formation of
silicon nitride layers (Si3N4) NH3 dissociates on the surface and reacts with the
dangling bonds giving Si-NH2 and Si-H termination Silicon nitride is stable on the
surface up to 1100degC above this temperature it desorbs as Si2N
Phosphorous Phosphine (PH3) adsorbs dissociatively as Si-PH2 and Si-H and behaviour is similar to
that of ammonia
Arsenic A monolayer of As is often deposited on silicon as a substrate for GaAs growth Arsenic
forms dimers on Si(100) creating dimer rows which are parallel (perpendicular) to the
underlying silicon dimer rows when it is deposited at high (low) temperature On
Si(111) Arsenic breaks the 7x7 reconstruction giving a 1x1 pattern This 1x1 As
monolayer acts as a passivating layer and can resist further As adsorption oxygen and
air
Antimony and
Bismuth
Due to their large covalent radii Sb and Bi form only short dimer rows on silicon and
only at high temperatures
Group 16
Oxygen
Oxygen can oxidize or etch silicon according to the reactions
Si(s) + O2 rarr SiO2 (s)
Si(s) + frac12O2 rarr SiO(g)uarr
Silicon oxidation has already been described in the previous chapter Molecular oxygen
adsorption on silicon is mostly dissociative O atoms break the Si-Si bonds forming an
Si-O-Si bridge or a peroxy bridge Si-O-O-Si Hydrided silicon surfaces do not have
surface dangling bonds and are thus more protected from oxygen attack Etching is
favoured at high temperatures and low oxygen pressures [3] At T gt700degC silicon oxide
on silicon decomposes as SiO(g)
Water
H2O adsorbs dissociatively on silicon easily on Si (100) and with more difficulty on
Si(111) The molecule decomposes giving Si-OH and Si-H on adjacent sites
64 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Group 17
Fluorine
Fluorine adsorbs and reacts easily on silicon dissociating on the surface and forming
Si-F bonds It can then penetrate the surface and attack polarised Si-Si backbonds The
surface is etched with production of SiF4 and SiF2
Chlorine
Chlorine reacts aggressively with silicon etching the surface Cl2 and F2 are often used
commercially to etch silicon usually in a plasma Atomic Cl and Cl2 adsorb on Si(100)
up to saturation The most stable configuration seems to be a Cl atom bonded to each Si
atom of a surface dimer in a symmetric structure but metastable configurations with
two Cl atoms bonded to a buckled dimer and even a central Cl atom bridged across a
dimer have been observed
On Si(111) silicon mono- di- and tri-chlorides are formed In particular heating at
Tgt400degC a uniform Si-Cl monolayer is obtained and the 7x7 reconstruction changes
into a bulk like 1x1 lattice The surface transition induced by Cl adsorption is fully
reversible The 7x7 reconstruction can be restored on desorbing the chlorine by heating
at Tgt1100degC The 1x1 domains nucleate at the lower terrace side of steps the 7x7 at the
upper step edge
Bromine
Like hydrogen bromine maintains the 2x1 structure of Si(100) forming Si-Br bonds
with the dangling bonds of the silicon dimer rows At high exposure some etching of
silicon by formation of volatile SiBr3 species has been observed
Surface diffusion on silicon
In the previous chapter we observed nano-island growth on silicon surfaces with native
oxide and surfaces contaminated with organic impurities While this phenomenon has been
observed in several experiments the formation and growth mechanisms of nanoislands has
still not been satisfactorily described theoretically Various substances have been
hypothesised as nanoisland nucleating agents such as organic or metallic contaminants
present on the surface or oxide clusters (see previous chapter) The island growth process
has been attributed by various workers to lsquosome sort of kinetic instabilityrsquo [4] to the
pinning of step flow by SiC clusters [5] or to a mesoscopic atom flux from areas with low
step density to areas of higher step density [6]
One mechanism proposed to explain surface roughening and island growth is the Ehrlich-
Schwoebel effect ie the presence of an energetic barrier that reflects atoms approaching a
VPalermo 65
Molecule (L
O2
H2O MeOH
CO CO2 CH4
clean surface
downward
weak on si
The aim o
surface an
To obtain
the ones
spontaneou
first obser
surface W
nanoisland
Experime
We used S
removed f
41 mixtur
water and
whole hea
1200degC re
kept below
Table 62 Treatment island density and island volume for each molecule tested
Dose angmuir)
Heating time at 800degC (min)
Island density (microm-2)
Ave Volume (nm3)
Equivalent layer thickness (nm)
280 10 no islands - -
90 10 no islands - - 30 10 2000 19 004
180 10 360 516 019 30 60 no islands - -
180 10 no islands - - - 10 no islands - -
step [7] Recent experiments however indicate that the Schwoebel effect is
licon [48]
f our experiments was to study the reaction of simple molecules with the silicon
d to check the ability of these molecules to nucleate nanoisland growth
a better understanding of the process we chose simple molecules starting from
which are more likely to be present as traces in UHV chambers where
s nanoisland growth is often observed after sample heating Using STM we
ved at the atomic level the possible interactions of these molecules with the
e then heated the surface and checked the density and dimensions of the
s produced
ntal procedures and results
i(111) wafers p-doped 0015 Ω cm Gross contamination and particles were
rom the sample surface with a standard Pirana etch (15 min dip in a H2S4H2O2
e at T=80degC) After this the samples were thoroughly washed in ultra-pure
introduced into the UHV system After degassing for 5 hours at 600degC the
ting stage was allowed to cool down and a sequence of rapid flashes to 1100deg-
moved the native oxide layer and cleaned the surface Pressure during flashes is
1x10-9 mbar
66 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Surface flatness and cleanliness were checked with STM and the contaminant molecules
introduced into the UHV system through a leak valve Doses of several Langmuirs were
used for the experiments (1 Langmuir=10-6 torr sec-1) Pressure was kept constant at 10-7
mbar during adsorption The STM tip was retracted during gas deposition to avoid any tip
shielding effect After the adsorption process the leak valve was closed the base pressure
in the UHV chamber returning to below 10-10 mbar and the presence of adsorbed
molecules checked on the surface with STM Finally the sample was heated at 800degC for
several minutes to allow nanoisland growth Several different areas were examined on each
sample
Table 62 summarizes the dose of exposure and the heating time for each molecule Other
experiments were carried out using smaller dosing times but only the significant results
are reported here
As expected oxygen and water react with the surface oxidising it The adsorption process
can be easily followed with the STM by lowering the tip from time to time and taking an
image No effect of the STM tip is observed over the scanned areas Upon molecule
adsorption the order of the crystalline surface rapidly degenerates and after several
minutes the surface looks completely covered by irregular atomic-sized protrusions
corresponding to Si-OH or Si-O-Si species On heating to 800degC the contaminating species
were easily desorbed through SiO formation the surface became clean and the crystal
surface periodicity re-established No residues or nanoislands were observed
Amongst the molecules tested was methanol CH3OH reacts readily with the surface and
covers it with a disorderd layer already after sim 30 Langmuirs exposure (fig 61)
Fig 61 Si(111) surface during methanol adsorption at 0 1 and 5 minutes respectively (corresponding to 0 6 and 30 L exposure) The dark irregular lines are borders between different 7x7 domains Image size 60x60 nm
VPalermo 67
Synchroton radiation photoemission spectroscopy experiments by Carbone et al [9]
showed that methanol reacts with the rest atoms of the 7x7 silicon cell already after 1
Langmuir exposure but that longer exposures are needed for methanol to react with the
remaining silicon atoms Methanol adsorbs dissociatively forming SiOCH3 and Si-CHx
species on the surface
After methanol adsorption the sample is heated to 800degC and observed again with STM
As in the case of water and oxygen the heating restores the crystalline surface and the 7x7
pattern returns visible But this time quite a dense array of nanoislands is observed on the
surface (fig 62) The islands have an average diameter of 20 nm and a height of sim2 nm
At temperatures above 400degC the Si-OCH3 and Si-CHx species decompose oxygen and
hydrogen are desorbed and according to [9] carbon atoms remain as SiC dispersed
uniformly on the surface Our STM measurements indicate that the carbide does not
randomly cover the surface but that C atoms are concentrated at the nanoislands Rough
calculations based on island volume suggest that the islands must be composed of a SixC1-x
alloy with x varying between 05 and 1
Following these results we expected carbon monoxide to behave in a similar manner
adsorbing onto the surface and with nanoisland growth However no adsorption was
observed with STM and the surface looked perfectly clean even after a dose of more than
100 L of carbon monoxide Some species did adsorb onto the surface though because
after 10 minutes subsequent heating nanoisland growth was observed Island density in
this case was only 18 of the density obtained with methanol and island dimensions are
Fig62 Nanoislands on Si(111) created after 30 L adsorption of methanol and 10 min heating at 800degC Image size 200x200 nm
68 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
much bigger (see table 62 and fig 63) To check whether the islands had been nucleated
by some other contaminant present in the vacuum chamber we heated a blank sample
without introducing any molecule No island growth was observed Exposing the surface to
other simple molecules including carbon dioxide and methane also gave no nanoisland
formation
We can estimate the amount of carbon deposited on the surface during gas dosing The
total flux of molecules hitting the surface is obtained from the Hertz-Knudsen formula
TkmPF
π2=
Fig 63 STM images of Si(111) after exposure to methanol (left) and carbonmonoxide (right) contaminating molecules and subsequent nanoisland growthImage size 400x400 nm The size distribution of nanoislands is shown below eachimage
VPalermo 69
Fig64 Left Si(111) after 180 L contamination of CO and 16 hours heating at 800degC Right Si(111)without CO contamination after 16 hours heating at 800degC Image size 250x250 nm Vertical scale is thesame for both images z-ranges are 35 and 5 nm respectively
where P is the pressure in Pascal k=138x10-23 J K-1 is the Boltzmann constant T is the
temperature and m is the molecular weight (sim32 for methanol and sim28 for CO) in
kilograms
Using this formula we obtain a flux of 027 molecules nm-2 sec-1 for methanol and 029
molecules nm-2 sec-1 for CO After ten minutes exposure at 10-7 mbar pressure more than
150 molecules will have hit each square nanometer of the surface Surface density of
Si(111) 7x7 is sim16 atomsnm2 so each surface atom will be hit by several molecules which
could react or be adsorbed even assuming a sticking coefficient much lower than unity
To explain nanoisland growth we hypothesize that CO molecules adsorb molecularly and
-5
0
5
10
15
20
25
30
0 20 40 60 80 100 120
nm
nm
10 min40 min16 h16 h- no CO
Fig65 STM profiles of nanoislandsgrown for different annealing times The lower curve correspond to the surface heated at 16 hours without CO contamination
70 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Table 63 Island density and volume for the CO-contaminated surface at different heating times For each sample the thickness of a uniform layer having the same volume per unit area of the islands is calculated
Molecule Dose
(Langmuir) Heating time at
800degC (min) Island density
(microm-2) Ave Volume
(nm3) Equivalent layer thickness (nm)
CO 180 10 362 517 019 CO 180 40 550 1072 059 CO 180 16 h 505 3675 186
clean surface - 16 h rough - -
very weakly on Si(111) moving rapidly on the surface Because of their rapid diffusion
the STM is not able to image the CO molecules on the silicon [10]
Increasing the temperature part of the CO molecules will surely desorb but some of them
will dissociatively adsorb onto the surface The oxygen will then desorb as SiO while
carbon will remain fixed forming very stable Si-C bonds
CO decomposition will be favored at surface defects surface steps or in the proximity of
already formed SiC clusters An increased reactivity of the CO molecule on a surface in
correspondence with phase boundaries has already been observed on Pt surfaces [10]
In this case a reduced number of nucleation centres will react with the CO molecules
yielding large and fewer islands with respect to the case of methanol which does not have
the possibility of travelling long distances over the surface
Other factors are likely to influence the process of island nucleation after CO adsorption
the co-adsorption of other molecules which can slow down CO and favour its
decomposition cannot be ruled out as well as the formation of new defects at higher
temperatures Measurements with a variable temperature STM or with some other surface
analysis technique are planned in the future to confirm the proposed mechanism
When the surface was further annealed the nanoislands grew in size After 16 hours
heating very large nanoislands with diameters of sim35 nm and heights of sim20 nm became
visible (fig64 left) As a comparison when the same surface was heated without
nucleation centres deriving from contamination an irregular surface was obtained (fig64
right)
While the nanoisland density seemed to reach an asymptotic limit of 500 islandsmicrom2
(table 63) island size continued to grow with time (fig65) even when no further carbon
was supplied to the surface This indicates that even though the initial nucleating core of
the nanoisland is likely to be an SixC1-x alloy further growth is due to silicon atoms
diffusing from the crystal and being adsorbed by the growing island The final volume
VPalermo 71
occupied by the islands corresponds roughly to a 18 nm thick overall layer of removed
silicon
It is noteworthy that the clean sample after 16 hours annealing even though smoother than
the CO-contaminated one looked much more disordered with hardly any flat area visible
on nanometres scale In the presence of nanoislands however further annealing increases
island size but keeps part of the surface quite flat and ordered with flat areas visible
between the islands Thus the nanoislands can be imagined to act as ldquoimpurity sinksrdquo for
further contaminants approaching the surface yielding a greater but more ordered surface
roughening with respect to the uncontaminated silicon surface
Circular arrays of nanoislands
We used the voids described in the previous chapter formed during oxide layer
decomposition as nanoscopic masks to test the validity of the hypothesis outlined above
The small circular areas uncovered during oxide decomposition can act as ldquoskating rinksrdquo
where contaminants and diffusing silicon atoms can move freely over the surface while
the oxide layer all around and higher than the void area as well as being chemically
bonded to the surface silicon atoms will hinder surface diffusion Furthermore the oxide
passivating layer will at least in some cases prevent contaminating molecules from
adsorbing and decomposing on the surface In this way selective nanoisland growth inside
void areas can take place
The experimental procedure consisted of the following steps
1) An oxide-passivated sample was heated for 10 minutes at 800degC Oxide decomposition
began in correspondence with surface defects and spread laterally creating naked areas
of silicon
2) While the void enlarges the same defect that initiated oxide decomposition behaved as
an attractor for moving silicon atoms and nucleated the growth of a nanoisland at the
void centre
3) The void surface was exposed to contaminating molecules
4) The sample was then heated again at 800degC During the second heating the void
continued to enlarge and new nanoislands were created The freshly uncovered silicon
surface provided mobile silicon atoms to feed nanoisland growth
72 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
5) When the heating continued for long enough all the oxide was desorbed and a clean
silicon surface covered by circular groups of nanoislands was obtained
Figure 66 shows the different stages of the process
We thus heated an oxidized sample at 800degC for 10 minutes in UHV to create multiple
voids in the oxide layer After this we exposed the sample to the contaminating molecules
Then we further annealed the surface at the same temperature to promote island growth
The first time we observed nanoisland growth in the void area was by chance after heating
for a second time samples which had been stored for 2 weeks in vacuum Even at pressures
as low as 10-11 mbar some molecules will adsorb onto the surface slowly creating defects
and thus potential nanoisland nucleation centres It is possible to observe with STM that
the silicon crystal surface looks very clean just after a high temperature flash but even
after only a few days randomly adsorbed species will be observable on the surface
Unfortunately it is not possible to identify these species simply from in situ STM
measurements
Mass spectrometry measurements show that the residual gas contaminants in vacuum are
usually He Ar H2 CH4 CO CO2 and N2 [11] Some of these molecules such as
hydrogen will stick to the silicon surface at room temperature but will simply desorb
during annealing without nucleating nanoislands Others like CH4 or CO2 will not react
with the surface to generate nanoislands as demonstrated previously The best candidate
for contamination of samples stored in UHV seems thus carbon monoxide has the ability
321
4 5
Fig 66 Selective nanoisland growth within oxide voids See text for details
VPalermo 73
as shown before to nucleate nanoislands although co-adsorption mechanisms can not be
ruled out Fig 67 (left) shows the circular groups of nano-islands obtained after heating
void-covered samples stored for 2 weeks in UHV Small islands are observed in the former
void area The void itself has enlarged slightly uncovering a clean oxide-free area of the
silicon surface Apart from the central island created during the first period of heating the
surrounding ones have dimensions decreasing from the void border to the inner area of the
void
This confirms that most of the material needed for island growth comes from the silicon
atoms diffusing from the freshly uncovered areas around the original void The outer
nucleating centres are nearest to the silicon atom source and thus generate larger islands
For a faster process we directly exposed the void to a significant concentration of CO
Fig67 (centre) shows the islands grown after a 180 Langmuir exposure and subsequent
heating The islands are preferentially located around the original void perimeter the rest
of the original oxide-free surface remaining untouched This suggests that although the
whole surface is exposed to CO the gas only interacts chemically with the surface at the
SiSiO2 interface around the oxide-free void perimeter The SiSiO2 interface thus provides
preferential nucleation sites for nanoisland growth in the successive thermal annealing
step This recalls a similar phenomenon observed in the catalytic oxidation of CO on the
Pt(111) surface where CO molecules reacted with adsorbed O species only along the
perimeters of oxygen islands [10]
On repeating the CO exposure and thermal annealing steps a second circle of nano-islands
Fig67 Left circular area of nanoislands obtained from a sample stored 2 weeks in UHV 500x500 nmCentre nanoisland circle obtained after exposing the voids to 180 L of CO and heating for 10 min at800degC 400x400 nm Right two concentric circles obtained with further CO adsorption followed byheating 650x650 nm
74 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
was formed around the newly-expanded surface void perimeter (fig67 right) This
process can in principle be repeated again to obtain a series of concentric groups of
nanoislands the only limit depending on the minimal distance between neighbouring
expanding voids We have thus observed that contaminants adsorbed at low pressure over
long periods of time generate random island growth over the whole void area while CO
molecules adsorbed at relatively high pressures over short periods yield selective
nanoisland growth at the former void border Several theories can explain the different
results CH4 and CO2 as mentioned above did not nucleate islands when when the silicon
surface was exposed to them for short periods However the lower surface mobility of
these contaminants and the greater time available for reaction with the surface could be
the cause of nucleation of SiC clusters over the whole oxide-free void area and not only at
its border The coadsorption of water another UHV residual gas with CO on the naked
silicon areas could perhaps allow a surface reaction similar to that between CO and the
SiSiO2 interface during CO exposure Another explanation for the formation of these
random nucleation sites within the voids could be the coadsorption of residual carbon-
containing species present in the UHV system with hydrogen the major residual gas in
stainless steel UHV systems
The possibility of decorating silicon oxide borders with nanoislands using the high
mobility and selective decomposition of CO molecules suggests interesting applications
for nanofabrication processes that will be discussed below
Silicon nanowire creation on Si(100)
We performed most of the experiments described above on silicon (111) 7x7
reconstructed because this surface is easily prepared in UHV and above all because it is
isotropic and has no preferential directions for atom diffusion The 2x1 reconstruction of
silicon (100) even though it has a much simpler unit cell than Si(111) 7x7 shows a more
complicated morphology with lots of monoatomic and biatomic steps and dimer rows
parallel to each other The orientation of the dimer rows changes by 90deg on alternate
atomic layers Surface diffusion in this case is much more complicated because atoms will
experience the effects of the surface anisotropy and diffuse preferentially along or across
the dimer rows The diffusion energy of a silicon atom as example is 06 eV along a
dimer row and 085 eV across different dimer rows This while being a complication
VPalermo 75
offers interesting possibilities for the creation of ordered structures The monoatomic steps
and the dimer rows can for example act as templates for the formation of elongated
structures of composition similar to the islands described previously
On heating a Si(100) surface oxide decomposition and void growth takes place as
described for Si(111) In some cases the void shape reflects the surface symmetry
especially for very thin oxide layers and the voids often have a slightly squared shape
Nanoisland growth is also influenced by the substrate symmetry and the islands look
rectangular as irregular clusters surrounded by concentric patterns of monoatomic steps
The whole structure reminds vaguely a ldquozigguratrdquo a kind of stepped pyramid found in
Mesopotamia (fig68 inset)
Apart from nanoisland growth in some cases we observed spontaneous formation of
nanowire-like SiC structures on the Si(100) surface The wires were generated after
exposing a Si(100) surface in vacuum to traces of CO followed by sample annealing at
800degC for 15 minutes The wires exhibit lengths ranging from 10 to 100 nm and average
widths of ca 5 nm Although being randomly positioned on the surface they are perfectly
aligned along the crystal axes of the substrate (Fig68)
Several descending monoatomic steps can be observed around each nano-line and nano-
island Very often one of the line extremities coincided with an island
Nanoline
Nanoisland
Fig 68 A Si(100) surface covered by nanoislands and nanowires The monoatomic steps present on the surface are visible The two insets show a typical nanoisland and a nanowire obtained on this surface The typical ldquodimer-row ldquostripes are visible along the sides of the nanowire
76 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Couples of parallel lines starting from the same nano-island were also observed At high
resolution (fig 69) the lines showed a lamellar periodic structure with a periodicity of
around 4 nm and frequent defects Scanning Tunneling Spectroscopy measurements
showed no significant difference between the line and the surrounding silicon surface
The proposed formation mechanism for this kind of structure involves two stages in the
first one organic contaminating molecules adsorb and diffuse on the surface from the
vacuum environment stopping preferentially at surface singularities including steps and
defects
Several substances such as atomic carbon carbon monoxide ethylene and fullerene
decompose when adsorbed on silicon surfaces at temperatures above 500degC and create
localised defects on the surface inducing strain deformations and a metastable surface
reconstruction [12]
If these molecules remain attached to the steps they can diffuse rapidly along step border
until they meet a line and decompose In this way long lines of SiC defects parallel to the
surface steps can be produced In the second phase the clusters of silicon carbide can
-005
0
005
01
015
02
025
03
035
04
-3 -2 -1 0 1 2 3
V
dId
V lt
dId
Vgt
linesilicon
Current- Voltage behaviour measured with STS Fourier analysis of wire periodicity
asymp 4 Aring
High resolution image of a line
Fig 69 Top high resolution image of a nanowire showing the lamellar structure Fourier analysis (bottom left) shows that the lamellae periodicity is sim 04 nm bottom right the IV characteristics measured with the STM tip on the line and the silicon surface
VPalermo 77
25-600degC
Final linear structure
600-800degC
Contaminant Silicon
Fig 610 Schematic representation of nanowire formation mechanism
locally inhibit the spontaneous surface diffusion of silicon atoms acting as templates for
the reorganisation of surface steps (fig 610)
The final morphology shows disordered lines and islands each surrounded by a complex
pattern of silicon monatomic layers
At 800degC silicon atoms are highly mobile on the surface Due to atom diffusion the
surface reorganizes with a step-flow mechanism The presence of a fixed line of
contaminants blocks atom diffusion and step flow generating the monoatomic steps
around each wire (fig 611) SiC clusters act as lsquopinning sitesrsquo on silicon atom surface
diffusion and can be purposely used to direct the localised growth of nano-islands [13]
Although the possibility of using methanol and carbon monoxide as nano-island precursors
has been demonstrated the exact nature of the substances nucleating nano-wire growth is
still unknown
Further experiments will be needed to confirm these hypotheses as well as to identify the
contaminating species nucleating nanowire growth Carbon monoxide is a main candidate
for nanowire growth but the real nucleation process is likely to be a complex one
78 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Step flowblocked
Fig 611 Representation of step flow blocked by the presence of a nanowire 105x105nm
Bibliography
[1] Surface-Chemistry Of Silicon Waltenburg HN Yates JT Chemical Reviews 95 (5)
1589-1673 Jul-Aug 1995
[2] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[4] Growth Of Si On The Si(111) Surface Lanczycki Cj Kotlyar R Fu E Yang Yn
Williams Ed Das Sarma S Physical Review B 57 (20) 13132-13148 May 15 1998
[5] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
VPalermo 79
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990
[6] Nanoscale roughening of Si(001) by oxide desorption in ultrahigh vacuum Gray SM
Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology B 14 (2)
1043-1047 Mar-Apr 1996
[7] Step Motion On Crystal Surfaces Schwoebel Rl Journal Of Applied Physics 40 614-
618 Feb 1969
[8] Scanning Tunneling Microscopy Investigation At High Temperatures Of Islands And
Holes On Si(111)7x7 In Real Time Evidence For Diffusion-Limited Decay
Hildebrandt S Kraus A Kulla R Wilhelmi G Hanbucken M Neddermeyer H Surface
Science 486 (1-2) 24-32 Jul 1 2001
[9] Methanol Adsorption On Si(111)-(7x7) Investigated By Core-Line Photoemission And
Mass Spectrometry Of Photodesorbed Ions Carbone M Piancastelli Mn Zanoni R
Comtet G Dujardin G Hellner L Surface Science 370 (1) L179-L184 Jan 1 1997
[10] Atomic And Macroscopic Reaction Rates Of A Surface-Catalyzed Reaction
Wintterlin J Volkening S Janssens Tvw Zambelli T Ertl G Science 278 (5345)
1931-1934 Dec 12 1997
[11] Redhead PA Hobson JP Kornelsen EV The Physical Basis Of Ultrahigh
Vacuum Chapter 12 (Chapman amp Hall London 1968)
[12] The Si(001) C(4 X 4) Surface Reconstruction A Comprehensive Experimental
Study Norenberg H Briggs Gad Surface Science 430 (1-3) 154-164 Jun 21 1999
[13] Production Of Nanostructures Of Silicon On Silicon By Atomic Self-Organization
Observed By Scanning Tunneling Microscopy Jones D Palermo V Applied Physics
Letters 80 (4) 673-675 Jan 28 2002
80 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 81
7 Conclusions and proposals for future work It is very difficult to foresee the future of nanoscience and nanotechnology even over the
next few years exciting discoveries are continuously made opening up new possibilities
and progress in this sector is now one of the fastest and most varied So it is impossible to
predict whether computers will in the future be made with nanowires nanotubes single
electron devices or some kind of quantum-based transistor What does seem clear
however is that silicon devices can still further extend their levels of miniaturization to
remain competitive for at least the next ten years [1] During this period new emerging
technologies will reach scientific maturity and arrive at the production lines
Even when new technology prototypes do become available for full-scale production the
astronomical cost of changing from silicon to new technologies will be a major concern
Moreover a vast amount of extremely detailed knowledge is already available on all
aspects of silicon technology Thus the possibility of integrating innovative
nanotechnologies with standard CMOS silicon technology already seems an attractive
prospect [2]
For this and other reasons we have focussed our research efforts on surface phenomena on
silicon wafers which could be potentially useful for the development of silicon-compatible
nano-devices
Some important conclusions can be drawn from the work described in this thesis
bull Silicon surfaces of different crystal faces were studied at atomic resolution in ultra-
high-vacuum following transformations induced on the surface by chemical (etching)
and physico-chemical (organic contamination heating) treatments
82 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull The presence of surface contaminants induces at high temperature the growth of
nanometre-sized islands and lines on silicon
bull Final island densities and dimensions strongly depend upon the chemical nature of the
contaminant molecules and the treatments used Molecules interacting only weakly
with silicon will desorb upon heating whereas molecules which decompose easily on
Si will give widespread random nano-island nucleation Between these two extreme
cases complex behaviours of diffusion and subsequent decomposition are possible
bull Nanoscopic naked silicon voids were produced in oxide layers through carefully-
controlled thermal annealing It was found that void density and size depends upon
oxide purity surface contamination levels and annealing conditions
bull The growth of nano-islands and nano-voids was exploited simultaneously and the
voids used as nanoscopic masks to control the positions of growing islands Complex
silicon-on-silicon nano-structures were obtained with multi-step growth processes
bull Finally the use of macroscopic treatments and simple physical processes to produce
large quantities of nanometre-sized structures on silicon surfaces was demonstrated
These processes and the structures produced are of potential interest to the
microelectronics industry
We showed in previous chapters how it is possible to modify silicon surfaces using
chemical and physical methods The use of self-organization for the production of Si-on-Si
20 nm
Fig 71 Left SEM image of field emitter arrays made by conventional lithography [3] Center 3-D STM image of a self-organized void with a central nano-island Right Lateral schematic view of a field emitter device
VPalermo 83
nano-structures has the advantage with respect to other emerging techniques of being
silicon compatible and thus is a potential candidate for its implementation in the
production of new devices Furthermore this technique can create billions of strucures on a
wafer in only a few minutes being thus applicable to large-scale production
Even when ordered structures are obtained these methods still show a poor reproducibility
and control Chemically modified surfaces look very disordered at high magnification
with trenches and holes etched into the surface and progress has recently been made in
producing very small structures electrochemically [4] Oxide decomposition with
nanoisland growth in UHV can also yield ordered structures The void holes with
nanoislands located at their centres although quite variable in dimensions are more similar
to engineered devices than to a spontaneous random surface-roughening process In fact
there is a surprising similarity between the nanovoid-nanoisland structures and well-known
commercially available devices called field emitter arrays (FEA) These are small tips used
as electron guns in various kinds of displays and other devices Both the commercial
devices and our self-organized structures consist of a protruding tip at the centre of
microscopic holes on an insulating layer covering a semiconductor surface (see fig71)
Of course the order and size uniformity of commercial FEAs is much better than that
obtained with our process but it is encouraging to see that a self-organized structure
obtained with simple heating has a very similar surface topography and composition of
commercial working devices obtained after many expensive stages of optical lithography
deposition and etching
Nano-island density and size can to some extent be controlled and different
contaminants can be adsorbed onto the surface to modify the growth process In the future
with a better understanding of diffusion dynamics and growth mechanisms a multi-stage
fabrication process can be envisaged where purposely-designed molecules are adsorbed
onto silicon surfaces to diffuse organize in ordered arrays and favour surface
reorganization at the nanometre scale The same molecules could be used to deliver
dopants to obtain for example highly conducting nanostructures on otherwise undoped
highly resistive silicon The selective formation of nanostructures at the SiSiO2 border
can be used to scale down mesoscopic patterns to the nanometric level A silicon oxide
layer can be easily patterned on the surface with normal optical lithography and then
nanostructures can be built following the pattern edge In this way the production of very
small nanowire based devices may be possible (fig 72)
84 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
D
Gate
S
D
S
SiO2
SiO2
2 Heating nanowire creation
and oxide removal
3 Source-drain deposition by conventional lithography
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
4 Insulating layer (oxide) and gate (metal) deposition
Fig 72 Schematic drawing of the construction of a self-organized nanowire-based transistor
Surface diffusion and decomposition of molecules can be sensitive to even smaller surface
features The nanoscopic lines formed on Si(100) are very regular and neat if their
nucleation and growth mechanism could be controlled ordered networks of nanowires
could be made on silicon
VPalermo 85
The techniques developed here are still experimental ones the careful control and
reproducibility of each step is still an issue and further studies will be needed to clearly
understand the dynamics of formation of these structures Nonetheless the results obtained
so far open up the possibilities of producing novel devices such as for example nano-
transistors (see scheme in fig73 where the mechanism described in fig 72 is applied for
the fabrication of a device array) The possibility of using the border of a lithographic
pattern to draw nanolines can in principle allow the creation of nanodevices using normal
IC manufacturing techniques
The results obtained and their description in this thesis are not the first examples of the use
of self-organization for the production of ordered structures and will certainly not be the
last In our opinion the most important outcome of these findings is that it is possible to
use surface diffusion to overcome the inherent limits of lithographic techniques in
microdevice production processes and to force properly selected molecules to react with
the surface in a spatially differentiated manner at particular sites thereby creating ordered
series of nanostructures The role of the surface dynamics of the substrate atoms
themselves is extremely important in this process
Outstanding results have been and continue to be obtained in nanoscience and
nanotechnology research promising the emergence of new production technologies Sooner
or later those emerging technologies will have to face the issues of cost and compatibility
with the enormous investment made in existing technologies and processing facilities The
commercial production of nanostructures based on surface diffusion processes could in
principle be cheap simple and compatible with existing technology Moreover the
processes of diffusion in bulk silicon is well-known in the microelectronics industry which
has both the know-how and the equipment to better understand and exploit the surface
diffusion processes on silicon surfaces studied here
Extremely important and innovative results are often obtained by exploiting very simple
ideas and in our opinion physical processes such as surface diffusion used to create
complex nanostructures on silicon surfaces is a prime example
86 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
2 Heating nanowire creation and
oxide removal
3 Source-drain deposition by conventional lithography
4 Insulating layer (oxide) and gate (metal) deposition
Fig 73 Schematic macroscopic outline of the construction of a self-organized nanowire-based transistor
VPalermo 87
Bibliography
[1] P Gargini Intel Technology Coordinator ldquoFrom Microelectronics To
Nanotechnologiesrdquo Invited Campus Colloquia Lecture CNR Research Area Bologna
February 6th 2003
[2] Toward A Hybrid Micro-Nanoelectronics Cerofolini Gf Ferla G Journal Of
Nanoparticle Research 4 (3) 185-191 Jun 2002
[3] Recent Progress In Field Emitter Array Development For High Performance
Applications Temple D Materials Science amp Engineering R-Reports 24 (5) 185-239
Jan 25 1999
[4] Silicon Dioxide Micropillars For Sieving Fabricated By Macroporous Silicon-Based
Micromachining Izuo S Ohji H French Pj Tsutsumi K Kimata M Sensors And
Materials 14 (5) 239-251 2002 Electrochemical Etching In HF Solution For Silicon
Micromachining Barillaro G Nannini A Piotto M Sensors And Actuators A-Physical
102 (1-2) 195-201 Dec 1 2002
88 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 89
Acknowledgements
This work would not have been possible without the help and guidance of Dr Derek Jones
he introduced me to the use of STM and other techniques found the funding for this PhD
and helped me with the thousands of small and big problems I had to face during my
research I am grateful to Prof Alberto Ripamonti too for being the supervisor of this
thesis Thanks are also due to Dr Giancarlo Seconi director of ISOF-CNR where most of
this work was carried out Financial support from the Italian National Research Council
(CNR) is also gratefully acknowledged
My passion for surface science began when Fabio Biscarini showed me that it was possible
to really see the atoms and for this I will always thank him
During these years I had the luck to collaborate with many good scientists among those I
would like to cite Enrichetta Susi Massimo Cocchi and Anna Mazzone from CNR
Claudio Zannoni Anna Cavallini Daniela Cavalcoli and Antonio Castaldini from the
University of Bologna Sergio Pizzini and Maurizio Acciarri from the University of Milan
One of the most stimulating periods of my PhD was the one I spent at the Steacie Institute
for Molecular Sciences in Ottawa working with Robert Wolkow Dan Wayner Greg
Lopinski and Peter Kruse I thank them for discussing together many exciting scientific
ideas and trying to put some of them into practice
Special thanks go to my colleagues Paolo Samorigrave and Stefano De Cesari for long and
useful discussions on Science Life and their interactions often with the help of a good
pint
Finally I would like to thank Vassilia Gaetano Simone Claudia Silvia Letizia
Alessandro Angela and Sebastien even though not directly involved in this work they
have been over these three years a continuous source of happiness and support
Bologna March 2003
90 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
SCIENTIFIC PUBLICATIONS bull Lateral diffusion of titanium disilicide as a way to contacting
hybrid Si-organic nanostructures Palermo V Buchanan M Bezinger A Wolkow RA APPLIED PHYSICS LETTERS 2002 v81 p 3636 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v6 issue 20) bull Surface modifications in Si after Rapid Thermal Annealing Castaldini A Cavalcoli D Cavallini A Jones D Palermo V Susi E JOURNAL OF THE ELECTROCHEMICAL SOCIETY 2002 v 149 pG633 bull Production of nanostructures of silicon on silicon by atomic self-
organisation observed by scanning tunnelling microscopy Jones D Palermo V APPLIED PHYSICS LETTERS 2002 v 80 p 673 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v5 issue 5) bull Nucleation of nanostructures from surface defects on silicon Palermo V Jones D SOLID STATE PHENOMENA 2002 v 82-84 p 687 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo MATERIALS SCIENCE AND ENGINEERING B 2002 v88 (2-3) p220 bull Morphological changes of Si[100] surface after treatment with
concentrated and dilute HF Palermo V Jones D MATERIALS SCIENCE IN SEMICONDUCTOR PROCESSING 2001 v 4 p 437 bull Electrical and structural properties of processed silicon surfaces Susi E Cavallini A Castaldini A Cavalcoli D Jones D Palermo V ldquoRECENT RESEARCH DEVELOPMENTS IN VACUUM SCIENCE amp TECHNOLOGYrdquo 2001 v 3 p189 bull Numerical Solutions of the Stochastic Equations of Crystal
Growth Mazzone A M Palermo V INTERNATIONAL JOURNAL OF MODERN PHYSICS C 2000 v 11 Part 1 p195-204 bull Advances in silicon surface characterisation using light beam
injection techniques
VPalermo 91
Acciarri M Pizzini S Simone G Jones D Palermo V MATERIALS SCIENCE AND ENGINEERING B 2000 V73 (1-3) p 235 - 239 bull Abrupt orientational changes for liquid crystals adsorbed on a
graphite surface Palermo V Biscarini F Zannoni C PHYSICAL REVIEW -SERIES E- 1998 V 57 NUMBER 3A p R2519-R2522 Oral presentations bull The use of oxide desorption and surface diffusion for the
creation of silicon on silicon nanostructures Palermo V Jones D 1ST NATIONAL WORKSHOP ON CURRENT TRENDS IN NANOTECHNOLOGIES Catania (Italy) Feb 2002 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo V E-MRS Spring Meeting Strasbourg (France) June 2001 bull STM study of surface transformations on silicon during UHV
annealing Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999 bull Advances in silicon surface characterisation using light beam
injection technique Acciarri M Pizzini S Simone G Jones D Palermo V E_MRS SPRING MEETING June 1999 Posters bull Spontaneous nano-wire growth on silicon Palermo V Jones D SMARTON workshop Leuven (Belgium) October 2002 bull Ordered circles of nano-islands on silicon from CO adsorption Palermo V Jones D TRENDS IN NANOTECHNOLOGY 2002 Santiago de Compostela (Spain) September 2002 bull Production of nanostructures of silicon on silicon by atomic self-
organisation Palermo V Jones D EUROMAT Rimini (Italy) June 2001 bull Etching holes and anisotropic corrosion on silicon [100] Palermo V Jones D Susi E Asoli B SILICON WORKSHOP Genova (Italy) February 2001
92 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull Morphological and electrical characteristics of damaged silicon surfaces
Susi E Castaldini A Cavalcoli D Cavallini A Jones D Palermo V SILICON WORKSHOP Genova (Italy) February 2001 bull Effect of HF etching on the roughness of a silicon surface Palermo V Jones D NATIONAL CONFERENCE ON PHYSICS OF MATTER Genova (Italy) June 2000 bull Effect of Native Oxide Desorption upon the Surface Morphology of Si[100] by STM and LEED Palermo V Jones D SILICON WORKSHOP Genova (Italy) February 2000 bull Characterisation of silicon surfaces for microelectronics through STM measurements Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999
VPalermo 93
Contact Vincenzo Palermo ISOF- Institute for Organic Synthesis and Photoreactivity Via Gobetti 101 40129 Bologna ITALY Tel +39-051-6398336 Fax +39-051-6398349 Mail palermoisofcnrit
94 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Index
1 Introduction nanotechnology and the future of computers
2 Silicon surfaces
3 STM and other surface analysis techniques
4 Surface modification of silicon in liquid Nano-hole creation
5 Surface modification of silicon in vacuum void creation and oxide desorption
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and
nano-lines
7 Conclusions and proposals for future work
VPalermo 9
silicon technology
Below is a brief summary of the most recent developments in nanotechnology and
nanoscience
Actual trends in nanotechnology
Perhaps the most fascinating idea for nanodevice construction is to use one single
molecule working as a complete device The first molecular diodes (ie molecules
conducting current only in one direction) were created in 1997 in 1999 a molecular fuse
and a molecular transistor were demonstrated although there was no possibility of wiring
these devices to external contacts In April 2001 James Heath and his group at UCLA
fabricated an array of overlapping crossbars and placed a small molecule of rotaxane
between each crossbar (fig 13 left) This composite molecule is made up of two
component parts the main rod-like molecular axis and a mobile ring ldquothreadedrdquo on it like a
bead on a necklace and can function as a molecular switch A working 16-bit memory
circuit was constructed using these molecules For a brief review of these works see [4] In
June 2002 a single molecule transistor was built by connecting an organic molecule to two
metal contacts the molecule contained one or two atoms of a transition metal (cobalt or
vanadium) forming the active region of the device supported by an organic backbone [5]
Fig13 Working nanodevices Left schematic representation of rotaxane molecules between crossed nanowires [4] Right SEM image of semiconductor nanowires forming a small circuit [9]
10 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another approach to nanodevice fabrication has become possible through the discovery of
carbon nanotubes which were observed for the first time in 1991 by a Japanese electron
microscopist studying the material deposited during arc-evaporation synthesis of fullerenes
[6] They consist of a graphite-like carbon seamless cylinder with a diameter of several
nanometers and lengths of up to a millimetre Carbon nanotubes are very stable can
behave as metals or semiconductors and can host other molecular or ionic species thus
modifying their electrical behaviour In 2001 Avouris and his group reported the first
circuit made with a single nanotube [7] A few months later Cees Dekker presented a
nanotube-based transistor able to amplify an input signal by a factor of ten and built
several logic circuits using these nanotube transistors [8]
One problem with carbon nanotubes is that it is very difficult to control their electronic
properties ie their metallic or semiconducting behaviour An alternative to carbon
nanotubes are semiconductor nanowires Silicon nanowires can be made using a laser to
vaporize the silicon together with a metal catalyst like iron or gold The vapour condenses
in nanosized drops of silicon and metal from which the wires slowly grow out as more
silicon is adsorbed In 2001 a group at Harvard University [9] created a transistor by
crossing two different nanowires After this the same group arranged four nanowires in a
noughts and crosses grid creating something like a 4-bit memory (fig 13 right) Even
metallic nanowires made of platinum and silver can be used in a crossed configuration to
store information [10]
There are some issues common to all these new technologies though First it is difficult to
imagine these methods applied to large-scale production Up to now the insertion of a
molecule between two electrodes is an occasional lucky event while nanotubes and wires
have to be positioned on the surface creating the appropriate contacts on them manually
The large-scale production of integrated circuits using these building blocks will not be
straightforward and does not seem imminent
Another issue is of an economic and not a scientific or technological nature Since 1965
the cost of IC manufacturing plants has sky-rocketed If the cost of semiconductor
production plants continues to rise exponentially in a few years such plants will cost up to
$20 billion This is a sizeable investment even for large companies such as IBM or Intel
For this reason it is likely that IC companies will resist changing to completely new
technologies closing down their existing plants As it is clear that silicon will remain the
VPalermo 11
fundamental raw material of the IC industry for the foreseeable future nanotechnology
developments for microelectronics will need to be silicon-compatible In these early days
of nanotechnology the most valid approach would seem to be the addition of molecular
functions to existing silicon technology ndash using the latter as a foundation on which to build
Fabrication of self-organised structures on silicon
The possibility of using the phenomenon of atomic or molecular self-organization to create
nanostructures on silicon has already been demonstrated The clean silicon surface shows
in some cases a high degree of order and complex surface reconstruction as will be
described later Several different ordered structures form spontaneously on this surface
such as series of monatomic steps or boundaries between reconstructed areas It has been
demonstrated that it is possible to use these structures to fabricate ordered nanodots and
nanolines on the surface [11] More recently well-defined nanometric patterns have been
obtained with selective etching of silicon using nitric oxide [12]
In this study the possibility of creating different types of nanostructures on the silicon
surface is explored Methods had to be developed which were
- Simple They must not need complex masks or lithographic steps to create the
structure but rather exploit self-organisation phenomena
- High resolution the silicon surface modifications should be on a scalelength of below
100 nm
- Fast billions of nanostructures have to form over the whole surface simultaneously to
be compatible with large-scale production requirements
- Cheap they must not require expensive equipment (such as e-beam lithography) but
exploit simple chemical andor physical treatments to produce nanostructures on the
silicon surface
In Chapter 2 the main characteristics of silicon are described Chapter 3 provides a
summary of the techniques used for this research Chapter 4 examines the chemical etching
of silicon in different liquid environments and the effects of this etching on the surface at a
nanoscopic level with the creation of nanoholes
12 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Chapter 5 describes the growth in ultra-high vacuum (UHV) of nanoscopic voids and
islands on the silicon surface and the effect of surface oxide on this growth
Chapter 6 discusses the modification of silicon surfaces in UHV following the adsorption
of molecules and thermal heating to produce nanoislands and nanolines on silicon
The overall conclusions of our work are summarised in Chapter 7
Finally we will give some conclusions based on the results obtained and discuss possible
applications of the methods developed
Bibliography
[1] From wwwintelcom
[2] From wwwsematechorg
[3] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
[4] Molecules Get Wired Service RF Science 294 (5551) 2442-2443 Dec 21 2001
[5] Coulomb Blockade And The Kondo Effect In Single-Atom Transistors Park J
Pasupathy AN Goldsmith JI Chang C Yaish Y Petta JR Rinkoski M Sethna JP
Abruna HD Mceuen PL Ralph DC Nature 417 (6890) 722-725 Jun 13 2002 Kondo
Resonance In A Single-Molecule Transistor Liang WJ Shores MP Bockrath M Long
JR Park H Nature 417 (6890) 725-729 Jun 13 2002 Nanotechnology - Electronics
And The Single Atom De Franceschi S Kouwenhoven L Nature 417 (6890) 701-702
Jun 13 2002
[6] Smallest Carbon Nanotube Ajayan PM Ijima S Nature 358 (6381) 23-23 Jul 2 1992
[7] Carbon Nanotube Inter- And Intramolecular Logic Gates Derycke V Martel R
Appenzeller J Nano Letters 1 (9) 453-456 Sep 2001
[8] Logic Circuits With Carbon Nanotube Transistors Bachtold A Hadley P Nakanishi T
Dekker C Science 294 (5545) 1317-1320 Nov 2001
VPalermo 13
[9] Logic Gates And Computation From Assembled Nanowire Building Blocks Huang Y
Duan XF Cui Y Lauhon LJ Kim Kh Lieber CM Science 294 (5545) 1313-1317
Nov 9 2001
[10] Formation And Disappearance Of A Nanoscale Silver Cluster Realized By Solid
Electrochemical Reaction Terabe K Nakayama T Hasegawa T Aono M Journal Of
Applied Physics 91 (12) 10110-10114 Jun 15 2002
[11] Fabrication And Integration Of Nanostructures On Si Surfaces Ogino T Hibino H
Homma Y Kobayashi Y Prabhakaran K Sumitomo K Omi H Accounts Of Chemical
Research 32 (5) 447-454 May 1999
[12] Ultrafine And Well-Defined Patterns On Silicon Through Reaction Selectivity
Prabhakaran K Hibino H Ogino T Advanced Materials 14 (19) 1418-1421 Oct 2
2002
14 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 15
2 Silicon surfaces
The name silicon (silicio in Italian) comes from the latin word silex Amorphous silicon
was first isolated by Berzelius in 1824 by reaction of potassium with silicon tetrafluoride
Thirty years later the first crystalline silicon was prepared Silicon makes up 25 of
earthrsquos crust and is the second most abundant element after oxygen Elemental silicon is
not found in nature occurring as silicon oxide (sand quartz amethyst flint etc) or
silicates (asbestos clay mica etc) Perhaps no other element and its compounds has such
a wide range of uses Silicon compounds such as sand and clay are used in the building
industry as refractory materials for high-temperature applications and for enamels and
pottery Silica is the main component of glass silicon carbide is an important abrasive and
silicones are commonly used polymers and lubricants
Here the most interesting use of silicon of course is for the production of
microelectronics devices For this application silicon of high purity (999999) and of
high crystallinity is needed Table 21 lists some of the physical characteristics of silicon
High purity polycrystalline silicon is produced by the reaction of gaseous trichlorosilane
with hydrogen in a furnace Then to prepare a single-crystal of silicon the so-called
Czochralski method is commonly used
Polycrystalline silicon is melted in a quartz furnace at 1415degC in an argon atmosphere
Then a seed of single-crystal silicon is lowered into contact with the melt and slowly
pulled out In this way the crystal grows and a crystalline cylindrical ingot several metres
long is created from the initial seed
After cooling down the ingot is sliced into thin silicon wafers The wafer surfaces are
polished using a counter-rotating lapping machine in an Al2O3 slurry until the surface is
very flat and shiny ready for the lithographic processes
16 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another way to obtain single crystal silicon is the Floating Zone (FZ) method in which a
silicon cylinder is slowly passed through a heating ring The area inside the ring melts and
solidifies smoothly crystallising as it comes out of the ring yielding a single silicon crystal
Microelectronic devices are built on the silicon surface which is the surface of interest
here Unfortunately silicon surfaces are normally quite dirty and uneven at the atomic
scale Atmospheric oxygen and humidity react with silicon surfaces creating a thin layer of
oxide (called ldquonative oxiderdquo) which is usually irregular and full of defects Different kinds
of contaminants also adsorb onto the surface These are usually small organic molecules
and microscopic dust particles A clean surface on exposure to the atmosphere is
completely covered with gas molecules in less than 10-9 seconds If the pressure is
reduced letrsquos say to 10-6 mbar this time increases to 1 second This is the reason why to
study a clean surface we have to work in UHV at pressures below 10-10 mbar
The atoms in the silicon crystal have a diamond-like structure each atom having 4 bonds
in a tetrahedral sp3 arrangement with bond angles of 10947 degrees At the crystal
surface some atoms will have non-bonding orbitals ldquodanglingrdquo in the vacuum ie sp3
orbitals with a lone electron which are highly reactive These orbitals are known as
dangling bonds To minimize surface energy the surface will reorganize by decreasing
the number of dangling bonds
Table 21 Physical data of silicon [1]
Atomic Weight 2809 Lattice constant (A) 543095 Crystal structure Face-centered cubic
(diamond) Melting point 1415 degC
Density (gcm3) 2328 Boiling point 2355degC Atomscm3 50E22 Minority carrier
lifetime (s) 25E-3
Dielectric Constant 119 Specific heat (Jg degC)
07
Breakdown field (Vcm)
~3E5 Thermal conductivity (Wcm degC)
15
Electron affinity x(V)
405 Vapour pressure (Pa) 1 at 1650degC 1E-6 at 900deg C
Energy gap (eV) at 300K
112 Reactivity Inert to acids Attacked by halogens and alkaline
solutions Intrinsic carrier
conc (cm-3) 145E10 Oxidation states +4 -4
Intrinsic Debye Length (microm)
24 Energy of a Si-Si bond (eV)
232
Intrinsic resistivity (Ω-cm)
23E5
VPalermo 17
Dangling bond densities and positions and thus the type of surface reconstruction will
depend upon crystal orientation as well as the temperature and kinetics of the system
Fig 21 shows a drawing of the main faces of a silicon crystal The angle α between any
(11n) face and the (100) face can be calculated from 2cos 2 += nnα The angle
between any (11n) face and the (111) face can be calculated from
)2(3)2(cos 2 ++= nnα
The chemistry and physics of the faces are very different a brief description will be given
for the most important orientations
Si (100)
On the (100) surface each atom has two Si-Si bonds connecting it to the bulk and two
dangling bonds pointing outward Surface energy is reduced by the dimerisation of the
surface atoms through overlap interaction of one dangling bond per atom forming rows of
dimers aligned along the (110) direction This is the well-known ldquo2x1rdquo reconstruction of
this silicon surface The symmetric dimers would make the silicon surface metallic but to
reduce surface stress the dimers tend to buckle and the surface is thus semiconductive It
took several years to understand that the dimers are buckled because at room temperature
Fig 21 Schematic view of the principal orientations of a silicon surface Surface atoms are white bulk atoms are black dangling bonds are gray [2]
18 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
they shift easily from one buckling direction to the other thus appearing symmetric under
STM observation Fig 22 shows an STM image of the 2x1 reconstructed surface
Even almost perfect (100) surfaces have a certain number of monoatomic steps and the
dimer rows on atomic layers are aligned at 90deg to those on adjacent layers Dimer rows are
thus perpendicular or parallel to the step When the dimers on the upper side of the step
are parallel to the step the step is called SA if they are perpendicular the step is called SB
Because of this symmetry SA step edges will be smoother compared to the more broken
fragmented edges of the SB steps
A common defect on the Si(100) surface is the presence of nickel contamination which
appear as missing dimers in STM images This type of contamination is so critical that
even if the silicon sample is only briefly brought into contact with stainless steel tools
(tweezers for example) the 2x1 reconstruction of the surface can be blocked
Silicon atoms can diffuse easily over the silicon surface as monomers and dimers
especially at elevated temperatures The anisotropy due to the 2x1 reconstruction causes a
difference in the diffusion energies of adsorbates over the surface Diffusion of these
silicon species along dimer rows for example will be much easier A list of diffusion
energies for monomers and dimers is given below [2]
Diffusion on Si(100) 2x1 Ed (eV) Monomers along dimer rows 06 Monomers across dimer rows 085 Monomer formation energy 18 Dimer along dimer rows 11 Dimer across dimer rows 15 Dimer formation energy 26 Dimer binding energy 076 Vacancies along dimer rows 17 Vacancies across dimer rows 19
SA
SB
Fig 22 STM image of a 2x1reconstructed silicon surfaceshowing the dimer rows and steps40x35 nm Nickel-induced defectsare visible as dark spots SA and SBsteps are indicated
VPalermo 19
So the diffusion energy for both monomers and dimers is nearly 40 greater if they have
to cross a dimer row This difference reduces to sim10 for vacancy diffusion
Si(113)
The (113) surface can be imaged as a sequence of alternating (100) and (111)-like
structures with two and one dangling bonds on alternate atoms respectively Interest in the
(113) surface is scientific as it has been used to study the energetics of the (100) and (111)
surfaces as well as for surface adsorption experiments
Si(100) surfaces can easily develop (113) facets
Si(111)
This surface besides being the first one imaged with STM with atomic resolution is one of
the most studied because it is the best cleavage face of silicon and because it shows one
of the most complex and elegant reconstructions in surface science
All Si-Si bonds in the silicon crystal are perpendicular to a (111) plane so this face will
have the lowest number of dangling bonds created per unit area In fact each Si atom on a
(111) surface shows a single dangling bond oriented perpendicular to the surface and
bonded to three back atoms These three bonds for each surface atom account for the great
chemical and physical stability of the Si(111) surface Surface energy is 009 eV Aring-2
compared to 015 eV Aring-2 for Si(100)
For energy minimization this surface reconstructs forming a huge 7x7 lattice cell
containing 102 atoms described by the Dimer-Adatom-Stacking fault model (DAS) For a
detailed description of cell structure see fig 23
The cell described by this model is very complex being composed of three kinds of atoms
adatoms rest atoms and corner hole atoms Furthermore a subsurface stacking fault is
present in one half of the cell making the two halves of the unit cell look different under
STM (Fig 24) It took 26 years of research to completely understand the exact structure of
the 7x7 reconstruction
Cleaving a silicon crystal along a (111) plane produces a metastable 2x1 reconstruction
the 7x7 reconstruction is easily obtained by flashing at high temperature in UHV At T gt
20 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
830degC a disordered 1x1 phase covers the surface Cooling down to 800degC leads to the
formation of the 7x7 phase If the cooling process is too rapid small 7x7 domains nucleate
and a disordered 1x1 phase is preserved between domain boundaries
Si(110)
Even though as mentioned before the (111) plane is the favoured cleaving plane of
silicon thin (100) commercial wafers will not break along this plane because the angle
between (100) and (111) is too far from 90deg (see table 22) Instead they will break along
the (110) plane because it is perpendicular to the (100) surface Each surface atom on
Si(110) has a Si-Si bond pointing downward one dangling bond pointing outward and two
Si-Si bonds parallel to the surface in a zig-zag pattern (see fig 21) Cleaved (110)
surfaces are disordered but upon annealing at high temperatures an ordered complex 16x2
Fig23 Scheme of the 7x7 DAS model [2] In each unit cell there are 9 dimers 12 adatoms and a stacking layer fault The force driving this complex reconstruction is the minimization of dangling bonddensity The DAS model shows the lowest number of dangling bonds (19) of all possiblereconstructions 12 dangling bonds are at the adatoms 6 at the rest atoms and 1 at the corner hole atom This surface is metallic
VPalermo 21
reconstruction takes place The surface appears as a series of long ridges and valleys
parallel to each other Eventually tilted facets of orientation (17 15 1) can form on this
surface The adsorption of Ge atoms on this surface leads to the formation of self-
assembled nanowires [3]
Table 22 Angles in degrees between different silicon faces [2]
Orientation
113
110
111
100
100
2524
9000
5474
0
111
2950
3526
0
110
6476
0
113
0
Fig 24 STM image of a Si(111)surface with 7x7 reconstruction A unitcell with its adatoms is highlightedImage size 13x13 nm
22 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] Weast RC Handbook Of Chemistry And Physics (Chemical Rubber Co Cleveland
1972)
[2] Dabrowski J Mussig H Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] The Structure Of Clean And SiGe-Covered Si(110) Surfaces Butz R Luth H Surface
Science 365 (3) 807-816 Oct 1 1996
VPalermo 23
3 STM and other surface analysis
techniques
Scanning Tunneling Microscopy
Since the invention of the optical microscope at the end of the 16th century the possibility
of examining surfaces at higher and higher magnification has fascinated mankind
Development of the technique continued and towards the end of the 19th century optical
microscopes were as good as todays standard instruments The physical limits of the
wavelength of visible light (350-800nm) had been reached
In the 1920s de Broglie showed that electrons can behave like waves and the use of these
particles for imaging with much higher resolution soon followed Atomic resolution using
this technique is only possible in the transmission mode with extremely carefully prepared
samples
In 1982 using the peculiar properties of piezoelectric materials Binnig and Rohrer brought
a metallic tip very very close to a silicon surface and scanned it across an extremly small
area (fig31) The tunneling of electrons from the tip into the sample or vice versa allowed
them to obtain a local density of electronic states (LDOS) map of that surface Although
theory (which treated the extreme point of the tip as a sphere) then excluded the possibility
of atomic resolution following a tip crash into the surface Binnig and Rohrer began to
observe the LDOS with atomic resolution For this discovery and their following work
they were awarded the Nobel Prize in Physics in 1986
24 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The basic principle of STM is very simple A metallic tip is scanned over a surface without
making ohmic contact and a tunneling current passes between the tip and the surface An
electronic circuit keeps this current constant by raising and lowering the tip during the
scan In this way recording the tip height at each point a three-dimensional image of local
density of electronic states (LDOS) of the surface can be obtained To explain the
extremely high resolution attainable by this simple technique quantum theory is needed
According to classical physics the current will flow between sample and tip only if they
are in physical contact If there is a vacuum gap between the two the electrons will simply
remain confined for example within the surface without the possibility of passing into the
tip
In quantum physics however the electrons have a certain probability of passing
(tunneling) across the gap appearing on the other side of the gap in this way reaching the
tip It can be shown that the probability of an electron tunneling through a gap of thickness
z is
kzep 22)0( minusprop ψ h
φmk
2= (1)
where ψ(0) is the electron wavefunction at the surface-gap border m is the electron
mass=91x10-28 g and φ is the work function of the metal (ie the energy required to
remove an electron from that material For silicon it is 48 eV) The tunneling current thus
decays exponentially with z and is extremely sensitive to topographical imperfections
present on the scanned surface A rough formula giving the current as a function of z is [1]
zFS eEVI φρ 0251)( minusprop
Fig 31 Binnig and Rohrer with the first STM Image from IBM [2]
VPalermo 25
where ρs(EF) is the local density of states at the Fermi level on the given surface For
example the formula predicts that for silicon an increase in tip-surface distance of 1 Aring
will give a 95 decrease in tunneling current
This huge dependence of tunneling current upon the distance allows detection even of the
sub-nanometre changes in height given by the single atoms of which the surface is
composed and thus to resolve them in the LDOS images Of course this description of the
tunneling process is oversimplified and for a more accurate one the electronic states of
the tip of the sample and their interaction have to be taken into account Fig 32 shows a
schematic representation of the interaction between tip and sample orbitals
The exponential decay of current with distance also yields high lateral resolution If the tip
is approximated as a sphere of radius R and the current passing at the minimum tip-sample
distance is I0 then the current passing at a lateral distance x from this point will be
Rxk
eII 22
0
2minus
=
Assuming a tip radius of 100 nm the current is concentrated in an area sim15 nm wide at tip
apex
Very sharp tips with even smaller curvature radii can be produced in several ways Simple
Fig32 Schematic view of tip-sampleorbitals interaction a) no interaction b)equilibrium c) sample positive d) tippositive [1]
26 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
tungsten wires mechanically cut are capable of obtaining atomic resolution on graphite in
air but for more disordered and rougher samples sharper and more reproducible tips are
needed
STM tips are mostly made by electrochemically etching a W or Pt-Ir wire The tips we
used were prepared using methods based on the work of Fotino [3]
A tungsten wire 038 nm diameter is immersed in a KOH solution(06M) with a thicker
tungsten wire used as a counterelectrode The cathodic and anodic reactions involved in the
etching are
Cathode 6H2O + 6 e- rarr 3H2(g) + 6 OH-
Anode W(s) + 8 OH- rarr WO42- + 4H2O + 6 e-
A potential of 3V ac is applied to the tungsten and the wire is immersed in the solution
until a constant current of sim100 mA is established The etching rapidly removes metal
shaping the wire end as a sharp tip When the potential reaches 11 V the coarse tip etching
is finished The wire is then removed from the solution carefully inserted into an
insulating plastic tube leaving only the tip exposed and re-immersed in the solution with
the tip pointing upwards A more gentle etching is thus made to reduce tip radius Usually
5 minutes etching at 07 V ac is used In this configuration very small hydrogen bubbles
formed on the tip sides sliding upwards with a ldquohoningrdquo effect on the tip
This procedure yields extremely sharp and reproducible tips at the microscopic level
After the etching the tip is thoroughly rinsed in ultrapure water then dipped into
concentrated HF to remove surface oxides and hydroxides [4] The tip is dried with
nitrogen inserted into the UHV system and degassed overnight at sim150degC
The possibility of measuring sub-nanometric distances would be useless without being able
to control tip movement over such a minute scale To scan the tip over the surface a
piezoelectric scanner is used Piezos are usually made of an alloy of PbZrO3 and PbTiO3 a
material which contracts or expands when a voltage is applied to it The Omicron
instrument used in our laboratory has three such piezo scanners for xy and z tip motion
allowing one to scan the tip over the surface with sub-Aringngstrom precision (fig 33)
To isolate the instrument from ambient vibrations the whole STM stage is suspended upon
four springs which eliminate all frequencies above 1 Hz and surrounded by a crown of
VPalermo 27
copper wings and fixed magnets Parasitic currents generated by the magnets into the
copper wings contrast every movement of the stage and efficiently block stage vibrations
The STM can be used not only to explore surface topography but to measure the IV
characteristics of single atoms or molecules on the surface (Scanning Tunneling
Spectroscopy or STS) Furthermore it can be used to modify the surface with voltage
bursts digging into it or delicately moving single molecules or atoms over a surface [5] It
can be used in vacuum in air and with proper lateral isolation of the tip even in liquids
A major drawback of STM is that it works only on conducting and semiconducting
surfaces and thus cannot be used on many surfaces of biological and chemical interest
Another instrument more suitable for these and other applications is the Atomic Force
Microscope (AFM)
Atomic Force Microscopy
AFM was invented in 1986 by Binnig Quate and Gerber after calculating the possibility of
building a cantilever with a force constant of the same order of magnitude as that of a
chemical bond
In AFM a tip mounted on a microscopic cantilever (usually made of Si3N4 fabricated with
optical lithography) is brought close to a surface When the tip touches the surface the
cantilever is very slightly deflected upwards The movement is measured by observing the
Y-PIEZO
SILICON SAMPLE
Fig 33 A picture of theSTM used for theexperiments The tripodpiezo scanner is shown
Z-PIEZO
X-PIEZO
TIP
28 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
deviation of a laser beam hitting the upper face of the cantilever Fig 34 provides a
schematic view of the principle of AFM
The typical force constant of the cantilever varies from 00006 to 2 Nm the typical
resonance frequency is 3 to 120 kHz The AFM tip can apply a force on the sample of up
to 10-9 N The AFM can be used on conductive or insulating surfaces in vacuum air or
liquids Furthermore the tip can be modified to sense electrostatic potentials (electric force
microscopy) or magnetic fields (magnetic force microscopy) it can even be functionalized
with complex molecules such as proteins to interact with biological surfaces
A drawback of the AFM is that the force it exerts can damage the surface under
observation especially if the sample is soft (as in the case of cell membranes for
example) This problem can be overcome using the instrument in tapping mode (where the
tip does not move laterally during its brief contact with the surface) or in non-contact mode
in which the tip oscillates above the surface during the scan and the changes in its
frequency due to interaction with surface are monitored The shifts in the oscillating
frequency of the cantilever due to tip-sample interaction are then used for imaging the
surface In this mode interaction of the tip with the surface is minimal and soft samples
can be imaged
STM and AFM are the main techniques used for this work A brief description of other
techniques used occasionally is given below
Fig 34 Scheme of an atomic force microscope
VPalermo 29
Low Energy Electron Diffraction (LEED)
Electrons with energies in the 20-500 eV range are diffracted by a crystalline surface the
diffraction peaks are visualized on a fluorescent screen This technique probes the long
range order of the surface up to a depth of several nanometres
X-Ray Photoelectron Spectroscopy (XPS)
XPS allows both qualitative and quantitative chemical analysis of the elements present on
or near the sample surface
An X-ray source is used to photoionize the atoms on a surface and produce photoelectrons
By measuring the kinetic energy of the photoelectrons the binding energy of the electronic
levels can be calculated This energy will depend on the chemical environment of the
surface atoms
Although the soft X-rays used penetrate to a depth of ~2000Aring the sampling depth of the
technique is determined by the mean free path of the photoelectrons which allows their
escape from only the first 10-100Aring
Secondary Ion Mass Spectroscopy (SIMS)
High and low energy ions (primary ions) are used to bombard a sample and remove surface
atoms and ions The ionic fragments removed (secondary ions) are then analysed by a mass
spectrometer The surface can be consumed during the measurement and profiles obtained
giving concentrations of the materials composing the sample at different depths (depth
profiles)
A popular variant of SIMS is TOF-SIMS In this technique the secondary ion masses are
measured by a time-of-flight (TOF) measurement The secondary ions generated by the
bombarding primary ions are accelerated to a constant kinetic energy and then move
through a field-free space before they reach the detector where their intensity is measured
as a function of flight time Since ions with different masses have different velocities at a
30 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
given kinetic energy the measured flight times of the ions can easily be converted to their
masses The static nature of this latter technique allows mass spectroscopy surface analysis
with minimal damage to the surface
Bibliography
[1] Chen CJ Introduction To Scanning Tunneling Microscopy (Oxford University Press
Oxford 1993)
[2] From wwwibmcom
[3] Tip Sharpening By Normal And Reverse Electrochemical Etching Fotino M Review
Of Scientific Instruments 64 (1) 159-167 Jan 1993
[4] A Convenient Method For Removing Surface Oxides From Tungsten STM Tips
Hockett LA Creager SE Review Of Scientific Instruments 64 (1) 263-264 Jan 1993
[5] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
VPalermo 31
4 Surface modification of silicon in liquid
Nano-hole creation
Liquid treatments of silicon wafers are very common in the integrated circuit (IC)
manufacturing industry They are used to clean and improve surface uniformity to create
and etch protective oxide layers and to remove photo-resist layers
Crystalline silicon with its native oxide layer is very stable and is resistant to many acids
It is easily attacked by hydrofluoric acid (HF) and alkaline solutions
The thin (~2 nm) passivating layer of native oxide (SiO2) is formed on exposure to the
atmosphere This surface layer contains many defects and contaminants so it is usually
chemically stripped and substituted with a better chemically-formed protective oxide
The most common silicon cleaning procedure is the RCA method named after the Radio
Corporation of America [1] It consists of two steps in the first one the surface is treated
with a hot alkaline solution (H2OH2O2NH4OH 411) to remove particles from the
surface following this a hot acidic solution (H2OH2O2HCl 411) is used to remove
metal contamination Other well-known cleaning methods are IMEC (a sequence of
cleaning steps in H2OO3 and dilute HF) or the Pirana etch (a hot 41 mixture of H2SO4
H2O2)
32 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The standard RCA clean removes surface contaminants etches the native oxide and
oxidizes the silicon surface leaving a uniform layer of silicon oxide which better protects
the surface from further contamination
Etching with fluorine-based solutions
Hydrofluoric acid is one of the most common reagents used in the treatment of silicon
wafers both in the research field and in industrial processes A rapid dip in dilute HF is the
simplest way to remove the native oxide from Si(100) and leaves the surface passivated by
a layer of Si-H bonds Because of the low polarization of Si-H bonds the Si-H layer is
stable even for several days protecting the surface from contamination It has often been
assumed that this short etch does not significantly change the surface morphology of the
silicon substrate[2] even though a prolonged dip in dilute HF leads to surface roughening
[3]
Although dilute HF roughens the Si(100) surface at the atomic scale [45] immersion in
concentrated HF (49) etches the surface oxide without attacking the Si surface
uncovering in this way the buried SiSiO2 interface The final counter-intuitive result is
that dilute HF etches the silicon while concentrated HF leaves the crystalline silicon
untouched [3]
Etching Si with fluorine-containing solutions at different concentrations and pH can
produce different morphologies from rough surfaces to flat nearly ideal Si-H terminated
surfaces
Hessel et al and Higashi et al demonstrated in 1991 that very flat Si(111) surfaces can be
obtained using 40 NH4F while etching with HF always yields rough surfaces The
surface becomes smoother because the etchant rapidly attacks Si atoms at step borders
thus removing surface kinks and irregularities in a step-flow mechanism [6 7] Later on
even smoother and more perfect surfaces were obtained by removing oxygen from the
solution after it was discovered that oxygen dissolved in 40 NH4F initiates the formation
of triangular etch pits It was not possible to obtain flat surfaces by etching Si(100) with
ammonium fluoride solutions which leads to the formation of small 2x1 dimer-row
reconstructed (100) terraces together with (111) facets [8]
This difference is caused by the different hydride terminations prevailing on the (100) and
(111) faces While the ideal Si(111)-H surface is monohydride terminated the more
VPalermo 33
reactive dihydrides predominate on the Si(100)-H surface making it more vulnerable to
etching The etching reaction is thus strongly anisotropic etching (100) facets faster than
(111) thus producing (111) microfaceting on Si(100) crystals
A more uniform Si(100) surface can however be prepared by etching at low pH with an
HFHCl mixture [9] or by using very dilute HF solutions and ultrapure water with low
dissolved oxygen and carbon contents [10]
Electrochemical etching can also be used applying anodic or cathodic bias to the silicon
to obtain different morphologies [11] by varying the potential isotropic or anisotropic
etching is observed The aforementioned results show that despite the simplicity of the
reactants fluoride etching of silicon is quite a complex reaction
Fig 41 Chemical etching of silicon
HO+H
H HO
H
H
H
FSi
Si
SiSi
+H2O
-OH-
+F-
H2O H H
Si H
F OH
H Si
Si Si
-H2
+H2O
H
H
H
FSi
Si
SiSi
H
H
H
OHSi
Si
SiSi
H
H
H
H Si
Si
Si Si
Etching mechanism of silicon
HF rapidly dissolves the SiO2 passivating layer on silicon leaving the surface almost
completely hydrogenated [12] After this two different types of reactions etch the silicon
simultaneously one chemical and the other electrochemical [13] The overall etching
mechanism can be schematized in two stages (see Figure 41)
i) Si-H bonds are substituted by Si-F or Si-OH bonds creating a partial charge on the
surface silicon atom and polarizing its Si-Si backbonds
ii) these polarised backbonds are then more easily attacked by HF or H2O After
rupture of the Si-Si bond the atom is removed leaving behind new Si-H
terminations and the reaction can start again
34 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
These reactions take place although at different rates on both Si(100) and Si(111)
Stage i) is usually the rate-determining step of the reaction and the stability of the Si-H
bonds depends upon the pH the concentration of nucleophilic species in solution and an
eventual potential applied to the crystal
For pH gt5 as in the case of concentrated NH4F solutions reaction begins with attack by
water to give Si-OH (step ArarrC) The -OH group is rapidly substituted by fluorine with
polarization of the underlying Si-Si bonds These bonds are then easily attacked by water
the silicon atom being released into solution as HSiF(OH)2 The Si-OH rarr Si-F substitution
is not fundamental for the reaction and etching can proceed even for Si-OH terminated
atoms but XPS measurements showed the presence of a certain number of Si-F bonds
remaining Furthermore fluorine seems to have a catalytic effect on Si-H substitution as
indicated by the dependence of the etch rate upon the F- concentration at least for pH
values between 4 and 8
Si-F bonds can be easily removed by a water rinse In the case of strongly alkaline
solutions (pH=14) OH- groups act directly as nucleophiles and no fluorine is needed to
catalyze Si-H bond rupture
At pH lt4 almost no free F- ions are present in solution and the etch rate is very small at
pH lt2 all etching reactions are very slow and this explains the stability of Si crystals in
concentrated (50 ww) HF solutions
Si Si
Si Si
H
H
H
H Si
Si
Si Si
H
H
H Si
Si
SiSi
H
H
H
OH-H+ -e
(F-)
-H+
-e
+H2O hellipas in fig1
Fig 42 Electrochemical etching of silicon by H2O
The electrochemical etching of silicon involves electron transfer from the surface atoms to
the valence or conduction band of the crystal (according to the type of doping of the
silicon) Several studies have been carried out at different electric potentials In the case of
p-silicon or n-silicon at anodic potentials where positive charges (h+) are available in the
crystal a silicon-centred radical can be created by capture of a hole from the bulk (fig
42) The F- ion does not participate in the reaction but is thought to contribute through
VPalermo 35
electrostatic interactions by lowering the energy of the interaction step After the
formation of the Si-OH group the reaction proceeds as shown in the scheme of fig 41
Matsumura et al [4] proposed that not only water but HF2- molecules also play a major role
in electrochemical etching of silicon leaving on the surface Si-F terminated bonds which
can be immediately attacked in an autocatalytic process (fig 43)
In the electrochemical reactions described above an external potential is applied to the
silicon crystal The chemical and electrochemical reactions in any case take place
simultaneously most of the time with the chemical path predominating at high pH Even
when no external potential is applied to the silicon partial electrochemical reactions can
take place at different ldquocathodicrdquo and ldquoanodicrdquo sites on the surface with an internal charge
exchange which ensures neutrality [11] This macroscopic silicon etching and hydrogen
bubble formation can lead to visually observable patterns on the surface when Si(100) is
immersed in ammonium fluoride even without applying a potential
Si Si
Si Si
H
H
H
F Si
Si
Si SiH
F F-H+ -2e
+HF2-
H
HF
SiSi
SiSi + F
F F
FH
H
H
F -H+ -2e
+HF2-
Fig 43 Autocatalytic electrochemical etching of silicon by HF2-
Inhomogeneities on silicon surfaces caused by electrochemical reactions and charge
transfer have been studied extensively because they are of fundamental importance in the
formation of porous silicon
