i
Optimization of the etching parameters of the ion
milling system Nordiko 3600
Diminish of redeposition on micro-devices
André Filipe Rodrigues Augusto
Dissertação para obtenção do Grau de Mestre em
Engenharia Física Tecnológica
Júri
Presidente: Professor João Carlos Carvalho de Sá Seixas
Orientadora: Professora Susana Isabel Pinheiro Cardoso de Freitas
Vogais: Professor Paulo Jorge Peixeiro de Freitas
Setembro de 2007
ii
Abstract
The aim of this thesis is to optimize the etching parameters using the ion milling system
Nordiko3600 at INESC-MN. The magnetic and electrical properties of the magnetic tunnel
junctions are analysed for different etching conditions. The efforts of optimization are focus on
the second etching, where the geometry and area of the junction pillar is defined. This is the
critical point for junction manufacturing, since the magnetic characteristics and above all the
electrical signal are strongly conditioned by it. The electrical signal is characterized in terms of
the so-called tunnel magnetoresistence, TMR, which is defined as the percentage relative
resistance change along a variable magnetic applied field.
The redeposition of conductive material on the sidewalls of the tunnel junction drives to a
decreasing of tunnelling transport across the barrier, reducing the TMR signal. For junctions
with low resistance this is a higher matter of concern. To decrease redeposition during the ion
milling etching two approaches are exploited. The first one is to vary the incident angle of the
Argon-etching beam with the sample, removing by etching part of the redeposited material
gather on the sidewalls. The angle between beam and sample surface ranges from 40º-90º.
The second approach is to use different times of oxidation via oxygen plasma after
performing the etching. By oxidation of the lateral material, less conduction paths are found
for the electrons, which enhances the tunnelling effect. As is shown in this thesis the TMR
increases dramatically with oxidation time, however the resistance also increases. Therefore
the oxidation time is conditioned by how critical is the resistance for a certain application. 50
seconds time oxidation via oxygen plasma has already become a standard procedure in the
magnetic tunnel junction’s process at INESC-MN.
KEYWORDS - Ion milling etching, magnetic tunnel junction, redeposition, plasma oxidation,
etching angle.
iii
Resumo
Esta tese tem como objectivo optimizar os parâmetros de gravação por feixe iónico,
utilizando o sistema Nordiko3600 do INESC-MN. As propriedades magnéticas e eléctricas
das junções de efeito de túnel são analisadas para diferentes condições de gravação. Os
esforços de optimização centram-se na segunda gravação, onde a geometria e a área do
pilar da junção são definidas. Este é o ponto crítico na fabricação da junção, uma vez que
condiciona as caracteristicas magnéticas e acima de tudo o sinal eléctrico. O sinal eléctrico é
condicionado em termos da magnetoresistência, TMR, que é definida como a percentagem
relativa da variação da resistência.ao longo da aplicação de um campo magnético variável.
A redeposição de material condutor nas paredes da junção por efeito de túnel conduz a uma
diminuição da condução por efeito de túnel na barreira, reduzindo o sinal TMR. Para junções
de baixa resistência isto constitui um grande problema. Para diminuir a redeposição durante
o gravação por feixe iónico duas direcções são escolhidas. A primeira, consiste em variar o
ângulo de incidência do feixe de Árgon que grava com a amostra, removendo assim, o
material redepositado nas paredes da junção. Os ângulos usados estão entre 40º-90º. A
segunda, usa diferentes tempos de oxidação via plasma de oxigénio depois do gravação ser
efectuada. Desta forma os materiais condutores oxidam-se, tornando-se não condutores.
Como é mostrado nesta tese o TMR aumenta dramaticamente com o tempo de oxidação, no
entanto a resistência também aumenta. Como tal, o tempo de oxidação é condicionado pela
importância que a resistência tiver numa certa aplicação. 50 segundos de oxidação via
plasma de oxigénio já faz parte do normal processo de fabrico de junções no INESC-MN.
PALAVRAS CHAVE – Gravação por feixe de iões, juncão de efeito de túnel, redeposição,
oxidação a plasma, ângulo de gravação.
iv
Agradecimentos
No Inverno sou budista, E no Verão sou nudista. Poema “A Minha Religião” em “O segredo de Joe Gould”, Joseph Mitchel
À primeira vista a escolha deste poema pode provocar uma certa estranheza, surpresa e
acima de tudo incompreensão. No entanto, este poema, retirado de uma pequena maravilha
da literatura contemporânea Nova Iorquina, traduz brilhantemente em dois versos o espírito
que me acompanhou ao longo desta tese e nos últimos anos do meu curso. Tendo em conta
a natureza do trabalho experimental realizado no INESC-MN, onde a busca de perfeição é
feita à custa da optimização continuada dos métodos, num contra relógio, onde o tempo ou
falta dele, revela-se muitas vezes como o nosso principal inimigo, muitos diriam, e
provavelmente vocês concordariam, que as palavras de Cesário Verde,
Se eu não morresse nunca! E eternamente o Buscasse e conseguisse a perfeição das coisas! Cesário Verde, Estação de Metro da Cidade Universitária
lidas e relidas por muitos no seu quotidiano literário lisboeta seriam bem mais adequadas.
Porém, a realidade da minha tese é bem mais abrangente do que as condicionantes
intrínsecas ao trabalho experimental, muitos outros factores como a família e os amigos são
igualmente parte importante desta tese. A verdade é que a minha tese foi feita de pequenos
Invernos e Verões, sem qualquer sentimento de dicotomia, mas sim de complementaridade.
Onde muitas vezes a necessidade se impôs à vontade.
Aos meus amigos (Verão) que me ofereceram uma cerveja e que recusei, não deixando por
isso de continuar a oferecê-las, agradeço e muito, pois receberam sempre um não de braços
abertos, percebendo que por vezes temos que ser budistas, como Joe Gould. para que o
trabalho seja feito.
À minha família, que não veio em primeiro lugar por uma pura questão de estética literária.
Tenho que destacar os meus pais, devido à sua amizade, abertura, excentricidade saudável
e acima de tudo, por me terem incutido ao longo da minha vida a conhecida expressão
anglo-saxónica “Thinking out of the box”. À minha irmã, por me fazer sentir sempre bem
v
comigo mesmo. E à minha avó pelos 20!, que de quando em quando escondia no meio dos
meus livros, os quais foram responsáveis por muitas idas ao cinema seguidas de “algumas”
loirinhas bem fresquinhas. Obrigado Avó!
Como budista tenho que agradecer à minha orientadora, a professora Susana, por todo o
apoio que me deu, ao professor Freitas por partilhar o seu profundo conhecimento nesta e
noutras áreas, e finalmente (“last but not least”) um muito obrigado a todos os meus colegas
que sem qualquer obrigação em muito contribuíram para realização desta tese.
Lisboa . Setembro de 2007
vi
Contents
1. INTRODUCTION ................................................................................................................................1
2. SPIN DEPENDENT TUNNELING ...................................................................................................3
2.1.SIGNAL AS RESISTANCE VARIATION.............................................................................5
2.2.MAGNETIC RESPONSE..........................................................................................................7
3. MAGNETIC TUNNEL JUNCTION ARCHITECTURE ..............................................................9
3.1.LAYERS ROLE ........................................................................................................................12
3.2.LAYERS MAGNETIC CHARACTERIZATION ...............................................................14
4. MAGNETIC TUNNEL JUNCTION FABRICATION ................................................................16
4.1.FILM DEPOSITION ................................................................................................................16
4.2.ETCHING PROCESS ..............................................................................................................19
4.3.LITHOGRAPHY AND PATTERN TRANSFER ................................................................24
4.4.PROCESS STEPS .....................................................................................................................27
5. ELECTRICAL CHARACTERIZATION ......................................................................................37
6. REDEPOSITION STUDY.................................................................................................................39
6.1.ETCHING ANGLES ................................................................................................................42
6.2.OXIDATION VIA OXYGEN PLASMA ...............................................................................46
7. CONCLUSIONS .................................................................................................................................52
8. APPENDIX ..........................................................................................................................................53
8.1.RUN SHEETS ............................................................................................................................53
8.2.PAPER ........................................................................................................................................70
9. BIBLIOGRAPHY...............................................................................................................................81
vii
Index of Figures
Figure 1 - In the MRAM, information stored in one of the magnetic layers, can be erased by a current Ix
through the top conductive paths creating an in-plane magnetic field (Hx). However, to address only a single element also the bottom lines are used to create a magnetic field Hy to reduce the effective switch field [3]. __________________________________________________________________________ 1
Figure 2 – Tunnelling between the two ferromagnetic is governing by the density of states at Fermi level when a small bias voltage V is applied across the junction of thickness t and height ! (depends on material characteristics) [3].___________________________________________________________ 3
Figure 3 - Band diagram of the tunneling process between ferromagnetic electrodes. The process is shown for an applied bias voltage to the electrodes, with tunneling occurring with spin conservation between equivalent spin states [6]. _____________________________________________________ 4
Figure 4 - (Left) Resistance circuit of the parallel state of magnetization. (Right) Resistance circuit of the antiparallel state of magnetization [3]. ________________________________________________ 5
Figure 5 – Hysteresis loop curve [15]. __________________________________________________ 7
Figure 6 - Evolution of tunnel junctions engineered for MRAM applications. (a) Basic MTJ structure. (b) MTJ structure with a ferromagnetic layer pinned by an antiferromagnetic layer (c) MTJ structure with synthetic antiferromagnetic pinned layer (SAF). (d) Structure in which both the pinned and free elements consist of antiferromagnetically coupled pairs [7]. __________________________________________ 9
Figure 7 - Tunneling at 295 K in a CoFe–Al2O3–Co junction [7]. ______________________________ 9
Figure 8 – Hysteresis curves from samples with SAF. Sample TJ1268_7 with 5Å thick AlOx barrier. Sample N2TJ68_B with 8Å thick MgO barrier. As is noticed the TMR with MgO barrier is much higher.____________________________________________________________Error! Bookmark not defined.
