logic cmos scaling - applied materials
TRANSCRIPT
LOGIC CMOS SCALING
IN THE NEXT DECADE & BEYOND
AARON THEAN , Ph.D.
Mar 2014
APPLICATION TECHNOLOGY
TRENDS
3
© IMEC 2014
Sensor / MtM
Networks
Smart Mobile
Devices
Data Centers
& Servers
EMERGING DATA INFRASTRUCTURE
4
© IMEC 2014
1.5EB 2.6EB
+81%
GLOBAL MOBILE DATA FORECAST (CISCO)
• Astounding demand on Data Processing & Exchange
foreseen for the next 5 years and beyond?
15.9EB
6x
5
© IMEC 2014
GLOBAL MOBILE DATA FORECAST (CISCO)
• Continuous demand for devices to support data traffic driven by
growing population & information consumption ?
2014 Global Pop. ~ 7B
2020 Global Pop. ~ 8B
6
© IMEC 2014
PROCESS TECHNOLOGY EXPECTATIONS
Expected Leading CMOS technology
in product
22nm/20nm
28nm
14nm?
20nm?
10nm?
14nm?
• How can advanced CMOS technologies enable increasing
applications?
Mixed-Signal RF, Low-Cost
CMOS (Mixed-Tech)
7
© IMEC 2014
Sensor / MtM
Networks
Smart Mobile
Devices
Data Centers
& Servers
EMERGING DATA INFRASTRUCTURE
8
© IMEC 2014
Sensor / MtM
Networks
Smart Mobile
Devices
Heat
• Performance-centric
• Heat-Dissipation Limited
• User-Experience centric • Battery & Form-Factor
constraints
• Perf-Power Balanced Battery
• Ubiquitous Deployment • Ultra-Low Power
• Ambient Energy Sources?
• Low Cost
SYSTEM-DEVICE-PROCESS SPECIALIZATIONS
Data Centers
& Servers
9
© IMEC 2014
Sensor / MtM
Networks
Smart Mobile
Devices
Data Centers
& Servers
SYSTEM-DEVICE-PROCESS SPECIALIZATIONS
10
© IMEC 2014
PROCESS TECHNOLOGY
TRENDS
11
© IMEC 2014
1 µm 0.8 µm 0.5 µm 0.35 µm 250 180 130 90 65 45 32 22 14 10 7 5
INFLECTIONS: DISRUPTIONS OR ENABLERS?
Copper Interconnect Shallow-Trench Isolation
Embedded SiGe SD pFET
Trench contact + Silicide last
1st Gen RMG High-k
Gate-First MGHK &
SiGe channel 2nd Gen RMG High-k
Local interconnect
FinFET
LELE Metal Interconnect
Ultra-Thin SOI
Spacer-Defined Multiple Patterning
R&D in
Progress
BOTH: THEY ARE INNOVATIONS
Semiconductor Industry Technology nodes
INNOVATION-TO-PRODUCT R&D CYCLES
45nm
2001 2003 2005 2007 1999
32nm/28nm
22nm/20nm
2009 2011
14nm DEV. RESEARCH PROD
14nm DEV. RESEARCH PROD
14nm DEV. RESEARCH PROD
1990-1995
High-k Metal Gate Proposed
1998-2000 FinFET Proposed & Demo’d
•High-k Metal Gate in products
14nm/16nm 14nm DEV. RESEARCH PROD
•FinFET/Trigate in products
2013
• Approx. 10 years or more to bring some new game-changing
processes to product ... Research needs to start early.
10nm
7 nm
5 nm
2014 2016 2018 2020 2012
3 nm
2022 2024
• IMEC: Shared-Resources for Precompetitive (& Early) R&D
14nm DEV. RESEARCH PROD
14nm DEV. RESEARCH PROD
N+1 FinFET platform
- Power/performance
- Area (193i vs EUV)
- Cost
14nm DEV. RESEARCH PROD
14nm DEV. RESEARCH PROD
IMEC
IMEC
IMEC
N+3: device, circuit & physical
design exploration
N+2: process modules/ feature
screening
N+4: New materials, Beyond-
CMOS device concepts, &
System implications
IMEC: PROCESS R&D PIPELINE
14
© IMEC 2014
1 µm 0.8 µm 0.5 µm 0.35 µm 250 180 130 90 65 45 32 22 14 10 7 5
imec founded
What ’s next ?
SOI
FinFET
Bulk FinFET
FUSI MG
HKMG RMG-HK
SiGe Epi rSi Epi
Ge
MOSFET
GaAs on Ge
Si Epi GAA-SOI
PD/FD-SOI TFET
SiGe pFET
SiGe channel
Buried Spacer-defined Fins
Vertical SiGe FET
today
IMEC: INNOVATIVE PROCESS OPTIONS
Semiconductor Industry Technology nodes
15
© IMEC 2014
WHAT’S NEXT?
• Logic Devices R&D
• Extending Si
• Beyond Si
• Managing Variability
• Beyond CMOS
• Power, Density, Cost Scalings
• The challenges of the transistor
• Conclusion
16
© IMEC 2014
POWER, DENSITY, & COST SCALINGS “THE CHALLENGES OF THE TRANSISTOR”
17
© IMEC 2014
POWER CRISIS: THE NEED FOR VDD SCALING
Po
wer
(W/C
hip
) S
up
ply
Vo
ltage
,Vd
d (V)
Year
0.5
0.7
Mobile mProc:
Power-source limited
• Higher-Performance & Mobile Processors are power limited
22/20nm 32/28nm
45nm
130nm 90/65nm
CMOS Technologies
• Vdd scaling slowing – What are the process limiters?