Pore formation on silicon
When Si(100) or Si(111) are etched under anodic bias in fluorine-based solutions
microscopic pores form on their surface Several different morphologies of pores have
been observed with pore diameters ranging from 10 nm to several microns with depths of
several microns [14] Pore shape is very variable too ranging from ordered straight pores
to chaotic networks of branched pores (fig 44) Porous silicon has been known since the
fifties but it was only in 1990 that interest in this material increased following the
discovery that porous silicon layers were able to emit bright red light This led to a large
amount of research and now different classes of micropores can be reproducibly created
36 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Fig 44 Different types of Silicon micropores From ref [14]
mostly for optical and micromachining applications However there is still no unified
theory able to explain the nucleation and growth mechanism of all the different kinds of
pores
We will give a short description of some of these theories for more detail see Parkhutik et
al [15]
One model explains pore nucleation on the basis of physical processes such as hole
positive charge migration ion transport to the surface and small perturbations on the
silicon surface modelled as Fourier components The system is shown to be unstable and
some spatial frequencies that lead to pore nucleation evolve from the etching process
A second model focusses on stationary pore growth without explaining the nucleation
stage According to this model silicon dissolves preferentially at pore edges because h+
charges are attracted by the stronger electric field present at these edges
A third class of models explains pore growth as a Diffusion Limited Aggregation (DLA)
process where the random walk of h+ charge carriers through the depleted layer present at
the silicon-liquid interface controls pore shape
Finally the model by Carstensen Cristophersen and Foll [16] proposes that areas of the
surface of some characteristic size LCO are etched by synchronized ldquocurrent burstsrdquo in the
flow of h+ charges These bursts dissolve silicon through cyclic stages of surface oxidation
oxide removal and hydrogen passivation Areas where a burst has recently taken place are
less passivated and thus more likely to be etched again in this way the pore bottom
continues to dissolve while the pore walls are passivated and are thus less favourable areas
towards current bursts
VPalermo 37
EXPERIMENTAL RESULTS
In the following sections we will show some experimental results obtained from STM and
AFM measurements of fluorine-treated Si(100) surfaces In the first part the results of
mild etching using concentrated and dilute HF solutions at low pH are presented In the
second part the results of etching at high pH using ammonium fluoride are presented and
the mechanism of pore formation discussed
Etching of Si(100) in dilute and concentrated HF
Samples were cut from different areas of an 8-inch diameter p-doped silicon(100) wafer
(10 Ω-cm) supplied by MEMC Electronic Materials Each series of STM measurements
was carried out over at least six different areas on at least two identical samples Low
Electron Energy Diffraction (LEED) was used to check the surface cleanliness of the
samples before STM measurements
Table 41 summarizes the different treatments of each sample After etching with
electronic grade HF each sample underwent a final rinse in Ultra-Pure Water (UPW
resistivity gt18 MΩ-cm) Both the HF and the UPW were allowed to flow continuously
over the sample surface Some samples were not etched with HF at all but just washed with
UPW to observe the morphology of the native oxide layer (~2 nm thick) covering the
surface All of these processes were carried out under nitrogen to limit reoxidation and the
samples were then introduced from the nitrogen atmosphere directly into the vacuum
chamber and degassed overnight at sim150degC before LEED and STM measurements
STM images were obtained from each sample using the same measurement parameters
(sample bias 4 V feedback current 1 nA scan speed 800 nm s-1) These parameters
38 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
provided a satisfactory level of reproducibility for all the samples Measurements were
made over an area of 500x500 nm (image size 500x500 pixel) Slope correction was
carried out by subtracting row-wise and column-wise fitted slopes from the entire image
which gave better results than the simple subtraction of a fitted plane especially for the
rougher samples Following slope correction the rms roughness
sum minus=xy
hyxhN
22 ))((1σ
and the 2-D Fourier transform
)(22
2
)(4
)( vyuxi
xyeyxhvuF +sum∆
= π
π
were calculated for each image where N2 is the number of pixels composing the image
h(xy) is the surface height at each point ∆ is the distance between points h is the mean
height and u v are the spatial frequencies The radial power spectrum PS(f) of the STM image data is obtained from the angular
average of the squared Fourier transform with f 2 = u2 + v2
Fig 45 shows the STM images obtained from the various samples Sample A still covered
with its native oxide layer shows an irregular surface with RMS roughness of ~05 nm
(see Table 41) Observing the sample with LEED gave no diffraction pattern even at
relatively high incident electron energies because of the surface oxide coverage After 1
min etching in dilute HF (sample B) the morphology is similar to the original one though
Table 41 Sample treatments average RMS roughness and integrated area of power spectra
Sample Treatment RMS roughness (nm)
PS area (f lt 01 nm-1)
PS area (f gt 01 nm-1)
A Rapid dip in water 053 plusmn 013 366 032
B 1 min in HF 5 + 10 min in water 051 plusmn 008 272 034
C 30 min in HF 5 + 10 min in water 062 plusmn 008 1064 038
D 5 sec in HF 49 + 10 min in water 042 plusmn 004 183 022
VPalermo 39
Fig 45 STM images of each group of samples showing the topography of the silicon surface A) noetching original oxide surface B) after 1 min etching in dilute HF C) after 30 min etching in diluteHF D) after dipping in concentrated HF Grey scale indicates height of the surface from lower (black) to higher (white) The images are 250x250 nm ie representative portions of the images usedfor the roughness measurement and PSD analysis
some of the larger features have disappeared and the image quality is better maybe due to
improved tunnelling due to the cleaner surface The RMS roughness is comparable to that
of the original surface Clear diffraction patterns are visible using LEED though at quite
high energies (200 eV) After prolonged etching (sample C) the RMS roughness increases
to 062 nm and a long-range corrugation is visible on the surface even if the LEED pattern
is good
The samples dipped in concentrated HF (D) reveal the bare SiSiO2 interface which has a
disordered aspect and protrusions over a wide range of dimensions The quality of the
STM images of sample D is very good probably due to the cleanliness of the surface
40 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 E -0 2
1 E -0 1
1 E + 0 0
1 E + 0 1
1 E + 0 2
1 E + 0 3
0 0 0 0 0 1 0 1 0 1 0 0f (1 nm )
nm^4
A a s re c e iv e d
B e tc h e d 1 min
C e tc h e d 3 0 min
D e tc h e d HF 4 9
Fig 46 Log-Log plot of the averaged power spectra of the STM images for all the samples
which gives a more stable tunnelling junction The LEED pattern is excellent showing
clear diffraction peaks at energies as low as 37 eV comparable to that obtained after high
temperature cleaning in UHV
Fig 46 shows the power spectra of the samples The high frequency and low frequency
areas of the power specturm are considered separately Table 41 shows for each sample
together with the roughness the integrated area of the power spectrum for the high and low
frequency part
We first examine the differences between the samples in the low frequency part of the
spectrum (lt01 nm-1) In this spatial range the short dip in concentrated or dilute HF
(samples BD) seems to lower the roughness of the sample removing some of the larger
features of the native oxide surface visible in Fig 45a and thus decreasing the
corresponding part of the power spectrum Sample C however etched with HF for 30 min
shows an increase in roughness peaking at 0012 nm-1 (~85 nm corresponding to the
typical dimensions of the corrugations visible in Fig 45c) On this scale sample C has a
PS density of 435 nm4 compared to 101 nm4 for sample A
Samples A B and C have the same PS in the high frequency range showing that the fine
structure of the surface is not changed by the HF etching Sample D on the other hand
shows a decrease of roughness for frequencies gt 015 nm-1 This effect could be due to the
VPalermo 41
improved surface cleanliness after etching with concentrated HF which would give a more
stable STM junction thus reducing the high frequency noise in the image
Fig 45 and the analysis of the power spectrum of each sample shows that a rapid dip in
HF removes the native oxide but does not lead to major changes in the morphology of the
Si surface its only effect being the removal of some of the larger features present on the
original surface Prolonged etching on the other hand increases the RMS roughness of the
surface
Etching of Si(100) in concentrated ammonium fluoride and nano-hole creation
Two different types of commercial p-doped Si(100) wafers (2Ω-cm and 10Ω-cm) from
MEMC were used Several different samples of 10x5 mm were immersed for 10 minutes
in 40 electronic grade NH4F solution under agitation Previous works used low
temperatures or anodic potentials applied to the silicon to avoid gaseous hydrogen
production and to obtain a uniform surface but in our experiment we worked at room
temperature to check the influence of hydrogen bubbles on surface morphology During
the etching the stirring was sufficient to provide a uniform concentration of reagents over
the whole sample surface but not to mechanically remove the hydrogen bubbles from the
silicon surface
After the etching each sample was rinsed in ultra-pure water to remove any etching
residues and observed with STM AFM and optical microscopy The AFM measurements
were made in air while for STM measurements the samples were rapidly dried with
nitrogen and inserted into the vacuum system to avoid surface reoxidation After insertion
into the vacuum surface cleanliness was checked with LEED and the surface morphology
observed by STM Parameters for STM measurements were sample bias +4 V 1 nA
current 16 Hz scan rate The images obtained were stable and reproducible over several
days Scan parameters for AFM were 20 nN force and 1 Hz scan rate
Some of the samples were cleaned with an RCA standard clean [1] before NH4F etching to
check the influence of possible surface contaminants on the final results Eight different
samples were prepared and more than sixty STM images of the samples were taken at
different points of the various samples
42 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
(110)
(110)
Fig47 abc) STMtopographic images of differentetching morphologies Eachimage is 500x500 nm Z-ranges are 10 10 and 18 nmrespectively d) STM image of a bridgecreated by etching of the lowerlayers of silicon (black arrow)Image is 250x250x6 nm
After ~2 min of immersion in the solution hydrogen bubbles become visible on the sample
surface The production is slow and the bubbles are quite stable on the sample without
detaching Thus some areas of the surface are masked from the liquid etching action
STM observations (fig 47) show that at the nanometer scale the surface is unevenly
covered with holes of radii ranging from 10 to 200 nm with depths of 2-4 nm These holes
have a wide range of different shapes and distributions In most cases the surface was
covered with a uniform distribution of round-shaped holes (fig47a) indicating isotropic
etching The dimension and the density of the holes changed greatly from sample to
sample and even over the surface of a single sample In some cases the etching was
anisotropic yielding nearly square holes and layered structures as shown in fig 47b
Square holes have been previously observed in cases where the etching speed in the (110)
direction is significantly smaller than in the (100) direction [17]
Over these areas (fig 47b and especially 47c) it is clearly visible how once the surface
had been attacked the reaction continued to preferentially remove atoms at step
irregularities (kink atoms) straightening step edges Eventually the exposed underlying
silicon was also attacked and further holes created inside the previously etched larger
ones It was not possible to detect monatomic steps on this kind of surface The smallest
step height observed was ~15 nm corresponding to several atomic layers In the image
shown in Figure 47a the etching was not strong and created only anisotropic holes on the
surface In fig 47b and c the stronger etch proceeded laterally for several tens of
VPalermo 43
nanometres leaving straight steps several tens of nanometers long and roughly rectangular
holes as expected given the structure of the (100) crystal face In some cases a
significative underetch is observed and the formation of suspended bridges and tunnels
can be deduced from the STM images (fig 47d)
The formation of branched pores and suspended structures has been attributed during pore
formation to diffusion limited aggregation effects where the h+ charge carriers necessary
for silicon etching have a higher probability of reacting at pore bottoms than reaching the
upper part of the silicon surface In the case of very deep pores quantum wire effects have
been invoked to explain the pore growth mechanism [15] In our case though the pores
formed were very shallow the underetch depth being only a few nanometres on pores of
sim100 nm width Thus more than diffusion effects the main contribution to the
underetching process must come from anisotropic etching and some kind of autocatalytic
reaction path analogous to the one described by Matsumura et al [4] with some areas of
the silicon surface hydrogenated and thus less vulnerable to etching
Pre-treatment with RCA cleaning has no effect on the final morphology and this seems to
exclude pore nucleation being caused by presence of metallic or organic surface
contaminants
The morphology and the distribution density of the pits was quite uniform over
microscopic areas of the sample but changes were observed over the millimetre scale This
suggests that etching intensity is influenced by some large-scale parameter
Large-area measurements made with AFM or with an optical microscope (fig 48)
showed that the inhomogeneity of surface etching can be correlated with the masking
action of the bubbles While the fluoride dissolved the silicon hydrogen bubbles formed
by the reaction covered some areas of the surface thus blocking the etching over that area
generating macroscopic steps at the bubble-liquid border As the reaction proceeded more
hydrogen accumulated and the bubble diameter increased producing in this way a circular
pattern of steps The increase in bubble diameter was not continuous with time otherwise a
uniform surface slope gradient would have been obtained The formation of this circular
ldquoetching staircaserdquo indicates that the bubble growth was stepwise the bubble accumulating
more and more hydrogen without enlarging across the surface until it relaxed increasing
its diameter stepwise and covering more silicon The circular structures in fig 48a are not
co-axial and their asymmetry could derive from physical processes due to stirring or
irregularities on the surface
44 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
a b Fig 48 a) optical micrograph of etching patterns on Si(100) created by NH4F 12x09 mm b) AFM image of the circles border xy range is 40x24 microm z-range is 30 nm
The step structure was not destroyed by the etching even after the bubble detached from
the surface but on the contrary the etching process seemed to be influenced by the
presence of the step
Observing in detail a series of steps (fig 48b) a quite deep trench is visible at the base of
each step A close-up image of a step and the corresponding line profile of the trench is
shown in fig 49 The trench is sim5 nm deep with respect to the lower surface compared to
a step height of 22 nm
A similar structure has been recently obtained with electrochemical etching of p-type
Si(100) in 4 HF [18] in which a ldquocurrent burstrdquo etching model previously described
was assumed for silicon dissolution In that experiment the trench was created at the
border of silicon nitride masks and began to grow after a nucleation stage Preferential
trench etching was along the (110) direction and the trench growth was explained as an
effect of mechanical stress induced by the nitride mask and of electric field enhanced
dissolution which depended upon an external applied potential
While it is clear that in our system the gas bubbles have a masking effect similar to a
classical solid nitride mask it is unlikely that hydrogen present on the surface can induce a
significant stress in the silicon lattice as in the case of a nitride mask Furthermore no
external field was applied to drive preferential etching at the trench site
It has been proposed [19] that the cathodic and anodic part of the etching reaction
(hydrogen production and silicon oxidative etching respectively) take place at different
points on the surface with a net charge transfer between the different areas In this case
the highest reaction rates will correspond to the silicon area surrounding the bubble border
where a high number of positive charges will be available for the reaction Furthermore a
sharp trench extending into the silicon crystal will be a preferential electrostatic attractor
VPalermo 45
Fig 49 AFM image of the etched surface showing a step created on the surface by bubblemasking A stronger etching action is visible on the right side of the step as well as theprotected area on the upper side of the step (indicated by the arrows) Image is 10x10 micromz-range is 30 nm The profile on the right is taken from the central area of the image
for the h+ charge carriers coming from other ldquocathodic areasrdquo of the sample either from
other regions on the surface or from the back of the silicon chip [18]
In the areas where the hydrogen bubble had detached and the surface was exposed to the
etching the reaction was not uniform in the neighbourhood of the steps It is possible to
observe (fig 48b and 49) an area on the upper side of the step where less or even no
etching at all seems to have taken place as if the step was able to protect the surface from
etching While etching on the lower side with trench formation can be attributed to the
presence of the bubble the surface on the upper step side can be etched only after bubble
detachment so no masking effect can account for this result However a further
preferential attraction of h+ charge carriers from the already formed trench can be
hypothised electrochemically shielding the surrounding area from further etching If this is
true the shielding effect would be very strong with a relatively shallow 5 nm-deep trench
protecting an area of sim1 microm parallel to the step
To summarise the etching of Si(100) in NH4F is a complex process in which different
reaction paths both chemical and electrochemical co-exist Hydrogen bubbles formed by
the reaction act as a mask on the surface and create etching paths and inhomogeneous
etching of the surface Different kinds of pores are observed on the surface and in some
cases the anisotropy of the process is so strong as to give square-shaped holes and
underetching
46 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The diffusion of h+ charge carriers in the crystal is one of the main rate-determining steps
of the reaction and leads to the formation of a deep trench immediately outside the bubble
perimeter These trenches act as charge collectors and reduce the etching of the upper step
surface in the proximity of the steps
Bibliography
[1] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[2] Spectroscopic Ellipsometry Studies Of HF Treated Si (100) Surfaces Yao H Woollam
Ja Alterovitz SA Applied Physics Letters 62 (25) 3324-3326 Jun 21 1993 Influence
Of HF-H2O2 Treatment On Si(100) And Si(111) Surfaces Graf D Bauermayer S
Schnegg A Journal Of Applied Physics 74 (3) 1679-1683 Aug 1 1993 Kinetics Of
Oxidation On Hydrogen-Terminated Si(100) And (111) Surfaces Stored In Air Miura
T Niwano M Shoji D Miyamoto N Journal Of Applied Physics 79 (8) 4373-4380
Part 1 Apr 15 1996
[3] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995 Structure Of The Stepped SiSiO2 Interface After Thermal-Oxidation -
Investigations With Scanning Tunneling Microscopy And Spot-Profile Analysis Of
Low-Energy Electron-Diffraction Pietsch GJ Kohler U Jusko O Henzler M Hahn
PO Applied Physics Letters 60 (11) 1321-1323 Mar 16 1992
[4] Enhanced Etching Rate Of Silicon In Fluoride Containing Solutions At pH 64
Matsumura M Fukidome H Journal Of The Electrochemical Society 143 (8) 2683-
2686 Aug 1996
[5] A Study Comparing Measurements Of Roughness Of Silicon And SiO2 Surfaces And
Interfaces Using Scanning Probe Microscopy And Neutron Reflectivity Crossley A
Sofield CJ Goff JP Lake ACI Hutchings MT Menelle A Journal Of Non-Crystalline
Solids 187 221-226 Jul 1995
VPalermo 47
[6] Step-Flow Mechanism Versus Pit Corrosion - Scanning-Tunneling Microscopy
Observations On Wet Etching Of Si(111) By Hf Solutions Hessel HE Feltz A Reiter
M Memmert U Behm RJ Chemical Physics Letters 186 (2-3) 275-280 Nov 8 1991
[7] Comparison Of Si(111) Surfaces Prepared Using Aqueous-Solutions Of NH4F Versus
HF Higashi GS Becker RS Chabal YJ Becker AJ Applied Physics Letters 58 (15)
1656-1658 Apr 15 1991
[8] Wet Chemical Etching Of Si(100) Surfaces In Concentrated NH4F Solution -
Formation Of (2x1)H Reconstructed Si(100) Terraces Versus (111) Faceting Neuwald
U Hessel HE Feltz A Memmert U Behm RJ Surface Science 296 (1) L8-L14 Oct
10 1993
[9] Ideal Hydrogen Termination Of Si(001) Surface By Wet-Chemical Preparation Morita
Y Tokumoto H Applied Physics Letters 67 (18) 2654-2656 Oct 30 1995
[10] Atomic Structures Of Hydrogen-Terminated Si(001) Surfaces After Wet Cleaning
By Scanning Tunneling Microscopy Endo K Arima K Kataoka T Oshikane Y Inoue
H Mori Y Applied Physics Letters 73 (13) 1853-1855 Sep 28 1998
[11] On The Potential-Dependent Etching Of Si(111) In Aqueous NH4F Solution
Houbertz R Memmert U Behm RJ Surface Science 396 (1-3) 198-211 Jan 20 1998
[12] Etching Process Of SiO2 By HF Molecules Hoshino T Nishioka Y Journal Of
Chemical Physics 111 (5) 2109-2114 Aug 1 1999
[13] Etching Mechanism And Atomic-Structure Of H-Si(111) Surfaces Prepared In
NH4F Allongue P Kieling V Gerischer H Electrochimica Acta 40 (10) 1353-1360
Jul 1995
[14] Pore Formation Mechanisms For The Si-HF System Carstensen J Christophersen
M Foll H Materials Science And Engineering B-Solid State Materials For Advanced
Technology 69 23-28 Sp Iss Si Jan 19 2000
[15] Porous Silicon - Mechanisms Of Growth And Applications Parkhutik V Solid-
State Electronics 43 (6) 1121-1141 Jun 1999
[16] Formation And Application Of Porous Silicon Foll H Christophersen M
Carstensen J Hasse G Materials Science amp Engineering R-Reports 39 (4) 93-141 Nov
1 2002
[17] Pore Morphology And The Mechanism Of Pore Formation In N-Type Silicon
Searson PC Macaulay JM Ross FM Journal Of Applied Physics 72 (1) 253-258 Jul 1
1992
48 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
[18] Deep Electrochemical Trench Etching With Organic Hydrofluoric Electrolytes
Christophersen M Merz P Quenzer J Carstensen J Foll H Sensors And Actuators A-
Physical 88 (3) 241-246 Mar 5 2001
[19] Morphology Of Anodically Etched Si(111) Surfaces - A Structural Comparison Of
NH4F Versus HF Etching Houbertz R Memmert U Behm RJ Journal Of Vacuum
Science amp Technology B 12 (6) 3145-3148 Nov-Dec 1994
VPalermo 49
5 Surface modification of silicon in vacuum void creation and oxide desorption
The main reason for the huge success of silicon in the microelectronics industry is not due
to its superior properties as a semiconductor Other materials for example germanium
have better qualities such as higher mobility of charge carriers and lower noise levels
which would allow the construction of faster and higher performance devices
The widespread use of silicon however is mainly due to the outstanding characteristics of
its oxide Silicon dioxide (SiO2) is a very good electrical insulator easy to form
chemically and thermally stable and is compatible with lithographic and metal deposition
processes Germanium oxide on the contrary is too reactive to be used
Even the use of Si(100) substrates for nearly all microelectronic devices is dictated by
oxide quality The (111) face of silicon crystal can be easily cleaved and flattened and
almost atomically perfect surfaces can be obtained with simple chemical procedures (as
described above) But the density of interfacial defects is highest for Si(111)-SiO2
interfaces and lowest for Si(100)-SiO2 ones so microchips will continue to be fabricated
on Si(100) wafers
SiO2 (silica) is present in 95 of the earthrsquos minerals in different allotropic forms such as
quartz tridymite and cristobalite In the bulk each silicon atom is bonded to four oxygens
in a Si-O-Si tri-dimensional network Si-O bonds are 016 nm long and form an angle
ranging from 120deg to 150deg
Three typical intrinsic defects are present in SiO2 The so-called Ersquo centres are oxygen
vacancies with a hole localised on a silicon atom with only three Si-O bonds
50 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
O3Si +SiO3 Whereas the PR (peroxy radical) defects are holes trapped by a charged
peroxy moiety with a O3Si-O-O+ SiO3 structure The NBOHC (non-bridging oxygen hole
centres) derive from water or hydrogen contamination and are schematized as O3Si- O- H-
O-SiO3
The atomic structure of the Si-SiO2 interface varies enormously Local domains resembling
the tridymite and the cristobalite structure of silica are present but it seems that only 10
of the interface is ordered [1] Far from the interface the SiO2 bulk is completely
disordered The passage from bulk Si to stoichiometric SiO2 passes through a non-
stoichiometric SiOx layer 07 nm thick
When a clean silicon surface is exposed to atmospheric oxygen a thin sim2 nm thick layer
of native oxide forms spontaneously which is usually removed and substituted with
thicker better quality oxide layers before further processing
Silicon is usually oxidised by thermal annealing at temperatures between 800deg and 1100deg
in an atmosphere of pure O2 with some water eventually added to increase oxidation speed
Thermal oxides made in pure oxygen (dry oxides) grow more slowly than oxides produced
in an oxygen-water atmosphere (wet oxides) but are usually of better quality
According to the Deal-Groove formula the time t needed to grow an oxide of thickness X
is given by [1]
1
212minus
minusminus
+=
ABXBXt α
where the constant B and BA decrease exponentially with temperature as
minus
kTEexp
with activation energies for dry oxidation of EB =123 and EBA =20 eV respectively
EB is related to the diffusion of oxygen in silicon while the value of EBA is interpreted as
the energy required to break a Si-Si bond The exponent α is 1 for wet oxidation and 0 for
oxidation at high temperatures and low oxygen pressures It has intermediate values for dry
oxidation This formula does not work well for low values of X at the initial stages of
oxidation and usually empirical corrections are used
An interesting characteristic of silicon is that at high temperature and in vacuum oxygen
can actually etch the silicon crystal giving gaseous products and the oxidized layer present
on the crystal becomes unstable (fig 51) [2]
VPalermo 51
1E-10
1E-09
1E-08
1E-07
1E-06
1E-05
1E-04
1E-03
1E-02
1E-01
0607080911112
1000T (1K)
P (T
orr)
SiO2 + Si rarr 2SiO(g) Oxide decomposition
Si+O2 rarr SiO2(s) oxide formation Fig51 phase diagram of the
oxygen-silicon system
Silicon oxidation apart from the initial nucleation stages at the monolayer level proceeds
uniformly over the whole surface with a planar reaction front moving from the surface
into the bulk
If heated under low oxygen partial pressure (vacuum or inert atmosphere) SiO2 is known to
decompose following the reaction
SiO2 + Si rarr 2SiOuarr (1)
The reaction begins with nucleation at defect points on the SiSiO2 interface and proceeds
in a spatially inhomogeneous manner with the formation of large voids on the oxide
surface [3]
Several studies have been made on the dynamics of void growth both on thick [4] and thin
[5] layers of SiO2 The process has been used to decorate otherwise unobservable defects at
the SiSiO2 interface [6] or to grow nanoislands of silicon on the void surface [7] It has
been suggested that the defects acting as nucleation centres could be metallic contaminants
present on the native surface which aggregate and catalyze SiO2 decomposition [8]
After oxide desorption the silicon surface is very rough In particular on Si(100) square
islands are observed several nanometres high which act as lsquopinning sitesrsquo on the motion of
monatomic steps on the Si surface It has been proposed that these islands can be composed
of silicon carbide coming from the organic contaminants present on the surface [9] or of
oxide clusters during partial reoxidation [10]
In the following section some experimental results obtained studying voids and nanoisland
growth are presented
52 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nanoisland growth on silicon
We used two different samples the first rougher than the second (i) p-type Czochralski Si
(100) with a RMS roughness of 05 nm shown in fig52a and (ii) p-type epitaxial Si(100)
with a RMS roughness of 019 nm shown in fig 52b The roughness was measured with
STM on the lsquoas receivedrsquo samples on different areas of 500x500 nm Every sample was
covered by a layer of native oxide ~2 nm thick
Some of the samples (AC) were introduced into the UHV system without any cleaning
others (BD) were dipped in concentrated HF (49) to remove the surface oxide without
etching the silicon [11]
After degassing the samples were heated resistively increasing the temperature slowly to
keep the pressure within the 10-10 mbar range during heating Each sample was held at
900˚C for 30 min For some of the samples the heating was stopped at lower temperatures
to obtain incomplete oxide desorption and surface roughening thus enabling observation
of the different steps of the process Table 51 summarizes the different treatments for each
group of samples
Fig52 Original surfaces of Czochralski (a) and epitaxial (b) silicon Every image is 500x500 nm
VPalermo 53
Table 51 Summary of different treatements
Silicon type Oxide removed
Heating Islands density (microm-2)
Av Island Volume (nm3)
A Czochralski No 30rsquo at 900˚C 24x103 353
B Czochralski Yes ldquo 20x104 71
C Epitaxial No ldquo 28x102 2450
D Epitaxial Yes ldquo 23x104 63
Fig53 ab) Surface roughening on group A samples after heating 30 min at 800deg and900deg respectively in presence of an oxide layer Image size 500x500 nm cd) Surface roughening on group B samples after heating 30 min at 700deg and 900degrespectively after removing the oxide layer with HF Image size 250x250 nm
Fig 52a is the native oxide surface of sample A quite disordered and irregular This
surface was stable when heated to 700˚C while at higher temperatures the oxide began to
desorb Fig 53a was taken after heating the sample at 800˚C for 30 min
54 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The oxide began to desorb in correspondence with defect points creating oxide-free
surface voids which enlarge radially uncovering the silicon surface Some surface silicon
atoms are removed through reaction (1) Mobile silicon atoms moving randomly over the
surface aggregate forming a nanocrystal at the initial defect point surrounded by a stable
flat oxide-free surface with some steps The initial SiSiO2 interface as observed after
oxide etching with HF [12] is similar to the one shown in fig 52a and thus the surface
observed in the void area is a completely new one coming from surface reorganization
The reaction continues leading to a coalescence of the voids the final result of the process
is shown in fig 53b with protruding islands aligned along the main crystal axes and flat
areas with some steps LEED measurements confirmed that this surface is crystalline
silicon Atomic resolution was obtained on the flat areas of the surface showing the
typical 2x1 reconstruction of Si(100) Figures 53c and 53d show the evolution of the
roughening on sample B after chemically removing the native oxide layer before
introducing the sample into UHV As mentioned above the initial morphology is similar to
that shown in fig 52a though the oxide has been removed as confirmed by LEED
measurements Fig 53c shows the surface after heating at 700˚ for 30 min Without the
oxide layer the surface began to reorganize at lower temperatures and the small irregular
grains of the original surface begin to grow and become more rectangular After heating to
900˚C (fig 53d) the surface is flat with small rectangular islands The shape of the islands
is similar to that shown in fig 53b but the lateral island dimension is one order of
magnitude smaller
To assess the importance of the initial surface on the process the same treatments were
repeated starting from flat epitaxial silicon (samples CD) The initial epitaxial surface is
smooth with periodic steps due to a small miscut angle (fig52b) On heating the epitaxial
silicon without etching (sample C) relatively large islands are obtained similar to the ones
shown in fig 53b where the ldquostep-pinningrdquo effect of the islands is evident The average
dimension of sample C islands was sim2500 nm3 while only 353 nm3 for sample A Island
density decreased from 24000 to 280 microm-2 This result is maybe due to a lower
concentration of defects in the oxide layer of the epitaxial silicon which allows the
expanding void to grow much larger before coalescence with neighbouring voids the
nanostructures within them thus reaching greater dimensions To verify this hypothesis we
applied the same treatment to clean not-so clean and purposely contaminated samples To
check that the defects inducing void nucleation and nanoisland growth derive from carbon
VPalermo 55
contamination we used electronic-grade methanol to contaminate the silicon surface
Figure 54 shows the results of this comparison
State-of-the-art cleaned samples obtained with multi-step RCA cleaning [13] showed the
development of very few voids the oxide desorption leaving large smooth areas of oxide-
and nanostructure-free silicon as shown in Fig 54a where the presence of residual oxide
can be seen Samples cleaned with a simpler one-stage Pirana cleaning solution showed a
higher number of islands per unit area (Fig 54b) and samples purposely contaminated
with methanol (Fig 54c) had the highest density of island nucleation of all three samples
When the native oxide had been chemically removed from the epitaxial silicon before
heating (sample D) the island size and density is similar to the standard as-received
silicon case (sample B) Thus starting from a flat substrate the process of surface
roughening and island creation was the same even when the oxide layer had been
previously removed
From these results some important indications can be drawn The first is that oxide
desorption temperature depends upon the contamination levels of the oxide layers (fig 54)
with desorption at lower temperatures for contaminated surfaces
Second once the oxide layer is removed the silicon atoms become mobile on the surface
at relatively low temperatures (~700˚C) Their initial configuration is disordered and has a
high surface energy so they crystallize in the small islands shown in fig53d reducing
their surface area and lowering the energy of the system When the oxide is present the
surface cannot reorganize at 700˚C and is stable At 800˚C the oxide begins to desorb in
correspondence with defect points on the surface As the void area enlarges a small
cba Fig 54 The effect of contamination on nano-island production Gradient-filtered STM images of (a)sample cleaned two times with RCA 400x400 nm (b) Sample cleaned with a Pirana solution 500x500nm (c) sample contaminated with MeOH before insertion into the UHV system 500x500 nm Annealingtime is 40 min at 800degC for all the samples
56 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
crystallite grows at the centre of the voids Further oxide desorption uncovers larger areas
freeing more silicon atoms which feed the initial island making it bigger
Fig 55 shows the scaled size distribution of the islands for each group of samples Every
distribution has been obtained using several images of different samples for each group
The size distribution of all the samples follows an exponential decay and the decay seems
the same for all the samples though the average volume of the islands differs by more than
one order of magnitude (see Table 51 for details)
The point defects which catalyze void nucleation can be metallic contaminants present on
the initial oxide surface [8] or organic contaminants which at high temperatures can
form SiC nanocrystals [14] We used Scanning Tunneling Spectroscopy (STS) to look for
differences between the islands and the surrounding flat silicon surface but no difference
was found Furthermore the total island volume per surface unit is very high (more than
104 nm3microm2) and it seems unlikely that such a huge volume could consist of surface
contaminants Islands of pure silicon on silicon have been grown without evidence of
surface contamination[14]
To summarise the formation of nano-sized islands on silicon through surface diffusion
was studied At high temperatures the oxide layer covering the surface decomposes non
uniformly and circular voids of clean silicon are created The presence of the oxide layer
blocks surface silicon atomic motion and surface reorganization except within the voids
causing the growth of islands more than 10 nm high and 30 nm wide on the silicon surface
According to STM and LEED measurements we can say that the islands are mostly
Fig 55 Size distribution (scaled) for the samples described in table 1 The line is a reference for the eye
VPalermo 57
composed of silicon with traces of other substances such as carbon Final island densities
and dimensions depend upon the initial purity of the oxide layer
On the other hand if the oxide is removed chemically before heating the island growth
process is not so localised and takes place simultaneously over the whole surface yielding
smaller and more numerous islands
58 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[2] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[3] High-Temperature SiO2 Decomposition At The Sio2Si Interface Tromp R Rubloff
GW Balk P Legoues FK Physical Review Letters 55 2332-2335 Nov 1985
[4] Defect Microchemistry At The SiO2Si Interface Rubloff GW Hofmann K Liher M
Young DR Physical Review Letters 582379-2382 Jun 1987 Defect Formation In
Thermal SiO2 By High Temperature Annealing Hofmann K Rubloff GW Mccorkle
RA Applied Physics Letters 49 1525-1527 Dec 1986 Kinetics Of High-Temperature
Thermal Decomposition Of SiO2 On Si(100) Liher M Lewis JE Rubloff GW Journal
Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 5 1559-1562 Aug
1987
[5] Thermal-Decomposition Of Very Thin Oxide Layers On Si(111) Kobayashi Y Sugii
K Journal Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 10 (4)
2308-2313 Part 3 Jul-Aug 1992 Scanning Tunneling Microscope Study On Mid-
Desorption Stages Of Native Oxides On Si(111) Kobayashi Y Sugii K Journal Of
Vacuum Science amp Technology B 9 (2) 748-751 Part 2 Mar-Apr 1991 Controlled
Growth Of SiO2 Tunnel Barrier And Crystalline Si Quantum Wells For Si Resonant
Tunneling Diodes Wei Y Wallace RM Seabaugh AC Journal Of Applied Physics 81
(9) 6415-6424 May 1 1997
[6] Defect Formation In SiO2Si(100) By Metal Diffusion And Reaction Liher M
Dallaporta H Lewis Je Appl Phys Lett 53 589-591 Aug 1988 SiO2 Film
Decomposition Reaction Initiated By Carbon Impurities Located At A Si- SiO2
Interface Raider Si Herd Sr Walkup Re Applied Physics Letters 59 (19) 2424-2426
Nov 4 1991
[7] Nanometer-Scale Si Selective Epitaxial Growth On Si(001) Surfaces Using The
Thermal Decomposition Of Ultrathin Oxide Films Fujita K Watanabe H Ichikawa M
Applied Physics Letters 70 (21) 2807-2809 May 26 1997 Pyramidal Si Nanocrystals
VPalermo 59
With A Quasiequilibrium Shape Selectively Grown On Si(001) Windows In Ultrathin
SiO2 Films Shibata M Nitta Y Fujita K Ichikawa M Physical Review B 61 (11)
7499-7504 Mar 15 2000
[8] Stacking-Fault-Induced Defect Creation In SiO2 On Si(100) Liher M Bronner Gb
Lewis Je Appl Phys Lett 52 1982-1985 May 1988
[9] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990 AFM And XPS Characterization Of The
Si(111) Surface After Thermal-Treatment Lamontagne B Guay D Roy D Sporken R
Caudano R Applied Surface Science 90 (4) 481-487 Dec 1995
[10] Scanning-Tunneling-Microscopy Study Of Oxide Nucleation And Oxidation-
Induced Roughening At Elevated-Temperatures On The Si(001)-(2x1) Surface Seiple
JV Pelz JP Physical Review Letters 73 (7) 999-1002 Aug 15 1994 Evolution Of
Atomic-Scale Roughening On Si(001)-(2x1) Surfaces Resulting From High-
Temperature Oxidation Seiple JV Pelz JP Journal Of Vacuum Science amp Technology
A-Vacuum Surfaces And Films 13 (3) 772-776 Part 1 May-Jun 1995
[11] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995
[12] Morphological Changes Of The Si [100] Surface After Treatment With
Concentrated And Diluted HF Palermo V Jones D Materials Science In
Semiconductor Processing 4 (5) 437-441 Oct 2001
[13] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[14] Nanoscale Roughening Of Si(001) By Oxide Desorption In Ultrahigh Vacuum
Gray SM Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology
B 14 (2) 1043-1047 Mar-Apr 1996
60 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 61
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and nano-lines
When the silicon surface is not protected by a native oxide layer or by a hydrogenated
passivating layer it is very reactive The surface chemistry of both Si(100) and Si(111) has
been extensively studied starting from clean surfaces prepared in UHV because of its great
relevance to the microelectronics industry and its technological interest The diffusion and
reaction of molecules and atoms on silicon is an interesting scientific problem on its own
apart from technological issues because silicon reconstruction yields a very complex and
anisotropic surface
We give here a brief summary of the adsorption behaviour of several elements on Si
surfaces The summary is not comprehensive and only the most interesting characteristics
for each substance are given
Table 61 Summary of adsorption behaviour of atoms and molecules on silicon surfaces [12]
Hydrogen
Molecular H2 shows low reactivity towards silicon while atomic hydrogen easily forms
Si-H bonds and can even break Si-Si bonds
Alkali metals
Alkali metals diffuse rapidly into SiO2 and can damage silicon-based transistors
Transition
metals
All transition metals apart from gold and silver react with Si forming metal silicides
62 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nickel A very common contaminant of silicon Nickel from even slight contact with stainless
steel tweezers can contaminate silicon samples forming its silicide and is very difficult
to remove even at high temperatures because it segregates on the silicon surface
Palladium Forms silicides especially Pd2Si and can be used to make contacts possessing a low
Schottky barrier
Titanium Widely used to fabricate contacts for silicon-based microdevices TiSi2 is one of the
more stable and highly conductive silicides
Tungsten The silicide is more stable than TiSi2 on polysilicon and is thus used for polysilicon
gate contacts
Platinum Silicide is used for bipolar transistors PtSi contacts on n-silicon give Schottky diodes
Cobalt Cobalt silicide gives better quality contacts than TiSi2 at the nanometrelevel but is less
used at the moment because it consumes too much silicon during its growth
Iron Silicides have been studied for potential optoelectronic applications because FeSi2 has a
directly accessible bandgap and is optically active
Group 13
Group 13 metals are used in IC technology to dope silicon (giving p-type doping) and
to make III-V type semiconductors When evaporated on Si(100) or Si(111) they react
strongly with the surface giving surface reconstruction and kink formation at steps
Boron Widely used for doping Usually deposited with decomposition of boron hydrides
Sticking coefficient of simple hydrides is very low so decaboranes (B10H14) are used for
deposition
Aluminum Aluminum is widely used for microelectronic contacts but in some cases it reacts with
silicon diffusing into the crystal and with SiO2 stealing oxygen atoms to form Al2O3
Group 14
Carbon Carbon can form a composite semiconductor with silicon (silicon carbide) widely used
to fabricate integrated circuits for use at elevated temperatures or in the presence of
ionising radiation When carbon is deposited on silicon usually polycrystalline films
with properties between diamond and graphite are obtained
CO adsorbs molecularly on silicon without breakage of the C=O bond Other
unsaturated hydrocarbons do not react with silicon Even very reactive strained
molecules like cyclopropane have low sticking coefficient Ethylene acetylene and
propylene adsorb molecularly on Si(100) each molecule sitting on a silicon dimer
interacting with the dangling bonds of the surface Acetylene also adsorbs molecularly
on Si(111) on the 7x7 reconstructed surface preferential adsorption on centre atoms
with respect to corner atoms is observed Benzene and other aromatic ring compounds
adsorb molecularly on the surface They can form σ or π-type bonds with the silicon
laying parallel or tilted respect to the surface according to the bond type
Silicon Silicon atoms can be deposited on the surface with molecular beam epitaxy or
VPalermo 63
decomposition of silanes and chlorosilanes Silanes with single Si-Si bonds have a high
sticking probability and dissociate upon adsorption giving trihydrides and
monohydrides
Chlorosilanes have high sticking coefficients too and decompose on the surface but
they can etch the surface through the reaction SiCl4 + Si rarr 2SiCl2
Germanium Germanium has a lattice constant similar to silicon (Ge lc is 4 larger than Si) so that
various alloys of Si1-xGex can be formed The growth of Germanium on silicon is quite
peculiar with several monolayers adsorbing uniformly on the surface followed by
island formation (Stranski-Krastanov growth)
Group 15
Nitrogen Exposure to ammonia and nitrogen gas at high temperatures leads to the formation of
silicon nitride layers (Si3N4) NH3 dissociates on the surface and reacts with the
dangling bonds giving Si-NH2 and Si-H termination Silicon nitride is stable on the
surface up to 1100degC above this temperature it desorbs as Si2N
Phosphorous Phosphine (PH3) adsorbs dissociatively as Si-PH2 and Si-H and behaviour is similar to
that of ammonia
Arsenic A monolayer of As is often deposited on silicon as a substrate for GaAs growth Arsenic
forms dimers on Si(100) creating dimer rows which are parallel (perpendicular) to the
underlying silicon dimer rows when it is deposited at high (low) temperature On
Si(111) Arsenic breaks the 7x7 reconstruction giving a 1x1 pattern This 1x1 As
monolayer acts as a passivating layer and can resist further As adsorption oxygen and
air
Antimony and
Bismuth
Due to their large covalent radii Sb and Bi form only short dimer rows on silicon and
only at high temperatures
Group 16
Oxygen
Oxygen can oxidize or etch silicon according to the reactions
Si(s) + O2 rarr SiO2 (s)
Si(s) + frac12O2 rarr SiO(g)uarr
Silicon oxidation has already been described in the previous chapter Molecular oxygen
adsorption on silicon is mostly dissociative O atoms break the Si-Si bonds forming an
Si-O-Si bridge or a peroxy bridge Si-O-O-Si Hydrided silicon surfaces do not have
surface dangling bonds and are thus more protected from oxygen attack Etching is
favoured at high temperatures and low oxygen pressures [3] At T gt700degC silicon oxide
on silicon decomposes as SiO(g)
Water
H2O adsorbs dissociatively on silicon easily on Si (100) and with more difficulty on
Si(111) The molecule decomposes giving Si-OH and Si-H on adjacent sites
64 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Group 17
Fluorine
Fluorine adsorbs and reacts easily on silicon dissociating on the surface and forming
Si-F bonds It can then penetrate the surface and attack polarised Si-Si backbonds The
surface is etched with production of SiF4 and SiF2
Chlorine
Chlorine reacts aggressively with silicon etching the surface Cl2 and F2 are often used
commercially to etch silicon usually in a plasma Atomic Cl and Cl2 adsorb on Si(100)
up to saturation The most stable configuration seems to be a Cl atom bonded to each Si
atom of a surface dimer in a symmetric structure but metastable configurations with
two Cl atoms bonded to a buckled dimer and even a central Cl atom bridged across a
dimer have been observed
On Si(111) silicon mono- di- and tri-chlorides are formed In particular heating at
Tgt400degC a uniform Si-Cl monolayer is obtained and the 7x7 reconstruction changes
into a bulk like 1x1 lattice The surface transition induced by Cl adsorption is fully
reversible The 7x7 reconstruction can be restored on desorbing the chlorine by heating
at Tgt1100degC The 1x1 domains nucleate at the lower terrace side of steps the 7x7 at the
upper step edge
Bromine
Like hydrogen bromine maintains the 2x1 structure of Si(100) forming Si-Br bonds
with the dangling bonds of the silicon dimer rows At high exposure some etching of
silicon by formation of volatile SiBr3 species has been observed
Surface diffusion on silicon
In the previous chapter we observed nano-island growth on silicon surfaces with native
oxide and surfaces contaminated with organic impurities While this phenomenon has been
observed in several experiments the formation and growth mechanisms of nanoislands has
still not been satisfactorily described theoretically Various substances have been
hypothesised as nanoisland nucleating agents such as organic or metallic contaminants
present on the surface or oxide clusters (see previous chapter) The island growth process
has been attributed by various workers to lsquosome sort of kinetic instabilityrsquo [4] to the
pinning of step flow by SiC clusters [5] or to a mesoscopic atom flux from areas with low
step density to areas of higher step density [6]
One mechanism proposed to explain surface roughening and island growth is the Ehrlich-
Schwoebel effect ie the presence of an energetic barrier that reflects atoms approaching a
VPalermo 65
Molecule (L
O2
H2O MeOH
CO CO2 CH4
clean surface
downward
weak on si
The aim o
surface an
To obtain
the ones
spontaneou
first obser
surface W
nanoisland
Experime
We used S
removed f
41 mixtur
water and
whole hea
1200degC re
kept below
Table 62 Treatment island density and island volume for each molecule tested
Dose angmuir)
Heating time at 800degC (min)
Island density (microm-2)
Ave Volume (nm3)
Equivalent layer thickness (nm)
280 10 no islands - -
90 10 no islands - - 30 10 2000 19 004
180 10 360 516 019 30 60 no islands - -
180 10 no islands - - - 10 no islands - -
step [7] Recent experiments however indicate that the Schwoebel effect is
licon [48]
f our experiments was to study the reaction of simple molecules with the silicon
d to check the ability of these molecules to nucleate nanoisland growth
a better understanding of the process we chose simple molecules starting from
which are more likely to be present as traces in UHV chambers where
s nanoisland growth is often observed after sample heating Using STM we
ved at the atomic level the possible interactions of these molecules with the
e then heated the surface and checked the density and dimensions of the
s produced
ntal procedures and results
i(111) wafers p-doped 0015 Ω cm Gross contamination and particles were
rom the sample surface with a standard Pirana etch (15 min dip in a H2S4H2O2
e at T=80degC) After this the samples were thoroughly washed in ultra-pure
introduced into the UHV system After degassing for 5 hours at 600degC the
ting stage was allowed to cool down and a sequence of rapid flashes to 1100deg-
moved the native oxide layer and cleaned the surface Pressure during flashes is
1x10-9 mbar
66 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Surface flatness and cleanliness were checked with STM and the contaminant molecules
introduced into the UHV system through a leak valve Doses of several Langmuirs were
used for the experiments (1 Langmuir=10-6 torr sec-1) Pressure was kept constant at 10-7
mbar during adsorption The STM tip was retracted during gas deposition to avoid any tip
shielding effect After the adsorption process the leak valve was closed the base pressure
in the UHV chamber returning to below 10-10 mbar and the presence of adsorbed
molecules checked on the surface with STM Finally the sample was heated at 800degC for
several minutes to allow nanoisland growth Several different areas were examined on each
sample
Table 62 summarizes the dose of exposure and the heating time for each molecule Other
experiments were carried out using smaller dosing times but only the significant results
are reported here
As expected oxygen and water react with the surface oxidising it The adsorption process
can be easily followed with the STM by lowering the tip from time to time and taking an
image No effect of the STM tip is observed over the scanned areas Upon molecule
adsorption the order of the crystalline surface rapidly degenerates and after several
minutes the surface looks completely covered by irregular atomic-sized protrusions
corresponding to Si-OH or Si-O-Si species On heating to 800degC the contaminating species
were easily desorbed through SiO formation the surface became clean and the crystal
surface periodicity re-established No residues or nanoislands were observed
Amongst the molecules tested was methanol CH3OH reacts readily with the surface and
covers it with a disorderd layer already after sim 30 Langmuirs exposure (fig 61)
Fig 61 Si(111) surface during methanol adsorption at 0 1 and 5 minutes respectively (corresponding to 0 6 and 30 L exposure) The dark irregular lines are borders between different 7x7 domains Image size 60x60 nm
VPalermo 67
Synchroton radiation photoemission spectroscopy experiments by Carbone et al [9]
showed that methanol reacts with the rest atoms of the 7x7 silicon cell already after 1
Langmuir exposure but that longer exposures are needed for methanol to react with the
remaining silicon atoms Methanol adsorbs dissociatively forming SiOCH3 and Si-CHx
species on the surface
After methanol adsorption the sample is heated to 800degC and observed again with STM
As in the case of water and oxygen the heating restores the crystalline surface and the 7x7
pattern returns visible But this time quite a dense array of nanoislands is observed on the
surface (fig 62) The islands have an average diameter of 20 nm and a height of sim2 nm
At temperatures above 400degC the Si-OCH3 and Si-CHx species decompose oxygen and
hydrogen are desorbed and according to [9] carbon atoms remain as SiC dispersed
uniformly on the surface Our STM measurements indicate that the carbide does not
randomly cover the surface but that C atoms are concentrated at the nanoislands Rough
calculations based on island volume suggest that the islands must be composed of a SixC1-x
alloy with x varying between 05 and 1
Following these results we expected carbon monoxide to behave in a similar manner
adsorbing onto the surface and with nanoisland growth However no adsorption was
observed with STM and the surface looked perfectly clean even after a dose of more than
100 L of carbon monoxide Some species did adsorb onto the surface though because
after 10 minutes subsequent heating nanoisland growth was observed Island density in
this case was only 18 of the density obtained with methanol and island dimensions are
Fig62 Nanoislands on Si(111) created after 30 L adsorption of methanol and 10 min heating at 800degC Image size 200x200 nm
68 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
much bigger (see table 62 and fig 63) To check whether the islands had been nucleated
by some other contaminant present in the vacuum chamber we heated a blank sample
without introducing any molecule No island growth was observed Exposing the surface to
other simple molecules including carbon dioxide and methane also gave no nanoisland
formation
We can estimate the amount of carbon deposited on the surface during gas dosing The
total flux of molecules hitting the surface is obtained from the Hertz-Knudsen formula
TkmPF
π2=
Fig 63 STM images of Si(111) after exposure to methanol (left) and carbonmonoxide (right) contaminating molecules and subsequent nanoisland growthImage size 400x400 nm The size distribution of nanoislands is shown below eachimage
VPalermo 69
Fig64 Left Si(111) after 180 L contamination of CO and 16 hours heating at 800degC Right Si(111)without CO contamination after 16 hours heating at 800degC Image size 250x250 nm Vertical scale is thesame for both images z-ranges are 35 and 5 nm respectively
where P is the pressure in Pascal k=138x10-23 J K-1 is the Boltzmann constant T is the
temperature and m is the molecular weight (sim32 for methanol and sim28 for CO) in
kilograms
Using this formula we obtain a flux of 027 molecules nm-2 sec-1 for methanol and 029
molecules nm-2 sec-1 for CO After ten minutes exposure at 10-7 mbar pressure more than
150 molecules will have hit each square nanometer of the surface Surface density of
Si(111) 7x7 is sim16 atomsnm2 so each surface atom will be hit by several molecules which
could react or be adsorbed even assuming a sticking coefficient much lower than unity
To explain nanoisland growth we hypothesize that CO molecules adsorb molecularly and
-5
0
5
10
15
20
25
30
0 20 40 60 80 100 120
nm
nm
10 min40 min16 h16 h- no CO
Fig65 STM profiles of nanoislandsgrown for different annealing times The lower curve correspond to the surface heated at 16 hours without CO contamination
70 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Table 63 Island density and volume for the CO-contaminated surface at different heating times For each sample the thickness of a uniform layer having the same volume per unit area of the islands is calculated
Molecule Dose
(Langmuir) Heating time at
800degC (min) Island density
(microm-2) Ave Volume
(nm3) Equivalent layer thickness (nm)
CO 180 10 362 517 019 CO 180 40 550 1072 059 CO 180 16 h 505 3675 186
clean surface - 16 h rough - -
very weakly on Si(111) moving rapidly on the surface Because of their rapid diffusion
the STM is not able to image the CO molecules on the silicon [10]
Increasing the temperature part of the CO molecules will surely desorb but some of them
will dissociatively adsorb onto the surface The oxygen will then desorb as SiO while
carbon will remain fixed forming very stable Si-C bonds
CO decomposition will be favored at surface defects surface steps or in the proximity of
already formed SiC clusters An increased reactivity of the CO molecule on a surface in
correspondence with phase boundaries has already been observed on Pt surfaces [10]
In this case a reduced number of nucleation centres will react with the CO molecules
yielding large and fewer islands with respect to the case of methanol which does not have
the possibility of travelling long distances over the surface
Other factors are likely to influence the process of island nucleation after CO adsorption
the co-adsorption of other molecules which can slow down CO and favour its
decomposition cannot be ruled out as well as the formation of new defects at higher
temperatures Measurements with a variable temperature STM or with some other surface
analysis technique are planned in the future to confirm the proposed mechanism
When the surface was further annealed the nanoislands grew in size After 16 hours
heating very large nanoislands with diameters of sim35 nm and heights of sim20 nm became
visible (fig64 left) As a comparison when the same surface was heated without
nucleation centres deriving from contamination an irregular surface was obtained (fig64
right)
While the nanoisland density seemed to reach an asymptotic limit of 500 islandsmicrom2
(table 63) island size continued to grow with time (fig65) even when no further carbon
was supplied to the surface This indicates that even though the initial nucleating core of
the nanoisland is likely to be an SixC1-x alloy further growth is due to silicon atoms
diffusing from the crystal and being adsorbed by the growing island The final volume
VPalermo 71
occupied by the islands corresponds roughly to a 18 nm thick overall layer of removed
silicon
It is noteworthy that the clean sample after 16 hours annealing even though smoother than
the CO-contaminated one looked much more disordered with hardly any flat area visible
on nanometres scale In the presence of nanoislands however further annealing increases
island size but keeps part of the surface quite flat and ordered with flat areas visible
between the islands Thus the nanoislands can be imagined to act as ldquoimpurity sinksrdquo for
further contaminants approaching the surface yielding a greater but more ordered surface
roughening with respect to the uncontaminated silicon surface
Circular arrays of nanoislands
We used the voids described in the previous chapter formed during oxide layer
decomposition as nanoscopic masks to test the validity of the hypothesis outlined above
The small circular areas uncovered during oxide decomposition can act as ldquoskating rinksrdquo
where contaminants and diffusing silicon atoms can move freely over the surface while
the oxide layer all around and higher than the void area as well as being chemically
bonded to the surface silicon atoms will hinder surface diffusion Furthermore the oxide
passivating layer will at least in some cases prevent contaminating molecules from
adsorbing and decomposing on the surface In this way selective nanoisland growth inside
void areas can take place
The experimental procedure consisted of the following steps
1) An oxide-passivated sample was heated for 10 minutes at 800degC Oxide decomposition
began in correspondence with surface defects and spread laterally creating naked areas
of silicon
2) While the void enlarges the same defect that initiated oxide decomposition behaved as
an attractor for moving silicon atoms and nucleated the growth of a nanoisland at the
void centre
3) The void surface was exposed to contaminating molecules
4) The sample was then heated again at 800degC During the second heating the void
continued to enlarge and new nanoislands were created The freshly uncovered silicon
surface provided mobile silicon atoms to feed nanoisland growth
72 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
5) When the heating continued for long enough all the oxide was desorbed and a clean
silicon surface covered by circular groups of nanoislands was obtained
Figure 66 shows the different stages of the process
We thus heated an oxidized sample at 800degC for 10 minutes in UHV to create multiple
voids in the oxide layer After this we exposed the sample to the contaminating molecules
Then we further annealed the surface at the same temperature to promote island growth
The first time we observed nanoisland growth in the void area was by chance after heating
for a second time samples which had been stored for 2 weeks in vacuum Even at pressures
as low as 10-11 mbar some molecules will adsorb onto the surface slowly creating defects
and thus potential nanoisland nucleation centres It is possible to observe with STM that
the silicon crystal surface looks very clean just after a high temperature flash but even
after only a few days randomly adsorbed species will be observable on the surface
Unfortunately it is not possible to identify these species simply from in situ STM
measurements
Mass spectrometry measurements show that the residual gas contaminants in vacuum are
usually He Ar H2 CH4 CO CO2 and N2 [11] Some of these molecules such as
hydrogen will stick to the silicon surface at room temperature but will simply desorb
during annealing without nucleating nanoislands Others like CH4 or CO2 will not react
with the surface to generate nanoislands as demonstrated previously The best candidate
for contamination of samples stored in UHV seems thus carbon monoxide has the ability
321
4 5
Fig 66 Selective nanoisland growth within oxide voids See text for details
VPalermo 73
as shown before to nucleate nanoislands although co-adsorption mechanisms can not be
ruled out Fig 67 (left) shows the circular groups of nano-islands obtained after heating
void-covered samples stored for 2 weeks in UHV Small islands are observed in the former
void area The void itself has enlarged slightly uncovering a clean oxide-free area of the
silicon surface Apart from the central island created during the first period of heating the
surrounding ones have dimensions decreasing from the void border to the inner area of the
void
This confirms that most of the material needed for island growth comes from the silicon
atoms diffusing from the freshly uncovered areas around the original void The outer
nucleating centres are nearest to the silicon atom source and thus generate larger islands
For a faster process we directly exposed the void to a significant concentration of CO
Fig67 (centre) shows the islands grown after a 180 Langmuir exposure and subsequent
heating The islands are preferentially located around the original void perimeter the rest
of the original oxide-free surface remaining untouched This suggests that although the
whole surface is exposed to CO the gas only interacts chemically with the surface at the
SiSiO2 interface around the oxide-free void perimeter The SiSiO2 interface thus provides
preferential nucleation sites for nanoisland growth in the successive thermal annealing
step This recalls a similar phenomenon observed in the catalytic oxidation of CO on the
Pt(111) surface where CO molecules reacted with adsorbed O species only along the
perimeters of oxygen islands [10]
On repeating the CO exposure and thermal annealing steps a second circle of nano-islands
Fig67 Left circular area of nanoislands obtained from a sample stored 2 weeks in UHV 500x500 nmCentre nanoisland circle obtained after exposing the voids to 180 L of CO and heating for 10 min at800degC 400x400 nm Right two concentric circles obtained with further CO adsorption followed byheating 650x650 nm
74 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
was formed around the newly-expanded surface void perimeter (fig67 right) This
process can in principle be repeated again to obtain a series of concentric groups of
nanoislands the only limit depending on the minimal distance between neighbouring
expanding voids We have thus observed that contaminants adsorbed at low pressure over
long periods of time generate random island growth over the whole void area while CO
molecules adsorbed at relatively high pressures over short periods yield selective
nanoisland growth at the former void border Several theories can explain the different
results CH4 and CO2 as mentioned above did not nucleate islands when when the silicon
surface was exposed to them for short periods However the lower surface mobility of
these contaminants and the greater time available for reaction with the surface could be
the cause of nucleation of SiC clusters over the whole oxide-free void area and not only at
its border The coadsorption of water another UHV residual gas with CO on the naked
silicon areas could perhaps allow a surface reaction similar to that between CO and the
SiSiO2 interface during CO exposure Another explanation for the formation of these
random nucleation sites within the voids could be the coadsorption of residual carbon-
containing species present in the UHV system with hydrogen the major residual gas in
stainless steel UHV systems
The possibility of decorating silicon oxide borders with nanoislands using the high
mobility and selective decomposition of CO molecules suggests interesting applications
for nanofabrication processes that will be discussed below
Silicon nanowire creation on Si(100)
We performed most of the experiments described above on silicon (111) 7x7
reconstructed because this surface is easily prepared in UHV and above all because it is
isotropic and has no preferential directions for atom diffusion The 2x1 reconstruction of
silicon (100) even though it has a much simpler unit cell than Si(111) 7x7 shows a more
complicated morphology with lots of monoatomic and biatomic steps and dimer rows
parallel to each other The orientation of the dimer rows changes by 90deg on alternate
atomic layers Surface diffusion in this case is much more complicated because atoms will
experience the effects of the surface anisotropy and diffuse preferentially along or across
the dimer rows The diffusion energy of a silicon atom as example is 06 eV along a
dimer row and 085 eV across different dimer rows This while being a complication
VPalermo 75
offers interesting possibilities for the creation of ordered structures The monoatomic steps
and the dimer rows can for example act as templates for the formation of elongated
structures of composition similar to the islands described previously
On heating a Si(100) surface oxide decomposition and void growth takes place as
described for Si(111) In some cases the void shape reflects the surface symmetry
especially for very thin oxide layers and the voids often have a slightly squared shape
Nanoisland growth is also influenced by the substrate symmetry and the islands look
rectangular as irregular clusters surrounded by concentric patterns of monoatomic steps
The whole structure reminds vaguely a ldquozigguratrdquo a kind of stepped pyramid found in
Mesopotamia (fig68 inset)
Apart from nanoisland growth in some cases we observed spontaneous formation of
nanowire-like SiC structures on the Si(100) surface The wires were generated after
exposing a Si(100) surface in vacuum to traces of CO followed by sample annealing at
800degC for 15 minutes The wires exhibit lengths ranging from 10 to 100 nm and average
widths of ca 5 nm Although being randomly positioned on the surface they are perfectly
aligned along the crystal axes of the substrate (Fig68)
Several descending monoatomic steps can be observed around each nano-line and nano-
island Very often one of the line extremities coincided with an island
Nanoline
Nanoisland
Fig 68 A Si(100) surface covered by nanoislands and nanowires The monoatomic steps present on the surface are visible The two insets show a typical nanoisland and a nanowire obtained on this surface The typical ldquodimer-row ldquostripes are visible along the sides of the nanowire
76 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Couples of parallel lines starting from the same nano-island were also observed At high
resolution (fig 69) the lines showed a lamellar periodic structure with a periodicity of
around 4 nm and frequent defects Scanning Tunneling Spectroscopy measurements
showed no significant difference between the line and the surrounding silicon surface
The proposed formation mechanism for this kind of structure involves two stages in the
first one organic contaminating molecules adsorb and diffuse on the surface from the
vacuum environment stopping preferentially at surface singularities including steps and
defects
Several substances such as atomic carbon carbon monoxide ethylene and fullerene
decompose when adsorbed on silicon surfaces at temperatures above 500degC and create
localised defects on the surface inducing strain deformations and a metastable surface
reconstruction [12]
If these molecules remain attached to the steps they can diffuse rapidly along step border
until they meet a line and decompose In this way long lines of SiC defects parallel to the
surface steps can be produced In the second phase the clusters of silicon carbide can
-005
0
005
01
015
02
025
03
035
04
-3 -2 -1 0 1 2 3
V
dId
V lt
dId
Vgt
linesilicon
Current- Voltage behaviour measured with STS Fourier analysis of wire periodicity
asymp 4 Aring
High resolution image of a line
Fig 69 Top high resolution image of a nanowire showing the lamellar structure Fourier analysis (bottom left) shows that the lamellae periodicity is sim 04 nm bottom right the IV characteristics measured with the STM tip on the line and the silicon surface
VPalermo 77
25-600degC
Final linear structure
600-800degC
Contaminant Silicon
Fig 610 Schematic representation of nanowire formation mechanism
locally inhibit the spontaneous surface diffusion of silicon atoms acting as templates for
the reorganisation of surface steps (fig 610)
The final morphology shows disordered lines and islands each surrounded by a complex
pattern of silicon monatomic layers
At 800degC silicon atoms are highly mobile on the surface Due to atom diffusion the
surface reorganizes with a step-flow mechanism The presence of a fixed line of
contaminants blocks atom diffusion and step flow generating the monoatomic steps
around each wire (fig 611) SiC clusters act as lsquopinning sitesrsquo on silicon atom surface
diffusion and can be purposely used to direct the localised growth of nano-islands [13]
Although the possibility of using methanol and carbon monoxide as nano-island precursors
has been demonstrated the exact nature of the substances nucleating nano-wire growth is
still unknown
Further experiments will be needed to confirm these hypotheses as well as to identify the
contaminating species nucleating nanowire growth Carbon monoxide is a main candidate
for nanowire growth but the real nucleation process is likely to be a complex one
78 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Step flowblocked
Fig 611 Representation of step flow blocked by the presence of a nanowire 105x105nm
Bibliography
[1] Surface-Chemistry Of Silicon Waltenburg HN Yates JT Chemical Reviews 95 (5)
1589-1673 Jul-Aug 1995
[2] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[4] Growth Of Si On The Si(111) Surface Lanczycki Cj Kotlyar R Fu E Yang Yn
Williams Ed Das Sarma S Physical Review B 57 (20) 13132-13148 May 15 1998
[5] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
VPalermo 79
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990
[6] Nanoscale roughening of Si(001) by oxide desorption in ultrahigh vacuum Gray SM
Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology B 14 (2)
1043-1047 Mar-Apr 1996
[7] Step Motion On Crystal Surfaces Schwoebel Rl Journal Of Applied Physics 40 614-
618 Feb 1969
[8] Scanning Tunneling Microscopy Investigation At High Temperatures Of Islands And
Holes On Si(111)7x7 In Real Time Evidence For Diffusion-Limited Decay
Hildebrandt S Kraus A Kulla R Wilhelmi G Hanbucken M Neddermeyer H Surface
Science 486 (1-2) 24-32 Jul 1 2001
[9] Methanol Adsorption On Si(111)-(7x7) Investigated By Core-Line Photoemission And
Mass Spectrometry Of Photodesorbed Ions Carbone M Piancastelli Mn Zanoni R
Comtet G Dujardin G Hellner L Surface Science 370 (1) L179-L184 Jan 1 1997
[10] Atomic And Macroscopic Reaction Rates Of A Surface-Catalyzed Reaction
Wintterlin J Volkening S Janssens Tvw Zambelli T Ertl G Science 278 (5345)
1931-1934 Dec 12 1997
[11] Redhead PA Hobson JP Kornelsen EV The Physical Basis Of Ultrahigh
Vacuum Chapter 12 (Chapman amp Hall London 1968)
[12] The Si(001) C(4 X 4) Surface Reconstruction A Comprehensive Experimental
Study Norenberg H Briggs Gad Surface Science 430 (1-3) 154-164 Jun 21 1999
[13] Production Of Nanostructures Of Silicon On Silicon By Atomic Self-Organization
Observed By Scanning Tunneling Microscopy Jones D Palermo V Applied Physics
Letters 80 (4) 673-675 Jan 28 2002
80 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 81
7 Conclusions and proposals for future work It is very difficult to foresee the future of nanoscience and nanotechnology even over the
next few years exciting discoveries are continuously made opening up new possibilities
and progress in this sector is now one of the fastest and most varied So it is impossible to
predict whether computers will in the future be made with nanowires nanotubes single
electron devices or some kind of quantum-based transistor What does seem clear
however is that silicon devices can still further extend their levels of miniaturization to
remain competitive for at least the next ten years [1] During this period new emerging
technologies will reach scientific maturity and arrive at the production lines
Even when new technology prototypes do become available for full-scale production the
astronomical cost of changing from silicon to new technologies will be a major concern
Moreover a vast amount of extremely detailed knowledge is already available on all
aspects of silicon technology Thus the possibility of integrating innovative
nanotechnologies with standard CMOS silicon technology already seems an attractive
prospect [2]
For this and other reasons we have focussed our research efforts on surface phenomena on
silicon wafers which could be potentially useful for the development of silicon-compatible
nano-devices
Some important conclusions can be drawn from the work described in this thesis
bull Silicon surfaces of different crystal faces were studied at atomic resolution in ultra-
high-vacuum following transformations induced on the surface by chemical (etching)
and physico-chemical (organic contamination heating) treatments
82 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull The presence of surface contaminants induces at high temperature the growth of
nanometre-sized islands and lines on silicon
bull Final island densities and dimensions strongly depend upon the chemical nature of the
contaminant molecules and the treatments used Molecules interacting only weakly
with silicon will desorb upon heating whereas molecules which decompose easily on
Si will give widespread random nano-island nucleation Between these two extreme
cases complex behaviours of diffusion and subsequent decomposition are possible
bull Nanoscopic naked silicon voids were produced in oxide layers through carefully-
controlled thermal annealing It was found that void density and size depends upon
oxide purity surface contamination levels and annealing conditions
bull The growth of nano-islands and nano-voids was exploited simultaneously and the
voids used as nanoscopic masks to control the positions of growing islands Complex
silicon-on-silicon nano-structures were obtained with multi-step growth processes
bull Finally the use of macroscopic treatments and simple physical processes to produce
large quantities of nanometre-sized structures on silicon surfaces was demonstrated
These processes and the structures produced are of potential interest to the
microelectronics industry
We showed in previous chapters how it is possible to modify silicon surfaces using
chemical and physical methods The use of self-organization for the production of Si-on-Si
20 nm
Fig 71 Left SEM image of field emitter arrays made by conventional lithography [3] Center 3-D STM image of a self-organized void with a central nano-island Right Lateral schematic view of a field emitter device
VPalermo 83
nano-structures has the advantage with respect to other emerging techniques of being
silicon compatible and thus is a potential candidate for its implementation in the
production of new devices Furthermore this technique can create billions of strucures on a
wafer in only a few minutes being thus applicable to large-scale production
Even when ordered structures are obtained these methods still show a poor reproducibility
and control Chemically modified surfaces look very disordered at high magnification
with trenches and holes etched into the surface and progress has recently been made in
producing very small structures electrochemically [4] Oxide decomposition with
nanoisland growth in UHV can also yield ordered structures The void holes with
nanoislands located at their centres although quite variable in dimensions are more similar
to engineered devices than to a spontaneous random surface-roughening process In fact
there is a surprising similarity between the nanovoid-nanoisland structures and well-known
commercially available devices called field emitter arrays (FEA) These are small tips used
as electron guns in various kinds of displays and other devices Both the commercial
devices and our self-organized structures consist of a protruding tip at the centre of
microscopic holes on an insulating layer covering a semiconductor surface (see fig71)
Of course the order and size uniformity of commercial FEAs is much better than that
obtained with our process but it is encouraging to see that a self-organized structure
obtained with simple heating has a very similar surface topography and composition of
commercial working devices obtained after many expensive stages of optical lithography
deposition and etching
Nano-island density and size can to some extent be controlled and different
contaminants can be adsorbed onto the surface to modify the growth process In the future
with a better understanding of diffusion dynamics and growth mechanisms a multi-stage
fabrication process can be envisaged where purposely-designed molecules are adsorbed
onto silicon surfaces to diffuse organize in ordered arrays and favour surface
reorganization at the nanometre scale The same molecules could be used to deliver
dopants to obtain for example highly conducting nanostructures on otherwise undoped
highly resistive silicon The selective formation of nanostructures at the SiSiO2 border
can be used to scale down mesoscopic patterns to the nanometric level A silicon oxide
layer can be easily patterned on the surface with normal optical lithography and then
nanostructures can be built following the pattern edge In this way the production of very
small nanowire based devices may be possible (fig 72)
84 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
D
Gate
S
D
S
SiO2
SiO2
2 Heating nanowire creation
and oxide removal
3 Source-drain deposition by conventional lithography
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
4 Insulating layer (oxide) and gate (metal) deposition
Fig 72 Schematic drawing of the construction of a self-organized nanowire-based transistor
Surface diffusion and decomposition of molecules can be sensitive to even smaller surface
features The nanoscopic lines formed on Si(100) are very regular and neat if their
nucleation and growth mechanism could be controlled ordered networks of nanowires
could be made on silicon
VPalermo 85
The techniques developed here are still experimental ones the careful control and
reproducibility of each step is still an issue and further studies will be needed to clearly
understand the dynamics of formation of these structures Nonetheless the results obtained
so far open up the possibilities of producing novel devices such as for example nano-
transistors (see scheme in fig73 where the mechanism described in fig 72 is applied for
the fabrication of a device array) The possibility of using the border of a lithographic
pattern to draw nanolines can in principle allow the creation of nanodevices using normal
IC manufacturing techniques
The results obtained and their description in this thesis are not the first examples of the use
of self-organization for the production of ordered structures and will certainly not be the
last In our opinion the most important outcome of these findings is that it is possible to
use surface diffusion to overcome the inherent limits of lithographic techniques in
microdevice production processes and to force properly selected molecules to react with
the surface in a spatially differentiated manner at particular sites thereby creating ordered
series of nanostructures The role of the surface dynamics of the substrate atoms
themselves is extremely important in this process
Outstanding results have been and continue to be obtained in nanoscience and
nanotechnology research promising the emergence of new production technologies Sooner
or later those emerging technologies will have to face the issues of cost and compatibility
with the enormous investment made in existing technologies and processing facilities The
commercial production of nanostructures based on surface diffusion processes could in
principle be cheap simple and compatible with existing technology Moreover the
processes of diffusion in bulk silicon is well-known in the microelectronics industry which
has both the know-how and the equipment to better understand and exploit the surface
diffusion processes on silicon surfaces studied here
Extremely important and innovative results are often obtained by exploiting very simple
ideas and in our opinion physical processes such as surface diffusion used to create
complex nanostructures on silicon surfaces is a prime example
86 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
2 Heating nanowire creation and
oxide removal
3 Source-drain deposition by conventional lithography
4 Insulating layer (oxide) and gate (metal) deposition
Fig 73 Schematic macroscopic outline of the construction of a self-organized nanowire-based transistor
VPalermo 87
Bibliography
[1] P Gargini Intel Technology Coordinator ldquoFrom Microelectronics To
Nanotechnologiesrdquo Invited Campus Colloquia Lecture CNR Research Area Bologna
February 6th 2003
[2] Toward A Hybrid Micro-Nanoelectronics Cerofolini Gf Ferla G Journal Of
Nanoparticle Research 4 (3) 185-191 Jun 2002
[3] Recent Progress In Field Emitter Array Development For High Performance
Applications Temple D Materials Science amp Engineering R-Reports 24 (5) 185-239
Jan 25 1999
[4] Silicon Dioxide Micropillars For Sieving Fabricated By Macroporous Silicon-Based
Micromachining Izuo S Ohji H French Pj Tsutsumi K Kimata M Sensors And
Materials 14 (5) 239-251 2002 Electrochemical Etching In HF Solution For Silicon
Micromachining Barillaro G Nannini A Piotto M Sensors And Actuators A-Physical
102 (1-2) 195-201 Dec 1 2002
88 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 89
Acknowledgements
This work would not have been possible without the help and guidance of Dr Derek Jones
he introduced me to the use of STM and other techniques found the funding for this PhD
and helped me with the thousands of small and big problems I had to face during my
research I am grateful to Prof Alberto Ripamonti too for being the supervisor of this
thesis Thanks are also due to Dr Giancarlo Seconi director of ISOF-CNR where most of
this work was carried out Financial support from the Italian National Research Council
(CNR) is also gratefully acknowledged
My passion for surface science began when Fabio Biscarini showed me that it was possible
to really see the atoms and for this I will always thank him
During these years I had the luck to collaborate with many good scientists among those I