Figure 9 - TMR ratio of MTJs with an MgO barrier. At room temperature (RT) a 500% signal is achieved by Tohoku-Hitachi in 2007. At 5K temperature that value raises to 1010% [8].___________________ 11
Figure 10 – Layers from samples N2TJ68 (left) and TJ1268 (right). All samples N2TJ68 (A,B,C,D) were deposited at same time at sputtering deposition machine, Nordiko2000, while the 6-inch sample TJ1268 was deposited at ion beam deposition machine, Nordiko3000., both at INESC-MN. ______________ 12
Figure 11 - Transmission electron microscopy (TEM) picture from the sample TJ1268, with AlOx barrier taken at Glasgow University. _________________________________________________________ 13
Figure 12 - Picture of the VSM model 880 at INESC-MN. __________________________________ 14
Figure 13 – Typical VSM measurement of a bulk magnetic tunnel junction sample with SAF after annealing at 320ºC (set temperature) for 1 hour. Analysing the curve from the negative to the positive field direction: First, all layers are aligned with the negative field (magnetic moment is minimum), then when the field becomes less negative, the SAF ferromagnetic 1 layer (near the barrier) rotates due to the coupling interaction between the ferromagnetic layers of the SAF. When the zero field is reached the free layer rotates following the field direction. Increasing the positive field the SAF ferromagnetic 2 layer (near the antiferromagnetic layer) rotates, and all layers become align with positive field (magnetic moment is maximum)..The slope difference between the SAF ferromagnetic layers rotation is related with their differences in thickness. The SAF ferromagnetic 1 layer is thicker than the SAF ferromagnetic 2 layer, so its moment is bigger and the rotation slower, consequently the slope is higher. _________ 15
Figure 14 - Schematic view of the deposition chamber of the sputtering deposition system, Nordiko2000 used to deposit the 1-inch MgO barrier samples. There are 6 magnetrons and a water-cooled substrate table with 12stations. A permanent magnet array in one of the stations provides a 30Oe field during the deposition. The samples N2TJ68_A, N2TJ68_B, N2TJ68_C and N2TJ68_D, were deposited at same time [6]. _________________________________________________________________________ 17
Figure 15 - Schematic view of the deposition chamber of the ion beam deposition system Nordiko3000 used to deposit the 6-inch TJ1268 sample. The system has two ion sources allowing for deposition or ion milling in the same chamber. A permanent magnet array in substrate holder provides a 40Oe field during the deposition [6]. ____________________________________________________________ 18
viii
Figure 16 – Schematic of etching processes [12]. ________________________________________ 20
Figure 17 – Consequences of an isotropic and anisotropic etching on the final shape [12]._________ 20
Figure 18 - Dry processes etchings with plasma, with special enhance on physical processes sputter etching and Ion milling [11].__________________________________________________________ 20
Figure 19 – (Left) Outside view of Nordiko 3600 ion milling system. (Right) Inside view of Nordiko 3600 system and its dimensions. The sample is placed on the table at wafer level. The table can rotate, giving homogeneity to the etching and be levelled, controlling in this way the angle between the samples and the etching direction, ". _____________________________________________________________ 21
Figure 20 – Close view of the etching gun and the etching direction at Nordiko3600______________ 22
Figure 21 – Etch rate dependence on surface area of etchable material (A) and on feature aspect ratio (B) [11]. _________________________________________________________________________ 23
Figure 22 - SVG Resist coater and developer track: Track system for spin coating of photoresist (1.2 to 2 "m thick) and for development of post-exposed wafers (cassettes of 25 Si wafers of 6 inch). The room has a filtered light (yellow light) to not interfere with photoresist properties [5].___________________ 25
Figure 23 - Heidelberg Instruments Direct Write Laser Lithography System: Direct write lithography system utilizing a HeCdlaser (#=442 nm (g-line)/ write lens NA= 0.85) capable of critical dimensions down to 0.8 "m. System works with software masks [11].___________________________________ 25
Figure 24 - Etching process: The material to pattern is deposited on the substrate (1). After lithography the photoresist covers the material (2) that is protected during the etching step (3). After photoresist removal (resiststrip) the material has the same lithographed pattern (4) [6]._____________________ 26
Figure 25 - Lift-off process: The substrate is coated with photoresist (1) and the pattern is exposed (2). The new layer deposited on the sample only adheres to the uncovered areas (3).The material on the photoresist is ’lifted-off ’ upon chemical removal of the photoresist and the desired pattern is obtained [6]. _____________________________________________________________________________ 26
Figure 26 - MTJ structure after the first etching that defines the bottom electrode. A – Side view; B -Top view; C- microscope picture from sample N2TJ68 after etching and resiststrip. __________________ 27
Figure 27 - A – Side view after 2nd etch; B -Top view after 2nd etch; C- microscope picture with filter (green) from sample N2TJ68 with 1x5 µm2 area after 2nd exposure. This filter is needed because if the developing time was not enough, it’s possible after observation to develop more time, until the junction’s area gets the right. If unfiltered light is used such isn’t possible because the properties of the photoresist would change. ____________________________________________________________________ 28
Figure 28 - Side view of Nordiko 3600 machine with chamber door removed. The so-called etching angle, ", is the angle between etching direction (red line) and the substrate holder direction (light blue line).____________________________________________________________________________ 29
Figure 29 - Side view of the junction area. A - after insulating passivation layer deposition; B - after liftoff opening the contact to the junction area and pads. The photoresist from the second etch step is removed after the passivation layer deposition. This liftoff process opens a via in the insulating layer aligned with the top junction electrode. A via is also opened to the larger contact pad areas in an similar way. C- Microscope top view image, from TJ1268 after liftoff (no oxide on lighter regions). _________ 30
Figure 30 - UHV II machine for oxide deposition by sputtering from a SiO2 or Al2O3 ceramic target. This system is used for the oxide layer deposition in the tunnel junction process. The machine is installed in a class 10000 clean room. Details on the machine are provided in the table on left. A - chamber; B – Ar bottle for sputtering [5]. _____________________________________________________________ 31
Figure 31 - Top view of the sample N2TJ68 after expose and develop.________________________ 31
Figure 32 - Top view of the Nordiko 7000. There are 4 process modules: flash anneal (module1), soft sputter etching (module2),TiW(N2) sputter deposition (module3) and AlSiCu sputter deposition (module4) [6]. ____________________________________________________________________ 32
Figure 33 – A - Cross-section view of the contact leads defined by lift-off of the Al metallization. B- Top view of the sample N2TJ68 after Al lift-off, with the top electrode contact pads and lead. __________ 32
Figure 34 – Final outlook of the samples after patterning. (Right) sample N2TJ68 deposited on a glass substrate, with 3 columns, each one with 9x20 junctions. (Left) sample TJ1268 deposited on a silicon substrate. The sample has several designs, 1 column with 9x19 junctions, matrix of junctions, lines and squares matrixes. _________________________________________________________________ 32
Figure 35 – TMR as function of the RxA after annealing the sample N2TJ68_B at 280ºC for 1h and after annealing the same sample at 320ºC for 1h._____________________________________________ 33
ix
Figure 36 – Heating and cooling steps in the big annealing setup. All TJ1268 samples were annealed with these temperature conditions, but only for 30 minutes. The sample N2TJ68_B was first annealed with these conditions and then with the conditions set in Figure 38. ___________________________ 34
Figure 37 – Scheme and picture of the big annealing setup. ________________________________ 34
Figure 38 - Heating and cooling steps in the small annealing setup. All N2TJ68 samples were annealed with these conditions. ______________________________________________________________ 35
Figure 39 - The temperature calibration curve used in small annealing.________________________ 35
Figure 40 - Pictures of the small annealing setup. The removable glass chamber surrounds the copper basis, and the magnet coil separation is adjusted to set the magnetic field at the region of the sample [5]. _____________________________________________________________________________ 36
Figure 41 – Transfer curve from the sample N2TJ68_B, obtain from the electrical measure. _______ 37
Figure 42 – Setup used to measure the transfer curves of the tunnel junctions [5]. _______________ 38
Figure 43 – The 2-contact and the 4-contact configurations schemes [5]. ______________________ 38
Figure 44 – Schematic of a tunnel junction in the second etch with material redeposition (orange spots) on sidewalls, shorting parts of the barrier. _______________________________________________ 39
Figure 45 – Electrical circuit proposed to simulate the redeposition effect on the TMR. The model consists in a perfect junction resistance, RTJ, shorted by a parallel resistance, RS, which depends on the amount of redeposition [6]. __________________________________________________________ 39
Figure 46 – Magnetic tunnel junction TJ1268 deposited on 6-inch silicon wafer size at Nordiko3000 machine, sliced by diamond saw in 12 smaller samples (dies with 2x2 cm2). The samples structure is represented in figure 9, and has 5Å thick AlOx barrier. The picture on right is from 6-inch sample GTJ301 previous processed and not from TJ1268, but the dies design is the same. ______________ 42
Figure 47 - Design mask layout with the test structures to be etched by ion beam etching. A- Lines width from 1 to 100 µm; B – MTJ; C- Matrix of rectangles 3"m x 6"m; D – Matrix of large rectangles; E – Matrix of MTJ. ____________________________________________________________________ 43
Figure 48 – Plot TMR vs RxA obtained from samples with same structure (TJ1268_6, TJ1268_8, TJ1268_9, TJ1269_11), see figure 9 (right), but with different etching angles during the junction pillar definition, see Table 5. _____________________________________________________________ 44
Figure 49 - TEM pictures from patterned parts of the 6-inch wafer magnetic tunnel junction TJ1268. In both samples the etching didn’t reach the AlOx barrier. These samples are not represented in the plots. (Top) The sample was etched with one angle (one slope). (Bottom) The sample was etched with two angles (two slopes).________________________________________________________________ 45
Figure 50 – Schematic of lateral oxidation with oxygen plasma. The angle between the sample and the plasma is 10º. ____________________________________________________________________ 46
Figure 51 - Plot TMR vs RxA obtained from samples with same structure (TJ1268_6, TJ1268_4, TJ1268_5, TJ1269_7), see __________________________________________________________ 47
Figure 52 – Is the same plot that is in Figure 51, but ignores the 3 points with larger resistance. The reasons of this choice are explained above. _____________________________________________ 48
Figure 53 – Comparison between samples with similar oxidation times. The samples are TJ1268_4 with 20s of oxidation and TJ1268_5 with 50s. Little difference can be noticed. ______________________ 48
Figure 54 - Plot TMR vs RxA from samples with MgO barrier (N2TJ68_A, N2TJ68_B, N2TJ68_C), with different times of oxygen plasma oxidation. For 100 s of oxidation a cluster is formed with little increase of the resistance and with considerable improvement on the TMR signal. The etching conditions are standard and equal for all N2TJ68 samples. All samples have the same junction’s areas. The areas range is from 3 "m2 to 15"m2. In these experiments the areas of 1 "m2 and 2 "m2 aren’t used in order to avoid some doubts about the real size of those areas. The causes are related with the developing time, where a broad range of areas makes hard the decision when to stop the developing process. If it’s stopped to soon the larger areas could not be well defined, if it is decided to proceed with more time of developing the smaller areas features could vanish (it isn’t unusual).__________________________ 50
Figure 55 – Plot TMR vs RxA from sample N2TJ68_B. Each colour is related with the junction area. As observed the redeposition is not directly linked with the area junction, since for all junctions’ sizes the TMR values are widely spread, as well the resistances. ____________________________________ 50
Figure 56 – (Left) TMR sensor map with a colour scale indicating the TMR value of each MTJ sensor, that goes from a lower intensity represented by red colour to higher intensity, the blue colour. Each square represents a magnetic tunnel junction sensor. In total there are 20x9 sensors. The position #4 in
x
the lines is never measured because the needles are broken for that position in the probe card of the automatic measure setup (KLA). (Right) Magnetoresistance histogram and statistics, the numbers on top of each column is the number of junctions with the same TMR. ___________________________ 51
xi
Index of Tables
Table 1 - Standard set values for magnetic tunnel junction deposition at PVD Nordiko 2000. ..................17
Table 2 - Standard set values for magnetic tunnel junction deposition at IBD Nordiko 3000. ....................18
Table 3 – Etching rates for certain conditions of the assist gun and assist neutralizer at Nordiko3600 machine. The assist gun is responsible to create the plasma and the assist neutralizer produces a beam of electrons to neutralize the incoming charge ions in order to prevent over-charge of the sample and sparkles. The etching rate also depends on the angle of etching. The pan angle value it is always 10º less then the real angle, due to the inclination of the etching gun. ....................................................................22
Table 4 – Annealing conditions used on samples TJ1268 and N2TJ68. The sample N2TJ68_B, was twice annealed, first for 1h at 280ºC and then for 1h at 320ºC. .........................................................................36
Table 5 – Etching conditions used during the pillar definition for samples with same structure (TJ1268_6, TJ1268_8, TJ1268_9, TJ1269_11), see figure 9 (right). The total time of etching is 400 s. The etching angle conditions used for sample TJ1268_6, are the standard ones used at INESC-MN. ...43
Table 6 – Standard values read in the reactor chamber of Nordiko 3600 during the etching process. The sub pan value, X, depends on the desired etching angle, its value is always 10º less then the real angle, due to the inclination of the etching gun. The sub rotn is the speed that the support table rotates in order to have a homogeneous etching in all sample. ......................................................................................................44
Table 7 – Oxidation times (Y) after etch the sample for 300s at 70º and 100s at 40º. The sample TJ1268_6 is the standard sample used to compare. ............................................................................................46
Table 8 - Standard values read in the reactor chamber of Nordiko 3600 during oxidation beam step. The reading values for the etching steps are given in Table 6. The current used in oxidation step is much lower than the one used to etch; the reason is to prevent the lateral etching during the oxidation that would decrease the junction’s area...........................................................................................................................47
Table 9 – O2 plasma oxidation times used in the samples .................................................................................49
xii
List of Acronyms
INESC-MN - Instituto de Engenharia de Sistemas e Computadores - Microsistemas e Nanotecnologias
TMR - Tunnel Magnetoresistence
MTJ – Magnetic Tunnel Junction
SAF – Synthetic Antiferromagnetic
FM – Ferromagnetic
MRAM – Magnetic Random Access Memory
DRAM – Dynamic Random Access Memory
SRAM – Static Random Access Memory
EEPROM – Electrically Erasable Programmable Read Only Memory
RT – Room Temperature
PVD – Physical Vapour Deposition
CVD – Chemical Vapour Deposition
IBD – Ion Beam Deposition
DWL – Direct Write Laser
1
1. Introduction Nowadays society, the metallic magnetic materials are very common, and applications are
spread out in various products and devices. In middle of nineties, a small revolution took
place in the field of magnetism and electrical transport. Jagadeesh Moodera from MIT
measured ferromagnetic-insulator-ferromagnetic tunneling in CoFe /Al2O3 /Co and NiFe
junctions at room temperature with 11.8% tunneling magnetoresistance [2]. This phenomena,
where two magnetic layers separated by a very thin insulating film (barrier) reveals a large
resistance changes, due to spin-dependent tunneling effect, was already known since M.
Julliere discovered it using Fe-Ge-Co junctions back in seventies [1], still it was never verified
before at room temperature. This breakthrough permitted to give one step ahead in the
development of magnetoresistive random access memory with attributes that are competitive
with semiconductor memory. The large magnetoresistance signal of this material enables fast
memory-read operations. In addition, the memory is nonvolatile (the information remains
stored when the power is turned off) because the information is stored in the magnetic state
of the bit. MRAM has similar speeds to SRAM, similar density of DRAM but much lower
power consumption than DRAM, and is much faster and suffers no degradation over time in
comparison to Flash memory. It is this combination of features that some suggest make it the
"universal memory", able to replace SRAM, DRAM, EEPROM and Flash. This explains the
huge amount of research being carried out into developing it. The large signal also makes
magnetic tunnel junction material an attractive candidate for magnetic-media read heads and
other types of sensor applications, such as biosensors.
Figure 1 - In the MRAM, information stored in one of the magnetic layers, can be erased by a current Ix through the top conductive paths creating an in-plane magnetic field (Hx). However, to address only a single element also the bottom lines are used to create a magnetic field Hy to reduce the effective switch field [3].
Presently, one of the goals in tunnel junction research is to decrease the junction resistance.
The main reason is that lower resistances allow better signal to noise ratios and faster access
times. The resistance can be lowered reducing the thickness or the barrier height. The first
2
possibility is preferred, as it is easier to control the thickness of the insulator, than to change
the material’s barrier height. The thicknesses can only be reduced to a certain point, where
the barrier starts to be discontinuous. Therefore, research has been initiated looking at the
possibility of finding materials with lower barrier height. At lower resistances the effect of
redeposition on the junction sidewalls becomes even a greater issue of concern, because it
will reduce the effective resistance of the tunnel junction in a more effective way, and
consequently the TMR. In this thesis lateral oxidation via oxygen plasma is used after the
second etching in order to reduce the redeposition by oxidizing the redeposited conductive
material gather on sidewalls of the junction. These results are a part of a paper, “Ion Beam
Assisted deposition of MgO barriers for magnetic tunnel junctions” [4], submitted to the
Journal of Applied Physics on September 12th 2007.
3
2. Spin Dependent Tunneling The spin-dependent tunneling effect is due to the wave-like nature of electrons, that allows a
penetration of electron waves into an insulating barrier region, Figure 2. As a result, there is a
finite probability for an electron to tunnel through the energetically forbidden barrier. When a
finite bias voltage is applied between two ferromagnetic (FM) electrodes separated by an
insulating film barrier (I), electrons are able to tunnel from filled electron state (FM1) towards
unoccupied states in the second electrode (FM2).
Figure 2 – Tunnelling between the two ferromagnetic is governing by the density of states at Fermi level when a small bias voltage V is applied across the junction of thickness t and height ! (depends on material characteristics) [3].
