10/03/05 1 FLCC Schottky-Barrier Engineering for Low-Resistance Contacts Pankaj Kalra, Hideki Takeuchi, Tsu-Jae King Department of Electrical Engineering and Computer Sciences University of California, Berkeley, CA Oct 03, 2005
10/03/05
1
FLCC
Schottky-Barrier Engineering for Low-Resistance Contacts
Pankaj Kalra, Hideki Takeuchi, Tsu-Jae KingDepartment of Electrical Engineering and Computer Sciences
University of California, Berkeley, CA
Oct 03, 2005
10/03/05
2
FLCC
Outline
• Introduction
• Characterization Schemes
• Si1-xGex Source/Drain
• Dopant Segregation
• Strain
• Summary
10/03/05
3
FLCC
Parasitic Resistance Components• Rsl Silicide sheet resistance
• Rcsd Contact Resistance
• Rsd Silicon Sheet Resistance
• Rov Overlap resistance
Gate
RslRchRov
RcsdRsd
• Parasitic resistance must be <10% of total FET resistance
• CMOS scaling– reduces channel resistance ∝ 1/L– increases contact resistance Source: Prof. Jason Woo, UCLA
10/03/05
4
FLCC
Contact Resistance Scalingcontact area
Al-Si
Heavily doped Si
reduction of contact area
SALICIDEcontact area
WPlugMSix
contact area
WPlugMSix
– Change in contact scheme (adoption of SALICIDE) has extended the contact scaling
– Due to the reduction of active area, silicide/Si contact resistance is now an issue
10/03/05
5
FLCC
Int’l Technology Roadmap for Semiconductors (2004 update)
Target
2006 2008 2010 2012 2014 2016 2018
10-8
10-7
Ozturk et al., IEDM 2002
ρ c (oh
m-c
m2 )
– New materials and processes are needed
10/03/05
6
FLCC
Impact of Rc on FinFETH. Kam and T.-J. King, 2004 Silicon Nanoelectronics Workshop
Top Contact
DRAIN
GATE
SOURCE
DRAIN
GATE
SOURCE
End Contact
DRAIN
GATE
SOURCE
DRAIN
GATE
SOURCE
Wrapped Contact
DRAIN
GATE
SOURCE
0
400
800
1200
0 0.1 0.2 0.3 0.4 0.5 0.6
Wrapped Contact, ρ c = 0 Wrapped Contact
End Contact
Top Contactρc=10-8 Ω-cm2
Lgate = 18 nm, Leff = 22 nm
34% reduction
16% reduction
• Parasitic resistances dominate FinFET performance• ρC<10-8 Ω-cm2 required
Vds
Ids
10/03/05
7
FLCC
Specific Contact ResistivityBarrier Height and Active Dopant Concentration
⎟⎠
⎞⎜⎝
⎛∝
Nm B
c
φερh
*4exp
Metal Silicon Body
φB
c
cco A
R ρ=
ρc = Contact resistivty
Ac = Contact Area
– Dopant concentration, N– Barrier height, φB
• Fermi-level pinning results in:– Barrier height independent of metal work function
10/03/05
8
FLCC
Approaches to Lowering ρc• Material engineering
– SiGe source/drain• smaller bandgap smaller Schottky barrier
ρc ~10-8 Ω-cm2 for Ni germanosilicides on SiGe• lower resistivity
• Barrier height tuning– image force lowering by dopant segregation
– strain-induced φB reduction
• Fermi-level de-pinning by interface engineering– insertion of insulator layer– selenium passivation
A. Yagishita et al., SSDM, 2003
M. Ozturk et al, IEDM, 2002
M. Tao et al., APL, 2003
A. Kinoshita et al., Symp. VLSI Technology, 2004
D. Connelly et al., IEEE Trans. Nanotech., 2004
10/03/05
9
FLCC
Research Objective• To understand the mechanisms for tuning the effective
Schottky barrier height of a metallic electrode, to guide the engineering of contact-formation processes– Low-φB contacts for reduced parasitic resistance
GateGate
Silicon Substrate
TBOX
TSi
SiO2
SOI
GateGate
Silicon Substrate
TBOX
TSi
SiO2
SOI– Demonstrate fully silicidedsource/drain UTB MOSFETswith improved Idsat by reducing ρc (to <10-8 Ω-cm2) for silicide-to-silicon contacts Schematic Cross-section of
Silicide S/D UTB MOSFET
silicide
10/03/05
10
FLCC
Outline
• Introduction
• Characterization Schemes
• Si1-xGex Source/ Drain
• Dopant Segregation
• Strain
• Summary
10/03/05
11
FLCC