would like to cite Enrichetta Susi Massimo Cocchi and Anna Mazzone from CNR
Claudio Zannoni Anna Cavallini Daniela Cavalcoli and Antonio Castaldini from the
University of Bologna Sergio Pizzini and Maurizio Acciarri from the University of Milan
One of the most stimulating periods of my PhD was the one I spent at the Steacie Institute
for Molecular Sciences in Ottawa working with Robert Wolkow Dan Wayner Greg
Lopinski and Peter Kruse I thank them for discussing together many exciting scientific
ideas and trying to put some of them into practice
Special thanks go to my colleagues Paolo Samorigrave and Stefano De Cesari for long and
useful discussions on Science Life and their interactions often with the help of a good
pint
Finally I would like to thank Vassilia Gaetano Simone Claudia Silvia Letizia
Alessandro Angela and Sebastien even though not directly involved in this work they
have been over these three years a continuous source of happiness and support
Bologna March 2003
90 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
SCIENTIFIC PUBLICATIONS bull Lateral diffusion of titanium disilicide as a way to contacting
hybrid Si-organic nanostructures Palermo V Buchanan M Bezinger A Wolkow RA APPLIED PHYSICS LETTERS 2002 v81 p 3636 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v6 issue 20) bull Surface modifications in Si after Rapid Thermal Annealing Castaldini A Cavalcoli D Cavallini A Jones D Palermo V Susi E JOURNAL OF THE ELECTROCHEMICAL SOCIETY 2002 v 149 pG633 bull Production of nanostructures of silicon on silicon by atomic self-
organisation observed by scanning tunnelling microscopy Jones D Palermo V APPLIED PHYSICS LETTERS 2002 v 80 p 673 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v5 issue 5) bull Nucleation of nanostructures from surface defects on silicon Palermo V Jones D SOLID STATE PHENOMENA 2002 v 82-84 p 687 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo MATERIALS SCIENCE AND ENGINEERING B 2002 v88 (2-3) p220 bull Morphological changes of Si[100] surface after treatment with
concentrated and dilute HF Palermo V Jones D MATERIALS SCIENCE IN SEMICONDUCTOR PROCESSING 2001 v 4 p 437 bull Electrical and structural properties of processed silicon surfaces Susi E Cavallini A Castaldini A Cavalcoli D Jones D Palermo V ldquoRECENT RESEARCH DEVELOPMENTS IN VACUUM SCIENCE amp TECHNOLOGYrdquo 2001 v 3 p189 bull Numerical Solutions of the Stochastic Equations of Crystal
Growth Mazzone A M Palermo V INTERNATIONAL JOURNAL OF MODERN PHYSICS C 2000 v 11 Part 1 p195-204 bull Advances in silicon surface characterisation using light beam
injection techniques
VPalermo 91
Acciarri M Pizzini S Simone G Jones D Palermo V MATERIALS SCIENCE AND ENGINEERING B 2000 V73 (1-3) p 235 - 239 bull Abrupt orientational changes for liquid crystals adsorbed on a
graphite surface Palermo V Biscarini F Zannoni C PHYSICAL REVIEW -SERIES E- 1998 V 57 NUMBER 3A p R2519-R2522 Oral presentations bull The use of oxide desorption and surface diffusion for the
creation of silicon on silicon nanostructures Palermo V Jones D 1ST NATIONAL WORKSHOP ON CURRENT TRENDS IN NANOTECHNOLOGIES Catania (Italy) Feb 2002 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo V E-MRS Spring Meeting Strasbourg (France) June 2001 bull STM study of surface transformations on silicon during UHV
annealing Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999 bull Advances in silicon surface characterisation using light beam
injection technique Acciarri M Pizzini S Simone G Jones D Palermo V E_MRS SPRING MEETING June 1999 Posters bull Spontaneous nano-wire growth on silicon Palermo V Jones D SMARTON workshop Leuven (Belgium) October 2002 bull Ordered circles of nano-islands on silicon from CO adsorption Palermo V Jones D TRENDS IN NANOTECHNOLOGY 2002 Santiago de Compostela (Spain) September 2002 bull Production of nanostructures of silicon on silicon by atomic self-
organisation Palermo V Jones D EUROMAT Rimini (Italy) June 2001 bull Etching holes and anisotropic corrosion on silicon [100] Palermo V Jones D Susi E Asoli B SILICON WORKSHOP Genova (Italy) February 2001
92 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull Morphological and electrical characteristics of damaged silicon surfaces
Susi E Castaldini A Cavalcoli D Cavallini A Jones D Palermo V SILICON WORKSHOP Genova (Italy) February 2001 bull Effect of HF etching on the roughness of a silicon surface Palermo V Jones D NATIONAL CONFERENCE ON PHYSICS OF MATTER Genova (Italy) June 2000 bull Effect of Native Oxide Desorption upon the Surface Morphology of Si[100] by STM and LEED Palermo V Jones D SILICON WORKSHOP Genova (Italy) February 2000 bull Characterisation of silicon surfaces for microelectronics through STM measurements Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999
VPalermo 93
Contact Vincenzo Palermo ISOF- Institute for Organic Synthesis and Photoreactivity Via Gobetti 101 40129 Bologna ITALY Tel +39-051-6398336 Fax +39-051-6398349 Mail palermoisofcnrit
94 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Index
1 Introduction nanotechnology and the future of computers
2 Silicon surfaces
3 STM and other surface analysis techniques
4 Surface modification of silicon in liquid Nano-hole creation
5 Surface modification of silicon in vacuum void creation and oxide desorption
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and
nano-lines
7 Conclusions and proposals for future work
10 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another approach to nanodevice fabrication has become possible through the discovery of
carbon nanotubes which were observed for the first time in 1991 by a Japanese electron
microscopist studying the material deposited during arc-evaporation synthesis of fullerenes
[6] They consist of a graphite-like carbon seamless cylinder with a diameter of several
nanometers and lengths of up to a millimetre Carbon nanotubes are very stable can
behave as metals or semiconductors and can host other molecular or ionic species thus
modifying their electrical behaviour In 2001 Avouris and his group reported the first
circuit made with a single nanotube [7] A few months later Cees Dekker presented a
nanotube-based transistor able to amplify an input signal by a factor of ten and built
several logic circuits using these nanotube transistors [8]
One problem with carbon nanotubes is that it is very difficult to control their electronic
properties ie their metallic or semiconducting behaviour An alternative to carbon
nanotubes are semiconductor nanowires Silicon nanowires can be made using a laser to
vaporize the silicon together with a metal catalyst like iron or gold The vapour condenses
in nanosized drops of silicon and metal from which the wires slowly grow out as more
silicon is adsorbed In 2001 a group at Harvard University [9] created a transistor by
crossing two different nanowires After this the same group arranged four nanowires in a
noughts and crosses grid creating something like a 4-bit memory (fig 13 right) Even
metallic nanowires made of platinum and silver can be used in a crossed configuration to
store information [10]
There are some issues common to all these new technologies though First it is difficult to
imagine these methods applied to large-scale production Up to now the insertion of a
molecule between two electrodes is an occasional lucky event while nanotubes and wires
have to be positioned on the surface creating the appropriate contacts on them manually
The large-scale production of integrated circuits using these building blocks will not be
straightforward and does not seem imminent
Another issue is of an economic and not a scientific or technological nature Since 1965
the cost of IC manufacturing plants has sky-rocketed If the cost of semiconductor
production plants continues to rise exponentially in a few years such plants will cost up to
$20 billion This is a sizeable investment even for large companies such as IBM or Intel
For this reason it is likely that IC companies will resist changing to completely new
technologies closing down their existing plants As it is clear that silicon will remain the
VPalermo 11
fundamental raw material of the IC industry for the foreseeable future nanotechnology
developments for microelectronics will need to be silicon-compatible In these early days
of nanotechnology the most valid approach would seem to be the addition of molecular
functions to existing silicon technology ndash using the latter as a foundation on which to build
Fabrication of self-organised structures on silicon
The possibility of using the phenomenon of atomic or molecular self-organization to create
nanostructures on silicon has already been demonstrated The clean silicon surface shows
in some cases a high degree of order and complex surface reconstruction as will be
described later Several different ordered structures form spontaneously on this surface
such as series of monatomic steps or boundaries between reconstructed areas It has been
demonstrated that it is possible to use these structures to fabricate ordered nanodots and
nanolines on the surface [11] More recently well-defined nanometric patterns have been
obtained with selective etching of silicon using nitric oxide [12]
In this study the possibility of creating different types of nanostructures on the silicon
surface is explored Methods had to be developed which were
- Simple They must not need complex masks or lithographic steps to create the
structure but rather exploit self-organisation phenomena
- High resolution the silicon surface modifications should be on a scalelength of below
100 nm
- Fast billions of nanostructures have to form over the whole surface simultaneously to
be compatible with large-scale production requirements
- Cheap they must not require expensive equipment (such as e-beam lithography) but
exploit simple chemical andor physical treatments to produce nanostructures on the
silicon surface
In Chapter 2 the main characteristics of silicon are described Chapter 3 provides a
summary of the techniques used for this research Chapter 4 examines the chemical etching
of silicon in different liquid environments and the effects of this etching on the surface at a
nanoscopic level with the creation of nanoholes
12 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Chapter 5 describes the growth in ultra-high vacuum (UHV) of nanoscopic voids and
islands on the silicon surface and the effect of surface oxide on this growth
Chapter 6 discusses the modification of silicon surfaces in UHV following the adsorption
of molecules and thermal heating to produce nanoislands and nanolines on silicon
The overall conclusions of our work are summarised in Chapter 7
Finally we will give some conclusions based on the results obtained and discuss possible
applications of the methods developed
Bibliography
[1] From wwwintelcom
[2] From wwwsematechorg
[3] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
[4] Molecules Get Wired Service RF Science 294 (5551) 2442-2443 Dec 21 2001
[5] Coulomb Blockade And The Kondo Effect In Single-Atom Transistors Park J
Pasupathy AN Goldsmith JI Chang C Yaish Y Petta JR Rinkoski M Sethna JP
Abruna HD Mceuen PL Ralph DC Nature 417 (6890) 722-725 Jun 13 2002 Kondo
Resonance In A Single-Molecule Transistor Liang WJ Shores MP Bockrath M Long
JR Park H Nature 417 (6890) 725-729 Jun 13 2002 Nanotechnology - Electronics
And The Single Atom De Franceschi S Kouwenhoven L Nature 417 (6890) 701-702
Jun 13 2002
[6] Smallest Carbon Nanotube Ajayan PM Ijima S Nature 358 (6381) 23-23 Jul 2 1992
[7] Carbon Nanotube Inter- And Intramolecular Logic Gates Derycke V Martel R
Appenzeller J Nano Letters 1 (9) 453-456 Sep 2001
[8] Logic Circuits With Carbon Nanotube Transistors Bachtold A Hadley P Nakanishi T
Dekker C Science 294 (5545) 1317-1320 Nov 2001
VPalermo 13
[9] Logic Gates And Computation From Assembled Nanowire Building Blocks Huang Y
Duan XF Cui Y Lauhon LJ Kim Kh Lieber CM Science 294 (5545) 1313-1317
Nov 9 2001
[10] Formation And Disappearance Of A Nanoscale Silver Cluster Realized By Solid
Electrochemical Reaction Terabe K Nakayama T Hasegawa T Aono M Journal Of
Applied Physics 91 (12) 10110-10114 Jun 15 2002
[11] Fabrication And Integration Of Nanostructures On Si Surfaces Ogino T Hibino H
Homma Y Kobayashi Y Prabhakaran K Sumitomo K Omi H Accounts Of Chemical
Research 32 (5) 447-454 May 1999
[12] Ultrafine And Well-Defined Patterns On Silicon Through Reaction Selectivity
Prabhakaran K Hibino H Ogino T Advanced Materials 14 (19) 1418-1421 Oct 2
2002
14 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 15
2 Silicon surfaces
The name silicon (silicio in Italian) comes from the latin word silex Amorphous silicon
was first isolated by Berzelius in 1824 by reaction of potassium with silicon tetrafluoride
Thirty years later the first crystalline silicon was prepared Silicon makes up 25 of
earthrsquos crust and is the second most abundant element after oxygen Elemental silicon is
not found in nature occurring as silicon oxide (sand quartz amethyst flint etc) or
silicates (asbestos clay mica etc) Perhaps no other element and its compounds has such
a wide range of uses Silicon compounds such as sand and clay are used in the building
industry as refractory materials for high-temperature applications and for enamels and
pottery Silica is the main component of glass silicon carbide is an important abrasive and
silicones are commonly used polymers and lubricants
Here the most interesting use of silicon of course is for the production of
microelectronics devices For this application silicon of high purity (999999) and of
high crystallinity is needed Table 21 lists some of the physical characteristics of silicon
High purity polycrystalline silicon is produced by the reaction of gaseous trichlorosilane
with hydrogen in a furnace Then to prepare a single-crystal of silicon the so-called
Czochralski method is commonly used
Polycrystalline silicon is melted in a quartz furnace at 1415degC in an argon atmosphere
Then a seed of single-crystal silicon is lowered into contact with the melt and slowly
pulled out In this way the crystal grows and a crystalline cylindrical ingot several metres
long is created from the initial seed
After cooling down the ingot is sliced into thin silicon wafers The wafer surfaces are
polished using a counter-rotating lapping machine in an Al2O3 slurry until the surface is
very flat and shiny ready for the lithographic processes
16 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another way to obtain single crystal silicon is the Floating Zone (FZ) method in which a
silicon cylinder is slowly passed through a heating ring The area inside the ring melts and
solidifies smoothly crystallising as it comes out of the ring yielding a single silicon crystal
Microelectronic devices are built on the silicon surface which is the surface of interest
here Unfortunately silicon surfaces are normally quite dirty and uneven at the atomic
scale Atmospheric oxygen and humidity react with silicon surfaces creating a thin layer of
oxide (called ldquonative oxiderdquo) which is usually irregular and full of defects Different kinds
of contaminants also adsorb onto the surface These are usually small organic molecules
and microscopic dust particles A clean surface on exposure to the atmosphere is
completely covered with gas molecules in less than 10-9 seconds If the pressure is
reduced letrsquos say to 10-6 mbar this time increases to 1 second This is the reason why to
study a clean surface we have to work in UHV at pressures below 10-10 mbar
The atoms in the silicon crystal have a diamond-like structure each atom having 4 bonds
in a tetrahedral sp3 arrangement with bond angles of 10947 degrees At the crystal
surface some atoms will have non-bonding orbitals ldquodanglingrdquo in the vacuum ie sp3
orbitals with a lone electron which are highly reactive These orbitals are known as
dangling bonds To minimize surface energy the surface will reorganize by decreasing
the number of dangling bonds
Table 21 Physical data of silicon [1]
Atomic Weight 2809 Lattice constant (A) 543095 Crystal structure Face-centered cubic
(diamond) Melting point 1415 degC
Density (gcm3) 2328 Boiling point 2355degC Atomscm3 50E22 Minority carrier
lifetime (s) 25E-3
Dielectric Constant 119 Specific heat (Jg degC)
07
Breakdown field (Vcm)
~3E5 Thermal conductivity (Wcm degC)
15
Electron affinity x(V)
405 Vapour pressure (Pa) 1 at 1650degC 1E-6 at 900deg C
Energy gap (eV) at 300K
112 Reactivity Inert to acids Attacked by halogens and alkaline
solutions Intrinsic carrier
conc (cm-3) 145E10 Oxidation states +4 -4
Intrinsic Debye Length (microm)
24 Energy of a Si-Si bond (eV)
232
Intrinsic resistivity (Ω-cm)
23E5
VPalermo 17
Dangling bond densities and positions and thus the type of surface reconstruction will
depend upon crystal orientation as well as the temperature and kinetics of the system
Fig 21 shows a drawing of the main faces of a silicon crystal The angle α between any
(11n) face and the (100) face can be calculated from 2cos 2 += nnα The angle
between any (11n) face and the (111) face can be calculated from
)2(3)2(cos 2 ++= nnα
The chemistry and physics of the faces are very different a brief description will be given
for the most important orientations
Si (100)
On the (100) surface each atom has two Si-Si bonds connecting it to the bulk and two
dangling bonds pointing outward Surface energy is reduced by the dimerisation of the
surface atoms through overlap interaction of one dangling bond per atom forming rows of
dimers aligned along the (110) direction This is the well-known ldquo2x1rdquo reconstruction of
this silicon surface The symmetric dimers would make the silicon surface metallic but to
reduce surface stress the dimers tend to buckle and the surface is thus semiconductive It
took several years to understand that the dimers are buckled because at room temperature
Fig 21 Schematic view of the principal orientations of a silicon surface Surface atoms are white bulk atoms are black dangling bonds are gray [2]
18 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
they shift easily from one buckling direction to the other thus appearing symmetric under
STM observation Fig 22 shows an STM image of the 2x1 reconstructed surface
Even almost perfect (100) surfaces have a certain number of monoatomic steps and the
dimer rows on atomic layers are aligned at 90deg to those on adjacent layers Dimer rows are
thus perpendicular or parallel to the step When the dimers on the upper side of the step
are parallel to the step the step is called SA if they are perpendicular the step is called SB
Because of this symmetry SA step edges will be smoother compared to the more broken
fragmented edges of the SB steps
A common defect on the Si(100) surface is the presence of nickel contamination which
appear as missing dimers in STM images This type of contamination is so critical that
even if the silicon sample is only briefly brought into contact with stainless steel tools
(tweezers for example) the 2x1 reconstruction of the surface can be blocked
Silicon atoms can diffuse easily over the silicon surface as monomers and dimers
especially at elevated temperatures The anisotropy due to the 2x1 reconstruction causes a
difference in the diffusion energies of adsorbates over the surface Diffusion of these
silicon species along dimer rows for example will be much easier A list of diffusion
energies for monomers and dimers is given below [2]
Diffusion on Si(100) 2x1 Ed (eV) Monomers along dimer rows 06 Monomers across dimer rows 085 Monomer formation energy 18 Dimer along dimer rows 11 Dimer across dimer rows 15 Dimer formation energy 26 Dimer binding energy 076 Vacancies along dimer rows 17 Vacancies across dimer rows 19
SA
SB
Fig 22 STM image of a 2x1reconstructed silicon surfaceshowing the dimer rows and steps40x35 nm Nickel-induced defectsare visible as dark spots SA and SBsteps are indicated
VPalermo 19
So the diffusion energy for both monomers and dimers is nearly 40 greater if they have
to cross a dimer row This difference reduces to sim10 for vacancy diffusion
Si(113)
The (113) surface can be imaged as a sequence of alternating (100) and (111)-like
structures with two and one dangling bonds on alternate atoms respectively Interest in the
(113) surface is scientific as it has been used to study the energetics of the (100) and (111)
surfaces as well as for surface adsorption experiments
Si(100) surfaces can easily develop (113) facets
Si(111)
This surface besides being the first one imaged with STM with atomic resolution is one of
the most studied because it is the best cleavage face of silicon and because it shows one
of the most complex and elegant reconstructions in surface science
All Si-Si bonds in the silicon crystal are perpendicular to a (111) plane so this face will
have the lowest number of dangling bonds created per unit area In fact each Si atom on a
(111) surface shows a single dangling bond oriented perpendicular to the surface and
bonded to three back atoms These three bonds for each surface atom account for the great
chemical and physical stability of the Si(111) surface Surface energy is 009 eV Aring-2
compared to 015 eV Aring-2 for Si(100)
For energy minimization this surface reconstructs forming a huge 7x7 lattice cell
containing 102 atoms described by the Dimer-Adatom-Stacking fault model (DAS) For a
detailed description of cell structure see fig 23
The cell described by this model is very complex being composed of three kinds of atoms
adatoms rest atoms and corner hole atoms Furthermore a subsurface stacking fault is
present in one half of the cell making the two halves of the unit cell look different under
STM (Fig 24) It took 26 years of research to completely understand the exact structure of
the 7x7 reconstruction
Cleaving a silicon crystal along a (111) plane produces a metastable 2x1 reconstruction
the 7x7 reconstruction is easily obtained by flashing at high temperature in UHV At T gt
20 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
830degC a disordered 1x1 phase covers the surface Cooling down to 800degC leads to the
formation of the 7x7 phase If the cooling process is too rapid small 7x7 domains nucleate
and a disordered 1x1 phase is preserved between domain boundaries
Si(110)
Even though as mentioned before the (111) plane is the favoured cleaving plane of
silicon thin (100) commercial wafers will not break along this plane because the angle
between (100) and (111) is too far from 90deg (see table 22) Instead they will break along
the (110) plane because it is perpendicular to the (100) surface Each surface atom on
Si(110) has a Si-Si bond pointing downward one dangling bond pointing outward and two
Si-Si bonds parallel to the surface in a zig-zag pattern (see fig 21) Cleaved (110)
surfaces are disordered but upon annealing at high temperatures an ordered complex 16x2
Fig23 Scheme of the 7x7 DAS model [2] In each unit cell there are 9 dimers 12 adatoms and a stacking layer fault The force driving this complex reconstruction is the minimization of dangling bonddensity The DAS model shows the lowest number of dangling bonds (19) of all possiblereconstructions 12 dangling bonds are at the adatoms 6 at the rest atoms and 1 at the corner hole atom This surface is metallic
VPalermo 21
reconstruction takes place The surface appears as a series of long ridges and valleys
parallel to each other Eventually tilted facets of orientation (17 15 1) can form on this
surface The adsorption of Ge atoms on this surface leads to the formation of self-
assembled nanowires [3]
Table 22 Angles in degrees between different silicon faces [2]
Orientation
113
110
111
100
100
2524
9000
5474
0
111
2950
3526
0
110
6476
0
113
0
Fig 24 STM image of a Si(111)surface with 7x7 reconstruction A unitcell with its adatoms is highlightedImage size 13x13 nm
22 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] Weast RC Handbook Of Chemistry And Physics (Chemical Rubber Co Cleveland
1972)
[2] Dabrowski J Mussig H Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] The Structure Of Clean And SiGe-Covered Si(110) Surfaces Butz R Luth H Surface
Science 365 (3) 807-816 Oct 1 1996
VPalermo 23
3 STM and other surface analysis
techniques
Scanning Tunneling Microscopy
Since the invention of the optical microscope at the end of the 16th century the possibility
of examining surfaces at higher and higher magnification has fascinated mankind
Development of the technique continued and towards the end of the 19th century optical
microscopes were as good as todays standard instruments The physical limits of the
wavelength of visible light (350-800nm) had been reached
In the 1920s de Broglie showed that electrons can behave like waves and the use of these
particles for imaging with much higher resolution soon followed Atomic resolution using
this technique is only possible in the transmission mode with extremely carefully prepared
samples
In 1982 using the peculiar properties of piezoelectric materials Binnig and Rohrer brought
a metallic tip very very close to a silicon surface and scanned it across an extremly small
area (fig31) The tunneling of electrons from the tip into the sample or vice versa allowed
them to obtain a local density of electronic states (LDOS) map of that surface Although
theory (which treated the extreme point of the tip as a sphere) then excluded the possibility
of atomic resolution following a tip crash into the surface Binnig and Rohrer began to
observe the LDOS with atomic resolution For this discovery and their following work
they were awarded the Nobel Prize in Physics in 1986
24 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The basic principle of STM is very simple A metallic tip is scanned over a surface without
making ohmic contact and a tunneling current passes between the tip and the surface An
electronic circuit keeps this current constant by raising and lowering the tip during the
scan In this way recording the tip height at each point a three-dimensional image of local
density of electronic states (LDOS) of the surface can be obtained To explain the
extremely high resolution attainable by this simple technique quantum theory is needed
According to classical physics the current will flow between sample and tip only if they
are in physical contact If there is a vacuum gap between the two the electrons will simply
remain confined for example within the surface without the possibility of passing into the
tip
In quantum physics however the electrons have a certain probability of passing
(tunneling) across the gap appearing on the other side of the gap in this way reaching the
tip It can be shown that the probability of an electron tunneling through a gap of thickness
z is
kzep 22)0( minusprop ψ h
φmk
2= (1)
where ψ(0) is the electron wavefunction at the surface-gap border m is the electron
mass=91x10-28 g and φ is the work function of the metal (ie the energy required to
remove an electron from that material For silicon it is 48 eV) The tunneling current thus
decays exponentially with z and is extremely sensitive to topographical imperfections
present on the scanned surface A rough formula giving the current as a function of z is [1]
zFS eEVI φρ 0251)( minusprop
Fig 31 Binnig and Rohrer with the first STM Image from IBM [2]
VPalermo 25
where ρs(EF) is the local density of states at the Fermi level on the given surface For
example the formula predicts that for silicon an increase in tip-surface distance of 1 Aring
will give a 95 decrease in tunneling current
This huge dependence of tunneling current upon the distance allows detection even of the
sub-nanometre changes in height given by the single atoms of which the surface is
composed and thus to resolve them in the LDOS images Of course this description of the
tunneling process is oversimplified and for a more accurate one the electronic states of
the tip of the sample and their interaction have to be taken into account Fig 32 shows a
schematic representation of the interaction between tip and sample orbitals
The exponential decay of current with distance also yields high lateral resolution If the tip
is approximated as a sphere of radius R and the current passing at the minimum tip-sample
distance is I0 then the current passing at a lateral distance x from this point will be
Rxk
eII 22
0
2minus
=
Assuming a tip radius of 100 nm the current is concentrated in an area sim15 nm wide at tip
apex
Very sharp tips with even smaller curvature radii can be produced in several ways Simple
Fig32 Schematic view of tip-sampleorbitals interaction a) no interaction b)equilibrium c) sample positive d) tippositive [1]
26 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
tungsten wires mechanically cut are capable of obtaining atomic resolution on graphite in
air but for more disordered and rougher samples sharper and more reproducible tips are
needed
STM tips are mostly made by electrochemically etching a W or Pt-Ir wire The tips we
used were prepared using methods based on the work of Fotino [3]
A tungsten wire 038 nm diameter is immersed in a KOH solution(06M) with a thicker
tungsten wire used as a counterelectrode The cathodic and anodic reactions involved in the
etching are
Cathode 6H2O + 6 e- rarr 3H2(g) + 6 OH-
Anode W(s) + 8 OH- rarr WO42- + 4H2O + 6 e-
A potential of 3V ac is applied to the tungsten and the wire is immersed in the solution
until a constant current of sim100 mA is established The etching rapidly removes metal
shaping the wire end as a sharp tip When the potential reaches 11 V the coarse tip etching
is finished The wire is then removed from the solution carefully inserted into an
insulating plastic tube leaving only the tip exposed and re-immersed in the solution with
the tip pointing upwards A more gentle etching is thus made to reduce tip radius Usually
5 minutes etching at 07 V ac is used In this configuration very small hydrogen bubbles
formed on the tip sides sliding upwards with a ldquohoningrdquo effect on the tip
This procedure yields extremely sharp and reproducible tips at the microscopic level
After the etching the tip is thoroughly rinsed in ultrapure water then dipped into
concentrated HF to remove surface oxides and hydroxides [4] The tip is dried with
nitrogen inserted into the UHV system and degassed overnight at sim150degC
The possibility of measuring sub-nanometric distances would be useless without being able
to control tip movement over such a minute scale To scan the tip over the surface a
piezoelectric scanner is used Piezos are usually made of an alloy of PbZrO3 and PbTiO3 a
material which contracts or expands when a voltage is applied to it The Omicron
instrument used in our laboratory has three such piezo scanners for xy and z tip motion
allowing one to scan the tip over the surface with sub-Aringngstrom precision (fig 33)
To isolate the instrument from ambient vibrations the whole STM stage is suspended upon
four springs which eliminate all frequencies above 1 Hz and surrounded by a crown of
VPalermo 27
copper wings and fixed magnets Parasitic currents generated by the magnets into the
copper wings contrast every movement of the stage and efficiently block stage vibrations
The STM can be used not only to explore surface topography but to measure the IV
characteristics of single atoms or molecules on the surface (Scanning Tunneling
Spectroscopy or STS) Furthermore it can be used to modify the surface with voltage
bursts digging into it or delicately moving single molecules or atoms over a surface [5] It
can be used in vacuum in air and with proper lateral isolation of the tip even in liquids
A major drawback of STM is that it works only on conducting and semiconducting
surfaces and thus cannot be used on many surfaces of biological and chemical interest
Another instrument more suitable for these and other applications is the Atomic Force
Microscope (AFM)
Atomic Force Microscopy
AFM was invented in 1986 by Binnig Quate and Gerber after calculating the possibility of
building a cantilever with a force constant of the same order of magnitude as that of a
chemical bond
In AFM a tip mounted on a microscopic cantilever (usually made of Si3N4 fabricated with
optical lithography) is brought close to a surface When the tip touches the surface the
cantilever is very slightly deflected upwards The movement is measured by observing the
Y-PIEZO
SILICON SAMPLE
Fig 33 A picture of theSTM used for theexperiments The tripodpiezo scanner is shown
Z-PIEZO
X-PIEZO
TIP
28 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
deviation of a laser beam hitting the upper face of the cantilever Fig 34 provides a
schematic view of the principle of AFM
The typical force constant of the cantilever varies from 00006 to 2 Nm the typical
resonance frequency is 3 to 120 kHz The AFM tip can apply a force on the sample of up
to 10-9 N The AFM can be used on conductive or insulating surfaces in vacuum air or
liquids Furthermore the tip can be modified to sense electrostatic potentials (electric force
microscopy) or magnetic fields (magnetic force microscopy) it can even be functionalized
with complex molecules such as proteins to interact with biological surfaces
A drawback of the AFM is that the force it exerts can damage the surface under
observation especially if the sample is soft (as in the case of cell membranes for
example) This problem can be overcome using the instrument in tapping mode (where the
tip does not move laterally during its brief contact with the surface) or in non-contact mode
in which the tip oscillates above the surface during the scan and the changes in its
frequency due to interaction with surface are monitored The shifts in the oscillating
frequency of the cantilever due to tip-sample interaction are then used for imaging the
surface In this mode interaction of the tip with the surface is minimal and soft samples
can be imaged
STM and AFM are the main techniques used for this work A brief description of other
techniques used occasionally is given below
Fig 34 Scheme of an atomic force microscope
VPalermo 29
Low Energy Electron Diffraction (LEED)
Electrons with energies in the 20-500 eV range are diffracted by a crystalline surface the
diffraction peaks are visualized on a fluorescent screen This technique probes the long
range order of the surface up to a depth of several nanometres
X-Ray Photoelectron Spectroscopy (XPS)
XPS allows both qualitative and quantitative chemical analysis of the elements present on
or near the sample surface
An X-ray source is used to photoionize the atoms on a surface and produce photoelectrons
By measuring the kinetic energy of the photoelectrons the binding energy of the electronic
levels can be calculated This energy will depend on the chemical environment of the
surface atoms
Although the soft X-rays used penetrate to a depth of ~2000Aring the sampling depth of the
technique is determined by the mean free path of the photoelectrons which allows their
escape from only the first 10-100Aring
Secondary Ion Mass Spectroscopy (SIMS)
High and low energy ions (primary ions) are used to bombard a sample and remove surface
atoms and ions The ionic fragments removed (secondary ions) are then analysed by a mass
spectrometer The surface can be consumed during the measurement and profiles obtained
giving concentrations of the materials composing the sample at different depths (depth
profiles)
A popular variant of SIMS is TOF-SIMS In this technique the secondary ion masses are
measured by a time-of-flight (TOF) measurement The secondary ions generated by the
bombarding primary ions are accelerated to a constant kinetic energy and then move
through a field-free space before they reach the detector where their intensity is measured
as a function of flight time Since ions with different masses have different velocities at a
30 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
given kinetic energy the measured flight times of the ions can easily be converted to their
masses The static nature of this latter technique allows mass spectroscopy surface analysis
with minimal damage to the surface
Bibliography
[1] Chen CJ Introduction To Scanning Tunneling Microscopy (Oxford University Press
Oxford 1993)
[2] From wwwibmcom
[3] Tip Sharpening By Normal And Reverse Electrochemical Etching Fotino M Review
Of Scientific Instruments 64 (1) 159-167 Jan 1993
[4] A Convenient Method For Removing Surface Oxides From Tungsten STM Tips
Hockett LA Creager SE Review Of Scientific Instruments 64 (1) 263-264 Jan 1993
[5] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
VPalermo 31
4 Surface modification of silicon in liquid
Nano-hole creation
Liquid treatments of silicon wafers are very common in the integrated circuit (IC)
manufacturing industry They are used to clean and improve surface uniformity to create
and etch protective oxide layers and to remove photo-resist layers
Crystalline silicon with its native oxide layer is very stable and is resistant to many acids
It is easily attacked by hydrofluoric acid (HF) and alkaline solutions
The thin (~2 nm) passivating layer of native oxide (SiO2) is formed on exposure to the
atmosphere This surface layer contains many defects and contaminants so it is usually
chemically stripped and substituted with a better chemically-formed protective oxide
The most common silicon cleaning procedure is the RCA method named after the Radio
Corporation of America [1] It consists of two steps in the first one the surface is treated
with a hot alkaline solution (H2OH2O2NH4OH 411) to remove particles from the
surface following this a hot acidic solution (H2OH2O2HCl 411) is used to remove
metal contamination Other well-known cleaning methods are IMEC (a sequence of
cleaning steps in H2OO3 and dilute HF) or the Pirana etch (a hot 41 mixture of H2SO4
H2O2)
32 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The standard RCA clean removes surface contaminants etches the native oxide and
oxidizes the silicon surface leaving a uniform layer of silicon oxide which better protects
the surface from further contamination
Etching with fluorine-based solutions
Hydrofluoric acid is one of the most common reagents used in the treatment of silicon
wafers both in the research field and in industrial processes A rapid dip in dilute HF is the
simplest way to remove the native oxide from Si(100) and leaves the surface passivated by
a layer of Si-H bonds Because of the low polarization of Si-H bonds the Si-H layer is
stable even for several days protecting the surface from contamination It has often been
assumed that this short etch does not significantly change the surface morphology of the
silicon substrate[2] even though a prolonged dip in dilute HF leads to surface roughening
[3]
Although dilute HF roughens the Si(100) surface at the atomic scale [45] immersion in
concentrated HF (49) etches the surface oxide without attacking the Si surface
uncovering in this way the buried SiSiO2 interface The final counter-intuitive result is
that dilute HF etches the silicon while concentrated HF leaves the crystalline silicon
untouched [3]
Etching Si with fluorine-containing solutions at different concentrations and pH can
produce different morphologies from rough surfaces to flat nearly ideal Si-H terminated
surfaces
Hessel et al and Higashi et al demonstrated in 1991 that very flat Si(111) surfaces can be
obtained using 40 NH4F while etching with HF always yields rough surfaces The
surface becomes smoother because the etchant rapidly attacks Si atoms at step borders
thus removing surface kinks and irregularities in a step-flow mechanism [6 7] Later on
even smoother and more perfect surfaces were obtained by removing oxygen from the
solution after it was discovered that oxygen dissolved in 40 NH4F initiates the formation
of triangular etch pits It was not possible to obtain flat surfaces by etching Si(100) with
ammonium fluoride solutions which leads to the formation of small 2x1 dimer-row
reconstructed (100) terraces together with (111) facets [8]
This difference is caused by the different hydride terminations prevailing on the (100) and
(111) faces While the ideal Si(111)-H surface is monohydride terminated the more
VPalermo 33
reactive dihydrides predominate on the Si(100)-H surface making it more vulnerable to
etching The etching reaction is thus strongly anisotropic etching (100) facets faster than
(111) thus producing (111) microfaceting on Si(100) crystals
A more uniform Si(100) surface can however be prepared by etching at low pH with an
HFHCl mixture [9] or by using very dilute HF solutions and ultrapure water with low
dissolved oxygen and carbon contents [10]
Electrochemical etching can also be used applying anodic or cathodic bias to the silicon
to obtain different morphologies [11] by varying the potential isotropic or anisotropic
etching is observed The aforementioned results show that despite the simplicity of the
reactants fluoride etching of silicon is quite a complex reaction
Fig 41 Chemical etching of silicon
HO+H
H HO
H
H
H
FSi
Si
SiSi
+H2O
-OH-
+F-
H2O H H
Si H
F OH
H Si
Si Si
-H2
+H2O
H
H
H
FSi
Si
SiSi
H
H
H
OHSi
Si
SiSi
H
H
H
H Si
Si
Si Si
Etching mechanism of silicon
HF rapidly dissolves the SiO2 passivating layer on silicon leaving the surface almost
completely hydrogenated [12] After this two different types of reactions etch the silicon
simultaneously one chemical and the other electrochemical [13] The overall etching
mechanism can be schematized in two stages (see Figure 41)
i) Si-H bonds are substituted by Si-F or Si-OH bonds creating a partial charge on the
surface silicon atom and polarizing its Si-Si backbonds
ii) these polarised backbonds are then more easily attacked by HF or H2O After
rupture of the Si-Si bond the atom is removed leaving behind new Si-H
terminations and the reaction can start again
34 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
These reactions take place although at different rates on both Si(100) and Si(111)
Stage i) is usually the rate-determining step of the reaction and the stability of the Si-H
bonds depends upon the pH the concentration of nucleophilic species in solution and an
eventual potential applied to the crystal
For pH gt5 as in the case of concentrated NH4F solutions reaction begins with attack by
water to give Si-OH (step ArarrC) The -OH group is rapidly substituted by fluorine with
polarization of the underlying Si-Si bonds These bonds are then easily attacked by water
the silicon atom being released into solution as HSiF(OH)2 The Si-OH rarr Si-F substitution
is not fundamental for the reaction and etching can proceed even for Si-OH terminated
atoms but XPS measurements showed the presence of a certain number of Si-F bonds
remaining Furthermore fluorine seems to have a catalytic effect on Si-H substitution as
indicated by the dependence of the etch rate upon the F- concentration at least for pH
values between 4 and 8
Si-F bonds can be easily removed by a water rinse In the case of strongly alkaline
solutions (pH=14) OH- groups act directly as nucleophiles and no fluorine is needed to
catalyze Si-H bond rupture
At pH lt4 almost no free F- ions are present in solution and the etch rate is very small at
pH lt2 all etching reactions are very slow and this explains the stability of Si crystals in
concentrated (50 ww) HF solutions
Si Si
Si Si
H
H
H
H Si
Si
Si Si
H
H
H Si
Si
SiSi
H
H
H
OH-H+ -e
(F-)
-H+
-e
+H2O hellipas in fig1
Fig 42 Electrochemical etching of silicon by H2O
The electrochemical etching of silicon involves electron transfer from the surface atoms to
the valence or conduction band of the crystal (according to the type of doping of the
silicon) Several studies have been carried out at different electric potentials In the case of
p-silicon or n-silicon at anodic potentials where positive charges (h+) are available in the
crystal a silicon-centred radical can be created by capture of a hole from the bulk (fig
42) The F- ion does not participate in the reaction but is thought to contribute through
VPalermo 35
electrostatic interactions by lowering the energy of the interaction step After the
formation of the Si-OH group the reaction proceeds as shown in the scheme of fig 41
Matsumura et al [4] proposed that not only water but HF2- molecules also play a major role
in electrochemical etching of silicon leaving on the surface Si-F terminated bonds which
can be immediately attacked in an autocatalytic process (fig 43)
In the electrochemical reactions described above an external potential is applied to the
silicon crystal The chemical and electrochemical reactions in any case take place
simultaneously most of the time with the chemical path predominating at high pH Even
when no external potential is applied to the silicon partial electrochemical reactions can
take place at different ldquocathodicrdquo and ldquoanodicrdquo sites on the surface with an internal charge
exchange which ensures neutrality [11] This macroscopic silicon etching and hydrogen
bubble formation can lead to visually observable patterns on the surface when Si(100) is
immersed in ammonium fluoride even without applying a potential
Si Si
Si Si
H
H
H
F Si
Si
Si SiH
F F-H+ -2e
+HF2-
H
HF
SiSi
SiSi + F
F F
FH
H
H
F -H+ -2e
+HF2-
Fig 43 Autocatalytic electrochemical etching of silicon by HF2-
Inhomogeneities on silicon surfaces caused by electrochemical reactions and charge
transfer have been studied extensively because they are of fundamental importance in the
formation of porous silicon
Pore formation on silicon
When Si(100) or Si(111) are etched under anodic bias in fluorine-based solutions
microscopic pores form on their surface Several different morphologies of pores have
been observed with pore diameters ranging from 10 nm to several microns with depths of
several microns [14] Pore shape is very variable too ranging from ordered straight pores
to chaotic networks of branched pores (fig 44) Porous silicon has been known since the
fifties but it was only in 1990 that interest in this material increased following the
discovery that porous silicon layers were able to emit bright red light This led to a large
amount of research and now different classes of micropores can be reproducibly created
36 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Fig 44 Different types of Silicon micropores From ref [14]
mostly for optical and micromachining applications However there is still no unified
theory able to explain the nucleation and growth mechanism of all the different kinds of
pores
We will give a short description of some of these theories for more detail see Parkhutik et
al [15]
One model explains pore nucleation on the basis of physical processes such as hole
positive charge migration ion transport to the surface and small perturbations on the
silicon surface modelled as Fourier components The system is shown to be unstable and
some spatial frequencies that lead to pore nucleation evolve from the etching process
A second model focusses on stationary pore growth without explaining the nucleation
stage According to this model silicon dissolves preferentially at pore edges because h+
charges are attracted by the stronger electric field present at these edges
A third class of models explains pore growth as a Diffusion Limited Aggregation (DLA)
process where the random walk of h+ charge carriers through the depleted layer present at
the silicon-liquid interface controls pore shape
Finally the model by Carstensen Cristophersen and Foll [16] proposes that areas of the
surface of some characteristic size LCO are etched by synchronized ldquocurrent burstsrdquo in the
flow of h+ charges These bursts dissolve silicon through cyclic stages of surface oxidation
oxide removal and hydrogen passivation Areas where a burst has recently taken place are
less passivated and thus more likely to be etched again in this way the pore bottom
continues to dissolve while the pore walls are passivated and are thus less favourable areas
towards current bursts
VPalermo 37
EXPERIMENTAL RESULTS
In the following sections we will show some experimental results obtained from STM and
AFM measurements of fluorine-treated Si(100) surfaces In the first part the results of
mild etching using concentrated and dilute HF solutions at low pH are presented In the
second part the results of etching at high pH using ammonium fluoride are presented and
the mechanism of pore formation discussed
Etching of Si(100) in dilute and concentrated HF
Samples were cut from different areas of an 8-inch diameter p-doped silicon(100) wafer
(10 Ω-cm) supplied by MEMC Electronic Materials Each series of STM measurements
was carried out over at least six different areas on at least two identical samples Low
Electron Energy Diffraction (LEED) was used to check the surface cleanliness of the
samples before STM measurements
Table 41 summarizes the different treatments of each sample After etching with
electronic grade HF each sample underwent a final rinse in Ultra-Pure Water (UPW
resistivity gt18 MΩ-cm) Both the HF and the UPW were allowed to flow continuously
over the sample surface Some samples were not etched with HF at all but just washed with
UPW to observe the morphology of the native oxide layer (~2 nm thick) covering the
surface All of these processes were carried out under nitrogen to limit reoxidation and the
samples were then introduced from the nitrogen atmosphere directly into the vacuum
chamber and degassed overnight at sim150degC before LEED and STM measurements
STM images were obtained from each sample using the same measurement parameters
(sample bias 4 V feedback current 1 nA scan speed 800 nm s-1) These parameters
38 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
provided a satisfactory level of reproducibility for all the samples Measurements were
made over an area of 500x500 nm (image size 500x500 pixel) Slope correction was
carried out by subtracting row-wise and column-wise fitted slopes from the entire image
which gave better results than the simple subtraction of a fitted plane especially for the
rougher samples Following slope correction the rms roughness
sum minus=xy
hyxhN
22 ))((1σ
and the 2-D Fourier transform
)(22
2
)(4
)( vyuxi
xyeyxhvuF +sum∆
= π
π
were calculated for each image where N2 is the number of pixels composing the image
h(xy) is the surface height at each point ∆ is the distance between points h is the mean
height and u v are the spatial frequencies The radial power spectrum PS(f) of the STM image data is obtained from the angular
average of the squared Fourier transform with f 2 = u2 + v2
Fig 45 shows the STM images obtained from the various samples Sample A still covered
with its native oxide layer shows an irregular surface with RMS roughness of ~05 nm
(see Table 41) Observing the sample with LEED gave no diffraction pattern even at
relatively high incident electron energies because of the surface oxide coverage After 1
min etching in dilute HF (sample B) the morphology is similar to the original one though
Table 41 Sample treatments average RMS roughness and integrated area of power spectra
Sample Treatment RMS roughness (nm)
PS area (f lt 01 nm-1)
PS area (f gt 01 nm-1)
A Rapid dip in water 053 plusmn 013 366 032
B 1 min in HF 5 + 10 min in water 051 plusmn 008 272 034
C 30 min in HF 5 + 10 min in water 062 plusmn 008 1064 038
D 5 sec in HF 49 + 10 min in water 042 plusmn 004 183 022
VPalermo 39
Fig 45 STM images of each group of samples showing the topography of the silicon surface A) noetching original oxide surface B) after 1 min etching in dilute HF C) after 30 min etching in diluteHF D) after dipping in concentrated HF Grey scale indicates height of the surface from lower (black) to higher (white) The images are 250x250 nm ie representative portions of the images usedfor the roughness measurement and PSD analysis
some of the larger features have disappeared and the image quality is better maybe due to
improved tunnelling due to the cleaner surface The RMS roughness is comparable to that
of the original surface Clear diffraction patterns are visible using LEED though at quite
high energies (200 eV) After prolonged etching (sample C) the RMS roughness increases
to 062 nm and a long-range corrugation is visible on the surface even if the LEED pattern
is good
The samples dipped in concentrated HF (D) reveal the bare SiSiO2 interface which has a
disordered aspect and protrusions over a wide range of dimensions The quality of the
STM images of sample D is very good probably due to the cleanliness of the surface
40 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 E -0 2
1 E -0 1
1 E + 0 0
1 E + 0 1
1 E + 0 2
1 E + 0 3
0 0 0 0 0 1 0 1 0 1 0 0f (1 nm )
nm^4
A a s re c e iv e d
B e tc h e d 1 min
C e tc h e d 3 0 min
D e tc h e d HF 4 9
Fig 46 Log-Log plot of the averaged power spectra of the STM images for all the samples
which gives a more stable tunnelling junction The LEED pattern is excellent showing
clear diffraction peaks at energies as low as 37 eV comparable to that obtained after high
temperature cleaning in UHV
Fig 46 shows the power spectra of the samples The high frequency and low frequency
areas of the power specturm are considered separately Table 41 shows for each sample
together with the roughness the integrated area of the power spectrum for the high and low
frequency part
We first examine the differences between the samples in the low frequency part of the
spectrum (lt01 nm-1) In this spatial range the short dip in concentrated or dilute HF
(samples BD) seems to lower the roughness of the sample removing some of the larger
features of the native oxide surface visible in Fig 45a and thus decreasing the
corresponding part of the power spectrum Sample C however etched with HF for 30 min
shows an increase in roughness peaking at 0012 nm-1 (~85 nm corresponding to the
typical dimensions of the corrugations visible in Fig 45c) On this scale sample C has a
PS density of 435 nm4 compared to 101 nm4 for sample A
Samples A B and C have the same PS in the high frequency range showing that the fine
structure of the surface is not changed by the HF etching Sample D on the other hand
shows a decrease of roughness for frequencies gt 015 nm-1 This effect could be due to the
VPalermo 41
improved surface cleanliness after etching with concentrated HF which would give a more
stable STM junction thus reducing the high frequency noise in the image
Fig 45 and the analysis of the power spectrum of each sample shows that a rapid dip in
HF removes the native oxide but does not lead to major changes in the morphology of the
Si surface its only effect being the removal of some of the larger features present on the
original surface Prolonged etching on the other hand increases the RMS roughness of the
surface
Etching of Si(100) in concentrated ammonium fluoride and nano-hole creation
Two different types of commercial p-doped Si(100) wafers (2Ω-cm and 10Ω-cm) from
MEMC were used Several different samples of 10x5 mm were immersed for 10 minutes
in 40 electronic grade NH4F solution under agitation Previous works used low
temperatures or anodic potentials applied to the silicon to avoid gaseous hydrogen
production and to obtain a uniform surface but in our experiment we worked at room
temperature to check the influence of hydrogen bubbles on surface morphology During
the etching the stirring was sufficient to provide a uniform concentration of reagents over
the whole sample surface but not to mechanically remove the hydrogen bubbles from the
silicon surface
After the etching each sample was rinsed in ultra-pure water to remove any etching
residues and observed with STM AFM and optical microscopy The AFM measurements
were made in air while for STM measurements the samples were rapidly dried with
nitrogen and inserted into the vacuum system to avoid surface reoxidation After insertion
into the vacuum surface cleanliness was checked with LEED and the surface morphology
observed by STM Parameters for STM measurements were sample bias +4 V 1 nA
current 16 Hz scan rate The images obtained were stable and reproducible over several
days Scan parameters for AFM were 20 nN force and 1 Hz scan rate
Some of the samples were cleaned with an RCA standard clean [1] before NH4F etching to
check the influence of possible surface contaminants on the final results Eight different
samples were prepared and more than sixty STM images of the samples were taken at
different points of the various samples
42 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
(110)
(110)
Fig47 abc) STMtopographic images of differentetching morphologies Eachimage is 500x500 nm Z-ranges are 10 10 and 18 nmrespectively d) STM image of a bridgecreated by etching of the lowerlayers of silicon (black arrow)Image is 250x250x6 nm
After ~2 min of immersion in the solution hydrogen bubbles become visible on the sample
surface The production is slow and the bubbles are quite stable on the sample without
detaching Thus some areas of the surface are masked from the liquid etching action
STM observations (fig 47) show that at the nanometer scale the surface is unevenly
covered with holes of radii ranging from 10 to 200 nm with depths of 2-4 nm These holes
have a wide range of different shapes and distributions In most cases the surface was
covered with a uniform distribution of round-shaped holes (fig47a) indicating isotropic
etching The dimension and the density of the holes changed greatly from sample to
sample and even over the surface of a single sample In some cases the etching was
anisotropic yielding nearly square holes and layered structures as shown in fig 47b
Square holes have been previously observed in cases where the etching speed in the (110)
direction is significantly smaller than in the (100) direction [17]
Over these areas (fig 47b and especially 47c) it is clearly visible how once the surface
had been attacked the reaction continued to preferentially remove atoms at step
irregularities (kink atoms) straightening step edges Eventually the exposed underlying
silicon was also attacked and further holes created inside the previously etched larger
ones It was not possible to detect monatomic steps on this kind of surface The smallest
step height observed was ~15 nm corresponding to several atomic layers In the image
shown in Figure 47a the etching was not strong and created only anisotropic holes on the
surface In fig 47b and c the stronger etch proceeded laterally for several tens of
VPalermo 43
nanometres leaving straight steps several tens of nanometers long and roughly rectangular
holes as expected given the structure of the (100) crystal face In some cases a
significative underetch is observed and the formation of suspended bridges and tunnels
can be deduced from the STM images (fig 47d)
The formation of branched pores and suspended structures has been attributed during pore
formation to diffusion limited aggregation effects where the h+ charge carriers necessary
for silicon etching have a higher probability of reacting at pore bottoms than reaching the
upper part of the silicon surface In the case of very deep pores quantum wire effects have
been invoked to explain the pore growth mechanism [15] In our case though the pores
formed were very shallow the underetch depth being only a few nanometres on pores of
sim100 nm width Thus more than diffusion effects the main contribution to the
underetching process must come from anisotropic etching and some kind of autocatalytic
reaction path analogous to the one described by Matsumura et al [4] with some areas of
the silicon surface hydrogenated and thus less vulnerable to etching
Pre-treatment with RCA cleaning has no effect on the final morphology and this seems to
exclude pore nucleation being caused by presence of metallic or organic surface
contaminants
The morphology and the distribution density of the pits was quite uniform over
microscopic areas of the sample but changes were observed over the millimetre scale This
suggests that etching intensity is influenced by some large-scale parameter
Large-area measurements made with AFM or with an optical microscope (fig 48)
showed that the inhomogeneity of surface etching can be correlated with the masking
action of the bubbles While the fluoride dissolved the silicon hydrogen bubbles formed
by the reaction covered some areas of the surface thus blocking the etching over that area
generating macroscopic steps at the bubble-liquid border As the reaction proceeded more
hydrogen accumulated and the bubble diameter increased producing in this way a circular
pattern of steps The increase in bubble diameter was not continuous with time otherwise a
uniform surface slope gradient would have been obtained The formation of this circular
ldquoetching staircaserdquo indicates that the bubble growth was stepwise the bubble accumulating
more and more hydrogen without enlarging across the surface until it relaxed increasing
its diameter stepwise and covering more silicon The circular structures in fig 48a are not
co-axial and their asymmetry could derive from physical processes due to stirring or
irregularities on the surface
44 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
a b Fig 48 a) optical micrograph of etching patterns on Si(100) created by NH4F 12x09 mm b) AFM image of the circles border xy range is 40x24 microm z-range is 30 nm
The step structure was not destroyed by the etching even after the bubble detached from
the surface but on the contrary the etching process seemed to be influenced by the
presence of the step
Observing in detail a series of steps (fig 48b) a quite deep trench is visible at the base of
each step A close-up image of a step and the corresponding line profile of the trench is
shown in fig 49 The trench is sim5 nm deep with respect to the lower surface compared to
a step height of 22 nm
A similar structure has been recently obtained with electrochemical etching of p-type
Si(100) in 4 HF [18] in which a ldquocurrent burstrdquo etching model previously described
was assumed for silicon dissolution In that experiment the trench was created at the
border of silicon nitride masks and began to grow after a nucleation stage Preferential
trench etching was along the (110) direction and the trench growth was explained as an
effect of mechanical stress induced by the nitride mask and of electric field enhanced
dissolution which depended upon an external applied potential
While it is clear that in our system the gas bubbles have a masking effect similar to a
classical solid nitride mask it is unlikely that hydrogen present on the surface can induce a
significant stress in the silicon lattice as in the case of a nitride mask Furthermore no
external field was applied to drive preferential etching at the trench site
It has been proposed [19] that the cathodic and anodic part of the etching reaction
(hydrogen production and silicon oxidative etching respectively) take place at different
points on the surface with a net charge transfer between the different areas In this case
the highest reaction rates will correspond to the silicon area surrounding the bubble border
where a high number of positive charges will be available for the reaction Furthermore a
sharp trench extending into the silicon crystal will be a preferential electrostatic attractor
VPalermo 45
Fig 49 AFM image of the etched surface showing a step created on the surface by bubblemasking A stronger etching action is visible on the right side of the step as well as theprotected area on the upper side of the step (indicated by the arrows) Image is 10x10 micromz-range is 30 nm The profile on the right is taken from the central area of the image
for the h+ charge carriers coming from other ldquocathodic areasrdquo of the sample either from
other regions on the surface or from the back of the silicon chip [18]
In the areas where the hydrogen bubble had detached and the surface was exposed to the
etching the reaction was not uniform in the neighbourhood of the steps It is possible to
observe (fig 48b and 49) an area on the upper side of the step where less or even no
etching at all seems to have taken place as if the step was able to protect the surface from
etching While etching on the lower side with trench formation can be attributed to the
presence of the bubble the surface on the upper step side can be etched only after bubble
detachment so no masking effect can account for this result However a further
preferential attraction of h+ charge carriers from the already formed trench can be
hypothised electrochemically shielding the surrounding area from further etching If this is
true the shielding effect would be very strong with a relatively shallow 5 nm-deep trench
protecting an area of sim1 microm parallel to the step
To summarise the etching of Si(100) in NH4F is a complex process in which different
reaction paths both chemical and electrochemical co-exist Hydrogen bubbles formed by
the reaction act as a mask on the surface and create etching paths and inhomogeneous
etching of the surface Different kinds of pores are observed on the surface and in some
cases the anisotropy of the process is so strong as to give square-shaped holes and
underetching
46 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The diffusion of h+ charge carriers in the crystal is one of the main rate-determining steps
of the reaction and leads to the formation of a deep trench immediately outside the bubble
perimeter These trenches act as charge collectors and reduce the etching of the upper step
surface in the proximity of the steps
Bibliography
[1] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[2] Spectroscopic Ellipsometry Studies Of HF Treated Si (100) Surfaces Yao H Woollam
Ja Alterovitz SA Applied Physics Letters 62 (25) 3324-3326 Jun 21 1993 Influence
Of HF-H2O2 Treatment On Si(100) And Si(111) Surfaces Graf D Bauermayer S
Schnegg A Journal Of Applied Physics 74 (3) 1679-1683 Aug 1 1993 Kinetics Of
Oxidation On Hydrogen-Terminated Si(100) And (111) Surfaces Stored In Air Miura
T Niwano M Shoji D Miyamoto N Journal Of Applied Physics 79 (8) 4373-4380
Part 1 Apr 15 1996
[3] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995 Structure Of The Stepped SiSiO2 Interface After Thermal-Oxidation -
Investigations With Scanning Tunneling Microscopy And Spot-Profile Analysis Of
Low-Energy Electron-Diffraction Pietsch GJ Kohler U Jusko O Henzler M Hahn
PO Applied Physics Letters 60 (11) 1321-1323 Mar 16 1992
[4] Enhanced Etching Rate Of Silicon In Fluoride Containing Solutions At pH 64
Matsumura M Fukidome H Journal Of The Electrochemical Society 143 (8) 2683-
2686 Aug 1996
[5] A Study Comparing Measurements Of Roughness Of Silicon And SiO2 Surfaces And
Interfaces Using Scanning Probe Microscopy And Neutron Reflectivity Crossley A
Sofield CJ Goff JP Lake ACI Hutchings MT Menelle A Journal Of Non-Crystalline
Solids 187 221-226 Jul 1995
VPalermo 47
[6] Step-Flow Mechanism Versus Pit Corrosion - Scanning-Tunneling Microscopy
Observations On Wet Etching Of Si(111) By Hf Solutions Hessel HE Feltz A Reiter
M Memmert U Behm RJ Chemical Physics Letters 186 (2-3) 275-280 Nov 8 1991
[7] Comparison Of Si(111) Surfaces Prepared Using Aqueous-Solutions Of NH4F Versus
HF Higashi GS Becker RS Chabal YJ Becker AJ Applied Physics Letters 58 (15)
1656-1658 Apr 15 1991
[8] Wet Chemical Etching Of Si(100) Surfaces In Concentrated NH4F Solution -
Formation Of (2x1)H Reconstructed Si(100) Terraces Versus (111) Faceting Neuwald
U Hessel HE Feltz A Memmert U Behm RJ Surface Science 296 (1) L8-L14 Oct
10 1993
[9] Ideal Hydrogen Termination Of Si(001) Surface By Wet-Chemical Preparation Morita
Y Tokumoto H Applied Physics Letters 67 (18) 2654-2656 Oct 30 1995
[10] Atomic Structures Of Hydrogen-Terminated Si(001) Surfaces After Wet Cleaning
By Scanning Tunneling Microscopy Endo K Arima K Kataoka T Oshikane Y Inoue
H Mori Y Applied Physics Letters 73 (13) 1853-1855 Sep 28 1998
[11] On The Potential-Dependent Etching Of Si(111) In Aqueous NH4F Solution
Houbertz R Memmert U Behm RJ Surface Science 396 (1-3) 198-211 Jan 20 1998
[12] Etching Process Of SiO2 By HF Molecules Hoshino T Nishioka Y Journal Of
Chemical Physics 111 (5) 2109-2114 Aug 1 1999
[13] Etching Mechanism And Atomic-Structure Of H-Si(111) Surfaces Prepared In
NH4F Allongue P Kieling V Gerischer H Electrochimica Acta 40 (10) 1353-1360
Jul 1995
[14] Pore Formation Mechanisms For The Si-HF System Carstensen J Christophersen
M Foll H Materials Science And Engineering B-Solid State Materials For Advanced
Technology 69 23-28 Sp Iss Si Jan 19 2000
[15] Porous Silicon - Mechanisms Of Growth And Applications Parkhutik V Solid-
State Electronics 43 (6) 1121-1141 Jun 1999
[16] Formation And Application Of Porous Silicon Foll H Christophersen M
Carstensen J Hasse G Materials Science amp Engineering R-Reports 39 (4) 93-141 Nov
1 2002
[17] Pore Morphology And The Mechanism Of Pore Formation In N-Type Silicon
Searson PC Macaulay JM Ross FM Journal Of Applied Physics 72 (1) 253-258 Jul 1
1992
48 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
[18] Deep Electrochemical Trench Etching With Organic Hydrofluoric Electrolytes
Christophersen M Merz P Quenzer J Carstensen J Foll H Sensors And Actuators A-
Physical 88 (3) 241-246 Mar 5 2001
[19] Morphology Of Anodically Etched Si(111) Surfaces - A Structural Comparison Of
NH4F Versus HF Etching Houbertz R Memmert U Behm RJ Journal Of Vacuum
Science amp Technology B 12 (6) 3145-3148 Nov-Dec 1994
VPalermo 49
5 Surface modification of silicon in vacuum void creation and oxide desorption
The main reason for the huge success of silicon in the microelectronics industry is not due
to its superior properties as a semiconductor Other materials for example germanium
have better qualities such as higher mobility of charge carriers and lower noise levels
which would allow the construction of faster and higher performance devices
The widespread use of silicon however is mainly due to the outstanding characteristics of
its oxide Silicon dioxide (SiO2) is a very good electrical insulator easy to form
chemically and thermally stable and is compatible with lithographic and metal deposition
processes Germanium oxide on the contrary is too reactive to be used
Even the use of Si(100) substrates for nearly all microelectronic devices is dictated by
oxide quality The (111) face of silicon crystal can be easily cleaved and flattened and
almost atomically perfect surfaces can be obtained with simple chemical procedures (as
described above) But the density of interfacial defects is highest for Si(111)-SiO2
interfaces and lowest for Si(100)-SiO2 ones so microchips will continue to be fabricated
on Si(100) wafers
SiO2 (silica) is present in 95 of the earthrsquos minerals in different allotropic forms such as
quartz tridymite and cristobalite In the bulk each silicon atom is bonded to four oxygens
in a Si-O-Si tri-dimensional network Si-O bonds are 016 nm long and form an angle
ranging from 120deg to 150deg
Three typical intrinsic defects are present in SiO2 The so-called Ersquo centres are oxygen
vacancies with a hole localised on a silicon atom with only three Si-O bonds
50 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
O3Si +SiO3 Whereas the PR (peroxy radical) defects are holes trapped by a charged
peroxy moiety with a O3Si-O-O+ SiO3 structure The NBOHC (non-bridging oxygen hole
centres) derive from water or hydrogen contamination and are schematized as O3Si- O- H-
O-SiO3
The atomic structure of the Si-SiO2 interface varies enormously Local domains resembling
the tridymite and the cristobalite structure of silica are present but it seems that only 10
of the interface is ordered [1] Far from the interface the SiO2 bulk is completely
disordered The passage from bulk Si to stoichiometric SiO2 passes through a non-
stoichiometric SiOx layer 07 nm thick
When a clean silicon surface is exposed to atmospheric oxygen a thin sim2 nm thick layer
of native oxide forms spontaneously which is usually removed and substituted with
thicker better quality oxide layers before further processing
Silicon is usually oxidised by thermal annealing at temperatures between 800deg and 1100deg
in an atmosphere of pure O2 with some water eventually added to increase oxidation speed
Thermal oxides made in pure oxygen (dry oxides) grow more slowly than oxides produced
in an oxygen-water atmosphere (wet oxides) but are usually of better quality
According to the Deal-Groove formula the time t needed to grow an oxide of thickness X
is given by [1]
1
212minus
minusminus
+=
ABXBXt α
where the constant B and BA decrease exponentially with temperature as
minus
kTEexp
with activation energies for dry oxidation of EB =123 and EBA =20 eV respectively
EB is related to the diffusion of oxygen in silicon while the value of EBA is interpreted as
the energy required to break a Si-Si bond The exponent α is 1 for wet oxidation and 0 for
oxidation at high temperatures and low oxygen pressures It has intermediate values for dry
oxidation This formula does not work well for low values of X at the initial stages of
oxidation and usually empirical corrections are used
An interesting characteristic of silicon is that at high temperature and in vacuum oxygen
can actually etch the silicon crystal giving gaseous products and the oxidized layer present
on the crystal becomes unstable (fig 51) [2]
VPalermo 51
1E-10
1E-09
1E-08
1E-07
1E-06
1E-05
1E-04
1E-03
1E-02
1E-01
0607080911112
1000T (1K)
P (T
orr)
SiO2 + Si rarr 2SiO(g) Oxide decomposition
Si+O2 rarr SiO2(s) oxide formation Fig51 phase diagram of the
oxygen-silicon system
Silicon oxidation apart from the initial nucleation stages at the monolayer level proceeds
uniformly over the whole surface with a planar reaction front moving from the surface
into the bulk
If heated under low oxygen partial pressure (vacuum or inert atmosphere) SiO2 is known to
decompose following the reaction
SiO2 + Si rarr 2SiOuarr (1)
The reaction begins with nucleation at defect points on the SiSiO2 interface and proceeds
in a spatially inhomogeneous manner with the formation of large voids on the oxide
surface [3]
Several studies have been made on the dynamics of void growth both on thick [4] and thin
[5] layers of SiO2 The process has been used to decorate otherwise unobservable defects at
the SiSiO2 interface [6] or to grow nanoislands of silicon on the void surface [7] It has
been suggested that the defects acting as nucleation centres could be metallic contaminants
present on the native surface which aggregate and catalyze SiO2 decomposition [8]
After oxide desorption the silicon surface is very rough In particular on Si(100) square
islands are observed several nanometres high which act as lsquopinning sitesrsquo on the motion of
monatomic steps on the Si surface It has been proposed that these islands can be composed
of silicon carbide coming from the organic contaminants present on the surface [9] or of
oxide clusters during partial reoxidation [10]
In the following section some experimental results obtained studying voids and nanoisland
growth are presented
52 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nanoisland growth on silicon
We used two different samples the first rougher than the second (i) p-type Czochralski Si
(100) with a RMS roughness of 05 nm shown in fig52a and (ii) p-type epitaxial Si(100)
with a RMS roughness of 019 nm shown in fig 52b The roughness was measured with
STM on the lsquoas receivedrsquo samples on different areas of 500x500 nm Every sample was
covered by a layer of native oxide ~2 nm thick
Some of the samples (AC) were introduced into the UHV system without any cleaning
others (BD) were dipped in concentrated HF (49) to remove the surface oxide without
etching the silicon [11]
After degassing the samples were heated resistively increasing the temperature slowly to
keep the pressure within the 10-10 mbar range during heating Each sample was held at
900˚C for 30 min For some of the samples the heating was stopped at lower temperatures
to obtain incomplete oxide desorption and surface roughening thus enabling observation
of the different steps of the process Table 51 summarizes the different treatments for each
group of samples
Fig52 Original surfaces of Czochralski (a) and epitaxial (b) silicon Every image is 500x500 nm
VPalermo 53
Table 51 Summary of different treatements
Silicon type Oxide removed
Heating Islands density (microm-2)
Av Island Volume (nm3)
A Czochralski No 30rsquo at 900˚C 24x103 353
B Czochralski Yes ldquo 20x104 71
C Epitaxial No ldquo 28x102 2450
D Epitaxial Yes ldquo 23x104 63
Fig53 ab) Surface roughening on group A samples after heating 30 min at 800deg and900deg respectively in presence of an oxide layer Image size 500x500 nm cd) Surface roughening on group B samples after heating 30 min at 700deg and 900degrespectively after removing the oxide layer with HF Image size 250x250 nm
Fig 52a is the native oxide surface of sample A quite disordered and irregular This
surface was stable when heated to 700˚C while at higher temperatures the oxide began to
desorb Fig 53a was taken after heating the sample at 800˚C for 30 min
54 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The oxide began to desorb in correspondence with defect points creating oxide-free
surface voids which enlarge radially uncovering the silicon surface Some surface silicon
atoms are removed through reaction (1) Mobile silicon atoms moving randomly over the
surface aggregate forming a nanocrystal at the initial defect point surrounded by a stable
flat oxide-free surface with some steps The initial SiSiO2 interface as observed after
oxide etching with HF [12] is similar to the one shown in fig 52a and thus the surface
observed in the void area is a completely new one coming from surface reorganization
The reaction continues leading to a coalescence of the voids the final result of the process
is shown in fig 53b with protruding islands aligned along the main crystal axes and flat
areas with some steps LEED measurements confirmed that this surface is crystalline
silicon Atomic resolution was obtained on the flat areas of the surface showing the
typical 2x1 reconstruction of Si(100) Figures 53c and 53d show the evolution of the
roughening on sample B after chemically removing the native oxide layer before
introducing the sample into UHV As mentioned above the initial morphology is similar to
that shown in fig 52a though the oxide has been removed as confirmed by LEED
measurements Fig 53c shows the surface after heating at 700˚ for 30 min Without the
oxide layer the surface began to reorganize at lower temperatures and the small irregular
grains of the original surface begin to grow and become more rectangular After heating to
900˚C (fig 53d) the surface is flat with small rectangular islands The shape of the islands
is similar to that shown in fig 53b but the lateral island dimension is one order of
magnitude smaller
To assess the importance of the initial surface on the process the same treatments were
repeated starting from flat epitaxial silicon (samples CD) The initial epitaxial surface is
smooth with periodic steps due to a small miscut angle (fig52b) On heating the epitaxial
silicon without etching (sample C) relatively large islands are obtained similar to the ones
shown in fig 53b where the ldquostep-pinningrdquo effect of the islands is evident The average
dimension of sample C islands was sim2500 nm3 while only 353 nm3 for sample A Island
density decreased from 24000 to 280 microm-2 This result is maybe due to a lower
concentration of defects in the oxide layer of the epitaxial silicon which allows the
expanding void to grow much larger before coalescence with neighbouring voids the
nanostructures within them thus reaching greater dimensions To verify this hypothesis we
applied the same treatment to clean not-so clean and purposely contaminated samples To
check that the defects inducing void nucleation and nanoisland growth derive from carbon
VPalermo 55
contamination we used electronic-grade methanol to contaminate the silicon surface
Figure 54 shows the results of this comparison
State-of-the-art cleaned samples obtained with multi-step RCA cleaning [13] showed the
development of very few voids the oxide desorption leaving large smooth areas of oxide-
and nanostructure-free silicon as shown in Fig 54a where the presence of residual oxide
can be seen Samples cleaned with a simpler one-stage Pirana cleaning solution showed a
higher number of islands per unit area (Fig 54b) and samples purposely contaminated
with methanol (Fig 54c) had the highest density of island nucleation of all three samples
When the native oxide had been chemically removed from the epitaxial silicon before
heating (sample D) the island size and density is similar to the standard as-received
silicon case (sample B) Thus starting from a flat substrate the process of surface
roughening and island creation was the same even when the oxide layer had been
previously removed
From these results some important indications can be drawn The first is that oxide
desorption temperature depends upon the contamination levels of the oxide layers (fig 54)
with desorption at lower temperatures for contaminated surfaces
Second once the oxide layer is removed the silicon atoms become mobile on the surface
at relatively low temperatures (~700˚C) Their initial configuration is disordered and has a
high surface energy so they crystallize in the small islands shown in fig53d reducing
their surface area and lowering the energy of the system When the oxide is present the
surface cannot reorganize at 700˚C and is stable At 800˚C the oxide begins to desorb in
correspondence with defect points on the surface As the void area enlarges a small
cba Fig 54 The effect of contamination on nano-island production Gradient-filtered STM images of (a)sample cleaned two times with RCA 400x400 nm (b) Sample cleaned with a Pirana solution 500x500nm (c) sample contaminated with MeOH before insertion into the UHV system 500x500 nm Annealingtime is 40 min at 800degC for all the samples
56 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
crystallite grows at the centre of the voids Further oxide desorption uncovers larger areas
freeing more silicon atoms which feed the initial island making it bigger
Fig 55 shows the scaled size distribution of the islands for each group of samples Every
distribution has been obtained using several images of different samples for each group
The size distribution of all the samples follows an exponential decay and the decay seems
the same for all the samples though the average volume of the islands differs by more than
one order of magnitude (see Table 51 for details)
The point defects which catalyze void nucleation can be metallic contaminants present on
the initial oxide surface [8] or organic contaminants which at high temperatures can
form SiC nanocrystals [14] We used Scanning Tunneling Spectroscopy (STS) to look for
differences between the islands and the surrounding flat silicon surface but no difference
was found Furthermore the total island volume per surface unit is very high (more than
104 nm3microm2) and it seems unlikely that such a huge volume could consist of surface
contaminants Islands of pure silicon on silicon have been grown without evidence of
surface contamination[14]
To summarise the formation of nano-sized islands on silicon through surface diffusion
was studied At high temperatures the oxide layer covering the surface decomposes non
uniformly and circular voids of clean silicon are created The presence of the oxide layer
blocks surface silicon atomic motion and surface reorganization except within the voids
causing the growth of islands more than 10 nm high and 30 nm wide on the silicon surface
According to STM and LEED measurements we can say that the islands are mostly
Fig 55 Size distribution (scaled) for the samples described in table 1 The line is a reference for the eye
VPalermo 57
composed of silicon with traces of other substances such as carbon Final island densities
and dimensions depend upon the initial purity of the oxide layer
On the other hand if the oxide is removed chemically before heating the island growth
process is not so localised and takes place simultaneously over the whole surface yielding
smaller and more numerous islands
58 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[2] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[3] High-Temperature SiO2 Decomposition At The Sio2Si Interface Tromp R Rubloff
GW Balk P Legoues FK Physical Review Letters 55 2332-2335 Nov 1985
[4] Defect Microchemistry At The SiO2Si Interface Rubloff GW Hofmann K Liher M
Young DR Physical Review Letters 582379-2382 Jun 1987 Defect Formation In
Thermal SiO2 By High Temperature Annealing Hofmann K Rubloff GW Mccorkle
RA Applied Physics Letters 49 1525-1527 Dec 1986 Kinetics Of High-Temperature
Thermal Decomposition Of SiO2 On Si(100) Liher M Lewis JE Rubloff GW Journal
Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 5 1559-1562 Aug
1987
[5] Thermal-Decomposition Of Very Thin Oxide Layers On Si(111) Kobayashi Y Sugii
K Journal Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 10 (4)
2308-2313 Part 3 Jul-Aug 1992 Scanning Tunneling Microscope Study On Mid-
Desorption Stages Of Native Oxides On Si(111) Kobayashi Y Sugii K Journal Of
Vacuum Science amp Technology B 9 (2) 748-751 Part 2 Mar-Apr 1991 Controlled
Growth Of SiO2 Tunnel Barrier And Crystalline Si Quantum Wells For Si Resonant
Tunneling Diodes Wei Y Wallace RM Seabaugh AC Journal Of Applied Physics 81
(9) 6415-6424 May 1 1997
[6] Defect Formation In SiO2Si(100) By Metal Diffusion And Reaction Liher M
Dallaporta H Lewis Je Appl Phys Lett 53 589-591 Aug 1988 SiO2 Film
Decomposition Reaction Initiated By Carbon Impurities Located At A Si- SiO2
Interface Raider Si Herd Sr Walkup Re Applied Physics Letters 59 (19) 2424-2426
Nov 4 1991
[7] Nanometer-Scale Si Selective Epitaxial Growth On Si(001) Surfaces Using The
Thermal Decomposition Of Ultrathin Oxide Films Fujita K Watanabe H Ichikawa M
Applied Physics Letters 70 (21) 2807-2809 May 26 1997 Pyramidal Si Nanocrystals
VPalermo 59
With A Quasiequilibrium Shape Selectively Grown On Si(001) Windows In Ultrathin
SiO2 Films Shibata M Nitta Y Fujita K Ichikawa M Physical Review B 61 (11)
7499-7504 Mar 15 2000
[8] Stacking-Fault-Induced Defect Creation In SiO2 On Si(100) Liher M Bronner Gb
Lewis Je Appl Phys Lett 52 1982-1985 May 1988
[9] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990 AFM And XPS Characterization Of The
Si(111) Surface After Thermal-Treatment Lamontagne B Guay D Roy D Sporken R
Caudano R Applied Surface Science 90 (4) 481-487 Dec 1995
[10] Scanning-Tunneling-Microscopy Study Of Oxide Nucleation And Oxidation-
Induced Roughening At Elevated-Temperatures On The Si(001)-(2x1) Surface Seiple
JV Pelz JP Physical Review Letters 73 (7) 999-1002 Aug 15 1994 Evolution Of
Atomic-Scale Roughening On Si(001)-(2x1) Surfaces Resulting From High-
Temperature Oxidation Seiple JV Pelz JP Journal Of Vacuum Science amp Technology
A-Vacuum Surfaces And Films 13 (3) 772-776 Part 1 May-Jun 1995
[11] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995
[12] Morphological Changes Of The Si [100] Surface After Treatment With
Concentrated And Diluted HF Palermo V Jones D Materials Science In
Semiconductor Processing 4 (5) 437-441 Oct 2001