Therefore, the contribution to the tunneling current is directly proportionally to the density of
states in FM1 times the density of states in FM2, both at the Fermi level. The FM1/ I /FM2
device is called magnetic tunnel junction (MTJ). Assuming that the spin is conserved during
the tunneling process, which appears to be a reasonable assumption when tunneling across
ultra-thin clean oxide barriers [3]. Then a spin up and spin down current can be discriminated,
with magnitudes that strongly depend on the orientation of magnetization of the two
ferromagnetic electrodes. For better description of these phenomena, is useful to use the
quantity I/V known as the tunnel conductance, G, that is defined as the inverse of the
tunneling resistance, R, therefore proportional to the current, and like the current, to the
density of states of the two electrodes.
When both the ferromagnetic electrodes have the same magnetization direction, majority
electrons are tunneling into the majority band, minorities to minority band. In this case, the
conductance for parallel alignment, Gp, can be written as:
Gp$ U1U2+D1D2 ( 1 )
FM2
FM1 FM2
FM1
4
where Ui is spin up density of states in electrode i=1,2 and Di the spin down density of states
in electrode i=1,2. In contrast, when ferromagnetic electrodes have opposite magnetization
directions, the electrons are tunneling from minority into majority bands and vice-versa. The
conductance for antiparallel alignment, Gap, can be define as
Gap$ U1D2+D1U2 ( 2 )
For ferromagnetic materials at the Fermi level there is an imbalance in the density of states of
spin up U and spin down D direction that creates the magnetic moment in these materials.
This means that U >D or vice-versa. Consequently, the conductance between electrodes with
parallel magnetization directions is higher than the antiparallel configuration, Gp > Gap.
Figure 3 - Band diagram of the tunneling process between ferromagnetic electrodes. The process is shown for an applied bias voltage to the electrodes, with tunneling occurring with spin conservation between equivalent spin states [6].
The relative difference in density of states at the Fermi level is often expressed in the so-
called tunneling spin polarization, P:
!
Pi=Di"U
i
Di+U
i
( 3 )
The substitution in equations (1) and (2) allows writing the relative conductance change as
!
"G
G=Gp #Gap
Gp
=2P
1P2
1+ P1P2
=Gp #Gap
Gap
=2P
1P2
1# P1P2
( 4 )
5
2.1. Signal as Resistance Variation
Magnetic tunnel junctions are usually characterized in terms of resistance (R=1/G). Then the
high conductance parallel magnetization state corresponds to low resistance (Rp=1/Gp), and
the low conductance antiparallel magnetization state corresponds to high resistance
(Rap=1/Gap).
Figure 4 - (Left) Resistance circuit of the parallel state of magnetization. (Right) Resistance circuit of the antiparallel state of magnetization [3].
So, using the previous equations
!
Rap " Rp
Rap
=2P
1P2
1+ P1P2
Rap " Rp
Rp
=2P
1P2
1" P1P2
( 5 )
The tunnel magnetoresistance (TMR) signal is defined as the second of these equations. The
first one is sometimes called as ’pessimistic’ TMR as it results in lower TMR values [6]. These
assumptions lead to a TMR signal only dependent on the electrodes spin polarization.
Throughout this work the signal is calculated as the change in resistance over the low
resistance value Rp
!
TMR =Rap " Rp
Rp
=2P
1P2
1" P1P2
( 6 )
In this approach to the tunneling process, the electrodes are considered to be independent.
The ferromagnetic layers to be used as electrodes should be high polarization materials,
which allow tunnel junctions of higher TMR signal.
1/U1U2
1/D1D2
1/U1D2
1/D1U2
6
Physically the TMR signal is a change in junction resistance that occurs when the angle
between the electrodes magnetization, %, is changed by a magnetic field. As previously
addressed, the resistance changes from lowest resistance (Rp) for parallel alignment to
highest ( Rp+&R ) for antiparellel alignment. For any angle in between, the resistance
corresponds to a fraction of parallel (%=0º) and antiparallel (%=180º) alignments. Thus the
total resistance R can be expressed as:
!
R = Rp +"R
2(1# cos$) ( 7 )
The junction resistance is inversely proportional to the junction area, A. Such comes from the
fact that the number of electrons tunneling is directly proportional to the total electrode area,
giving rise to higher tunneling current for larger areas and therefore lower resistance [6]. This
outcome is very important, because it allows comparing junctions with different areas, since
the so-called resistance area product, RxA, is expected to be fairly constant. This result is
extensively used to study the redeposition. Several times, the term “resistance” is used when
actually RxA product is meant.
7
2.2. Magnetic Response The polarization is not the only important parameter to study the magnetic tunnel junctions.
The response to a magnetic field is equally important since this provides the control over the
signal response.
When a ferromagnetic material is magnetized in one direction, it will not relax back to zero
magnetization when the imposed magnetizing field is removed. The amount of magnetization
it retains at zero driving field is called its remanence, Mr. It must be driven back to zero by a
field in the opposite direction; the amount of reverse driving field required to demagnetize it is
called its coercivity, Hc. If an alternating magnetic field is applied to the material, its
magnetization will trace out a loop called a hysteresis loop. This property of ferromagnetic
materials is useful as a magnetic "memory", because it is important to have a single
resistance state (magnetization) for each value of the applied field. For some sensors, for
instance biosensors, this coercivity is not desirable, since very high sensibility to very low
fields is required, still it is important to have a single resistance state (magnetization) for each
value of the applied field.
Figure 5 – Hysteresis loop curve [15].
The most common ferromagnetic materials used in magnetic sensors are based on magnetic
transition metals like Ni, Fe, Co and their alloys. All these materials belong to the so-called
soft magnetic materials, because of their low coercive field, and high permeability,µ, allowing
high sensibility to small magnetic fields. This low coercivity allows for magnetization reversal
in small magnetic fields ('100 Oe). The hysteresis loops of these materials are largely
determined by their anisotropy, K. The anisotropy can result from the crystalline structure,
and or from the shape anisotropy of the patterned element. The coercive field of a magnetic
-Hc Hc
8
layer with uniaxial anisotropy and total magnetization M measured along the easy axis
direction is determined by
!
Hc
=2K
M ( 8 )
where the total magnetization of the layer is determined from the magnetic moment, m, and
the volume, V of the material, M = mV . The easy axis can be set upon deposition in a
magnetic field, resulting in uniaxial anisotropy.
9
3. Magnetic Tunnel Junction Architecture
For a clear understanding of the architecture of a magnetic tunnel junction, it is useful to
address some of the history and reasons behind its evolution, especially its adaptation for
MRAM application. Figure 6 illustrates structures used to engineer the response of magnetic
tunnel junctions in ways beneficial for memory applications.
Figure 6 - Evolution of tunnel junctions engineered for MRAM applications. (a) Basic MTJ structure. (b) MTJ structure with a ferromagnetic layer pinned by an antiferromagnetic layer (c) MTJ structure with synthetic antiferromagnetic pinned layer (SAF). (d) Structure in which both the pinned and free elements consist of antiferromagnetically coupled pairs [7].
Figure 6(a) is the basic magnetic tunnel junction structure, used for the early studies, with
specific characteristics illustrated below.
Figure 7 - Tunneling at 295 K in a CoFe–Al2O3–Co junction [7].
10
In principle this structure could be made to work for a memory if the coercivity of one of the
layers, a ‘‘reference’’ layer, were much higher than that of the other. However several
difficulties are pointed out. First, the field excursion would have to be restricted to being lower
than a maximum value so that the high-coercivity layer would never be disturbed. Even so, it
is possible that repeated low-field excursions could reverse small domains in the higher-
coercivity reference layer that have no way of returning to their original state.
The possibility of disturbing the reference layer could be avoided by pinning one the magnetic
electrodes via exchange coupling to an adjacent antiferromagnet, as illustrated in Figure 6(b).
In this case only the free layer electrode reacts to the field. The TMR response reflects the
hysteresis of the other so-called free layer and has a response curve more suitable for
memory.
The magnetic offset caused by fields emanating from the pinned layer can be avoided by
replacing a simple pinned layer with a synthetic antiferromagnetic (SAF) pinned layer, which
consists of a pair of ferromagnetic layers antiferromagnetically coupled through a ruthenium
(Ru) spacer layer. The lower layer in this artificial antiferromagnet is pinned via exchange
bias, as shown in Figure 6(b). This flux closure increases the magnetic stability of the pinned
layer and reduces coupling to the free layer.
Figure 8 – Hysteresis curves from samples with SAF. Sample TJ1268_7 with 5Å thick AlOx barrier. Sample N2TJ68_B with 8Å thick MgO barrier. As is noticed the TMR with MgO barrier is much higher.
AlOx Barrier
11
Figure 8 - TMR ratio of MTJs with an MgO barrier. At room temperature (RT) a 500% signal is achieved by Tohoku-Hitachi in 2007. At temperature of 5K that value raises to 1010% [8].
12
3.1. Layers Role As explained before the MTJ structure cannot be described as being a simple structure of two
ferromagnetic electrodes separated by a thin insulating barrier. Both electrodes are in reality
comprised of several layers each with it’s own objective, some of them were already
addressed.
Figure 9 – Layers from samples N2TJ68 (left) and TJ1268 (right). All samples N2TJ68 (A,B,C,D) were deposited at same time at sputtering deposition machine, Nordiko2000, while the 6-inch sample TJ1268 was deposited at ion beam deposition machine, Nordiko3000, both at INESC-MN.
Apart from the evident role of the insulating layer, to provide the tunneling barrier, the other
layers can be divided in the following groups. The buffer layer, which the main purpose is
providing a low resistance contact to the junction. The ferromagnetic layer at each side of the
barrier provides the electron spin polarization at the base of the spin dependent tunneling
effect. Since the tunnel junction signal also depends on the relative magnetization directions
of the ferromagnetic layers, the antiferromagnetic layer and the SAF layers used have the aim
of keeping the magnetization direction of the ferromagnetic fixed via exchange-coupling to set
the reference direction and it’s magnetization should only rotate for fields much higher than
the reversal fields of the free layer. The capping layer is used to protect the full structure from
external elements, either chemicals used in processing or oxidation in ambient atmosphere,
in particular the TiWN2 is also used as an anti-reflecting, very useful during the direct writing
laser.
Passivation
Top Electrode
Barrier
Bottom Electrode Pinned
Buffer
Oxide Layer
Substrate
SAF
Barrier
Passivation
Free Ferromagnetic
SAF Ferromagnetic 1
Antiferromagnetic
Substrate
SAF Ferromagnetic 2
13
Figure 10 - Transmission electron microscopy (TEM) picture from the sample TJ1268, with AlOx barrier taken at Glasgow University, Department of Astronomy and Physics.
AlOx barrier
14
3.2. Layers Magnetic Characterization
Right after the layers deposition, the magnetic tunnel junction can be characterized
magnetically, using the vibrating sample magnetometer (VSM). The sample is attached to a
rod that vibrates along the vertical axis. The sample at the end of the rod is displaced in a
uniform field. The flux changes created by the movement of the sample produces a voltage in
a set of pick-up coils proportional to the flux rate of change (Faraday’s law). The signal from
the pick-up coils is fed to a lock-in amplifier that uses the rod displacement as reference
signal. The maximum field between the magnets is 13 kOe, for a pole separation of 1.8 cm
and it has 1x10-6 emu of sensitivity [5].
Figure 11 - Picture of the VSM model 880 at INESC-MN.
It is especially useful to interpret transport measurements, acting as an independent
characterization of the junction’s magnetic behavior. It provides information on the exchange
coupling field of the pinned layer, the coercive fields of the free layer and if these two layers
are decoupled and acting individually. There is also information on the magnetic moment of
the film structure.
15
Figure 12 – Typical VSM measurement of a bulk magnetic tunnel junction sample with SAF after annealing at
320ºC (set temperature) for 1 hour. Analysing the curve from the negative to the positive field direction: First, all
layers are aligned with the negative field (magnetic moment is minimum), then when the field becomes less
negative, the SAF ferromagnetic 1 layer (near the barrier) rotates due to the coupling interaction between the
ferromagnetic layers of the SAF. When the zero field is reached the free layer rotates following the field direction.
Increasing the positive field the SAF ferromagnetic 2 layer (near the antiferromagnetic layer) rotates, and all layers
become align with positive field (magnetic moment is maximum)..The slope difference between the SAF
ferromagnetic layers rotation is related with their differences in thickness. The SAF ferromagnetic 1 layer is
thicker than the SAF ferromagnetic 2 layer, so its moment is bigger and the rotation slower, consequently the slope
is higher.
( ( (
( ) (
) ) (
) ) )
* Free layer * SAF Ferromagnetic 1 * SAF Ferromagnetic 2
16
4. Magnetic Tunnel Junction Fabrication
During the fabrication of a tunnel junction two distinguish processes play a key role in the final
outcome, the deposition and the etching. While deposition is the formation of a thin layer by
sticking of gas phase species on the surface, the etching is the removal of material by
particles collision and/or by formation of volatile species (“inverse of deposition”) [12]. The
propose of this thesis is to optimize the etching conditions used for a crucial step of MTJ
fabrication, therefore this process will be take in hand with more detail. However, being
deposition a key process, it is useful to address some considerations about it. These
processes are used jointly with lithography to pattern the magnetic tunnel junctions to become
possible to test their properties.
4.1. Film Deposition Deposition can occur by physical or chemical means (ex: Chemical Vapor Deposition) [16].
Concerning the magnetic tunnel junctions manufactured at INESC-MN, physical processes
are used in deposition, such as sputtering and ion beam deposition. The physical process
behind these two methods is similar, as in both techniques the material is sputtered from a
target by accelerated ions impacting on the target surface. In the collision process, the energy
of the incident ion is transferred to a target atom, and if high enough, the atom is sputtered
from the target in the direction of the sample, and thus deposited on the substrate. The
incident ions should be from an inert gas (e.g. Ar, Kr, Xe) to exclude a chemical reaction with
the material, unless this is wanted, for example, depositing oxides where oxygen might also
be added. It is desired that the process take place at the lowest possible pressure, so less of
the inert gas ions are incorporated in the deposited film.
