Transcript Slide 1

SiNANO Workshop
Carrier mobility enhancement in
strained silicon germanium channels
David Leadley
University of Warwick
Collaborators
Warwick
Tim Grasby, Andy Dobbie, Chris Beer,
Jon Parsons, Evan Parker, Terry Whall
IMEC
Gareth Nicholas, Marc Meuris, M Heyns,
P Zimmerman, Matty Caymax, ++
SINANO partners
KTH, Udine, Chalmers, AMO
Why is mobility still critical ?
e
 *
m
High mobility - light mass and minimal scattering
ITRS – long term years
By 2020, Lg=5nm, Vdd=0.7V
Mobility enhancement factor of 1.04 each node
Ballistic enhancement factor x2
Double gate structures – light doping
Why add Ge?
ADVANTAGES
– Smaller bandgap
– Lighter hole mass
– Strain – splits bands and reduces scattering
BUT…
– Native oxide no good
– Band-to-band tunnelling
Pseudomorphic Si0.64Ge0.36 on Si
2
Effective Hole Mobility (cm /Vs)
200
150
Mobility doubled
SiGe
100
Si Control
50
10m x 160m
0
0.10
0.15
0.20
0.25
0.30
Effective Field (MV/cm)
0.35
0.40
Impact Ionisation
101
I body
100
Si Control
I source
M-1 (no units)
10-1
SiGe
10-2
10-3
10-4
Vgs = -1.5 V
10-5
0.0
-2.0
-4.0
-6.0
-8.0
Vint (V)
Reduced impact ionisation for SiGe, despite higher mobility
Nicholas et al, Electronics Letters 41, 20052074
High-k gated SiGe
Deposited HfO2
Si cap – oxidises to thin SiO2 interlayer
Metal
HfO2
SiO2
SiGe
Si
< 1nm
Lowest EOT 12Å
Metal gate/HfO2 gate stack development for
Si(cap)/Si0.8Ge0.2 channel
1e+0
1013
1e-1
as-deposited
425C, 30min, N2/H2
1e-2
1e-3
-1
Average Dit (cm eV )
800C, 30sec, Ar
950C, 30sec, Ar
1e-4
-2
Jg (A/cm2)
1012
1e-5
1e-6
1e-7
1e-8
1011
1e-9
as-deposited
425C, 30min, N2/H2
950C, 30sec, Ar
1010
Si Control
SiGe 1nm Cap SiGe 2nm Cap SiGe 3nm Cap
950C, 30sec, Ar
1e-10
Thick TiN
Thin TiN
W
1e-11
1e-12
-4
-3
-2
-1
0
1
2
3
4
Vg (V)
Interface state density (W gate)
Gate leakage
Warwick, AMO, Chalmers
High-k/metal gated Si0.8Ge0.2
surface p-channel devices
0.10
Silicon
SiGe 1nm cap
SiGe 3nm cap
SiGe 4nm cap
120
100
2
Effective Mobility (cm /Vs)
0.08
Drain Current (A/m)
140
Silicon
SiGe 1nm cap
SiGe 3nm cap
SiGe 4nm cap
0.06
0.04
80
60
40
0.02
20
L = 50m, W = 50m
Vds = -50mV
L = 50m, W = 50m
0
0.00
-3
-2
-1
Gate Overdrive (V)
0
0.1
0.2
0.3
0.4
0.5
Effective Field (MV/cm)
........enhancements beating Intel (IEDM 2004)
KTH, Warwick, Chalmers, AMO
Effective Mobility
Best SiGe (25%) devices show mobility enhancement over silicon control
Mobility degradation compared to universal – interface roughness and Coulomb
scattering important
160
Si Universal
Si Control
SiGe 1nm Cap
SiGe 2nm Cap
SiGe 3nm Cap
SiGe 4nm Cap
SiGe 5nm Cap
120
2
Effective Mobility (cm /Vs)
140
100
80
60
40
20
0
0.03
0.05
0.07
0.1
0.2
0.3
0.5
0.7
1
Effective Field (MV/cm)
KTH, Warwick, Chalmers, AMO
HOT SiGe
Hybrid Orientation Technology
– n on (100), p on (110) in Si, Yang et al. IEDM 2003
– in Si, no variation with orientation for long channels,
but <110> best for short channels, Saito et al VLSI 2006
– 50% enhanced Idsat for SiGe (110) over Si (100),
3.3x for mobility, Liu et al VLSI 2006
[001]
[011]
[011]
(100) substrate
[110]
(110) substrate
NovelPMOSFETs,
p-channel/substrate
V =-0.1 V, L=25 µm, W=4.6orientations
µm
DS
<112> <111>
<110>
<110>
<100>
0,5
<112>
<100>
<110>
<100>
<112>
<110>
gm [µS/µm]
0,4
0,3
Si Ge
0.8
0,2
0,1
0.2
(110) t =4.5 nm
ox
Si (100) t =2.6
nm
= 3.2
nm
ox
0
0
30
60
90
120
150
180
Rotation [deg]
… enhancement beating Liu et al (VLSI Symp 2006)
KTH, Warwick, Chalmers, AMO
Unstrained Ge pMOS
High performance Ge pMOS devices using a Si-compatible process flow
Zimmerman et al. IEDM 2006
4nm HfO2
2μm Ge on Si substrate
12Å EOT, TDD 107-108 cm-2
PMOS fabricated with Lg ~125 nm
IMEC
Unstrained Ge
Ge
+ anneal
Si
IMEC
Nit spread 0.2-8x1012cm-2
- affects Ids, gm and Vt
Uniformity restored after anneal
IMEC
Thinner EOT?
Si
Mobility increase with thinner HfO2
- less charge trapping
Starts to leak when too thin !
Ragnarsson et al. IEEE TED 53, 1657 ( 2006)
IMEC
Mobility in unstrained Ge
μ > 300 cm2/Vs
Correction required for Rsd in short channel devices Lg<0.25μm
IMEC
Investigating trapped charge
Dit varies across wafer 1011 - 2x1012cm-2
Vt  Dit
1
 Peak