[13] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[14] Nanoscale Roughening Of Si(001) By Oxide Desorption In Ultrahigh Vacuum
Gray SM Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology
B 14 (2) 1043-1047 Mar-Apr 1996
60 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 61
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and nano-lines
When the silicon surface is not protected by a native oxide layer or by a hydrogenated
passivating layer it is very reactive The surface chemistry of both Si(100) and Si(111) has
been extensively studied starting from clean surfaces prepared in UHV because of its great
relevance to the microelectronics industry and its technological interest The diffusion and
reaction of molecules and atoms on silicon is an interesting scientific problem on its own
apart from technological issues because silicon reconstruction yields a very complex and
anisotropic surface
We give here a brief summary of the adsorption behaviour of several elements on Si
surfaces The summary is not comprehensive and only the most interesting characteristics
for each substance are given
Table 61 Summary of adsorption behaviour of atoms and molecules on silicon surfaces [12]
Hydrogen
Molecular H2 shows low reactivity towards silicon while atomic hydrogen easily forms
Si-H bonds and can even break Si-Si bonds
Alkali metals
Alkali metals diffuse rapidly into SiO2 and can damage silicon-based transistors
Transition
metals
All transition metals apart from gold and silver react with Si forming metal silicides
62 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nickel A very common contaminant of silicon Nickel from even slight contact with stainless
steel tweezers can contaminate silicon samples forming its silicide and is very difficult
to remove even at high temperatures because it segregates on the silicon surface
Palladium Forms silicides especially Pd2Si and can be used to make contacts possessing a low
Schottky barrier
Titanium Widely used to fabricate contacts for silicon-based microdevices TiSi2 is one of the
more stable and highly conductive silicides
Tungsten The silicide is more stable than TiSi2 on polysilicon and is thus used for polysilicon
gate contacts
Platinum Silicide is used for bipolar transistors PtSi contacts on n-silicon give Schottky diodes
Cobalt Cobalt silicide gives better quality contacts than TiSi2 at the nanometrelevel but is less
used at the moment because it consumes too much silicon during its growth
Iron Silicides have been studied for potential optoelectronic applications because FeSi2 has a
directly accessible bandgap and is optically active
Group 13
Group 13 metals are used in IC technology to dope silicon (giving p-type doping) and
to make III-V type semiconductors When evaporated on Si(100) or Si(111) they react
strongly with the surface giving surface reconstruction and kink formation at steps
Boron Widely used for doping Usually deposited with decomposition of boron hydrides
Sticking coefficient of simple hydrides is very low so decaboranes (B10H14) are used for
deposition
Aluminum Aluminum is widely used for microelectronic contacts but in some cases it reacts with
silicon diffusing into the crystal and with SiO2 stealing oxygen atoms to form Al2O3
Group 14
Carbon Carbon can form a composite semiconductor with silicon (silicon carbide) widely used
to fabricate integrated circuits for use at elevated temperatures or in the presence of
ionising radiation When carbon is deposited on silicon usually polycrystalline films
with properties between diamond and graphite are obtained
CO adsorbs molecularly on silicon without breakage of the C=O bond Other
unsaturated hydrocarbons do not react with silicon Even very reactive strained
molecules like cyclopropane have low sticking coefficient Ethylene acetylene and
propylene adsorb molecularly on Si(100) each molecule sitting on a silicon dimer
interacting with the dangling bonds of the surface Acetylene also adsorbs molecularly
on Si(111) on the 7x7 reconstructed surface preferential adsorption on centre atoms
with respect to corner atoms is observed Benzene and other aromatic ring compounds
adsorb molecularly on the surface They can form σ or π-type bonds with the silicon
laying parallel or tilted respect to the surface according to the bond type
Silicon Silicon atoms can be deposited on the surface with molecular beam epitaxy or
VPalermo 63
decomposition of silanes and chlorosilanes Silanes with single Si-Si bonds have a high
sticking probability and dissociate upon adsorption giving trihydrides and
monohydrides
Chlorosilanes have high sticking coefficients too and decompose on the surface but
they can etch the surface through the reaction SiCl4 + Si rarr 2SiCl2
Germanium Germanium has a lattice constant similar to silicon (Ge lc is 4 larger than Si) so that
various alloys of Si1-xGex can be formed The growth of Germanium on silicon is quite
peculiar with several monolayers adsorbing uniformly on the surface followed by
island formation (Stranski-Krastanov growth)
Group 15
Nitrogen Exposure to ammonia and nitrogen gas at high temperatures leads to the formation of
silicon nitride layers (Si3N4) NH3 dissociates on the surface and reacts with the
dangling bonds giving Si-NH2 and Si-H termination Silicon nitride is stable on the
surface up to 1100degC above this temperature it desorbs as Si2N
Phosphorous Phosphine (PH3) adsorbs dissociatively as Si-PH2 and Si-H and behaviour is similar to
that of ammonia
Arsenic A monolayer of As is often deposited on silicon as a substrate for GaAs growth Arsenic
forms dimers on Si(100) creating dimer rows which are parallel (perpendicular) to the
underlying silicon dimer rows when it is deposited at high (low) temperature On
Si(111) Arsenic breaks the 7x7 reconstruction giving a 1x1 pattern This 1x1 As
monolayer acts as a passivating layer and can resist further As adsorption oxygen and
air
Antimony and
Bismuth
Due to their large covalent radii Sb and Bi form only short dimer rows on silicon and
only at high temperatures
Group 16
Oxygen
Oxygen can oxidize or etch silicon according to the reactions
Si(s) + O2 rarr SiO2 (s)
Si(s) + frac12O2 rarr SiO(g)uarr
Silicon oxidation has already been described in the previous chapter Molecular oxygen
adsorption on silicon is mostly dissociative O atoms break the Si-Si bonds forming an
Si-O-Si bridge or a peroxy bridge Si-O-O-Si Hydrided silicon surfaces do not have
surface dangling bonds and are thus more protected from oxygen attack Etching is
favoured at high temperatures and low oxygen pressures [3] At T gt700degC silicon oxide
on silicon decomposes as SiO(g)
Water
H2O adsorbs dissociatively on silicon easily on Si (100) and with more difficulty on
Si(111) The molecule decomposes giving Si-OH and Si-H on adjacent sites
64 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Group 17
Fluorine
Fluorine adsorbs and reacts easily on silicon dissociating on the surface and forming
Si-F bonds It can then penetrate the surface and attack polarised Si-Si backbonds The
surface is etched with production of SiF4 and SiF2
Chlorine
Chlorine reacts aggressively with silicon etching the surface Cl2 and F2 are often used
commercially to etch silicon usually in a plasma Atomic Cl and Cl2 adsorb on Si(100)
up to saturation The most stable configuration seems to be a Cl atom bonded to each Si
atom of a surface dimer in a symmetric structure but metastable configurations with
two Cl atoms bonded to a buckled dimer and even a central Cl atom bridged across a
dimer have been observed
On Si(111) silicon mono- di- and tri-chlorides are formed In particular heating at
Tgt400degC a uniform Si-Cl monolayer is obtained and the 7x7 reconstruction changes
into a bulk like 1x1 lattice The surface transition induced by Cl adsorption is fully
reversible The 7x7 reconstruction can be restored on desorbing the chlorine by heating
at Tgt1100degC The 1x1 domains nucleate at the lower terrace side of steps the 7x7 at the
upper step edge
Bromine
Like hydrogen bromine maintains the 2x1 structure of Si(100) forming Si-Br bonds
with the dangling bonds of the silicon dimer rows At high exposure some etching of
silicon by formation of volatile SiBr3 species has been observed
Surface diffusion on silicon
In the previous chapter we observed nano-island growth on silicon surfaces with native
oxide and surfaces contaminated with organic impurities While this phenomenon has been
observed in several experiments the formation and growth mechanisms of nanoislands has
still not been satisfactorily described theoretically Various substances have been
hypothesised as nanoisland nucleating agents such as organic or metallic contaminants
present on the surface or oxide clusters (see previous chapter) The island growth process
has been attributed by various workers to lsquosome sort of kinetic instabilityrsquo [4] to the
pinning of step flow by SiC clusters [5] or to a mesoscopic atom flux from areas with low
step density to areas of higher step density [6]
One mechanism proposed to explain surface roughening and island growth is the Ehrlich-
Schwoebel effect ie the presence of an energetic barrier that reflects atoms approaching a
VPalermo 65
Molecule (L
O2
H2O MeOH
CO CO2 CH4
clean surface
downward
weak on si
The aim o
surface an
To obtain
the ones
spontaneou
first obser
surface W
nanoisland
Experime
We used S
removed f
41 mixtur
water and
whole hea
1200degC re
kept below
Table 62 Treatment island density and island volume for each molecule tested
Dose angmuir)
Heating time at 800degC (min)
Island density (microm-2)
Ave Volume (nm3)
Equivalent layer thickness (nm)
280 10 no islands - -
90 10 no islands - - 30 10 2000 19 004
180 10 360 516 019 30 60 no islands - -
180 10 no islands - - - 10 no islands - -
step [7] Recent experiments however indicate that the Schwoebel effect is
licon [48]
f our experiments was to study the reaction of simple molecules with the silicon
d to check the ability of these molecules to nucleate nanoisland growth
a better understanding of the process we chose simple molecules starting from
which are more likely to be present as traces in UHV chambers where
s nanoisland growth is often observed after sample heating Using STM we
ved at the atomic level the possible interactions of these molecules with the
e then heated the surface and checked the density and dimensions of the
s produced
ntal procedures and results
i(111) wafers p-doped 0015 Ω cm Gross contamination and particles were
rom the sample surface with a standard Pirana etch (15 min dip in a H2S4H2O2
e at T=80degC) After this the samples were thoroughly washed in ultra-pure
introduced into the UHV system After degassing for 5 hours at 600degC the
ting stage was allowed to cool down and a sequence of rapid flashes to 1100deg-
moved the native oxide layer and cleaned the surface Pressure during flashes is
1x10-9 mbar
66 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Surface flatness and cleanliness were checked with STM and the contaminant molecules
introduced into the UHV system through a leak valve Doses of several Langmuirs were
used for the experiments (1 Langmuir=10-6 torr sec-1) Pressure was kept constant at 10-7
mbar during adsorption The STM tip was retracted during gas deposition to avoid any tip
shielding effect After the adsorption process the leak valve was closed the base pressure
in the UHV chamber returning to below 10-10 mbar and the presence of adsorbed
molecules checked on the surface with STM Finally the sample was heated at 800degC for
several minutes to allow nanoisland growth Several different areas were examined on each
sample
Table 62 summarizes the dose of exposure and the heating time for each molecule Other
experiments were carried out using smaller dosing times but only the significant results
are reported here
As expected oxygen and water react with the surface oxidising it The adsorption process
can be easily followed with the STM by lowering the tip from time to time and taking an
image No effect of the STM tip is observed over the scanned areas Upon molecule
adsorption the order of the crystalline surface rapidly degenerates and after several
minutes the surface looks completely covered by irregular atomic-sized protrusions
corresponding to Si-OH or Si-O-Si species On heating to 800degC the contaminating species
were easily desorbed through SiO formation the surface became clean and the crystal
surface periodicity re-established No residues or nanoislands were observed
Amongst the molecules tested was methanol CH3OH reacts readily with the surface and
covers it with a disorderd layer already after sim 30 Langmuirs exposure (fig 61)
Fig 61 Si(111) surface during methanol adsorption at 0 1 and 5 minutes respectively (corresponding to 0 6 and 30 L exposure) The dark irregular lines are borders between different 7x7 domains Image size 60x60 nm
VPalermo 67
Synchroton radiation photoemission spectroscopy experiments by Carbone et al [9]
showed that methanol reacts with the rest atoms of the 7x7 silicon cell already after 1
Langmuir exposure but that longer exposures are needed for methanol to react with the
remaining silicon atoms Methanol adsorbs dissociatively forming SiOCH3 and Si-CHx
species on the surface
After methanol adsorption the sample is heated to 800degC and observed again with STM
As in the case of water and oxygen the heating restores the crystalline surface and the 7x7
pattern returns visible But this time quite a dense array of nanoislands is observed on the
surface (fig 62) The islands have an average diameter of 20 nm and a height of sim2 nm
At temperatures above 400degC the Si-OCH3 and Si-CHx species decompose oxygen and
hydrogen are desorbed and according to [9] carbon atoms remain as SiC dispersed
uniformly on the surface Our STM measurements indicate that the carbide does not
randomly cover the surface but that C atoms are concentrated at the nanoislands Rough
calculations based on island volume suggest that the islands must be composed of a SixC1-x
alloy with x varying between 05 and 1
Following these results we expected carbon monoxide to behave in a similar manner
adsorbing onto the surface and with nanoisland growth However no adsorption was
observed with STM and the surface looked perfectly clean even after a dose of more than
100 L of carbon monoxide Some species did adsorb onto the surface though because
after 10 minutes subsequent heating nanoisland growth was observed Island density in
this case was only 18 of the density obtained with methanol and island dimensions are
Fig62 Nanoislands on Si(111) created after 30 L adsorption of methanol and 10 min heating at 800degC Image size 200x200 nm
68 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
much bigger (see table 62 and fig 63) To check whether the islands had been nucleated
by some other contaminant present in the vacuum chamber we heated a blank sample
without introducing any molecule No island growth was observed Exposing the surface to
other simple molecules including carbon dioxide and methane also gave no nanoisland
formation
We can estimate the amount of carbon deposited on the surface during gas dosing The
total flux of molecules hitting the surface is obtained from the Hertz-Knudsen formula
TkmPF
π2=
Fig 63 STM images of Si(111) after exposure to methanol (left) and carbonmonoxide (right) contaminating molecules and subsequent nanoisland growthImage size 400x400 nm The size distribution of nanoislands is shown below eachimage
VPalermo 69
Fig64 Left Si(111) after 180 L contamination of CO and 16 hours heating at 800degC Right Si(111)without CO contamination after 16 hours heating at 800degC Image size 250x250 nm Vertical scale is thesame for both images z-ranges are 35 and 5 nm respectively
where P is the pressure in Pascal k=138x10-23 J K-1 is the Boltzmann constant T is the
temperature and m is the molecular weight (sim32 for methanol and sim28 for CO) in
kilograms
Using this formula we obtain a flux of 027 molecules nm-2 sec-1 for methanol and 029
molecules nm-2 sec-1 for CO After ten minutes exposure at 10-7 mbar pressure more than
150 molecules will have hit each square nanometer of the surface Surface density of
Si(111) 7x7 is sim16 atomsnm2 so each surface atom will be hit by several molecules which
could react or be adsorbed even assuming a sticking coefficient much lower than unity
To explain nanoisland growth we hypothesize that CO molecules adsorb molecularly and
-5
0
5
10
15
20
25
30
0 20 40 60 80 100 120
nm
nm
10 min40 min16 h16 h- no CO
Fig65 STM profiles of nanoislandsgrown for different annealing times The lower curve correspond to the surface heated at 16 hours without CO contamination
70 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Table 63 Island density and volume for the CO-contaminated surface at different heating times For each sample the thickness of a uniform layer having the same volume per unit area of the islands is calculated
Molecule Dose
(Langmuir) Heating time at
800degC (min) Island density
(microm-2) Ave Volume
(nm3) Equivalent layer thickness (nm)
CO 180 10 362 517 019 CO 180 40 550 1072 059 CO 180 16 h 505 3675 186
clean surface - 16 h rough - -
very weakly on Si(111) moving rapidly on the surface Because of their rapid diffusion
the STM is not able to image the CO molecules on the silicon [10]
Increasing the temperature part of the CO molecules will surely desorb but some of them
will dissociatively adsorb onto the surface The oxygen will then desorb as SiO while
carbon will remain fixed forming very stable Si-C bonds
CO decomposition will be favored at surface defects surface steps or in the proximity of
already formed SiC clusters An increased reactivity of the CO molecule on a surface in
correspondence with phase boundaries has already been observed on Pt surfaces [10]
In this case a reduced number of nucleation centres will react with the CO molecules
yielding large and fewer islands with respect to the case of methanol which does not have
the possibility of travelling long distances over the surface
Other factors are likely to influence the process of island nucleation after CO adsorption
the co-adsorption of other molecules which can slow down CO and favour its
decomposition cannot be ruled out as well as the formation of new defects at higher
temperatures Measurements with a variable temperature STM or with some other surface
analysis technique are planned in the future to confirm the proposed mechanism
When the surface was further annealed the nanoislands grew in size After 16 hours
heating very large nanoislands with diameters of sim35 nm and heights of sim20 nm became
visible (fig64 left) As a comparison when the same surface was heated without
nucleation centres deriving from contamination an irregular surface was obtained (fig64
right)
While the nanoisland density seemed to reach an asymptotic limit of 500 islandsmicrom2
(table 63) island size continued to grow with time (fig65) even when no further carbon
was supplied to the surface This indicates that even though the initial nucleating core of
the nanoisland is likely to be an SixC1-x alloy further growth is due to silicon atoms
diffusing from the crystal and being adsorbed by the growing island The final volume
VPalermo 71
occupied by the islands corresponds roughly to a 18 nm thick overall layer of removed
silicon
It is noteworthy that the clean sample after 16 hours annealing even though smoother than
the CO-contaminated one looked much more disordered with hardly any flat area visible
on nanometres scale In the presence of nanoislands however further annealing increases
island size but keeps part of the surface quite flat and ordered with flat areas visible
between the islands Thus the nanoislands can be imagined to act as ldquoimpurity sinksrdquo for
further contaminants approaching the surface yielding a greater but more ordered surface
roughening with respect to the uncontaminated silicon surface
Circular arrays of nanoislands
We used the voids described in the previous chapter formed during oxide layer
decomposition as nanoscopic masks to test the validity of the hypothesis outlined above
The small circular areas uncovered during oxide decomposition can act as ldquoskating rinksrdquo
where contaminants and diffusing silicon atoms can move freely over the surface while
the oxide layer all around and higher than the void area as well as being chemically
bonded to the surface silicon atoms will hinder surface diffusion Furthermore the oxide
passivating layer will at least in some cases prevent contaminating molecules from
adsorbing and decomposing on the surface In this way selective nanoisland growth inside
void areas can take place
The experimental procedure consisted of the following steps
1) An oxide-passivated sample was heated for 10 minutes at 800degC Oxide decomposition
began in correspondence with surface defects and spread laterally creating naked areas
of silicon
2) While the void enlarges the same defect that initiated oxide decomposition behaved as
an attractor for moving silicon atoms and nucleated the growth of a nanoisland at the
void centre
3) The void surface was exposed to contaminating molecules
4) The sample was then heated again at 800degC During the second heating the void
continued to enlarge and new nanoislands were created The freshly uncovered silicon
surface provided mobile silicon atoms to feed nanoisland growth
72 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
5) When the heating continued for long enough all the oxide was desorbed and a clean
silicon surface covered by circular groups of nanoislands was obtained
Figure 66 shows the different stages of the process
We thus heated an oxidized sample at 800degC for 10 minutes in UHV to create multiple
voids in the oxide layer After this we exposed the sample to the contaminating molecules
Then we further annealed the surface at the same temperature to promote island growth
The first time we observed nanoisland growth in the void area was by chance after heating
for a second time samples which had been stored for 2 weeks in vacuum Even at pressures
as low as 10-11 mbar some molecules will adsorb onto the surface slowly creating defects
and thus potential nanoisland nucleation centres It is possible to observe with STM that
the silicon crystal surface looks very clean just after a high temperature flash but even
after only a few days randomly adsorbed species will be observable on the surface
Unfortunately it is not possible to identify these species simply from in situ STM
measurements
Mass spectrometry measurements show that the residual gas contaminants in vacuum are
usually He Ar H2 CH4 CO CO2 and N2 [11] Some of these molecules such as
hydrogen will stick to the silicon surface at room temperature but will simply desorb
during annealing without nucleating nanoislands Others like CH4 or CO2 will not react
with the surface to generate nanoislands as demonstrated previously The best candidate
for contamination of samples stored in UHV seems thus carbon monoxide has the ability
321
4 5
Fig 66 Selective nanoisland growth within oxide voids See text for details
VPalermo 73
as shown before to nucleate nanoislands although co-adsorption mechanisms can not be
ruled out Fig 67 (left) shows the circular groups of nano-islands obtained after heating
void-covered samples stored for 2 weeks in UHV Small islands are observed in the former
void area The void itself has enlarged slightly uncovering a clean oxide-free area of the
silicon surface Apart from the central island created during the first period of heating the
surrounding ones have dimensions decreasing from the void border to the inner area of the
void
This confirms that most of the material needed for island growth comes from the silicon
atoms diffusing from the freshly uncovered areas around the original void The outer
nucleating centres are nearest to the silicon atom source and thus generate larger islands
For a faster process we directly exposed the void to a significant concentration of CO
Fig67 (centre) shows the islands grown after a 180 Langmuir exposure and subsequent
heating The islands are preferentially located around the original void perimeter the rest
of the original oxide-free surface remaining untouched This suggests that although the
whole surface is exposed to CO the gas only interacts chemically with the surface at the
SiSiO2 interface around the oxide-free void perimeter The SiSiO2 interface thus provides
preferential nucleation sites for nanoisland growth in the successive thermal annealing
step This recalls a similar phenomenon observed in the catalytic oxidation of CO on the
Pt(111) surface where CO molecules reacted with adsorbed O species only along the
perimeters of oxygen islands [10]
On repeating the CO exposure and thermal annealing steps a second circle of nano-islands
Fig67 Left circular area of nanoislands obtained from a sample stored 2 weeks in UHV 500x500 nmCentre nanoisland circle obtained after exposing the voids to 180 L of CO and heating for 10 min at800degC 400x400 nm Right two concentric circles obtained with further CO adsorption followed byheating 650x650 nm
74 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
was formed around the newly-expanded surface void perimeter (fig67 right) This
process can in principle be repeated again to obtain a series of concentric groups of
nanoislands the only limit depending on the minimal distance between neighbouring
expanding voids We have thus observed that contaminants adsorbed at low pressure over
long periods of time generate random island growth over the whole void area while CO
molecules adsorbed at relatively high pressures over short periods yield selective
nanoisland growth at the former void border Several theories can explain the different
results CH4 and CO2 as mentioned above did not nucleate islands when when the silicon
surface was exposed to them for short periods However the lower surface mobility of
these contaminants and the greater time available for reaction with the surface could be
the cause of nucleation of SiC clusters over the whole oxide-free void area and not only at
its border The coadsorption of water another UHV residual gas with CO on the naked
silicon areas could perhaps allow a surface reaction similar to that between CO and the
SiSiO2 interface during CO exposure Another explanation for the formation of these
random nucleation sites within the voids could be the coadsorption of residual carbon-
containing species present in the UHV system with hydrogen the major residual gas in
stainless steel UHV systems
The possibility of decorating silicon oxide borders with nanoislands using the high
mobility and selective decomposition of CO molecules suggests interesting applications
for nanofabrication processes that will be discussed below
Silicon nanowire creation on Si(100)
We performed most of the experiments described above on silicon (111) 7x7
reconstructed because this surface is easily prepared in UHV and above all because it is
isotropic and has no preferential directions for atom diffusion The 2x1 reconstruction of
silicon (100) even though it has a much simpler unit cell than Si(111) 7x7 shows a more
complicated morphology with lots of monoatomic and biatomic steps and dimer rows
parallel to each other The orientation of the dimer rows changes by 90deg on alternate
atomic layers Surface diffusion in this case is much more complicated because atoms will
experience the effects of the surface anisotropy and diffuse preferentially along or across
the dimer rows The diffusion energy of a silicon atom as example is 06 eV along a
dimer row and 085 eV across different dimer rows This while being a complication
VPalermo 75
offers interesting possibilities for the creation of ordered structures The monoatomic steps
and the dimer rows can for example act as templates for the formation of elongated
structures of composition similar to the islands described previously
On heating a Si(100) surface oxide decomposition and void growth takes place as
described for Si(111) In some cases the void shape reflects the surface symmetry
especially for very thin oxide layers and the voids often have a slightly squared shape
Nanoisland growth is also influenced by the substrate symmetry and the islands look
rectangular as irregular clusters surrounded by concentric patterns of monoatomic steps
The whole structure reminds vaguely a ldquozigguratrdquo a kind of stepped pyramid found in
Mesopotamia (fig68 inset)
Apart from nanoisland growth in some cases we observed spontaneous formation of
nanowire-like SiC structures on the Si(100) surface The wires were generated after
exposing a Si(100) surface in vacuum to traces of CO followed by sample annealing at
800degC for 15 minutes The wires exhibit lengths ranging from 10 to 100 nm and average
widths of ca 5 nm Although being randomly positioned on the surface they are perfectly
aligned along the crystal axes of the substrate (Fig68)
Several descending monoatomic steps can be observed around each nano-line and nano-
island Very often one of the line extremities coincided with an island
Nanoline
Nanoisland
Fig 68 A Si(100) surface covered by nanoislands and nanowires The monoatomic steps present on the surface are visible The two insets show a typical nanoisland and a nanowire obtained on this surface The typical ldquodimer-row ldquostripes are visible along the sides of the nanowire
76 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Couples of parallel lines starting from the same nano-island were also observed At high
resolution (fig 69) the lines showed a lamellar periodic structure with a periodicity of
around 4 nm and frequent defects Scanning Tunneling Spectroscopy measurements
showed no significant difference between the line and the surrounding silicon surface
The proposed formation mechanism for this kind of structure involves two stages in the
first one organic contaminating molecules adsorb and diffuse on the surface from the
vacuum environment stopping preferentially at surface singularities including steps and
defects
Several substances such as atomic carbon carbon monoxide ethylene and fullerene
decompose when adsorbed on silicon surfaces at temperatures above 500degC and create
localised defects on the surface inducing strain deformations and a metastable surface
reconstruction [12]
If these molecules remain attached to the steps they can diffuse rapidly along step border
until they meet a line and decompose In this way long lines of SiC defects parallel to the
surface steps can be produced In the second phase the clusters of silicon carbide can
-005
0
005
01
015
02
025
03
035
04
-3 -2 -1 0 1 2 3
V
dId
V lt
dId
Vgt
linesilicon
Current- Voltage behaviour measured with STS Fourier analysis of wire periodicity
asymp 4 Aring
High resolution image of a line
Fig 69 Top high resolution image of a nanowire showing the lamellar structure Fourier analysis (bottom left) shows that the lamellae periodicity is sim 04 nm bottom right the IV characteristics measured with the STM tip on the line and the silicon surface
VPalermo 77
25-600degC
Final linear structure
600-800degC
Contaminant Silicon
Fig 610 Schematic representation of nanowire formation mechanism
locally inhibit the spontaneous surface diffusion of silicon atoms acting as templates for
the reorganisation of surface steps (fig 610)
The final morphology shows disordered lines and islands each surrounded by a complex
pattern of silicon monatomic layers
At 800degC silicon atoms are highly mobile on the surface Due to atom diffusion the
surface reorganizes with a step-flow mechanism The presence of a fixed line of
contaminants blocks atom diffusion and step flow generating the monoatomic steps
around each wire (fig 611) SiC clusters act as lsquopinning sitesrsquo on silicon atom surface
diffusion and can be purposely used to direct the localised growth of nano-islands [13]
Although the possibility of using methanol and carbon monoxide as nano-island precursors
has been demonstrated the exact nature of the substances nucleating nano-wire growth is
still unknown
Further experiments will be needed to confirm these hypotheses as well as to identify the
contaminating species nucleating nanowire growth Carbon monoxide is a main candidate
for nanowire growth but the real nucleation process is likely to be a complex one
78 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Step flowblocked
Fig 611 Representation of step flow blocked by the presence of a nanowire 105x105nm
Bibliography
[1] Surface-Chemistry Of Silicon Waltenburg HN Yates JT Chemical Reviews 95 (5)
1589-1673 Jul-Aug 1995
[2] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[4] Growth Of Si On The Si(111) Surface Lanczycki Cj Kotlyar R Fu E Yang Yn
Williams Ed Das Sarma S Physical Review B 57 (20) 13132-13148 May 15 1998
[5] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
VPalermo 79
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990
[6] Nanoscale roughening of Si(001) by oxide desorption in ultrahigh vacuum Gray SM
Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology B 14 (2)
1043-1047 Mar-Apr 1996
[7] Step Motion On Crystal Surfaces Schwoebel Rl Journal Of Applied Physics 40 614-
618 Feb 1969
[8] Scanning Tunneling Microscopy Investigation At High Temperatures Of Islands And
Holes On Si(111)7x7 In Real Time Evidence For Diffusion-Limited Decay
Hildebrandt S Kraus A Kulla R Wilhelmi G Hanbucken M Neddermeyer H Surface
Science 486 (1-2) 24-32 Jul 1 2001
[9] Methanol Adsorption On Si(111)-(7x7) Investigated By Core-Line Photoemission And
Mass Spectrometry Of Photodesorbed Ions Carbone M Piancastelli Mn Zanoni R
Comtet G Dujardin G Hellner L Surface Science 370 (1) L179-L184 Jan 1 1997
[10] Atomic And Macroscopic Reaction Rates Of A Surface-Catalyzed Reaction
Wintterlin J Volkening S Janssens Tvw Zambelli T Ertl G Science 278 (5345)
1931-1934 Dec 12 1997
[11] Redhead PA Hobson JP Kornelsen EV The Physical Basis Of Ultrahigh
Vacuum Chapter 12 (Chapman amp Hall London 1968)
[12] The Si(001) C(4 X 4) Surface Reconstruction A Comprehensive Experimental
Study Norenberg H Briggs Gad Surface Science 430 (1-3) 154-164 Jun 21 1999
[13] Production Of Nanostructures Of Silicon On Silicon By Atomic Self-Organization
Observed By Scanning Tunneling Microscopy Jones D Palermo V Applied Physics
Letters 80 (4) 673-675 Jan 28 2002
80 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 81
7 Conclusions and proposals for future work It is very difficult to foresee the future of nanoscience and nanotechnology even over the
next few years exciting discoveries are continuously made opening up new possibilities
and progress in this sector is now one of the fastest and most varied So it is impossible to
predict whether computers will in the future be made with nanowires nanotubes single
electron devices or some kind of quantum-based transistor What does seem clear
however is that silicon devices can still further extend their levels of miniaturization to
remain competitive for at least the next ten years [1] During this period new emerging
technologies will reach scientific maturity and arrive at the production lines
Even when new technology prototypes do become available for full-scale production the
astronomical cost of changing from silicon to new technologies will be a major concern
Moreover a vast amount of extremely detailed knowledge is already available on all
aspects of silicon technology Thus the possibility of integrating innovative
nanotechnologies with standard CMOS silicon technology already seems an attractive
prospect [2]
For this and other reasons we have focussed our research efforts on surface phenomena on
silicon wafers which could be potentially useful for the development of silicon-compatible
nano-devices
Some important conclusions can be drawn from the work described in this thesis
bull Silicon surfaces of different crystal faces were studied at atomic resolution in ultra-
high-vacuum following transformations induced on the surface by chemical (etching)
and physico-chemical (organic contamination heating) treatments
82 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull The presence of surface contaminants induces at high temperature the growth of
nanometre-sized islands and lines on silicon
bull Final island densities and dimensions strongly depend upon the chemical nature of the
contaminant molecules and the treatments used Molecules interacting only weakly
with silicon will desorb upon heating whereas molecules which decompose easily on
Si will give widespread random nano-island nucleation Between these two extreme
cases complex behaviours of diffusion and subsequent decomposition are possible
bull Nanoscopic naked silicon voids were produced in oxide layers through carefully-
controlled thermal annealing It was found that void density and size depends upon
oxide purity surface contamination levels and annealing conditions
bull The growth of nano-islands and nano-voids was exploited simultaneously and the
voids used as nanoscopic masks to control the positions of growing islands Complex
silicon-on-silicon nano-structures were obtained with multi-step growth processes
bull Finally the use of macroscopic treatments and simple physical processes to produce
large quantities of nanometre-sized structures on silicon surfaces was demonstrated
These processes and the structures produced are of potential interest to the
microelectronics industry
We showed in previous chapters how it is possible to modify silicon surfaces using
chemical and physical methods The use of self-organization for the production of Si-on-Si
20 nm
Fig 71 Left SEM image of field emitter arrays made by conventional lithography [3] Center 3-D STM image of a self-organized void with a central nano-island Right Lateral schematic view of a field emitter device
VPalermo 83
nano-structures has the advantage with respect to other emerging techniques of being
silicon compatible and thus is a potential candidate for its implementation in the
production of new devices Furthermore this technique can create billions of strucures on a
wafer in only a few minutes being thus applicable to large-scale production
Even when ordered structures are obtained these methods still show a poor reproducibility
and control Chemically modified surfaces look very disordered at high magnification
with trenches and holes etched into the surface and progress has recently been made in
producing very small structures electrochemically [4] Oxide decomposition with
nanoisland growth in UHV can also yield ordered structures The void holes with
nanoislands located at their centres although quite variable in dimensions are more similar
to engineered devices than to a spontaneous random surface-roughening process In fact
there is a surprising similarity between the nanovoid-nanoisland structures and well-known
commercially available devices called field emitter arrays (FEA) These are small tips used
as electron guns in various kinds of displays and other devices Both the commercial
devices and our self-organized structures consist of a protruding tip at the centre of
microscopic holes on an insulating layer covering a semiconductor surface (see fig71)
Of course the order and size uniformity of commercial FEAs is much better than that
obtained with our process but it is encouraging to see that a self-organized structure
obtained with simple heating has a very similar surface topography and composition of
commercial working devices obtained after many expensive stages of optical lithography
deposition and etching
Nano-island density and size can to some extent be controlled and different
contaminants can be adsorbed onto the surface to modify the growth process In the future
with a better understanding of diffusion dynamics and growth mechanisms a multi-stage
fabrication process can be envisaged where purposely-designed molecules are adsorbed
onto silicon surfaces to diffuse organize in ordered arrays and favour surface
reorganization at the nanometre scale The same molecules could be used to deliver
dopants to obtain for example highly conducting nanostructures on otherwise undoped
highly resistive silicon The selective formation of nanostructures at the SiSiO2 border
can be used to scale down mesoscopic patterns to the nanometric level A silicon oxide
layer can be easily patterned on the surface with normal optical lithography and then
nanostructures can be built following the pattern edge In this way the production of very
small nanowire based devices may be possible (fig 72)
84 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
D
Gate
S
D
S
SiO2
SiO2
2 Heating nanowire creation
and oxide removal
3 Source-drain deposition by conventional lithography
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
4 Insulating layer (oxide) and gate (metal) deposition
Fig 72 Schematic drawing of the construction of a self-organized nanowire-based transistor
Surface diffusion and decomposition of molecules can be sensitive to even smaller surface
features The nanoscopic lines formed on Si(100) are very regular and neat if their
nucleation and growth mechanism could be controlled ordered networks of nanowires
could be made on silicon
VPalermo 85
The techniques developed here are still experimental ones the careful control and
reproducibility of each step is still an issue and further studies will be needed to clearly
understand the dynamics of formation of these structures Nonetheless the results obtained
so far open up the possibilities of producing novel devices such as for example nano-
transistors (see scheme in fig73 where the mechanism described in fig 72 is applied for
the fabrication of a device array) The possibility of using the border of a lithographic
pattern to draw nanolines can in principle allow the creation of nanodevices using normal
IC manufacturing techniques
The results obtained and their description in this thesis are not the first examples of the use
of self-organization for the production of ordered structures and will certainly not be the
last In our opinion the most important outcome of these findings is that it is possible to
use surface diffusion to overcome the inherent limits of lithographic techniques in
microdevice production processes and to force properly selected molecules to react with
the surface in a spatially differentiated manner at particular sites thereby creating ordered
series of nanostructures The role of the surface dynamics of the substrate atoms
themselves is extremely important in this process
Outstanding results have been and continue to be obtained in nanoscience and
nanotechnology research promising the emergence of new production technologies Sooner
or later those emerging technologies will have to face the issues of cost and compatibility
with the enormous investment made in existing technologies and processing facilities The
commercial production of nanostructures based on surface diffusion processes could in
principle be cheap simple and compatible with existing technology Moreover the
processes of diffusion in bulk silicon is well-known in the microelectronics industry which
has both the know-how and the equipment to better understand and exploit the surface
diffusion processes on silicon surfaces studied here
Extremely important and innovative results are often obtained by exploiting very simple
ideas and in our opinion physical processes such as surface diffusion used to create
complex nanostructures on silicon surfaces is a prime example
86 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
2 Heating nanowire creation and
oxide removal
3 Source-drain deposition by conventional lithography
4 Insulating layer (oxide) and gate (metal) deposition
Fig 73 Schematic macroscopic outline of the construction of a self-organized nanowire-based transistor
VPalermo 87
Bibliography
[1] P Gargini Intel Technology Coordinator ldquoFrom Microelectronics To
Nanotechnologiesrdquo Invited Campus Colloquia Lecture CNR Research Area Bologna
February 6th 2003
[2] Toward A Hybrid Micro-Nanoelectronics Cerofolini Gf Ferla G Journal Of
Nanoparticle Research 4 (3) 185-191 Jun 2002
[3] Recent Progress In Field Emitter Array Development For High Performance
Applications Temple D Materials Science amp Engineering R-Reports 24 (5) 185-239
Jan 25 1999
[4] Silicon Dioxide Micropillars For Sieving Fabricated By Macroporous Silicon-Based
Micromachining Izuo S Ohji H French Pj Tsutsumi K Kimata M Sensors And
Materials 14 (5) 239-251 2002 Electrochemical Etching In HF Solution For Silicon
Micromachining Barillaro G Nannini A Piotto M Sensors And Actuators A-Physical
102 (1-2) 195-201 Dec 1 2002
88 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 89
Acknowledgements
This work would not have been possible without the help and guidance of Dr Derek Jones
he introduced me to the use of STM and other techniques found the funding for this PhD
and helped me with the thousands of small and big problems I had to face during my
research I am grateful to Prof Alberto Ripamonti too for being the supervisor of this
thesis Thanks are also due to Dr Giancarlo Seconi director of ISOF-CNR where most of
this work was carried out Financial support from the Italian National Research Council
(CNR) is also gratefully acknowledged
My passion for surface science began when Fabio Biscarini showed me that it was possible
to really see the atoms and for this I will always thank him
During these years I had the luck to collaborate with many good scientists among those I
would like to cite Enrichetta Susi Massimo Cocchi and Anna Mazzone from CNR
Claudio Zannoni Anna Cavallini Daniela Cavalcoli and Antonio Castaldini from the
University of Bologna Sergio Pizzini and Maurizio Acciarri from the University of Milan
One of the most stimulating periods of my PhD was the one I spent at the Steacie Institute
for Molecular Sciences in Ottawa working with Robert Wolkow Dan Wayner Greg
Lopinski and Peter Kruse I thank them for discussing together many exciting scientific
ideas and trying to put some of them into practice
Special thanks go to my colleagues Paolo Samorigrave and Stefano De Cesari for long and
useful discussions on Science Life and their interactions often with the help of a good
pint
Finally I would like to thank Vassilia Gaetano Simone Claudia Silvia Letizia
Alessandro Angela and Sebastien even though not directly involved in this work they
have been over these three years a continuous source of happiness and support
Bologna March 2003
90 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
SCIENTIFIC PUBLICATIONS bull Lateral diffusion of titanium disilicide as a way to contacting
hybrid Si-organic nanostructures Palermo V Buchanan M Bezinger A Wolkow RA APPLIED PHYSICS LETTERS 2002 v81 p 3636 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v6 issue 20) bull Surface modifications in Si after Rapid Thermal Annealing Castaldini A Cavalcoli D Cavallini A Jones D Palermo V Susi E JOURNAL OF THE ELECTROCHEMICAL SOCIETY 2002 v 149 pG633 bull Production of nanostructures of silicon on silicon by atomic self-
organisation observed by scanning tunnelling microscopy Jones D Palermo V APPLIED PHYSICS LETTERS 2002 v 80 p 673 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v5 issue 5) bull Nucleation of nanostructures from surface defects on silicon Palermo V Jones D SOLID STATE PHENOMENA 2002 v 82-84 p 687 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo MATERIALS SCIENCE AND ENGINEERING B 2002 v88 (2-3) p220 bull Morphological changes of Si[100] surface after treatment with
concentrated and dilute HF Palermo V Jones D MATERIALS SCIENCE IN SEMICONDUCTOR PROCESSING 2001 v 4 p 437 bull Electrical and structural properties of processed silicon surfaces Susi E Cavallini A Castaldini A Cavalcoli D Jones D Palermo V ldquoRECENT RESEARCH DEVELOPMENTS IN VACUUM SCIENCE amp TECHNOLOGYrdquo 2001 v 3 p189 bull Numerical Solutions of the Stochastic Equations of Crystal
Growth Mazzone A M Palermo V INTERNATIONAL JOURNAL OF MODERN PHYSICS C 2000 v 11 Part 1 p195-204 bull Advances in silicon surface characterisation using light beam
injection techniques
VPalermo 91
Acciarri M Pizzini S Simone G Jones D Palermo V MATERIALS SCIENCE AND ENGINEERING B 2000 V73 (1-3) p 235 - 239 bull Abrupt orientational changes for liquid crystals adsorbed on a
graphite surface Palermo V Biscarini F Zannoni C PHYSICAL REVIEW -SERIES E- 1998 V 57 NUMBER 3A p R2519-R2522 Oral presentations bull The use of oxide desorption and surface diffusion for the
creation of silicon on silicon nanostructures Palermo V Jones D 1ST NATIONAL WORKSHOP ON CURRENT TRENDS IN NANOTECHNOLOGIES Catania (Italy) Feb 2002 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo V E-MRS Spring Meeting Strasbourg (France) June 2001 bull STM study of surface transformations on silicon during UHV
annealing Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999 bull Advances in silicon surface characterisation using light beam
injection technique Acciarri M Pizzini S Simone G Jones D Palermo V E_MRS SPRING MEETING June 1999 Posters bull Spontaneous nano-wire growth on silicon Palermo V Jones D SMARTON workshop Leuven (Belgium) October 2002 bull Ordered circles of nano-islands on silicon from CO adsorption Palermo V Jones D TRENDS IN NANOTECHNOLOGY 2002 Santiago de Compostela (Spain) September 2002 bull Production of nanostructures of silicon on silicon by atomic self-
organisation Palermo V Jones D EUROMAT Rimini (Italy) June 2001 bull Etching holes and anisotropic corrosion on silicon [100] Palermo V Jones D Susi E Asoli B SILICON WORKSHOP Genova (Italy) February 2001
92 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull Morphological and electrical characteristics of damaged silicon surfaces
Susi E Castaldini A Cavalcoli D Cavallini A Jones D Palermo V SILICON WORKSHOP Genova (Italy) February 2001 bull Effect of HF etching on the roughness of a silicon surface Palermo V Jones D NATIONAL CONFERENCE ON PHYSICS OF MATTER Genova (Italy) June 2000 bull Effect of Native Oxide Desorption upon the Surface Morphology of Si[100] by STM and LEED Palermo V Jones D SILICON WORKSHOP Genova (Italy) February 2000 bull Characterisation of silicon surfaces for microelectronics through STM measurements Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999
VPalermo 93
Contact Vincenzo Palermo ISOF- Institute for Organic Synthesis and Photoreactivity Via Gobetti 101 40129 Bologna ITALY Tel +39-051-6398336 Fax +39-051-6398349 Mail palermoisofcnrit
94 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Index
1 Introduction nanotechnology and the future of computers
2 Silicon surfaces
3 STM and other surface analysis techniques
4 Surface modification of silicon in liquid Nano-hole creation
5 Surface modification of silicon in vacuum void creation and oxide desorption
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and
nano-lines
7 Conclusions and proposals for future work
VPalermo 11
fundamental raw material of the IC industry for the foreseeable future nanotechnology
developments for microelectronics will need to be silicon-compatible In these early days
of nanotechnology the most valid approach would seem to be the addition of molecular
functions to existing silicon technology ndash using the latter as a foundation on which to build
Fabrication of self-organised structures on silicon
The possibility of using the phenomenon of atomic or molecular self-organization to create
nanostructures on silicon has already been demonstrated The clean silicon surface shows
in some cases a high degree of order and complex surface reconstruction as will be
described later Several different ordered structures form spontaneously on this surface
such as series of monatomic steps or boundaries between reconstructed areas It has been
demonstrated that it is possible to use these structures to fabricate ordered nanodots and
nanolines on the surface [11] More recently well-defined nanometric patterns have been
obtained with selective etching of silicon using nitric oxide [12]
In this study the possibility of creating different types of nanostructures on the silicon
surface is explored Methods had to be developed which were
- Simple They must not need complex masks or lithographic steps to create the
structure but rather exploit self-organisation phenomena
- High resolution the silicon surface modifications should be on a scalelength of below
100 nm
- Fast billions of nanostructures have to form over the whole surface simultaneously to
be compatible with large-scale production requirements
- Cheap they must not require expensive equipment (such as e-beam lithography) but
exploit simple chemical andor physical treatments to produce nanostructures on the
silicon surface
In Chapter 2 the main characteristics of silicon are described Chapter 3 provides a
summary of the techniques used for this research Chapter 4 examines the chemical etching
of silicon in different liquid environments and the effects of this etching on the surface at a
nanoscopic level with the creation of nanoholes
12 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Chapter 5 describes the growth in ultra-high vacuum (UHV) of nanoscopic voids and
islands on the silicon surface and the effect of surface oxide on this growth
Chapter 6 discusses the modification of silicon surfaces in UHV following the adsorption
of molecules and thermal heating to produce nanoislands and nanolines on silicon
The overall conclusions of our work are summarised in Chapter 7
Finally we will give some conclusions based on the results obtained and discuss possible
applications of the methods developed
Bibliography
[1] From wwwintelcom
[2] From wwwsematechorg
[3] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
[4] Molecules Get Wired Service RF Science 294 (5551) 2442-2443 Dec 21 2001
[5] Coulomb Blockade And The Kondo Effect In Single-Atom Transistors Park J
Pasupathy AN Goldsmith JI Chang C Yaish Y Petta JR Rinkoski M Sethna JP
Abruna HD Mceuen PL Ralph DC Nature 417 (6890) 722-725 Jun 13 2002 Kondo
Resonance In A Single-Molecule Transistor Liang WJ Shores MP Bockrath M Long
JR Park H Nature 417 (6890) 725-729 Jun 13 2002 Nanotechnology - Electronics
And The Single Atom De Franceschi S Kouwenhoven L Nature 417 (6890) 701-702
Jun 13 2002
[6] Smallest Carbon Nanotube Ajayan PM Ijima S Nature 358 (6381) 23-23 Jul 2 1992
[7] Carbon Nanotube Inter- And Intramolecular Logic Gates Derycke V Martel R
Appenzeller J Nano Letters 1 (9) 453-456 Sep 2001
[8] Logic Circuits With Carbon Nanotube Transistors Bachtold A Hadley P Nakanishi T
Dekker C Science 294 (5545) 1317-1320 Nov 2001
VPalermo 13
[9] Logic Gates And Computation From Assembled Nanowire Building Blocks Huang Y
Duan XF Cui Y Lauhon LJ Kim Kh Lieber CM Science 294 (5545) 1313-1317
Nov 9 2001
[10] Formation And Disappearance Of A Nanoscale Silver Cluster Realized By Solid
Electrochemical Reaction Terabe K Nakayama T Hasegawa T Aono M Journal Of
Applied Physics 91 (12) 10110-10114 Jun 15 2002
[11] Fabrication And Integration Of Nanostructures On Si Surfaces Ogino T Hibino H
Homma Y Kobayashi Y Prabhakaran K Sumitomo K Omi H Accounts Of Chemical
Research 32 (5) 447-454 May 1999
[12] Ultrafine And Well-Defined Patterns On Silicon Through Reaction Selectivity
Prabhakaran K Hibino H Ogino T Advanced Materials 14 (19) 1418-1421 Oct 2
2002
14 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 15
2 Silicon surfaces
The name silicon (silicio in Italian) comes from the latin word silex Amorphous silicon
was first isolated by Berzelius in 1824 by reaction of potassium with silicon tetrafluoride
Thirty years later the first crystalline silicon was prepared Silicon makes up 25 of
earthrsquos crust and is the second most abundant element after oxygen Elemental silicon is
not found in nature occurring as silicon oxide (sand quartz amethyst flint etc) or
silicates (asbestos clay mica etc) Perhaps no other element and its compounds has such
a wide range of uses Silicon compounds such as sand and clay are used in the building
industry as refractory materials for high-temperature applications and for enamels and
pottery Silica is the main component of glass silicon carbide is an important abrasive and
silicones are commonly used polymers and lubricants
Here the most interesting use of silicon of course is for the production of
microelectronics devices For this application silicon of high purity (999999) and of
high crystallinity is needed Table 21 lists some of the physical characteristics of silicon
High purity polycrystalline silicon is produced by the reaction of gaseous trichlorosilane
with hydrogen in a furnace Then to prepare a single-crystal of silicon the so-called
Czochralski method is commonly used
Polycrystalline silicon is melted in a quartz furnace at 1415degC in an argon atmosphere
Then a seed of single-crystal silicon is lowered into contact with the melt and slowly
pulled out In this way the crystal grows and a crystalline cylindrical ingot several metres
long is created from the initial seed
After cooling down the ingot is sliced into thin silicon wafers The wafer surfaces are
polished using a counter-rotating lapping machine in an Al2O3 slurry until the surface is
very flat and shiny ready for the lithographic processes
16 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another way to obtain single crystal silicon is the Floating Zone (FZ) method in which a
silicon cylinder is slowly passed through a heating ring The area inside the ring melts and
solidifies smoothly crystallising as it comes out of the ring yielding a single silicon crystal
Microelectronic devices are built on the silicon surface which is the surface of interest
here Unfortunately silicon surfaces are normally quite dirty and uneven at the atomic
scale Atmospheric oxygen and humidity react with silicon surfaces creating a thin layer of
oxide (called ldquonative oxiderdquo) which is usually irregular and full of defects Different kinds
of contaminants also adsorb onto the surface These are usually small organic molecules
and microscopic dust particles A clean surface on exposure to the atmosphere is
completely covered with gas molecules in less than 10-9 seconds If the pressure is
reduced letrsquos say to 10-6 mbar this time increases to 1 second This is the reason why to
study a clean surface we have to work in UHV at pressures below 10-10 mbar
The atoms in the silicon crystal have a diamond-like structure each atom having 4 bonds
in a tetrahedral sp3 arrangement with bond angles of 10947 degrees At the crystal
surface some atoms will have non-bonding orbitals ldquodanglingrdquo in the vacuum ie sp3
orbitals with a lone electron which are highly reactive These orbitals are known as
dangling bonds To minimize surface energy the surface will reorganize by decreasing
the number of dangling bonds
Table 21 Physical data of silicon [1]
Atomic Weight 2809 Lattice constant (A) 543095 Crystal structure Face-centered cubic
(diamond) Melting point 1415 degC
Density (gcm3) 2328 Boiling point 2355degC Atomscm3 50E22 Minority carrier
lifetime (s) 25E-3
Dielectric Constant 119 Specific heat (Jg degC)
07
Breakdown field (Vcm)
~3E5 Thermal conductivity (Wcm degC)
15
Electron affinity x(V)
405 Vapour pressure (Pa) 1 at 1650degC 1E-6 at 900deg C
Energy gap (eV) at 300K
112 Reactivity Inert to acids Attacked by halogens and alkaline
solutions Intrinsic carrier
conc (cm-3) 145E10 Oxidation states +4 -4
Intrinsic Debye Length (microm)
24 Energy of a Si-Si bond (eV)
232
Intrinsic resistivity (Ω-cm)
23E5
VPalermo 17
Dangling bond densities and positions and thus the type of surface reconstruction will
depend upon crystal orientation as well as the temperature and kinetics of the system
Fig 21 shows a drawing of the main faces of a silicon crystal The angle α between any
(11n) face and the (100) face can be calculated from 2cos 2 += nnα The angle
between any (11n) face and the (111) face can be calculated from
)2(3)2(cos 2 ++= nnα
The chemistry and physics of the faces are very different a brief description will be given
for the most important orientations
Si (100)
On the (100) surface each atom has two Si-Si bonds connecting it to the bulk and two
dangling bonds pointing outward Surface energy is reduced by the dimerisation of the
surface atoms through overlap interaction of one dangling bond per atom forming rows of
dimers aligned along the (110) direction This is the well-known ldquo2x1rdquo reconstruction of
this silicon surface The symmetric dimers would make the silicon surface metallic but to
reduce surface stress the dimers tend to buckle and the surface is thus semiconductive It
took several years to understand that the dimers are buckled because at room temperature
Fig 21 Schematic view of the principal orientations of a silicon surface Surface atoms are white bulk atoms are black dangling bonds are gray [2]
18 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
they shift easily from one buckling direction to the other thus appearing symmetric under
STM observation Fig 22 shows an STM image of the 2x1 reconstructed surface
Even almost perfect (100) surfaces have a certain number of monoatomic steps and the
dimer rows on atomic layers are aligned at 90deg to those on adjacent layers Dimer rows are
thus perpendicular or parallel to the step When the dimers on the upper side of the step
are parallel to the step the step is called SA if they are perpendicular the step is called SB
Because of this symmetry SA step edges will be smoother compared to the more broken
fragmented edges of the SB steps
A common defect on the Si(100) surface is the presence of nickel contamination which
appear as missing dimers in STM images This type of contamination is so critical that
even if the silicon sample is only briefly brought into contact with stainless steel tools
(tweezers for example) the 2x1 reconstruction of the surface can be blocked
Silicon atoms can diffuse easily over the silicon surface as monomers and dimers
especially at elevated temperatures The anisotropy due to the 2x1 reconstruction causes a
difference in the diffusion energies of adsorbates over the surface Diffusion of these
silicon species along dimer rows for example will be much easier A list of diffusion
energies for monomers and dimers is given below [2]
Diffusion on Si(100) 2x1 Ed (eV) Monomers along dimer rows 06 Monomers across dimer rows 085 Monomer formation energy 18 Dimer along dimer rows 11 Dimer across dimer rows 15 Dimer formation energy 26 Dimer binding energy 076 Vacancies along dimer rows 17 Vacancies across dimer rows 19
SA
SB
Fig 22 STM image of a 2x1reconstructed silicon surfaceshowing the dimer rows and steps40x35 nm Nickel-induced defectsare visible as dark spots SA and SBsteps are indicated
VPalermo 19
So the diffusion energy for both monomers and dimers is nearly 40 greater if they have
to cross a dimer row This difference reduces to sim10 for vacancy diffusion
Si(113)
The (113) surface can be imaged as a sequence of alternating (100) and (111)-like
structures with two and one dangling bonds on alternate atoms respectively Interest in the
(113) surface is scientific as it has been used to study the energetics of the (100) and (111)
surfaces as well as for surface adsorption experiments
Si(100) surfaces can easily develop (113) facets
Si(111)
This surface besides being the first one imaged with STM with atomic resolution is one of
the most studied because it is the best cleavage face of silicon and because it shows one
of the most complex and elegant reconstructions in surface science
All Si-Si bonds in the silicon crystal are perpendicular to a (111) plane so this face will
have the lowest number of dangling bonds created per unit area In fact each Si atom on a
(111) surface shows a single dangling bond oriented perpendicular to the surface and
bonded to three back atoms These three bonds for each surface atom account for the great
chemical and physical stability of the Si(111) surface Surface energy is 009 eV Aring-2
compared to 015 eV Aring-2 for Si(100)
For energy minimization this surface reconstructs forming a huge 7x7 lattice cell
containing 102 atoms described by the Dimer-Adatom-Stacking fault model (DAS) For a
detailed description of cell structure see fig 23
The cell described by this model is very complex being composed of three kinds of atoms
adatoms rest atoms and corner hole atoms Furthermore a subsurface stacking fault is
present in one half of the cell making the two halves of the unit cell look different under
STM (Fig 24) It took 26 years of research to completely understand the exact structure of
the 7x7 reconstruction
Cleaving a silicon crystal along a (111) plane produces a metastable 2x1 reconstruction
the 7x7 reconstruction is easily obtained by flashing at high temperature in UHV At T gt
20 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
830degC a disordered 1x1 phase covers the surface Cooling down to 800degC leads to the
formation of the 7x7 phase If the cooling process is too rapid small 7x7 domains nucleate
and a disordered 1x1 phase is preserved between domain boundaries
Si(110)
Even though as mentioned before the (111) plane is the favoured cleaving plane of
silicon thin (100) commercial wafers will not break along this plane because the angle
between (100) and (111) is too far from 90deg (see table 22) Instead they will break along
the (110) plane because it is perpendicular to the (100) surface Each surface atom on
Si(110) has a Si-Si bond pointing downward one dangling bond pointing outward and two
Si-Si bonds parallel to the surface in a zig-zag pattern (see fig 21) Cleaved (110)
surfaces are disordered but upon annealing at high temperatures an ordered complex 16x2
Fig23 Scheme of the 7x7 DAS model [2] In each unit cell there are 9 dimers 12 adatoms and a stacking layer fault The force driving this complex reconstruction is the minimization of dangling bonddensity The DAS model shows the lowest number of dangling bonds (19) of all possiblereconstructions 12 dangling bonds are at the adatoms 6 at the rest atoms and 1 at the corner hole atom This surface is metallic
VPalermo 21
reconstruction takes place The surface appears as a series of long ridges and valleys
parallel to each other Eventually tilted facets of orientation (17 15 1) can form on this
surface The adsorption of Ge atoms on this surface leads to the formation of self-
assembled nanowires [3]
Table 22 Angles in degrees between different silicon faces [2]
Orientation
113
110
111
100
100
2524
9000
5474
0
111
2950
3526
0
110
6476
0
113
0
Fig 24 STM image of a Si(111)surface with 7x7 reconstruction A unitcell with its adatoms is highlightedImage size 13x13 nm
22 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] Weast RC Handbook Of Chemistry And Physics (Chemical Rubber Co Cleveland
1972)
[2] Dabrowski J Mussig H Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] The Structure Of Clean And SiGe-Covered Si(110) Surfaces Butz R Luth H Surface
Science 365 (3) 807-816 Oct 1 1996
VPalermo 23
3 STM and other surface analysis
techniques
Scanning Tunneling Microscopy
Since the invention of the optical microscope at the end of the 16th century the possibility
of examining surfaces at higher and higher magnification has fascinated mankind
Development of the technique continued and towards the end of the 19th century optical
microscopes were as good as todays standard instruments The physical limits of the
wavelength of visible light (350-800nm) had been reached
In the 1920s de Broglie showed that electrons can behave like waves and the use of these
particles for imaging with much higher resolution soon followed Atomic resolution using
this technique is only possible in the transmission mode with extremely carefully prepared
samples
In 1982 using the peculiar properties of piezoelectric materials Binnig and Rohrer brought
a metallic tip very very close to a silicon surface and scanned it across an extremly small
area (fig31) The tunneling of electrons from the tip into the sample or vice versa allowed
them to obtain a local density of electronic states (LDOS) map of that surface Although
theory (which treated the extreme point of the tip as a sphere) then excluded the possibility
of atomic resolution following a tip crash into the surface Binnig and Rohrer began to
observe the LDOS with atomic resolution For this discovery and their following work
they were awarded the Nobel Prize in Physics in 1986
24 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The basic principle of STM is very simple A metallic tip is scanned over a surface without
making ohmic contact and a tunneling current passes between the tip and the surface An
electronic circuit keeps this current constant by raising and lowering the tip during the
scan In this way recording the tip height at each point a three-dimensional image of local
density of electronic states (LDOS) of the surface can be obtained To explain the
extremely high resolution attainable by this simple technique quantum theory is needed
According to classical physics the current will flow between sample and tip only if they
are in physical contact If there is a vacuum gap between the two the electrons will simply
remain confined for example within the surface without the possibility of passing into the
tip
In quantum physics however the electrons have a certain probability of passing
(tunneling) across the gap appearing on the other side of the gap in this way reaching the
tip It can be shown that the probability of an electron tunneling through a gap of thickness
z is
kzep 22)0( minusprop ψ h
φmk
2= (1)
where ψ(0) is the electron wavefunction at the surface-gap border m is the electron
mass=91x10-28 g and φ is the work function of the metal (ie the energy required to
remove an electron from that material For silicon it is 48 eV) The tunneling current thus
decays exponentially with z and is extremely sensitive to topographical imperfections
present on the scanned surface A rough formula giving the current as a function of z is [1]
zFS eEVI φρ 0251)( minusprop
Fig 31 Binnig and Rohrer with the first STM Image from IBM [2]
VPalermo 25
where ρs(EF) is the local density of states at the Fermi level on the given surface For
example the formula predicts that for silicon an increase in tip-surface distance of 1 Aring
will give a 95 decrease in tunneling current
This huge dependence of tunneling current upon the distance allows detection even of the
sub-nanometre changes in height given by the single atoms of which the surface is
composed and thus to resolve them in the LDOS images Of course this description of the
tunneling process is oversimplified and for a more accurate one the electronic states of
the tip of the sample and their interaction have to be taken into account Fig 32 shows a
schematic representation of the interaction between tip and sample orbitals
The exponential decay of current with distance also yields high lateral resolution If the tip
is approximated as a sphere of radius R and the current passing at the minimum tip-sample
distance is I0 then the current passing at a lateral distance x from this point will be
Rxk
eII 22
0
2minus
=
Assuming a tip radius of 100 nm the current is concentrated in an area sim15 nm wide at tip
apex
Very sharp tips with even smaller curvature radii can be produced in several ways Simple
Fig32 Schematic view of tip-sampleorbitals interaction a) no interaction b)equilibrium c) sample positive d) tippositive [1]
26 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
tungsten wires mechanically cut are capable of obtaining atomic resolution on graphite in
air but for more disordered and rougher samples sharper and more reproducible tips are
needed
STM tips are mostly made by electrochemically etching a W or Pt-Ir wire The tips we
used were prepared using methods based on the work of Fotino [3]
A tungsten wire 038 nm diameter is immersed in a KOH solution(06M) with a thicker
tungsten wire used as a counterelectrode The cathodic and anodic reactions involved in the
etching are
Cathode 6H2O + 6 e- rarr 3H2(g) + 6 OH-
Anode W(s) + 8 OH- rarr WO42- + 4H2O + 6 e-
A potential of 3V ac is applied to the tungsten and the wire is immersed in the solution
until a constant current of sim100 mA is established The etching rapidly removes metal
shaping the wire end as a sharp tip When the potential reaches 11 V the coarse tip etching
is finished The wire is then removed from the solution carefully inserted into an
insulating plastic tube leaving only the tip exposed and re-immersed in the solution with
the tip pointing upwards A more gentle etching is thus made to reduce tip radius Usually
5 minutes etching at 07 V ac is used In this configuration very small hydrogen bubbles
formed on the tip sides sliding upwards with a ldquohoningrdquo effect on the tip
This procedure yields extremely sharp and reproducible tips at the microscopic level
After the etching the tip is thoroughly rinsed in ultrapure water then dipped into
concentrated HF to remove surface oxides and hydroxides [4] The tip is dried with
nitrogen inserted into the UHV system and degassed overnight at sim150degC
The possibility of measuring sub-nanometric distances would be useless without being able
to control tip movement over such a minute scale To scan the tip over the surface a
piezoelectric scanner is used Piezos are usually made of an alloy of PbZrO3 and PbTiO3 a
material which contracts or expands when a voltage is applied to it The Omicron
instrument used in our laboratory has three such piezo scanners for xy and z tip motion
allowing one to scan the tip over the surface with sub-Aringngstrom precision (fig 33)
To isolate the instrument from ambient vibrations the whole STM stage is suspended upon
four springs which eliminate all frequencies above 1 Hz and surrounded by a crown of
VPalermo 27
copper wings and fixed magnets Parasitic currents generated by the magnets into the
copper wings contrast every movement of the stage and efficiently block stage vibrations
The STM can be used not only to explore surface topography but to measure the IV
characteristics of single atoms or molecules on the surface (Scanning Tunneling
Spectroscopy or STS) Furthermore it can be used to modify the surface with voltage
bursts digging into it or delicately moving single molecules or atoms over a surface [5] It
can be used in vacuum in air and with proper lateral isolation of the tip even in liquids
A major drawback of STM is that it works only on conducting and semiconducting
surfaces and thus cannot be used on many surfaces of biological and chemical interest
Another instrument more suitable for these and other applications is the Atomic Force
Microscope (AFM)
Atomic Force Microscopy
AFM was invented in 1986 by Binnig Quate and Gerber after calculating the possibility of
building a cantilever with a force constant of the same order of magnitude as that of a
chemical bond
In AFM a tip mounted on a microscopic cantilever (usually made of Si3N4 fabricated with
optical lithography) is brought close to a surface When the tip touches the surface the
cantilever is very slightly deflected upwards The movement is measured by observing the
Y-PIEZO
SILICON SAMPLE
Fig 33 A picture of theSTM used for theexperiments The tripodpiezo scanner is shown
Z-PIEZO
X-PIEZO
TIP
28 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
deviation of a laser beam hitting the upper face of the cantilever Fig 34 provides a
schematic view of the principle of AFM
The typical force constant of the cantilever varies from 00006 to 2 Nm the typical
resonance frequency is 3 to 120 kHz The AFM tip can apply a force on the sample of up
to 10-9 N The AFM can be used on conductive or insulating surfaces in vacuum air or
liquids Furthermore the tip can be modified to sense electrostatic potentials (electric force
microscopy) or magnetic fields (magnetic force microscopy) it can even be functionalized
with complex molecules such as proteins to interact with biological surfaces
A drawback of the AFM is that the force it exerts can damage the surface under
observation especially if the sample is soft (as in the case of cell membranes for
example) This problem can be overcome using the instrument in tapping mode (where the
tip does not move laterally during its brief contact with the surface) or in non-contact mode
in which the tip oscillates above the surface during the scan and the changes in its
frequency due to interaction with surface are monitored The shifts in the oscillating
frequency of the cantilever due to tip-sample interaction are then used for imaging the
surface In this mode interaction of the tip with the surface is minimal and soft samples
can be imaged
STM and AFM are the main techniques used for this work A brief description of other
techniques used occasionally is given below
Fig 34 Scheme of an atomic force microscope
VPalermo 29
Low Energy Electron Diffraction (LEED)
Electrons with energies in the 20-500 eV range are diffracted by a crystalline surface the
diffraction peaks are visualized on a fluorescent screen This technique probes the long
range order of the surface up to a depth of several nanometres
X-Ray Photoelectron Spectroscopy (XPS)
XPS allows both qualitative and quantitative chemical analysis of the elements present on
or near the sample surface
An X-ray source is used to photoionize the atoms on a surface and produce photoelectrons
By measuring the kinetic energy of the photoelectrons the binding energy of the electronic
levels can be calculated This energy will depend on the chemical environment of the
surface atoms
Although the soft X-rays used penetrate to a depth of ~2000Aring the sampling depth of the
technique is determined by the mean free path of the photoelectrons which allows their
escape from only the first 10-100Aring
Secondary Ion Mass Spectroscopy (SIMS)
High and low energy ions (primary ions) are used to bombard a sample and remove surface
atoms and ions The ionic fragments removed (secondary ions) are then analysed by a mass
spectrometer The surface can be consumed during the measurement and profiles obtained
giving concentrations of the materials composing the sample at different depths (depth
profiles)
A popular variant of SIMS is TOF-SIMS In this technique the secondary ion masses are
measured by a time-of-flight (TOF) measurement The secondary ions generated by the
bombarding primary ions are accelerated to a constant kinetic energy and then move
through a field-free space before they reach the detector where their intensity is measured
as a function of flight time Since ions with different masses have different velocities at a
30 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
given kinetic energy the measured flight times of the ions can easily be converted to their
masses The static nature of this latter technique allows mass spectroscopy surface analysis
with minimal damage to the surface
Bibliography
[1] Chen CJ Introduction To Scanning Tunneling Microscopy (Oxford University Press
Oxford 1993)
[2] From wwwibmcom
[3] Tip Sharpening By Normal And Reverse Electrochemical Etching Fotino M Review
Of Scientific Instruments 64 (1) 159-167 Jan 1993
[4] A Convenient Method For Removing Surface Oxides From Tungsten STM Tips
Hockett LA Creager SE Review Of Scientific Instruments 64 (1) 263-264 Jan 1993
[5] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
VPalermo 31
4 Surface modification of silicon in liquid
Nano-hole creation
Liquid treatments of silicon wafers are very common in the integrated circuit (IC)
manufacturing industry They are used to clean and improve surface uniformity to create
and etch protective oxide layers and to remove photo-resist layers
Crystalline silicon with its native oxide layer is very stable and is resistant to many acids
It is easily attacked by hydrofluoric acid (HF) and alkaline solutions
The thin (~2 nm) passivating layer of native oxide (SiO2) is formed on exposure to the
atmosphere This surface layer contains many defects and contaminants so it is usually
chemically stripped and substituted with a better chemically-formed protective oxide
The most common silicon cleaning procedure is the RCA method named after the Radio
Corporation of America [1] It consists of two steps in the first one the surface is treated
with a hot alkaline solution (H2OH2O2NH4OH 411) to remove particles from the
surface following this a hot acidic solution (H2OH2O2HCl 411) is used to remove
metal contamination Other well-known cleaning methods are IMEC (a sequence of
cleaning steps in H2OO3 and dilute HF) or the Pirana etch (a hot 41 mixture of H2SO4
H2O2)
32 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The standard RCA clean removes surface contaminants etches the native oxide and
oxidizes the silicon surface leaving a uniform layer of silicon oxide which better protects
the surface from further contamination
Etching with fluorine-based solutions
Hydrofluoric acid is one of the most common reagents used in the treatment of silicon
wafers both in the research field and in industrial processes A rapid dip in dilute HF is the
simplest way to remove the native oxide from Si(100) and leaves the surface passivated by
a layer of Si-H bonds Because of the low polarization of Si-H bonds the Si-H layer is
stable even for several days protecting the surface from contamination It has often been
assumed that this short etch does not significantly change the surface morphology of the
silicon substrate[2] even though a prolonged dip in dilute HF leads to surface roughening
[3]
Although dilute HF roughens the Si(100) surface at the atomic scale [45] immersion in
concentrated HF (49) etches the surface oxide without attacking the Si surface
uncovering in this way the buried SiSiO2 interface The final counter-intuitive result is
that dilute HF etches the silicon while concentrated HF leaves the crystalline silicon
untouched [3]
Etching Si with fluorine-containing solutions at different concentrations and pH can
produce different morphologies from rough surfaces to flat nearly ideal Si-H terminated
surfaces
Hessel et al and Higashi et al demonstrated in 1991 that very flat Si(111) surfaces can be
obtained using 40 NH4F while etching with HF always yields rough surfaces The
surface becomes smoother because the etchant rapidly attacks Si atoms at step borders
thus removing surface kinks and irregularities in a step-flow mechanism [6 7] Later on
even smoother and more perfect surfaces were obtained by removing oxygen from the
solution after it was discovered that oxygen dissolved in 40 NH4F initiates the formation
of triangular etch pits It was not possible to obtain flat surfaces by etching Si(100) with
ammonium fluoride solutions which leads to the formation of small 2x1 dimer-row
reconstructed (100) terraces together with (111) facets [8]
This difference is caused by the different hydride terminations prevailing on the (100) and
(111) faces While the ideal Si(111)-H surface is monohydride terminated the more
VPalermo 33
reactive dihydrides predominate on the Si(100)-H surface making it more vulnerable to
etching The etching reaction is thus strongly anisotropic etching (100) facets faster than
(111) thus producing (111) microfaceting on Si(100) crystals
A more uniform Si(100) surface can however be prepared by etching at low pH with an
HFHCl mixture [9] or by using very dilute HF solutions and ultrapure water with low
dissolved oxygen and carbon contents [10]
Electrochemical etching can also be used applying anodic or cathodic bias to the silicon
to obtain different morphologies [11] by varying the potential isotropic or anisotropic
etching is observed The aforementioned results show that despite the simplicity of the
reactants fluoride etching of silicon is quite a complex reaction
Fig 41 Chemical etching of silicon
HO+H
H HO
H
H
H
FSi
Si
SiSi
+H2O
-OH-
+F-
H2O H H
Si H
F OH
H Si
Si Si
-H2
+H2O
H
H
H
FSi
Si
SiSi
H
H
H
OHSi
Si
SiSi
H
H
H
H Si
Si
Si Si
Etching mechanism of silicon
HF rapidly dissolves the SiO2 passivating layer on silicon leaving the surface almost
completely hydrogenated [12] After this two different types of reactions etch the silicon
simultaneously one chemical and the other electrochemical [13] The overall etching
mechanism can be schematized in two stages (see Figure 41)
i) Si-H bonds are substituted by Si-F or Si-OH bonds creating a partial charge on the
surface silicon atom and polarizing its Si-Si backbonds
ii) these polarised backbonds are then more easily attacked by HF or H2O After
rupture of the Si-Si bond the atom is removed leaving behind new Si-H
terminations and the reaction can start again
34 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
These reactions take place although at different rates on both Si(100) and Si(111)
Stage i) is usually the rate-determining step of the reaction and the stability of the Si-H
bonds depends upon the pH the concentration of nucleophilic species in solution and an
eventual potential applied to the crystal
For pH gt5 as in the case of concentrated NH4F solutions reaction begins with attack by
water to give Si-OH (step ArarrC) The -OH group is rapidly substituted by fluorine with
polarization of the underlying Si-Si bonds These bonds are then easily attacked by water
the silicon atom being released into solution as HSiF(OH)2 The Si-OH rarr Si-F substitution
is not fundamental for the reaction and etching can proceed even for Si-OH terminated
atoms but XPS measurements showed the presence of a certain number of Si-F bonds
remaining Furthermore fluorine seems to have a catalytic effect on Si-H substitution as
indicated by the dependence of the etch rate upon the F- concentration at least for pH
values between 4 and 8
Si-F bonds can be easily removed by a water rinse In the case of strongly alkaline
solutions (pH=14) OH- groups act directly as nucleophiles and no fluorine is needed to
catalyze Si-H bond rupture
At pH lt4 almost no free F- ions are present in solution and the etch rate is very small at
pH lt2 all etching reactions are very slow and this explains the stability of Si crystals in
concentrated (50 ww) HF solutions
Si Si
Si Si
H
H
H
H Si
Si
Si Si
H
H
H Si
Si
SiSi
H
H
H
OH-H+ -e
(F-)
-H+
-e
+H2O hellipas in fig1
Fig 42 Electrochemical etching of silicon by H2O
The electrochemical etching of silicon involves electron transfer from the surface atoms to
the valence or conduction band of the crystal (according to the type of doping of the
silicon) Several studies have been carried out at different electric potentials In the case of
p-silicon or n-silicon at anodic potentials where positive charges (h+) are available in the
crystal a silicon-centred radical can be created by capture of a hole from the bulk (fig
42) The F- ion does not participate in the reaction but is thought to contribute through
VPalermo 35
electrostatic interactions by lowering the energy of the interaction step After the
formation of the Si-OH group the reaction proceeds as shown in the scheme of fig 41
Matsumura et al [4] proposed that not only water but HF2- molecules also play a major role
in electrochemical etching of silicon leaving on the surface Si-F terminated bonds which
can be immediately attacked in an autocatalytic process (fig 43)
In the electrochemical reactions described above an external potential is applied to the