The major difference between these two deposition processes resides in the place where the
ions responsible for the sputtering are generated. In sputtering all action, including the ions
generation, takes place in the chamber where the substrate and target are, Figure 13. In ion
beam deposition the ions are originate from an ion source apart from the deposition
17
chamber.The plasma is confined to what is called the ’deposition gun’, Figure 14, away from
the target. [17]. This allows deposition pressures in the 10-5 - 10-4 Torr range, one order of
magnitude lower than what is achievable in magnetron sputtering systems. The ions are
extracted from the source by a first inner grid with negative bias. The next grid has a positive
bias and it is the potential difference to the third external grounded grid that determines final
ion energy. The beam hitting the target has a reduced ion density, compared to magnetron
sputtering, resulting in lower deposition rates. This can actually be beneficial to control the
thickness of very thin layers <10Å [6].
Figure 13 - Schematic view of the deposition chamber of the sputtering deposition system, Nordiko2000 used to deposit the 1-inch MgO barrier samples. There are 6 magnetrons and a water-cooled substrate table with 12 stations. A permanent magnet array in one of the stations provides a 30Oe field during the deposition. The samples N2TJ68_A, N2TJ68_B, N2TJ68_C and N2TJ68_D, were deposited at same time [6].
Film Deposition Conditions
Set Values Material Deposition rate
(Å/s)
Voltage
Current
Ar Flow
Base Pressure
LoadLock Pressure
Dep. Pressure
MgO Dep.Pressure
332 V
40 mA
9.8 sccm
7x10-9 Torr
1x10-5 Torr
5.1 mTorr
20 mTorr
P V D
N 2 0 0 0
Ta
Ru
MnPt
CoFe
CoFeB
MgO
0.52
0.40
1.1
1.1
0.33
0.18
Table 1 - Standard set values for magnetic tunnel junction deposition at PVD Nordiko 2000.
18
Figure 14 - Schematic view of the deposition chamber of the ion beam deposition system Nordiko3000 used to deposit the 6-inch TJ1268 sample. The system has two ion sources allowing for deposition or ion milling in the same chamber. A permanent magnet array in substrate holder provides a 40Oe field during the deposition [6].
Film Deposition Conditions
Set Values Material Deposition rate @ 22mA
(Å/s)
Power
V +
V –
Current
Xe Flow
Rotation
Pan
Base Pressure
Dep. Pressure
115 W
+1200V
-300 V
22 mA*
1.6 sccm
50%
80º
6x10-8 Torr
3.5x10-5 Torr
I B D
N 3 0 0 0
Al
CoFe
CoFeFe
CoFeB10
CoFeB20
MnIr
Mg
NiFe
Ru
Ta
0.21 (0.12 @15mA)
0.23
0.20
0.18
0.14
0.23
0.60
0.24
0.19
0.15
*Grid Set 24 mA, but real value 22 mA
Table 2 - Standard set values for magnetic tunnel junction deposition at IBD Nordiko 3000.
The most challenging step in tunnel junction deposition is the insulating barrier. The barrier is
required to be continuous and without pinholes at thicknesses below 20 Å, to insure current
conduction through tunneling only. It must be a non-magnetic oxide to prevent spin flips
events in the tunneling process and form a stable oxide that preserves the integrity of the
ferromagnetic electrodes.
19
4.2. Etching Process As previously explained the material is etched as result of particles collision and/or by
formation of volatile species. Thus, the etching processes can be described as having a
physical and/or a chemical nature. Depending on its nature some etching mechanism can be
discriminated (see Figure 15) [12].
Physical - ejection of atoms from surfaces due to energy ion bombardment.
- unselective process- only depends on the surface binding energy and the
masses of the targets and projectiles
- low etch rate - one atom per incident ion
- anisotropic process - very sensitive to the angle of incidence of the ion
- only process that can remove involatile products from the surface
Chemical - discharge supplies gas phase etchant atoms or molecules that chemical react
with the surface to form gas phase products. The etchant can be also a chemical solution
(wet-etch), where no plasma is need (very common and useful process).
- highly chemical selective
- isotropic
- etch products must be volatiles
- high etch rate - flux of etchants can be high in processing discharges.
- However the rate is limited by the complex set of reactions at the surface
leading to formation of etch product.
Ion-enhanced energy driven – can combine the both effects: etchants atoms and energetic
ions.
Ion-enhanced inhibitor - the discharge supplies etchants, energetic ions, and inhibitor
precursor molecules that adsorb or deposit on substrate to form a protective layer or polymer
film.
20
Figure 15 – Schematic of etching processes [12].
The geometry and the area of a magnetic tunnel junction is strongly related with its electrical
and magnetic response, therefore it is required an anisotropic etching process for a good
shape control.
Figure 16 – Consequences of an isotropic and anisotropic etching on the final shape [12].
A magnetic tunnel junction is a compound of different materials with different chemical
properties. Therefore, a chemical selective process has no use in this case. So, taking in
account the need of an anisotropic process with no chemical selectivity use, the obvious
choice is to use a physical process, the sputtering or the ion beam milling, both dry etching
processes with plasma. Plasma has some properties that makes it interesting from etching
point of view, such as, gas phase production of reactive species (electrons, ions, activated
neutrals), low thermal budget and surface ion bombardment.
Figure 17- Dry processes etchings with plasma, with special enhance on physical processes sputter etching and Ion milling [11].
21
The main difference between these two physical processes is the control over the ions
incidence angle. In the sputter etch process the electric field imposes an almost normal
incidence of Ar+ ions at the surface of the sample, which cannot be changed. Not all ions
arrive at 90º to the surface, since the relatively high gas pressure results in some collision and
deflection that decreases the directionality of the incoming ions. This creates a lateral
displacement component of the etched material causing redeposition at the sidewalls of the
magnetic tunnel junction pillar. Since the angle of incidence cannot be changed, there is no
direct way to control the redeposition process. In ion beam milling, the ion beam direction is
defined by the acceleration grids at the exit of the plasma source. Lower gas pressure in the
chamber, allows less dispersion in the beam direction due to ion interaction. The incidence
angle of the ions on the sample can be controlled changing the angle of the substrate table.
Concerning the magnetic tunnel junctions fabrication, the etching process used and optimized
is the ion milling system Nordiko3600 at INESC-MN.
Figure 18 – (Left) Outside view of Nordiko 3600 ion milling system. (Right) Inside view of Nordiko 3600 system and its dimensions. The sample is placed on the table at wafer level. The table can rotate, giving homogeneity to the etching and be levelled, controlling in this way the angle between the samples and the etching direction, ".
Opening the door chamber
22
Figure 19 – Close view of the etching gun and the etching direction at Nordiko3600
At Nordiko3600 the etch rate varies with the etching direction, as shown in Table 3.
Assist Gun Standard Reading
Assist Neutralizer
Standard Reading
Pan Angle Etching rate
(Å/s)
Grid 1 Volts + 735 V Current 130 mA 60º 1.05
Grid 1 Current 105 mA Argon Flow 3.0 sccm
30º 1.15
Grid 2 Volts - 350V
Argon Flow 10 sccm
Table 3 – Etching rates for certain conditions of the assist gun and assist neutralizer at Nordiko3600 machine. The assist gun is responsible to create the plasma and the assist neutralizer produces a beam of electrons to neutralize the incoming charge ions in order to prevent over-charge of the sample and sparkles. The etching rate also depends on the angle of etching. The pan angle value it is always 10º less then the real angle, due to the inclination of the etching gun.
The etch rate doesn’t depend only on the machine conditions or etching angle, there are
some factors related with the sample that is being processed. For instance, the etch rate of a
material depends on the surface area of the etchable material, a large unmasked area
exposed to the beam consumes more etch species than a single trench, Figure 20(A).
Moreover, the feature aspect size ratio has also influence on etch rate Figure 20(B).
23
Figure 20 – Etch rate dependence on surface area of etchable material (A) and on feature aspect ratio (B) [11].
A B
24
4.3. Lithography and Pattern Transfer
Tunnel junction sensors need some patterning process to be tested for their electrical
properties. Being a sensor where current flows perpendicular to the plane, the measurement
requires an electrical contact to the top and bottom electrodes.
Lithography (Photolithography or optical lithography) is a process used in microfabrication to
selectively remove parts of a thin layer. It uses light to transfer a geometric pattern from a
mask to a light-sensitive polymer, photoresist, on the substrate. The photoresist layer will
protect the covered areas from etch processes, Figure 23, or prevent new deposited layers to
adhere in those areas, liftoff, Figure 24. There are different systems of lithography. In this
thesis is used a direct-write laser, DWL 2.0 from Heidelberg, using a 442nm HeCd laser to
write a pattern on 1.5 µm thick photoresist. The ultimate resolution of this system is 0.8 µm
and alignment precision of 0.25 µm. The advantage of direct laser write is that the laser scan
is software controlled, masks can be easily changed at no cost, however the exposure is
slower comparing with other system (ex: hard masks), the exposure time depends on the
area to be exposed. The following steps resume the lithography process:
Step 0: Draw a mask in a CAD program and transfer it to the lithography system.
Step 1: Vapor prime: Dehydration bake to remove adsorbed water from the substrate
surface and coat it with a surfactant (HMDS:hexamethyldisilazane) to promote a good
photoresist adhesion to the substrate (photoresist adhesion to Si wafers is poor).
Step 2: Coating the substrate with 1.5 µm thick photoresist by spinning and soft
baking the resist. The spinning gives the resist thickness and the soft baking remove
solvents and stress promoting its adhesion to wafer.
Coating parameters (recipe 6/2): spin at 2.8 krpm for 40 sec.,acceleration 50 krpm
(soft) bake at 85ºC for 60 sec.
25
Figure 21 - SVG Resist coater and developer track: Track system for spin coating of photoresist (1.2 to 2 "m thick) and for development of post-exposed wafers (cassettes of 25 Si wafers of 6 inch). The room has a filtered light (yellow light) to not interfere with photoresist properties [5].
Step 3: Laser exposure (DWL): The laser is swept over the area to be exposed,
turning on and off, as defined in a mask pattern, breaking the chemical bonds of the
polymer.
Figure 22 - Heidelberg Instruments Direct Write Laser Lithography System: Direct write lithography system utilizing a HeCdlaser (#=442 nm (g-line)/ write lens NA= 0.85) capable of critical dimensions down to 0.8 "m. System works with software masks [11].
Step 4: Developing: For positive photoresist, removes the unexposed areas of the
photoresist, leaving the pattern to be transferred on the exposed photoresist. If
negative photoresist is used the areas removed are the exposed ones.
Developer parameters (recipe 6/2): Bake at 110ºC for 60s
Cool for 30s
Developer for 60s
26
After pattern the photoresist layer, two possible patterning processes are used, an etch
process that patterns the already existing material on the sample, and the second process,
called lift-off, that patterns new layers being deposited. In order to obtain a tunnel junction
fully capable to be measured several lithography steps are need.
Figure 23 - Etching process: The material to pattern is deposited on the substrate (1). After lithography the photoresist covers the material (2) that is protected during the etching step (3). After photoresist removal (resiststrip) the material has the same lithographed pattern (4) [6].
Figure 24- Lift-off process: The substrate is coated with photoresist (1) and the pattern is exposed (2). The new layer deposited on the sample only adheres to the uncovered areas (3).The material on the photoresist is ’lifted-off ’ upon chemical removal of the photoresist and the desired pattern is obtained [6].
27
4.4. Process Steps Step 1. Bottom electrode definition
The first step in the lithographic tunnel junction fabrication is the definition of the bottom
electrode lead and contact areas.
Step 1.1 - The photoresist is patterned to cover the bottom electrode lead and
contact areas. The side view in Figure 25, shows the photoresist protected area
during the etch step. The top view illustrates the lead area where the junction will be
defined and the two large contact areas to this lead. These larger areas will provide
an outside contact point to the bottom electrode of the junction. The narrow section
will act as the bottom electrode lead, providing the current path to the bottom
electrode at the junction area. The alignment mark is used to align a bottom layer
with the top layer, they have a specific design that is recognize by the DWL software,
that allows the 0.25 "m alignment precision between consecutive layers mention
before.
Figure 25 - MTJ structure after the first etching that defines the bottom electrode. A – Side view; B -Top view; C- microscope picture from sample N2TJ68 after etching and resiststrip.
Step 1.2 - The material uncovered by the photoresist pattern is etched away by ion
beam milling. The junction has to be etched until it reaches the non-conductive layer.
On glass substrate, the MTJ has to be totally etched until the glass, for silicon
substrates it has to be etched until the oxide layer above the silicon substrate. In this
way the short–cut between the several bottom electrodes is prevented.
Step 1.3 – Resiststrip with hot Microstrip (photoresist solvent) and ultrasounds. Rinse with IPA and DI water.
Top electrode
Bottom electrode
Barrier
Photoresist
Substrate
A
B C
Alignment marks
28
Step 2. Junction area definition
The second step is to define the junction geometry and area. This is the most crucial and
sensitive step in all junction fabrication process.
Step 2.1 - The photoresist is patterned to cover a small area on the bottom lead and
also at the large bottom contact areas. In this particularly lithography step, the laser
energy used during the exposure should be a little bit smaller than the one use for
others cases, because the dimensions required are much smaller, in this way the
developer process can be better controlled.
Step 2.2 - The unprotected material is removed by ion milling etch. Since the bottom
electrode is to serve as bottom contact lead to the junction area, the etch must be
stopped after the tunnel barrier has been reached and before the bottom electrode
material is completely removed. When over-etch takes place, redeposition of the
etched metallic material on the sidewall can effectively short the two electrodes. Thus,
the etch should be stopped once the barrier is reached, but exact timing is difficult and
some amount of over-etch will always occur. At Nordiko 3600 machine the amount of
material etched is control by time, Table 3. However, to be sure that the junction is
reached, it should be placed next to the real sample a calibration sample that is no
more than, the top electrode structure deposited on glass. When the glass is reached
(easy to see at naked eye), it means at least the top electrode was removed by
etching on both samples. The etch is interrupted at several points in time to verify if
the material on the calibration sample has already been removed. The etch is stopped
once this happens. The side view of the sample at this point is shown below.