1
Coulomb
1
 Peak
1
Coulomb

1
 SR

1
 Phonon
 number of charges,nit
 Vt
Warwick, IMEC
Interface
charge
density
Warwick, IMEC
Extract Dit in 3 ways at 300K:
i. from Vt – average all energies
ii. subthreshold slope – specific energy
iii. charge pumping
Mobility modelling at 4K
1
1
1
At low T, with confinement


only have one HH subband.  Peak Coulomb  SurfaceRoughness
1
Coulomb
 ni
1
 SR
 2 2 K F4
Fit using ni and  as parameters ...
Warwick, IMEC
Modelled 4K mobility
As grown
Annealed
ni agrees with values from Dit
Δ decreases after anneal from 0.6nm to 0.5nm,
hence μSR increases by 35%
μ300K must also depend on surface roughness
Warwick, IMEC
Strain tuning buffers
 Strained-Si by local strain now in production
 Si1-yGey virtual substrates for global strain
o y~0.2 for s-Si
o y~0.5 for s-Si1-xGex
o y~0.8 for s-Ge
Need low TDD and zero pile-up
Terrace Graded VS – better than
industrial quality
20% Ge
XTEM
TDD ≤ 105cm-2, PUD=0
10x10um2 AFM image indicating
RMS roughness 1.5nm
Warwick, LETI, Jeulich
State-of-the-art sSi electron mobility
from TG-VSs
y = 15 -27%
strain 0.6 – 1.1%
State of the art
F. Driussi et al., ULIS 2007
Warwick, Udine, KTH
sGe global platform
- 80% terrace grade
TDD=3x105/cm-2, PUD = 0, RMS ≈ 8nm
Warwick, FZ-Jeulich, LETI
Novel thin VS for sGe
Strained Si
Strained Ge
300nm Si0.2Ge0.8 Grade
Seed layer
Si Sub
20x20 um2 AFM image.
1.3nm RMS,
fully relaxed 80% platform
TDD ≈ 106/cm2 range
Warwick
Strained Ge pMOS
150nm TiN
4nm ALD HfO2
Positive Vt due to gate workfunction
ASM, Warwick, IMEC
Record Ge mobility via CMOS process
Effective Hole Mobility (cm2/Vs)
700
25 nm strained Ge
600
Peak μ = 650 cm2/Vs
500
15 nm
400 strained Ge
300
1 nm Si
Unstrained Ge
200
2 nm Si
100
Si/SiO2 universal
0
0
0.2
0.4
0.6
0.8
1
Effective Field (MV/cm)
G. Nicholas et al., IEEE EDL 28, 825 (2007)
Warwick, IMEC
Further comments on s-Ge
Full band Monte Carlo calculations, incl. BTBT, SCE etc.
show best prospect for pMOS …
… strained-Ge directly on Si !
Krishnamohan et al (Stanford) VLSI (2006)
For s-SiGe OI, Idsat enhancement in short channels
exceeds μ increase in long channel.
Tsutomu Tezuka et al. VLSI (2006)
nMOS still a problem for Ge …
… but tensile strain Si0.1Ge0.9 on relaxed Ge promising.
Summary
Addition of Ge improves mobility
High-k makes Ge viable
Mobility enhancements relevant for
nanometre scale devices
Terrace grading – new generation of virtual
substrates
30%
TDD ≈ 4 x 104 cm-2
PUD = 0
Warwick
LG 30%
≈105 cm-2 TDD
Pile-up ≈1 cm-1
High stability sSi layers
Relaxation (%)
Linear Grading
Terrace Grading
Strained Si thickness (nm)
J. Parsons et al., ULIS 2007
Warwick, FZ-Jeulich
Extended Stacking Faults
Defect etch image of 20nm
strained silicon layer
J. Parsons et al., ULIS 2007
PVTEM image of 180nm
strained silicon layer
Warwick, FZ-Jeulich
Extended Stacking Faults
“Perfect”
dislocation
Stacking fault
Trailing (30°)
Leading (90°)
“partial” dislocation
“Partial” dislocation
S-Si
SiGe
Stacking faults
– effective misfit dislocation blockers?
Defect etch images of 30nm strained silicon layer
J. Parsons et al., ULIS 2007
Warwick, FZ-Jeulich
50% terrace graded (for sSi)
TDD = 2x105 cm-2
PUD = 0
Warwick
≈ 20% TG - slower growth
(+sSi)
TDD = 4 x 104/cm2
PUD = 0
Schottky Barrier MOSFET- world best?
Vds = -1.4 V
1.0E+3
1.0E+2
Tunnelling f(m*)
1.0E+1
1.0E+0
Id (mA/mm)
1.0E-1
1.0E-2
1.0E-3
1.0E-4
1.0E-5
1.0E-6
Thermionic
emmision
1.0E-7
1.0E-8
1.0E-9
1.0E-10
-3.0
-2.5
-2.0
-1.5
-1.0
-0.5
Vg (V)
Measurement
D.J. Pearman et al., IEEE Trans ED (accepted 2006)
Modelling
UCL, Warwick, Glasgow