silicon crystal The chemical and electrochemical reactions in any case take place
simultaneously most of the time with the chemical path predominating at high pH Even
when no external potential is applied to the silicon partial electrochemical reactions can
take place at different ldquocathodicrdquo and ldquoanodicrdquo sites on the surface with an internal charge
exchange which ensures neutrality [11] This macroscopic silicon etching and hydrogen
bubble formation can lead to visually observable patterns on the surface when Si(100) is
immersed in ammonium fluoride even without applying a potential
Si Si
Si Si
H
H
H
F Si
Si
Si SiH
F F-H+ -2e
+HF2-
H
HF
SiSi
SiSi + F
F F
FH
H
H
F -H+ -2e
+HF2-
Fig 43 Autocatalytic electrochemical etching of silicon by HF2-
Inhomogeneities on silicon surfaces caused by electrochemical reactions and charge
transfer have been studied extensively because they are of fundamental importance in the
formation of porous silicon
Pore formation on silicon
When Si(100) or Si(111) are etched under anodic bias in fluorine-based solutions
microscopic pores form on their surface Several different morphologies of pores have
been observed with pore diameters ranging from 10 nm to several microns with depths of
several microns [14] Pore shape is very variable too ranging from ordered straight pores
to chaotic networks of branched pores (fig 44) Porous silicon has been known since the
fifties but it was only in 1990 that interest in this material increased following the
discovery that porous silicon layers were able to emit bright red light This led to a large
amount of research and now different classes of micropores can be reproducibly created
36 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Fig 44 Different types of Silicon micropores From ref [14]
mostly for optical and micromachining applications However there is still no unified
theory able to explain the nucleation and growth mechanism of all the different kinds of
pores
We will give a short description of some of these theories for more detail see Parkhutik et
al [15]
One model explains pore nucleation on the basis of physical processes such as hole
positive charge migration ion transport to the surface and small perturbations on the
silicon surface modelled as Fourier components The system is shown to be unstable and
some spatial frequencies that lead to pore nucleation evolve from the etching process
A second model focusses on stationary pore growth without explaining the nucleation
stage According to this model silicon dissolves preferentially at pore edges because h+
charges are attracted by the stronger electric field present at these edges
A third class of models explains pore growth as a Diffusion Limited Aggregation (DLA)
process where the random walk of h+ charge carriers through the depleted layer present at
the silicon-liquid interface controls pore shape
Finally the model by Carstensen Cristophersen and Foll [16] proposes that areas of the
surface of some characteristic size LCO are etched by synchronized ldquocurrent burstsrdquo in the
flow of h+ charges These bursts dissolve silicon through cyclic stages of surface oxidation
oxide removal and hydrogen passivation Areas where a burst has recently taken place are
less passivated and thus more likely to be etched again in this way the pore bottom
continues to dissolve while the pore walls are passivated and are thus less favourable areas
towards current bursts
VPalermo 37
EXPERIMENTAL RESULTS
In the following sections we will show some experimental results obtained from STM and
AFM measurements of fluorine-treated Si(100) surfaces In the first part the results of
mild etching using concentrated and dilute HF solutions at low pH are presented In the
second part the results of etching at high pH using ammonium fluoride are presented and
the mechanism of pore formation discussed
Etching of Si(100) in dilute and concentrated HF
Samples were cut from different areas of an 8-inch diameter p-doped silicon(100) wafer
(10 Ω-cm) supplied by MEMC Electronic Materials Each series of STM measurements
was carried out over at least six different areas on at least two identical samples Low
Electron Energy Diffraction (LEED) was used to check the surface cleanliness of the
samples before STM measurements
Table 41 summarizes the different treatments of each sample After etching with
electronic grade HF each sample underwent a final rinse in Ultra-Pure Water (UPW
resistivity gt18 MΩ-cm) Both the HF and the UPW were allowed to flow continuously
over the sample surface Some samples were not etched with HF at all but just washed with
UPW to observe the morphology of the native oxide layer (~2 nm thick) covering the
surface All of these processes were carried out under nitrogen to limit reoxidation and the
samples were then introduced from the nitrogen atmosphere directly into the vacuum
chamber and degassed overnight at sim150degC before LEED and STM measurements
STM images were obtained from each sample using the same measurement parameters
(sample bias 4 V feedback current 1 nA scan speed 800 nm s-1) These parameters
38 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
provided a satisfactory level of reproducibility for all the samples Measurements were
made over an area of 500x500 nm (image size 500x500 pixel) Slope correction was
carried out by subtracting row-wise and column-wise fitted slopes from the entire image
which gave better results than the simple subtraction of a fitted plane especially for the
rougher samples Following slope correction the rms roughness
sum minus=xy
hyxhN
22 ))((1σ
and the 2-D Fourier transform
)(22
2
)(4
)( vyuxi
xyeyxhvuF +sum∆
= π
π
were calculated for each image where N2 is the number of pixels composing the image
h(xy) is the surface height at each point ∆ is the distance between points h is the mean
height and u v are the spatial frequencies The radial power spectrum PS(f) of the STM image data is obtained from the angular
average of the squared Fourier transform with f 2 = u2 + v2
Fig 45 shows the STM images obtained from the various samples Sample A still covered
with its native oxide layer shows an irregular surface with RMS roughness of ~05 nm
(see Table 41) Observing the sample with LEED gave no diffraction pattern even at
relatively high incident electron energies because of the surface oxide coverage After 1
min etching in dilute HF (sample B) the morphology is similar to the original one though
Table 41 Sample treatments average RMS roughness and integrated area of power spectra
Sample Treatment RMS roughness (nm)
PS area (f lt 01 nm-1)
PS area (f gt 01 nm-1)
A Rapid dip in water 053 plusmn 013 366 032
B 1 min in HF 5 + 10 min in water 051 plusmn 008 272 034
C 30 min in HF 5 + 10 min in water 062 plusmn 008 1064 038
D 5 sec in HF 49 + 10 min in water 042 plusmn 004 183 022
VPalermo 39
Fig 45 STM images of each group of samples showing the topography of the silicon surface A) noetching original oxide surface B) after 1 min etching in dilute HF C) after 30 min etching in diluteHF D) after dipping in concentrated HF Grey scale indicates height of the surface from lower (black) to higher (white) The images are 250x250 nm ie representative portions of the images usedfor the roughness measurement and PSD analysis
some of the larger features have disappeared and the image quality is better maybe due to
improved tunnelling due to the cleaner surface The RMS roughness is comparable to that
of the original surface Clear diffraction patterns are visible using LEED though at quite
high energies (200 eV) After prolonged etching (sample C) the RMS roughness increases
to 062 nm and a long-range corrugation is visible on the surface even if the LEED pattern
is good
The samples dipped in concentrated HF (D) reveal the bare SiSiO2 interface which has a
disordered aspect and protrusions over a wide range of dimensions The quality of the
STM images of sample D is very good probably due to the cleanliness of the surface
40 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 E -0 2
1 E -0 1
1 E + 0 0
1 E + 0 1
1 E + 0 2
1 E + 0 3
0 0 0 0 0 1 0 1 0 1 0 0f (1 nm )
nm^4
A a s re c e iv e d
B e tc h e d 1 min
C e tc h e d 3 0 min
D e tc h e d HF 4 9
Fig 46 Log-Log plot of the averaged power spectra of the STM images for all the samples
which gives a more stable tunnelling junction The LEED pattern is excellent showing
clear diffraction peaks at energies as low as 37 eV comparable to that obtained after high
temperature cleaning in UHV
Fig 46 shows the power spectra of the samples The high frequency and low frequency
areas of the power specturm are considered separately Table 41 shows for each sample
together with the roughness the integrated area of the power spectrum for the high and low
frequency part
We first examine the differences between the samples in the low frequency part of the
spectrum (lt01 nm-1) In this spatial range the short dip in concentrated or dilute HF
(samples BD) seems to lower the roughness of the sample removing some of the larger
features of the native oxide surface visible in Fig 45a and thus decreasing the
corresponding part of the power spectrum Sample C however etched with HF for 30 min
shows an increase in roughness peaking at 0012 nm-1 (~85 nm corresponding to the
typical dimensions of the corrugations visible in Fig 45c) On this scale sample C has a
PS density of 435 nm4 compared to 101 nm4 for sample A
Samples A B and C have the same PS in the high frequency range showing that the fine
structure of the surface is not changed by the HF etching Sample D on the other hand
shows a decrease of roughness for frequencies gt 015 nm-1 This effect could be due to the
VPalermo 41
improved surface cleanliness after etching with concentrated HF which would give a more
stable STM junction thus reducing the high frequency noise in the image
Fig 45 and the analysis of the power spectrum of each sample shows that a rapid dip in
HF removes the native oxide but does not lead to major changes in the morphology of the
Si surface its only effect being the removal of some of the larger features present on the
original surface Prolonged etching on the other hand increases the RMS roughness of the
surface
Etching of Si(100) in concentrated ammonium fluoride and nano-hole creation
Two different types of commercial p-doped Si(100) wafers (2Ω-cm and 10Ω-cm) from
MEMC were used Several different samples of 10x5 mm were immersed for 10 minutes
in 40 electronic grade NH4F solution under agitation Previous works used low
temperatures or anodic potentials applied to the silicon to avoid gaseous hydrogen
production and to obtain a uniform surface but in our experiment we worked at room
temperature to check the influence of hydrogen bubbles on surface morphology During
the etching the stirring was sufficient to provide a uniform concentration of reagents over
the whole sample surface but not to mechanically remove the hydrogen bubbles from the
silicon surface
After the etching each sample was rinsed in ultra-pure water to remove any etching
residues and observed with STM AFM and optical microscopy The AFM measurements
were made in air while for STM measurements the samples were rapidly dried with
nitrogen and inserted into the vacuum system to avoid surface reoxidation After insertion
into the vacuum surface cleanliness was checked with LEED and the surface morphology
observed by STM Parameters for STM measurements were sample bias +4 V 1 nA
current 16 Hz scan rate The images obtained were stable and reproducible over several
days Scan parameters for AFM were 20 nN force and 1 Hz scan rate
Some of the samples were cleaned with an RCA standard clean [1] before NH4F etching to
check the influence of possible surface contaminants on the final results Eight different
samples were prepared and more than sixty STM images of the samples were taken at
different points of the various samples
42 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
(110)
(110)
Fig47 abc) STMtopographic images of differentetching morphologies Eachimage is 500x500 nm Z-ranges are 10 10 and 18 nmrespectively d) STM image of a bridgecreated by etching of the lowerlayers of silicon (black arrow)Image is 250x250x6 nm
After ~2 min of immersion in the solution hydrogen bubbles become visible on the sample
surface The production is slow and the bubbles are quite stable on the sample without
detaching Thus some areas of the surface are masked from the liquid etching action
STM observations (fig 47) show that at the nanometer scale the surface is unevenly
covered with holes of radii ranging from 10 to 200 nm with depths of 2-4 nm These holes
have a wide range of different shapes and distributions In most cases the surface was
covered with a uniform distribution of round-shaped holes (fig47a) indicating isotropic
etching The dimension and the density of the holes changed greatly from sample to
sample and even over the surface of a single sample In some cases the etching was
anisotropic yielding nearly square holes and layered structures as shown in fig 47b
Square holes have been previously observed in cases where the etching speed in the (110)
direction is significantly smaller than in the (100) direction [17]
Over these areas (fig 47b and especially 47c) it is clearly visible how once the surface
had been attacked the reaction continued to preferentially remove atoms at step
irregularities (kink atoms) straightening step edges Eventually the exposed underlying
silicon was also attacked and further holes created inside the previously etched larger
ones It was not possible to detect monatomic steps on this kind of surface The smallest
step height observed was ~15 nm corresponding to several atomic layers In the image
shown in Figure 47a the etching was not strong and created only anisotropic holes on the
surface In fig 47b and c the stronger etch proceeded laterally for several tens of
VPalermo 43
nanometres leaving straight steps several tens of nanometers long and roughly rectangular
holes as expected given the structure of the (100) crystal face In some cases a
significative underetch is observed and the formation of suspended bridges and tunnels
can be deduced from the STM images (fig 47d)
The formation of branched pores and suspended structures has been attributed during pore
formation to diffusion limited aggregation effects where the h+ charge carriers necessary
for silicon etching have a higher probability of reacting at pore bottoms than reaching the
upper part of the silicon surface In the case of very deep pores quantum wire effects have
been invoked to explain the pore growth mechanism [15] In our case though the pores
formed were very shallow the underetch depth being only a few nanometres on pores of
sim100 nm width Thus more than diffusion effects the main contribution to the
underetching process must come from anisotropic etching and some kind of autocatalytic
reaction path analogous to the one described by Matsumura et al [4] with some areas of
the silicon surface hydrogenated and thus less vulnerable to etching
Pre-treatment with RCA cleaning has no effect on the final morphology and this seems to
exclude pore nucleation being caused by presence of metallic or organic surface
contaminants
The morphology and the distribution density of the pits was quite uniform over
microscopic areas of the sample but changes were observed over the millimetre scale This
suggests that etching intensity is influenced by some large-scale parameter
Large-area measurements made with AFM or with an optical microscope (fig 48)
showed that the inhomogeneity of surface etching can be correlated with the masking
action of the bubbles While the fluoride dissolved the silicon hydrogen bubbles formed
by the reaction covered some areas of the surface thus blocking the etching over that area
generating macroscopic steps at the bubble-liquid border As the reaction proceeded more
hydrogen accumulated and the bubble diameter increased producing in this way a circular
pattern of steps The increase in bubble diameter was not continuous with time otherwise a
uniform surface slope gradient would have been obtained The formation of this circular
ldquoetching staircaserdquo indicates that the bubble growth was stepwise the bubble accumulating
more and more hydrogen without enlarging across the surface until it relaxed increasing
its diameter stepwise and covering more silicon The circular structures in fig 48a are not
co-axial and their asymmetry could derive from physical processes due to stirring or
irregularities on the surface
44 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
a b Fig 48 a) optical micrograph of etching patterns on Si(100) created by NH4F 12x09 mm b) AFM image of the circles border xy range is 40x24 microm z-range is 30 nm
The step structure was not destroyed by the etching even after the bubble detached from
the surface but on the contrary the etching process seemed to be influenced by the
presence of the step
Observing in detail a series of steps (fig 48b) a quite deep trench is visible at the base of
each step A close-up image of a step and the corresponding line profile of the trench is
shown in fig 49 The trench is sim5 nm deep with respect to the lower surface compared to
a step height of 22 nm
A similar structure has been recently obtained with electrochemical etching of p-type
Si(100) in 4 HF [18] in which a ldquocurrent burstrdquo etching model previously described
was assumed for silicon dissolution In that experiment the trench was created at the
border of silicon nitride masks and began to grow after a nucleation stage Preferential
trench etching was along the (110) direction and the trench growth was explained as an
effect of mechanical stress induced by the nitride mask and of electric field enhanced
dissolution which depended upon an external applied potential
While it is clear that in our system the gas bubbles have a masking effect similar to a
classical solid nitride mask it is unlikely that hydrogen present on the surface can induce a
significant stress in the silicon lattice as in the case of a nitride mask Furthermore no
external field was applied to drive preferential etching at the trench site
It has been proposed [19] that the cathodic and anodic part of the etching reaction
(hydrogen production and silicon oxidative etching respectively) take place at different
points on the surface with a net charge transfer between the different areas In this case
the highest reaction rates will correspond to the silicon area surrounding the bubble border
where a high number of positive charges will be available for the reaction Furthermore a
sharp trench extending into the silicon crystal will be a preferential electrostatic attractor
VPalermo 45
Fig 49 AFM image of the etched surface showing a step created on the surface by bubblemasking A stronger etching action is visible on the right side of the step as well as theprotected area on the upper side of the step (indicated by the arrows) Image is 10x10 micromz-range is 30 nm The profile on the right is taken from the central area of the image
for the h+ charge carriers coming from other ldquocathodic areasrdquo of the sample either from
other regions on the surface or from the back of the silicon chip [18]
In the areas where the hydrogen bubble had detached and the surface was exposed to the
etching the reaction was not uniform in the neighbourhood of the steps It is possible to
observe (fig 48b and 49) an area on the upper side of the step where less or even no
etching at all seems to have taken place as if the step was able to protect the surface from
etching While etching on the lower side with trench formation can be attributed to the
presence of the bubble the surface on the upper step side can be etched only after bubble
detachment so no masking effect can account for this result However a further
preferential attraction of h+ charge carriers from the already formed trench can be
hypothised electrochemically shielding the surrounding area from further etching If this is
true the shielding effect would be very strong with a relatively shallow 5 nm-deep trench
protecting an area of sim1 microm parallel to the step
To summarise the etching of Si(100) in NH4F is a complex process in which different
reaction paths both chemical and electrochemical co-exist Hydrogen bubbles formed by
the reaction act as a mask on the surface and create etching paths and inhomogeneous
etching of the surface Different kinds of pores are observed on the surface and in some
cases the anisotropy of the process is so strong as to give square-shaped holes and
underetching
46 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The diffusion of h+ charge carriers in the crystal is one of the main rate-determining steps
of the reaction and leads to the formation of a deep trench immediately outside the bubble
perimeter These trenches act as charge collectors and reduce the etching of the upper step
surface in the proximity of the steps
Bibliography
[1] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[2] Spectroscopic Ellipsometry Studies Of HF Treated Si (100) Surfaces Yao H Woollam
Ja Alterovitz SA Applied Physics Letters 62 (25) 3324-3326 Jun 21 1993 Influence
Of HF-H2O2 Treatment On Si(100) And Si(111) Surfaces Graf D Bauermayer S
Schnegg A Journal Of Applied Physics 74 (3) 1679-1683 Aug 1 1993 Kinetics Of
Oxidation On Hydrogen-Terminated Si(100) And (111) Surfaces Stored In Air Miura
T Niwano M Shoji D Miyamoto N Journal Of Applied Physics 79 (8) 4373-4380
Part 1 Apr 15 1996
[3] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995 Structure Of The Stepped SiSiO2 Interface After Thermal-Oxidation -
Investigations With Scanning Tunneling Microscopy And Spot-Profile Analysis Of
Low-Energy Electron-Diffraction Pietsch GJ Kohler U Jusko O Henzler M Hahn
PO Applied Physics Letters 60 (11) 1321-1323 Mar 16 1992
[4] Enhanced Etching Rate Of Silicon In Fluoride Containing Solutions At pH 64
Matsumura M Fukidome H Journal Of The Electrochemical Society 143 (8) 2683-
2686 Aug 1996
[5] A Study Comparing Measurements Of Roughness Of Silicon And SiO2 Surfaces And
Interfaces Using Scanning Probe Microscopy And Neutron Reflectivity Crossley A
Sofield CJ Goff JP Lake ACI Hutchings MT Menelle A Journal Of Non-Crystalline
Solids 187 221-226 Jul 1995
VPalermo 47
[6] Step-Flow Mechanism Versus Pit Corrosion - Scanning-Tunneling Microscopy
Observations On Wet Etching Of Si(111) By Hf Solutions Hessel HE Feltz A Reiter
M Memmert U Behm RJ Chemical Physics Letters 186 (2-3) 275-280 Nov 8 1991
[7] Comparison Of Si(111) Surfaces Prepared Using Aqueous-Solutions Of NH4F Versus
HF Higashi GS Becker RS Chabal YJ Becker AJ Applied Physics Letters 58 (15)
1656-1658 Apr 15 1991
[8] Wet Chemical Etching Of Si(100) Surfaces In Concentrated NH4F Solution -
Formation Of (2x1)H Reconstructed Si(100) Terraces Versus (111) Faceting Neuwald
U Hessel HE Feltz A Memmert U Behm RJ Surface Science 296 (1) L8-L14 Oct
10 1993
[9] Ideal Hydrogen Termination Of Si(001) Surface By Wet-Chemical Preparation Morita
Y Tokumoto H Applied Physics Letters 67 (18) 2654-2656 Oct 30 1995
[10] Atomic Structures Of Hydrogen-Terminated Si(001) Surfaces After Wet Cleaning
By Scanning Tunneling Microscopy Endo K Arima K Kataoka T Oshikane Y Inoue
H Mori Y Applied Physics Letters 73 (13) 1853-1855 Sep 28 1998
[11] On The Potential-Dependent Etching Of Si(111) In Aqueous NH4F Solution
Houbertz R Memmert U Behm RJ Surface Science 396 (1-3) 198-211 Jan 20 1998
[12] Etching Process Of SiO2 By HF Molecules Hoshino T Nishioka Y Journal Of
Chemical Physics 111 (5) 2109-2114 Aug 1 1999
[13] Etching Mechanism And Atomic-Structure Of H-Si(111) Surfaces Prepared In
NH4F Allongue P Kieling V Gerischer H Electrochimica Acta 40 (10) 1353-1360
Jul 1995
[14] Pore Formation Mechanisms For The Si-HF System Carstensen J Christophersen
M Foll H Materials Science And Engineering B-Solid State Materials For Advanced
Technology 69 23-28 Sp Iss Si Jan 19 2000
[15] Porous Silicon - Mechanisms Of Growth And Applications Parkhutik V Solid-
State Electronics 43 (6) 1121-1141 Jun 1999
[16] Formation And Application Of Porous Silicon Foll H Christophersen M
Carstensen J Hasse G Materials Science amp Engineering R-Reports 39 (4) 93-141 Nov
1 2002
[17] Pore Morphology And The Mechanism Of Pore Formation In N-Type Silicon
Searson PC Macaulay JM Ross FM Journal Of Applied Physics 72 (1) 253-258 Jul 1
1992
48 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
[18] Deep Electrochemical Trench Etching With Organic Hydrofluoric Electrolytes
Christophersen M Merz P Quenzer J Carstensen J Foll H Sensors And Actuators A-
Physical 88 (3) 241-246 Mar 5 2001
[19] Morphology Of Anodically Etched Si(111) Surfaces - A Structural Comparison Of
NH4F Versus HF Etching Houbertz R Memmert U Behm RJ Journal Of Vacuum
Science amp Technology B 12 (6) 3145-3148 Nov-Dec 1994
VPalermo 49
5 Surface modification of silicon in vacuum void creation and oxide desorption
The main reason for the huge success of silicon in the microelectronics industry is not due
to its superior properties as a semiconductor Other materials for example germanium
have better qualities such as higher mobility of charge carriers and lower noise levels
which would allow the construction of faster and higher performance devices
The widespread use of silicon however is mainly due to the outstanding characteristics of
its oxide Silicon dioxide (SiO2) is a very good electrical insulator easy to form
chemically and thermally stable and is compatible with lithographic and metal deposition
processes Germanium oxide on the contrary is too reactive to be used
Even the use of Si(100) substrates for nearly all microelectronic devices is dictated by
oxide quality The (111) face of silicon crystal can be easily cleaved and flattened and
almost atomically perfect surfaces can be obtained with simple chemical procedures (as
described above) But the density of interfacial defects is highest for Si(111)-SiO2
interfaces and lowest for Si(100)-SiO2 ones so microchips will continue to be fabricated
on Si(100) wafers
SiO2 (silica) is present in 95 of the earthrsquos minerals in different allotropic forms such as
quartz tridymite and cristobalite In the bulk each silicon atom is bonded to four oxygens
in a Si-O-Si tri-dimensional network Si-O bonds are 016 nm long and form an angle
ranging from 120deg to 150deg
Three typical intrinsic defects are present in SiO2 The so-called Ersquo centres are oxygen
vacancies with a hole localised on a silicon atom with only three Si-O bonds
50 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
O3Si +SiO3 Whereas the PR (peroxy radical) defects are holes trapped by a charged
peroxy moiety with a O3Si-O-O+ SiO3 structure The NBOHC (non-bridging oxygen hole
centres) derive from water or hydrogen contamination and are schematized as O3Si- O- H-
O-SiO3
The atomic structure of the Si-SiO2 interface varies enormously Local domains resembling
the tridymite and the cristobalite structure of silica are present but it seems that only 10
of the interface is ordered [1] Far from the interface the SiO2 bulk is completely
disordered The passage from bulk Si to stoichiometric SiO2 passes through a non-
stoichiometric SiOx layer 07 nm thick
When a clean silicon surface is exposed to atmospheric oxygen a thin sim2 nm thick layer
of native oxide forms spontaneously which is usually removed and substituted with
thicker better quality oxide layers before further processing
Silicon is usually oxidised by thermal annealing at temperatures between 800deg and 1100deg
in an atmosphere of pure O2 with some water eventually added to increase oxidation speed
Thermal oxides made in pure oxygen (dry oxides) grow more slowly than oxides produced
in an oxygen-water atmosphere (wet oxides) but are usually of better quality
According to the Deal-Groove formula the time t needed to grow an oxide of thickness X
is given by [1]
1
212minus
minusminus
+=
ABXBXt α
where the constant B and BA decrease exponentially with temperature as
minus
kTEexp
with activation energies for dry oxidation of EB =123 and EBA =20 eV respectively
EB is related to the diffusion of oxygen in silicon while the value of EBA is interpreted as
the energy required to break a Si-Si bond The exponent α is 1 for wet oxidation and 0 for
oxidation at high temperatures and low oxygen pressures It has intermediate values for dry
oxidation This formula does not work well for low values of X at the initial stages of
oxidation and usually empirical corrections are used
An interesting characteristic of silicon is that at high temperature and in vacuum oxygen
can actually etch the silicon crystal giving gaseous products and the oxidized layer present
on the crystal becomes unstable (fig 51) [2]
VPalermo 51
1E-10
1E-09
1E-08
1E-07
1E-06
1E-05
1E-04
1E-03
1E-02
1E-01
0607080911112
1000T (1K)
P (T
orr)
SiO2 + Si rarr 2SiO(g) Oxide decomposition
Si+O2 rarr SiO2(s) oxide formation Fig51 phase diagram of the
oxygen-silicon system
Silicon oxidation apart from the initial nucleation stages at the monolayer level proceeds
uniformly over the whole surface with a planar reaction front moving from the surface
into the bulk
If heated under low oxygen partial pressure (vacuum or inert atmosphere) SiO2 is known to
decompose following the reaction
SiO2 + Si rarr 2SiOuarr (1)
The reaction begins with nucleation at defect points on the SiSiO2 interface and proceeds
in a spatially inhomogeneous manner with the formation of large voids on the oxide
surface [3]
Several studies have been made on the dynamics of void growth both on thick [4] and thin
[5] layers of SiO2 The process has been used to decorate otherwise unobservable defects at
the SiSiO2 interface [6] or to grow nanoislands of silicon on the void surface [7] It has
been suggested that the defects acting as nucleation centres could be metallic contaminants
present on the native surface which aggregate and catalyze SiO2 decomposition [8]
After oxide desorption the silicon surface is very rough In particular on Si(100) square
islands are observed several nanometres high which act as lsquopinning sitesrsquo on the motion of
monatomic steps on the Si surface It has been proposed that these islands can be composed
of silicon carbide coming from the organic contaminants present on the surface [9] or of
oxide clusters during partial reoxidation [10]
In the following section some experimental results obtained studying voids and nanoisland
growth are presented
52 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nanoisland growth on silicon
We used two different samples the first rougher than the second (i) p-type Czochralski Si
(100) with a RMS roughness of 05 nm shown in fig52a and (ii) p-type epitaxial Si(100)
with a RMS roughness of 019 nm shown in fig 52b The roughness was measured with
STM on the lsquoas receivedrsquo samples on different areas of 500x500 nm Every sample was
covered by a layer of native oxide ~2 nm thick
Some of the samples (AC) were introduced into the UHV system without any cleaning
others (BD) were dipped in concentrated HF (49) to remove the surface oxide without
etching the silicon [11]
After degassing the samples were heated resistively increasing the temperature slowly to
keep the pressure within the 10-10 mbar range during heating Each sample was held at
900˚C for 30 min For some of the samples the heating was stopped at lower temperatures
to obtain incomplete oxide desorption and surface roughening thus enabling observation
of the different steps of the process Table 51 summarizes the different treatments for each
group of samples
Fig52 Original surfaces of Czochralski (a) and epitaxial (b) silicon Every image is 500x500 nm
VPalermo 53
Table 51 Summary of different treatements
Silicon type Oxide removed
Heating Islands density (microm-2)
Av Island Volume (nm3)
A Czochralski No 30rsquo at 900˚C 24x103 353
B Czochralski Yes ldquo 20x104 71
C Epitaxial No ldquo 28x102 2450
D Epitaxial Yes ldquo 23x104 63
Fig53 ab) Surface roughening on group A samples after heating 30 min at 800deg and900deg respectively in presence of an oxide layer Image size 500x500 nm cd) Surface roughening on group B samples after heating 30 min at 700deg and 900degrespectively after removing the oxide layer with HF Image size 250x250 nm
Fig 52a is the native oxide surface of sample A quite disordered and irregular This
surface was stable when heated to 700˚C while at higher temperatures the oxide began to
desorb Fig 53a was taken after heating the sample at 800˚C for 30 min
54 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The oxide began to desorb in correspondence with defect points creating oxide-free
surface voids which enlarge radially uncovering the silicon surface Some surface silicon
atoms are removed through reaction (1) Mobile silicon atoms moving randomly over the
surface aggregate forming a nanocrystal at the initial defect point surrounded by a stable
flat oxide-free surface with some steps The initial SiSiO2 interface as observed after
oxide etching with HF [12] is similar to the one shown in fig 52a and thus the surface
observed in the void area is a completely new one coming from surface reorganization
The reaction continues leading to a coalescence of the voids the final result of the process
is shown in fig 53b with protruding islands aligned along the main crystal axes and flat
areas with some steps LEED measurements confirmed that this surface is crystalline
silicon Atomic resolution was obtained on the flat areas of the surface showing the
typical 2x1 reconstruction of Si(100) Figures 53c and 53d show the evolution of the
roughening on sample B after chemically removing the native oxide layer before
introducing the sample into UHV As mentioned above the initial morphology is similar to
that shown in fig 52a though the oxide has been removed as confirmed by LEED
measurements Fig 53c shows the surface after heating at 700˚ for 30 min Without the
oxide layer the surface began to reorganize at lower temperatures and the small irregular
grains of the original surface begin to grow and become more rectangular After heating to
900˚C (fig 53d) the surface is flat with small rectangular islands The shape of the islands
is similar to that shown in fig 53b but the lateral island dimension is one order of
magnitude smaller
To assess the importance of the initial surface on the process the same treatments were
repeated starting from flat epitaxial silicon (samples CD) The initial epitaxial surface is
smooth with periodic steps due to a small miscut angle (fig52b) On heating the epitaxial
silicon without etching (sample C) relatively large islands are obtained similar to the ones
shown in fig 53b where the ldquostep-pinningrdquo effect of the islands is evident The average
dimension of sample C islands was sim2500 nm3 while only 353 nm3 for sample A Island
density decreased from 24000 to 280 microm-2 This result is maybe due to a lower
concentration of defects in the oxide layer of the epitaxial silicon which allows the
expanding void to grow much larger before coalescence with neighbouring voids the
nanostructures within them thus reaching greater dimensions To verify this hypothesis we
applied the same treatment to clean not-so clean and purposely contaminated samples To
check that the defects inducing void nucleation and nanoisland growth derive from carbon
VPalermo 55
contamination we used electronic-grade methanol to contaminate the silicon surface
Figure 54 shows the results of this comparison
State-of-the-art cleaned samples obtained with multi-step RCA cleaning [13] showed the
development of very few voids the oxide desorption leaving large smooth areas of oxide-
and nanostructure-free silicon as shown in Fig 54a where the presence of residual oxide
can be seen Samples cleaned with a simpler one-stage Pirana cleaning solution showed a
higher number of islands per unit area (Fig 54b) and samples purposely contaminated
with methanol (Fig 54c) had the highest density of island nucleation of all three samples
When the native oxide had been chemically removed from the epitaxial silicon before
heating (sample D) the island size and density is similar to the standard as-received
silicon case (sample B) Thus starting from a flat substrate the process of surface
roughening and island creation was the same even when the oxide layer had been
previously removed
From these results some important indications can be drawn The first is that oxide
desorption temperature depends upon the contamination levels of the oxide layers (fig 54)
with desorption at lower temperatures for contaminated surfaces
Second once the oxide layer is removed the silicon atoms become mobile on the surface
at relatively low temperatures (~700˚C) Their initial configuration is disordered and has a
high surface energy so they crystallize in the small islands shown in fig53d reducing
their surface area and lowering the energy of the system When the oxide is present the
surface cannot reorganize at 700˚C and is stable At 800˚C the oxide begins to desorb in
correspondence with defect points on the surface As the void area enlarges a small
cba Fig 54 The effect of contamination on nano-island production Gradient-filtered STM images of (a)sample cleaned two times with RCA 400x400 nm (b) Sample cleaned with a Pirana solution 500x500nm (c) sample contaminated with MeOH before insertion into the UHV system 500x500 nm Annealingtime is 40 min at 800degC for all the samples
56 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
crystallite grows at the centre of the voids Further oxide desorption uncovers larger areas
freeing more silicon atoms which feed the initial island making it bigger
Fig 55 shows the scaled size distribution of the islands for each group of samples Every
distribution has been obtained using several images of different samples for each group
The size distribution of all the samples follows an exponential decay and the decay seems
the same for all the samples though the average volume of the islands differs by more than
one order of magnitude (see Table 51 for details)
The point defects which catalyze void nucleation can be metallic contaminants present on
the initial oxide surface [8] or organic contaminants which at high temperatures can
form SiC nanocrystals [14] We used Scanning Tunneling Spectroscopy (STS) to look for
differences between the islands and the surrounding flat silicon surface but no difference
was found Furthermore the total island volume per surface unit is very high (more than
104 nm3microm2) and it seems unlikely that such a huge volume could consist of surface
contaminants Islands of pure silicon on silicon have been grown without evidence of
surface contamination[14]
To summarise the formation of nano-sized islands on silicon through surface diffusion
was studied At high temperatures the oxide layer covering the surface decomposes non
uniformly and circular voids of clean silicon are created The presence of the oxide layer
blocks surface silicon atomic motion and surface reorganization except within the voids
causing the growth of islands more than 10 nm high and 30 nm wide on the silicon surface
According to STM and LEED measurements we can say that the islands are mostly
Fig 55 Size distribution (scaled) for the samples described in table 1 The line is a reference for the eye
VPalermo 57
composed of silicon with traces of other substances such as carbon Final island densities
and dimensions depend upon the initial purity of the oxide layer
On the other hand if the oxide is removed chemically before heating the island growth
process is not so localised and takes place simultaneously over the whole surface yielding
smaller and more numerous islands
58 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[2] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[3] High-Temperature SiO2 Decomposition At The Sio2Si Interface Tromp R Rubloff
GW Balk P Legoues FK Physical Review Letters 55 2332-2335 Nov 1985
[4] Defect Microchemistry At The SiO2Si Interface Rubloff GW Hofmann K Liher M
Young DR Physical Review Letters 582379-2382 Jun 1987 Defect Formation In
Thermal SiO2 By High Temperature Annealing Hofmann K Rubloff GW Mccorkle
RA Applied Physics Letters 49 1525-1527 Dec 1986 Kinetics Of High-Temperature
Thermal Decomposition Of SiO2 On Si(100) Liher M Lewis JE Rubloff GW Journal
Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 5 1559-1562 Aug
1987
[5] Thermal-Decomposition Of Very Thin Oxide Layers On Si(111) Kobayashi Y Sugii
K Journal Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 10 (4)
2308-2313 Part 3 Jul-Aug 1992 Scanning Tunneling Microscope Study On Mid-
Desorption Stages Of Native Oxides On Si(111) Kobayashi Y Sugii K Journal Of
Vacuum Science amp Technology B 9 (2) 748-751 Part 2 Mar-Apr 1991 Controlled
Growth Of SiO2 Tunnel Barrier And Crystalline Si Quantum Wells For Si Resonant
Tunneling Diodes Wei Y Wallace RM Seabaugh AC Journal Of Applied Physics 81
(9) 6415-6424 May 1 1997
[6] Defect Formation In SiO2Si(100) By Metal Diffusion And Reaction Liher M
Dallaporta H Lewis Je Appl Phys Lett 53 589-591 Aug 1988 SiO2 Film
Decomposition Reaction Initiated By Carbon Impurities Located At A Si- SiO2
Interface Raider Si Herd Sr Walkup Re Applied Physics Letters 59 (19) 2424-2426
Nov 4 1991
[7] Nanometer-Scale Si Selective Epitaxial Growth On Si(001) Surfaces Using The
Thermal Decomposition Of Ultrathin Oxide Films Fujita K Watanabe H Ichikawa M
Applied Physics Letters 70 (21) 2807-2809 May 26 1997 Pyramidal Si Nanocrystals
VPalermo 59
With A Quasiequilibrium Shape Selectively Grown On Si(001) Windows In Ultrathin
SiO2 Films Shibata M Nitta Y Fujita K Ichikawa M Physical Review B 61 (11)
7499-7504 Mar 15 2000
[8] Stacking-Fault-Induced Defect Creation In SiO2 On Si(100) Liher M Bronner Gb
Lewis Je Appl Phys Lett 52 1982-1985 May 1988
[9] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990 AFM And XPS Characterization Of The
Si(111) Surface After Thermal-Treatment Lamontagne B Guay D Roy D Sporken R
Caudano R Applied Surface Science 90 (4) 481-487 Dec 1995
[10] Scanning-Tunneling-Microscopy Study Of Oxide Nucleation And Oxidation-
Induced Roughening At Elevated-Temperatures On The Si(001)-(2x1) Surface Seiple
JV Pelz JP Physical Review Letters 73 (7) 999-1002 Aug 15 1994 Evolution Of
Atomic-Scale Roughening On Si(001)-(2x1) Surfaces Resulting From High-
Temperature Oxidation Seiple JV Pelz JP Journal Of Vacuum Science amp Technology
A-Vacuum Surfaces And Films 13 (3) 772-776 Part 1 May-Jun 1995
[11] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995
[12] Morphological Changes Of The Si [100] Surface After Treatment With
Concentrated And Diluted HF Palermo V Jones D Materials Science In
Semiconductor Processing 4 (5) 437-441 Oct 2001
[13] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[14] Nanoscale Roughening Of Si(001) By Oxide Desorption In Ultrahigh Vacuum
Gray SM Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology
B 14 (2) 1043-1047 Mar-Apr 1996
60 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 61
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and nano-lines
When the silicon surface is not protected by a native oxide layer or by a hydrogenated
passivating layer it is very reactive The surface chemistry of both Si(100) and Si(111) has
been extensively studied starting from clean surfaces prepared in UHV because of its great
relevance to the microelectronics industry and its technological interest The diffusion and
reaction of molecules and atoms on silicon is an interesting scientific problem on its own
apart from technological issues because silicon reconstruction yields a very complex and
anisotropic surface
We give here a brief summary of the adsorption behaviour of several elements on Si
surfaces The summary is not comprehensive and only the most interesting characteristics
for each substance are given
Table 61 Summary of adsorption behaviour of atoms and molecules on silicon surfaces [12]
Hydrogen
Molecular H2 shows low reactivity towards silicon while atomic hydrogen easily forms
Si-H bonds and can even break Si-Si bonds
Alkali metals
Alkali metals diffuse rapidly into SiO2 and can damage silicon-based transistors
Transition
metals
All transition metals apart from gold and silver react with Si forming metal silicides
62 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nickel A very common contaminant of silicon Nickel from even slight contact with stainless
steel tweezers can contaminate silicon samples forming its silicide and is very difficult
to remove even at high temperatures because it segregates on the silicon surface
Palladium Forms silicides especially Pd2Si and can be used to make contacts possessing a low
Schottky barrier
Titanium Widely used to fabricate contacts for silicon-based microdevices TiSi2 is one of the
more stable and highly conductive silicides
Tungsten The silicide is more stable than TiSi2 on polysilicon and is thus used for polysilicon
gate contacts
Platinum Silicide is used for bipolar transistors PtSi contacts on n-silicon give Schottky diodes
Cobalt Cobalt silicide gives better quality contacts than TiSi2 at the nanometrelevel but is less
used at the moment because it consumes too much silicon during its growth
Iron Silicides have been studied for potential optoelectronic applications because FeSi2 has a
directly accessible bandgap and is optically active
Group 13
Group 13 metals are used in IC technology to dope silicon (giving p-type doping) and
to make III-V type semiconductors When evaporated on Si(100) or Si(111) they react
strongly with the surface giving surface reconstruction and kink formation at steps
Boron Widely used for doping Usually deposited with decomposition of boron hydrides
Sticking coefficient of simple hydrides is very low so decaboranes (B10H14) are used for
deposition
Aluminum Aluminum is widely used for microelectronic contacts but in some cases it reacts with
silicon diffusing into the crystal and with SiO2 stealing oxygen atoms to form Al2O3
Group 14
Carbon Carbon can form a composite semiconductor with silicon (silicon carbide) widely used
to fabricate integrated circuits for use at elevated temperatures or in the presence of
ionising radiation When carbon is deposited on silicon usually polycrystalline films
with properties between diamond and graphite are obtained
CO adsorbs molecularly on silicon without breakage of the C=O bond Other
unsaturated hydrocarbons do not react with silicon Even very reactive strained
molecules like cyclopropane have low sticking coefficient Ethylene acetylene and
propylene adsorb molecularly on Si(100) each molecule sitting on a silicon dimer
interacting with the dangling bonds of the surface Acetylene also adsorbs molecularly
on Si(111) on the 7x7 reconstructed surface preferential adsorption on centre atoms
with respect to corner atoms is observed Benzene and other aromatic ring compounds
adsorb molecularly on the surface They can form σ or π-type bonds with the silicon
laying parallel or tilted respect to the surface according to the bond type
Silicon Silicon atoms can be deposited on the surface with molecular beam epitaxy or
VPalermo 63
decomposition of silanes and chlorosilanes Silanes with single Si-Si bonds have a high
sticking probability and dissociate upon adsorption giving trihydrides and
monohydrides
Chlorosilanes have high sticking coefficients too and decompose on the surface but
they can etch the surface through the reaction SiCl4 + Si rarr 2SiCl2
Germanium Germanium has a lattice constant similar to silicon (Ge lc is 4 larger than Si) so that
various alloys of Si1-xGex can be formed The growth of Germanium on silicon is quite
peculiar with several monolayers adsorbing uniformly on the surface followed by
island formation (Stranski-Krastanov growth)
Group 15
Nitrogen Exposure to ammonia and nitrogen gas at high temperatures leads to the formation of
silicon nitride layers (Si3N4) NH3 dissociates on the surface and reacts with the
dangling bonds giving Si-NH2 and Si-H termination Silicon nitride is stable on the
surface up to 1100degC above this temperature it desorbs as Si2N
Phosphorous Phosphine (PH3) adsorbs dissociatively as Si-PH2 and Si-H and behaviour is similar to
that of ammonia
Arsenic A monolayer of As is often deposited on silicon as a substrate for GaAs growth Arsenic
forms dimers on Si(100) creating dimer rows which are parallel (perpendicular) to the
underlying silicon dimer rows when it is deposited at high (low) temperature On
Si(111) Arsenic breaks the 7x7 reconstruction giving a 1x1 pattern This 1x1 As
monolayer acts as a passivating layer and can resist further As adsorption oxygen and
air
Antimony and
Bismuth
Due to their large covalent radii Sb and Bi form only short dimer rows on silicon and
only at high temperatures
Group 16
Oxygen
Oxygen can oxidize or etch silicon according to the reactions
Si(s) + O2 rarr SiO2 (s)
Si(s) + frac12O2 rarr SiO(g)uarr
Silicon oxidation has already been described in the previous chapter Molecular oxygen
adsorption on silicon is mostly dissociative O atoms break the Si-Si bonds forming an
Si-O-Si bridge or a peroxy bridge Si-O-O-Si Hydrided silicon surfaces do not have
surface dangling bonds and are thus more protected from oxygen attack Etching is
favoured at high temperatures and low oxygen pressures [3] At T gt700degC silicon oxide
on silicon decomposes as SiO(g)
Water
H2O adsorbs dissociatively on silicon easily on Si (100) and with more difficulty on
Si(111) The molecule decomposes giving Si-OH and Si-H on adjacent sites
64 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Group 17
Fluorine
Fluorine adsorbs and reacts easily on silicon dissociating on the surface and forming
Si-F bonds It can then penetrate the surface and attack polarised Si-Si backbonds The
surface is etched with production of SiF4 and SiF2
Chlorine
Chlorine reacts aggressively with silicon etching the surface Cl2 and F2 are often used
commercially to etch silicon usually in a plasma Atomic Cl and Cl2 adsorb on Si(100)
up to saturation The most stable configuration seems to be a Cl atom bonded to each Si
atom of a surface dimer in a symmetric structure but metastable configurations with
two Cl atoms bonded to a buckled dimer and even a central Cl atom bridged across a
dimer have been observed
On Si(111) silicon mono- di- and tri-chlorides are formed In particular heating at
Tgt400degC a uniform Si-Cl monolayer is obtained and the 7x7 reconstruction changes
into a bulk like 1x1 lattice The surface transition induced by Cl adsorption is fully
reversible The 7x7 reconstruction can be restored on desorbing the chlorine by heating
at Tgt1100degC The 1x1 domains nucleate at the lower terrace side of steps the 7x7 at the
upper step edge
Bromine
Like hydrogen bromine maintains the 2x1 structure of Si(100) forming Si-Br bonds
with the dangling bonds of the silicon dimer rows At high exposure some etching of
silicon by formation of volatile SiBr3 species has been observed
Surface diffusion on silicon
In the previous chapter we observed nano-island growth on silicon surfaces with native
oxide and surfaces contaminated with organic impurities While this phenomenon has been
observed in several experiments the formation and growth mechanisms of nanoislands has
still not been satisfactorily described theoretically Various substances have been
hypothesised as nanoisland nucleating agents such as organic or metallic contaminants
present on the surface or oxide clusters (see previous chapter) The island growth process
has been attributed by various workers to lsquosome sort of kinetic instabilityrsquo [4] to the
pinning of step flow by SiC clusters [5] or to a mesoscopic atom flux from areas with low
step density to areas of higher step density [6]
One mechanism proposed to explain surface roughening and island growth is the Ehrlich-
Schwoebel effect ie the presence of an energetic barrier that reflects atoms approaching a
VPalermo 65
Molecule (L
O2
H2O MeOH
CO CO2 CH4
clean surface
downward
weak on si
The aim o
surface an
To obtain
the ones
spontaneou
first obser
surface W
nanoisland
Experime
We used S
removed f
41 mixtur
water and
whole hea
1200degC re
kept below
Table 62 Treatment island density and island volume for each molecule tested
Dose angmuir)
Heating time at 800degC (min)
Island density (microm-2)
Ave Volume (nm3)
Equivalent layer thickness (nm)
280 10 no islands - -
90 10 no islands - - 30 10 2000 19 004
180 10 360 516 019 30 60 no islands - -
180 10 no islands - - - 10 no islands - -
step [7] Recent experiments however indicate that the Schwoebel effect is
licon [48]
f our experiments was to study the reaction of simple molecules with the silicon
d to check the ability of these molecules to nucleate nanoisland growth
a better understanding of the process we chose simple molecules starting from
which are more likely to be present as traces in UHV chambers where
s nanoisland growth is often observed after sample heating Using STM we
ved at the atomic level the possible interactions of these molecules with the
e then heated the surface and checked the density and dimensions of the
s produced
ntal procedures and results
i(111) wafers p-doped 0015 Ω cm Gross contamination and particles were
rom the sample surface with a standard Pirana etch (15 min dip in a H2S4H2O2
e at T=80degC) After this the samples were thoroughly washed in ultra-pure
introduced into the UHV system After degassing for 5 hours at 600degC the
ting stage was allowed to cool down and a sequence of rapid flashes to 1100deg-
moved the native oxide layer and cleaned the surface Pressure during flashes is
1x10-9 mbar
66 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Surface flatness and cleanliness were checked with STM and the contaminant molecules
introduced into the UHV system through a leak valve Doses of several Langmuirs were
used for the experiments (1 Langmuir=10-6 torr sec-1) Pressure was kept constant at 10-7
mbar during adsorption The STM tip was retracted during gas deposition to avoid any tip
shielding effect After the adsorption process the leak valve was closed the base pressure
in the UHV chamber returning to below 10-10 mbar and the presence of adsorbed
molecules checked on the surface with STM Finally the sample was heated at 800degC for
several minutes to allow nanoisland growth Several different areas were examined on each
sample
Table 62 summarizes the dose of exposure and the heating time for each molecule Other
experiments were carried out using smaller dosing times but only the significant results
are reported here
As expected oxygen and water react with the surface oxidising it The adsorption process
can be easily followed with the STM by lowering the tip from time to time and taking an
image No effect of the STM tip is observed over the scanned areas Upon molecule
adsorption the order of the crystalline surface rapidly degenerates and after several
minutes the surface looks completely covered by irregular atomic-sized protrusions
corresponding to Si-OH or Si-O-Si species On heating to 800degC the contaminating species
were easily desorbed through SiO formation the surface became clean and the crystal
surface periodicity re-established No residues or nanoislands were observed
Amongst the molecules tested was methanol CH3OH reacts readily with the surface and
covers it with a disorderd layer already after sim 30 Langmuirs exposure (fig 61)
Fig 61 Si(111) surface during methanol adsorption at 0 1 and 5 minutes respectively (corresponding to 0 6 and 30 L exposure) The dark irregular lines are borders between different 7x7 domains Image size 60x60 nm
VPalermo 67
Synchroton radiation photoemission spectroscopy experiments by Carbone et al [9]
showed that methanol reacts with the rest atoms of the 7x7 silicon cell already after 1
Langmuir exposure but that longer exposures are needed for methanol to react with the
remaining silicon atoms Methanol adsorbs dissociatively forming SiOCH3 and Si-CHx
species on the surface
After methanol adsorption the sample is heated to 800degC and observed again with STM
As in the case of water and oxygen the heating restores the crystalline surface and the 7x7
pattern returns visible But this time quite a dense array of nanoislands is observed on the
surface (fig 62) The islands have an average diameter of 20 nm and a height of sim2 nm
At temperatures above 400degC the Si-OCH3 and Si-CHx species decompose oxygen and
hydrogen are desorbed and according to [9] carbon atoms remain as SiC dispersed
uniformly on the surface Our STM measurements indicate that the carbide does not
randomly cover the surface but that C atoms are concentrated at the nanoislands Rough
calculations based on island volume suggest that the islands must be composed of a SixC1-x
alloy with x varying between 05 and 1
Following these results we expected carbon monoxide to behave in a similar manner
adsorbing onto the surface and with nanoisland growth However no adsorption was
observed with STM and the surface looked perfectly clean even after a dose of more than
100 L of carbon monoxide Some species did adsorb onto the surface though because
after 10 minutes subsequent heating nanoisland growth was observed Island density in
this case was only 18 of the density obtained with methanol and island dimensions are
Fig62 Nanoislands on Si(111) created after 30 L adsorption of methanol and 10 min heating at 800degC Image size 200x200 nm
68 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
much bigger (see table 62 and fig 63) To check whether the islands had been nucleated
by some other contaminant present in the vacuum chamber we heated a blank sample
without introducing any molecule No island growth was observed Exposing the surface to
other simple molecules including carbon dioxide and methane also gave no nanoisland
formation
We can estimate the amount of carbon deposited on the surface during gas dosing The
total flux of molecules hitting the surface is obtained from the Hertz-Knudsen formula
TkmPF
π2=
Fig 63 STM images of Si(111) after exposure to methanol (left) and carbonmonoxide (right) contaminating molecules and subsequent nanoisland growthImage size 400x400 nm The size distribution of nanoislands is shown below eachimage
VPalermo 69
Fig64 Left Si(111) after 180 L contamination of CO and 16 hours heating at 800degC Right Si(111)without CO contamination after 16 hours heating at 800degC Image size 250x250 nm Vertical scale is thesame for both images z-ranges are 35 and 5 nm respectively
where P is the pressure in Pascal k=138x10-23 J K-1 is the Boltzmann constant T is the
temperature and m is the molecular weight (sim32 for methanol and sim28 for CO) in
kilograms
Using this formula we obtain a flux of 027 molecules nm-2 sec-1 for methanol and 029
molecules nm-2 sec-1 for CO After ten minutes exposure at 10-7 mbar pressure more than
150 molecules will have hit each square nanometer of the surface Surface density of
Si(111) 7x7 is sim16 atomsnm2 so each surface atom will be hit by several molecules which
could react or be adsorbed even assuming a sticking coefficient much lower than unity
To explain nanoisland growth we hypothesize that CO molecules adsorb molecularly and
-5
0
5
10
15
20
25
30
0 20 40 60 80 100 120
nm
nm
10 min40 min16 h16 h- no CO
Fig65 STM profiles of nanoislandsgrown for different annealing times The lower curve correspond to the surface heated at 16 hours without CO contamination
70 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Table 63 Island density and volume for the CO-contaminated surface at different heating times For each sample the thickness of a uniform layer having the same volume per unit area of the islands is calculated
Molecule Dose
(Langmuir) Heating time at
800degC (min) Island density
(microm-2) Ave Volume
(nm3) Equivalent layer thickness (nm)
CO 180 10 362 517 019 CO 180 40 550 1072 059 CO 180 16 h 505 3675 186
clean surface - 16 h rough - -
very weakly on Si(111) moving rapidly on the surface Because of their rapid diffusion
the STM is not able to image the CO molecules on the silicon [10]
Increasing the temperature part of the CO molecules will surely desorb but some of them
will dissociatively adsorb onto the surface The oxygen will then desorb as SiO while
carbon will remain fixed forming very stable Si-C bonds
CO decomposition will be favored at surface defects surface steps or in the proximity of
already formed SiC clusters An increased reactivity of the CO molecule on a surface in
correspondence with phase boundaries has already been observed on Pt surfaces [10]
In this case a reduced number of nucleation centres will react with the CO molecules
yielding large and fewer islands with respect to the case of methanol which does not have
the possibility of travelling long distances over the surface
Other factors are likely to influence the process of island nucleation after CO adsorption
the co-adsorption of other molecules which can slow down CO and favour its
decomposition cannot be ruled out as well as the formation of new defects at higher
temperatures Measurements with a variable temperature STM or with some other surface
analysis technique are planned in the future to confirm the proposed mechanism
When the surface was further annealed the nanoislands grew in size After 16 hours
heating very large nanoislands with diameters of sim35 nm and heights of sim20 nm became
visible (fig64 left) As a comparison when the same surface was heated without
nucleation centres deriving from contamination an irregular surface was obtained (fig64
right)
While the nanoisland density seemed to reach an asymptotic limit of 500 islandsmicrom2
(table 63) island size continued to grow with time (fig65) even when no further carbon
was supplied to the surface This indicates that even though the initial nucleating core of
the nanoisland is likely to be an SixC1-x alloy further growth is due to silicon atoms
diffusing from the crystal and being adsorbed by the growing island The final volume
VPalermo 71
occupied by the islands corresponds roughly to a 18 nm thick overall layer of removed
silicon
It is noteworthy that the clean sample after 16 hours annealing even though smoother than
the CO-contaminated one looked much more disordered with hardly any flat area visible
on nanometres scale In the presence of nanoislands however further annealing increases
island size but keeps part of the surface quite flat and ordered with flat areas visible
between the islands Thus the nanoislands can be imagined to act as ldquoimpurity sinksrdquo for
further contaminants approaching the surface yielding a greater but more ordered surface
roughening with respect to the uncontaminated silicon surface
Circular arrays of nanoislands
We used the voids described in the previous chapter formed during oxide layer
decomposition as nanoscopic masks to test the validity of the hypothesis outlined above
The small circular areas uncovered during oxide decomposition can act as ldquoskating rinksrdquo
where contaminants and diffusing silicon atoms can move freely over the surface while
the oxide layer all around and higher than the void area as well as being chemically
bonded to the surface silicon atoms will hinder surface diffusion Furthermore the oxide
passivating layer will at least in some cases prevent contaminating molecules from
adsorbing and decomposing on the surface In this way selective nanoisland growth inside
void areas can take place
The experimental procedure consisted of the following steps
1) An oxide-passivated sample was heated for 10 minutes at 800degC Oxide decomposition
began in correspondence with surface defects and spread laterally creating naked areas
of silicon
2) While the void enlarges the same defect that initiated oxide decomposition behaved as
an attractor for moving silicon atoms and nucleated the growth of a nanoisland at the
void centre
3) The void surface was exposed to contaminating molecules
4) The sample was then heated again at 800degC During the second heating the void
continued to enlarge and new nanoislands were created The freshly uncovered silicon
surface provided mobile silicon atoms to feed nanoisland growth
72 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
5) When the heating continued for long enough all the oxide was desorbed and a clean
silicon surface covered by circular groups of nanoislands was obtained
Figure 66 shows the different stages of the process
We thus heated an oxidized sample at 800degC for 10 minutes in UHV to create multiple
voids in the oxide layer After this we exposed the sample to the contaminating molecules
Then we further annealed the surface at the same temperature to promote island growth
The first time we observed nanoisland growth in the void area was by chance after heating
for a second time samples which had been stored for 2 weeks in vacuum Even at pressures
as low as 10-11 mbar some molecules will adsorb onto the surface slowly creating defects
and thus potential nanoisland nucleation centres It is possible to observe with STM that
the silicon crystal surface looks very clean just after a high temperature flash but even
after only a few days randomly adsorbed species will be observable on the surface
Unfortunately it is not possible to identify these species simply from in situ STM
measurements
Mass spectrometry measurements show that the residual gas contaminants in vacuum are
usually He Ar H2 CH4 CO CO2 and N2 [11] Some of these molecules such as
hydrogen will stick to the silicon surface at room temperature but will simply desorb
during annealing without nucleating nanoislands Others like CH4 or CO2 will not react
with the surface to generate nanoislands as demonstrated previously The best candidate
for contamination of samples stored in UHV seems thus carbon monoxide has the ability
321
4 5
Fig 66 Selective nanoisland growth within oxide voids See text for details
VPalermo 73
as shown before to nucleate nanoislands although co-adsorption mechanisms can not be
ruled out Fig 67 (left) shows the circular groups of nano-islands obtained after heating
void-covered samples stored for 2 weeks in UHV Small islands are observed in the former
void area The void itself has enlarged slightly uncovering a clean oxide-free area of the
silicon surface Apart from the central island created during the first period of heating the
surrounding ones have dimensions decreasing from the void border to the inner area of the
void
This confirms that most of the material needed for island growth comes from the silicon
atoms diffusing from the freshly uncovered areas around the original void The outer
nucleating centres are nearest to the silicon atom source and thus generate larger islands
For a faster process we directly exposed the void to a significant concentration of CO
Fig67 (centre) shows the islands grown after a 180 Langmuir exposure and subsequent
heating The islands are preferentially located around the original void perimeter the rest
of the original oxide-free surface remaining untouched This suggests that although the
whole surface is exposed to CO the gas only interacts chemically with the surface at the
SiSiO2 interface around the oxide-free void perimeter The SiSiO2 interface thus provides
preferential nucleation sites for nanoisland growth in the successive thermal annealing
step This recalls a similar phenomenon observed in the catalytic oxidation of CO on the
Pt(111) surface where CO molecules reacted with adsorbed O species only along the
perimeters of oxygen islands [10]
On repeating the CO exposure and thermal annealing steps a second circle of nano-islands
Fig67 Left circular area of nanoislands obtained from a sample stored 2 weeks in UHV 500x500 nmCentre nanoisland circle obtained after exposing the voids to 180 L of CO and heating for 10 min at800degC 400x400 nm Right two concentric circles obtained with further CO adsorption followed byheating 650x650 nm
74 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
was formed around the newly-expanded surface void perimeter (fig67 right) This
process can in principle be repeated again to obtain a series of concentric groups of
nanoislands the only limit depending on the minimal distance between neighbouring
expanding voids We have thus observed that contaminants adsorbed at low pressure over
long periods of time generate random island growth over the whole void area while CO
molecules adsorbed at relatively high pressures over short periods yield selective
nanoisland growth at the former void border Several theories can explain the different
results CH4 and CO2 as mentioned above did not nucleate islands when when the silicon
surface was exposed to them for short periods However the lower surface mobility of
these contaminants and the greater time available for reaction with the surface could be
the cause of nucleation of SiC clusters over the whole oxide-free void area and not only at
its border The coadsorption of water another UHV residual gas with CO on the naked
silicon areas could perhaps allow a surface reaction similar to that between CO and the
SiSiO2 interface during CO exposure Another explanation for the formation of these
random nucleation sites within the voids could be the coadsorption of residual carbon-
containing species present in the UHV system with hydrogen the major residual gas in
stainless steel UHV systems
The possibility of decorating silicon oxide borders with nanoislands using the high
mobility and selective decomposition of CO molecules suggests interesting applications
for nanofabrication processes that will be discussed below
Silicon nanowire creation on Si(100)
We performed most of the experiments described above on silicon (111) 7x7
reconstructed because this surface is easily prepared in UHV and above all because it is
isotropic and has no preferential directions for atom diffusion The 2x1 reconstruction of
silicon (100) even though it has a much simpler unit cell than Si(111) 7x7 shows a more
complicated morphology with lots of monoatomic and biatomic steps and dimer rows
parallel to each other The orientation of the dimer rows changes by 90deg on alternate
atomic layers Surface diffusion in this case is much more complicated because atoms will
experience the effects of the surface anisotropy and diffuse preferentially along or across
the dimer rows The diffusion energy of a silicon atom as example is 06 eV along a
dimer row and 085 eV across different dimer rows This while being a complication
VPalermo 75
offers interesting possibilities for the creation of ordered structures The monoatomic steps
and the dimer rows can for example act as templates for the formation of elongated
structures of composition similar to the islands described previously
On heating a Si(100) surface oxide decomposition and void growth takes place as
described for Si(111) In some cases the void shape reflects the surface symmetry
especially for very thin oxide layers and the voids often have a slightly squared shape
Nanoisland growth is also influenced by the substrate symmetry and the islands look
rectangular as irregular clusters surrounded by concentric patterns of monoatomic steps
The whole structure reminds vaguely a ldquozigguratrdquo a kind of stepped pyramid found in
Mesopotamia (fig68 inset)
Apart from nanoisland growth in some cases we observed spontaneous formation of
nanowire-like SiC structures on the Si(100) surface The wires were generated after
exposing a Si(100) surface in vacuum to traces of CO followed by sample annealing at
800degC for 15 minutes The wires exhibit lengths ranging from 10 to 100 nm and average
widths of ca 5 nm Although being randomly positioned on the surface they are perfectly
aligned along the crystal axes of the substrate (Fig68)
Several descending monoatomic steps can be observed around each nano-line and nano-
island Very often one of the line extremities coincided with an island
Nanoline
Nanoisland
Fig 68 A Si(100) surface covered by nanoislands and nanowires The monoatomic steps present on the surface are visible The two insets show a typical nanoisland and a nanowire obtained on this surface The typical ldquodimer-row ldquostripes are visible along the sides of the nanowire
76 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Couples of parallel lines starting from the same nano-island were also observed At high
resolution (fig 69) the lines showed a lamellar periodic structure with a periodicity of
around 4 nm and frequent defects Scanning Tunneling Spectroscopy measurements
showed no significant difference between the line and the surrounding silicon surface
The proposed formation mechanism for this kind of structure involves two stages in the
first one organic contaminating molecules adsorb and diffuse on the surface from the
vacuum environment stopping preferentially at surface singularities including steps and
defects
Several substances such as atomic carbon carbon monoxide ethylene and fullerene
decompose when adsorbed on silicon surfaces at temperatures above 500degC and create
localised defects on the surface inducing strain deformations and a metastable surface
reconstruction [12]
If these molecules remain attached to the steps they can diffuse rapidly along step border
until they meet a line and decompose In this way long lines of SiC defects parallel to the
surface steps can be produced In the second phase the clusters of silicon carbide can
-005
0
005
01
015
02
025
03
035
04
-3 -2 -1 0 1 2 3
V
dId
V lt
dId
Vgt
linesilicon
Current- Voltage behaviour measured with STS Fourier analysis of wire periodicity
asymp 4 Aring
High resolution image of a line
Fig 69 Top high resolution image of a nanowire showing the lamellar structure Fourier analysis (bottom left) shows that the lamellae periodicity is sim 04 nm bottom right the IV characteristics measured with the STM tip on the line and the silicon surface
VPalermo 77
25-600degC
Final linear structure
600-800degC
Contaminant Silicon
Fig 610 Schematic representation of nanowire formation mechanism
locally inhibit the spontaneous surface diffusion of silicon atoms acting as templates for
the reorganisation of surface steps (fig 610)
The final morphology shows disordered lines and islands each surrounded by a complex
pattern of silicon monatomic layers
At 800degC silicon atoms are highly mobile on the surface Due to atom diffusion the
surface reorganizes with a step-flow mechanism The presence of a fixed line of
contaminants blocks atom diffusion and step flow generating the monoatomic steps
around each wire (fig 611) SiC clusters act as lsquopinning sitesrsquo on silicon atom surface
diffusion and can be purposely used to direct the localised growth of nano-islands [13]
Although the possibility of using methanol and carbon monoxide as nano-island precursors
has been demonstrated the exact nature of the substances nucleating nano-wire growth is
still unknown
Further experiments will be needed to confirm these hypotheses as well as to identify the
contaminating species nucleating nanowire growth Carbon monoxide is a main candidate
for nanowire growth but the real nucleation process is likely to be a complex one
78 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Step flowblocked
Fig 611 Representation of step flow blocked by the presence of a nanowire 105x105nm
Bibliography
[1] Surface-Chemistry Of Silicon Waltenburg HN Yates JT Chemical Reviews 95 (5)
1589-1673 Jul-Aug 1995
[2] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[4] Growth Of Si On The Si(111) Surface Lanczycki Cj Kotlyar R Fu E Yang Yn
Williams Ed Das Sarma S Physical Review B 57 (20) 13132-13148 May 15 1998
[5] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
VPalermo 79
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990
[6] Nanoscale roughening of Si(001) by oxide desorption in ultrahigh vacuum Gray SM
Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology B 14 (2)
1043-1047 Mar-Apr 1996
[7] Step Motion On Crystal Surfaces Schwoebel Rl Journal Of Applied Physics 40 614-
618 Feb 1969
[8] Scanning Tunneling Microscopy Investigation At High Temperatures Of Islands And
Holes On Si(111)7x7 In Real Time Evidence For Diffusion-Limited Decay
Hildebrandt S Kraus A Kulla R Wilhelmi G Hanbucken M Neddermeyer H Surface
Science 486 (1-2) 24-32 Jul 1 2001
[9] Methanol Adsorption On Si(111)-(7x7) Investigated By Core-Line Photoemission And
Mass Spectrometry Of Photodesorbed Ions Carbone M Piancastelli Mn Zanoni R
Comtet G Dujardin G Hellner L Surface Science 370 (1) L179-L184 Jan 1 1997
[10] Atomic And Macroscopic Reaction Rates Of A Surface-Catalyzed Reaction
Wintterlin J Volkening S Janssens Tvw Zambelli T Ertl G Science 278 (5345)
1931-1934 Dec 12 1997
[11] Redhead PA Hobson JP Kornelsen EV The Physical Basis Of Ultrahigh
Vacuum Chapter 12 (Chapman amp Hall London 1968)
[12] The Si(001) C(4 X 4) Surface Reconstruction A Comprehensive Experimental
Study Norenberg H Briggs Gad Surface Science 430 (1-3) 154-164 Jun 21 1999
[13] Production Of Nanostructures Of Silicon On Silicon By Atomic Self-Organization
Observed By Scanning Tunneling Microscopy Jones D Palermo V Applied Physics
Letters 80 (4) 673-675 Jan 28 2002
80 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 81
7 Conclusions and proposals for future work It is very difficult to foresee the future of nanoscience and nanotechnology even over the
next few years exciting discoveries are continuously made opening up new possibilities
and progress in this sector is now one of the fastest and most varied So it is impossible to
predict whether computers will in the future be made with nanowires nanotubes single
electron devices or some kind of quantum-based transistor What does seem clear
however is that silicon devices can still further extend their levels of miniaturization to
remain competitive for at least the next ten years [1] During this period new emerging
technologies will reach scientific maturity and arrive at the production lines
Even when new technology prototypes do become available for full-scale production the
astronomical cost of changing from silicon to new technologies will be a major concern
Moreover a vast amount of extremely detailed knowledge is already available on all
aspects of silicon technology Thus the possibility of integrating innovative
nanotechnologies with standard CMOS silicon technology already seems an attractive
prospect [2]
For this and other reasons we have focussed our research efforts on surface phenomena on
silicon wafers which could be potentially useful for the development of silicon-compatible
nano-devices
Some important conclusions can be drawn from the work described in this thesis
bull Silicon surfaces of different crystal faces were studied at atomic resolution in ultra-
high-vacuum following transformations induced on the surface by chemical (etching)
and physico-chemical (organic contamination heating) treatments
82 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull The presence of surface contaminants induces at high temperature the growth of
nanometre-sized islands and lines on silicon
bull Final island densities and dimensions strongly depend upon the chemical nature of the
contaminant molecules and the treatments used Molecules interacting only weakly
with silicon will desorb upon heating whereas molecules which decompose easily on
Si will give widespread random nano-island nucleation Between these two extreme
cases complex behaviours of diffusion and subsequent decomposition are possible
bull Nanoscopic naked silicon voids were produced in oxide layers through carefully-
controlled thermal annealing It was found that void density and size depends upon
oxide purity surface contamination levels and annealing conditions
bull The growth of nano-islands and nano-voids was exploited simultaneously and the
voids used as nanoscopic masks to control the positions of growing islands Complex
silicon-on-silicon nano-structures were obtained with multi-step growth processes
bull Finally the use of macroscopic treatments and simple physical processes to produce
large quantities of nanometre-sized structures on silicon surfaces was demonstrated
These processes and the structures produced are of potential interest to the
microelectronics industry
We showed in previous chapters how it is possible to modify silicon surfaces using
chemical and physical methods The use of self-organization for the production of Si-on-Si
20 nm
Fig 71 Left SEM image of field emitter arrays made by conventional lithography [3] Center 3-D STM image of a self-organized void with a central nano-island Right Lateral schematic view of a field emitter device
VPalermo 83
nano-structures has the advantage with respect to other emerging techniques of being
silicon compatible and thus is a potential candidate for its implementation in the
production of new devices Furthermore this technique can create billions of strucures on a
wafer in only a few minutes being thus applicable to large-scale production
Even when ordered structures are obtained these methods still show a poor reproducibility
and control Chemically modified surfaces look very disordered at high magnification
with trenches and holes etched into the surface and progress has recently been made in
producing very small structures electrochemically [4] Oxide decomposition with
nanoisland growth in UHV can also yield ordered structures The void holes with
nanoislands located at their centres although quite variable in dimensions are more similar
to engineered devices than to a spontaneous random surface-roughening process In fact
there is a surprising similarity between the nanovoid-nanoisland structures and well-known
commercially available devices called field emitter arrays (FEA) These are small tips used
as electron guns in various kinds of displays and other devices Both the commercial
devices and our self-organized structures consist of a protruding tip at the centre of
microscopic holes on an insulating layer covering a semiconductor surface (see fig71)
Of course the order and size uniformity of commercial FEAs is much better than that
obtained with our process but it is encouraging to see that a self-organized structure
obtained with simple heating has a very similar surface topography and composition of
commercial working devices obtained after many expensive stages of optical lithography
deposition and etching
Nano-island density and size can to some extent be controlled and different
contaminants can be adsorbed onto the surface to modify the growth process In the future
with a better understanding of diffusion dynamics and growth mechanisms a multi-stage
fabrication process can be envisaged where purposely-designed molecules are adsorbed
onto silicon surfaces to diffuse organize in ordered arrays and favour surface
reorganization at the nanometre scale The same molecules could be used to deliver
dopants to obtain for example highly conducting nanostructures on otherwise undoped
highly resistive silicon The selective formation of nanostructures at the SiSiO2 border
can be used to scale down mesoscopic patterns to the nanometric level A silicon oxide
layer can be easily patterned on the surface with normal optical lithography and then
nanostructures can be built following the pattern edge In this way the production of very
small nanowire based devices may be possible (fig 72)
84 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
D
Gate
S
D
S
SiO2
SiO2
2 Heating nanowire creation
and oxide removal
3 Source-drain deposition by conventional lithography
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
4 Insulating layer (oxide) and gate (metal) deposition
Fig 72 Schematic drawing of the construction of a self-organized nanowire-based transistor
Surface diffusion and decomposition of molecules can be sensitive to even smaller surface
features The nanoscopic lines formed on Si(100) are very regular and neat if their
nucleation and growth mechanism could be controlled ordered networks of nanowires
could be made on silicon
VPalermo 85
The techniques developed here are still experimental ones the careful control and
reproducibility of each step is still an issue and further studies will be needed to clearly
understand the dynamics of formation of these structures Nonetheless the results obtained
so far open up the possibilities of producing novel devices such as for example nano-
transistors (see scheme in fig73 where the mechanism described in fig 72 is applied for
the fabrication of a device array) The possibility of using the border of a lithographic
pattern to draw nanolines can in principle allow the creation of nanodevices using normal
IC manufacturing techniques
The results obtained and their description in this thesis are not the first examples of the use
of self-organization for the production of ordered structures and will certainly not be the
last In our opinion the most important outcome of these findings is that it is possible to
use surface diffusion to overcome the inherent limits of lithographic techniques in
microdevice production processes and to force properly selected molecules to react with
the surface in a spatially differentiated manner at particular sites thereby creating ordered
series of nanostructures The role of the surface dynamics of the substrate atoms
themselves is extremely important in this process
Outstanding results have been and continue to be obtained in nanoscience and
nanotechnology research promising the emergence of new production technologies Sooner
or later those emerging technologies will have to face the issues of cost and compatibility
with the enormous investment made in existing technologies and processing facilities The
commercial production of nanostructures based on surface diffusion processes could in
principle be cheap simple and compatible with existing technology Moreover the
processes of diffusion in bulk silicon is well-known in the microelectronics industry which
has both the know-how and the equipment to better understand and exploit the surface
diffusion processes on silicon surfaces studied here
Extremely important and innovative results are often obtained by exploiting very simple
ideas and in our opinion physical processes such as surface diffusion used to create
complex nanostructures on silicon surfaces is a prime example
86 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
2 Heating nanowire creation and
oxide removal
3 Source-drain deposition by conventional lithography
4 Insulating layer (oxide) and gate (metal) deposition
Fig 73 Schematic macroscopic outline of the construction of a self-organized nanowire-based transistor
VPalermo 87
Bibliography
[1] P Gargini Intel Technology Coordinator ldquoFrom Microelectronics To
Nanotechnologiesrdquo Invited Campus Colloquia Lecture CNR Research Area Bologna
February 6th 2003
[2] Toward A Hybrid Micro-Nanoelectronics Cerofolini Gf Ferla G Journal Of
Nanoparticle Research 4 (3) 185-191 Jun 2002
[3] Recent Progress In Field Emitter Array Development For High Performance
Applications Temple D Materials Science amp Engineering R-Reports 24 (5) 185-239
Jan 25 1999
[4] Silicon Dioxide Micropillars For Sieving Fabricated By Macroporous Silicon-Based
Micromachining Izuo S Ohji H French Pj Tsutsumi K Kimata M Sensors And
Materials 14 (5) 239-251 2002 Electrochemical Etching In HF Solution For Silicon