Figure 26 - A – Side view after 2nd etch; B -Top view after 2nd etch; C- microscope picture with filter (green) from sample N2TJ68 with 1x5 µm2 area after 2nd exposure. This filter is needed because if the developing time was not enough, it’s possible after observation to develop more time, until the junction’s area gets right. If unfiltered light is used such isn’t possible because the properties of the photoresist would change.
Area: 1x5 µm2
A C
Bottom
electrode pads
Junction’s area
B
29
To ensured that no material remains on the sidewall shorting the barrier. The etch has to be
performed not in one step, but in fact in two steps at different angles, ".
Figure 27 - Side view of Nordiko 3600 machine with chamber door removed. The so-called etching angle, ", is the angle between etching direction (red line) and the substrate holder direction (light blue line).
Step 2.2.A - Etch the junction stack down until the barrier at an angle
near 90º, defining the junction geometry and area. The results
reported in this thesis were obtained using the 70º angle (60º at
N3600 machine software). This angle was optimized at INESC-MN
for a similar machine, Nordiko 3000 [6]. The angle optimization for
N3600 was not done, because small variations around that value
don’t have critical impact on the final outcome.
Step 2.2.B - Etch the barrier and a small part of the bottom electrode
with an angle next to 45º, in order to clean the redeposited particles
from the sidewalls. At INESC-MN the standard angle used is 40º(less
"
70 cm
Etching
direction
30
10º at N3600 machine software), and the etch rate is a little higher
than at 70º, around 1.15 Å/s. Since the aim of this thesis is to
optimize the junction definition step, several angles were tried to see
if redeposition is reduced. As shall later be shown, sidewall material
redeposition can have serious consequences on the tunnel junction’s
electrical properties. Besides the angles study, sidewalls oxidation
with oxygen plasma is tried as discussed ahead.
Step 3. Insulating layer deposition
In this step an insulating layer Al2O3 covers the junction sidewalls and the bottom electrode.
This avoids the shortcut between top and bottom electrodes and also forces the current to
pass through the barrier from the top of it and not by the sides. As for, the thickness of the
insulating layer has to be higher than the barrier and top electrode together. Normally is used
at INESC-MN a 500Å thick layer deposited by sputtering at UHV2.
Figure 28 - Side view of the junction area. A - after insulating passivation layer deposition; B - after liftoff opening the contact to the junction area and pads. The photoresist from the second etch step is removed after the passivation layer deposition. This liftoff process opens a via in the insulating layer aligned with the top junction electrode. A via is also opened to the larger contact pad areas in an similar way. C- Microscope top view image, from TJ1268 after liftoff (no oxide on lighter regions).
This liftoff process opens a via in the insulating layer aligned with the top junction electrode
and with contact pads areas, by removing the existing photoresist. Due to the small
dimensions of the junction’s areas this process can take 2 or more days for the smaller
junction’s areas, the process is much faster for contact pad areas. During the liftoff process
with microstrip (chemical used to remove photoresist) in a low hot bath, ultrasounds should
be used to speed up the process.
B A
Al2O3
Insulator Top electrode via
C
Alignment mark
for e-beam
31
Figure 29 – UHV2 machine for oxide deposition by sputtering from a SiO2 or Al2O3 ceramic target. This system is used for the oxide layer deposition in the tunnel junction process. The machine is installed in a class 10000 clean room. Details on the machine are provided in the table on right. A - chamber; B – Argon bottle for sputtering [5].
Step 4. Top electrode metallization
After the liftoff of the insulator material, the via to the junction area and the bottom contact are
open. The metallization step provides the lead and contact pads for the top electrode.
Step 4.1- The photoresist is patterned to cover the sample, except the top lead and
the large top contact areas (in those areas AlSiCu and TiWN2 will remain deposited
after the liftoff).
Figure 30 - Top view of the sample N2TJ68 after expose and develop.
Step 4.2 - Deposition of AlSiCu (98.5% of Al, 1% of Si and 0.5% of Cu) and TiWN2 by
sputtering at Nordiko 7000, Figure 31. The deposition is a sequence of steps. The
first one is a 30s sputter etch step to clean the surface where the metal is going to be
deposited (decrease the resistance). Then 3000Å AlSiCu followed by 150Å TiWN2
are deposited on the sample.
Step 4.3 – Liftoff with hot Microstrip and ultrasounds. Rinse with IPA and DI water.
Base pressure 6 x10-8 Torr (turbo pump) Sample dimension Maximum F6 inch
Automatic machine No Sample rotation No Sample mounting Samples facing up
Not clamped Shutter No (no target presputtering) Loadlock No
Sample rotation No Target presputtering no (no shutter ) Heated substrate table No
Targets One target facing down f6 inch
Deposition conditions 200 W (RF), 45 sccm Ar, 6 mTorr Film deposition rates 0.38 Å/s (SiO2)
0.2 Å/s (Al2O3) Gas lines Ar
No Photoresist
Photoresist
A
B
32
Figure 31 - Top view of the Nordiko 7000. There are 4 process modules: flash anneal (module1), soft sputter etching (module2), TiWN2 sputter deposition (module3) and AlSiCu sputter deposition (module4) [6].
The TiWN2 protects the AlSiCu layer from oxidation during the liftoff in the photoresist solvent
(microstrip or acetone). The metal is removed from the sample, except where there is no
photoresist below (top lead and the large top contact areas). This metallization step and the
final view of the fabricated sample are illustrated in Figure 33.
Figure 32 – A - Cross-section view of the contact leads defined by lift-off of the Al metallization. B- Top view of the sample N2TJ68 after Al lift-off, with the top electrode contact pads and lead.
Figure 33 – Final outlook of the samples after patterning. (Right) sample N2TJ68 deposited on a glass substrate, with 3 columns, each one with 9x20 junctions. (Left) sample TJ1268 deposited on a silicon substrate. The sample has several designs, 1 column with 9x19 junctions, matrix of junctions, lines and squares matrixes.
A
Bottom electrode pad
Top electrode pad
B
33
Step5. Annealing
Finished the patterning, the magnetic tunnel junctions are annealed under the influence of an
external magnetic field. There is a large difference between the as-deposited and the low-
temperature annealed material. In case of bottom pinned tunnel junctions no exchange is
verified before annealing. After a low-temperature anneal the TMR increases significantly and
the resistance decreases slightly. As is shown in Figure 34, the annealing temperature has a
major impact in the TMR. The sample N2TJ68_B is a magnetic tunnel junction with an 8Å
thick MgO barrier deposited by sputtering in Nordiko 2000 at INESC-MN.
Figure 34 – TMR as function of the RxA after annealing the sample N2TJ68_B at 280ºC for 1h and after annealing the same sample at 320ºC for 1h.
The annealing process is responsible for the re-crystallization of the as-deposited material
that presents some defects in its crystalline structure and orientation. After the re-
crystallization grain growth can occur. In case of magnetic tunnel junctions, this re-
crystallization happens in different temperatures (so-called annealing temperatures), because
of the different constituents materials of the junction. In the junction with an MgO barrier the
materials in question are the CoFeB and the MgO (see Figure 10). Their re-crystallization
temperatures are around 320ºC and 360ºC, respectively. Besides the re-crystallization, the
annealing when carry out together with an external magnetic field, improves the TMR by
forcing a desired magnetization in the magnetic layers of the junction, optimizing the spin
polarisation at the barrier interfaces. In this case the target layer is the antiferromagnetic
layer responsible for the exchange coupling. The antiferromagnetic layer has a blocking
temperature, at which the exchange field decreases to zero, the junction is cooled down in an
external magnetic field, the pinned layer is oriented by the external field and maintained by
the antiferromagnetic layer after the field has been removed. The blocking temperature is
34
lower than the others temperatures mentioned before; thus an external magnetic field is
always needed during cooling down step otherwise there isn’t a well define preferential
magnetic direction, and the spin polarization would perform poorly. The set annealing
temperature and time are limited, because after a certain temperature and time the TMR
drops very quickly (typically 360ºC), as result of material diffusion.
The N2TJ68_B was annealed first at 280ºC for 1h with a 4ºC/min raise temperature step in
the newest and big annealing setup at INESC-MN, see Figure 36.
Figure 35 – Heating and cooling steps in the big annealing setup. All TJ1268 samples were annealed with these temperature conditions, but only for 30 minutes. The sample N2TJ68_B was first annealed with these conditions and then with the conditions set in Figure 38.
Figure 36 – Scheme and picture of the big annealing setup.
Then the same sample was sliced to fit in the small annealing setup, being annealed at 320ºC
(set temperature) for 1h, with a 4000 Oe field created by a current (14% of the maximum
current) in the coils. The raising time was 45 min.
1h at 280ºC
4ºC/min
Room Temperature
Annealing Temperature
Annealing Time Cool down Heating
Connection to turbomolecular pump, pressure sensor and temperature sensor
Magneto 1 Tesla
Sample entrance
Furnace
Magnetic Field Direction
35
Figure 37 - Heating and cooling steps in the small annealing setup. All N2TJ68 samples were annealed with these conditions.
The samples (maximum ~ 10 mm wide) are mounted on top of a copper block, using Apiezon
grease to improve the thermal contact. The heat source is a halogen lamp (100 W, 12V),
placed inside the copper block. The temperature is measured by a thermocouple and the
controller sets the lamp On/Off cycles in order to reach the set temperature after the set rising
time, and to maintain the annealing temperature within ±5ºC deviation. The temperature
calibration curve used in small annealing is plotted in Figure 38. The rising time depends on
the annealing temperature and is typically set to 45 minutes for Tannealing of 300-350ºC, and
the cooling is done without temperature control. During the heating treatments the sample is
kept inside a removable glass chamber, pumped down to ~10-6 Torr by a turbo pump. An
external magnetic field is applied during heating and cooling down. Two water-cooled
magnets create the field, and the distance between poles and current in the coils can be
selected in order to achieve up to 5000 Oe at the sample region (pole separation of 4 cm and
coil current of 14% of the maximum). For the tunnel junction annealing, the magnetic field
was set to 4000 Oe open gap, with 4 cm separation between coils).
Figure 38 - The temperature calibration curve used in the small annealing.
1h at 320ºC
45min
Room Temperature
Annealing Temperature
Annealing Time Cool down Heating
36
Figure 39 - Pictures of the small annealing setup. The removable glass chamber surrounds the copper basis, and the magnet coil separation is adjusted to set the magnetic field at the region of the sample [5].
Samples Annealing set conditions
TJ1268 30 minutes @ 280 ºC
N2TJ68_B* 60 minutes @ 280 ºC
N2TJ68 including N2TJ68_B* 60 minutes @ 320 ºC
Table 4 – Annealing conditions used on samples TJ1268 and N2TJ68. The sample N2TJ68_B, was twice annealed, first for 1h at 280ºC and then for 1h at 320ºC.
Sample Copper basis
Glass Chamber
Magnet coils
Thermocouple
(calibration)
Thermocouple
(controller)
Lamp
37
5. Electrical Characterization Finished the fabrication, the tunnel junctions are ready to be characterized for their electrical
properties. For the transport measurement, a direct electrical response of the tunnel junction
in a magnetic field is recorded. TMR signal vs. resistance area (RxA) in the patterned element
were obtained from these measurements. In this way redeposition could be studied, as it will
be observed ahead. The curve in Figure 40, the so-called transfer curve, is the typical curve
that is obtained from the electrical measurements.
Figure 40 – Transfer curve from the sample N2TJ68_B, obtain from the electrical measure.
The resistance of the device is monitored as an external magnetic field is varied between two
extreme values. The resistance goes from its lowest value, for parallel aligned pinned and
free layer magnetizations, to the maximum value when the magnetizations are anti-parallel.
The TMR is calculated using the equation 6, where TMR is defined as the percentage relative
resistance change. These measurements allow comparing values between junctions with
same structure but different areas and shapes. From this study it is possible to compare if
there was more or less redeposition, as is discussed further on. The measurements take
place in the setup illustrated in Figure 42.
38
Figure 41 – Setup used to measure the transfer curves of the tunnel junctions [5].
All transfer curves measurements are carried out in a 4-contact configuration, according to
the schemes of Figure 41 and Figure 42. A 2-contact configuration (V and I connections done
at the same pad) could be also used. However, while the 4-contact configuration measures
only the voltage change across the tunnel junction, V=I.Rjunction, in the 2-contact the voltage
change is measured across both the electrodes and the junction, V=I.(Rjunction+Rtop+Rbot).
When the electrode resistance is of the order of the junction resistance the measurements
should be done with 4-contacts.
Figure 42 – The 2-contact and the 4-contact configurations schemes [5].
Field direction
optical
microscope
Coils
(~140 Oe with ~8 cm
separation between coils)
Probes
(magnetically
fixed)
Probes
(vacuum fixed)
protection
switch
lamp
sample
Setup IIb)
V +
V-
I+
I-
V+ I+
V- I-
ZOOM IN
Field direction
39
6. Redeposition study The redeposition of conductive material on the junction’s sidewalls during the etching reduces
the effective resistance of the tunnel junction, creating shortcuts around the barrier. The
diminution of the effective resistance drives to a loss of the TMR signal. At lower resistances,
this loss becomes quite critical.
Figure 43 – Schematic of a tunnel junction in the second etch with material redeposition (orange spots) on sidewalls, shorting parts of the barrier.
The redeposition phenomena can be described in terms of resistances by an electrical model,
where a parallel resistance, RS, shorts a perfect junction, RTJ.
Figure 44 – Electrical circuit proposed to simulate the redeposition effect on the TMR. The model consists in a perfect junction resistance, RTJ, shorted by a parallel resistance, RS, which depends on the amount of redeposition [6].
The equivalent resistance of the electrical circuit, corresponds to the measured resistance,
RM, and is given by:
!
RM
=RSRTJ
RS
+ RTJ
( 9 )
Top electrode
Bottom Electrode
Barrier Redeposited conductive material
40
The equation 9 presents two limiting cases. The first one, the redeposition is small and the RS
tends to infinite or is much larger than the junction resistance. Then RS is negligible compared
to RTJ and the measured resistance will be equal to the real junction resistance. On the other
extreme, if there is a lot of redeposition, the parallel resistance becomes much smaller than
the junction resistance and RS tends to zero. In this last case the measured resistance tends
to zero.