Contact Resistance Measurement
M. Ozturk et al, IEDM, 2002
Minimum measurable resistance is ~10 ΩNeed very small contact holes to determine ρc accurately
below 10-8 Ω-cm-2
10/03/05
12
FLCC
Test structuresmetal pad
active area
• Fabrication of Kelvin structures– Evaluation of contact resistance
• Fabrication of diode– Measure schottky barrier height
Kelvin Structure: Plan View
metalSiO2
lightly-doped Si
Diode: Schematic Cross section
10/03/05
13
FLCC
Fabrication of Sub-0.25µm Contacts
• Contacts were fabricated using DUV stepperASML5500/90; Cymer KrF excimer laser (l=248nm)
85mJ/cm265mJ/cm2 70mJ/cm2 75mJ/cm2 80mJ/cm2
after litho:
188nm 247nm 261nm 268nm 275nm
after etch:
171nm 211nm
10/03/05
14
FLCC
φB extraction • Measure diode I-V
characteristic at different temperatures
1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6
1E-9
1E-8
1E-7
1E-6
J/T2
1000/T
( ) ( )FBneF V
kTqAA
TI
−−≈⎟⎠⎞
⎜⎝⎛ φ**
2 lnln
10/03/05
15
FLCC
Outline• Introduction
• Characterization Schemes
• Si1-xGex Source/ Drain
• Dopant Segregation
• Strain
• Summary
(This work is sponsored by the FLCC project )
10/03/05
16
FLCC
Dopant Behavior in Ultra-Thin SOI
• P-channel thin-body FETs exhibit higher series resistanceDifferent behaviors of B and P in ultra-thin Si– dopant segregation to
interface(s), or into surrounding oxide?TSi=10nm
F.-L. Yang et al., 2004 Symp. VLSI Technology
TSi=80nm
D. Ha et al., IEDM 2004
10/03/05
17
FLCC
Epitaxial Si1-xGex Source/Drain• Epitaxial Si1-xGex source/ drain regions for lowering
Rseries and inducing compressive strain to enhance hole mobility– Conventional approach for epitaxial growth of Si1-xGex is
not possible for thin-body devices because there is not sufficient crystalline substrate after S/D etchback
Si1-xGex Si1-xGexSi1-xGex Si1-xGex
Silicon Substrate
Source Drain
SiO2
SOI
GateGate
Silicon Substrate
Source Drain
SiO2
SOI
GateGate
10/03/05
18
FLCC
Approach
• Develop a process for selectively forming strained Si1-xGex-in-SOI by intermixing Ge & Si– Study the intermixing of Ge with SOI films
• effects of anneal temperature, time, boron doping
– Investigate strain in the resultant Si1-xGex alloy
– Characterize metal-to-Si1-xGex contact resistance
– Germano-silicidation of Si1-xGex to achieve dopant pile-up (to study φB reduction)
10/03/05
19
FLCC
Advantages of This Approach• Selective deposition of Ge by conventional LPCVD
– GeH4 gas, 320oC, 200mT– high process throughput (batch process)
low cost
Ge Source Ge Drain
TSi = 3 nm
Gate
SiO2Si
XTEM of UTB MOSFET w/ raised Ge S/D
Y.-K. Choi et al., IEEE Electron Device Lett., Vol. 22, p. 447, 2001
10/03/05
20
FLCC
Ge/Si Interface Preparation• Selective Ge deposition in LPCVD furnace requires a
clean silicon surface – Interface preparation is critical
• Native oxide removal methods– in-situ HF vapor clean– in-situ HF vapor clean and Hydrogen bake– HF dip followed by Hydrogen bake– in-situ cleaning by GeH4
– Si I/I after Ge deposition• for breaking up any native oxide at the Ge/Si interface
10/03/05
21
FLCC
In-Situ GeH4 CleaningGeO
Silicon Substrate
GeH4
native oxide
H2O
Silicon Substrate
H2
Silicon Substrate
Decreased GeH4 flow to diffuse Ge on the surface
M. Moslehi, Proceedings of SPIE-The international Society for Optical Engineering, vol. 1393, pp.90-108, 1991.
Silicon Substrate
Clean surface
10/03/05
22
FLCC
Si Ion Implantation to Break Up Native Oxide
• Si+ implant has been used to break up the native oxide barrier for Solid-Phase Epitaxy (SPE)Y. C. Yeo et al., IEEE Transactions on Electron Devices, Vol. 49, No. 2, pp.279-286, 2002.