Micromachining Barillaro G Nannini A Piotto M Sensors And Actuators A-Physical
102 (1-2) 195-201 Dec 1 2002
88 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 89
Acknowledgements
This work would not have been possible without the help and guidance of Dr Derek Jones
he introduced me to the use of STM and other techniques found the funding for this PhD
and helped me with the thousands of small and big problems I had to face during my
research I am grateful to Prof Alberto Ripamonti too for being the supervisor of this
thesis Thanks are also due to Dr Giancarlo Seconi director of ISOF-CNR where most of
this work was carried out Financial support from the Italian National Research Council
(CNR) is also gratefully acknowledged
My passion for surface science began when Fabio Biscarini showed me that it was possible
to really see the atoms and for this I will always thank him
During these years I had the luck to collaborate with many good scientists among those I
would like to cite Enrichetta Susi Massimo Cocchi and Anna Mazzone from CNR
Claudio Zannoni Anna Cavallini Daniela Cavalcoli and Antonio Castaldini from the
University of Bologna Sergio Pizzini and Maurizio Acciarri from the University of Milan
One of the most stimulating periods of my PhD was the one I spent at the Steacie Institute
for Molecular Sciences in Ottawa working with Robert Wolkow Dan Wayner Greg
Lopinski and Peter Kruse I thank them for discussing together many exciting scientific
ideas and trying to put some of them into practice
Special thanks go to my colleagues Paolo Samorigrave and Stefano De Cesari for long and
useful discussions on Science Life and their interactions often with the help of a good
pint
Finally I would like to thank Vassilia Gaetano Simone Claudia Silvia Letizia
Alessandro Angela and Sebastien even though not directly involved in this work they
have been over these three years a continuous source of happiness and support
Bologna March 2003
90 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
SCIENTIFIC PUBLICATIONS bull Lateral diffusion of titanium disilicide as a way to contacting
hybrid Si-organic nanostructures Palermo V Buchanan M Bezinger A Wolkow RA APPLIED PHYSICS LETTERS 2002 v81 p 3636 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v6 issue 20) bull Surface modifications in Si after Rapid Thermal Annealing Castaldini A Cavalcoli D Cavallini A Jones D Palermo V Susi E JOURNAL OF THE ELECTROCHEMICAL SOCIETY 2002 v 149 pG633 bull Production of nanostructures of silicon on silicon by atomic self-
organisation observed by scanning tunnelling microscopy Jones D Palermo V APPLIED PHYSICS LETTERS 2002 v 80 p 673 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v5 issue 5) bull Nucleation of nanostructures from surface defects on silicon Palermo V Jones D SOLID STATE PHENOMENA 2002 v 82-84 p 687 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo MATERIALS SCIENCE AND ENGINEERING B 2002 v88 (2-3) p220 bull Morphological changes of Si[100] surface after treatment with
concentrated and dilute HF Palermo V Jones D MATERIALS SCIENCE IN SEMICONDUCTOR PROCESSING 2001 v 4 p 437 bull Electrical and structural properties of processed silicon surfaces Susi E Cavallini A Castaldini A Cavalcoli D Jones D Palermo V ldquoRECENT RESEARCH DEVELOPMENTS IN VACUUM SCIENCE amp TECHNOLOGYrdquo 2001 v 3 p189 bull Numerical Solutions of the Stochastic Equations of Crystal
Growth Mazzone A M Palermo V INTERNATIONAL JOURNAL OF MODERN PHYSICS C 2000 v 11 Part 1 p195-204 bull Advances in silicon surface characterisation using light beam
injection techniques
VPalermo 91
Acciarri M Pizzini S Simone G Jones D Palermo V MATERIALS SCIENCE AND ENGINEERING B 2000 V73 (1-3) p 235 - 239 bull Abrupt orientational changes for liquid crystals adsorbed on a
graphite surface Palermo V Biscarini F Zannoni C PHYSICAL REVIEW -SERIES E- 1998 V 57 NUMBER 3A p R2519-R2522 Oral presentations bull The use of oxide desorption and surface diffusion for the
creation of silicon on silicon nanostructures Palermo V Jones D 1ST NATIONAL WORKSHOP ON CURRENT TRENDS IN NANOTECHNOLOGIES Catania (Italy) Feb 2002 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo V E-MRS Spring Meeting Strasbourg (France) June 2001 bull STM study of surface transformations on silicon during UHV
annealing Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999 bull Advances in silicon surface characterisation using light beam
injection technique Acciarri M Pizzini S Simone G Jones D Palermo V E_MRS SPRING MEETING June 1999 Posters bull Spontaneous nano-wire growth on silicon Palermo V Jones D SMARTON workshop Leuven (Belgium) October 2002 bull Ordered circles of nano-islands on silicon from CO adsorption Palermo V Jones D TRENDS IN NANOTECHNOLOGY 2002 Santiago de Compostela (Spain) September 2002 bull Production of nanostructures of silicon on silicon by atomic self-
organisation Palermo V Jones D EUROMAT Rimini (Italy) June 2001 bull Etching holes and anisotropic corrosion on silicon [100] Palermo V Jones D Susi E Asoli B SILICON WORKSHOP Genova (Italy) February 2001
92 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull Morphological and electrical characteristics of damaged silicon surfaces
Susi E Castaldini A Cavalcoli D Cavallini A Jones D Palermo V SILICON WORKSHOP Genova (Italy) February 2001 bull Effect of HF etching on the roughness of a silicon surface Palermo V Jones D NATIONAL CONFERENCE ON PHYSICS OF MATTER Genova (Italy) June 2000 bull Effect of Native Oxide Desorption upon the Surface Morphology of Si[100] by STM and LEED Palermo V Jones D SILICON WORKSHOP Genova (Italy) February 2000 bull Characterisation of silicon surfaces for microelectronics through STM measurements Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999
VPalermo 93
Contact Vincenzo Palermo ISOF- Institute for Organic Synthesis and Photoreactivity Via Gobetti 101 40129 Bologna ITALY Tel +39-051-6398336 Fax +39-051-6398349 Mail palermoisofcnrit
94 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Index
1 Introduction nanotechnology and the future of computers
2 Silicon surfaces
3 STM and other surface analysis techniques
4 Surface modification of silicon in liquid Nano-hole creation
5 Surface modification of silicon in vacuum void creation and oxide desorption
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and
nano-lines
7 Conclusions and proposals for future work
12 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Chapter 5 describes the growth in ultra-high vacuum (UHV) of nanoscopic voids and
islands on the silicon surface and the effect of surface oxide on this growth
Chapter 6 discusses the modification of silicon surfaces in UHV following the adsorption
of molecules and thermal heating to produce nanoislands and nanolines on silicon
The overall conclusions of our work are summarised in Chapter 7
Finally we will give some conclusions based on the results obtained and discuss possible
applications of the methods developed
Bibliography
[1] From wwwintelcom
[2] From wwwsematechorg
[3] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
[4] Molecules Get Wired Service RF Science 294 (5551) 2442-2443 Dec 21 2001
[5] Coulomb Blockade And The Kondo Effect In Single-Atom Transistors Park J
Pasupathy AN Goldsmith JI Chang C Yaish Y Petta JR Rinkoski M Sethna JP
Abruna HD Mceuen PL Ralph DC Nature 417 (6890) 722-725 Jun 13 2002 Kondo
Resonance In A Single-Molecule Transistor Liang WJ Shores MP Bockrath M Long
JR Park H Nature 417 (6890) 725-729 Jun 13 2002 Nanotechnology - Electronics
And The Single Atom De Franceschi S Kouwenhoven L Nature 417 (6890) 701-702
Jun 13 2002
[6] Smallest Carbon Nanotube Ajayan PM Ijima S Nature 358 (6381) 23-23 Jul 2 1992
[7] Carbon Nanotube Inter- And Intramolecular Logic Gates Derycke V Martel R
Appenzeller J Nano Letters 1 (9) 453-456 Sep 2001
[8] Logic Circuits With Carbon Nanotube Transistors Bachtold A Hadley P Nakanishi T
Dekker C Science 294 (5545) 1317-1320 Nov 2001
VPalermo 13
[9] Logic Gates And Computation From Assembled Nanowire Building Blocks Huang Y
Duan XF Cui Y Lauhon LJ Kim Kh Lieber CM Science 294 (5545) 1313-1317
Nov 9 2001
[10] Formation And Disappearance Of A Nanoscale Silver Cluster Realized By Solid
Electrochemical Reaction Terabe K Nakayama T Hasegawa T Aono M Journal Of
Applied Physics 91 (12) 10110-10114 Jun 15 2002
[11] Fabrication And Integration Of Nanostructures On Si Surfaces Ogino T Hibino H
Homma Y Kobayashi Y Prabhakaran K Sumitomo K Omi H Accounts Of Chemical
Research 32 (5) 447-454 May 1999
[12] Ultrafine And Well-Defined Patterns On Silicon Through Reaction Selectivity
Prabhakaran K Hibino H Ogino T Advanced Materials 14 (19) 1418-1421 Oct 2
2002
14 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 15
2 Silicon surfaces
The name silicon (silicio in Italian) comes from the latin word silex Amorphous silicon
was first isolated by Berzelius in 1824 by reaction of potassium with silicon tetrafluoride
Thirty years later the first crystalline silicon was prepared Silicon makes up 25 of
earthrsquos crust and is the second most abundant element after oxygen Elemental silicon is
not found in nature occurring as silicon oxide (sand quartz amethyst flint etc) or
silicates (asbestos clay mica etc) Perhaps no other element and its compounds has such
a wide range of uses Silicon compounds such as sand and clay are used in the building
industry as refractory materials for high-temperature applications and for enamels and
pottery Silica is the main component of glass silicon carbide is an important abrasive and
silicones are commonly used polymers and lubricants
Here the most interesting use of silicon of course is for the production of
microelectronics devices For this application silicon of high purity (999999) and of
high crystallinity is needed Table 21 lists some of the physical characteristics of silicon
High purity polycrystalline silicon is produced by the reaction of gaseous trichlorosilane
with hydrogen in a furnace Then to prepare a single-crystal of silicon the so-called
Czochralski method is commonly used
Polycrystalline silicon is melted in a quartz furnace at 1415degC in an argon atmosphere
Then a seed of single-crystal silicon is lowered into contact with the melt and slowly
pulled out In this way the crystal grows and a crystalline cylindrical ingot several metres
long is created from the initial seed
After cooling down the ingot is sliced into thin silicon wafers The wafer surfaces are
polished using a counter-rotating lapping machine in an Al2O3 slurry until the surface is
very flat and shiny ready for the lithographic processes
16 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another way to obtain single crystal silicon is the Floating Zone (FZ) method in which a
silicon cylinder is slowly passed through a heating ring The area inside the ring melts and
solidifies smoothly crystallising as it comes out of the ring yielding a single silicon crystal
Microelectronic devices are built on the silicon surface which is the surface of interest
here Unfortunately silicon surfaces are normally quite dirty and uneven at the atomic
scale Atmospheric oxygen and humidity react with silicon surfaces creating a thin layer of
oxide (called ldquonative oxiderdquo) which is usually irregular and full of defects Different kinds
of contaminants also adsorb onto the surface These are usually small organic molecules
and microscopic dust particles A clean surface on exposure to the atmosphere is
completely covered with gas molecules in less than 10-9 seconds If the pressure is
reduced letrsquos say to 10-6 mbar this time increases to 1 second This is the reason why to
study a clean surface we have to work in UHV at pressures below 10-10 mbar
The atoms in the silicon crystal have a diamond-like structure each atom having 4 bonds
in a tetrahedral sp3 arrangement with bond angles of 10947 degrees At the crystal
surface some atoms will have non-bonding orbitals ldquodanglingrdquo in the vacuum ie sp3
orbitals with a lone electron which are highly reactive These orbitals are known as
dangling bonds To minimize surface energy the surface will reorganize by decreasing
the number of dangling bonds
Table 21 Physical data of silicon [1]
Atomic Weight 2809 Lattice constant (A) 543095 Crystal structure Face-centered cubic
(diamond) Melting point 1415 degC
Density (gcm3) 2328 Boiling point 2355degC Atomscm3 50E22 Minority carrier
lifetime (s) 25E-3
Dielectric Constant 119 Specific heat (Jg degC)
07
Breakdown field (Vcm)
~3E5 Thermal conductivity (Wcm degC)
15
Electron affinity x(V)
405 Vapour pressure (Pa) 1 at 1650degC 1E-6 at 900deg C
Energy gap (eV) at 300K
112 Reactivity Inert to acids Attacked by halogens and alkaline
solutions Intrinsic carrier
conc (cm-3) 145E10 Oxidation states +4 -4
Intrinsic Debye Length (microm)
24 Energy of a Si-Si bond (eV)
232
Intrinsic resistivity (Ω-cm)
23E5
VPalermo 17
Dangling bond densities and positions and thus the type of surface reconstruction will
depend upon crystal orientation as well as the temperature and kinetics of the system
Fig 21 shows a drawing of the main faces of a silicon crystal The angle α between any
(11n) face and the (100) face can be calculated from 2cos 2 += nnα The angle
between any (11n) face and the (111) face can be calculated from
)2(3)2(cos 2 ++= nnα
The chemistry and physics of the faces are very different a brief description will be given
for the most important orientations
Si (100)
On the (100) surface each atom has two Si-Si bonds connecting it to the bulk and two
dangling bonds pointing outward Surface energy is reduced by the dimerisation of the
surface atoms through overlap interaction of one dangling bond per atom forming rows of
dimers aligned along the (110) direction This is the well-known ldquo2x1rdquo reconstruction of
this silicon surface The symmetric dimers would make the silicon surface metallic but to
reduce surface stress the dimers tend to buckle and the surface is thus semiconductive It
took several years to understand that the dimers are buckled because at room temperature
Fig 21 Schematic view of the principal orientations of a silicon surface Surface atoms are white bulk atoms are black dangling bonds are gray [2]
18 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
they shift easily from one buckling direction to the other thus appearing symmetric under
STM observation Fig 22 shows an STM image of the 2x1 reconstructed surface
Even almost perfect (100) surfaces have a certain number of monoatomic steps and the
dimer rows on atomic layers are aligned at 90deg to those on adjacent layers Dimer rows are
thus perpendicular or parallel to the step When the dimers on the upper side of the step
are parallel to the step the step is called SA if they are perpendicular the step is called SB
Because of this symmetry SA step edges will be smoother compared to the more broken
fragmented edges of the SB steps
A common defect on the Si(100) surface is the presence of nickel contamination which
appear as missing dimers in STM images This type of contamination is so critical that
even if the silicon sample is only briefly brought into contact with stainless steel tools
(tweezers for example) the 2x1 reconstruction of the surface can be blocked
Silicon atoms can diffuse easily over the silicon surface as monomers and dimers
especially at elevated temperatures The anisotropy due to the 2x1 reconstruction causes a
difference in the diffusion energies of adsorbates over the surface Diffusion of these
silicon species along dimer rows for example will be much easier A list of diffusion
energies for monomers and dimers is given below [2]
Diffusion on Si(100) 2x1 Ed (eV) Monomers along dimer rows 06 Monomers across dimer rows 085 Monomer formation energy 18 Dimer along dimer rows 11 Dimer across dimer rows 15 Dimer formation energy 26 Dimer binding energy 076 Vacancies along dimer rows 17 Vacancies across dimer rows 19
SA
SB
Fig 22 STM image of a 2x1reconstructed silicon surfaceshowing the dimer rows and steps40x35 nm Nickel-induced defectsare visible as dark spots SA and SBsteps are indicated
VPalermo 19
So the diffusion energy for both monomers and dimers is nearly 40 greater if they have
to cross a dimer row This difference reduces to sim10 for vacancy diffusion
Si(113)
The (113) surface can be imaged as a sequence of alternating (100) and (111)-like
structures with two and one dangling bonds on alternate atoms respectively Interest in the
(113) surface is scientific as it has been used to study the energetics of the (100) and (111)
surfaces as well as for surface adsorption experiments
Si(100) surfaces can easily develop (113) facets
Si(111)
This surface besides being the first one imaged with STM with atomic resolution is one of
the most studied because it is the best cleavage face of silicon and because it shows one
of the most complex and elegant reconstructions in surface science
All Si-Si bonds in the silicon crystal are perpendicular to a (111) plane so this face will
have the lowest number of dangling bonds created per unit area In fact each Si atom on a
(111) surface shows a single dangling bond oriented perpendicular to the surface and
bonded to three back atoms These three bonds for each surface atom account for the great
chemical and physical stability of the Si(111) surface Surface energy is 009 eV Aring-2
compared to 015 eV Aring-2 for Si(100)
For energy minimization this surface reconstructs forming a huge 7x7 lattice cell
containing 102 atoms described by the Dimer-Adatom-Stacking fault model (DAS) For a
detailed description of cell structure see fig 23
The cell described by this model is very complex being composed of three kinds of atoms
adatoms rest atoms and corner hole atoms Furthermore a subsurface stacking fault is
present in one half of the cell making the two halves of the unit cell look different under
STM (Fig 24) It took 26 years of research to completely understand the exact structure of
the 7x7 reconstruction
Cleaving a silicon crystal along a (111) plane produces a metastable 2x1 reconstruction
the 7x7 reconstruction is easily obtained by flashing at high temperature in UHV At T gt
20 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
830degC a disordered 1x1 phase covers the surface Cooling down to 800degC leads to the
formation of the 7x7 phase If the cooling process is too rapid small 7x7 domains nucleate
and a disordered 1x1 phase is preserved between domain boundaries
Si(110)
Even though as mentioned before the (111) plane is the favoured cleaving plane of
silicon thin (100) commercial wafers will not break along this plane because the angle
between (100) and (111) is too far from 90deg (see table 22) Instead they will break along
the (110) plane because it is perpendicular to the (100) surface Each surface atom on
Si(110) has a Si-Si bond pointing downward one dangling bond pointing outward and two
Si-Si bonds parallel to the surface in a zig-zag pattern (see fig 21) Cleaved (110)
surfaces are disordered but upon annealing at high temperatures an ordered complex 16x2
Fig23 Scheme of the 7x7 DAS model [2] In each unit cell there are 9 dimers 12 adatoms and a stacking layer fault The force driving this complex reconstruction is the minimization of dangling bonddensity The DAS model shows the lowest number of dangling bonds (19) of all possiblereconstructions 12 dangling bonds are at the adatoms 6 at the rest atoms and 1 at the corner hole atom This surface is metallic
VPalermo 21
reconstruction takes place The surface appears as a series of long ridges and valleys
parallel to each other Eventually tilted facets of orientation (17 15 1) can form on this
surface The adsorption of Ge atoms on this surface leads to the formation of self-
assembled nanowires [3]
Table 22 Angles in degrees between different silicon faces [2]
Orientation
113
110
111
100
100
2524
9000
5474
0
111
2950
3526
0
110
6476
0
113
0
Fig 24 STM image of a Si(111)surface with 7x7 reconstruction A unitcell with its adatoms is highlightedImage size 13x13 nm
22 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] Weast RC Handbook Of Chemistry And Physics (Chemical Rubber Co Cleveland
1972)
[2] Dabrowski J Mussig H Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] The Structure Of Clean And SiGe-Covered Si(110) Surfaces Butz R Luth H Surface
Science 365 (3) 807-816 Oct 1 1996
VPalermo 23
3 STM and other surface analysis
techniques
Scanning Tunneling Microscopy
Since the invention of the optical microscope at the end of the 16th century the possibility
of examining surfaces at higher and higher magnification has fascinated mankind
Development of the technique continued and towards the end of the 19th century optical
microscopes were as good as todays standard instruments The physical limits of the
wavelength of visible light (350-800nm) had been reached
In the 1920s de Broglie showed that electrons can behave like waves and the use of these
particles for imaging with much higher resolution soon followed Atomic resolution using
this technique is only possible in the transmission mode with extremely carefully prepared
samples
In 1982 using the peculiar properties of piezoelectric materials Binnig and Rohrer brought
a metallic tip very very close to a silicon surface and scanned it across an extremly small
area (fig31) The tunneling of electrons from the tip into the sample or vice versa allowed
them to obtain a local density of electronic states (LDOS) map of that surface Although
theory (which treated the extreme point of the tip as a sphere) then excluded the possibility
of atomic resolution following a tip crash into the surface Binnig and Rohrer began to
observe the LDOS with atomic resolution For this discovery and their following work
they were awarded the Nobel Prize in Physics in 1986
24 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The basic principle of STM is very simple A metallic tip is scanned over a surface without
making ohmic contact and a tunneling current passes between the tip and the surface An
electronic circuit keeps this current constant by raising and lowering the tip during the
scan In this way recording the tip height at each point a three-dimensional image of local
density of electronic states (LDOS) of the surface can be obtained To explain the
extremely high resolution attainable by this simple technique quantum theory is needed
According to classical physics the current will flow between sample and tip only if they
are in physical contact If there is a vacuum gap between the two the electrons will simply
remain confined for example within the surface without the possibility of passing into the
tip
In quantum physics however the electrons have a certain probability of passing
(tunneling) across the gap appearing on the other side of the gap in this way reaching the
tip It can be shown that the probability of an electron tunneling through a gap of thickness
z is
kzep 22)0( minusprop ψ h
φmk
2= (1)
where ψ(0) is the electron wavefunction at the surface-gap border m is the electron
mass=91x10-28 g and φ is the work function of the metal (ie the energy required to
remove an electron from that material For silicon it is 48 eV) The tunneling current thus
decays exponentially with z and is extremely sensitive to topographical imperfections
present on the scanned surface A rough formula giving the current as a function of z is [1]
zFS eEVI φρ 0251)( minusprop
Fig 31 Binnig and Rohrer with the first STM Image from IBM [2]
VPalermo 25
where ρs(EF) is the local density of states at the Fermi level on the given surface For
example the formula predicts that for silicon an increase in tip-surface distance of 1 Aring
will give a 95 decrease in tunneling current
This huge dependence of tunneling current upon the distance allows detection even of the
sub-nanometre changes in height given by the single atoms of which the surface is
composed and thus to resolve them in the LDOS images Of course this description of the
tunneling process is oversimplified and for a more accurate one the electronic states of
the tip of the sample and their interaction have to be taken into account Fig 32 shows a
schematic representation of the interaction between tip and sample orbitals
The exponential decay of current with distance also yields high lateral resolution If the tip
is approximated as a sphere of radius R and the current passing at the minimum tip-sample
distance is I0 then the current passing at a lateral distance x from this point will be
Rxk
eII 22
0
2minus
=
Assuming a tip radius of 100 nm the current is concentrated in an area sim15 nm wide at tip
apex
Very sharp tips with even smaller curvature radii can be produced in several ways Simple
Fig32 Schematic view of tip-sampleorbitals interaction a) no interaction b)equilibrium c) sample positive d) tippositive [1]
26 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
tungsten wires mechanically cut are capable of obtaining atomic resolution on graphite in
air but for more disordered and rougher samples sharper and more reproducible tips are
needed
STM tips are mostly made by electrochemically etching a W or Pt-Ir wire The tips we
used were prepared using methods based on the work of Fotino [3]
A tungsten wire 038 nm diameter is immersed in a KOH solution(06M) with a thicker
tungsten wire used as a counterelectrode The cathodic and anodic reactions involved in the
etching are
Cathode 6H2O + 6 e- rarr 3H2(g) + 6 OH-
Anode W(s) + 8 OH- rarr WO42- + 4H2O + 6 e-
A potential of 3V ac is applied to the tungsten and the wire is immersed in the solution
until a constant current of sim100 mA is established The etching rapidly removes metal
shaping the wire end as a sharp tip When the potential reaches 11 V the coarse tip etching
is finished The wire is then removed from the solution carefully inserted into an
insulating plastic tube leaving only the tip exposed and re-immersed in the solution with
the tip pointing upwards A more gentle etching is thus made to reduce tip radius Usually
5 minutes etching at 07 V ac is used In this configuration very small hydrogen bubbles
formed on the tip sides sliding upwards with a ldquohoningrdquo effect on the tip
This procedure yields extremely sharp and reproducible tips at the microscopic level
After the etching the tip is thoroughly rinsed in ultrapure water then dipped into
concentrated HF to remove surface oxides and hydroxides [4] The tip is dried with
nitrogen inserted into the UHV system and degassed overnight at sim150degC
The possibility of measuring sub-nanometric distances would be useless without being able
to control tip movement over such a minute scale To scan the tip over the surface a
piezoelectric scanner is used Piezos are usually made of an alloy of PbZrO3 and PbTiO3 a
material which contracts or expands when a voltage is applied to it The Omicron
instrument used in our laboratory has three such piezo scanners for xy and z tip motion
allowing one to scan the tip over the surface with sub-Aringngstrom precision (fig 33)
To isolate the instrument from ambient vibrations the whole STM stage is suspended upon
four springs which eliminate all frequencies above 1 Hz and surrounded by a crown of
VPalermo 27
copper wings and fixed magnets Parasitic currents generated by the magnets into the
copper wings contrast every movement of the stage and efficiently block stage vibrations
The STM can be used not only to explore surface topography but to measure the IV
characteristics of single atoms or molecules on the surface (Scanning Tunneling
Spectroscopy or STS) Furthermore it can be used to modify the surface with voltage
bursts digging into it or delicately moving single molecules or atoms over a surface [5] It
can be used in vacuum in air and with proper lateral isolation of the tip even in liquids
A major drawback of STM is that it works only on conducting and semiconducting
surfaces and thus cannot be used on many surfaces of biological and chemical interest
Another instrument more suitable for these and other applications is the Atomic Force
Microscope (AFM)
Atomic Force Microscopy
AFM was invented in 1986 by Binnig Quate and Gerber after calculating the possibility of
building a cantilever with a force constant of the same order of magnitude as that of a
chemical bond
In AFM a tip mounted on a microscopic cantilever (usually made of Si3N4 fabricated with
optical lithography) is brought close to a surface When the tip touches the surface the
cantilever is very slightly deflected upwards The movement is measured by observing the
Y-PIEZO
SILICON SAMPLE
Fig 33 A picture of theSTM used for theexperiments The tripodpiezo scanner is shown
Z-PIEZO
X-PIEZO
TIP
28 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
deviation of a laser beam hitting the upper face of the cantilever Fig 34 provides a
schematic view of the principle of AFM
The typical force constant of the cantilever varies from 00006 to 2 Nm the typical
resonance frequency is 3 to 120 kHz The AFM tip can apply a force on the sample of up
to 10-9 N The AFM can be used on conductive or insulating surfaces in vacuum air or
liquids Furthermore the tip can be modified to sense electrostatic potentials (electric force
microscopy) or magnetic fields (magnetic force microscopy) it can even be functionalized
with complex molecules such as proteins to interact with biological surfaces
A drawback of the AFM is that the force it exerts can damage the surface under
observation especially if the sample is soft (as in the case of cell membranes for
example) This problem can be overcome using the instrument in tapping mode (where the
tip does not move laterally during its brief contact with the surface) or in non-contact mode
in which the tip oscillates above the surface during the scan and the changes in its
frequency due to interaction with surface are monitored The shifts in the oscillating
frequency of the cantilever due to tip-sample interaction are then used for imaging the
surface In this mode interaction of the tip with the surface is minimal and soft samples
can be imaged
STM and AFM are the main techniques used for this work A brief description of other
techniques used occasionally is given below
Fig 34 Scheme of an atomic force microscope
VPalermo 29
Low Energy Electron Diffraction (LEED)
Electrons with energies in the 20-500 eV range are diffracted by a crystalline surface the
diffraction peaks are visualized on a fluorescent screen This technique probes the long
range order of the surface up to a depth of several nanometres
X-Ray Photoelectron Spectroscopy (XPS)
XPS allows both qualitative and quantitative chemical analysis of the elements present on
or near the sample surface
An X-ray source is used to photoionize the atoms on a surface and produce photoelectrons
By measuring the kinetic energy of the photoelectrons the binding energy of the electronic
levels can be calculated This energy will depend on the chemical environment of the
surface atoms
Although the soft X-rays used penetrate to a depth of ~2000Aring the sampling depth of the
technique is determined by the mean free path of the photoelectrons which allows their
escape from only the first 10-100Aring
Secondary Ion Mass Spectroscopy (SIMS)
High and low energy ions (primary ions) are used to bombard a sample and remove surface
atoms and ions The ionic fragments removed (secondary ions) are then analysed by a mass
spectrometer The surface can be consumed during the measurement and profiles obtained
giving concentrations of the materials composing the sample at different depths (depth
profiles)
A popular variant of SIMS is TOF-SIMS In this technique the secondary ion masses are
measured by a time-of-flight (TOF) measurement The secondary ions generated by the
bombarding primary ions are accelerated to a constant kinetic energy and then move
through a field-free space before they reach the detector where their intensity is measured
as a function of flight time Since ions with different masses have different velocities at a
30 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
given kinetic energy the measured flight times of the ions can easily be converted to their
masses The static nature of this latter technique allows mass spectroscopy surface analysis
with minimal damage to the surface
Bibliography
[1] Chen CJ Introduction To Scanning Tunneling Microscopy (Oxford University Press
Oxford 1993)
[2] From wwwibmcom
[3] Tip Sharpening By Normal And Reverse Electrochemical Etching Fotino M Review
Of Scientific Instruments 64 (1) 159-167 Jan 1993
[4] A Convenient Method For Removing Surface Oxides From Tungsten STM Tips
Hockett LA Creager SE Review Of Scientific Instruments 64 (1) 263-264 Jan 1993
[5] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
VPalermo 31
4 Surface modification of silicon in liquid
Nano-hole creation
Liquid treatments of silicon wafers are very common in the integrated circuit (IC)
manufacturing industry They are used to clean and improve surface uniformity to create
and etch protective oxide layers and to remove photo-resist layers
Crystalline silicon with its native oxide layer is very stable and is resistant to many acids
It is easily attacked by hydrofluoric acid (HF) and alkaline solutions
The thin (~2 nm) passivating layer of native oxide (SiO2) is formed on exposure to the
atmosphere This surface layer contains many defects and contaminants so it is usually
chemically stripped and substituted with a better chemically-formed protective oxide
The most common silicon cleaning procedure is the RCA method named after the Radio
Corporation of America [1] It consists of two steps in the first one the surface is treated
with a hot alkaline solution (H2OH2O2NH4OH 411) to remove particles from the
surface following this a hot acidic solution (H2OH2O2HCl 411) is used to remove
metal contamination Other well-known cleaning methods are IMEC (a sequence of
cleaning steps in H2OO3 and dilute HF) or the Pirana etch (a hot 41 mixture of H2SO4
H2O2)
32 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The standard RCA clean removes surface contaminants etches the native oxide and
oxidizes the silicon surface leaving a uniform layer of silicon oxide which better protects
the surface from further contamination
Etching with fluorine-based solutions
Hydrofluoric acid is one of the most common reagents used in the treatment of silicon
wafers both in the research field and in industrial processes A rapid dip in dilute HF is the
simplest way to remove the native oxide from Si(100) and leaves the surface passivated by
a layer of Si-H bonds Because of the low polarization of Si-H bonds the Si-H layer is
stable even for several days protecting the surface from contamination It has often been
assumed that this short etch does not significantly change the surface morphology of the
silicon substrate[2] even though a prolonged dip in dilute HF leads to surface roughening
[3]
Although dilute HF roughens the Si(100) surface at the atomic scale [45] immersion in
concentrated HF (49) etches the surface oxide without attacking the Si surface
uncovering in this way the buried SiSiO2 interface The final counter-intuitive result is
that dilute HF etches the silicon while concentrated HF leaves the crystalline silicon
untouched [3]
Etching Si with fluorine-containing solutions at different concentrations and pH can
produce different morphologies from rough surfaces to flat nearly ideal Si-H terminated
surfaces
Hessel et al and Higashi et al demonstrated in 1991 that very flat Si(111) surfaces can be
obtained using 40 NH4F while etching with HF always yields rough surfaces The
surface becomes smoother because the etchant rapidly attacks Si atoms at step borders
thus removing surface kinks and irregularities in a step-flow mechanism [6 7] Later on
even smoother and more perfect surfaces were obtained by removing oxygen from the
solution after it was discovered that oxygen dissolved in 40 NH4F initiates the formation
of triangular etch pits It was not possible to obtain flat surfaces by etching Si(100) with
ammonium fluoride solutions which leads to the formation of small 2x1 dimer-row
reconstructed (100) terraces together with (111) facets [8]
This difference is caused by the different hydride terminations prevailing on the (100) and
(111) faces While the ideal Si(111)-H surface is monohydride terminated the more
VPalermo 33
reactive dihydrides predominate on the Si(100)-H surface making it more vulnerable to
etching The etching reaction is thus strongly anisotropic etching (100) facets faster than
(111) thus producing (111) microfaceting on Si(100) crystals
A more uniform Si(100) surface can however be prepared by etching at low pH with an
HFHCl mixture [9] or by using very dilute HF solutions and ultrapure water with low
dissolved oxygen and carbon contents [10]
Electrochemical etching can also be used applying anodic or cathodic bias to the silicon
to obtain different morphologies [11] by varying the potential isotropic or anisotropic
etching is observed The aforementioned results show that despite the simplicity of the
reactants fluoride etching of silicon is quite a complex reaction
Fig 41 Chemical etching of silicon
HO+H
H HO
H
H
H
FSi
Si
SiSi
+H2O
-OH-
+F-
H2O H H
Si H
F OH
H Si
Si Si
-H2
+H2O
H
H
H
FSi
Si
SiSi
H
H
H
OHSi
Si
SiSi
H
H
H
H Si
Si
Si Si
Etching mechanism of silicon
HF rapidly dissolves the SiO2 passivating layer on silicon leaving the surface almost
completely hydrogenated [12] After this two different types of reactions etch the silicon
simultaneously one chemical and the other electrochemical [13] The overall etching
mechanism can be schematized in two stages (see Figure 41)
i) Si-H bonds are substituted by Si-F or Si-OH bonds creating a partial charge on the
surface silicon atom and polarizing its Si-Si backbonds
ii) these polarised backbonds are then more easily attacked by HF or H2O After
rupture of the Si-Si bond the atom is removed leaving behind new Si-H
terminations and the reaction can start again
34 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
These reactions take place although at different rates on both Si(100) and Si(111)
Stage i) is usually the rate-determining step of the reaction and the stability of the Si-H
bonds depends upon the pH the concentration of nucleophilic species in solution and an
eventual potential applied to the crystal
For pH gt5 as in the case of concentrated NH4F solutions reaction begins with attack by
water to give Si-OH (step ArarrC) The -OH group is rapidly substituted by fluorine with
polarization of the underlying Si-Si bonds These bonds are then easily attacked by water
the silicon atom being released into solution as HSiF(OH)2 The Si-OH rarr Si-F substitution
is not fundamental for the reaction and etching can proceed even for Si-OH terminated
atoms but XPS measurements showed the presence of a certain number of Si-F bonds
remaining Furthermore fluorine seems to have a catalytic effect on Si-H substitution as
indicated by the dependence of the etch rate upon the F- concentration at least for pH
values between 4 and 8
Si-F bonds can be easily removed by a water rinse In the case of strongly alkaline
solutions (pH=14) OH- groups act directly as nucleophiles and no fluorine is needed to
catalyze Si-H bond rupture
At pH lt4 almost no free F- ions are present in solution and the etch rate is very small at
pH lt2 all etching reactions are very slow and this explains the stability of Si crystals in
concentrated (50 ww) HF solutions
Si Si
Si Si
H
H
H
H Si
Si
Si Si
H
H
H Si
Si
SiSi
H
H
H
OH-H+ -e
(F-)
-H+
-e
+H2O hellipas in fig1
Fig 42 Electrochemical etching of silicon by H2O
The electrochemical etching of silicon involves electron transfer from the surface atoms to
the valence or conduction band of the crystal (according to the type of doping of the
silicon) Several studies have been carried out at different electric potentials In the case of
p-silicon or n-silicon at anodic potentials where positive charges (h+) are available in the
crystal a silicon-centred radical can be created by capture of a hole from the bulk (fig
42) The F- ion does not participate in the reaction but is thought to contribute through
VPalermo 35
electrostatic interactions by lowering the energy of the interaction step After the
formation of the Si-OH group the reaction proceeds as shown in the scheme of fig 41
Matsumura et al [4] proposed that not only water but HF2- molecules also play a major role
in electrochemical etching of silicon leaving on the surface Si-F terminated bonds which
can be immediately attacked in an autocatalytic process (fig 43)
In the electrochemical reactions described above an external potential is applied to the
silicon crystal The chemical and electrochemical reactions in any case take place
simultaneously most of the time with the chemical path predominating at high pH Even
when no external potential is applied to the silicon partial electrochemical reactions can
take place at different ldquocathodicrdquo and ldquoanodicrdquo sites on the surface with an internal charge
exchange which ensures neutrality [11] This macroscopic silicon etching and hydrogen
bubble formation can lead to visually observable patterns on the surface when Si(100) is
immersed in ammonium fluoride even without applying a potential
Si Si
Si Si
H
H
H
F Si
Si
Si SiH
F F-H+ -2e
+HF2-
H
HF
SiSi
SiSi + F
F F
FH
H
H
F -H+ -2e
+HF2-
Fig 43 Autocatalytic electrochemical etching of silicon by HF2-
Inhomogeneities on silicon surfaces caused by electrochemical reactions and charge
transfer have been studied extensively because they are of fundamental importance in the
formation of porous silicon
Pore formation on silicon
When Si(100) or Si(111) are etched under anodic bias in fluorine-based solutions
microscopic pores form on their surface Several different morphologies of pores have
been observed with pore diameters ranging from 10 nm to several microns with depths of
several microns [14] Pore shape is very variable too ranging from ordered straight pores
to chaotic networks of branched pores (fig 44) Porous silicon has been known since the
fifties but it was only in 1990 that interest in this material increased following the
discovery that porous silicon layers were able to emit bright red light This led to a large
amount of research and now different classes of micropores can be reproducibly created
36 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Fig 44 Different types of Silicon micropores From ref [14]
mostly for optical and micromachining applications However there is still no unified
theory able to explain the nucleation and growth mechanism of all the different kinds of
pores
We will give a short description of some of these theories for more detail see Parkhutik et
al [15]
One model explains pore nucleation on the basis of physical processes such as hole
positive charge migration ion transport to the surface and small perturbations on the
silicon surface modelled as Fourier components The system is shown to be unstable and
some spatial frequencies that lead to pore nucleation evolve from the etching process
A second model focusses on stationary pore growth without explaining the nucleation
stage According to this model silicon dissolves preferentially at pore edges because h+
charges are attracted by the stronger electric field present at these edges
A third class of models explains pore growth as a Diffusion Limited Aggregation (DLA)
process where the random walk of h+ charge carriers through the depleted layer present at
the silicon-liquid interface controls pore shape
Finally the model by Carstensen Cristophersen and Foll [16] proposes that areas of the
surface of some characteristic size LCO are etched by synchronized ldquocurrent burstsrdquo in the
flow of h+ charges These bursts dissolve silicon through cyclic stages of surface oxidation
oxide removal and hydrogen passivation Areas where a burst has recently taken place are
less passivated and thus more likely to be etched again in this way the pore bottom
continues to dissolve while the pore walls are passivated and are thus less favourable areas
towards current bursts
VPalermo 37
EXPERIMENTAL RESULTS
In the following sections we will show some experimental results obtained from STM and
AFM measurements of fluorine-treated Si(100) surfaces In the first part the results of
mild etching using concentrated and dilute HF solutions at low pH are presented In the
second part the results of etching at high pH using ammonium fluoride are presented and
the mechanism of pore formation discussed
Etching of Si(100) in dilute and concentrated HF
Samples were cut from different areas of an 8-inch diameter p-doped silicon(100) wafer
(10 Ω-cm) supplied by MEMC Electronic Materials Each series of STM measurements
was carried out over at least six different areas on at least two identical samples Low
Electron Energy Diffraction (LEED) was used to check the surface cleanliness of the
samples before STM measurements
Table 41 summarizes the different treatments of each sample After etching with
electronic grade HF each sample underwent a final rinse in Ultra-Pure Water (UPW
resistivity gt18 MΩ-cm) Both the HF and the UPW were allowed to flow continuously
over the sample surface Some samples were not etched with HF at all but just washed with
UPW to observe the morphology of the native oxide layer (~2 nm thick) covering the
surface All of these processes were carried out under nitrogen to limit reoxidation and the
samples were then introduced from the nitrogen atmosphere directly into the vacuum
chamber and degassed overnight at sim150degC before LEED and STM measurements
STM images were obtained from each sample using the same measurement parameters
(sample bias 4 V feedback current 1 nA scan speed 800 nm s-1) These parameters
38 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
provided a satisfactory level of reproducibility for all the samples Measurements were
made over an area of 500x500 nm (image size 500x500 pixel) Slope correction was
carried out by subtracting row-wise and column-wise fitted slopes from the entire image
which gave better results than the simple subtraction of a fitted plane especially for the
rougher samples Following slope correction the rms roughness
sum minus=xy
hyxhN
22 ))((1σ
and the 2-D Fourier transform
)(22
2
)(4
)( vyuxi
xyeyxhvuF +sum∆
= π
π
were calculated for each image where N2 is the number of pixels composing the image
h(xy) is the surface height at each point ∆ is the distance between points h is the mean
height and u v are the spatial frequencies The radial power spectrum PS(f) of the STM image data is obtained from the angular
average of the squared Fourier transform with f 2 = u2 + v2
Fig 45 shows the STM images obtained from the various samples Sample A still covered
with its native oxide layer shows an irregular surface with RMS roughness of ~05 nm
(see Table 41) Observing the sample with LEED gave no diffraction pattern even at
relatively high incident electron energies because of the surface oxide coverage After 1
min etching in dilute HF (sample B) the morphology is similar to the original one though
Table 41 Sample treatments average RMS roughness and integrated area of power spectra
Sample Treatment RMS roughness (nm)
PS area (f lt 01 nm-1)
PS area (f gt 01 nm-1)
A Rapid dip in water 053 plusmn 013 366 032
B 1 min in HF 5 + 10 min in water 051 plusmn 008 272 034
C 30 min in HF 5 + 10 min in water 062 plusmn 008 1064 038
D 5 sec in HF 49 + 10 min in water 042 plusmn 004 183 022
VPalermo 39
Fig 45 STM images of each group of samples showing the topography of the silicon surface A) noetching original oxide surface B) after 1 min etching in dilute HF C) after 30 min etching in diluteHF D) after dipping in concentrated HF Grey scale indicates height of the surface from lower (black) to higher (white) The images are 250x250 nm ie representative portions of the images usedfor the roughness measurement and PSD analysis
some of the larger features have disappeared and the image quality is better maybe due to
improved tunnelling due to the cleaner surface The RMS roughness is comparable to that
of the original surface Clear diffraction patterns are visible using LEED though at quite
high energies (200 eV) After prolonged etching (sample C) the RMS roughness increases
to 062 nm and a long-range corrugation is visible on the surface even if the LEED pattern
is good
The samples dipped in concentrated HF (D) reveal the bare SiSiO2 interface which has a
disordered aspect and protrusions over a wide range of dimensions The quality of the
STM images of sample D is very good probably due to the cleanliness of the surface
40 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 E -0 2
1 E -0 1
1 E + 0 0
1 E + 0 1
1 E + 0 2
1 E + 0 3
0 0 0 0 0 1 0 1 0 1 0 0f (1 nm )
nm^4
A a s re c e iv e d
B e tc h e d 1 min
C e tc h e d 3 0 min
D e tc h e d HF 4 9
Fig 46 Log-Log plot of the averaged power spectra of the STM images for all the samples
which gives a more stable tunnelling junction The LEED pattern is excellent showing
clear diffraction peaks at energies as low as 37 eV comparable to that obtained after high
temperature cleaning in UHV
Fig 46 shows the power spectra of the samples The high frequency and low frequency
areas of the power specturm are considered separately Table 41 shows for each sample
together with the roughness the integrated area of the power spectrum for the high and low
frequency part
We first examine the differences between the samples in the low frequency part of the
spectrum (lt01 nm-1) In this spatial range the short dip in concentrated or dilute HF
(samples BD) seems to lower the roughness of the sample removing some of the larger
features of the native oxide surface visible in Fig 45a and thus decreasing the
corresponding part of the power spectrum Sample C however etched with HF for 30 min
shows an increase in roughness peaking at 0012 nm-1 (~85 nm corresponding to the
typical dimensions of the corrugations visible in Fig 45c) On this scale sample C has a
PS density of 435 nm4 compared to 101 nm4 for sample A
Samples A B and C have the same PS in the high frequency range showing that the fine
structure of the surface is not changed by the HF etching Sample D on the other hand
shows a decrease of roughness for frequencies gt 015 nm-1 This effect could be due to the
VPalermo 41
improved surface cleanliness after etching with concentrated HF which would give a more
stable STM junction thus reducing the high frequency noise in the image
Fig 45 and the analysis of the power spectrum of each sample shows that a rapid dip in
HF removes the native oxide but does not lead to major changes in the morphology of the
Si surface its only effect being the removal of some of the larger features present on the
original surface Prolonged etching on the other hand increases the RMS roughness of the
surface
Etching of Si(100) in concentrated ammonium fluoride and nano-hole creation
Two different types of commercial p-doped Si(100) wafers (2Ω-cm and 10Ω-cm) from
MEMC were used Several different samples of 10x5 mm were immersed for 10 minutes
in 40 electronic grade NH4F solution under agitation Previous works used low
temperatures or anodic potentials applied to the silicon to avoid gaseous hydrogen
production and to obtain a uniform surface but in our experiment we worked at room
temperature to check the influence of hydrogen bubbles on surface morphology During
the etching the stirring was sufficient to provide a uniform concentration of reagents over
the whole sample surface but not to mechanically remove the hydrogen bubbles from the
silicon surface
After the etching each sample was rinsed in ultra-pure water to remove any etching
residues and observed with STM AFM and optical microscopy The AFM measurements
were made in air while for STM measurements the samples were rapidly dried with
nitrogen and inserted into the vacuum system to avoid surface reoxidation After insertion
into the vacuum surface cleanliness was checked with LEED and the surface morphology
observed by STM Parameters for STM measurements were sample bias +4 V 1 nA
current 16 Hz scan rate The images obtained were stable and reproducible over several
days Scan parameters for AFM were 20 nN force and 1 Hz scan rate
Some of the samples were cleaned with an RCA standard clean [1] before NH4F etching to
check the influence of possible surface contaminants on the final results Eight different
samples were prepared and more than sixty STM images of the samples were taken at
different points of the various samples
42 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
(110)
(110)
Fig47 abc) STMtopographic images of differentetching morphologies Eachimage is 500x500 nm Z-ranges are 10 10 and 18 nmrespectively d) STM image of a bridgecreated by etching of the lowerlayers of silicon (black arrow)Image is 250x250x6 nm
After ~2 min of immersion in the solution hydrogen bubbles become visible on the sample
surface The production is slow and the bubbles are quite stable on the sample without
detaching Thus some areas of the surface are masked from the liquid etching action
STM observations (fig 47) show that at the nanometer scale the surface is unevenly
covered with holes of radii ranging from 10 to 200 nm with depths of 2-4 nm These holes
have a wide range of different shapes and distributions In most cases the surface was
covered with a uniform distribution of round-shaped holes (fig47a) indicating isotropic
etching The dimension and the density of the holes changed greatly from sample to
sample and even over the surface of a single sample In some cases the etching was
anisotropic yielding nearly square holes and layered structures as shown in fig 47b
Square holes have been previously observed in cases where the etching speed in the (110)
direction is significantly smaller than in the (100) direction [17]
Over these areas (fig 47b and especially 47c) it is clearly visible how once the surface
had been attacked the reaction continued to preferentially remove atoms at step
irregularities (kink atoms) straightening step edges Eventually the exposed underlying
silicon was also attacked and further holes created inside the previously etched larger
ones It was not possible to detect monatomic steps on this kind of surface The smallest
step height observed was ~15 nm corresponding to several atomic layers In the image
shown in Figure 47a the etching was not strong and created only anisotropic holes on the
surface In fig 47b and c the stronger etch proceeded laterally for several tens of
VPalermo 43
nanometres leaving straight steps several tens of nanometers long and roughly rectangular
holes as expected given the structure of the (100) crystal face In some cases a
significative underetch is observed and the formation of suspended bridges and tunnels
can be deduced from the STM images (fig 47d)
The formation of branched pores and suspended structures has been attributed during pore
formation to diffusion limited aggregation effects where the h+ charge carriers necessary
for silicon etching have a higher probability of reacting at pore bottoms than reaching the
upper part of the silicon surface In the case of very deep pores quantum wire effects have
been invoked to explain the pore growth mechanism [15] In our case though the pores
formed were very shallow the underetch depth being only a few nanometres on pores of
sim100 nm width Thus more than diffusion effects the main contribution to the
underetching process must come from anisotropic etching and some kind of autocatalytic
reaction path analogous to the one described by Matsumura et al [4] with some areas of
the silicon surface hydrogenated and thus less vulnerable to etching
Pre-treatment with RCA cleaning has no effect on the final morphology and this seems to
exclude pore nucleation being caused by presence of metallic or organic surface
contaminants
The morphology and the distribution density of the pits was quite uniform over
microscopic areas of the sample but changes were observed over the millimetre scale This
suggests that etching intensity is influenced by some large-scale parameter
Large-area measurements made with AFM or with an optical microscope (fig 48)
showed that the inhomogeneity of surface etching can be correlated with the masking
action of the bubbles While the fluoride dissolved the silicon hydrogen bubbles formed
by the reaction covered some areas of the surface thus blocking the etching over that area
generating macroscopic steps at the bubble-liquid border As the reaction proceeded more
hydrogen accumulated and the bubble diameter increased producing in this way a circular
pattern of steps The increase in bubble diameter was not continuous with time otherwise a
uniform surface slope gradient would have been obtained The formation of this circular
ldquoetching staircaserdquo indicates that the bubble growth was stepwise the bubble accumulating
more and more hydrogen without enlarging across the surface until it relaxed increasing
its diameter stepwise and covering more silicon The circular structures in fig 48a are not
co-axial and their asymmetry could derive from physical processes due to stirring or
irregularities on the surface
44 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
a b Fig 48 a) optical micrograph of etching patterns on Si(100) created by NH4F 12x09 mm b) AFM image of the circles border xy range is 40x24 microm z-range is 30 nm
The step structure was not destroyed by the etching even after the bubble detached from
the surface but on the contrary the etching process seemed to be influenced by the
presence of the step
Observing in detail a series of steps (fig 48b) a quite deep trench is visible at the base of
each step A close-up image of a step and the corresponding line profile of the trench is
shown in fig 49 The trench is sim5 nm deep with respect to the lower surface compared to
a step height of 22 nm
A similar structure has been recently obtained with electrochemical etching of p-type
Si(100) in 4 HF [18] in which a ldquocurrent burstrdquo etching model previously described
was assumed for silicon dissolution In that experiment the trench was created at the
border of silicon nitride masks and began to grow after a nucleation stage Preferential
trench etching was along the (110) direction and the trench growth was explained as an
effect of mechanical stress induced by the nitride mask and of electric field enhanced
dissolution which depended upon an external applied potential
While it is clear that in our system the gas bubbles have a masking effect similar to a
classical solid nitride mask it is unlikely that hydrogen present on the surface can induce a
significant stress in the silicon lattice as in the case of a nitride mask Furthermore no
external field was applied to drive preferential etching at the trench site
It has been proposed [19] that the cathodic and anodic part of the etching reaction
(hydrogen production and silicon oxidative etching respectively) take place at different
points on the surface with a net charge transfer between the different areas In this case
the highest reaction rates will correspond to the silicon area surrounding the bubble border
where a high number of positive charges will be available for the reaction Furthermore a
sharp trench extending into the silicon crystal will be a preferential electrostatic attractor
VPalermo 45
Fig 49 AFM image of the etched surface showing a step created on the surface by bubblemasking A stronger etching action is visible on the right side of the step as well as theprotected area on the upper side of the step (indicated by the arrows) Image is 10x10 micromz-range is 30 nm The profile on the right is taken from the central area of the image
for the h+ charge carriers coming from other ldquocathodic areasrdquo of the sample either from
other regions on the surface or from the back of the silicon chip [18]
In the areas where the hydrogen bubble had detached and the surface was exposed to the
etching the reaction was not uniform in the neighbourhood of the steps It is possible to
observe (fig 48b and 49) an area on the upper side of the step where less or even no
etching at all seems to have taken place as if the step was able to protect the surface from
etching While etching on the lower side with trench formation can be attributed to the
presence of the bubble the surface on the upper step side can be etched only after bubble
detachment so no masking effect can account for this result However a further
preferential attraction of h+ charge carriers from the already formed trench can be
hypothised electrochemically shielding the surrounding area from further etching If this is
true the shielding effect would be very strong with a relatively shallow 5 nm-deep trench
protecting an area of sim1 microm parallel to the step
To summarise the etching of Si(100) in NH4F is a complex process in which different
reaction paths both chemical and electrochemical co-exist Hydrogen bubbles formed by
the reaction act as a mask on the surface and create etching paths and inhomogeneous
etching of the surface Different kinds of pores are observed on the surface and in some
cases the anisotropy of the process is so strong as to give square-shaped holes and
underetching
46 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The diffusion of h+ charge carriers in the crystal is one of the main rate-determining steps
of the reaction and leads to the formation of a deep trench immediately outside the bubble
perimeter These trenches act as charge collectors and reduce the etching of the upper step
surface in the proximity of the steps
Bibliography
[1] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[2] Spectroscopic Ellipsometry Studies Of HF Treated Si (100) Surfaces Yao H Woollam
Ja Alterovitz SA Applied Physics Letters 62 (25) 3324-3326 Jun 21 1993 Influence
Of HF-H2O2 Treatment On Si(100) And Si(111) Surfaces Graf D Bauermayer S
Schnegg A Journal Of Applied Physics 74 (3) 1679-1683 Aug 1 1993 Kinetics Of
Oxidation On Hydrogen-Terminated Si(100) And (111) Surfaces Stored In Air Miura
T Niwano M Shoji D Miyamoto N Journal Of Applied Physics 79 (8) 4373-4380
Part 1 Apr 15 1996
[3] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995 Structure Of The Stepped SiSiO2 Interface After Thermal-Oxidation -
Investigations With Scanning Tunneling Microscopy And Spot-Profile Analysis Of
Low-Energy Electron-Diffraction Pietsch GJ Kohler U Jusko O Henzler M Hahn
PO Applied Physics Letters 60 (11) 1321-1323 Mar 16 1992
[4] Enhanced Etching Rate Of Silicon In Fluoride Containing Solutions At pH 64
Matsumura M Fukidome H Journal Of The Electrochemical Society 143 (8) 2683-
2686 Aug 1996
[5] A Study Comparing Measurements Of Roughness Of Silicon And SiO2 Surfaces And
Interfaces Using Scanning Probe Microscopy And Neutron Reflectivity Crossley A
Sofield CJ Goff JP Lake ACI Hutchings MT Menelle A Journal Of Non-Crystalline
Solids 187 221-226 Jul 1995
VPalermo 47
[6] Step-Flow Mechanism Versus Pit Corrosion - Scanning-Tunneling Microscopy
Observations On Wet Etching Of Si(111) By Hf Solutions Hessel HE Feltz A Reiter
M Memmert U Behm RJ Chemical Physics Letters 186 (2-3) 275-280 Nov 8 1991
[7] Comparison Of Si(111) Surfaces Prepared Using Aqueous-Solutions Of NH4F Versus
HF Higashi GS Becker RS Chabal YJ Becker AJ Applied Physics Letters 58 (15)
1656-1658 Apr 15 1991
[8] Wet Chemical Etching Of Si(100) Surfaces In Concentrated NH4F Solution -
Formation Of (2x1)H Reconstructed Si(100) Terraces Versus (111) Faceting Neuwald
U Hessel HE Feltz A Memmert U Behm RJ Surface Science 296 (1) L8-L14 Oct
10 1993
[9] Ideal Hydrogen Termination Of Si(001) Surface By Wet-Chemical Preparation Morita
Y Tokumoto H Applied Physics Letters 67 (18) 2654-2656 Oct 30 1995
[10] Atomic Structures Of Hydrogen-Terminated Si(001) Surfaces After Wet Cleaning
By Scanning Tunneling Microscopy Endo K Arima K Kataoka T Oshikane Y Inoue
H Mori Y Applied Physics Letters 73 (13) 1853-1855 Sep 28 1998
[11] On The Potential-Dependent Etching Of Si(111) In Aqueous NH4F Solution
Houbertz R Memmert U Behm RJ Surface Science 396 (1-3) 198-211 Jan 20 1998
[12] Etching Process Of SiO2 By HF Molecules Hoshino T Nishioka Y Journal Of
Chemical Physics 111 (5) 2109-2114 Aug 1 1999
[13] Etching Mechanism And Atomic-Structure Of H-Si(111) Surfaces Prepared In
NH4F Allongue P Kieling V Gerischer H Electrochimica Acta 40 (10) 1353-1360
Jul 1995
[14] Pore Formation Mechanisms For The Si-HF System Carstensen J Christophersen
M Foll H Materials Science And Engineering B-Solid State Materials For Advanced
Technology 69 23-28 Sp Iss Si Jan 19 2000
[15] Porous Silicon - Mechanisms Of Growth And Applications Parkhutik V Solid-
State Electronics 43 (6) 1121-1141 Jun 1999
[16] Formation And Application Of Porous Silicon Foll H Christophersen M
Carstensen J Hasse G Materials Science amp Engineering R-Reports 39 (4) 93-141 Nov
1 2002
[17] Pore Morphology And The Mechanism Of Pore Formation In N-Type Silicon
Searson PC Macaulay JM Ross FM Journal Of Applied Physics 72 (1) 253-258 Jul 1
1992
48 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
[18] Deep Electrochemical Trench Etching With Organic Hydrofluoric Electrolytes
Christophersen M Merz P Quenzer J Carstensen J Foll H Sensors And Actuators A-
Physical 88 (3) 241-246 Mar 5 2001
[19] Morphology Of Anodically Etched Si(111) Surfaces - A Structural Comparison Of
NH4F Versus HF Etching Houbertz R Memmert U Behm RJ Journal Of Vacuum
Science amp Technology B 12 (6) 3145-3148 Nov-Dec 1994
VPalermo 49
5 Surface modification of silicon in vacuum void creation and oxide desorption
The main reason for the huge success of silicon in the microelectronics industry is not due
to its superior properties as a semiconductor Other materials for example germanium
have better qualities such as higher mobility of charge carriers and lower noise levels
which would allow the construction of faster and higher performance devices
The widespread use of silicon however is mainly due to the outstanding characteristics of
its oxide Silicon dioxide (SiO2) is a very good electrical insulator easy to form
chemically and thermally stable and is compatible with lithographic and metal deposition
processes Germanium oxide on the contrary is too reactive to be used
Even the use of Si(100) substrates for nearly all microelectronic devices is dictated by
oxide quality The (111) face of silicon crystal can be easily cleaved and flattened and
almost atomically perfect surfaces can be obtained with simple chemical procedures (as
described above) But the density of interfacial defects is highest for Si(111)-SiO2
interfaces and lowest for Si(100)-SiO2 ones so microchips will continue to be fabricated
on Si(100) wafers
SiO2 (silica) is present in 95 of the earthrsquos minerals in different allotropic forms such as
quartz tridymite and cristobalite In the bulk each silicon atom is bonded to four oxygens
in a Si-O-Si tri-dimensional network Si-O bonds are 016 nm long and form an angle
ranging from 120deg to 150deg
Three typical intrinsic defects are present in SiO2 The so-called Ersquo centres are oxygen
vacancies with a hole localised on a silicon atom with only three Si-O bonds
50 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
O3Si +SiO3 Whereas the PR (peroxy radical) defects are holes trapped by a charged
peroxy moiety with a O3Si-O-O+ SiO3 structure The NBOHC (non-bridging oxygen hole
centres) derive from water or hydrogen contamination and are schematized as O3Si- O- H-
O-SiO3
The atomic structure of the Si-SiO2 interface varies enormously Local domains resembling
the tridymite and the cristobalite structure of silica are present but it seems that only 10
of the interface is ordered [1] Far from the interface the SiO2 bulk is completely
disordered The passage from bulk Si to stoichiometric SiO2 passes through a non-
stoichiometric SiOx layer 07 nm thick
When a clean silicon surface is exposed to atmospheric oxygen a thin sim2 nm thick layer
of native oxide forms spontaneously which is usually removed and substituted with
thicker better quality oxide layers before further processing
Silicon is usually oxidised by thermal annealing at temperatures between 800deg and 1100deg
in an atmosphere of pure O2 with some water eventually added to increase oxidation speed
Thermal oxides made in pure oxygen (dry oxides) grow more slowly than oxides produced
in an oxygen-water atmosphere (wet oxides) but are usually of better quality
According to the Deal-Groove formula the time t needed to grow an oxide of thickness X
is given by [1]
1
212minus
minusminus
+=
ABXBXt α
where the constant B and BA decrease exponentially with temperature as
minus
kTEexp
with activation energies for dry oxidation of EB =123 and EBA =20 eV respectively
EB is related to the diffusion of oxygen in silicon while the value of EBA is interpreted as
the energy required to break a Si-Si bond The exponent α is 1 for wet oxidation and 0 for
oxidation at high temperatures and low oxygen pressures It has intermediate values for dry
oxidation This formula does not work well for low values of X at the initial stages of
oxidation and usually empirical corrections are used
An interesting characteristic of silicon is that at high temperature and in vacuum oxygen
can actually etch the silicon crystal giving gaseous products and the oxidized layer present
on the crystal becomes unstable (fig 51) [2]
VPalermo 51
1E-10
1E-09
1E-08
1E-07
1E-06
1E-05
1E-04
1E-03
1E-02
1E-01
0607080911112
1000T (1K)
P (T
orr)
SiO2 + Si rarr 2SiO(g) Oxide decomposition
Si+O2 rarr SiO2(s) oxide formation Fig51 phase diagram of the
oxygen-silicon system
Silicon oxidation apart from the initial nucleation stages at the monolayer level proceeds
uniformly over the whole surface with a planar reaction front moving from the surface
into the bulk
If heated under low oxygen partial pressure (vacuum or inert atmosphere) SiO2 is known to
decompose following the reaction
SiO2 + Si rarr 2SiOuarr (1)
The reaction begins with nucleation at defect points on the SiSiO2 interface and proceeds
in a spatially inhomogeneous manner with the formation of large voids on the oxide
surface [3]
Several studies have been made on the dynamics of void growth both on thick [4] and thin
[5] layers of SiO2 The process has been used to decorate otherwise unobservable defects at
the SiSiO2 interface [6] or to grow nanoislands of silicon on the void surface [7] It has
been suggested that the defects acting as nucleation centres could be metallic contaminants
present on the native surface which aggregate and catalyze SiO2 decomposition [8]
After oxide desorption the silicon surface is very rough In particular on Si(100) square
islands are observed several nanometres high which act as lsquopinning sitesrsquo on the motion of
monatomic steps on the Si surface It has been proposed that these islands can be composed
of silicon carbide coming from the organic contaminants present on the surface [9] or of
oxide clusters during partial reoxidation [10]
In the following section some experimental results obtained studying voids and nanoisland
growth are presented
52 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nanoisland growth on silicon
We used two different samples the first rougher than the second (i) p-type Czochralski Si
(100) with a RMS roughness of 05 nm shown in fig52a and (ii) p-type epitaxial Si(100)
with a RMS roughness of 019 nm shown in fig 52b The roughness was measured with
STM on the lsquoas receivedrsquo samples on different areas of 500x500 nm Every sample was
covered by a layer of native oxide ~2 nm thick
Some of the samples (AC) were introduced into the UHV system without any cleaning
others (BD) were dipped in concentrated HF (49) to remove the surface oxide without
etching the silicon [11]
After degassing the samples were heated resistively increasing the temperature slowly to
keep the pressure within the 10-10 mbar range during heating Each sample was held at
900˚C for 30 min For some of the samples the heating was stopped at lower temperatures
to obtain incomplete oxide desorption and surface roughening thus enabling observation
of the different steps of the process Table 51 summarizes the different treatments for each
group of samples
Fig52 Original surfaces of Czochralski (a) and epitaxial (b) silicon Every image is 500x500 nm
VPalermo 53
Table 51 Summary of different treatements
Silicon type Oxide removed
Heating Islands density (microm-2)
Av Island Volume (nm3)
A Czochralski No 30rsquo at 900˚C 24x103 353
B Czochralski Yes ldquo 20x104 71
C Epitaxial No ldquo 28x102 2450
D Epitaxial Yes ldquo 23x104 63
Fig53 ab) Surface roughening on group A samples after heating 30 min at 800deg and900deg respectively in presence of an oxide layer Image size 500x500 nm cd) Surface roughening on group B samples after heating 30 min at 700deg and 900degrespectively after removing the oxide layer with HF Image size 250x250 nm
Fig 52a is the native oxide surface of sample A quite disordered and irregular This
surface was stable when heated to 700˚C while at higher temperatures the oxide began to
desorb Fig 53a was taken after heating the sample at 800˚C for 30 min
54 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The oxide began to desorb in correspondence with defect points creating oxide-free
surface voids which enlarge radially uncovering the silicon surface Some surface silicon
atoms are removed through reaction (1) Mobile silicon atoms moving randomly over the
surface aggregate forming a nanocrystal at the initial defect point surrounded by a stable
flat oxide-free surface with some steps The initial SiSiO2 interface as observed after
oxide etching with HF [12] is similar to the one shown in fig 52a and thus the surface
observed in the void area is a completely new one coming from surface reorganization
The reaction continues leading to a coalescence of the voids the final result of the process
is shown in fig 53b with protruding islands aligned along the main crystal axes and flat
areas with some steps LEED measurements confirmed that this surface is crystalline
silicon Atomic resolution was obtained on the flat areas of the surface showing the
typical 2x1 reconstruction of Si(100) Figures 53c and 53d show the evolution of the
roughening on sample B after chemically removing the native oxide layer before
introducing the sample into UHV As mentioned above the initial morphology is similar to
that shown in fig 52a though the oxide has been removed as confirmed by LEED
measurements Fig 53c shows the surface after heating at 700˚ for 30 min Without the
oxide layer the surface began to reorganize at lower temperatures and the small irregular
grains of the original surface begin to grow and become more rectangular After heating to
900˚C (fig 53d) the surface is flat with small rectangular islands The shape of the islands
is similar to that shown in fig 53b but the lateral island dimension is one order of
magnitude smaller
To assess the importance of the initial surface on the process the same treatments were
repeated starting from flat epitaxial silicon (samples CD) The initial epitaxial surface is
smooth with periodic steps due to a small miscut angle (fig52b) On heating the epitaxial
silicon without etching (sample C) relatively large islands are obtained similar to the ones
shown in fig 53b where the ldquostep-pinningrdquo effect of the islands is evident The average
dimension of sample C islands was sim2500 nm3 while only 353 nm3 for sample A Island
density decreased from 24000 to 280 microm-2 This result is maybe due to a lower
concentration of defects in the oxide layer of the epitaxial silicon which allows the
expanding void to grow much larger before coalescence with neighbouring voids the
nanostructures within them thus reaching greater dimensions To verify this hypothesis we
applied the same treatment to clean not-so clean and purposely contaminated samples To
check that the defects inducing void nucleation and nanoisland growth derive from carbon
VPalermo 55
contamination we used electronic-grade methanol to contaminate the silicon surface
Figure 54 shows the results of this comparison
State-of-the-art cleaned samples obtained with multi-step RCA cleaning [13] showed the
development of very few voids the oxide desorption leaving large smooth areas of oxide-
and nanostructure-free silicon as shown in Fig 54a where the presence of residual oxide
can be seen Samples cleaned with a simpler one-stage Pirana cleaning solution showed a
higher number of islands per unit area (Fig 54b) and samples purposely contaminated
with methanol (Fig 54c) had the highest density of island nucleation of all three samples
When the native oxide had been chemically removed from the epitaxial silicon before
heating (sample D) the island size and density is similar to the standard as-received
silicon case (sample B) Thus starting from a flat substrate the process of surface
roughening and island creation was the same even when the oxide layer had been
previously removed
From these results some important indications can be drawn The first is that oxide
desorption temperature depends upon the contamination levels of the oxide layers (fig 54)
with desorption at lower temperatures for contaminated surfaces
Second once the oxide layer is removed the silicon atoms become mobile on the surface
at relatively low temperatures (~700˚C) Their initial configuration is disordered and has a
high surface energy so they crystallize in the small islands shown in fig53d reducing
their surface area and lowering the energy of the system When the oxide is present the
surface cannot reorganize at 700˚C and is stable At 800˚C the oxide begins to desorb in
correspondence with defect points on the surface As the void area enlarges a small
cba Fig 54 The effect of contamination on nano-island production Gradient-filtered STM images of (a)sample cleaned two times with RCA 400x400 nm (b) Sample cleaned with a Pirana solution 500x500nm (c) sample contaminated with MeOH before insertion into the UHV system 500x500 nm Annealingtime is 40 min at 800degC for all the samples
56 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
crystallite grows at the centre of the voids Further oxide desorption uncovers larger areas
freeing more silicon atoms which feed the initial island making it bigger
Fig 55 shows the scaled size distribution of the islands for each group of samples Every
distribution has been obtained using several images of different samples for each group
The size distribution of all the samples follows an exponential decay and the decay seems
the same for all the samples though the average volume of the islands differs by more than
one order of magnitude (see Table 51 for details)
The point defects which catalyze void nucleation can be metallic contaminants present on
the initial oxide surface [8] or organic contaminants which at high temperatures can
form SiC nanocrystals [14] We used Scanning Tunneling Spectroscopy (STS) to look for
differences between the islands and the surrounding flat silicon surface but no difference
was found Furthermore the total island volume per surface unit is very high (more than
104 nm3microm2) and it seems unlikely that such a huge volume could consist of surface
contaminants Islands of pure silicon on silicon have been grown without evidence of
surface contamination[14]
To summarise the formation of nano-sized islands on silicon through surface diffusion
was studied At high temperatures the oxide layer covering the surface decomposes non
uniformly and circular voids of clean silicon are created The presence of the oxide layer
blocks surface silicon atomic motion and surface reorganization except within the voids
causing the growth of islands more than 10 nm high and 30 nm wide on the silicon surface
According to STM and LEED measurements we can say that the islands are mostly
Fig 55 Size distribution (scaled) for the samples described in table 1 The line is a reference for the eye
VPalermo 57
composed of silicon with traces of other substances such as carbon Final island densities
and dimensions depend upon the initial purity of the oxide layer
On the other hand if the oxide is removed chemically before heating the island growth
process is not so localised and takes place simultaneously over the whole surface yielding
smaller and more numerous islands
58 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[2] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[3] High-Temperature SiO2 Decomposition At The Sio2Si Interface Tromp R Rubloff
GW Balk P Legoues FK Physical Review Letters 55 2332-2335 Nov 1985
[4] Defect Microchemistry At The SiO2Si Interface Rubloff GW Hofmann K Liher M
Young DR Physical Review Letters 582379-2382 Jun 1987 Defect Formation In
Thermal SiO2 By High Temperature Annealing Hofmann K Rubloff GW Mccorkle
RA Applied Physics Letters 49 1525-1527 Dec 1986 Kinetics Of High-Temperature
Thermal Decomposition Of SiO2 On Si(100) Liher M Lewis JE Rubloff GW Journal
Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 5 1559-1562 Aug
1987
[5] Thermal-Decomposition Of Very Thin Oxide Layers On Si(111) Kobayashi Y Sugii
K Journal Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 10 (4)
2308-2313 Part 3 Jul-Aug 1992 Scanning Tunneling Microscope Study On Mid-
Desorption Stages Of Native Oxides On Si(111) Kobayashi Y Sugii K Journal Of
Vacuum Science amp Technology B 9 (2) 748-751 Part 2 Mar-Apr 1991 Controlled
Growth Of SiO2 Tunnel Barrier And Crystalline Si Quantum Wells For Si Resonant
Tunneling Diodes Wei Y Wallace RM Seabaugh AC Journal Of Applied Physics 81
(9) 6415-6424 May 1 1997
[6] Defect Formation In SiO2Si(100) By Metal Diffusion And Reaction Liher M
Dallaporta H Lewis Je Appl Phys Lett 53 589-591 Aug 1988 SiO2 Film
Decomposition Reaction Initiated By Carbon Impurities Located At A Si- SiO2
Interface Raider Si Herd Sr Walkup Re Applied Physics Letters 59 (19) 2424-2426
Nov 4 1991
[7] Nanometer-Scale Si Selective Epitaxial Growth On Si(001) Surfaces Using The
Thermal Decomposition Of Ultrathin Oxide Films Fujita K Watanabe H Ichikawa M
Applied Physics Letters 70 (21) 2807-2809 May 26 1997 Pyramidal Si Nanocrystals
VPalermo 59
With A Quasiequilibrium Shape Selectively Grown On Si(001) Windows In Ultrathin
SiO2 Films Shibata M Nitta Y Fujita K Ichikawa M Physical Review B 61 (11)
7499-7504 Mar 15 2000
[8] Stacking-Fault-Induced Defect Creation In SiO2 On Si(100) Liher M Bronner Gb
Lewis Je Appl Phys Lett 52 1982-1985 May 1988
[9] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990 AFM And XPS Characterization Of The
Si(111) Surface After Thermal-Treatment Lamontagne B Guay D Roy D Sporken R
Caudano R Applied Surface Science 90 (4) 481-487 Dec 1995
[10] Scanning-Tunneling-Microscopy Study Of Oxide Nucleation And Oxidation-
Induced Roughening At Elevated-Temperatures On The Si(001)-(2x1) Surface Seiple
JV Pelz JP Physical Review Letters 73 (7) 999-1002 Aug 15 1994 Evolution Of
Atomic-Scale Roughening On Si(001)-(2x1) Surfaces Resulting From High-
Temperature Oxidation Seiple JV Pelz JP Journal Of Vacuum Science amp Technology
A-Vacuum Surfaces And Films 13 (3) 772-776 Part 1 May-Jun 1995
[11] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995
[12] Morphological Changes Of The Si [100] Surface After Treatment With
Concentrated And Diluted HF Palermo V Jones D Materials Science In
Semiconductor Processing 4 (5) 437-441 Oct 2001
[13] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[14] Nanoscale Roughening Of Si(001) By Oxide Desorption In Ultrahigh Vacuum
Gray SM Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology
B 14 (2) 1043-1047 Mar-Apr 1996
60 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 61
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and nano-lines
When the silicon surface is not protected by a native oxide layer or by a hydrogenated
passivating layer it is very reactive The surface chemistry of both Si(100) and Si(111) has
been extensively studied starting from clean surfaces prepared in UHV because of its great
relevance to the microelectronics industry and its technological interest The diffusion and
reaction of molecules and atoms on silicon is an interesting scientific problem on its own
apart from technological issues because silicon reconstruction yields a very complex and
anisotropic surface
We give here a brief summary of the adsorption behaviour of several elements on Si
surfaces The summary is not comprehensive and only the most interesting characteristics
for each substance are given
Table 61 Summary of adsorption behaviour of atoms and molecules on silicon surfaces [12]
Hydrogen
Molecular H2 shows low reactivity towards silicon while atomic hydrogen easily forms
Si-H bonds and can even break Si-Si bonds
Alkali metals
Alkali metals diffuse rapidly into SiO2 and can damage silicon-based transistors
Transition
metals
All transition metals apart from gold and silver react with Si forming metal silicides
62 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nickel A very common contaminant of silicon Nickel from even slight contact with stainless
steel tweezers can contaminate silicon samples forming its silicide and is very difficult
to remove even at high temperatures because it segregates on the silicon surface
Palladium Forms silicides especially Pd2Si and can be used to make contacts possessing a low
Schottky barrier
Titanium Widely used to fabricate contacts for silicon-based microdevices TiSi2 is one of the
more stable and highly conductive silicides
Tungsten The silicide is more stable than TiSi2 on polysilicon and is thus used for polysilicon
gate contacts
Platinum Silicide is used for bipolar transistors PtSi contacts on n-silicon give Schottky diodes
Cobalt Cobalt silicide gives better quality contacts than TiSi2 at the nanometrelevel but is less
used at the moment because it consumes too much silicon during its growth
Iron Silicides have been studied for potential optoelectronic applications because FeSi2 has a
directly accessible bandgap and is optically active
Group 13
Group 13 metals are used in IC technology to dope silicon (giving p-type doping) and
to make III-V type semiconductors When evaporated on Si(100) or Si(111) they react
strongly with the surface giving surface reconstruction and kink formation at steps
Boron Widely used for doping Usually deposited with decomposition of boron hydrides
Sticking coefficient of simple hydrides is very low so decaboranes (B10H14) are used for
deposition
Aluminum Aluminum is widely used for microelectronic contacts but in some cases it reacts with
silicon diffusing into the crystal and with SiO2 stealing oxygen atoms to form Al2O3
Group 14
Carbon Carbon can form a composite semiconductor with silicon (silicon carbide) widely used
to fabricate integrated circuits for use at elevated temperatures or in the presence of
ionising radiation When carbon is deposited on silicon usually polycrystalline films
with properties between diamond and graphite are obtained
CO adsorbs molecularly on silicon without breakage of the C=O bond Other
unsaturated hydrocarbons do not react with silicon Even very reactive strained
molecules like cyclopropane have low sticking coefficient Ethylene acetylene and