The transfer curve is characterized by a hysteresis loop that has a low and a high resistance
values on the saturation regions. Therefore, the perfect junction resistance, RTJ, can take the
low resistance value, RL, or the high resistance value, RH. Following the same line of thought,
the measured junction resistance, RM, can be equalize to the measured low resistance, RML,
or to the measured high resistance, RMH.
!
RML
=RLRS
RL
+ RS
( 10 )
!
RMH
=RHRS
RH
+ RS
( 11 )
As a result, there is also a measured tunnelling magnetoresistance, TMRM, and a real TMR,
equation 12 and 13 respectively.
!
TMR =RH" R
L
RL
( 12 )
!
TMRM
=RMH
" RML
RML
( 13 )
Replacing the equations 10 and 11 in the equation 13,
!
TMRM
=
RHRS
RH
+ RS
"RLRS
RL
+ RS
RLRS
RL
+ RS
=RH" R
L
RL
RS
RH
+ RS
= TMRRS
RH
+ RS
( 14 )
41
The measured TMR will tend to zero when redeposition is very high, i.e. when RS tends to
zero. Because the RS depends on amount of redeposition, its value is not known, thus it is
useful to express it in terms of RML and RL.
!
RS
=RLRML
RL" R
ML
( 15 )
In order to diminish the redeposition effect two different methods are studied in this thesis.
The first one, consists to vary the etch angle, ". The reason is to prevent the redeposition of
conductive material near the junction. On the other hand, after performing the etching
responsible for the definition of the pillar, oxygen plasma is used to oxidize the redeposit
conductive material, becoming it non-conductive. Finally, the best results from both methods
are used together improving the outcome.
42
6.1. Etching Angles Concerning the angles study, it was undertaken having as support the results previous
obtained for a similar, but smaller ion beam system existing at INESC-MN, the Nordiko3000
machine. At the time a 40º etching angle was considered as being the optimized angle for the
junction pillar etching, where redeposition is minimum. Due to the size and design
differences of both machines a new study was required. The study is focus in the second
etch, the junction pillar definition. As is explained at chapter 4.4, this etching occurs in two
phases, the first one at an angle next to 90º until the barrier giving the junction shape, follow
by a etching at an angle next to 45º for a small period just to “clean” the redeposit material.
The angle that is going to be varied is the last one, since it is the most important from
redeposition point of view. Such angle cannot be much smaller than 45º, because otherwise,
the barrier starts to be considerable etched on sides, changing its geometry and active area.
For this study was used dies from a 6-inch sample TJ1268, each die was label with a number,
as represented in Figure 45.
Figure 45 – Magnetic tunnel junction TJ1268 deposited on 6-inch silicon wafer size at Nordiko3000 machine, sliced by diamond saw in 12 smaller samples (dies with 2x2 cm2). The samples structure is represented in figure 9, and has 5Å thick AlOx barrier. The picture on right is from 6-inch sample GTJ301 previous processed and not from TJ1268, but the dies design is the same.
All dies have the same mask layout, Figure 46. Besides the junctions, there are test
structures, such as small and large rectangles matrixes and lines arrays. These structures are
used to study shape profiles caused by etching and to measure magnetic properties at VSM.
3 2 1
6 5 4
9 8 7
12 11 10
+ 155mm 6-inch
43
Figure 46 - Design mask layout with the test structures to be etched by ion beam etching. A- Lines width from 1 to 100 µm; B – MTJ; C- Matrix of rectangles 3"m x 6"m; D – Matrix of large rectangles; E – Matrix of MTJ.
The dies used are 6, 8, 9 and 11. All samples have a total etching time of 400 s, which
corresponds to etch proximity 400 Å of material, depending on etching angles used, Table 3.
With the exception of the sample TJ1268_11, all samples were first etched for 300s at 70º,
defining in this way their shape, and then in a second step they were etched for 100s at a
certain angle (the variable under study) acting on redeposit material, see Table 5 In the
sample TJ1268_11, the etching was performed at once with an angle of 40º.
Sample TJ1268
#11
#9
#8
#6
Etching Angle :40º Etching Angle :70º Etching Angle :70º Etching Angle :70º 1st
(400 s) (300 s) (300 s) (300 s) Step
Etching Angle : 50º Etching Angle :60º Etching Angle :40º 2nd
(100 s) (100 s) (100 s) Step
Table 5– Etching conditions used during the pillar definition for samples with same structure (TJ1268_6, TJ1268_8, TJ1268_9, TJ1269_11), see figure 9 (right). The total time of etching is 400 s. The etching angle conditions used for sample TJ1268_6, are the standard ones used at INESC-MN.
The effect of redeposition can be studied by plotting the TMR as function of the resistance
area, RxA. The typical fingerprint for redeposition in this kind of plots is an abruptly drop on
the TMR signal for low resistance areas, and a plateau for higher resistances areas with a
larger TMR. As can be remembered from chapter 2.1, the TMR should be constant for the
different resistances, and the resistance should be inversely proportional to the area.
Therefore, being the resistance area and the TMR fairly constants, in case of few or non-
redeposition the plot obtained should be a cluster of points well identified.
A
B
C
D
E
44
Figure 47 – Plot TMR vs RxA obtained from samples with same structure (TJ1268_6, TJ1268_8, TJ1268_9, TJ1269_11), see figure 9 (right), but with different etching angles during the junction pillar definition, see table 5 .
The standard etch conditions used at Nordiko 3600 are shown in Table 6.The Nordiko 3600
running sequence is:
Batch: Junction_etch
Wafer recipe: etch_junction_top_electrode
Process step: Load wafer at 60º / etch pan 60 deg (300 s) / etch pan 30 deg (100s) /
end_junction_etch
Basically a batch can hold several functions called the wafer recipes; each wafer recipe is
built from several steps; each step acts on the machine.
Assist Gun Standard Reading
Assist Neutralizer Standard Reading
RF FWD Power 158 W Current -
RF REF Power 2 W Voltage 330 V
Grid 1 Volts + 724 V Argon Flow 3.0 sccm
Grid 1 Current + 105 mA
Grid 2 Volts - 345 V Sub Pan X deg
Grid 2 Current - 2.3 mA Sub Rotn 30 rpm
Argon Flow 10.2 sccm
Oxygen Flow -
Pressure 10-4 Torr
Table 6 – Standard values read in the reactor chamber of Nordiko 3600 during the etching process. The sub pan value, X, depends on the desired etching angle, its value is always 10º less then the real angle, due to the inclination of the etching gun. The sub rotn is the speed that the support table rotates in order to have a homogeneous etching in all sample.
45
Figure 48 - TEM pictures from patterned parts of the 6-inch wafer magnetic tunnel junction TJ1268. In both samples the etching didn’t reach the AlOx barrier. These samples are not represented in the plots. (Top) The sample was etched with one angle (one slope). (Bottom) The sample was etched with two angles (two slopes). Pictures took at University of Glasgow, Department of Astronomy and Physics.
Single slop side wall
AlOx barrier
Two slop side wall
46
6.2. Oxidation Via Oxygen Plasma
The tests with oxygen plasma oxidation were carried out at first place on samples with AlOx
barrier, the TJ1268 samples. The oxidation is performed after the ion milling etching take
place. For all samples, the etching angles are the same used on the sample TJ1268_6, i.e.
300s at 70º and 100s at 40º. After performing the etching, the table where the sample is
attached rotates to 0º, at this point the angle between the sample and the etching gun is 10º
Figure 49. The aim is to oxidize laterally the junction (avoiding the lateral etching using a
lower current than the one used in the etching steps), and doing so, the redeposit material.
Figure 49 – Schematic of lateral oxidation with oxygen plasma. The angle between the sample and the plasma is 10º.
The differences between samples are in the oxidation times (Y), after etching, as shown in
Table 7.
Samples
TJ1268 Oxidation (Y) time /s
# 6* 0
# 4 20
# 5 50
# 7 300
Table 7 – Oxidation times (Y) after etch the sample for 300s at 70º and 100s at 40º. The sample TJ1268_6 is the standard sample used to compare.
Oxygen plasma with 10º angle
47
The full running sequence, i.e. the etching steps followed by the oxidation step is shown next.
Batch: Junction_etch
Wafer recipe: etch_junction_top_electrode + O2
Process step: Load wafer at 60º / etch pan 60 deg (300s) / etch pan 30 deg (100s) /
oxidation beam (Y s) / end_junction_etch
The Table 8 has the standard readings at Nordiko3600 during the oxidation process, after performing the etching steps. The Figure 50 shows the TMR as function of resistance area.
Assist Gun Standard Reading
RF FWD Power 135 W
RF REF Power 1 W Sub Rotn 30 rpm
Grid 1 Volts + 101 V
Grid 1 Current + 45.2 mA Sub Pan 0 deg
Grid 2 Volts - 345 V
Grid 2 Current - 1.2 mA Pressure 10-4 Torr
Argon Flow 2.2 sccm
Oxygen Flow 20.1 sccm
Table 8 - Standard values read in the reactor chamber of Nordiko 3600 during oxidation beam step. The reading values for the etching steps are given in Table 6. The current used in oxidation step is much lower than the one used to etch; the reason is to prevent the lateral etching during the oxidation that would decrease the junction’s area.
Figure 50 - Plot TMR vs RxA obtained from samples with same structure (TJ1268_6, TJ1268_4, TJ1268_5, TJ1269_7), see Figure 10 (right), but with different times of oxygen plasma oxidation, see Table 7. For 300 s of oxidation a cluster is formed, however the resistance increases. All samples have the same junction’s areas. The areas range is from 1 "m2 to 25 "m2.
Cluster: Sample TJ1268_7
48
Since the points with higher resistances corresponds to junctions with areas of 1 and 2 "m2,
they can be ignored in order to visualize better the other points. The reason is during the
developing after lithography, smaller features near 1 "m2 size are difficult to obtain with the
right size, being its size in general, smaller than expected, and therefore its RxA should be
also smaller. Besides, the small features face several challenges in the liftoff process
acquiring a non-precise size.
Figure 51 – Is the same plot that is in Figure 51 but ignores the 3 points with larger resistance. The reasons of this choice are explained above.
The TMR increases with the oxidation time, and for 300s of oxidation a cluster is formed, as it
is desirable. However, the resistance also increases considerable. If the oxidation times are
close there is almost no difference on the distribution, as shown in Figure 52.
Figure 52 – Comparison between samples with similar oxidation times. The samples are TJ1268_4 with 20s of oxidation and TJ1268_5 with 50s. Little difference can be noticed.
Cluster
49
From these first results it seems clear that the angles tested don’t show any improvement
comparing with the previous angle optimized for the Nordiko3000 machine, and made
standard condition at INESC-MN. On the other hand, the results obtained with oxygen plasma
oxidation seem promissory. For that reason, the studies carried out after these preliminary
tests are focus mainly in testing oxidation conditions. After testing different conditions on
samples with an AlOx barrier, worldwide used since the early times of magnetic tunnel
junctions, new samples with an MgO barrier are tested, the N2TJ68 samples, see Figure 10
(left). The samples are expected to have considerable higher TMR, due to its MgO barrier
nature. The running sequence use for N2TJ68 samples is the following one:
Batch: Junction_etch
Wafer recipe: etch_junction_top_electrode + O2
Process step: Load Wafer/etch pan 60 deg (270s)/etch pan 30 deg (60s) / oxidation beam
(Ys) / end function
The total etching time is 330s, i.e. about 330Å of material is etched. The differences in the
samples N2TJ68 are in the O2 plasma oxidation times (Y), after etching, see table 9.
Sample Oxidation Beam Time Y (s)
A No O2
B 100
C 20
Table 9 – O2 plasma oxidation times used in the samples N2TJ68 A, B and C, after etching 270s at 70º and 60s at 40º.
In Figure 53 it is possible to identify a cluster and an improvement on the TMR values more
than 30%. The sample in question is N2TJ68_B with 100s of oxidation, the etching angle
condition are standard, equal for all N2TJ68 samples. There is a resistance increase with a
factor of 2.
50
Figure 53 - Plot TMR vs RxA from samples with MgO barrier (N2TJ68_A, N2TJ68_B, N2TJ68_C), with different times of oxygen plasma oxidation. For 100 s of oxidation a cluster is formed with little increase of the resistance and with considerable improvement on the TMR signal. The etching conditions are standard and equal for all N2TJ68 samples. All samples have the same junction’s areas. The areas range is from 3 "m2 to 15"m2. In these experiments the areas of 1 "m2 and 2 "m2 aren’t used in order to avoid some doubts about the real size of those areas. The causes are related with the developing time, where a broad range of areas makes hard the decision when to stop the developing process. If it’s stopped to soon the larger areas could not be well defined, if it is decided to proceed with more time of developing the smaller areas features could vanish (it isn’t unusual).
Figure 54 – Plot TMR vs RxA from sample N2TJ68_B. Each colour is related with the junction area. As observed the redeposition is not directly linked with the area junction, since for all junctions’ sizes the TMR values are widely spread, as well the resistances.
Cluster
51
From Figure 54, where each dot colour is related with its junction’s area, it can be deduced
that in this case, redeposition is not directly linked with the area junction, since for all
junction’s areas the TMR values are widely spread as well the resistances. For that reason
the slope observed doesn’t seem to depend on junction’s areas.
Figure 55 – (Left) TMR sensor map with a colour scale indicating the TMR value of each MTJ sensor, that goes from a lower intensity represented by red colour to higher intensity, the blue colour. Each square represents a magnetic tunnel junction sensor. In total there are 20x9 sensors. The position #4 in the lines is never measured because the needles are broken for that position in the probe card of the automatic measure setup (KLA). (Right) Magnetoresistance histogram and statistics, the numbers on top of each column is the number of junctions with the same TMR.
