Guidelines for MOSFET Device Guidelines for MOSFET Device
Optimization accounting Optimization accounting
for L
for L - - dependent Mobility dependent Mobility Degradation
Degradation
G. Bidal1,2, D. Fleury1,2, G. Ghibaudo2, F. Boeuf1 and T. Skotnicki1.
1STMicroelectronics, 850 rue Jean Monnet, 38920 Crolles Cedex, France;
2IMEP, 3 parvis Louis Néel, BP 257, 38016 Grenoble Cedex 1, France;
Outline Outline
• Introduction
• Methodology used in this work
– Low field apparent mobility – µ-degradation modeling
• Experimental results
– Impact of technological modules
• Guidelines for transport enhancement
– How close to ballistic ?
– Benchmark of studied technological modules
• Conclusion
Outline Outline
• Introduction
• Methodology used in this work
– Low field apparent mobility – µ-degradation modeling
• Experimental results
– Impact of technological modules
• Guidelines for transport enhancement
– How close to ballistic ?
– Benchmark of studied technological modules
• Conclusion
Metal gate
Mobility crisis in highly scaled Mobility crisis in highly scaled
devices devices
0 50 100 150 200 250 300 350 400 450
0.01 0.1 1
Effective Channel Length Leff (µm) Low Field Mobility µ0 (cm2 .V-1 .s-1 ) electrons
holes
squares: data from [1]
circles: data from [2] • Experimental evidence of
carriers’ mobility diminution as Lg is scaling down [K.M.Cao IEDM 99, K.Rim IEDM 02, M.Zilli EDL 07, Ramos ESSDERC 06]
• Observed on:
Poly-Si gate
SiO2 dielectric
High-K dielectric
Strained Unstrained
Doped channel Undoped channel Bulk
SOI FinFET
[2]: A.Cros et al., IEDM 06 [2]: F.Andrieu et al., VLSI 05
Motivation
1.00E-05 1.00E-04 1.00E-03 1.00E-02
0.01 0.1 1 10
µshort x2
Same long channel mobility
Velocity limit Mobility limit
log Leff (µm)
1.E- 1.E- 1.E- 1.E-
• If mobility is high enough and only in this case, the
mobility term 1/(µ0Elateral) can become negligible and the transport will be mainly
driven by the limiting velocity
• Any mobility enhancement will get us closer to the
limiting velocity
) /(
1 /
1
1
0
lim lateral
gt ox
dsat WC V v E
I = × + µ
From T.Skotnicki et al., IEEE TED’08
I
ON• Many possible explanations…
- long range CS from S/D
[M.Cassé et al., VLSI 2008]
- CS from the high-K [V.Barral et al., SNW 2008]
- neutral defects due to S/D I/I
[A.Cros et al., IEDM 2006]
- etc.
Purpose of this work
• … that may be mixed !
Final mobility
• Purpose of this work is not to diagnose the mechanisms of this mobility degradation, but to identify key technological modules that may help to reach a higher short channel mobility
How to ?
0 50 100 150 200 250 300 350 400 450
0.01 0.1 1
Effective Channel Length Leff (µm) Low Field Mobility µ0 (cm2 .V-1 .s-1 ) electrons
holes
squares: data from [1]
circles: data from [2]
• By fitting experimental results by an empirical model in order to provide a simple
benchmarking tool between the different technologies
[2]: A.Cros et al., IEDM 06 [2]: F.Andrieu et al., VLSI 05
Outline Outline
• Introduction
• Methodology used in this work
– Low field apparent mobility – µ-degradation modeling
• Experimental results
– Impact of technological modules
• Guidelines for transport enhancement
– How close to ballistic ?
– Benchmark of studied technological modules
• Conclusion
Low field mobility
0 20 40 60 80 100 120 140 160
0.0E+0 5.0E-7 1.0E-6 1.5E-6
Inversion Charge Density (C.cm-2) Effective Mobility µeff (cm2 .v-1 .s-1 )
Leff reduction = µeff reduction = µ0 reduction
Lines : Y-function Symbols: Split CV
µ0=µeff(Qinv≈0)
• Y function and split C(V) methods give similar results, except under Vth
• Leff shrink implies µ0 degradation and µeff degradation
2 2 1
0
. .
1 gt gt
eff V V
µ µ
θ
θ +
= +
eff d
inv
eff d
eff Q V W
L µ I
. .
= .
2 2 1
0
. .
1 gt gt
eff V V
µ µ
θ
θ +
= +
eff d
inv
eff d
eff Q V W
L µ I
. .