Si +
Silicon Substrate
Ge
Silicon Substrate
Ge
10/03/05
23
FLCC
Test Sample Process Flow• Starting wafers
– n-type, ρ=5-10 µΩ-cm• Cross-sectional TEM pattern
formation– CVD SiO2 deposition (52nm)
– Lithography– Oxide etching (90% dry + 10% wet)
Silicon substrate
Silicon substrate
CVD SiO2
SiO2 SiO2
Silicon substrate
Top-down view through optical microscope
10/03/05
24
FLCC
Process Flow (Cont’d)• Interface Preparation
– HF last / HF vapor• Selective Ge deposition (22nm)
– 320oC/200mTorr/100sccm– Capping layer (25nm)
• Si Implant– Si : 40keV, 1E15 cm-2
• Doping– B : 10keV, 2E15 cm-2
• Recrystallization – 500oC, 1 hour
• Intermixing anneal– 800oC/ 850oC, 1minute
SiO2 SiO2
Silicon substrateGe Ge Ge
Ge GeSiO2Ge SiO2
SiO2
Silicon Substrate
Silicon Substrate
SiO2 SiO2
SiO2
Si1-xGex
10/03/05
25
FLCC
Experimental Splits4 1
23Si Bon wafer splits = 4
B I/I4
none3
Si I/I2
Si I/I + B I/I1
B I/I4
none3
Si I/I2
Si I/I + B I/I1
Split Table:
xxxxxxxx850 C
xxxxxxxx800 CAnnealing
xxxxnone
xxxxSi I/I + B I/I
xxxxB I/I
xxxxSi I/II/I splits
xxxxxxxxHF vapor
xxxxxxxxHF LastCleaning
Wafer ID
xxxxxxxx850 C
xxxxxxxx800 CAnnealing
xxxxnone
xxxxSi I/I + B I/I
xxxxB I/I
xxxxSi I/II/I splits
xxxxxxxxHF vapor
xxxxxxxxHF LastCleaning
Wafer ID
10/03/05
26
FLCC
Determining Ge and B profiles(in collaboration with Prof. Haller’s group)
• Characterization of vertical and lateral co-diffusion of Ge and B– Available Options:
• Cross-sectional TEM with EDX nanoprobe• Cross-sectional TEM with EELS• Cross-sectional SEM with EDX nanoprobe
10/03/05
27
FLCC
Summary and Future GoalsSi1-xGex Source/Drain
• Fabrication of first batch is finished– Splits for Doped/ undoped Ge, interface preparation,
annealing conditions etc.
– Results of this batch are awaited
• Future Work:– Characterization of boron-doped Si1-xGex-on-insulator
resistance
– Characterization of metal-to-Si1-xGex contact resistance• Germano-silicidation of Si1-xGex to achieve dopant pile-up
for Schottky barrier height (φB) reduction
10/03/05
28
FLCC
Outline
• Introduction
• Characterization Schemes
• Si1-xGex Source/ Drain
• Dopant Segregation
• Strain
• Summary
(This work is sponsored in part by Intel Corp.)
10/03/05
29
FLCC
φB Reduction by Dopant Segregation• φB can be reduced by using an ultrathin (<10nm) heavily doped layer at the semiconductor surfaceJ. Shannon, Applied Physics Letters, Vol. 24, pp. 369-371, 1974.
– image force lowering (∆φ) due to surface electric field
N = dopant concentration in surface layer
a = width of heavily doped surface layerπεφ
4aNq
si
=∆
Such a thin heavily doped layer can be formed by silicidation-induced dopant segregation:A. Kinoshita et al., 2004 Symp. VLSI Technology, Digest of Technical Papers, pp. 168-169.
10/03/05
30
FLCC
ExperimentGoals: Confirm dopant segregation w/ NiSi
Investigate dopant activation
Process sequence:• starting Si wafer (n-type/ p-type)• deposit capping layer, CVD SiO2• blanket implantation (B- 20KeV, 1016cm-2 As- 80KeV, 6×1015cm-2)• spike annealing@1000C• strip capping layer• excimer laser annealing• Ni deposition• silicidation• strip unreacted Ni
starting substrate
Ni
unreacted Ni
NiSi
10/03/05
31
FLCC
Results: SIMS Analyses
0 20 40 60 80 100 120 140 1601E17
1E18
1E19
1E20
1E21
1E22
O
NiSi
As
Con
cent
ratio
n (a
tom
s/cm
-3)
Depth (nm)
1
10
100
1000
10000
100000
Cou
nts/
sec
– Dopant pile-up at silicide/ Si interface is seen for both As and B doping (only As shown here)
10/03/05
32
FLCC
Schottky Barrier Lowering• Image Force Effect
– Induced charges at the interface– Equivalent to an image charge
16mS
qxEπε
=
SemiconductorMetal
EF
EC
xqV
qφB
qVb
qφBO
q∆φ
xm W
-qψ(x)
qExx
qxPEs
−−=επ16
)(2
Eqs
•=∆⇒επ
φ4
10/03/05
33
FLCC
Tailoring the Surface Electric Field
• ∆φ depends on the surface electric field– Lightly doped substrate:
• Low E, barrier-lowering is sensitive to reverse bias– Heavily doped substrate:
• High E φB is reduced, but reverse current increasestunneling Ohmic contact !