propylene adsorb molecularly on Si(100) each molecule sitting on a silicon dimer
interacting with the dangling bonds of the surface Acetylene also adsorbs molecularly
on Si(111) on the 7x7 reconstructed surface preferential adsorption on centre atoms
with respect to corner atoms is observed Benzene and other aromatic ring compounds
adsorb molecularly on the surface They can form σ or π-type bonds with the silicon
laying parallel or tilted respect to the surface according to the bond type
Silicon Silicon atoms can be deposited on the surface with molecular beam epitaxy or
VPalermo 63
decomposition of silanes and chlorosilanes Silanes with single Si-Si bonds have a high
sticking probability and dissociate upon adsorption giving trihydrides and
monohydrides
Chlorosilanes have high sticking coefficients too and decompose on the surface but
they can etch the surface through the reaction SiCl4 + Si rarr 2SiCl2
Germanium Germanium has a lattice constant similar to silicon (Ge lc is 4 larger than Si) so that
various alloys of Si1-xGex can be formed The growth of Germanium on silicon is quite
peculiar with several monolayers adsorbing uniformly on the surface followed by
island formation (Stranski-Krastanov growth)
Group 15
Nitrogen Exposure to ammonia and nitrogen gas at high temperatures leads to the formation of
silicon nitride layers (Si3N4) NH3 dissociates on the surface and reacts with the
dangling bonds giving Si-NH2 and Si-H termination Silicon nitride is stable on the
surface up to 1100degC above this temperature it desorbs as Si2N
Phosphorous Phosphine (PH3) adsorbs dissociatively as Si-PH2 and Si-H and behaviour is similar to
that of ammonia
Arsenic A monolayer of As is often deposited on silicon as a substrate for GaAs growth Arsenic
forms dimers on Si(100) creating dimer rows which are parallel (perpendicular) to the
underlying silicon dimer rows when it is deposited at high (low) temperature On
Si(111) Arsenic breaks the 7x7 reconstruction giving a 1x1 pattern This 1x1 As
monolayer acts as a passivating layer and can resist further As adsorption oxygen and
air
Antimony and
Bismuth
Due to their large covalent radii Sb and Bi form only short dimer rows on silicon and
only at high temperatures
Group 16
Oxygen
Oxygen can oxidize or etch silicon according to the reactions
Si(s) + O2 rarr SiO2 (s)
Si(s) + frac12O2 rarr SiO(g)uarr
Silicon oxidation has already been described in the previous chapter Molecular oxygen
adsorption on silicon is mostly dissociative O atoms break the Si-Si bonds forming an
Si-O-Si bridge or a peroxy bridge Si-O-O-Si Hydrided silicon surfaces do not have
surface dangling bonds and are thus more protected from oxygen attack Etching is
favoured at high temperatures and low oxygen pressures [3] At T gt700degC silicon oxide
on silicon decomposes as SiO(g)
Water
H2O adsorbs dissociatively on silicon easily on Si (100) and with more difficulty on
Si(111) The molecule decomposes giving Si-OH and Si-H on adjacent sites
64 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Group 17
Fluorine
Fluorine adsorbs and reacts easily on silicon dissociating on the surface and forming
Si-F bonds It can then penetrate the surface and attack polarised Si-Si backbonds The
surface is etched with production of SiF4 and SiF2
Chlorine
Chlorine reacts aggressively with silicon etching the surface Cl2 and F2 are often used
commercially to etch silicon usually in a plasma Atomic Cl and Cl2 adsorb on Si(100)
up to saturation The most stable configuration seems to be a Cl atom bonded to each Si
atom of a surface dimer in a symmetric structure but metastable configurations with
two Cl atoms bonded to a buckled dimer and even a central Cl atom bridged across a
dimer have been observed
On Si(111) silicon mono- di- and tri-chlorides are formed In particular heating at
Tgt400degC a uniform Si-Cl monolayer is obtained and the 7x7 reconstruction changes
into a bulk like 1x1 lattice The surface transition induced by Cl adsorption is fully
reversible The 7x7 reconstruction can be restored on desorbing the chlorine by heating
at Tgt1100degC The 1x1 domains nucleate at the lower terrace side of steps the 7x7 at the
upper step edge
Bromine
Like hydrogen bromine maintains the 2x1 structure of Si(100) forming Si-Br bonds
with the dangling bonds of the silicon dimer rows At high exposure some etching of
silicon by formation of volatile SiBr3 species has been observed
Surface diffusion on silicon
In the previous chapter we observed nano-island growth on silicon surfaces with native
oxide and surfaces contaminated with organic impurities While this phenomenon has been
observed in several experiments the formation and growth mechanisms of nanoislands has
still not been satisfactorily described theoretically Various substances have been
hypothesised as nanoisland nucleating agents such as organic or metallic contaminants
present on the surface or oxide clusters (see previous chapter) The island growth process
has been attributed by various workers to lsquosome sort of kinetic instabilityrsquo [4] to the
pinning of step flow by SiC clusters [5] or to a mesoscopic atom flux from areas with low
step density to areas of higher step density [6]
One mechanism proposed to explain surface roughening and island growth is the Ehrlich-
Schwoebel effect ie the presence of an energetic barrier that reflects atoms approaching a
VPalermo 65
Molecule (L
O2
H2O MeOH
CO CO2 CH4
clean surface
downward
weak on si
The aim o
surface an
To obtain
the ones
spontaneou
first obser
surface W
nanoisland
Experime
We used S
removed f
41 mixtur
water and
whole hea
1200degC re
kept below
Table 62 Treatment island density and island volume for each molecule tested
Dose angmuir)
Heating time at 800degC (min)
Island density (microm-2)
Ave Volume (nm3)
Equivalent layer thickness (nm)
280 10 no islands - -
90 10 no islands - - 30 10 2000 19 004
180 10 360 516 019 30 60 no islands - -
180 10 no islands - - - 10 no islands - -
step [7] Recent experiments however indicate that the Schwoebel effect is
licon [48]
f our experiments was to study the reaction of simple molecules with the silicon
d to check the ability of these molecules to nucleate nanoisland growth
a better understanding of the process we chose simple molecules starting from
which are more likely to be present as traces in UHV chambers where
s nanoisland growth is often observed after sample heating Using STM we
ved at the atomic level the possible interactions of these molecules with the
e then heated the surface and checked the density and dimensions of the
s produced
ntal procedures and results
i(111) wafers p-doped 0015 Ω cm Gross contamination and particles were
rom the sample surface with a standard Pirana etch (15 min dip in a H2S4H2O2
e at T=80degC) After this the samples were thoroughly washed in ultra-pure
introduced into the UHV system After degassing for 5 hours at 600degC the
ting stage was allowed to cool down and a sequence of rapid flashes to 1100deg-
moved the native oxide layer and cleaned the surface Pressure during flashes is
1x10-9 mbar
66 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Surface flatness and cleanliness were checked with STM and the contaminant molecules
introduced into the UHV system through a leak valve Doses of several Langmuirs were
used for the experiments (1 Langmuir=10-6 torr sec-1) Pressure was kept constant at 10-7
mbar during adsorption The STM tip was retracted during gas deposition to avoid any tip
shielding effect After the adsorption process the leak valve was closed the base pressure
in the UHV chamber returning to below 10-10 mbar and the presence of adsorbed
molecules checked on the surface with STM Finally the sample was heated at 800degC for
several minutes to allow nanoisland growth Several different areas were examined on each
sample
Table 62 summarizes the dose of exposure and the heating time for each molecule Other
experiments were carried out using smaller dosing times but only the significant results
are reported here
As expected oxygen and water react with the surface oxidising it The adsorption process
can be easily followed with the STM by lowering the tip from time to time and taking an
image No effect of the STM tip is observed over the scanned areas Upon molecule
adsorption the order of the crystalline surface rapidly degenerates and after several
minutes the surface looks completely covered by irregular atomic-sized protrusions
corresponding to Si-OH or Si-O-Si species On heating to 800degC the contaminating species
were easily desorbed through SiO formation the surface became clean and the crystal
surface periodicity re-established No residues or nanoislands were observed
Amongst the molecules tested was methanol CH3OH reacts readily with the surface and
covers it with a disorderd layer already after sim 30 Langmuirs exposure (fig 61)
Fig 61 Si(111) surface during methanol adsorption at 0 1 and 5 minutes respectively (corresponding to 0 6 and 30 L exposure) The dark irregular lines are borders between different 7x7 domains Image size 60x60 nm
VPalermo 67
Synchroton radiation photoemission spectroscopy experiments by Carbone et al [9]
showed that methanol reacts with the rest atoms of the 7x7 silicon cell already after 1
Langmuir exposure but that longer exposures are needed for methanol to react with the
remaining silicon atoms Methanol adsorbs dissociatively forming SiOCH3 and Si-CHx
species on the surface
After methanol adsorption the sample is heated to 800degC and observed again with STM
As in the case of water and oxygen the heating restores the crystalline surface and the 7x7
pattern returns visible But this time quite a dense array of nanoislands is observed on the
surface (fig 62) The islands have an average diameter of 20 nm and a height of sim2 nm
At temperatures above 400degC the Si-OCH3 and Si-CHx species decompose oxygen and
hydrogen are desorbed and according to [9] carbon atoms remain as SiC dispersed
uniformly on the surface Our STM measurements indicate that the carbide does not
randomly cover the surface but that C atoms are concentrated at the nanoislands Rough
calculations based on island volume suggest that the islands must be composed of a SixC1-x
alloy with x varying between 05 and 1
Following these results we expected carbon monoxide to behave in a similar manner
adsorbing onto the surface and with nanoisland growth However no adsorption was
observed with STM and the surface looked perfectly clean even after a dose of more than
100 L of carbon monoxide Some species did adsorb onto the surface though because
after 10 minutes subsequent heating nanoisland growth was observed Island density in
this case was only 18 of the density obtained with methanol and island dimensions are
Fig62 Nanoislands on Si(111) created after 30 L adsorption of methanol and 10 min heating at 800degC Image size 200x200 nm
68 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
much bigger (see table 62 and fig 63) To check whether the islands had been nucleated
by some other contaminant present in the vacuum chamber we heated a blank sample
without introducing any molecule No island growth was observed Exposing the surface to
other simple molecules including carbon dioxide and methane also gave no nanoisland
formation
We can estimate the amount of carbon deposited on the surface during gas dosing The
total flux of molecules hitting the surface is obtained from the Hertz-Knudsen formula
TkmPF
π2=
Fig 63 STM images of Si(111) after exposure to methanol (left) and carbonmonoxide (right) contaminating molecules and subsequent nanoisland growthImage size 400x400 nm The size distribution of nanoislands is shown below eachimage
VPalermo 69
Fig64 Left Si(111) after 180 L contamination of CO and 16 hours heating at 800degC Right Si(111)without CO contamination after 16 hours heating at 800degC Image size 250x250 nm Vertical scale is thesame for both images z-ranges are 35 and 5 nm respectively
where P is the pressure in Pascal k=138x10-23 J K-1 is the Boltzmann constant T is the
temperature and m is the molecular weight (sim32 for methanol and sim28 for CO) in
kilograms
Using this formula we obtain a flux of 027 molecules nm-2 sec-1 for methanol and 029
molecules nm-2 sec-1 for CO After ten minutes exposure at 10-7 mbar pressure more than
150 molecules will have hit each square nanometer of the surface Surface density of
Si(111) 7x7 is sim16 atomsnm2 so each surface atom will be hit by several molecules which
could react or be adsorbed even assuming a sticking coefficient much lower than unity
To explain nanoisland growth we hypothesize that CO molecules adsorb molecularly and
-5
0
5
10
15
20
25
30
0 20 40 60 80 100 120
nm
nm
10 min40 min16 h16 h- no CO
Fig65 STM profiles of nanoislandsgrown for different annealing times The lower curve correspond to the surface heated at 16 hours without CO contamination
70 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Table 63 Island density and volume for the CO-contaminated surface at different heating times For each sample the thickness of a uniform layer having the same volume per unit area of the islands is calculated
Molecule Dose
(Langmuir) Heating time at
800degC (min) Island density
(microm-2) Ave Volume
(nm3) Equivalent layer thickness (nm)
CO 180 10 362 517 019 CO 180 40 550 1072 059 CO 180 16 h 505 3675 186
clean surface - 16 h rough - -
very weakly on Si(111) moving rapidly on the surface Because of their rapid diffusion
the STM is not able to image the CO molecules on the silicon [10]
Increasing the temperature part of the CO molecules will surely desorb but some of them
will dissociatively adsorb onto the surface The oxygen will then desorb as SiO while
carbon will remain fixed forming very stable Si-C bonds
CO decomposition will be favored at surface defects surface steps or in the proximity of
already formed SiC clusters An increased reactivity of the CO molecule on a surface in
correspondence with phase boundaries has already been observed on Pt surfaces [10]
In this case a reduced number of nucleation centres will react with the CO molecules
yielding large and fewer islands with respect to the case of methanol which does not have
the possibility of travelling long distances over the surface
Other factors are likely to influence the process of island nucleation after CO adsorption
the co-adsorption of other molecules which can slow down CO and favour its
decomposition cannot be ruled out as well as the formation of new defects at higher
temperatures Measurements with a variable temperature STM or with some other surface
analysis technique are planned in the future to confirm the proposed mechanism
When the surface was further annealed the nanoislands grew in size After 16 hours
heating very large nanoislands with diameters of sim35 nm and heights of sim20 nm became
visible (fig64 left) As a comparison when the same surface was heated without
nucleation centres deriving from contamination an irregular surface was obtained (fig64
right)
While the nanoisland density seemed to reach an asymptotic limit of 500 islandsmicrom2
(table 63) island size continued to grow with time (fig65) even when no further carbon
was supplied to the surface This indicates that even though the initial nucleating core of
the nanoisland is likely to be an SixC1-x alloy further growth is due to silicon atoms
diffusing from the crystal and being adsorbed by the growing island The final volume
VPalermo 71
occupied by the islands corresponds roughly to a 18 nm thick overall layer of removed
silicon
It is noteworthy that the clean sample after 16 hours annealing even though smoother than
the CO-contaminated one looked much more disordered with hardly any flat area visible
on nanometres scale In the presence of nanoislands however further annealing increases
island size but keeps part of the surface quite flat and ordered with flat areas visible
between the islands Thus the nanoislands can be imagined to act as ldquoimpurity sinksrdquo for
further contaminants approaching the surface yielding a greater but more ordered surface
roughening with respect to the uncontaminated silicon surface
Circular arrays of nanoislands
We used the voids described in the previous chapter formed during oxide layer
decomposition as nanoscopic masks to test the validity of the hypothesis outlined above
The small circular areas uncovered during oxide decomposition can act as ldquoskating rinksrdquo
where contaminants and diffusing silicon atoms can move freely over the surface while
the oxide layer all around and higher than the void area as well as being chemically
bonded to the surface silicon atoms will hinder surface diffusion Furthermore the oxide
passivating layer will at least in some cases prevent contaminating molecules from
adsorbing and decomposing on the surface In this way selective nanoisland growth inside
void areas can take place
The experimental procedure consisted of the following steps
1) An oxide-passivated sample was heated for 10 minutes at 800degC Oxide decomposition
began in correspondence with surface defects and spread laterally creating naked areas
of silicon
2) While the void enlarges the same defect that initiated oxide decomposition behaved as
an attractor for moving silicon atoms and nucleated the growth of a nanoisland at the
void centre
3) The void surface was exposed to contaminating molecules
4) The sample was then heated again at 800degC During the second heating the void
continued to enlarge and new nanoislands were created The freshly uncovered silicon
surface provided mobile silicon atoms to feed nanoisland growth
72 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
5) When the heating continued for long enough all the oxide was desorbed and a clean
silicon surface covered by circular groups of nanoislands was obtained
Figure 66 shows the different stages of the process
We thus heated an oxidized sample at 800degC for 10 minutes in UHV to create multiple
voids in the oxide layer After this we exposed the sample to the contaminating molecules
Then we further annealed the surface at the same temperature to promote island growth
The first time we observed nanoisland growth in the void area was by chance after heating
for a second time samples which had been stored for 2 weeks in vacuum Even at pressures
as low as 10-11 mbar some molecules will adsorb onto the surface slowly creating defects
and thus potential nanoisland nucleation centres It is possible to observe with STM that
the silicon crystal surface looks very clean just after a high temperature flash but even
after only a few days randomly adsorbed species will be observable on the surface
Unfortunately it is not possible to identify these species simply from in situ STM
measurements
Mass spectrometry measurements show that the residual gas contaminants in vacuum are
usually He Ar H2 CH4 CO CO2 and N2 [11] Some of these molecules such as
hydrogen will stick to the silicon surface at room temperature but will simply desorb
during annealing without nucleating nanoislands Others like CH4 or CO2 will not react
with the surface to generate nanoislands as demonstrated previously The best candidate
for contamination of samples stored in UHV seems thus carbon monoxide has the ability
321
4 5
Fig 66 Selective nanoisland growth within oxide voids See text for details
VPalermo 73
as shown before to nucleate nanoislands although co-adsorption mechanisms can not be
ruled out Fig 67 (left) shows the circular groups of nano-islands obtained after heating
void-covered samples stored for 2 weeks in UHV Small islands are observed in the former
void area The void itself has enlarged slightly uncovering a clean oxide-free area of the
silicon surface Apart from the central island created during the first period of heating the
surrounding ones have dimensions decreasing from the void border to the inner area of the
void
This confirms that most of the material needed for island growth comes from the silicon
atoms diffusing from the freshly uncovered areas around the original void The outer
nucleating centres are nearest to the silicon atom source and thus generate larger islands
For a faster process we directly exposed the void to a significant concentration of CO
Fig67 (centre) shows the islands grown after a 180 Langmuir exposure and subsequent
heating The islands are preferentially located around the original void perimeter the rest
of the original oxide-free surface remaining untouched This suggests that although the
whole surface is exposed to CO the gas only interacts chemically with the surface at the
SiSiO2 interface around the oxide-free void perimeter The SiSiO2 interface thus provides
preferential nucleation sites for nanoisland growth in the successive thermal annealing
step This recalls a similar phenomenon observed in the catalytic oxidation of CO on the
Pt(111) surface where CO molecules reacted with adsorbed O species only along the
perimeters of oxygen islands [10]
On repeating the CO exposure and thermal annealing steps a second circle of nano-islands
Fig67 Left circular area of nanoislands obtained from a sample stored 2 weeks in UHV 500x500 nmCentre nanoisland circle obtained after exposing the voids to 180 L of CO and heating for 10 min at800degC 400x400 nm Right two concentric circles obtained with further CO adsorption followed byheating 650x650 nm
74 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
was formed around the newly-expanded surface void perimeter (fig67 right) This
process can in principle be repeated again to obtain a series of concentric groups of
nanoislands the only limit depending on the minimal distance between neighbouring
expanding voids We have thus observed that contaminants adsorbed at low pressure over
long periods of time generate random island growth over the whole void area while CO
molecules adsorbed at relatively high pressures over short periods yield selective
nanoisland growth at the former void border Several theories can explain the different
results CH4 and CO2 as mentioned above did not nucleate islands when when the silicon
surface was exposed to them for short periods However the lower surface mobility of
these contaminants and the greater time available for reaction with the surface could be
the cause of nucleation of SiC clusters over the whole oxide-free void area and not only at
its border The coadsorption of water another UHV residual gas with CO on the naked
silicon areas could perhaps allow a surface reaction similar to that between CO and the
SiSiO2 interface during CO exposure Another explanation for the formation of these
random nucleation sites within the voids could be the coadsorption of residual carbon-
containing species present in the UHV system with hydrogen the major residual gas in
stainless steel UHV systems
The possibility of decorating silicon oxide borders with nanoislands using the high
mobility and selective decomposition of CO molecules suggests interesting applications
for nanofabrication processes that will be discussed below
Silicon nanowire creation on Si(100)
We performed most of the experiments described above on silicon (111) 7x7
reconstructed because this surface is easily prepared in UHV and above all because it is
isotropic and has no preferential directions for atom diffusion The 2x1 reconstruction of
silicon (100) even though it has a much simpler unit cell than Si(111) 7x7 shows a more
complicated morphology with lots of monoatomic and biatomic steps and dimer rows
parallel to each other The orientation of the dimer rows changes by 90deg on alternate
atomic layers Surface diffusion in this case is much more complicated because atoms will
experience the effects of the surface anisotropy and diffuse preferentially along or across
the dimer rows The diffusion energy of a silicon atom as example is 06 eV along a
dimer row and 085 eV across different dimer rows This while being a complication
VPalermo 75
offers interesting possibilities for the creation of ordered structures The monoatomic steps
and the dimer rows can for example act as templates for the formation of elongated
structures of composition similar to the islands described previously
On heating a Si(100) surface oxide decomposition and void growth takes place as
described for Si(111) In some cases the void shape reflects the surface symmetry
especially for very thin oxide layers and the voids often have a slightly squared shape
Nanoisland growth is also influenced by the substrate symmetry and the islands look
rectangular as irregular clusters surrounded by concentric patterns of monoatomic steps
The whole structure reminds vaguely a ldquozigguratrdquo a kind of stepped pyramid found in
Mesopotamia (fig68 inset)
Apart from nanoisland growth in some cases we observed spontaneous formation of
nanowire-like SiC structures on the Si(100) surface The wires were generated after
exposing a Si(100) surface in vacuum to traces of CO followed by sample annealing at
800degC for 15 minutes The wires exhibit lengths ranging from 10 to 100 nm and average
widths of ca 5 nm Although being randomly positioned on the surface they are perfectly
aligned along the crystal axes of the substrate (Fig68)
Several descending monoatomic steps can be observed around each nano-line and nano-
island Very often one of the line extremities coincided with an island
Nanoline
Nanoisland
Fig 68 A Si(100) surface covered by nanoislands and nanowires The monoatomic steps present on the surface are visible The two insets show a typical nanoisland and a nanowire obtained on this surface The typical ldquodimer-row ldquostripes are visible along the sides of the nanowire
76 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Couples of parallel lines starting from the same nano-island were also observed At high
resolution (fig 69) the lines showed a lamellar periodic structure with a periodicity of
around 4 nm and frequent defects Scanning Tunneling Spectroscopy measurements
showed no significant difference between the line and the surrounding silicon surface
The proposed formation mechanism for this kind of structure involves two stages in the
first one organic contaminating molecules adsorb and diffuse on the surface from the
vacuum environment stopping preferentially at surface singularities including steps and
defects
Several substances such as atomic carbon carbon monoxide ethylene and fullerene
decompose when adsorbed on silicon surfaces at temperatures above 500degC and create
localised defects on the surface inducing strain deformations and a metastable surface
reconstruction [12]
If these molecules remain attached to the steps they can diffuse rapidly along step border
until they meet a line and decompose In this way long lines of SiC defects parallel to the
surface steps can be produced In the second phase the clusters of silicon carbide can
-005
0
005
01
015
02
025
03
035
04
-3 -2 -1 0 1 2 3
V
dId
V lt
dId
Vgt
linesilicon
Current- Voltage behaviour measured with STS Fourier analysis of wire periodicity
asymp 4 Aring
High resolution image of a line
Fig 69 Top high resolution image of a nanowire showing the lamellar structure Fourier analysis (bottom left) shows that the lamellae periodicity is sim 04 nm bottom right the IV characteristics measured with the STM tip on the line and the silicon surface
VPalermo 77
25-600degC
Final linear structure
600-800degC
Contaminant Silicon
Fig 610 Schematic representation of nanowire formation mechanism
locally inhibit the spontaneous surface diffusion of silicon atoms acting as templates for
the reorganisation of surface steps (fig 610)
The final morphology shows disordered lines and islands each surrounded by a complex
pattern of silicon monatomic layers
At 800degC silicon atoms are highly mobile on the surface Due to atom diffusion the
surface reorganizes with a step-flow mechanism The presence of a fixed line of
contaminants blocks atom diffusion and step flow generating the monoatomic steps
around each wire (fig 611) SiC clusters act as lsquopinning sitesrsquo on silicon atom surface
diffusion and can be purposely used to direct the localised growth of nano-islands [13]
Although the possibility of using methanol and carbon monoxide as nano-island precursors
has been demonstrated the exact nature of the substances nucleating nano-wire growth is
still unknown
Further experiments will be needed to confirm these hypotheses as well as to identify the
contaminating species nucleating nanowire growth Carbon monoxide is a main candidate
for nanowire growth but the real nucleation process is likely to be a complex one
78 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Step flowblocked
Fig 611 Representation of step flow blocked by the presence of a nanowire 105x105nm
Bibliography
[1] Surface-Chemistry Of Silicon Waltenburg HN Yates JT Chemical Reviews 95 (5)
1589-1673 Jul-Aug 1995
[2] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[4] Growth Of Si On The Si(111) Surface Lanczycki Cj Kotlyar R Fu E Yang Yn
Williams Ed Das Sarma S Physical Review B 57 (20) 13132-13148 May 15 1998
[5] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
VPalermo 79
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990
[6] Nanoscale roughening of Si(001) by oxide desorption in ultrahigh vacuum Gray SM
Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology B 14 (2)
1043-1047 Mar-Apr 1996
[7] Step Motion On Crystal Surfaces Schwoebel Rl Journal Of Applied Physics 40 614-
618 Feb 1969
[8] Scanning Tunneling Microscopy Investigation At High Temperatures Of Islands And
Holes On Si(111)7x7 In Real Time Evidence For Diffusion-Limited Decay
Hildebrandt S Kraus A Kulla R Wilhelmi G Hanbucken M Neddermeyer H Surface
Science 486 (1-2) 24-32 Jul 1 2001
[9] Methanol Adsorption On Si(111)-(7x7) Investigated By Core-Line Photoemission And
Mass Spectrometry Of Photodesorbed Ions Carbone M Piancastelli Mn Zanoni R
Comtet G Dujardin G Hellner L Surface Science 370 (1) L179-L184 Jan 1 1997
[10] Atomic And Macroscopic Reaction Rates Of A Surface-Catalyzed Reaction
Wintterlin J Volkening S Janssens Tvw Zambelli T Ertl G Science 278 (5345)
1931-1934 Dec 12 1997
[11] Redhead PA Hobson JP Kornelsen EV The Physical Basis Of Ultrahigh
Vacuum Chapter 12 (Chapman amp Hall London 1968)
[12] The Si(001) C(4 X 4) Surface Reconstruction A Comprehensive Experimental
Study Norenberg H Briggs Gad Surface Science 430 (1-3) 154-164 Jun 21 1999
[13] Production Of Nanostructures Of Silicon On Silicon By Atomic Self-Organization
Observed By Scanning Tunneling Microscopy Jones D Palermo V Applied Physics
Letters 80 (4) 673-675 Jan 28 2002
80 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 81
7 Conclusions and proposals for future work It is very difficult to foresee the future of nanoscience and nanotechnology even over the
next few years exciting discoveries are continuously made opening up new possibilities
and progress in this sector is now one of the fastest and most varied So it is impossible to
predict whether computers will in the future be made with nanowires nanotubes single
electron devices or some kind of quantum-based transistor What does seem clear
however is that silicon devices can still further extend their levels of miniaturization to
remain competitive for at least the next ten years [1] During this period new emerging
technologies will reach scientific maturity and arrive at the production lines
Even when new technology prototypes do become available for full-scale production the
astronomical cost of changing from silicon to new technologies will be a major concern
Moreover a vast amount of extremely detailed knowledge is already available on all
aspects of silicon technology Thus the possibility of integrating innovative
nanotechnologies with standard CMOS silicon technology already seems an attractive
prospect [2]
For this and other reasons we have focussed our research efforts on surface phenomena on
silicon wafers which could be potentially useful for the development of silicon-compatible
nano-devices
Some important conclusions can be drawn from the work described in this thesis
bull Silicon surfaces of different crystal faces were studied at atomic resolution in ultra-
high-vacuum following transformations induced on the surface by chemical (etching)
and physico-chemical (organic contamination heating) treatments
82 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull The presence of surface contaminants induces at high temperature the growth of
nanometre-sized islands and lines on silicon
bull Final island densities and dimensions strongly depend upon the chemical nature of the
contaminant molecules and the treatments used Molecules interacting only weakly
with silicon will desorb upon heating whereas molecules which decompose easily on
Si will give widespread random nano-island nucleation Between these two extreme
cases complex behaviours of diffusion and subsequent decomposition are possible
bull Nanoscopic naked silicon voids were produced in oxide layers through carefully-
controlled thermal annealing It was found that void density and size depends upon
oxide purity surface contamination levels and annealing conditions
bull The growth of nano-islands and nano-voids was exploited simultaneously and the
voids used as nanoscopic masks to control the positions of growing islands Complex
silicon-on-silicon nano-structures were obtained with multi-step growth processes
bull Finally the use of macroscopic treatments and simple physical processes to produce
large quantities of nanometre-sized structures on silicon surfaces was demonstrated
These processes and the structures produced are of potential interest to the
microelectronics industry
We showed in previous chapters how it is possible to modify silicon surfaces using
chemical and physical methods The use of self-organization for the production of Si-on-Si
20 nm
Fig 71 Left SEM image of field emitter arrays made by conventional lithography [3] Center 3-D STM image of a self-organized void with a central nano-island Right Lateral schematic view of a field emitter device
VPalermo 83
nano-structures has the advantage with respect to other emerging techniques of being
silicon compatible and thus is a potential candidate for its implementation in the
production of new devices Furthermore this technique can create billions of strucures on a
wafer in only a few minutes being thus applicable to large-scale production
Even when ordered structures are obtained these methods still show a poor reproducibility
and control Chemically modified surfaces look very disordered at high magnification
with trenches and holes etched into the surface and progress has recently been made in
producing very small structures electrochemically [4] Oxide decomposition with
nanoisland growth in UHV can also yield ordered structures The void holes with
nanoislands located at their centres although quite variable in dimensions are more similar
to engineered devices than to a spontaneous random surface-roughening process In fact
there is a surprising similarity between the nanovoid-nanoisland structures and well-known
commercially available devices called field emitter arrays (FEA) These are small tips used
as electron guns in various kinds of displays and other devices Both the commercial
devices and our self-organized structures consist of a protruding tip at the centre of
microscopic holes on an insulating layer covering a semiconductor surface (see fig71)
Of course the order and size uniformity of commercial FEAs is much better than that
obtained with our process but it is encouraging to see that a self-organized structure
obtained with simple heating has a very similar surface topography and composition of
commercial working devices obtained after many expensive stages of optical lithography
deposition and etching
Nano-island density and size can to some extent be controlled and different
contaminants can be adsorbed onto the surface to modify the growth process In the future
with a better understanding of diffusion dynamics and growth mechanisms a multi-stage
fabrication process can be envisaged where purposely-designed molecules are adsorbed
onto silicon surfaces to diffuse organize in ordered arrays and favour surface
reorganization at the nanometre scale The same molecules could be used to deliver
dopants to obtain for example highly conducting nanostructures on otherwise undoped
highly resistive silicon The selective formation of nanostructures at the SiSiO2 border
can be used to scale down mesoscopic patterns to the nanometric level A silicon oxide
layer can be easily patterned on the surface with normal optical lithography and then
nanostructures can be built following the pattern edge In this way the production of very
small nanowire based devices may be possible (fig 72)
84 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
D
Gate
S
D
S
SiO2
SiO2
2 Heating nanowire creation
and oxide removal
3 Source-drain deposition by conventional lithography
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
4 Insulating layer (oxide) and gate (metal) deposition
Fig 72 Schematic drawing of the construction of a self-organized nanowire-based transistor
Surface diffusion and decomposition of molecules can be sensitive to even smaller surface
features The nanoscopic lines formed on Si(100) are very regular and neat if their
nucleation and growth mechanism could be controlled ordered networks of nanowires
could be made on silicon
VPalermo 85
The techniques developed here are still experimental ones the careful control and
reproducibility of each step is still an issue and further studies will be needed to clearly
understand the dynamics of formation of these structures Nonetheless the results obtained
so far open up the possibilities of producing novel devices such as for example nano-
transistors (see scheme in fig73 where the mechanism described in fig 72 is applied for
the fabrication of a device array) The possibility of using the border of a lithographic
pattern to draw nanolines can in principle allow the creation of nanodevices using normal
IC manufacturing techniques
The results obtained and their description in this thesis are not the first examples of the use
of self-organization for the production of ordered structures and will certainly not be the
last In our opinion the most important outcome of these findings is that it is possible to
use surface diffusion to overcome the inherent limits of lithographic techniques in
microdevice production processes and to force properly selected molecules to react with
the surface in a spatially differentiated manner at particular sites thereby creating ordered
series of nanostructures The role of the surface dynamics of the substrate atoms
themselves is extremely important in this process
Outstanding results have been and continue to be obtained in nanoscience and
nanotechnology research promising the emergence of new production technologies Sooner
or later those emerging technologies will have to face the issues of cost and compatibility
with the enormous investment made in existing technologies and processing facilities The
commercial production of nanostructures based on surface diffusion processes could in
principle be cheap simple and compatible with existing technology Moreover the
processes of diffusion in bulk silicon is well-known in the microelectronics industry which
has both the know-how and the equipment to better understand and exploit the surface
diffusion processes on silicon surfaces studied here
Extremely important and innovative results are often obtained by exploiting very simple
ideas and in our opinion physical processes such as surface diffusion used to create
complex nanostructures on silicon surfaces is a prime example
86 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
2 Heating nanowire creation and
oxide removal
3 Source-drain deposition by conventional lithography
4 Insulating layer (oxide) and gate (metal) deposition
Fig 73 Schematic macroscopic outline of the construction of a self-organized nanowire-based transistor
VPalermo 87
Bibliography
[1] P Gargini Intel Technology Coordinator ldquoFrom Microelectronics To
Nanotechnologiesrdquo Invited Campus Colloquia Lecture CNR Research Area Bologna
February 6th 2003
[2] Toward A Hybrid Micro-Nanoelectronics Cerofolini Gf Ferla G Journal Of
Nanoparticle Research 4 (3) 185-191 Jun 2002
[3] Recent Progress In Field Emitter Array Development For High Performance
Applications Temple D Materials Science amp Engineering R-Reports 24 (5) 185-239
Jan 25 1999
[4] Silicon Dioxide Micropillars For Sieving Fabricated By Macroporous Silicon-Based
Micromachining Izuo S Ohji H French Pj Tsutsumi K Kimata M Sensors And
Materials 14 (5) 239-251 2002 Electrochemical Etching In HF Solution For Silicon
Micromachining Barillaro G Nannini A Piotto M Sensors And Actuators A-Physical
102 (1-2) 195-201 Dec 1 2002
88 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 89
Acknowledgements
This work would not have been possible without the help and guidance of Dr Derek Jones
he introduced me to the use of STM and other techniques found the funding for this PhD
and helped me with the thousands of small and big problems I had to face during my
research I am grateful to Prof Alberto Ripamonti too for being the supervisor of this
thesis Thanks are also due to Dr Giancarlo Seconi director of ISOF-CNR where most of
this work was carried out Financial support from the Italian National Research Council
(CNR) is also gratefully acknowledged
My passion for surface science began when Fabio Biscarini showed me that it was possible
to really see the atoms and for this I will always thank him
During these years I had the luck to collaborate with many good scientists among those I
would like to cite Enrichetta Susi Massimo Cocchi and Anna Mazzone from CNR
Claudio Zannoni Anna Cavallini Daniela Cavalcoli and Antonio Castaldini from the
University of Bologna Sergio Pizzini and Maurizio Acciarri from the University of Milan
One of the most stimulating periods of my PhD was the one I spent at the Steacie Institute
for Molecular Sciences in Ottawa working with Robert Wolkow Dan Wayner Greg
Lopinski and Peter Kruse I thank them for discussing together many exciting scientific
ideas and trying to put some of them into practice
Special thanks go to my colleagues Paolo Samorigrave and Stefano De Cesari for long and
useful discussions on Science Life and their interactions often with the help of a good
pint
Finally I would like to thank Vassilia Gaetano Simone Claudia Silvia Letizia
Alessandro Angela and Sebastien even though not directly involved in this work they
have been over these three years a continuous source of happiness and support
Bologna March 2003
90 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
SCIENTIFIC PUBLICATIONS bull Lateral diffusion of titanium disilicide as a way to contacting
hybrid Si-organic nanostructures Palermo V Buchanan M Bezinger A Wolkow RA APPLIED PHYSICS LETTERS 2002 v81 p 3636 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v6 issue 20) bull Surface modifications in Si after Rapid Thermal Annealing Castaldini A Cavalcoli D Cavallini A Jones D Palermo V Susi E JOURNAL OF THE ELECTROCHEMICAL SOCIETY 2002 v 149 pG633 bull Production of nanostructures of silicon on silicon by atomic self-
organisation observed by scanning tunnelling microscopy Jones D Palermo V APPLIED PHYSICS LETTERS 2002 v 80 p 673 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v5 issue 5) bull Nucleation of nanostructures from surface defects on silicon Palermo V Jones D SOLID STATE PHENOMENA 2002 v 82-84 p 687 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo MATERIALS SCIENCE AND ENGINEERING B 2002 v88 (2-3) p220 bull Morphological changes of Si[100] surface after treatment with
concentrated and dilute HF Palermo V Jones D MATERIALS SCIENCE IN SEMICONDUCTOR PROCESSING 2001 v 4 p 437 bull Electrical and structural properties of processed silicon surfaces Susi E Cavallini A Castaldini A Cavalcoli D Jones D Palermo V ldquoRECENT RESEARCH DEVELOPMENTS IN VACUUM SCIENCE amp TECHNOLOGYrdquo 2001 v 3 p189 bull Numerical Solutions of the Stochastic Equations of Crystal
Growth Mazzone A M Palermo V INTERNATIONAL JOURNAL OF MODERN PHYSICS C 2000 v 11 Part 1 p195-204 bull Advances in silicon surface characterisation using light beam
injection techniques
VPalermo 91
Acciarri M Pizzini S Simone G Jones D Palermo V MATERIALS SCIENCE AND ENGINEERING B 2000 V73 (1-3) p 235 - 239 bull Abrupt orientational changes for liquid crystals adsorbed on a
graphite surface Palermo V Biscarini F Zannoni C PHYSICAL REVIEW -SERIES E- 1998 V 57 NUMBER 3A p R2519-R2522 Oral presentations bull The use of oxide desorption and surface diffusion for the
creation of silicon on silicon nanostructures Palermo V Jones D 1ST NATIONAL WORKSHOP ON CURRENT TRENDS IN NANOTECHNOLOGIES Catania (Italy) Feb 2002 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo V E-MRS Spring Meeting Strasbourg (France) June 2001 bull STM study of surface transformations on silicon during UHV
annealing Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999 bull Advances in silicon surface characterisation using light beam
injection technique Acciarri M Pizzini S Simone G Jones D Palermo V E_MRS SPRING MEETING June 1999 Posters bull Spontaneous nano-wire growth on silicon Palermo V Jones D SMARTON workshop Leuven (Belgium) October 2002 bull Ordered circles of nano-islands on silicon from CO adsorption Palermo V Jones D TRENDS IN NANOTECHNOLOGY 2002 Santiago de Compostela (Spain) September 2002 bull Production of nanostructures of silicon on silicon by atomic self-
organisation Palermo V Jones D EUROMAT Rimini (Italy) June 2001 bull Etching holes and anisotropic corrosion on silicon [100] Palermo V Jones D Susi E Asoli B SILICON WORKSHOP Genova (Italy) February 2001
92 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull Morphological and electrical characteristics of damaged silicon surfaces
Susi E Castaldini A Cavalcoli D Cavallini A Jones D Palermo V SILICON WORKSHOP Genova (Italy) February 2001 bull Effect of HF etching on the roughness of a silicon surface Palermo V Jones D NATIONAL CONFERENCE ON PHYSICS OF MATTER Genova (Italy) June 2000 bull Effect of Native Oxide Desorption upon the Surface Morphology of Si[100] by STM and LEED Palermo V Jones D SILICON WORKSHOP Genova (Italy) February 2000 bull Characterisation of silicon surfaces for microelectronics through STM measurements Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999
VPalermo 93
Contact Vincenzo Palermo ISOF- Institute for Organic Synthesis and Photoreactivity Via Gobetti 101 40129 Bologna ITALY Tel +39-051-6398336 Fax +39-051-6398349 Mail palermoisofcnrit
94 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Index
1 Introduction nanotechnology and the future of computers
2 Silicon surfaces
3 STM and other surface analysis techniques
4 Surface modification of silicon in liquid Nano-hole creation
5 Surface modification of silicon in vacuum void creation and oxide desorption
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and
nano-lines
7 Conclusions and proposals for future work
VPalermo 13
[9] Logic Gates And Computation From Assembled Nanowire Building Blocks Huang Y
Duan XF Cui Y Lauhon LJ Kim Kh Lieber CM Science 294 (5545) 1313-1317
Nov 9 2001
[10] Formation And Disappearance Of A Nanoscale Silver Cluster Realized By Solid
Electrochemical Reaction Terabe K Nakayama T Hasegawa T Aono M Journal Of
Applied Physics 91 (12) 10110-10114 Jun 15 2002
[11] Fabrication And Integration Of Nanostructures On Si Surfaces Ogino T Hibino H
Homma Y Kobayashi Y Prabhakaran K Sumitomo K Omi H Accounts Of Chemical
Research 32 (5) 447-454 May 1999
[12] Ultrafine And Well-Defined Patterns On Silicon Through Reaction Selectivity
Prabhakaran K Hibino H Ogino T Advanced Materials 14 (19) 1418-1421 Oct 2
2002
14 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 15
2 Silicon surfaces
The name silicon (silicio in Italian) comes from the latin word silex Amorphous silicon
was first isolated by Berzelius in 1824 by reaction of potassium with silicon tetrafluoride
Thirty years later the first crystalline silicon was prepared Silicon makes up 25 of
earthrsquos crust and is the second most abundant element after oxygen Elemental silicon is
not found in nature occurring as silicon oxide (sand quartz amethyst flint etc) or
silicates (asbestos clay mica etc) Perhaps no other element and its compounds has such
a wide range of uses Silicon compounds such as sand and clay are used in the building
industry as refractory materials for high-temperature applications and for enamels and
pottery Silica is the main component of glass silicon carbide is an important abrasive and
silicones are commonly used polymers and lubricants
Here the most interesting use of silicon of course is for the production of
microelectronics devices For this application silicon of high purity (999999) and of
high crystallinity is needed Table 21 lists some of the physical characteristics of silicon
High purity polycrystalline silicon is produced by the reaction of gaseous trichlorosilane
with hydrogen in a furnace Then to prepare a single-crystal of silicon the so-called
Czochralski method is commonly used
Polycrystalline silicon is melted in a quartz furnace at 1415degC in an argon atmosphere
Then a seed of single-crystal silicon is lowered into contact with the melt and slowly
pulled out In this way the crystal grows and a crystalline cylindrical ingot several metres
long is created from the initial seed
After cooling down the ingot is sliced into thin silicon wafers The wafer surfaces are
polished using a counter-rotating lapping machine in an Al2O3 slurry until the surface is
very flat and shiny ready for the lithographic processes
16 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Another way to obtain single crystal silicon is the Floating Zone (FZ) method in which a
silicon cylinder is slowly passed through a heating ring The area inside the ring melts and
solidifies smoothly crystallising as it comes out of the ring yielding a single silicon crystal
Microelectronic devices are built on the silicon surface which is the surface of interest
here Unfortunately silicon surfaces are normally quite dirty and uneven at the atomic
scale Atmospheric oxygen and humidity react with silicon surfaces creating a thin layer of
oxide (called ldquonative oxiderdquo) which is usually irregular and full of defects Different kinds
of contaminants also adsorb onto the surface These are usually small organic molecules
and microscopic dust particles A clean surface on exposure to the atmosphere is
completely covered with gas molecules in less than 10-9 seconds If the pressure is
reduced letrsquos say to 10-6 mbar this time increases to 1 second This is the reason why to
study a clean surface we have to work in UHV at pressures below 10-10 mbar
The atoms in the silicon crystal have a diamond-like structure each atom having 4 bonds
in a tetrahedral sp3 arrangement with bond angles of 10947 degrees At the crystal
surface some atoms will have non-bonding orbitals ldquodanglingrdquo in the vacuum ie sp3
orbitals with a lone electron which are highly reactive These orbitals are known as
dangling bonds To minimize surface energy the surface will reorganize by decreasing
the number of dangling bonds
Table 21 Physical data of silicon [1]
Atomic Weight 2809 Lattice constant (A) 543095 Crystal structure Face-centered cubic
(diamond) Melting point 1415 degC
Density (gcm3) 2328 Boiling point 2355degC Atomscm3 50E22 Minority carrier
lifetime (s) 25E-3
Dielectric Constant 119 Specific heat (Jg degC)
07
Breakdown field (Vcm)
~3E5 Thermal conductivity (Wcm degC)
15
Electron affinity x(V)
405 Vapour pressure (Pa) 1 at 1650degC 1E-6 at 900deg C
Energy gap (eV) at 300K
112 Reactivity Inert to acids Attacked by halogens and alkaline
solutions Intrinsic carrier
conc (cm-3) 145E10 Oxidation states +4 -4
Intrinsic Debye Length (microm)
24 Energy of a Si-Si bond (eV)
232
Intrinsic resistivity (Ω-cm)
23E5
VPalermo 17
Dangling bond densities and positions and thus the type of surface reconstruction will
depend upon crystal orientation as well as the temperature and kinetics of the system
Fig 21 shows a drawing of the main faces of a silicon crystal The angle α between any
(11n) face and the (100) face can be calculated from 2cos 2 += nnα The angle
between any (11n) face and the (111) face can be calculated from
)2(3)2(cos 2 ++= nnα
The chemistry and physics of the faces are very different a brief description will be given
for the most important orientations
Si (100)
On the (100) surface each atom has two Si-Si bonds connecting it to the bulk and two
dangling bonds pointing outward Surface energy is reduced by the dimerisation of the
surface atoms through overlap interaction of one dangling bond per atom forming rows of
dimers aligned along the (110) direction This is the well-known ldquo2x1rdquo reconstruction of
this silicon surface The symmetric dimers would make the silicon surface metallic but to
reduce surface stress the dimers tend to buckle and the surface is thus semiconductive It
took several years to understand that the dimers are buckled because at room temperature
Fig 21 Schematic view of the principal orientations of a silicon surface Surface atoms are white bulk atoms are black dangling bonds are gray [2]
18 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
they shift easily from one buckling direction to the other thus appearing symmetric under
STM observation Fig 22 shows an STM image of the 2x1 reconstructed surface
Even almost perfect (100) surfaces have a certain number of monoatomic steps and the
dimer rows on atomic layers are aligned at 90deg to those on adjacent layers Dimer rows are
thus perpendicular or parallel to the step When the dimers on the upper side of the step
are parallel to the step the step is called SA if they are perpendicular the step is called SB
Because of this symmetry SA step edges will be smoother compared to the more broken
fragmented edges of the SB steps
A common defect on the Si(100) surface is the presence of nickel contamination which
appear as missing dimers in STM images This type of contamination is so critical that
even if the silicon sample is only briefly brought into contact with stainless steel tools
(tweezers for example) the 2x1 reconstruction of the surface can be blocked
Silicon atoms can diffuse easily over the silicon surface as monomers and dimers
especially at elevated temperatures The anisotropy due to the 2x1 reconstruction causes a
difference in the diffusion energies of adsorbates over the surface Diffusion of these
silicon species along dimer rows for example will be much easier A list of diffusion
energies for monomers and dimers is given below [2]
Diffusion on Si(100) 2x1 Ed (eV) Monomers along dimer rows 06 Monomers across dimer rows 085 Monomer formation energy 18 Dimer along dimer rows 11 Dimer across dimer rows 15 Dimer formation energy 26 Dimer binding energy 076 Vacancies along dimer rows 17 Vacancies across dimer rows 19
SA
SB
Fig 22 STM image of a 2x1reconstructed silicon surfaceshowing the dimer rows and steps40x35 nm Nickel-induced defectsare visible as dark spots SA and SBsteps are indicated
VPalermo 19
So the diffusion energy for both monomers and dimers is nearly 40 greater if they have
to cross a dimer row This difference reduces to sim10 for vacancy diffusion
Si(113)
The (113) surface can be imaged as a sequence of alternating (100) and (111)-like
structures with two and one dangling bonds on alternate atoms respectively Interest in the
(113) surface is scientific as it has been used to study the energetics of the (100) and (111)
surfaces as well as for surface adsorption experiments
Si(100) surfaces can easily develop (113) facets
Si(111)
This surface besides being the first one imaged with STM with atomic resolution is one of
the most studied because it is the best cleavage face of silicon and because it shows one
of the most complex and elegant reconstructions in surface science
All Si-Si bonds in the silicon crystal are perpendicular to a (111) plane so this face will
have the lowest number of dangling bonds created per unit area In fact each Si atom on a
(111) surface shows a single dangling bond oriented perpendicular to the surface and
bonded to three back atoms These three bonds for each surface atom account for the great
chemical and physical stability of the Si(111) surface Surface energy is 009 eV Aring-2
compared to 015 eV Aring-2 for Si(100)
For energy minimization this surface reconstructs forming a huge 7x7 lattice cell
containing 102 atoms described by the Dimer-Adatom-Stacking fault model (DAS) For a
detailed description of cell structure see fig 23
The cell described by this model is very complex being composed of three kinds of atoms
adatoms rest atoms and corner hole atoms Furthermore a subsurface stacking fault is
present in one half of the cell making the two halves of the unit cell look different under
STM (Fig 24) It took 26 years of research to completely understand the exact structure of
the 7x7 reconstruction
Cleaving a silicon crystal along a (111) plane produces a metastable 2x1 reconstruction
the 7x7 reconstruction is easily obtained by flashing at high temperature in UHV At T gt
20 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
830degC a disordered 1x1 phase covers the surface Cooling down to 800degC leads to the
formation of the 7x7 phase If the cooling process is too rapid small 7x7 domains nucleate
and a disordered 1x1 phase is preserved between domain boundaries
Si(110)
Even though as mentioned before the (111) plane is the favoured cleaving plane of
silicon thin (100) commercial wafers will not break along this plane because the angle
between (100) and (111) is too far from 90deg (see table 22) Instead they will break along
the (110) plane because it is perpendicular to the (100) surface Each surface atom on
Si(110) has a Si-Si bond pointing downward one dangling bond pointing outward and two
Si-Si bonds parallel to the surface in a zig-zag pattern (see fig 21) Cleaved (110)
surfaces are disordered but upon annealing at high temperatures an ordered complex 16x2
Fig23 Scheme of the 7x7 DAS model [2] In each unit cell there are 9 dimers 12 adatoms and a stacking layer fault The force driving this complex reconstruction is the minimization of dangling bonddensity The DAS model shows the lowest number of dangling bonds (19) of all possiblereconstructions 12 dangling bonds are at the adatoms 6 at the rest atoms and 1 at the corner hole atom This surface is metallic
VPalermo 21
reconstruction takes place The surface appears as a series of long ridges and valleys
parallel to each other Eventually tilted facets of orientation (17 15 1) can form on this
surface The adsorption of Ge atoms on this surface leads to the formation of self-
assembled nanowires [3]
Table 22 Angles in degrees between different silicon faces [2]
Orientation
113
110
111
100
100
2524
9000
5474
0
111
2950
3526
0
110
6476
0
113
0
Fig 24 STM image of a Si(111)surface with 7x7 reconstruction A unitcell with its adatoms is highlightedImage size 13x13 nm
22 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] Weast RC Handbook Of Chemistry And Physics (Chemical Rubber Co Cleveland
1972)
[2] Dabrowski J Mussig H Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] The Structure Of Clean And SiGe-Covered Si(110) Surfaces Butz R Luth H Surface
Science 365 (3) 807-816 Oct 1 1996
VPalermo 23
3 STM and other surface analysis
techniques
Scanning Tunneling Microscopy
Since the invention of the optical microscope at the end of the 16th century the possibility
of examining surfaces at higher and higher magnification has fascinated mankind
Development of the technique continued and towards the end of the 19th century optical
microscopes were as good as todays standard instruments The physical limits of the
wavelength of visible light (350-800nm) had been reached
In the 1920s de Broglie showed that electrons can behave like waves and the use of these
particles for imaging with much higher resolution soon followed Atomic resolution using
this technique is only possible in the transmission mode with extremely carefully prepared
samples
In 1982 using the peculiar properties of piezoelectric materials Binnig and Rohrer brought
a metallic tip very very close to a silicon surface and scanned it across an extremly small
area (fig31) The tunneling of electrons from the tip into the sample or vice versa allowed
them to obtain a local density of electronic states (LDOS) map of that surface Although
theory (which treated the extreme point of the tip as a sphere) then excluded the possibility
of atomic resolution following a tip crash into the surface Binnig and Rohrer began to
observe the LDOS with atomic resolution For this discovery and their following work
they were awarded the Nobel Prize in Physics in 1986
24 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The basic principle of STM is very simple A metallic tip is scanned over a surface without
making ohmic contact and a tunneling current passes between the tip and the surface An
electronic circuit keeps this current constant by raising and lowering the tip during the
scan In this way recording the tip height at each point a three-dimensional image of local
density of electronic states (LDOS) of the surface can be obtained To explain the
extremely high resolution attainable by this simple technique quantum theory is needed
According to classical physics the current will flow between sample and tip only if they
are in physical contact If there is a vacuum gap between the two the electrons will simply
remain confined for example within the surface without the possibility of passing into the
tip
In quantum physics however the electrons have a certain probability of passing
(tunneling) across the gap appearing on the other side of the gap in this way reaching the
tip It can be shown that the probability of an electron tunneling through a gap of thickness
z is
kzep 22)0( minusprop ψ h
φmk
2= (1)
where ψ(0) is the electron wavefunction at the surface-gap border m is the electron
mass=91x10-28 g and φ is the work function of the metal (ie the energy required to
remove an electron from that material For silicon it is 48 eV) The tunneling current thus
decays exponentially with z and is extremely sensitive to topographical imperfections
present on the scanned surface A rough formula giving the current as a function of z is [1]
zFS eEVI φρ 0251)( minusprop
Fig 31 Binnig and Rohrer with the first STM Image from IBM [2]
VPalermo 25
where ρs(EF) is the local density of states at the Fermi level on the given surface For
example the formula predicts that for silicon an increase in tip-surface distance of 1 Aring
will give a 95 decrease in tunneling current
This huge dependence of tunneling current upon the distance allows detection even of the
sub-nanometre changes in height given by the single atoms of which the surface is
composed and thus to resolve them in the LDOS images Of course this description of the
tunneling process is oversimplified and for a more accurate one the electronic states of
the tip of the sample and their interaction have to be taken into account Fig 32 shows a
schematic representation of the interaction between tip and sample orbitals
The exponential decay of current with distance also yields high lateral resolution If the tip
is approximated as a sphere of radius R and the current passing at the minimum tip-sample
distance is I0 then the current passing at a lateral distance x from this point will be
Rxk
eII 22
0
2minus
=
Assuming a tip radius of 100 nm the current is concentrated in an area sim15 nm wide at tip
apex
Very sharp tips with even smaller curvature radii can be produced in several ways Simple
Fig32 Schematic view of tip-sampleorbitals interaction a) no interaction b)equilibrium c) sample positive d) tippositive [1]
26 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
tungsten wires mechanically cut are capable of obtaining atomic resolution on graphite in
air but for more disordered and rougher samples sharper and more reproducible tips are
needed
STM tips are mostly made by electrochemically etching a W or Pt-Ir wire The tips we
used were prepared using methods based on the work of Fotino [3]
A tungsten wire 038 nm diameter is immersed in a KOH solution(06M) with a thicker
tungsten wire used as a counterelectrode The cathodic and anodic reactions involved in the
etching are
Cathode 6H2O + 6 e- rarr 3H2(g) + 6 OH-
Anode W(s) + 8 OH- rarr WO42- + 4H2O + 6 e-
A potential of 3V ac is applied to the tungsten and the wire is immersed in the solution
until a constant current of sim100 mA is established The etching rapidly removes metal
shaping the wire end as a sharp tip When the potential reaches 11 V the coarse tip etching
is finished The wire is then removed from the solution carefully inserted into an
insulating plastic tube leaving only the tip exposed and re-immersed in the solution with
the tip pointing upwards A more gentle etching is thus made to reduce tip radius Usually
5 minutes etching at 07 V ac is used In this configuration very small hydrogen bubbles
formed on the tip sides sliding upwards with a ldquohoningrdquo effect on the tip
This procedure yields extremely sharp and reproducible tips at the microscopic level
After the etching the tip is thoroughly rinsed in ultrapure water then dipped into
concentrated HF to remove surface oxides and hydroxides [4] The tip is dried with
nitrogen inserted into the UHV system and degassed overnight at sim150degC
The possibility of measuring sub-nanometric distances would be useless without being able
to control tip movement over such a minute scale To scan the tip over the surface a
piezoelectric scanner is used Piezos are usually made of an alloy of PbZrO3 and PbTiO3 a
material which contracts or expands when a voltage is applied to it The Omicron
instrument used in our laboratory has three such piezo scanners for xy and z tip motion
allowing one to scan the tip over the surface with sub-Aringngstrom precision (fig 33)
To isolate the instrument from ambient vibrations the whole STM stage is suspended upon
four springs which eliminate all frequencies above 1 Hz and surrounded by a crown of
VPalermo 27
copper wings and fixed magnets Parasitic currents generated by the magnets into the
copper wings contrast every movement of the stage and efficiently block stage vibrations
The STM can be used not only to explore surface topography but to measure the IV
characteristics of single atoms or molecules on the surface (Scanning Tunneling
Spectroscopy or STS) Furthermore it can be used to modify the surface with voltage
bursts digging into it or delicately moving single molecules or atoms over a surface [5] It
can be used in vacuum in air and with proper lateral isolation of the tip even in liquids
A major drawback of STM is that it works only on conducting and semiconducting
surfaces and thus cannot be used on many surfaces of biological and chemical interest
Another instrument more suitable for these and other applications is the Atomic Force
Microscope (AFM)
Atomic Force Microscopy
AFM was invented in 1986 by Binnig Quate and Gerber after calculating the possibility of
building a cantilever with a force constant of the same order of magnitude as that of a
chemical bond
In AFM a tip mounted on a microscopic cantilever (usually made of Si3N4 fabricated with
optical lithography) is brought close to a surface When the tip touches the surface the
cantilever is very slightly deflected upwards The movement is measured by observing the
Y-PIEZO
SILICON SAMPLE
Fig 33 A picture of theSTM used for theexperiments The tripodpiezo scanner is shown
Z-PIEZO
X-PIEZO
TIP
28 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
deviation of a laser beam hitting the upper face of the cantilever Fig 34 provides a
schematic view of the principle of AFM
The typical force constant of the cantilever varies from 00006 to 2 Nm the typical
resonance frequency is 3 to 120 kHz The AFM tip can apply a force on the sample of up
to 10-9 N The AFM can be used on conductive or insulating surfaces in vacuum air or
liquids Furthermore the tip can be modified to sense electrostatic potentials (electric force
microscopy) or magnetic fields (magnetic force microscopy) it can even be functionalized
with complex molecules such as proteins to interact with biological surfaces
A drawback of the AFM is that the force it exerts can damage the surface under
observation especially if the sample is soft (as in the case of cell membranes for
example) This problem can be overcome using the instrument in tapping mode (where the
tip does not move laterally during its brief contact with the surface) or in non-contact mode
in which the tip oscillates above the surface during the scan and the changes in its
frequency due to interaction with surface are monitored The shifts in the oscillating
frequency of the cantilever due to tip-sample interaction are then used for imaging the
surface In this mode interaction of the tip with the surface is minimal and soft samples
can be imaged
STM and AFM are the main techniques used for this work A brief description of other
techniques used occasionally is given below
Fig 34 Scheme of an atomic force microscope
VPalermo 29
Low Energy Electron Diffraction (LEED)
Electrons with energies in the 20-500 eV range are diffracted by a crystalline surface the
diffraction peaks are visualized on a fluorescent screen This technique probes the long
range order of the surface up to a depth of several nanometres
X-Ray Photoelectron Spectroscopy (XPS)
XPS allows both qualitative and quantitative chemical analysis of the elements present on
or near the sample surface
An X-ray source is used to photoionize the atoms on a surface and produce photoelectrons
By measuring the kinetic energy of the photoelectrons the binding energy of the electronic
levels can be calculated This energy will depend on the chemical environment of the
surface atoms
Although the soft X-rays used penetrate to a depth of ~2000Aring the sampling depth of the
technique is determined by the mean free path of the photoelectrons which allows their
escape from only the first 10-100Aring
Secondary Ion Mass Spectroscopy (SIMS)
High and low energy ions (primary ions) are used to bombard a sample and remove surface
atoms and ions The ionic fragments removed (secondary ions) are then analysed by a mass
spectrometer The surface can be consumed during the measurement and profiles obtained
giving concentrations of the materials composing the sample at different depths (depth
profiles)
A popular variant of SIMS is TOF-SIMS In this technique the secondary ion masses are
measured by a time-of-flight (TOF) measurement The secondary ions generated by the
bombarding primary ions are accelerated to a constant kinetic energy and then move
through a field-free space before they reach the detector where their intensity is measured
as a function of flight time Since ions with different masses have different velocities at a
30 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
given kinetic energy the measured flight times of the ions can easily be converted to their
masses The static nature of this latter technique allows mass spectroscopy surface analysis
with minimal damage to the surface
Bibliography
[1] Chen CJ Introduction To Scanning Tunneling Microscopy (Oxford University Press
Oxford 1993)
[2] From wwwibmcom
[3] Tip Sharpening By Normal And Reverse Electrochemical Etching Fotino M Review
Of Scientific Instruments 64 (1) 159-167 Jan 1993
[4] A Convenient Method For Removing Surface Oxides From Tungsten STM Tips
Hockett LA Creager SE Review Of Scientific Instruments 64 (1) 263-264 Jan 1993
[5] Confinement Of Electrons In Quantum Corrals On A Metal Surface Crommie MF
Lutz CP Eigler DM Science 262 (5131) 218-220 Oct 8 1993
VPalermo 31
4 Surface modification of silicon in liquid
Nano-hole creation
Liquid treatments of silicon wafers are very common in the integrated circuit (IC)
manufacturing industry They are used to clean and improve surface uniformity to create
and etch protective oxide layers and to remove photo-resist layers
Crystalline silicon with its native oxide layer is very stable and is resistant to many acids
It is easily attacked by hydrofluoric acid (HF) and alkaline solutions
The thin (~2 nm) passivating layer of native oxide (SiO2) is formed on exposure to the
atmosphere This surface layer contains many defects and contaminants so it is usually
chemically stripped and substituted with a better chemically-formed protective oxide
The most common silicon cleaning procedure is the RCA method named after the Radio
Corporation of America [1] It consists of two steps in the first one the surface is treated
with a hot alkaline solution (H2OH2O2NH4OH 411) to remove particles from the
surface following this a hot acidic solution (H2OH2O2HCl 411) is used to remove
metal contamination Other well-known cleaning methods are IMEC (a sequence of
cleaning steps in H2OO3 and dilute HF) or the Pirana etch (a hot 41 mixture of H2SO4
H2O2)
32 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The standard RCA clean removes surface contaminants etches the native oxide and
oxidizes the silicon surface leaving a uniform layer of silicon oxide which better protects
the surface from further contamination
Etching with fluorine-based solutions
Hydrofluoric acid is one of the most common reagents used in the treatment of silicon
wafers both in the research field and in industrial processes A rapid dip in dilute HF is the
simplest way to remove the native oxide from Si(100) and leaves the surface passivated by
a layer of Si-H bonds Because of the low polarization of Si-H bonds the Si-H layer is
stable even for several days protecting the surface from contamination It has often been
assumed that this short etch does not significantly change the surface morphology of the
silicon substrate[2] even though a prolonged dip in dilute HF leads to surface roughening
[3]
Although dilute HF roughens the Si(100) surface at the atomic scale [45] immersion in
concentrated HF (49) etches the surface oxide without attacking the Si surface
uncovering in this way the buried SiSiO2 interface The final counter-intuitive result is
that dilute HF etches the silicon while concentrated HF leaves the crystalline silicon
untouched [3]
Etching Si with fluorine-containing solutions at different concentrations and pH can
produce different morphologies from rough surfaces to flat nearly ideal Si-H terminated
surfaces
Hessel et al and Higashi et al demonstrated in 1991 that very flat Si(111) surfaces can be
obtained using 40 NH4F while etching with HF always yields rough surfaces The
surface becomes smoother because the etchant rapidly attacks Si atoms at step borders
thus removing surface kinks and irregularities in a step-flow mechanism [6 7] Later on
even smoother and more perfect surfaces were obtained by removing oxygen from the
solution after it was discovered that oxygen dissolved in 40 NH4F initiates the formation
of triangular etch pits It was not possible to obtain flat surfaces by etching Si(100) with
ammonium fluoride solutions which leads to the formation of small 2x1 dimer-row
reconstructed (100) terraces together with (111) facets [8]
This difference is caused by the different hydride terminations prevailing on the (100) and
(111) faces While the ideal Si(111)-H surface is monohydride terminated the more
VPalermo 33
reactive dihydrides predominate on the Si(100)-H surface making it more vulnerable to
etching The etching reaction is thus strongly anisotropic etching (100) facets faster than
(111) thus producing (111) microfaceting on Si(100) crystals
A more uniform Si(100) surface can however be prepared by etching at low pH with an
HFHCl mixture [9] or by using very dilute HF solutions and ultrapure water with low
dissolved oxygen and carbon contents [10]
Electrochemical etching can also be used applying anodic or cathodic bias to the silicon
to obtain different morphologies [11] by varying the potential isotropic or anisotropic
etching is observed The aforementioned results show that despite the simplicity of the
reactants fluoride etching of silicon is quite a complex reaction
Fig 41 Chemical etching of silicon
HO+H
H HO
H
H
H
FSi
Si
SiSi
+H2O
-OH-
+F-
H2O H H
Si H
F OH
H Si
Si Si
-H2
+H2O
H
H
H
FSi
Si
SiSi
H
H
H
OHSi
Si
SiSi
H
H
H
H Si
Si
Si Si
Etching mechanism of silicon
HF rapidly dissolves the SiO2 passivating layer on silicon leaving the surface almost
completely hydrogenated [12] After this two different types of reactions etch the silicon
simultaneously one chemical and the other electrochemical [13] The overall etching
mechanism can be schematized in two stages (see Figure 41)
i) Si-H bonds are substituted by Si-F or Si-OH bonds creating a partial charge on the
surface silicon atom and polarizing its Si-Si backbonds
ii) these polarised backbonds are then more easily attacked by HF or H2O After
rupture of the Si-Si bond the atom is removed leaving behind new Si-H
terminations and the reaction can start again
34 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
These reactions take place although at different rates on both Si(100) and Si(111)
Stage i) is usually the rate-determining step of the reaction and the stability of the Si-H
bonds depends upon the pH the concentration of nucleophilic species in solution and an
eventual potential applied to the crystal
For pH gt5 as in the case of concentrated NH4F solutions reaction begins with attack by
water to give Si-OH (step ArarrC) The -OH group is rapidly substituted by fluorine with
polarization of the underlying Si-Si bonds These bonds are then easily attacked by water
the silicon atom being released into solution as HSiF(OH)2 The Si-OH rarr Si-F substitution
is not fundamental for the reaction and etching can proceed even for Si-OH terminated
atoms but XPS measurements showed the presence of a certain number of Si-F bonds
remaining Furthermore fluorine seems to have a catalytic effect on Si-H substitution as
indicated by the dependence of the etch rate upon the F- concentration at least for pH
values between 4 and 8
Si-F bonds can be easily removed by a water rinse In the case of strongly alkaline
solutions (pH=14) OH- groups act directly as nucleophiles and no fluorine is needed to
catalyze Si-H bond rupture
At pH lt4 almost no free F- ions are present in solution and the etch rate is very small at
pH lt2 all etching reactions are very slow and this explains the stability of Si crystals in
concentrated (50 ww) HF solutions
Si Si
Si Si
H
H
H
H Si
Si
Si Si
H
H
H Si
Si
SiSi
H
H
H
OH-H+ -e
(F-)
-H+
-e
+H2O hellipas in fig1
Fig 42 Electrochemical etching of silicon by H2O
The electrochemical etching of silicon involves electron transfer from the surface atoms to
the valence or conduction band of the crystal (according to the type of doping of the
silicon) Several studies have been carried out at different electric potentials In the case of
p-silicon or n-silicon at anodic potentials where positive charges (h+) are available in the
crystal a silicon-centred radical can be created by capture of a hole from the bulk (fig
42) The F- ion does not participate in the reaction but is thought to contribute through
VPalermo 35
electrostatic interactions by lowering the energy of the interaction step After the
formation of the Si-OH group the reaction proceeds as shown in the scheme of fig 41
Matsumura et al [4] proposed that not only water but HF2- molecules also play a major role
in electrochemical etching of silicon leaving on the surface Si-F terminated bonds which
can be immediately attacked in an autocatalytic process (fig 43)
In the electrochemical reactions described above an external potential is applied to the
silicon crystal The chemical and electrochemical reactions in any case take place
simultaneously most of the time with the chemical path predominating at high pH Even
when no external potential is applied to the silicon partial electrochemical reactions can
take place at different ldquocathodicrdquo and ldquoanodicrdquo sites on the surface with an internal charge
exchange which ensures neutrality [11] This macroscopic silicon etching and hydrogen
bubble formation can lead to visually observable patterns on the surface when Si(100) is
immersed in ammonium fluoride even without applying a potential
Si Si
Si Si
H
H
H
F Si
Si
Si SiH
F F-H+ -2e
+HF2-
H
HF
SiSi
SiSi + F
F F
FH
H
H
F -H+ -2e
+HF2-
Fig 43 Autocatalytic electrochemical etching of silicon by HF2-
Inhomogeneities on silicon surfaces caused by electrochemical reactions and charge
transfer have been studied extensively because they are of fundamental importance in the
formation of porous silicon
Pore formation on silicon
When Si(100) or Si(111) are etched under anodic bias in fluorine-based solutions
microscopic pores form on their surface Several different morphologies of pores have
been observed with pore diameters ranging from 10 nm to several microns with depths of
several microns [14] Pore shape is very variable too ranging from ordered straight pores
to chaotic networks of branched pores (fig 44) Porous silicon has been known since the
fifties but it was only in 1990 that interest in this material increased following the
discovery that porous silicon layers were able to emit bright red light This led to a large
amount of research and now different classes of micropores can be reproducibly created
36 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Fig 44 Different types of Silicon micropores From ref [14]
mostly for optical and micromachining applications However there is still no unified
theory able to explain the nucleation and growth mechanism of all the different kinds of
pores
We will give a short description of some of these theories for more detail see Parkhutik et
al [15]
One model explains pore nucleation on the basis of physical processes such as hole
positive charge migration ion transport to the surface and small perturbations on the
silicon surface modelled as Fourier components The system is shown to be unstable and
some spatial frequencies that lead to pore nucleation evolve from the etching process
A second model focusses on stationary pore growth without explaining the nucleation
stage According to this model silicon dissolves preferentially at pore edges because h+
charges are attracted by the stronger electric field present at these edges
A third class of models explains pore growth as a Diffusion Limited Aggregation (DLA)
process where the random walk of h+ charge carriers through the depleted layer present at
the silicon-liquid interface controls pore shape
Finally the model by Carstensen Cristophersen and Foll [16] proposes that areas of the
surface of some characteristic size LCO are etched by synchronized ldquocurrent burstsrdquo in the
flow of h+ charges These bursts dissolve silicon through cyclic stages of surface oxidation
oxide removal and hydrogen passivation Areas where a burst has recently taken place are
less passivated and thus more likely to be etched again in this way the pore bottom
continues to dissolve while the pore walls are passivated and are thus less favourable areas
towards current bursts
VPalermo 37
EXPERIMENTAL RESULTS
In the following sections we will show some experimental results obtained from STM and
AFM measurements of fluorine-treated Si(100) surfaces In the first part the results of
mild etching using concentrated and dilute HF solutions at low pH are presented In the
second part the results of etching at high pH using ammonium fluoride are presented and
the mechanism of pore formation discussed
Etching of Si(100) in dilute and concentrated HF
Samples were cut from different areas of an 8-inch diameter p-doped silicon(100) wafer
(10 Ω-cm) supplied by MEMC Electronic Materials Each series of STM measurements
was carried out over at least six different areas on at least two identical samples Low
Electron Energy Diffraction (LEED) was used to check the surface cleanliness of the
samples before STM measurements
Table 41 summarizes the different treatments of each sample After etching with
electronic grade HF each sample underwent a final rinse in Ultra-Pure Water (UPW
resistivity gt18 MΩ-cm) Both the HF and the UPW were allowed to flow continuously
over the sample surface Some samples were not etched with HF at all but just washed with
UPW to observe the morphology of the native oxide layer (~2 nm thick) covering the
surface All of these processes were carried out under nitrogen to limit reoxidation and the
samples were then introduced from the nitrogen atmosphere directly into the vacuum
chamber and degassed overnight at sim150degC before LEED and STM measurements
STM images were obtained from each sample using the same measurement parameters
(sample bias 4 V feedback current 1 nA scan speed 800 nm s-1) These parameters
38 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
provided a satisfactory level of reproducibility for all the samples Measurements were
made over an area of 500x500 nm (image size 500x500 pixel) Slope correction was
carried out by subtracting row-wise and column-wise fitted slopes from the entire image
which gave better results than the simple subtraction of a fitted plane especially for the
rougher samples Following slope correction the rms roughness
sum minus=xy
hyxhN
22 ))((1σ
and the 2-D Fourier transform
)(22
2
)(4
)( vyuxi
xyeyxhvuF +sum∆
= π
π
were calculated for each image where N2 is the number of pixels composing the image
h(xy) is the surface height at each point ∆ is the distance between points h is the mean
height and u v are the spatial frequencies The radial power spectrum PS(f) of the STM image data is obtained from the angular
average of the squared Fourier transform with f 2 = u2 + v2
Fig 45 shows the STM images obtained from the various samples Sample A still covered
with its native oxide layer shows an irregular surface with RMS roughness of ~05 nm
(see Table 41) Observing the sample with LEED gave no diffraction pattern even at
relatively high incident electron energies because of the surface oxide coverage After 1
min etching in dilute HF (sample B) the morphology is similar to the original one though
Table 41 Sample treatments average RMS roughness and integrated area of power spectra
Sample Treatment RMS roughness (nm)
PS area (f lt 01 nm-1)
PS area (f gt 01 nm-1)
A Rapid dip in water 053 plusmn 013 366 032
B 1 min in HF 5 + 10 min in water 051 plusmn 008 272 034
C 30 min in HF 5 + 10 min in water 062 plusmn 008 1064 038
D 5 sec in HF 49 + 10 min in water 042 plusmn 004 183 022
VPalermo 39
Fig 45 STM images of each group of samples showing the topography of the silicon surface A) noetching original oxide surface B) after 1 min etching in dilute HF C) after 30 min etching in diluteHF D) after dipping in concentrated HF Grey scale indicates height of the surface from lower (black) to higher (white) The images are 250x250 nm ie representative portions of the images usedfor the roughness measurement and PSD analysis
some of the larger features have disappeared and the image quality is better maybe due to
improved tunnelling due to the cleaner surface The RMS roughness is comparable to that