From TMR sensor map, most of sensors have similar TMR (colour like). The junction’s areas
from the upper half part of the column are repeated on the lower part. Taking only the lower
part, the homogeneity of TMR improves, indicating a localization issue and not a junction’s
area size effect. The TMR problem can be from deposition that affects the film homogeneity,
because the four N2TJ68 samples were deposited at same time in N2000 machine, and for
this machine there is no guarantees about its deposition homogeneity for samples bigger than
1-inch.
TMR sensor
Map map
T
MR
TMR Sensor Map TMR (%)
MTJ Sensors
52
7. Conclusions The aim of this thesis was to optimize the etching parameters using the ion milling system
Nordiko3600 at INESC-MN. The optimization main focus was to decrease the redeposition
during the ion milling etching. Two approaches were considered, the etching angle variation
and the lateral oxidation via oxygen plasma. From the first results with AlOx junctions, the
etching angle variation study was put aside, because when compared with the angles
previously optimized for the ion beam system Nordiko 3000, no significant advantages were
found. On the other hand, the results from lateral oxidation via oxygen plasma showed
considerable improvements in the TMR signal, and for higher oxidation times (300 seconds),
a cluster in TMR vs RxA plot was formed, see Figure 51, indicating low redeposition.
However the resistance increases by a factor of 2-3, which could represent a drawback for
certain applications. Similar times of oxidation have shown little difference between them in
the final outcome, as seen in Figure 52.
The oxidation via oxygen plasma also reveals to be quite efficient to prevent redeposition in
MgO junctions, confirmed by the cluster formed in Figure 53. In sample TJ1268_B with 100
seconds of oxidation, the TMR improved in average 30-40% regarding equal samples with no
oxidation. The resistance also increases by a factor of 2. Nevertheless, in case of the MgO
junctions this increase translates in 150-200 #, while for the AlOx junctions it is around 700-
800 #, due to the higher resistance of the last ones. For lower oxidation times, around 20-50
seconds the resistance increase is negligible. At INESC-MN the 50 seconds oxidation via
oxygen plasma has become part of the standard procedure of the magnetic tunnel junction’s
fabrication, see [4]. In future a fine-tuning has to be done, concerning the oxidation times, the
oxygen flow and the ion current. All oxidations were performed with 20 sccm oxygen flow and
45 mA current, only the oxidation times were varied. The current used in oxidation step has to
be considerable lower than the one used to etch; the reason is to prevent the lateral etching
during the oxidation that would affect the junction’s area. This is of major importance for sub-
micron sized junctions, and could not be tested within this thesis.
53
8. Appendix
8.1. Run sheets During this thesis were performed six runs. Unfortunately, two of them couldn’t be carried out
until the end. In run #1, after performing the first etch the 6-inch wafer showed signs of
corrosion, and the film seemed to start pilling off, that’s why that run was terminated. This was
an important drawback; since junctions take time to be deposited, plus the machines not
always are available to deposit junctions since other works are been carried out at same time.
In the run #5 the junction’s magnetic properties were not right. There was problem with one of
the targets during the deposition. The results obtained are mostly from the run #2 and #4. The
6-inch junction TJ1268 was processed in the run #2 and the samples N2TJ68 were
processed in the run #4.
54
Run Sheet: # 1 Responsible: André Augusto
Structure Sample: GTJ301 (on 6-inch Si wafer)
Junction Structure Ta (30Å) / Ru (150Å) / CoFeFe (150Å) / MnIr (150Å) / CoFeFe^7 (40Å) / (Mg (5Å) w/o Oxidation) x 1 + (Mg+ox (5Å)) x 5 / CoFeFe^7 (50Å) / Ta (100Å)
Barrier Oxidation Conditions
5x8.0 sccm(O2); 2.2x10^-4 Torr (Cryo Hv Open); 500''/step ; 200'' Mg preclean/step with wafer in the loadlock
Deposition Conditions Mg deposited using +7mA; +746V; CoFe1 :+25mA; +1280V; CoFe2,3 :+15mA; +998V Remaining layers: +22mA; +1200V
Deposition Conditions (TiWN2) Deposition Time: 30s
N7000 Power (W) Gas Flow (sccm) Pressure (mTorr)
Setpoint 70 40
50 3
Read 69 39
50.2 3.1
Deposition Conditions (Al2O3) Deposition Time: 32 min
UHV2 Power (W) Gas Flow (sccm) Pressure (Torr)
Setpoint 200 45 -
Read 200-199 44.9 4.0 x 10-3
(A) (B) Responsible: Susana Freitas (A) Machine: Nordiko 3000 (B) Machine: Nordiko 7000
(C) Responsible: Fernando Silva Machine: UHV2
Coating PR: 1.5 µm PR (Recipe 6/2) Developer : Recipe 6/2 Machine: DWL Mask Name: etchL1 X=-56000 , Y=0.0
- Cross Center: X=168 , Y=54 - Die: 20000 x 20000 "m - Die Frame: 25400 x 25400 "m - Vapor Prime: 5’ recipe-0 - Photo Resist: 1.5 "m - WAFER6.map/dwl/wa/fa
STEP 1.1. 1st Exposure – Bottom electrode definition Date: 24 /04 / 07
STEP 0 Junction Deposition (A) + Passivation Layer (B) + Oxide layer (C)
55
Optical Inspection:
Machine: N3600 Batch: Junction_etch Wafer recipe: etch_function_stack / (etch_pan 60 deg (300s) and cool-down 200s)x4 Etching Conditions:
Assist Gun: 170 W +725 V/-345 V 10 sccm Ar Assist Neut: 364 V 3 sccm Ar Sub Rotn: 30 rpm Sub Pan: 60 deg
Hot Micro-Strip + Ultrasonic Rinse with IPA + DI water Acetona bath Optical Inspection:
Comments: The sample seems to show signs of corrosion and the film is pilling off. The process has to be stopped.
STEP 1.2. 1st Ion Milling – Total Structure Etch Date: 25/04 /2007
STEP 1.3 Resist Strip Date: 26 / 04/ 2007
56
Run Sheet: # 2 Responsible: André Augusto
Responsible: Susana Freitas
Machine: Nordiko 3000
Coating PR: 1.5 µm PR (Recipe 6/2) Developer : Recipe 6/2 Machine: DWL Mask Name: aaEtchL1
X=-45000 , Y=0.0
- Cross Center: X=168 , Y=54 - Die: 20000 x 20000 "m - Die Frame: 30000 x 30000 "m - Vapor Prime 5 min. (Recipe - 0) - Energy: 35 % - Focus: 35 - Photo Resist: 1.5 "m - WAFER6PA.map/dwl/wa/fa - Development time : 60 sec
Structure Junction TJ1268 (6-inch size)
200’’ Etch 30º//Ta (30Å) / Ru (150Å)/ CoFeFea (150Å)/ MnIr (150Å)/ CoFeFeb (40Å) / Ru (8Å) / CoFeB20 (50Å) / (Al (5Å) + ox rem. plasma) /CoFeB20 (50Å) / Ru (100Å) / TiWN (150Å)
Substrate = Si 6-inch/Al2O3 1000 Å /Al 600 Å heated a deposited with 25 mA b deposited with 15 mA CoFeFe = Co56Fe4
Oxidation Conditions Spot_a/Spot_b/Spot_c/Spot_1 (5'')
110W; +0V/-0V; 5x8.0 sccm(O2); 2.2x10^-4 Torr (Cryo Hv Open)
STEP 1.1. 1st Exposure – Bottom electrode definition
STEP 0.A Junction Deposition Date: 3 / 12 / 2006
STEP 0.B CAD design and mask conversion
Micron-size (lithography to be done by DWL) - Minimum feature size: 1 µm - Line/rectangle width from 1 to 100 µm To measure the magnetic properties by VSM: Matrix of small rectangles: 3µm x 6µm Conversion step L1 – Bottom Electrode - Inverted L2 – Junction Pillar - Inverted L3 – Top Electrode - Not Inverted
3 2 1
6 5 4
9 8 7
12 11 10
57
Comments: All structures are OK
Machine: N3000 Batch: Junction_etch Wafer Recipe: etch-gun-stab / junction_etch / end etch
Read Assist Gun 60W / +500 V / -200 V / 8 sccm Ar Assist Neutralizer Not need Sub Rotn 40% (30 rpm) Sub Pan 70 degs
Time: 1600s
Hot Micro-Strip + Ultrasonic Rinse with IPA + DI water Total Time in Hot Micro-Strip : some hours Ultrassonic Time : few Optical Inspection:
Comments: All dies are OK. Inside the dies the majority of the features are also OK
STEP 1.2. 1st Ion Milling – Total Structure Etch
STEP 1.3 Resist Strip Date: 26 / 04/ 2007
58
Aim Study the impact of oxidation on redeposition
Coating PR: 1.5 µm PR (Recipe 6/2) Developer : Recipe 6/2
Machine: DWL Mask Name: aaEtchL2
- Cross Center: X=168 , Y=54 - Vapor Prime 5 min. (Recipe - 0) - Energy: 30 % - Focus: 35 - Photo Resist: 1.5 "m - Development time: 55” - WAFER6PA.map/dwl/wa/fa
Optical Inspection:
The 1 µm size features seem smaller and in some samples they simple don’t exist. Maybe do the exposure energy or development time.
Machine: N3600 Batch: Junction_etch Wafer recipe: etch_junction_top_electrode + O2* Process Step: Load wafer at 60º / etch pan 60 deg(300 s) / etch pan 30 deg (100s) / oxidation beam (x s)* / end_junction_etch
Samples Oxidation Beam time (x /s) 6* 0
4 20
5 50
7 300
* The wafer recipe and process step don’t contain the red part. The sample #6 is etched with the standard conditions.
STEP 2.1 2nd Exposure – Junction area definition: Samples 4,5,6,7
STEP 2.2 2nd Ion Milling – Junction area definition Date: 16 / 01 /07
X,Y e.a.
59
P=1,8E-4 Torr T=21ºC Sub Rotn: 30 rpm
Sample Step Read
60 deg
Assist Gun: 135W(250W) /1W/ 724V/ 35,9mA (48,7mA)/ 345V/ 0mA/10,2sccm Assist Neut: 0mA/ 398,4V/ 3,0 sccm
6 30 deg
Assist Gun: 142W/2W/ 724,3V/ 104,8mA / 345V/ 2mA/10,1sccm Assist Neut: 183mA/ 345,6V/ 3,0 sccm
60 deg Assist Gun: 199W/2W/ 725V/ 104,5mA/ 344,8V/ 0mA/10,2sccm Assist Neut: 178mA/ 344V/ 3,0 sccm
30 deg Assist Gun: 134W/2W/ 724,3V/ 104,6mA/ 345V/ 2,3mA/10,2sccm Assist Neut: 47mA/ 318,3V/ 3,0 sccm
4
O2 beam Assist Gun: 145W/1W/101V/ 45mA / 350V/ 2sccm Ar/ 20 sccm O2 Sub Pan: 0 deg Shutter Open: 20”
60 deg Assist Gun: 187W/2W/ 724,3V/ 104,3mA/ 344,8V/ 0mA/10,2sccm Assist Neut: 176mA/ 344V/ 3,0 sccm
30 deg Assist Gun: 125W/1W/ 725V/ 104,2mA/ 344,8V/ 2,2mA/10,2sccm Assist Neut: 116,5mA/ 319,5V/ 3,0 sccm
5
O2 beam Assist Gun: 147W/1W/101V/ 45mA / 345V/ 1,1mA/2,1sccm Ar/ 20 sccm O2 Sub Pan: 0 deg Shutter Open: 50”
60 deg Assist Gun: 239W/2W/ 724V/ 107,3mA/ 344,8V/ 0,2mA/10,2sccm Assist Neut: 177mA/ 342V/ 3,0 sccm
30 deg Assist Gun: 162W/1W/ 724,3V/ 105mA/ 344,8V/ 2,7mA/10,2sccm Assist Neut: 117mA/ 318,3V/ 3,0 sccm
7
O2 beam Assist Gun: 145W/1W/101V/ 45mA / 345V/1.3mA/ 2,2sccm Ar/ 20,1 sccm O2 Sub Pan: 0 deg Shutter Open: 300”
Machine: UHV2 Power /W Gas Flow /sccm P /Torr Setpoint 200 45
Read 200 45 3,3E-3
Time 27 min
Thickness verification Ellipsometer: n=1,64 t = 491 A Perfilometer: t = 500 A Hot u-strip + ultrasonic Rinse with IPA + DI water Time: 24h Optical inspection:
STEP 3.1. Insulating Layer Deposition- 500Å of Al2O3
STEP 3,2. Oxide Lift-Off
6 6
4 5 4
7
60
The features bigger than 2x1 or 3x1 seem OK. However, the ones that are smaller than the previous dimensions are close or are not seen on microscope (see visual inspection paper on run sheet). After etching and liftoff the samples present several kind of bubbles on the pads (except sample 6).
Coating PR: coat 1.5 µm PR (Recipe 6/2) Developer : Recipe 6/2
Machine: DWL Mask Name: aaEtchL3
- Cross Center: X=168 , Y=54 - Vapor Prime 5 min. (Recipe - 0) - Energy: 35 % - Focus: 35 - Photo Resist: 1.5 "m - Recipe: 6/2 - WAFER6PA.map/dwl/wa/fa
Optical inspection:
Comments: The majority is OK. Few are bad.
Machine: Nordiko 7000 Sequence 48 Module 2: F9, 60” F69, 30W, ROB99, 50.2sccm, 3mT
Module 4: F1, 120”, 3000 Å, AlSiCu 2.0kW, 410V, 4.9A, 50.4sccm, 3.0mT Module 3: F19, 27”, 150Â, TiWN2 0.5 kW, 431V, 1.2A, 50.65sccm, 3.1mT, 10scc
Hot µ-strip + ultrasonic Rinse with IPA + DI water
STEP 4.1 3rt Exposure – Top electrode metallization
STEP 4.2 Contact Leads Deposition (AlSiCu) Date: 25 / 01 /07
STEP 4.3 AlSiCu Liftoff
X,Y e.a.