• To retain Schottky junction properties and to achieve ∆φ that is insensitive to bias, a heavily doped surface layer that is fully depleted by the built-in potential is needed
10/03/05
34
FLCC
Fully-Depleted Doped Surface Layer
• The required electric field has been shown to be >5×105 V/cmJ. Shannon, Applied Physics Letters, Vol. 24, pp. 369-371, 1974.
– Maximum surface field arising from implantation of a symmetrical distribution of charge about range Rp:
p
bs R
VE ≅max,
)(0 fcBb EEqqV −−= φqφB
q∆φ
qφBO
q∆φ
higher carrier concentration
lower carrier concentration
qVb
For a metal contacting a lightly doped substrate, built-in potential Vb~0.45V
ARp 100<⇒
10/03/05
35
FLCC
ΦB Reduction Model• Maximum surface field, Es,max
NB
NiSi
Np
NiSiNB
Np
a
[ ])(max, aWNaNqE Bp
s
s −+=ε
aNqE p
s
s ε≅max,
πεφ
4aNq p
s
≅∆⇒
Expected barrier height lowering due to a thin highly doped surface layer:
1018 1019 10200.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
design space
a=100A0
a=50A0
Bar
rier l
ower
ing,
eV
active dopant conc. in surface layer (cm-3)
10/03/05
36
FLCC
Effect of Interface States• Metal-induced gap states (MIGS)
– Penetration of wave function from the metal into the forbidden energy gap of Si
qφBO
Ef
Metal Semiconductor
Metallic wave penetration in metal-semiconductor system
λ
ελ
επx
ss
eqQqExx
qxPE −−−−=16
)(2
Electron potential energy including the contributions of image force and MIGS K. Shenai et al., IEEE TED, Vol. ED-32, No. 4, pp. 793-799, 1985
Q = magnitude of surface state charge
λ = penetration depth of surface state charge
10/03/05
37
FLCC
∆φ Inverse Modeling Approach– Find the location of PE(x) minimum
– Total barrier lowering is given by
– Solve Poisson’s equation to find Ψ0(x)
– Extract Q and λ from measured forward I-V characteristics
– Predict total barrier lowering from the model • Np, work-function difference, Q, and λ are input parameters
0)(16 2
2
=+−⇒ − λ
επεmx
s
m
ms
eqQxqEx
q0))(( =
= mxxxPE
dxd
λ
ελψψ
πεφ
mx
s
m
ms
eQxx
q −+−−=∆ )()0(16 00
s
xxερψ )()(0
2 −=∇
10/03/05
38
FLCC
Determining Active Dopant Concentration• Spreading Resistance Probe (SRP):
– within 1/2 the probe spacing (~10µm) of the Si/ silicide interface, silicide starts affecting readings because of low resistanceNeed to remove NiSi selectively• surface roughness increase• difficult to find bevel edge
NiSi
Si
10/03/05
39
FLCC
Summary and Future WorkDopant Segregation
• Ni silicidation induced dopant segregation phenomenon confirmed
• A quantitative inverse-modeling approach has been established for determining the amount of Schottky-barrier lowering
• Future Work:– Fabrication and characterization of diode structures and
Kelvin structures– Application of dopant-segregation technique to improve
FinFET performance by reducing S/D contact resistance
10/03/05
40
FLCC
Outline
• Introduction
• Characterization Schemes
• Si1-xGex Source/ Drain
• Dopant Segregation
• Strain
• Summary
10/03/05
41
FLCC
φB Reduction by Si StrainA. Yagishita et al., SSDM 2003
– 1% bi-axial strain reduces φB by 0.1eV (ErSi1.7 S/D NMOSFET)
10/03/05
42
FLCC
Experimental Plan
K. Uchida et al., IEDM 2004
• Use a bending apparatus– apply uniaxial or biaxial bending
stress to Si chips
• Study φB reduction– Fabricate Schottky diodes and
contact test structures to measure effect of strain on ρc
10/03/05
43
FLCC
Outline
• Introduction
• Characterization Schemes
• Si1-xGex Source/ Drain
• Dopant Segregation
• Strain
• Summary
10/03/05
44
FLCC
Summary• Source/drain contact resistance can limit the
performance of nanoscale FETs– ρc ~10-9 Ω-cm2 will be required
• Approaches for reducing ρc include use of Si1-xGexin the source/drain regions, dopant segregation, and strain
• Work in progress will clarify the mechanisms for lowering the effective Schottky barrier height φB
Application to nanoscale thin-body FETs
10/03/05
45
FLCC
Acknowledgements
• Akira Hokazono (Toshiba Corporation)• Dr. Chi On Chui (Intel Corporation)
• Research funding from– UC Discovery Grant program and member companies
of the Feature-Level Compensation and Control (FLCC) project at UC-Berkeley
– Intel Corporation