of the original surface Clear diffraction patterns are visible using LEED though at quite
high energies (200 eV) After prolonged etching (sample C) the RMS roughness increases
to 062 nm and a long-range corrugation is visible on the surface even if the LEED pattern
is good
The samples dipped in concentrated HF (D) reveal the bare SiSiO2 interface which has a
disordered aspect and protrusions over a wide range of dimensions The quality of the
STM images of sample D is very good probably due to the cleanliness of the surface
40 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 E -0 2
1 E -0 1
1 E + 0 0
1 E + 0 1
1 E + 0 2
1 E + 0 3
0 0 0 0 0 1 0 1 0 1 0 0f (1 nm )
nm^4
A a s re c e iv e d
B e tc h e d 1 min
C e tc h e d 3 0 min
D e tc h e d HF 4 9
Fig 46 Log-Log plot of the averaged power spectra of the STM images for all the samples
which gives a more stable tunnelling junction The LEED pattern is excellent showing
clear diffraction peaks at energies as low as 37 eV comparable to that obtained after high
temperature cleaning in UHV
Fig 46 shows the power spectra of the samples The high frequency and low frequency
areas of the power specturm are considered separately Table 41 shows for each sample
together with the roughness the integrated area of the power spectrum for the high and low
frequency part
We first examine the differences between the samples in the low frequency part of the
spectrum (lt01 nm-1) In this spatial range the short dip in concentrated or dilute HF
(samples BD) seems to lower the roughness of the sample removing some of the larger
features of the native oxide surface visible in Fig 45a and thus decreasing the
corresponding part of the power spectrum Sample C however etched with HF for 30 min
shows an increase in roughness peaking at 0012 nm-1 (~85 nm corresponding to the
typical dimensions of the corrugations visible in Fig 45c) On this scale sample C has a
PS density of 435 nm4 compared to 101 nm4 for sample A
Samples A B and C have the same PS in the high frequency range showing that the fine
structure of the surface is not changed by the HF etching Sample D on the other hand
shows a decrease of roughness for frequencies gt 015 nm-1 This effect could be due to the
VPalermo 41
improved surface cleanliness after etching with concentrated HF which would give a more
stable STM junction thus reducing the high frequency noise in the image
Fig 45 and the analysis of the power spectrum of each sample shows that a rapid dip in
HF removes the native oxide but does not lead to major changes in the morphology of the
Si surface its only effect being the removal of some of the larger features present on the
original surface Prolonged etching on the other hand increases the RMS roughness of the
surface
Etching of Si(100) in concentrated ammonium fluoride and nano-hole creation
Two different types of commercial p-doped Si(100) wafers (2Ω-cm and 10Ω-cm) from
MEMC were used Several different samples of 10x5 mm were immersed for 10 minutes
in 40 electronic grade NH4F solution under agitation Previous works used low
temperatures or anodic potentials applied to the silicon to avoid gaseous hydrogen
production and to obtain a uniform surface but in our experiment we worked at room
temperature to check the influence of hydrogen bubbles on surface morphology During
the etching the stirring was sufficient to provide a uniform concentration of reagents over
the whole sample surface but not to mechanically remove the hydrogen bubbles from the
silicon surface
After the etching each sample was rinsed in ultra-pure water to remove any etching
residues and observed with STM AFM and optical microscopy The AFM measurements
were made in air while for STM measurements the samples were rapidly dried with
nitrogen and inserted into the vacuum system to avoid surface reoxidation After insertion
into the vacuum surface cleanliness was checked with LEED and the surface morphology
observed by STM Parameters for STM measurements were sample bias +4 V 1 nA
current 16 Hz scan rate The images obtained were stable and reproducible over several
days Scan parameters for AFM were 20 nN force and 1 Hz scan rate
Some of the samples were cleaned with an RCA standard clean [1] before NH4F etching to
check the influence of possible surface contaminants on the final results Eight different
samples were prepared and more than sixty STM images of the samples were taken at
different points of the various samples
42 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
(110)
(110)
Fig47 abc) STMtopographic images of differentetching morphologies Eachimage is 500x500 nm Z-ranges are 10 10 and 18 nmrespectively d) STM image of a bridgecreated by etching of the lowerlayers of silicon (black arrow)Image is 250x250x6 nm
After ~2 min of immersion in the solution hydrogen bubbles become visible on the sample
surface The production is slow and the bubbles are quite stable on the sample without
detaching Thus some areas of the surface are masked from the liquid etching action
STM observations (fig 47) show that at the nanometer scale the surface is unevenly
covered with holes of radii ranging from 10 to 200 nm with depths of 2-4 nm These holes
have a wide range of different shapes and distributions In most cases the surface was
covered with a uniform distribution of round-shaped holes (fig47a) indicating isotropic
etching The dimension and the density of the holes changed greatly from sample to
sample and even over the surface of a single sample In some cases the etching was
anisotropic yielding nearly square holes and layered structures as shown in fig 47b
Square holes have been previously observed in cases where the etching speed in the (110)
direction is significantly smaller than in the (100) direction [17]
Over these areas (fig 47b and especially 47c) it is clearly visible how once the surface
had been attacked the reaction continued to preferentially remove atoms at step
irregularities (kink atoms) straightening step edges Eventually the exposed underlying
silicon was also attacked and further holes created inside the previously etched larger
ones It was not possible to detect monatomic steps on this kind of surface The smallest
step height observed was ~15 nm corresponding to several atomic layers In the image
shown in Figure 47a the etching was not strong and created only anisotropic holes on the
surface In fig 47b and c the stronger etch proceeded laterally for several tens of
VPalermo 43
nanometres leaving straight steps several tens of nanometers long and roughly rectangular
holes as expected given the structure of the (100) crystal face In some cases a
significative underetch is observed and the formation of suspended bridges and tunnels
can be deduced from the STM images (fig 47d)
The formation of branched pores and suspended structures has been attributed during pore
formation to diffusion limited aggregation effects where the h+ charge carriers necessary
for silicon etching have a higher probability of reacting at pore bottoms than reaching the
upper part of the silicon surface In the case of very deep pores quantum wire effects have
been invoked to explain the pore growth mechanism [15] In our case though the pores
formed were very shallow the underetch depth being only a few nanometres on pores of
sim100 nm width Thus more than diffusion effects the main contribution to the
underetching process must come from anisotropic etching and some kind of autocatalytic
reaction path analogous to the one described by Matsumura et al [4] with some areas of
the silicon surface hydrogenated and thus less vulnerable to etching
Pre-treatment with RCA cleaning has no effect on the final morphology and this seems to
exclude pore nucleation being caused by presence of metallic or organic surface
contaminants
The morphology and the distribution density of the pits was quite uniform over
microscopic areas of the sample but changes were observed over the millimetre scale This
suggests that etching intensity is influenced by some large-scale parameter
Large-area measurements made with AFM or with an optical microscope (fig 48)
showed that the inhomogeneity of surface etching can be correlated with the masking
action of the bubbles While the fluoride dissolved the silicon hydrogen bubbles formed
by the reaction covered some areas of the surface thus blocking the etching over that area
generating macroscopic steps at the bubble-liquid border As the reaction proceeded more
hydrogen accumulated and the bubble diameter increased producing in this way a circular
pattern of steps The increase in bubble diameter was not continuous with time otherwise a
uniform surface slope gradient would have been obtained The formation of this circular
ldquoetching staircaserdquo indicates that the bubble growth was stepwise the bubble accumulating
more and more hydrogen without enlarging across the surface until it relaxed increasing
its diameter stepwise and covering more silicon The circular structures in fig 48a are not
co-axial and their asymmetry could derive from physical processes due to stirring or
irregularities on the surface
44 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
a b Fig 48 a) optical micrograph of etching patterns on Si(100) created by NH4F 12x09 mm b) AFM image of the circles border xy range is 40x24 microm z-range is 30 nm
The step structure was not destroyed by the etching even after the bubble detached from
the surface but on the contrary the etching process seemed to be influenced by the
presence of the step
Observing in detail a series of steps (fig 48b) a quite deep trench is visible at the base of
each step A close-up image of a step and the corresponding line profile of the trench is
shown in fig 49 The trench is sim5 nm deep with respect to the lower surface compared to
a step height of 22 nm
A similar structure has been recently obtained with electrochemical etching of p-type
Si(100) in 4 HF [18] in which a ldquocurrent burstrdquo etching model previously described
was assumed for silicon dissolution In that experiment the trench was created at the
border of silicon nitride masks and began to grow after a nucleation stage Preferential
trench etching was along the (110) direction and the trench growth was explained as an
effect of mechanical stress induced by the nitride mask and of electric field enhanced
dissolution which depended upon an external applied potential
While it is clear that in our system the gas bubbles have a masking effect similar to a
classical solid nitride mask it is unlikely that hydrogen present on the surface can induce a
significant stress in the silicon lattice as in the case of a nitride mask Furthermore no
external field was applied to drive preferential etching at the trench site
It has been proposed [19] that the cathodic and anodic part of the etching reaction
(hydrogen production and silicon oxidative etching respectively) take place at different
points on the surface with a net charge transfer between the different areas In this case
the highest reaction rates will correspond to the silicon area surrounding the bubble border
where a high number of positive charges will be available for the reaction Furthermore a
sharp trench extending into the silicon crystal will be a preferential electrostatic attractor
VPalermo 45
Fig 49 AFM image of the etched surface showing a step created on the surface by bubblemasking A stronger etching action is visible on the right side of the step as well as theprotected area on the upper side of the step (indicated by the arrows) Image is 10x10 micromz-range is 30 nm The profile on the right is taken from the central area of the image
for the h+ charge carriers coming from other ldquocathodic areasrdquo of the sample either from
other regions on the surface or from the back of the silicon chip [18]
In the areas where the hydrogen bubble had detached and the surface was exposed to the
etching the reaction was not uniform in the neighbourhood of the steps It is possible to
observe (fig 48b and 49) an area on the upper side of the step where less or even no
etching at all seems to have taken place as if the step was able to protect the surface from
etching While etching on the lower side with trench formation can be attributed to the
presence of the bubble the surface on the upper step side can be etched only after bubble
detachment so no masking effect can account for this result However a further
preferential attraction of h+ charge carriers from the already formed trench can be
hypothised electrochemically shielding the surrounding area from further etching If this is
true the shielding effect would be very strong with a relatively shallow 5 nm-deep trench
protecting an area of sim1 microm parallel to the step
To summarise the etching of Si(100) in NH4F is a complex process in which different
reaction paths both chemical and electrochemical co-exist Hydrogen bubbles formed by
the reaction act as a mask on the surface and create etching paths and inhomogeneous
etching of the surface Different kinds of pores are observed on the surface and in some
cases the anisotropy of the process is so strong as to give square-shaped holes and
underetching
46 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The diffusion of h+ charge carriers in the crystal is one of the main rate-determining steps
of the reaction and leads to the formation of a deep trench immediately outside the bubble
perimeter These trenches act as charge collectors and reduce the etching of the upper step
surface in the proximity of the steps
Bibliography
[1] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[2] Spectroscopic Ellipsometry Studies Of HF Treated Si (100) Surfaces Yao H Woollam
Ja Alterovitz SA Applied Physics Letters 62 (25) 3324-3326 Jun 21 1993 Influence
Of HF-H2O2 Treatment On Si(100) And Si(111) Surfaces Graf D Bauermayer S
Schnegg A Journal Of Applied Physics 74 (3) 1679-1683 Aug 1 1993 Kinetics Of
Oxidation On Hydrogen-Terminated Si(100) And (111) Surfaces Stored In Air Miura
T Niwano M Shoji D Miyamoto N Journal Of Applied Physics 79 (8) 4373-4380
Part 1 Apr 15 1996
[3] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995 Structure Of The Stepped SiSiO2 Interface After Thermal-Oxidation -
Investigations With Scanning Tunneling Microscopy And Spot-Profile Analysis Of
Low-Energy Electron-Diffraction Pietsch GJ Kohler U Jusko O Henzler M Hahn
PO Applied Physics Letters 60 (11) 1321-1323 Mar 16 1992
[4] Enhanced Etching Rate Of Silicon In Fluoride Containing Solutions At pH 64
Matsumura M Fukidome H Journal Of The Electrochemical Society 143 (8) 2683-
2686 Aug 1996
[5] A Study Comparing Measurements Of Roughness Of Silicon And SiO2 Surfaces And
Interfaces Using Scanning Probe Microscopy And Neutron Reflectivity Crossley A
Sofield CJ Goff JP Lake ACI Hutchings MT Menelle A Journal Of Non-Crystalline
Solids 187 221-226 Jul 1995
VPalermo 47
[6] Step-Flow Mechanism Versus Pit Corrosion - Scanning-Tunneling Microscopy
Observations On Wet Etching Of Si(111) By Hf Solutions Hessel HE Feltz A Reiter
M Memmert U Behm RJ Chemical Physics Letters 186 (2-3) 275-280 Nov 8 1991
[7] Comparison Of Si(111) Surfaces Prepared Using Aqueous-Solutions Of NH4F Versus
HF Higashi GS Becker RS Chabal YJ Becker AJ Applied Physics Letters 58 (15)
1656-1658 Apr 15 1991
[8] Wet Chemical Etching Of Si(100) Surfaces In Concentrated NH4F Solution -
Formation Of (2x1)H Reconstructed Si(100) Terraces Versus (111) Faceting Neuwald
U Hessel HE Feltz A Memmert U Behm RJ Surface Science 296 (1) L8-L14 Oct
10 1993
[9] Ideal Hydrogen Termination Of Si(001) Surface By Wet-Chemical Preparation Morita
Y Tokumoto H Applied Physics Letters 67 (18) 2654-2656 Oct 30 1995
[10] Atomic Structures Of Hydrogen-Terminated Si(001) Surfaces After Wet Cleaning
By Scanning Tunneling Microscopy Endo K Arima K Kataoka T Oshikane Y Inoue
H Mori Y Applied Physics Letters 73 (13) 1853-1855 Sep 28 1998
[11] On The Potential-Dependent Etching Of Si(111) In Aqueous NH4F Solution
Houbertz R Memmert U Behm RJ Surface Science 396 (1-3) 198-211 Jan 20 1998
[12] Etching Process Of SiO2 By HF Molecules Hoshino T Nishioka Y Journal Of
Chemical Physics 111 (5) 2109-2114 Aug 1 1999
[13] Etching Mechanism And Atomic-Structure Of H-Si(111) Surfaces Prepared In
NH4F Allongue P Kieling V Gerischer H Electrochimica Acta 40 (10) 1353-1360
Jul 1995
[14] Pore Formation Mechanisms For The Si-HF System Carstensen J Christophersen
M Foll H Materials Science And Engineering B-Solid State Materials For Advanced
Technology 69 23-28 Sp Iss Si Jan 19 2000
[15] Porous Silicon - Mechanisms Of Growth And Applications Parkhutik V Solid-
State Electronics 43 (6) 1121-1141 Jun 1999
[16] Formation And Application Of Porous Silicon Foll H Christophersen M
Carstensen J Hasse G Materials Science amp Engineering R-Reports 39 (4) 93-141 Nov
1 2002
[17] Pore Morphology And The Mechanism Of Pore Formation In N-Type Silicon
Searson PC Macaulay JM Ross FM Journal Of Applied Physics 72 (1) 253-258 Jul 1
1992
48 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
[18] Deep Electrochemical Trench Etching With Organic Hydrofluoric Electrolytes
Christophersen M Merz P Quenzer J Carstensen J Foll H Sensors And Actuators A-
Physical 88 (3) 241-246 Mar 5 2001
[19] Morphology Of Anodically Etched Si(111) Surfaces - A Structural Comparison Of
NH4F Versus HF Etching Houbertz R Memmert U Behm RJ Journal Of Vacuum
Science amp Technology B 12 (6) 3145-3148 Nov-Dec 1994
VPalermo 49
5 Surface modification of silicon in vacuum void creation and oxide desorption
The main reason for the huge success of silicon in the microelectronics industry is not due
to its superior properties as a semiconductor Other materials for example germanium
have better qualities such as higher mobility of charge carriers and lower noise levels
which would allow the construction of faster and higher performance devices
The widespread use of silicon however is mainly due to the outstanding characteristics of
its oxide Silicon dioxide (SiO2) is a very good electrical insulator easy to form
chemically and thermally stable and is compatible with lithographic and metal deposition
processes Germanium oxide on the contrary is too reactive to be used
Even the use of Si(100) substrates for nearly all microelectronic devices is dictated by
oxide quality The (111) face of silicon crystal can be easily cleaved and flattened and
almost atomically perfect surfaces can be obtained with simple chemical procedures (as
described above) But the density of interfacial defects is highest for Si(111)-SiO2
interfaces and lowest for Si(100)-SiO2 ones so microchips will continue to be fabricated
on Si(100) wafers
SiO2 (silica) is present in 95 of the earthrsquos minerals in different allotropic forms such as
quartz tridymite and cristobalite In the bulk each silicon atom is bonded to four oxygens
in a Si-O-Si tri-dimensional network Si-O bonds are 016 nm long and form an angle
ranging from 120deg to 150deg
Three typical intrinsic defects are present in SiO2 The so-called Ersquo centres are oxygen
vacancies with a hole localised on a silicon atom with only three Si-O bonds
50 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
O3Si +SiO3 Whereas the PR (peroxy radical) defects are holes trapped by a charged
peroxy moiety with a O3Si-O-O+ SiO3 structure The NBOHC (non-bridging oxygen hole
centres) derive from water or hydrogen contamination and are schematized as O3Si- O- H-
O-SiO3
The atomic structure of the Si-SiO2 interface varies enormously Local domains resembling
the tridymite and the cristobalite structure of silica are present but it seems that only 10
of the interface is ordered [1] Far from the interface the SiO2 bulk is completely
disordered The passage from bulk Si to stoichiometric SiO2 passes through a non-
stoichiometric SiOx layer 07 nm thick
When a clean silicon surface is exposed to atmospheric oxygen a thin sim2 nm thick layer
of native oxide forms spontaneously which is usually removed and substituted with
thicker better quality oxide layers before further processing
Silicon is usually oxidised by thermal annealing at temperatures between 800deg and 1100deg
in an atmosphere of pure O2 with some water eventually added to increase oxidation speed
Thermal oxides made in pure oxygen (dry oxides) grow more slowly than oxides produced
in an oxygen-water atmosphere (wet oxides) but are usually of better quality
According to the Deal-Groove formula the time t needed to grow an oxide of thickness X
is given by [1]
1
212minus
minusminus
+=
ABXBXt α
where the constant B and BA decrease exponentially with temperature as
minus
kTEexp
with activation energies for dry oxidation of EB =123 and EBA =20 eV respectively
EB is related to the diffusion of oxygen in silicon while the value of EBA is interpreted as
the energy required to break a Si-Si bond The exponent α is 1 for wet oxidation and 0 for
oxidation at high temperatures and low oxygen pressures It has intermediate values for dry
oxidation This formula does not work well for low values of X at the initial stages of
oxidation and usually empirical corrections are used
An interesting characteristic of silicon is that at high temperature and in vacuum oxygen
can actually etch the silicon crystal giving gaseous products and the oxidized layer present
on the crystal becomes unstable (fig 51) [2]
VPalermo 51
1E-10
1E-09
1E-08
1E-07
1E-06
1E-05
1E-04
1E-03
1E-02
1E-01
0607080911112
1000T (1K)
P (T
orr)
SiO2 + Si rarr 2SiO(g) Oxide decomposition
Si+O2 rarr SiO2(s) oxide formation Fig51 phase diagram of the
oxygen-silicon system
Silicon oxidation apart from the initial nucleation stages at the monolayer level proceeds
uniformly over the whole surface with a planar reaction front moving from the surface
into the bulk
If heated under low oxygen partial pressure (vacuum or inert atmosphere) SiO2 is known to
decompose following the reaction
SiO2 + Si rarr 2SiOuarr (1)
The reaction begins with nucleation at defect points on the SiSiO2 interface and proceeds
in a spatially inhomogeneous manner with the formation of large voids on the oxide
surface [3]
Several studies have been made on the dynamics of void growth both on thick [4] and thin
[5] layers of SiO2 The process has been used to decorate otherwise unobservable defects at
the SiSiO2 interface [6] or to grow nanoislands of silicon on the void surface [7] It has
been suggested that the defects acting as nucleation centres could be metallic contaminants
present on the native surface which aggregate and catalyze SiO2 decomposition [8]
After oxide desorption the silicon surface is very rough In particular on Si(100) square
islands are observed several nanometres high which act as lsquopinning sitesrsquo on the motion of
monatomic steps on the Si surface It has been proposed that these islands can be composed
of silicon carbide coming from the organic contaminants present on the surface [9] or of
oxide clusters during partial reoxidation [10]
In the following section some experimental results obtained studying voids and nanoisland
growth are presented
52 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nanoisland growth on silicon
We used two different samples the first rougher than the second (i) p-type Czochralski Si
(100) with a RMS roughness of 05 nm shown in fig52a and (ii) p-type epitaxial Si(100)
with a RMS roughness of 019 nm shown in fig 52b The roughness was measured with
STM on the lsquoas receivedrsquo samples on different areas of 500x500 nm Every sample was
covered by a layer of native oxide ~2 nm thick
Some of the samples (AC) were introduced into the UHV system without any cleaning
others (BD) were dipped in concentrated HF (49) to remove the surface oxide without
etching the silicon [11]
After degassing the samples were heated resistively increasing the temperature slowly to
keep the pressure within the 10-10 mbar range during heating Each sample was held at
900˚C for 30 min For some of the samples the heating was stopped at lower temperatures
to obtain incomplete oxide desorption and surface roughening thus enabling observation
of the different steps of the process Table 51 summarizes the different treatments for each
group of samples
Fig52 Original surfaces of Czochralski (a) and epitaxial (b) silicon Every image is 500x500 nm
VPalermo 53
Table 51 Summary of different treatements
Silicon type Oxide removed
Heating Islands density (microm-2)
Av Island Volume (nm3)
A Czochralski No 30rsquo at 900˚C 24x103 353
B Czochralski Yes ldquo 20x104 71
C Epitaxial No ldquo 28x102 2450
D Epitaxial Yes ldquo 23x104 63
Fig53 ab) Surface roughening on group A samples after heating 30 min at 800deg and900deg respectively in presence of an oxide layer Image size 500x500 nm cd) Surface roughening on group B samples after heating 30 min at 700deg and 900degrespectively after removing the oxide layer with HF Image size 250x250 nm
Fig 52a is the native oxide surface of sample A quite disordered and irregular This
surface was stable when heated to 700˚C while at higher temperatures the oxide began to
desorb Fig 53a was taken after heating the sample at 800˚C for 30 min
54 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
The oxide began to desorb in correspondence with defect points creating oxide-free
surface voids which enlarge radially uncovering the silicon surface Some surface silicon
atoms are removed through reaction (1) Mobile silicon atoms moving randomly over the
surface aggregate forming a nanocrystal at the initial defect point surrounded by a stable
flat oxide-free surface with some steps The initial SiSiO2 interface as observed after
oxide etching with HF [12] is similar to the one shown in fig 52a and thus the surface
observed in the void area is a completely new one coming from surface reorganization
The reaction continues leading to a coalescence of the voids the final result of the process
is shown in fig 53b with protruding islands aligned along the main crystal axes and flat
areas with some steps LEED measurements confirmed that this surface is crystalline
silicon Atomic resolution was obtained on the flat areas of the surface showing the
typical 2x1 reconstruction of Si(100) Figures 53c and 53d show the evolution of the
roughening on sample B after chemically removing the native oxide layer before
introducing the sample into UHV As mentioned above the initial morphology is similar to
that shown in fig 52a though the oxide has been removed as confirmed by LEED
measurements Fig 53c shows the surface after heating at 700˚ for 30 min Without the
oxide layer the surface began to reorganize at lower temperatures and the small irregular
grains of the original surface begin to grow and become more rectangular After heating to
900˚C (fig 53d) the surface is flat with small rectangular islands The shape of the islands
is similar to that shown in fig 53b but the lateral island dimension is one order of
magnitude smaller
To assess the importance of the initial surface on the process the same treatments were
repeated starting from flat epitaxial silicon (samples CD) The initial epitaxial surface is
smooth with periodic steps due to a small miscut angle (fig52b) On heating the epitaxial
silicon without etching (sample C) relatively large islands are obtained similar to the ones
shown in fig 53b where the ldquostep-pinningrdquo effect of the islands is evident The average
dimension of sample C islands was sim2500 nm3 while only 353 nm3 for sample A Island
density decreased from 24000 to 280 microm-2 This result is maybe due to a lower
concentration of defects in the oxide layer of the epitaxial silicon which allows the
expanding void to grow much larger before coalescence with neighbouring voids the
nanostructures within them thus reaching greater dimensions To verify this hypothesis we
applied the same treatment to clean not-so clean and purposely contaminated samples To
check that the defects inducing void nucleation and nanoisland growth derive from carbon
VPalermo 55
contamination we used electronic-grade methanol to contaminate the silicon surface
Figure 54 shows the results of this comparison
State-of-the-art cleaned samples obtained with multi-step RCA cleaning [13] showed the
development of very few voids the oxide desorption leaving large smooth areas of oxide-
and nanostructure-free silicon as shown in Fig 54a where the presence of residual oxide
can be seen Samples cleaned with a simpler one-stage Pirana cleaning solution showed a
higher number of islands per unit area (Fig 54b) and samples purposely contaminated
with methanol (Fig 54c) had the highest density of island nucleation of all three samples
When the native oxide had been chemically removed from the epitaxial silicon before
heating (sample D) the island size and density is similar to the standard as-received
silicon case (sample B) Thus starting from a flat substrate the process of surface
roughening and island creation was the same even when the oxide layer had been
previously removed
From these results some important indications can be drawn The first is that oxide
desorption temperature depends upon the contamination levels of the oxide layers (fig 54)
with desorption at lower temperatures for contaminated surfaces
Second once the oxide layer is removed the silicon atoms become mobile on the surface
at relatively low temperatures (~700˚C) Their initial configuration is disordered and has a
high surface energy so they crystallize in the small islands shown in fig53d reducing
their surface area and lowering the energy of the system When the oxide is present the
surface cannot reorganize at 700˚C and is stable At 800˚C the oxide begins to desorb in
correspondence with defect points on the surface As the void area enlarges a small
cba Fig 54 The effect of contamination on nano-island production Gradient-filtered STM images of (a)sample cleaned two times with RCA 400x400 nm (b) Sample cleaned with a Pirana solution 500x500nm (c) sample contaminated with MeOH before insertion into the UHV system 500x500 nm Annealingtime is 40 min at 800degC for all the samples
56 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
crystallite grows at the centre of the voids Further oxide desorption uncovers larger areas
freeing more silicon atoms which feed the initial island making it bigger
Fig 55 shows the scaled size distribution of the islands for each group of samples Every
distribution has been obtained using several images of different samples for each group
The size distribution of all the samples follows an exponential decay and the decay seems
the same for all the samples though the average volume of the islands differs by more than
one order of magnitude (see Table 51 for details)
The point defects which catalyze void nucleation can be metallic contaminants present on
the initial oxide surface [8] or organic contaminants which at high temperatures can
form SiC nanocrystals [14] We used Scanning Tunneling Spectroscopy (STS) to look for
differences between the islands and the surrounding flat silicon surface but no difference
was found Furthermore the total island volume per surface unit is very high (more than
104 nm3microm2) and it seems unlikely that such a huge volume could consist of surface
contaminants Islands of pure silicon on silicon have been grown without evidence of
surface contamination[14]
To summarise the formation of nano-sized islands on silicon through surface diffusion
was studied At high temperatures the oxide layer covering the surface decomposes non
uniformly and circular voids of clean silicon are created The presence of the oxide layer
blocks surface silicon atomic motion and surface reorganization except within the voids
causing the growth of islands more than 10 nm high and 30 nm wide on the silicon surface
According to STM and LEED measurements we can say that the islands are mostly
Fig 55 Size distribution (scaled) for the samples described in table 1 The line is a reference for the eye
VPalermo 57
composed of silicon with traces of other substances such as carbon Final island densities
and dimensions depend upon the initial purity of the oxide layer
On the other hand if the oxide is removed chemically before heating the island growth
process is not so localised and takes place simultaneously over the whole surface yielding
smaller and more numerous islands
58 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Bibliography
[1] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[2] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[3] High-Temperature SiO2 Decomposition At The Sio2Si Interface Tromp R Rubloff
GW Balk P Legoues FK Physical Review Letters 55 2332-2335 Nov 1985
[4] Defect Microchemistry At The SiO2Si Interface Rubloff GW Hofmann K Liher M
Young DR Physical Review Letters 582379-2382 Jun 1987 Defect Formation In
Thermal SiO2 By High Temperature Annealing Hofmann K Rubloff GW Mccorkle
RA Applied Physics Letters 49 1525-1527 Dec 1986 Kinetics Of High-Temperature
Thermal Decomposition Of SiO2 On Si(100) Liher M Lewis JE Rubloff GW Journal
Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 5 1559-1562 Aug
1987
[5] Thermal-Decomposition Of Very Thin Oxide Layers On Si(111) Kobayashi Y Sugii
K Journal Of Vacuum Science amp Technology A-Vacuum Surfaces And Films 10 (4)
2308-2313 Part 3 Jul-Aug 1992 Scanning Tunneling Microscope Study On Mid-
Desorption Stages Of Native Oxides On Si(111) Kobayashi Y Sugii K Journal Of
Vacuum Science amp Technology B 9 (2) 748-751 Part 2 Mar-Apr 1991 Controlled
Growth Of SiO2 Tunnel Barrier And Crystalline Si Quantum Wells For Si Resonant
Tunneling Diodes Wei Y Wallace RM Seabaugh AC Journal Of Applied Physics 81
(9) 6415-6424 May 1 1997
[6] Defect Formation In SiO2Si(100) By Metal Diffusion And Reaction Liher M
Dallaporta H Lewis Je Appl Phys Lett 53 589-591 Aug 1988 SiO2 Film
Decomposition Reaction Initiated By Carbon Impurities Located At A Si- SiO2
Interface Raider Si Herd Sr Walkup Re Applied Physics Letters 59 (19) 2424-2426
Nov 4 1991
[7] Nanometer-Scale Si Selective Epitaxial Growth On Si(001) Surfaces Using The
Thermal Decomposition Of Ultrathin Oxide Films Fujita K Watanabe H Ichikawa M
Applied Physics Letters 70 (21) 2807-2809 May 26 1997 Pyramidal Si Nanocrystals
VPalermo 59
With A Quasiequilibrium Shape Selectively Grown On Si(001) Windows In Ultrathin
SiO2 Films Shibata M Nitta Y Fujita K Ichikawa M Physical Review B 61 (11)
7499-7504 Mar 15 2000
[8] Stacking-Fault-Induced Defect Creation In SiO2 On Si(100) Liher M Bronner Gb
Lewis Je Appl Phys Lett 52 1982-1985 May 1988
[9] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990 AFM And XPS Characterization Of The
Si(111) Surface After Thermal-Treatment Lamontagne B Guay D Roy D Sporken R
Caudano R Applied Surface Science 90 (4) 481-487 Dec 1995
[10] Scanning-Tunneling-Microscopy Study Of Oxide Nucleation And Oxidation-
Induced Roughening At Elevated-Temperatures On The Si(001)-(2x1) Surface Seiple
JV Pelz JP Physical Review Letters 73 (7) 999-1002 Aug 15 1994 Evolution Of
Atomic-Scale Roughening On Si(001)-(2x1) Surfaces Resulting From High-
Temperature Oxidation Seiple JV Pelz JP Journal Of Vacuum Science amp Technology
A-Vacuum Surfaces And Films 13 (3) 772-776 Part 1 May-Jun 1995
[11] Hydrogen On Si - Ubiquitous Surface Termination After Wet-Chemical Processing
Pietsch GJ Applied Physics A-Materials Science amp Processing 60 (4) 347-363 Apr
1995
[12] Morphological Changes Of The Si [100] Surface After Treatment With
Concentrated And Diluted HF Palermo V Jones D Materials Science In
Semiconductor Processing 4 (5) 437-441 Oct 2001
[13] The Evolution Of Silicon-Wafer Cleaning Technology Kern W Journal Of The
Electrochemical Society 137 (6) 1887-1892 Jun 1990
[14] Nanoscale Roughening Of Si(001) By Oxide Desorption In Ultrahigh Vacuum
Gray SM Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology
B 14 (2) 1043-1047 Mar-Apr 1996
60 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 61
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and nano-lines
When the silicon surface is not protected by a native oxide layer or by a hydrogenated
passivating layer it is very reactive The surface chemistry of both Si(100) and Si(111) has
been extensively studied starting from clean surfaces prepared in UHV because of its great
relevance to the microelectronics industry and its technological interest The diffusion and
reaction of molecules and atoms on silicon is an interesting scientific problem on its own
apart from technological issues because silicon reconstruction yields a very complex and
anisotropic surface
We give here a brief summary of the adsorption behaviour of several elements on Si
surfaces The summary is not comprehensive and only the most interesting characteristics
for each substance are given
Table 61 Summary of adsorption behaviour of atoms and molecules on silicon surfaces [12]
Hydrogen
Molecular H2 shows low reactivity towards silicon while atomic hydrogen easily forms
Si-H bonds and can even break Si-Si bonds
Alkali metals
Alkali metals diffuse rapidly into SiO2 and can damage silicon-based transistors
Transition
metals
All transition metals apart from gold and silver react with Si forming metal silicides
62 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Nickel A very common contaminant of silicon Nickel from even slight contact with stainless
steel tweezers can contaminate silicon samples forming its silicide and is very difficult
to remove even at high temperatures because it segregates on the silicon surface
Palladium Forms silicides especially Pd2Si and can be used to make contacts possessing a low
Schottky barrier
Titanium Widely used to fabricate contacts for silicon-based microdevices TiSi2 is one of the
more stable and highly conductive silicides
Tungsten The silicide is more stable than TiSi2 on polysilicon and is thus used for polysilicon
gate contacts
Platinum Silicide is used for bipolar transistors PtSi contacts on n-silicon give Schottky diodes
Cobalt Cobalt silicide gives better quality contacts than TiSi2 at the nanometrelevel but is less
used at the moment because it consumes too much silicon during its growth
Iron Silicides have been studied for potential optoelectronic applications because FeSi2 has a
directly accessible bandgap and is optically active
Group 13
Group 13 metals are used in IC technology to dope silicon (giving p-type doping) and
to make III-V type semiconductors When evaporated on Si(100) or Si(111) they react
strongly with the surface giving surface reconstruction and kink formation at steps
Boron Widely used for doping Usually deposited with decomposition of boron hydrides
Sticking coefficient of simple hydrides is very low so decaboranes (B10H14) are used for
deposition
Aluminum Aluminum is widely used for microelectronic contacts but in some cases it reacts with
silicon diffusing into the crystal and with SiO2 stealing oxygen atoms to form Al2O3
Group 14
Carbon Carbon can form a composite semiconductor with silicon (silicon carbide) widely used
to fabricate integrated circuits for use at elevated temperatures or in the presence of
ionising radiation When carbon is deposited on silicon usually polycrystalline films
with properties between diamond and graphite are obtained
CO adsorbs molecularly on silicon without breakage of the C=O bond Other
unsaturated hydrocarbons do not react with silicon Even very reactive strained
molecules like cyclopropane have low sticking coefficient Ethylene acetylene and
propylene adsorb molecularly on Si(100) each molecule sitting on a silicon dimer
interacting with the dangling bonds of the surface Acetylene also adsorbs molecularly
on Si(111) on the 7x7 reconstructed surface preferential adsorption on centre atoms
with respect to corner atoms is observed Benzene and other aromatic ring compounds
adsorb molecularly on the surface They can form σ or π-type bonds with the silicon
laying parallel or tilted respect to the surface according to the bond type
Silicon Silicon atoms can be deposited on the surface with molecular beam epitaxy or
VPalermo 63
decomposition of silanes and chlorosilanes Silanes with single Si-Si bonds have a high
sticking probability and dissociate upon adsorption giving trihydrides and
monohydrides
Chlorosilanes have high sticking coefficients too and decompose on the surface but
they can etch the surface through the reaction SiCl4 + Si rarr 2SiCl2
Germanium Germanium has a lattice constant similar to silicon (Ge lc is 4 larger than Si) so that
various alloys of Si1-xGex can be formed The growth of Germanium on silicon is quite
peculiar with several monolayers adsorbing uniformly on the surface followed by
island formation (Stranski-Krastanov growth)
Group 15
Nitrogen Exposure to ammonia and nitrogen gas at high temperatures leads to the formation of
silicon nitride layers (Si3N4) NH3 dissociates on the surface and reacts with the
dangling bonds giving Si-NH2 and Si-H termination Silicon nitride is stable on the
surface up to 1100degC above this temperature it desorbs as Si2N
Phosphorous Phosphine (PH3) adsorbs dissociatively as Si-PH2 and Si-H and behaviour is similar to
that of ammonia
Arsenic A monolayer of As is often deposited on silicon as a substrate for GaAs growth Arsenic
forms dimers on Si(100) creating dimer rows which are parallel (perpendicular) to the
underlying silicon dimer rows when it is deposited at high (low) temperature On
Si(111) Arsenic breaks the 7x7 reconstruction giving a 1x1 pattern This 1x1 As
monolayer acts as a passivating layer and can resist further As adsorption oxygen and
air
Antimony and
Bismuth
Due to their large covalent radii Sb and Bi form only short dimer rows on silicon and
only at high temperatures
Group 16
Oxygen
Oxygen can oxidize or etch silicon according to the reactions
Si(s) + O2 rarr SiO2 (s)
Si(s) + frac12O2 rarr SiO(g)uarr
Silicon oxidation has already been described in the previous chapter Molecular oxygen
adsorption on silicon is mostly dissociative O atoms break the Si-Si bonds forming an
Si-O-Si bridge or a peroxy bridge Si-O-O-Si Hydrided silicon surfaces do not have
surface dangling bonds and are thus more protected from oxygen attack Etching is
favoured at high temperatures and low oxygen pressures [3] At T gt700degC silicon oxide
on silicon decomposes as SiO(g)
Water
H2O adsorbs dissociatively on silicon easily on Si (100) and with more difficulty on
Si(111) The molecule decomposes giving Si-OH and Si-H on adjacent sites
64 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Group 17
Fluorine
Fluorine adsorbs and reacts easily on silicon dissociating on the surface and forming
Si-F bonds It can then penetrate the surface and attack polarised Si-Si backbonds The
surface is etched with production of SiF4 and SiF2
Chlorine
Chlorine reacts aggressively with silicon etching the surface Cl2 and F2 are often used
commercially to etch silicon usually in a plasma Atomic Cl and Cl2 adsorb on Si(100)
up to saturation The most stable configuration seems to be a Cl atom bonded to each Si
atom of a surface dimer in a symmetric structure but metastable configurations with
two Cl atoms bonded to a buckled dimer and even a central Cl atom bridged across a
dimer have been observed
On Si(111) silicon mono- di- and tri-chlorides are formed In particular heating at
Tgt400degC a uniform Si-Cl monolayer is obtained and the 7x7 reconstruction changes
into a bulk like 1x1 lattice The surface transition induced by Cl adsorption is fully
reversible The 7x7 reconstruction can be restored on desorbing the chlorine by heating
at Tgt1100degC The 1x1 domains nucleate at the lower terrace side of steps the 7x7 at the
upper step edge
Bromine
Like hydrogen bromine maintains the 2x1 structure of Si(100) forming Si-Br bonds
with the dangling bonds of the silicon dimer rows At high exposure some etching of
silicon by formation of volatile SiBr3 species has been observed
Surface diffusion on silicon
In the previous chapter we observed nano-island growth on silicon surfaces with native
oxide and surfaces contaminated with organic impurities While this phenomenon has been
observed in several experiments the formation and growth mechanisms of nanoislands has
still not been satisfactorily described theoretically Various substances have been
hypothesised as nanoisland nucleating agents such as organic or metallic contaminants
present on the surface or oxide clusters (see previous chapter) The island growth process
has been attributed by various workers to lsquosome sort of kinetic instabilityrsquo [4] to the
pinning of step flow by SiC clusters [5] or to a mesoscopic atom flux from areas with low
step density to areas of higher step density [6]
One mechanism proposed to explain surface roughening and island growth is the Ehrlich-
Schwoebel effect ie the presence of an energetic barrier that reflects atoms approaching a
VPalermo 65
Molecule (L
O2
H2O MeOH
CO CO2 CH4
clean surface
downward
weak on si
The aim o
surface an
To obtain
the ones
spontaneou
first obser
surface W
nanoisland
Experime
We used S
removed f
41 mixtur
water and
whole hea
1200degC re
kept below
Table 62 Treatment island density and island volume for each molecule tested
Dose angmuir)
Heating time at 800degC (min)
Island density (microm-2)
Ave Volume (nm3)
Equivalent layer thickness (nm)
280 10 no islands - -
90 10 no islands - - 30 10 2000 19 004
180 10 360 516 019 30 60 no islands - -
180 10 no islands - - - 10 no islands - -
step [7] Recent experiments however indicate that the Schwoebel effect is
licon [48]
f our experiments was to study the reaction of simple molecules with the silicon
d to check the ability of these molecules to nucleate nanoisland growth
a better understanding of the process we chose simple molecules starting from
which are more likely to be present as traces in UHV chambers where
s nanoisland growth is often observed after sample heating Using STM we
ved at the atomic level the possible interactions of these molecules with the
e then heated the surface and checked the density and dimensions of the
s produced
ntal procedures and results
i(111) wafers p-doped 0015 Ω cm Gross contamination and particles were
rom the sample surface with a standard Pirana etch (15 min dip in a H2S4H2O2
e at T=80degC) After this the samples were thoroughly washed in ultra-pure
introduced into the UHV system After degassing for 5 hours at 600degC the
ting stage was allowed to cool down and a sequence of rapid flashes to 1100deg-
moved the native oxide layer and cleaned the surface Pressure during flashes is
1x10-9 mbar
66 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Surface flatness and cleanliness were checked with STM and the contaminant molecules
introduced into the UHV system through a leak valve Doses of several Langmuirs were
used for the experiments (1 Langmuir=10-6 torr sec-1) Pressure was kept constant at 10-7
mbar during adsorption The STM tip was retracted during gas deposition to avoid any tip
shielding effect After the adsorption process the leak valve was closed the base pressure
in the UHV chamber returning to below 10-10 mbar and the presence of adsorbed
molecules checked on the surface with STM Finally the sample was heated at 800degC for
several minutes to allow nanoisland growth Several different areas were examined on each
sample
Table 62 summarizes the dose of exposure and the heating time for each molecule Other
experiments were carried out using smaller dosing times but only the significant results
are reported here
As expected oxygen and water react with the surface oxidising it The adsorption process
can be easily followed with the STM by lowering the tip from time to time and taking an
image No effect of the STM tip is observed over the scanned areas Upon molecule
adsorption the order of the crystalline surface rapidly degenerates and after several
minutes the surface looks completely covered by irregular atomic-sized protrusions
corresponding to Si-OH or Si-O-Si species On heating to 800degC the contaminating species
were easily desorbed through SiO formation the surface became clean and the crystal
surface periodicity re-established No residues or nanoislands were observed
Amongst the molecules tested was methanol CH3OH reacts readily with the surface and
covers it with a disorderd layer already after sim 30 Langmuirs exposure (fig 61)
Fig 61 Si(111) surface during methanol adsorption at 0 1 and 5 minutes respectively (corresponding to 0 6 and 30 L exposure) The dark irregular lines are borders between different 7x7 domains Image size 60x60 nm
VPalermo 67
Synchroton radiation photoemission spectroscopy experiments by Carbone et al [9]
showed that methanol reacts with the rest atoms of the 7x7 silicon cell already after 1
Langmuir exposure but that longer exposures are needed for methanol to react with the
remaining silicon atoms Methanol adsorbs dissociatively forming SiOCH3 and Si-CHx
species on the surface
After methanol adsorption the sample is heated to 800degC and observed again with STM
As in the case of water and oxygen the heating restores the crystalline surface and the 7x7
pattern returns visible But this time quite a dense array of nanoislands is observed on the
surface (fig 62) The islands have an average diameter of 20 nm and a height of sim2 nm
At temperatures above 400degC the Si-OCH3 and Si-CHx species decompose oxygen and
hydrogen are desorbed and according to [9] carbon atoms remain as SiC dispersed
uniformly on the surface Our STM measurements indicate that the carbide does not
randomly cover the surface but that C atoms are concentrated at the nanoislands Rough
calculations based on island volume suggest that the islands must be composed of a SixC1-x
alloy with x varying between 05 and 1
Following these results we expected carbon monoxide to behave in a similar manner
adsorbing onto the surface and with nanoisland growth However no adsorption was
observed with STM and the surface looked perfectly clean even after a dose of more than
100 L of carbon monoxide Some species did adsorb onto the surface though because
after 10 minutes subsequent heating nanoisland growth was observed Island density in
this case was only 18 of the density obtained with methanol and island dimensions are
Fig62 Nanoislands on Si(111) created after 30 L adsorption of methanol and 10 min heating at 800degC Image size 200x200 nm
68 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
much bigger (see table 62 and fig 63) To check whether the islands had been nucleated
by some other contaminant present in the vacuum chamber we heated a blank sample
without introducing any molecule No island growth was observed Exposing the surface to
other simple molecules including carbon dioxide and methane also gave no nanoisland
formation
We can estimate the amount of carbon deposited on the surface during gas dosing The
total flux of molecules hitting the surface is obtained from the Hertz-Knudsen formula
TkmPF
π2=
Fig 63 STM images of Si(111) after exposure to methanol (left) and carbonmonoxide (right) contaminating molecules and subsequent nanoisland growthImage size 400x400 nm The size distribution of nanoislands is shown below eachimage
VPalermo 69
Fig64 Left Si(111) after 180 L contamination of CO and 16 hours heating at 800degC Right Si(111)without CO contamination after 16 hours heating at 800degC Image size 250x250 nm Vertical scale is thesame for both images z-ranges are 35 and 5 nm respectively
where P is the pressure in Pascal k=138x10-23 J K-1 is the Boltzmann constant T is the
temperature and m is the molecular weight (sim32 for methanol and sim28 for CO) in
kilograms
Using this formula we obtain a flux of 027 molecules nm-2 sec-1 for methanol and 029
molecules nm-2 sec-1 for CO After ten minutes exposure at 10-7 mbar pressure more than
150 molecules will have hit each square nanometer of the surface Surface density of
Si(111) 7x7 is sim16 atomsnm2 so each surface atom will be hit by several molecules which
could react or be adsorbed even assuming a sticking coefficient much lower than unity
To explain nanoisland growth we hypothesize that CO molecules adsorb molecularly and
-5
0
5
10
15
20
25
30
0 20 40 60 80 100 120
nm
nm
10 min40 min16 h16 h- no CO
Fig65 STM profiles of nanoislandsgrown for different annealing times The lower curve correspond to the surface heated at 16 hours without CO contamination
70 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Table 63 Island density and volume for the CO-contaminated surface at different heating times For each sample the thickness of a uniform layer having the same volume per unit area of the islands is calculated
Molecule Dose
(Langmuir) Heating time at
800degC (min) Island density
(microm-2) Ave Volume
(nm3) Equivalent layer thickness (nm)
CO 180 10 362 517 019 CO 180 40 550 1072 059 CO 180 16 h 505 3675 186
clean surface - 16 h rough - -
very weakly on Si(111) moving rapidly on the surface Because of their rapid diffusion
the STM is not able to image the CO molecules on the silicon [10]
Increasing the temperature part of the CO molecules will surely desorb but some of them
will dissociatively adsorb onto the surface The oxygen will then desorb as SiO while
carbon will remain fixed forming very stable Si-C bonds
CO decomposition will be favored at surface defects surface steps or in the proximity of
already formed SiC clusters An increased reactivity of the CO molecule on a surface in
correspondence with phase boundaries has already been observed on Pt surfaces [10]
In this case a reduced number of nucleation centres will react with the CO molecules
yielding large and fewer islands with respect to the case of methanol which does not have
the possibility of travelling long distances over the surface
Other factors are likely to influence the process of island nucleation after CO adsorption
the co-adsorption of other molecules which can slow down CO and favour its
decomposition cannot be ruled out as well as the formation of new defects at higher
temperatures Measurements with a variable temperature STM or with some other surface
analysis technique are planned in the future to confirm the proposed mechanism
When the surface was further annealed the nanoislands grew in size After 16 hours
heating very large nanoislands with diameters of sim35 nm and heights of sim20 nm became
visible (fig64 left) As a comparison when the same surface was heated without
nucleation centres deriving from contamination an irregular surface was obtained (fig64
right)
While the nanoisland density seemed to reach an asymptotic limit of 500 islandsmicrom2
(table 63) island size continued to grow with time (fig65) even when no further carbon
was supplied to the surface This indicates that even though the initial nucleating core of
the nanoisland is likely to be an SixC1-x alloy further growth is due to silicon atoms
diffusing from the crystal and being adsorbed by the growing island The final volume
VPalermo 71
occupied by the islands corresponds roughly to a 18 nm thick overall layer of removed
silicon
It is noteworthy that the clean sample after 16 hours annealing even though smoother than
the CO-contaminated one looked much more disordered with hardly any flat area visible
on nanometres scale In the presence of nanoislands however further annealing increases
island size but keeps part of the surface quite flat and ordered with flat areas visible
between the islands Thus the nanoislands can be imagined to act as ldquoimpurity sinksrdquo for
further contaminants approaching the surface yielding a greater but more ordered surface
roughening with respect to the uncontaminated silicon surface
Circular arrays of nanoislands
We used the voids described in the previous chapter formed during oxide layer
decomposition as nanoscopic masks to test the validity of the hypothesis outlined above
The small circular areas uncovered during oxide decomposition can act as ldquoskating rinksrdquo
where contaminants and diffusing silicon atoms can move freely over the surface while
the oxide layer all around and higher than the void area as well as being chemically
bonded to the surface silicon atoms will hinder surface diffusion Furthermore the oxide
passivating layer will at least in some cases prevent contaminating molecules from
adsorbing and decomposing on the surface In this way selective nanoisland growth inside
void areas can take place
The experimental procedure consisted of the following steps
1) An oxide-passivated sample was heated for 10 minutes at 800degC Oxide decomposition
began in correspondence with surface defects and spread laterally creating naked areas
of silicon
2) While the void enlarges the same defect that initiated oxide decomposition behaved as
an attractor for moving silicon atoms and nucleated the growth of a nanoisland at the
void centre
3) The void surface was exposed to contaminating molecules
4) The sample was then heated again at 800degC During the second heating the void
continued to enlarge and new nanoislands were created The freshly uncovered silicon
surface provided mobile silicon atoms to feed nanoisland growth
72 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
5) When the heating continued for long enough all the oxide was desorbed and a clean
silicon surface covered by circular groups of nanoislands was obtained
Figure 66 shows the different stages of the process
We thus heated an oxidized sample at 800degC for 10 minutes in UHV to create multiple
voids in the oxide layer After this we exposed the sample to the contaminating molecules
Then we further annealed the surface at the same temperature to promote island growth
The first time we observed nanoisland growth in the void area was by chance after heating
for a second time samples which had been stored for 2 weeks in vacuum Even at pressures
as low as 10-11 mbar some molecules will adsorb onto the surface slowly creating defects
and thus potential nanoisland nucleation centres It is possible to observe with STM that
the silicon crystal surface looks very clean just after a high temperature flash but even
after only a few days randomly adsorbed species will be observable on the surface
Unfortunately it is not possible to identify these species simply from in situ STM
measurements
Mass spectrometry measurements show that the residual gas contaminants in vacuum are
usually He Ar H2 CH4 CO CO2 and N2 [11] Some of these molecules such as
hydrogen will stick to the silicon surface at room temperature but will simply desorb
during annealing without nucleating nanoislands Others like CH4 or CO2 will not react
with the surface to generate nanoislands as demonstrated previously The best candidate
for contamination of samples stored in UHV seems thus carbon monoxide has the ability
321
4 5
Fig 66 Selective nanoisland growth within oxide voids See text for details
VPalermo 73
as shown before to nucleate nanoislands although co-adsorption mechanisms can not be
ruled out Fig 67 (left) shows the circular groups of nano-islands obtained after heating
void-covered samples stored for 2 weeks in UHV Small islands are observed in the former
void area The void itself has enlarged slightly uncovering a clean oxide-free area of the
silicon surface Apart from the central island created during the first period of heating the
surrounding ones have dimensions decreasing from the void border to the inner area of the
void
This confirms that most of the material needed for island growth comes from the silicon
atoms diffusing from the freshly uncovered areas around the original void The outer
nucleating centres are nearest to the silicon atom source and thus generate larger islands
For a faster process we directly exposed the void to a significant concentration of CO
Fig67 (centre) shows the islands grown after a 180 Langmuir exposure and subsequent
heating The islands are preferentially located around the original void perimeter the rest
of the original oxide-free surface remaining untouched This suggests that although the
whole surface is exposed to CO the gas only interacts chemically with the surface at the
SiSiO2 interface around the oxide-free void perimeter The SiSiO2 interface thus provides
preferential nucleation sites for nanoisland growth in the successive thermal annealing
step This recalls a similar phenomenon observed in the catalytic oxidation of CO on the
Pt(111) surface where CO molecules reacted with adsorbed O species only along the
perimeters of oxygen islands [10]
On repeating the CO exposure and thermal annealing steps a second circle of nano-islands
Fig67 Left circular area of nanoislands obtained from a sample stored 2 weeks in UHV 500x500 nmCentre nanoisland circle obtained after exposing the voids to 180 L of CO and heating for 10 min at800degC 400x400 nm Right two concentric circles obtained with further CO adsorption followed byheating 650x650 nm
74 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
was formed around the newly-expanded surface void perimeter (fig67 right) This
process can in principle be repeated again to obtain a series of concentric groups of
nanoislands the only limit depending on the minimal distance between neighbouring
expanding voids We have thus observed that contaminants adsorbed at low pressure over
long periods of time generate random island growth over the whole void area while CO
molecules adsorbed at relatively high pressures over short periods yield selective
nanoisland growth at the former void border Several theories can explain the different
results CH4 and CO2 as mentioned above did not nucleate islands when when the silicon
surface was exposed to them for short periods However the lower surface mobility of
these contaminants and the greater time available for reaction with the surface could be
the cause of nucleation of SiC clusters over the whole oxide-free void area and not only at
its border The coadsorption of water another UHV residual gas with CO on the naked
silicon areas could perhaps allow a surface reaction similar to that between CO and the
SiSiO2 interface during CO exposure Another explanation for the formation of these
random nucleation sites within the voids could be the coadsorption of residual carbon-
containing species present in the UHV system with hydrogen the major residual gas in
stainless steel UHV systems
The possibility of decorating silicon oxide borders with nanoislands using the high
mobility and selective decomposition of CO molecules suggests interesting applications
for nanofabrication processes that will be discussed below
Silicon nanowire creation on Si(100)
We performed most of the experiments described above on silicon (111) 7x7
reconstructed because this surface is easily prepared in UHV and above all because it is
isotropic and has no preferential directions for atom diffusion The 2x1 reconstruction of
silicon (100) even though it has a much simpler unit cell than Si(111) 7x7 shows a more
complicated morphology with lots of monoatomic and biatomic steps and dimer rows
parallel to each other The orientation of the dimer rows changes by 90deg on alternate
atomic layers Surface diffusion in this case is much more complicated because atoms will
experience the effects of the surface anisotropy and diffuse preferentially along or across
the dimer rows The diffusion energy of a silicon atom as example is 06 eV along a
dimer row and 085 eV across different dimer rows This while being a complication
VPalermo 75
offers interesting possibilities for the creation of ordered structures The monoatomic steps
and the dimer rows can for example act as templates for the formation of elongated
structures of composition similar to the islands described previously
On heating a Si(100) surface oxide decomposition and void growth takes place as
described for Si(111) In some cases the void shape reflects the surface symmetry
especially for very thin oxide layers and the voids often have a slightly squared shape
Nanoisland growth is also influenced by the substrate symmetry and the islands look
rectangular as irregular clusters surrounded by concentric patterns of monoatomic steps
The whole structure reminds vaguely a ldquozigguratrdquo a kind of stepped pyramid found in
Mesopotamia (fig68 inset)
Apart from nanoisland growth in some cases we observed spontaneous formation of
nanowire-like SiC structures on the Si(100) surface The wires were generated after
exposing a Si(100) surface in vacuum to traces of CO followed by sample annealing at
800degC for 15 minutes The wires exhibit lengths ranging from 10 to 100 nm and average
widths of ca 5 nm Although being randomly positioned on the surface they are perfectly
aligned along the crystal axes of the substrate (Fig68)
Several descending monoatomic steps can be observed around each nano-line and nano-
island Very often one of the line extremities coincided with an island
Nanoline
Nanoisland
Fig 68 A Si(100) surface covered by nanoislands and nanowires The monoatomic steps present on the surface are visible The two insets show a typical nanoisland and a nanowire obtained on this surface The typical ldquodimer-row ldquostripes are visible along the sides of the nanowire
76 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Couples of parallel lines starting from the same nano-island were also observed At high
resolution (fig 69) the lines showed a lamellar periodic structure with a periodicity of
around 4 nm and frequent defects Scanning Tunneling Spectroscopy measurements
showed no significant difference between the line and the surrounding silicon surface
The proposed formation mechanism for this kind of structure involves two stages in the
first one organic contaminating molecules adsorb and diffuse on the surface from the
vacuum environment stopping preferentially at surface singularities including steps and
defects
Several substances such as atomic carbon carbon monoxide ethylene and fullerene
decompose when adsorbed on silicon surfaces at temperatures above 500degC and create
localised defects on the surface inducing strain deformations and a metastable surface
reconstruction [12]
If these molecules remain attached to the steps they can diffuse rapidly along step border
until they meet a line and decompose In this way long lines of SiC defects parallel to the
surface steps can be produced In the second phase the clusters of silicon carbide can
-005
0
005
01
015
02
025
03
035
04
-3 -2 -1 0 1 2 3
V
dId
V lt
dId
Vgt
linesilicon
Current- Voltage behaviour measured with STS Fourier analysis of wire periodicity
asymp 4 Aring
High resolution image of a line
Fig 69 Top high resolution image of a nanowire showing the lamellar structure Fourier analysis (bottom left) shows that the lamellae periodicity is sim 04 nm bottom right the IV characteristics measured with the STM tip on the line and the silicon surface
VPalermo 77
25-600degC
Final linear structure
600-800degC
Contaminant Silicon
Fig 610 Schematic representation of nanowire formation mechanism
locally inhibit the spontaneous surface diffusion of silicon atoms acting as templates for
the reorganisation of surface steps (fig 610)
The final morphology shows disordered lines and islands each surrounded by a complex
pattern of silicon monatomic layers
At 800degC silicon atoms are highly mobile on the surface Due to atom diffusion the
surface reorganizes with a step-flow mechanism The presence of a fixed line of
contaminants blocks atom diffusion and step flow generating the monoatomic steps
around each wire (fig 611) SiC clusters act as lsquopinning sitesrsquo on silicon atom surface
diffusion and can be purposely used to direct the localised growth of nano-islands [13]
Although the possibility of using methanol and carbon monoxide as nano-island precursors
has been demonstrated the exact nature of the substances nucleating nano-wire growth is
still unknown
Further experiments will be needed to confirm these hypotheses as well as to identify the
contaminating species nucleating nanowire growth Carbon monoxide is a main candidate
for nanowire growth but the real nucleation process is likely to be a complex one
78 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Step flowblocked
Fig 611 Representation of step flow blocked by the presence of a nanowire 105x105nm
Bibliography
[1] Surface-Chemistry Of Silicon Waltenburg HN Yates JT Chemical Reviews 95 (5)
1589-1673 Jul-Aug 1995
[2] J Dabrowski H Mussig Silicon Surfaces And Formation Of Interfaces (World
Scientific Publishing Singapore 2000)
[3] Oxidation And Roughening Of Silicon During Annealing In A Rapid Thermal
Processing Chamber Mohadjeri B Baklanov Mr Kondoh E Maex K Journal Of
Applied Physics 83 (7) 3614-3619 Apr 1 1998
[4] Growth Of Si On The Si(111) Surface Lanczycki Cj Kotlyar R Fu E Yang Yn
Williams Ed Das Sarma S Physical Review B 57 (20) 13132-13148 May 15 1998
[5] Scanning Tunneling Microscopy Study Of Si(001) And Si(110) Surface Structures
Resulting From Different Thermal Cleaning Treatments Dijkkamp D Van Loenen Ej
VPalermo 79
Hoeven Aj Dieleman J Journal Of Vacuum Science amp Technology A-Vacuum
Surfaces And Films 8 218-221 Jan 1990
[6] Nanoscale roughening of Si(001) by oxide desorption in ultrahigh vacuum Gray SM
Johansson MKJ Johansson LSO Journal Of Vacuum Science amp Technology B 14 (2)
1043-1047 Mar-Apr 1996
[7] Step Motion On Crystal Surfaces Schwoebel Rl Journal Of Applied Physics 40 614-
618 Feb 1969
[8] Scanning Tunneling Microscopy Investigation At High Temperatures Of Islands And
Holes On Si(111)7x7 In Real Time Evidence For Diffusion-Limited Decay
Hildebrandt S Kraus A Kulla R Wilhelmi G Hanbucken M Neddermeyer H Surface
Science 486 (1-2) 24-32 Jul 1 2001
[9] Methanol Adsorption On Si(111)-(7x7) Investigated By Core-Line Photoemission And
Mass Spectrometry Of Photodesorbed Ions Carbone M Piancastelli Mn Zanoni R
Comtet G Dujardin G Hellner L Surface Science 370 (1) L179-L184 Jan 1 1997
[10] Atomic And Macroscopic Reaction Rates Of A Surface-Catalyzed Reaction
Wintterlin J Volkening S Janssens Tvw Zambelli T Ertl G Science 278 (5345)
1931-1934 Dec 12 1997
[11] Redhead PA Hobson JP Kornelsen EV The Physical Basis Of Ultrahigh
Vacuum Chapter 12 (Chapman amp Hall London 1968)
[12] The Si(001) C(4 X 4) Surface Reconstruction A Comprehensive Experimental
Study Norenberg H Briggs Gad Surface Science 430 (1-3) 154-164 Jun 21 1999
[13] Production Of Nanostructures Of Silicon On Silicon By Atomic Self-Organization
Observed By Scanning Tunneling Microscopy Jones D Palermo V Applied Physics
Letters 80 (4) 673-675 Jan 28 2002
80 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 81
7 Conclusions and proposals for future work It is very difficult to foresee the future of nanoscience and nanotechnology even over the
next few years exciting discoveries are continuously made opening up new possibilities
and progress in this sector is now one of the fastest and most varied So it is impossible to
predict whether computers will in the future be made with nanowires nanotubes single
electron devices or some kind of quantum-based transistor What does seem clear
however is that silicon devices can still further extend their levels of miniaturization to
remain competitive for at least the next ten years [1] During this period new emerging
technologies will reach scientific maturity and arrive at the production lines
Even when new technology prototypes do become available for full-scale production the
astronomical cost of changing from silicon to new technologies will be a major concern
Moreover a vast amount of extremely detailed knowledge is already available on all
aspects of silicon technology Thus the possibility of integrating innovative
nanotechnologies with standard CMOS silicon technology already seems an attractive
prospect [2]
For this and other reasons we have focussed our research efforts on surface phenomena on
silicon wafers which could be potentially useful for the development of silicon-compatible
nano-devices
Some important conclusions can be drawn from the work described in this thesis
bull Silicon surfaces of different crystal faces were studied at atomic resolution in ultra-
high-vacuum following transformations induced on the surface by chemical (etching)
and physico-chemical (organic contamination heating) treatments
82 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull The presence of surface contaminants induces at high temperature the growth of
nanometre-sized islands and lines on silicon
bull Final island densities and dimensions strongly depend upon the chemical nature of the
contaminant molecules and the treatments used Molecules interacting only weakly
with silicon will desorb upon heating whereas molecules which decompose easily on
Si will give widespread random nano-island nucleation Between these two extreme
cases complex behaviours of diffusion and subsequent decomposition are possible
bull Nanoscopic naked silicon voids were produced in oxide layers through carefully-
controlled thermal annealing It was found that void density and size depends upon
oxide purity surface contamination levels and annealing conditions
bull The growth of nano-islands and nano-voids was exploited simultaneously and the
voids used as nanoscopic masks to control the positions of growing islands Complex
silicon-on-silicon nano-structures were obtained with multi-step growth processes
bull Finally the use of macroscopic treatments and simple physical processes to produce
large quantities of nanometre-sized structures on silicon surfaces was demonstrated
These processes and the structures produced are of potential interest to the
microelectronics industry
We showed in previous chapters how it is possible to modify silicon surfaces using
chemical and physical methods The use of self-organization for the production of Si-on-Si
20 nm
Fig 71 Left SEM image of field emitter arrays made by conventional lithography [3] Center 3-D STM image of a self-organized void with a central nano-island Right Lateral schematic view of a field emitter device
VPalermo 83
nano-structures has the advantage with respect to other emerging techniques of being
silicon compatible and thus is a potential candidate for its implementation in the
production of new devices Furthermore this technique can create billions of strucures on a
wafer in only a few minutes being thus applicable to large-scale production
Even when ordered structures are obtained these methods still show a poor reproducibility
and control Chemically modified surfaces look very disordered at high magnification
with trenches and holes etched into the surface and progress has recently been made in
producing very small structures electrochemically [4] Oxide decomposition with
nanoisland growth in UHV can also yield ordered structures The void holes with
nanoislands located at their centres although quite variable in dimensions are more similar
to engineered devices than to a spontaneous random surface-roughening process In fact
there is a surprising similarity between the nanovoid-nanoisland structures and well-known
commercially available devices called field emitter arrays (FEA) These are small tips used
as electron guns in various kinds of displays and other devices Both the commercial
devices and our self-organized structures consist of a protruding tip at the centre of
microscopic holes on an insulating layer covering a semiconductor surface (see fig71)
Of course the order and size uniformity of commercial FEAs is much better than that
obtained with our process but it is encouraging to see that a self-organized structure
obtained with simple heating has a very similar surface topography and composition of
commercial working devices obtained after many expensive stages of optical lithography
deposition and etching
Nano-island density and size can to some extent be controlled and different
contaminants can be adsorbed onto the surface to modify the growth process In the future
with a better understanding of diffusion dynamics and growth mechanisms a multi-stage
fabrication process can be envisaged where purposely-designed molecules are adsorbed
onto silicon surfaces to diffuse organize in ordered arrays and favour surface
reorganization at the nanometre scale The same molecules could be used to deliver
dopants to obtain for example highly conducting nanostructures on otherwise undoped
highly resistive silicon The selective formation of nanostructures at the SiSiO2 border
can be used to scale down mesoscopic patterns to the nanometric level A silicon oxide
layer can be easily patterned on the surface with normal optical lithography and then
nanostructures can be built following the pattern edge In this way the production of very
small nanowire based devices may be possible (fig 72)
84 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
D
Gate
S
D
S
SiO2
SiO2
2 Heating nanowire creation
and oxide removal
3 Source-drain deposition by conventional lithography
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
4 Insulating layer (oxide) and gate (metal) deposition
Fig 72 Schematic drawing of the construction of a self-organized nanowire-based transistor
Surface diffusion and decomposition of molecules can be sensitive to even smaller surface
features The nanoscopic lines formed on Si(100) are very regular and neat if their
nucleation and growth mechanism could be controlled ordered networks of nanowires
could be made on silicon
VPalermo 85
The techniques developed here are still experimental ones the careful control and
reproducibility of each step is still an issue and further studies will be needed to clearly
understand the dynamics of formation of these structures Nonetheless the results obtained
so far open up the possibilities of producing novel devices such as for example nano-
transistors (see scheme in fig73 where the mechanism described in fig 72 is applied for
the fabrication of a device array) The possibility of using the border of a lithographic
pattern to draw nanolines can in principle allow the creation of nanodevices using normal
IC manufacturing techniques
The results obtained and their description in this thesis are not the first examples of the use
of self-organization for the production of ordered structures and will certainly not be the
last In our opinion the most important outcome of these findings is that it is possible to
use surface diffusion to overcome the inherent limits of lithographic techniques in
microdevice production processes and to force properly selected molecules to react with
the surface in a spatially differentiated manner at particular sites thereby creating ordered
series of nanostructures The role of the surface dynamics of the substrate atoms
themselves is extremely important in this process
Outstanding results have been and continue to be obtained in nanoscience and
nanotechnology research promising the emergence of new production technologies Sooner
or later those emerging technologies will have to face the issues of cost and compatibility
with the enormous investment made in existing technologies and processing facilities The
commercial production of nanostructures based on surface diffusion processes could in
principle be cheap simple and compatible with existing technology Moreover the
processes of diffusion in bulk silicon is well-known in the microelectronics industry which
has both the know-how and the equipment to better understand and exploit the surface
diffusion processes on silicon surfaces studied here
Extremely important and innovative results are often obtained by exploiting very simple
ideas and in our opinion physical processes such as surface diffusion used to create
complex nanostructures on silicon surfaces is a prime example
86 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
1 Creation of a silicon-oxide pattern by conventional lithography Gas adorption and selective decomposition at the border
2 Heating nanowire creation and
oxide removal
3 Source-drain deposition by conventional lithography
4 Insulating layer (oxide) and gate (metal) deposition
Fig 73 Schematic macroscopic outline of the construction of a self-organized nanowire-based transistor
VPalermo 87
Bibliography
[1] P Gargini Intel Technology Coordinator ldquoFrom Microelectronics To
Nanotechnologiesrdquo Invited Campus Colloquia Lecture CNR Research Area Bologna
February 6th 2003
[2] Toward A Hybrid Micro-Nanoelectronics Cerofolini Gf Ferla G Journal Of
Nanoparticle Research 4 (3) 185-191 Jun 2002
[3] Recent Progress In Field Emitter Array Development For High Performance
Applications Temple D Materials Science amp Engineering R-Reports 24 (5) 185-239
Jan 25 1999
[4] Silicon Dioxide Micropillars For Sieving Fabricated By Macroporous Silicon-Based
Micromachining Izuo S Ohji H French Pj Tsutsumi K Kimata M Sensors And
Materials 14 (5) 239-251 2002 Electrochemical Etching In HF Solution For Silicon
Micromachining Barillaro G Nannini A Piotto M Sensors And Actuators A-Physical
102 (1-2) 195-201 Dec 1 2002
88 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
VPalermo 89
Acknowledgements
This work would not have been possible without the help and guidance of Dr Derek Jones
he introduced me to the use of STM and other techniques found the funding for this PhD
and helped me with the thousands of small and big problems I had to face during my
research I am grateful to Prof Alberto Ripamonti too for being the supervisor of this
thesis Thanks are also due to Dr Giancarlo Seconi director of ISOF-CNR where most of
this work was carried out Financial support from the Italian National Research Council
(CNR) is also gratefully acknowledged
My passion for surface science began when Fabio Biscarini showed me that it was possible
to really see the atoms and for this I will always thank him
During these years I had the luck to collaborate with many good scientists among those I
would like to cite Enrichetta Susi Massimo Cocchi and Anna Mazzone from CNR
Claudio Zannoni Anna Cavallini Daniela Cavalcoli and Antonio Castaldini from the
University of Bologna Sergio Pizzini and Maurizio Acciarri from the University of Milan
One of the most stimulating periods of my PhD was the one I spent at the Steacie Institute
for Molecular Sciences in Ottawa working with Robert Wolkow Dan Wayner Greg
Lopinski and Peter Kruse I thank them for discussing together many exciting scientific
ideas and trying to put some of them into practice
Special thanks go to my colleagues Paolo Samorigrave and Stefano De Cesari for long and
useful discussions on Science Life and their interactions often with the help of a good
pint
Finally I would like to thank Vassilia Gaetano Simone Claudia Silvia Letizia
Alessandro Angela and Sebastien even though not directly involved in this work they
have been over these three years a continuous source of happiness and support
Bologna March 2003
90 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
SCIENTIFIC PUBLICATIONS bull Lateral diffusion of titanium disilicide as a way to contacting
hybrid Si-organic nanostructures Palermo V Buchanan M Bezinger A Wolkow RA APPLIED PHYSICS LETTERS 2002 v81 p 3636 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v6 issue 20) bull Surface modifications in Si after Rapid Thermal Annealing Castaldini A Cavalcoli D Cavallini A Jones D Palermo V Susi E JOURNAL OF THE ELECTROCHEMICAL SOCIETY 2002 v 149 pG633 bull Production of nanostructures of silicon on silicon by atomic self-
organisation observed by scanning tunnelling microscopy Jones D Palermo V APPLIED PHYSICS LETTERS 2002 v 80 p 673 (this article has been selected for publication on the VIRTUAL JOURNAL OF NANOSCALE SCIENCE amp TECHNOLOGY 2002 v5 issue 5) bull Nucleation of nanostructures from surface defects on silicon Palermo V Jones D SOLID STATE PHENOMENA 2002 v 82-84 p 687 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo MATERIALS SCIENCE AND ENGINEERING B 2002 v88 (2-3) p220 bull Morphological changes of Si[100] surface after treatment with
concentrated and dilute HF Palermo V Jones D MATERIALS SCIENCE IN SEMICONDUCTOR PROCESSING 2001 v 4 p 437 bull Electrical and structural properties of processed silicon surfaces Susi E Cavallini A Castaldini A Cavalcoli D Jones D Palermo V ldquoRECENT RESEARCH DEVELOPMENTS IN VACUUM SCIENCE amp TECHNOLOGYrdquo 2001 v 3 p189 bull Numerical Solutions of the Stochastic Equations of Crystal
Growth Mazzone A M Palermo V INTERNATIONAL JOURNAL OF MODERN PHYSICS C 2000 v 11 Part 1 p195-204 bull Advances in silicon surface characterisation using light beam
injection techniques
VPalermo 91
Acciarri M Pizzini S Simone G Jones D Palermo V MATERIALS SCIENCE AND ENGINEERING B 2000 V73 (1-3) p 235 - 239 bull Abrupt orientational changes for liquid crystals adsorbed on a
graphite surface Palermo V Biscarini F Zannoni C PHYSICAL REVIEW -SERIES E- 1998 V 57 NUMBER 3A p R2519-R2522 Oral presentations bull The use of oxide desorption and surface diffusion for the
creation of silicon on silicon nanostructures Palermo V Jones D 1ST NATIONAL WORKSHOP ON CURRENT TRENDS IN NANOTECHNOLOGIES Catania (Italy) Feb 2002 bull Self-Organised Growth of Silicon Structures on Si(100) During
Oxide Desorption Jones D Palermo V E-MRS Spring Meeting Strasbourg (France) June 2001 bull STM study of surface transformations on silicon during UHV
annealing Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999 bull Advances in silicon surface characterisation using light beam
injection technique Acciarri M Pizzini S Simone G Jones D Palermo V E_MRS SPRING MEETING June 1999 Posters bull Spontaneous nano-wire growth on silicon Palermo V Jones D SMARTON workshop Leuven (Belgium) October 2002 bull Ordered circles of nano-islands on silicon from CO adsorption Palermo V Jones D TRENDS IN NANOTECHNOLOGY 2002 Santiago de Compostela (Spain) September 2002 bull Production of nanostructures of silicon on silicon by atomic self-
organisation Palermo V Jones D EUROMAT Rimini (Italy) June 2001 bull Etching holes and anisotropic corrosion on silicon [100] Palermo V Jones D Susi E Asoli B SILICON WORKSHOP Genova (Italy) February 2001
92 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
bull Morphological and electrical characteristics of damaged silicon surfaces
Susi E Castaldini A Cavalcoli D Cavallini A Jones D Palermo V SILICON WORKSHOP Genova (Italy) February 2001 bull Effect of HF etching on the roughness of a silicon surface Palermo V Jones D NATIONAL CONFERENCE ON PHYSICS OF MATTER Genova (Italy) June 2000 bull Effect of Native Oxide Desorption upon the Surface Morphology of Si[100] by STM and LEED Palermo V Jones D SILICON WORKSHOP Genova (Italy) February 2000 bull Characterisation of silicon surfaces for microelectronics through STM measurements Palermo V Jones D 4TH MULTINATIONAL CONGRESS ON ELECTRON MICROSCOPY Veszprem (Hungary) Sept 1999
VPalermo 93
Contact Vincenzo Palermo ISOF- Institute for Organic Synthesis and Photoreactivity Via Gobetti 101 40129 Bologna ITALY Tel +39-051-6398336 Fax +39-051-6398349 Mail palermoisofcnrit
94 Creation of Nanometre-Scale Islands Wires and Holes on Silicon Surfaces for Microelectronics
Index
1 Introduction nanotechnology and the future of computers
2 Silicon surfaces
3 STM and other surface analysis techniques
4 Surface modification of silicon in liquid Nano-hole creation
5 Surface modification of silicon in vacuum void creation and oxide desorption
6 Adsorption and diffusion of molecules on silicon creation of nano-islands and
nano-lines
7 Conclusions and proposals for future work