Shift: Sample 6
Particle of dust
61
Machine: big annealing setup Annealing conditions: 30 min@280ºC
Aim Study the impact of the etching angle variation
Coating PR: 1.5 µm PR (Recipe 6/2) Developer : Recipe 6/2
Machine: DWL Mask Name: aaEtchL2
- Cross Center: X=168 , Y=54 - Vapor Prime 5 min. (Recipe - 0) - Energy: 30 % - Focus: 35 - Photo Resist: 1.5 "m - Development time: 55” - WAFER6PA.map/dwl/wa/fa
Observations: #8 all junctions are there, the 1x1 seem little bit small #9 all junctions are there, the 1x1 seem better than #8 #11 same as #9 #3 all junctions are there, the 1x1 seem a little bit big
Machine: N3600 Batch: Junction_etch Wafer recipe: etch_junction_top_electrode P=1,8E-4 Torr T=21ºC Sub Rotn: 30 rpm
Sample #11 #9 #8 #3
Sub pan 30º (400 s)
Sub pan 60º (300 s) Sub pan 40º (100 s)
Sub pan 60º (300 s) Sub pan 50º (100 s)
Sub pan 60º (300 s) Sub pan 40º (100 s) Oxidation beam 20sccm (50 s)
STEP 5. Annealing Date 28 /02 / 07
45’ Tset = 280ºC
30’
STEP 2.1 2nd Exposure – Junction area definition: Samples 3,8,9,11
STEP 2.2 2nd Ion Milling – Junction area definition Date: 27 / 01 /07
X,Y e.a.
62
Responsible: Filipe Cardoso
Machine: UHV2 Power /W Gas Flow /sccm P /Torr Setpoint 200 45
Read 200 45 3,3E-3
Time 27 min
Hot u-strip + ultrasonic Rinse with IPA + DI water Time: 24h Optical inspection:
Comments: The sample #3 is the one less define in shape and size Details: #3 Areas bigger than expect: the junctions are 1µm side bigger Round corners (worst one) Not very well define Some of the big areas didn’t open
#8 Below 1µm side the junctions are about 20% smaller, some are not there Above 2µm side the junctions are well define #9 Junctions sizes are very close to the expected (betters)
(1x1)1.1x1.2 or 1x5)1.1x5.2) The smaller and thinners junctions, seem little bit unfocus
#11 Areas bigger than expect: the junctions are 1µm side bigger
(1x1)1.5x1.5 or 1x2)1.5x2.5 and then 1x5)2x6…) Junctions are well define, corners ± round (better than #3)
Coating PR: coat 1.5 µm PR (Recipe 6/2) Developer : Recipe 6/2
Machine: DWL Mask Name: aaEtchL3
- Cross Center: X=168 , Y=54 - Vapor Prime 5 min. (Recipe - 0) - Energy: 35 % - Focus: 35 - Photo Resist: 1.5 "m - Recipe: 6/2 - WAFER6PA.map/dwl/wa/fa
STEP 3.1. Insulating Layer Deposition- 500Å of Al2O3
STEP 3,2. Oxide Lift-Off Date: 06/02/07
STEP 4.1 3rt Exposure – Top electrode metallization
X,Y e.a.
#11 #3 #3 #3
63
Optical inspection:
Comments: They seem OK. The left squares matrix only appears after the liftoff.
Machine: Nordiko 7000 Sequence 48 Module 2: F9, 60” F69, 30W, ROB99, 50.2sccm, 3mT
Module 4: F1, 120”, 3000 Â 2.0kW, 410V, 4.9A, 50.4sccm, 3.0mT Module 3: F19, 27”/ 54”*, 150Â/ 300Â* 0.5 kW, 431V, 1.2A, 50.65sccm, 3.1mT, 10sccm
* First time was tried to deposit 300 Å of TiWN2, however there was a pressure problem, and system had to be reboot, so on second time was deposited 150Å of TiWN2, instead 300Â.
Hot µ-strip + ultrasonic Rinse with IPA + DI water
STEP 4.2 Contact Leads Deposition (AlSiCu) Date: 07 / 02 /07
STEP 4.3 AlSiCu Liftoff
Sample #3
Sample #8
64
Machine: big annealing setup Annealing conditions: 30 min@280ºC
Final outlook
STEP 5. Annealing Date 28 /02 / 07
45’ Tset = 280ºC
30’
Sample #9
Sample #11
65
Run Sheet: # 4 Responsible: André Augusto
Structure Substrate: Glass
Junction Bottom Electrode Barrier Top Electrode
N2 TJ68 4 samples (A,B,C,D)
Ta50/Ru180/Ta30/MnPt200/CoFe20/Ru9/ CoFeB30
MgO 8 A,
CoFeB 30/Ru50/Ta50/
TiWN2 150A
Responsible: Piotr/Zhao (A) Machine: Nordiko 2000 (B) Machine: Nordiko 7000
Seq.17 –Mod 2 funct. 9 (contetch) (60”) P=70W/40W, p=3mTorr, 50 sccm Ar Mod.3 funct.19 (svpassiv) (150A TiWN2, 27’’) 0.5 kW, 3mTorr, 50 sccm Ar + 10sccm N2
Coating PR: 1.5 µm PR (Recipe 6/2) Developer : Recipe 6/2
Machine: DWL Mask Name: amsmtj1
- Cross Center: X=168 , Y=54 - Vapor Prime 5 min. (Recipe - 0) - Energy: 47.5 % - Focus: 25 - Photo Resist: 1.5 "m - Development time: 60” - map:AMSION
Optical Inspection: Comments:
STEP 1.1. 1st Exposure – Bottom electrode definition Date: 24 /04 / 07
STEP 0 Junction Deposition (A) and passivation Layer( B)
The structures from A, B, C and D samples are well defined. They are quite clean, but there are some particles that can have influence on photoresist uniformity.
X,Y e.a.
66
Machine: N3600
Standard Etching Recipe (Junction Etch) : Etch pan 60°
Total thickness to etch: 867 A
Samples Conditions
A,B, C 3x(300” @ 60deg + 120 “ cool down)
D 4x(270” @ 60deg + 120” cool down)
Hot Micro-Strip + Ultrasonic Rinse with IPA + DI water Total Time in Hot Micro-Strip: some hours Ultrasounds Time : few Optical Inspection: Comments:
Coating PR: 1.5 µm PR (Recipe 6/2) Developer : Recipe 6/2
Machine: DWL Mask Name: aaAMSTJL2
- Cross Center: X=168 , Y=54 - Vapor Prime 5 min. (Recipe - 0) - Energy: 45 % - Focus: 25 - Photo Resist: 1.5 "m - map:AMSION
Sample Development time A, B, C 35”
D 30”+5”
STEP 1.2. 1st Ion Milling – Total Structure Etch Date: 25/04 /2007
STEP 1.3 Resist Strip Date: 26 / 04/ 2007
STEP 2.1 2nd Exposure – Junction area definition Date: 07 /05 /2007
All samples look OK.
X,Y e.a.
67
Optical Inspection:
Machine: N3600
Read Pressure (Torr) : 10-4
Wafer Recipe : #10: etch junction top electrode + O2
Process Step : Load Wafer/etch pan 60 deg (270”)/etch pan 30 deg (60”)/oxidation beam (Xs)/ end function
Total thickness with overetch (rate 1A/s): 330 A
Calibration Sample Structure
Piotr TiWN2150/CoFeB 30/Ru50/Ta50/MgO/CoFeB30
Sample Oxidation Beam X (s)
A No O2
B 100
C 20
D 300
Assist Gun / Neutralizer
Power (W)
V+ (V)
I+ (mA)
V- (V)
I- (mA)
Gas Flow (sccm)
I neutr (mA)
V neutr (V)
Sub Rotn (rpm)
Sub Pan (deg)
Read Values Etch steps 158 724 105 345 2.3
10.2 He 0.1 O2 116,5 330V 30 60/30
Read Values O2 step 135 101 45.2 345 1.2
2.2 He 20.1 O2 - - 30 0
Comments: The read values above written are similar for all samples.
-There was a problem with sample D, the arm stuck and the sample didn’t come out. Maybe the etching wasn’t done (the wafer didn’t enter in the main chamber)!!!!!!!!!!!!!!
STEP 2.2 2nd Ion Milling – Junction area definition
Majority all samples 1x5 B 1x5 B 1x3 A
Comments: All samples can have devices just a little bit bigger that the ones that were designed: more critical for 1"m side devices. Sample B The developing was not so good like others, but there is no critical issue. 1x3 um2 size devices can be 1x2 "m2.
68
Machine: UHV2
Deposition Time
AlO2 thickness Gas Flow Base
Pressure Power Source
27 min 500 A 45 sccm 3 mT 200 W
Hot u-strip + ultrasonic Rinse with IPA + DI water Optical inspection:
Coating PR: 1.5 µm PR (Recipe 6/2) Developer : Recipe 6/2
Machine: DWL Mask Name: amsmtj3
- Cross Center: X=168 , Y=54 - Vapor Prime 5 min. (Recipe - 0) - Energy: 47.5 % - Focus: 25 - Photo Resist: 1.5 "m - Development time: 60” - map:AMSION
Optical inspection:
STEP 3.1. Insulating Layer Deposition- 500Å of Al2O3 Date: 11 /05 / 2007
STEP 3.2. Oxide Lift-Off Date: 11 / 05 / 2007
STEP 4.1 3rt Exposure – Top electrode metallization
Comments: The samples A, B and C seem all ok. But the sample D faced several problems in the liftoff.
Sample A Sample D
Comments: All samples are OK.
X,Y e.a.
69
Machine: Nordiko 7000 Sequence 48 Module 2: F9, 60” F69, 30W, ROB99, 50.2sccm, 3mT
Module 4: F1, 120”, 3000 Å, AlSiCu 2.0kW, 410V, 4.9A, 50.4sccm, 3.0mT Module 3: F19, 27”, 150Â, TiWN2 0.5 kW, 431V, 1.2A, 50.65sccm, 3.1mT, 10scc
Hot µ-strip + ultrasonic Rinse with IPA + DI water Optical inspection:
Final outlook:
First the samples were annealed in the bigger setup at 280ºC for 1h and the TMR signal was lower than expected. Then the samples were sliced and annealed in smaller setup at 320ºC for 1h (field 4kOe), and the TMR improves a lot.
STEP 4.2. Contact Leads Deposition (AlSiCu)
STEP 4.3 AlSiCu Lift-Off
STEP 4 Annealing Date 07 /06 / 07
Comments: All samples OK.
1h at 320ºC
45min
Room Temperature
Annealing Temperature
Annealing Time Cool down Heating
70
8.2. Paper
71
72
73
74
75
76
77
78
79
80
81
9. Bibliography [1] Julliere, M.; Tunneling between ferromagnetic films. Phys. Lett. 54A: 225-226; (1975).
[2] Moodera, J. S. et al; Large magnetoresistance at room temperature in ferromagnetic thin film tunnel
junctions. Phys. Rev. Lett. 74: 3273–3276; (1995).
[3] Swagten H.J.M; Magnetism & Magnetic Course Syllabus 2005-2006; Faculteit Technische Natuurkunde; Technische Universiteit Eindhoven, Eindhoven, Holland.
[4] Cardoso S., Macedo R.J., Ferreira R., Augusto A., Wisniowski P.and Freitas P.P.; Ion Beam Assisted deposition of MgO barriers for magnetic tunnel junctions; submitted to the Journal of Applied Physics; INESC-MN; September 12th 2007.
[5] Freitas, Susana; PhD Thesis: Dual-Stripe GMR and Tunnel Junction Read Heads and Ion Beam Deposition and Oxidation of Tunnel Junctions; Universidade Técnica de Lisboa; Lisbon; 21st December.
[6] Sousa, Ricardo; PhD Thesis: Magnetic Random Access Memory (MRAM) based on Spin Dependent Tunnel Junctions Universidade Técnica de Lisboa; Lisbon; 2002.
[7] Gallagher, W. J; Parkin, S. S. P; Development of the magnetic tunnel junction MRAM at IBM: From first junctions to a 16-Mb MRAM demonstrator chip; IBM J. RES. & DEV; 50; 1; 2006.
[8] Hayakawa J., Ikeda S., Lee Y. M., Yamanouchi M., Chiba D., Dietl T., Ohno Y. and Matsukura F. (Laboratory for Nanoelectronics and Spintronics, Research Institute of Electrical Communication, Tohoku University; Japan Science and Technology Agency); The World of Spintronics; International Symposium on Advanced Magnetic Materials and Applications; Jeju, Korea, 2007.
[9] Hayakawa J., Ikeda S., Lee Y. M., Y. Ohno and Matsukura F.; Effect of electrode composition on the tunnel magnetoresistance of pseudo-spin-valve magnetic tunnel junction with a MgO tunnel barrier; Applied Physics Letters 90, (2007).
[10] Slaughter J.M; Chen E.Y; Whig R; Engel B.N; Janesky J; and Tehrani S; Magnetic Tunnel Junction Materials for Electronic Applications; JOM-e; 56(2); 2000.
[11] Freitas, Susana; Técnicas de Micro e Nanofabricação course handouts; Physics Department; Instituto Superior Técnico; Lisbon; 2007.
[12] Erwin, K; Plasma Processing course handouts; Dept. of Applied Physics; Eindhoven Univ. of Technology, 2005.
[13] Freitas, Paulo; Nanotecnologias e Nanoelectrónica course handouts; Physics Department; Instituto Superior Técnico; Lisbon; 2006-2007.
[14] Williams K; Hayes A; DiStefano S; Huang O; Ostan E; Reactive ion beam etching of ferroelectric materials using an RF inductively coupled ion beam source; Veeco Instruments Inc.; IEEE; 1996.
82
[15] http://hyperphysics.phy-astr.gsu.edu/hbase/hframe.html; Department of Physics and Astronomy; Georgia State University; 2005.
[16] Waser; R; Nanoelectronics and information technology – Advanced electronic materials and novel devices; Wiley-VCH; Germany; 2003.
[17] Davis M., Proudfoot G. and Pearson D.; Ion Beam vacuum sputtering apparatus and method; European Patent EP1212777; 2001.