= . Y-function:
Split C(V):
[J. Koomen et al., SSE, 1973]
[G. Ghibaudo et al., Electronics Letter, 1988]
[K.Romanjek et al., EDL, 2004]
Empirical Model
0 50 100 150 200 250 300 350 400 450
0.01 0.1 1 10
Effective channel length Leff (µm) Low field mobility µ0 (cm²V-1 s-1 )
αµ=0 αµ=0.05 αµ=0.1 αµ=1 µmax=µlong=200
µmax=µlong=300 µmax=µlong=400
eff µ
eff µ L
L µ
+ α
=
max 0
1 )
( Model : 1
(with 2 fitting
parameters)
eff µ
eff µ L
L µ
+ α
=
max 0
1 )
( 1
degradation factor
maximum mobility
• Zero degradation
corresponding to αµ =0 does not exist: αµ cannot be lower than
αµ,bal given by the apparent
mobility reduction due to ballistic transport
Fitting or not fitting ?
10 100 1000
0.01 0.1
effective channel length Leff (µm) low field mobility µ0 (cm².V-1 .s-1 )
symbols: measured lines: model
nM OS
• Perfect fit is obtained on
experimental data at short gate length
• Bulk and Ultra thin Body (UTB) results
• PolySi/SiON and Metal/High-K results
Outline Outline
• Introduction
• Methodology used in this work
– Low field apparent mobility – µ-degradation modeling
• Experimental results
– Impact of technological modules
• Guidelines for transport enhancement
– How close to ballistic ?
– Benchmark of studied technological modules
• Conclusion
Impact of SiON thickness
0 50 100 150 200 250 300 350 400 450
0.01 0.1 1
effective channel length Leff (µm) low field mobility µ0 (cm².V-1 .s-1 )
αµ=0.15 µmax=400
αµ=0.25 µmax=290 symbols: measured
lines: model
nM OS
Poly-Si gate
DPN SiON Si substrate
17Å
Si substrate Poly-Si gate
DPN SiON 12Å
Impact of metal gate material
0 50 100 150 200 250 300 350 400
0.01 0.1 1
effective channel length Le ff (µm) low field mobility µ0 (cm².V-1 .s-1 )
TaC gate TaN gate TiN gate αµ=0.18
µmax=360
αµ=0.28 µmax=300 symbols: measured
lines: model
Metal gate on
high-K TaC, TaN or
TiN
Impact of UTB doping
0 50 100 150 200 250 300
0.01 0.1 1
effective channel length Leff (µm) low field mobility µ0 (cm².V-1 .s-1 )
undoped doped αµ=0.30
µmax=300
αµ=0.31 µmax=150 symbols: measured
lines: model
nM OS
UTB doped
or undoped
Impact of junction formation
0 50 100 150 200 250 300 350 400 450 500
0.01 0.1 1
effective channel length Le ff (µm) low field mobility µ0 (cm².V-1 .s-1 )
optimized S/D I/I
@1000°C
S/D I/I @1010°C αµ=0.12
µmax=500
αµ=0.18 µmax=390 symbols: measured
lines: model
+ activation anneal Ion Implantation
Impact of strain
0 50 100 150 200 250
0.01 0.1 1
effective channel length Le ff (µm) low field mobility µ0 (cm².V-1 .s-1 )
αµ=0.10 µmax=200
αµ=0.15 µmax=90 w/ eSiGe stressors (PIS)
Ref. without stressors
discrepancy due to relaxed strain on long devices
symbols: measured lines: model
pM OS
eSiGe eSiGe
x y
Impact of crystal orientation
0 50 100 150 200 250
0.01 0.1 1
effective channel length Le ff (µm) low field mobility µ0 (cm².V-1 .s-1 )
αµ=0.16 µmax=230
αµ=0.15 µmax=90 (100)/<110>
(110)/<110>
symbols: measured lines: model
pM OS
(110)
<110>
L
g(100)
<110>
L
gOutline Outline
• Introduction
• Methodology used in this work
– Low field apparent mobility – µ-degradation modeling
• Experimental results
– Impact of technological modules
• Guidelines for transport enhancement
– How close to ballistic ?
– Benchmark of studied technological modules
• Conclusion
How close to the ballistic ?
• Boosting µ and reducing α is mandatory to reach high BR
0.025 0.075
0.125 0.175
0.225 0.275
25 275 525
0 10 20 30 40 50 60 70 80 90 100
Long channel mobility µmax
BallisticRate BR(%)
Deg rad
ation facto
r α
µ
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
0 200 400 600
µmax [cm2.V-1.s-1] αµ [nm.V.s.cm-2 ]
nMOS pMOS
αµ,bal for electrons=0.04 αµ,bal for holes=0.08
Final benchmark & Guidelines
X2
= ++
Crystal orientation
x3 +
++
Process Strain
x2
= ++
Channel doping
x2 +
+ Junctions
x2 to x4 ++
+ Gate stack
Merit factor η=µmax/αµ impact
on αααα
µ
impact on µmax Techno.
module
• Influence of each studied technological module is quantified