14/10nm
5/7nm
?
• How can we overcome them for post-14nm CMOS technologies?
High-Perf mProc:
Thermal Limited
(Adapted from various sources
Includes T.Masuhara, IEEE-SSCM 2013)
18
© IMEC 2014
LIMITERS OF VDD SCALING:
THE ROLE OF LOGIC TRANSISTORS
• Circuit performance loss & variability limits Vdd scaling
• Process & Transistor capability play critical roles
Rin
g O
scilla
tor
Dela
y (
s)
Vdd (V) 1.1 1.0 0.9 0.8 0.7
High Vt
LowVt L
og
(Id
s)
Ioff
High Vt
Vgs (V) 0 Vdd
Low Vt
1. Enhance Drive: High Mobility, Low Raccess Ceff
2. Improve
Electrostatics Sub-Vt swing & DIBL
Rapid performance loss
at low Voltage
Increased variability at
low Voltage
3. Reduce process &
material variations Random & Systematic
19
© IMEC 2014
High-Mobility Materials FinFET Stress Simulations
0
20
40
60
80
100
120
DR14
DR10
DR7
Mo
bilit
y in
cre
ase (
%)
CE
SL TiN
W g
ate
W c
on
tactS
/D s
tre
ss
TO
TA
L
N10
N7
N5
S/D Stressor
Strongest
Looking at scaling
10nm 5nm features sizes
Reducing benefit if
Lg scaling slows
MOBILITY ENHANCEMENT:
STRESSOR VS. ALTERNATE CHANNEL
Ge n/p FETs:
2-6x gains III-V :
10x gains
Narrow Band-
Gap limited
• Different Means = Different Trade-Offs
Dynamic Power
N20 Planar
Vdd=0.9V
N14
Vdd=0.9V
N10 Si FinFET
Vdd=0.8V
N7 Si FinFET
Vdd = 0.7V
N7 IIIV/Ge
Vdd = 0.7V
17%
35%
29%
25%
Leakage Power
N7 IIIV/Ge
Vdd = 0.7V
N7 Si FinFET
Vdd = 0.7V
N10 Si FinFET
Vdd=0.8V
29% 25% 30%
45%
POWER-PERFORMANCE BENCHMARK:
STRESSOR VS. ALTERNATE CHANNEL
Alt. channel offers more
performance – but scalability
needs to be assessed.
Performance gains from
FinFET wrt. planar
Si channel fin can scale to
maximize performance –
Limited by parasitics &
process capability.
• Value in assessing Si & Alternate Channel FinFET scalability
21
© IMEC 2014
N10 M1 DESIGN STYLES AND PATTERNING OPTIONS
2D
Targ
et
layer
• Restricted, 1D; 1.5D Vs unrestricted, 2D designs
• Complex 2D shapes (requiring stitching) are gradually simplified to a fully
unidirectional design style with 1D shapes only to enable scaling
• 1.5/1D have better TDDB performance due to the self-alignment
• 10-15% area penalty is inevitable due to the enforced regularity
48nm pitch
Unrestricted 2D
193 LE3 or EUV
48-45nm pitch
Simplified/Restricted 2D
SADP + block (193 / EUV)
40-45nm pitch
Restricted 1D
SADP + block (193LE2 / EUV)
SADP core
1.5D
block block (LE2)
1D
SADP core
IMEC, B. Vandewalle (SPIE 2014)
Fully-depleted
Channel for Improved
Electrostatics
Multi-gate FETs Ultra-Thin
SOI
Band-Engineered
Channel for
Enhanced Transport
High-Mobility
Channels
Gate-All-
Around,
Nanowires/
Tunnel FETs
(SiGe, Ge IIIV)
Novel Materials/
New Transport/
Extreme
Electrostatics
2D Materials
(Bi-layer
Graphene)
Quantum/
Spin Devices
• Material & Device Architecture Innovations: Enablers of continual scaling
Vdd 1.0-1.1V 0.9-1.0V 0.8-0.9V 0.7-0.8V 0.6-0.7V 0.5-0.6V < 0.5V
Strain &
Advanced Gate Stack
Engineering
SD/stressors Metal Gate
+High-k
MGHigh-k
Si substrate
MGHigh-k
Si substrate
45nm
32/28nm
22/20nm
14nm
10nm
7nm
5nm
Tech Node ...
• Feature Dimension & Voltage Scaling are concurrent drivers
IMEC LOGIC DEVICES R&D ACTIVITIES
23
© IMEC 2014
LOGIC DEVICES R&D
ENABLING PROCESS SPECIALIZATIONS
Fully-depleted
Channel for Improved
Electrostatics
Multi-gate FETs
Band-Engineered
Channel for
Enhanced Transport
High-Mobility
Channels
(SiGe, Ge IIIV)
Novel Materials/
New Transport/
Extreme
Electrostatics
2D Materials
(Bi-layer
Graphene)
Quantum/
Spin Devices
Vdd 1.0-1.1V 0.9-1.0V 0.8-0.9V 0.7-0.8V 0.6-0.7V 0.5-0.6V < 0.5V
Strain &
Advanced Gate Stack
Engineering
SD/stressors Metal Gate
+High-k
MGHigh-k
Si substrate
MGHigh-k
Si substrate
45nm
32/28nm
22/20nm
14nm
10nm
7nm
5nm
Tech Node ...
• Modules to improve Non-planar Si channel performance
IMEC LOGIC DEVICES R&D ACTIVITIES
• Enhancing &
Extending Si
FinFET Scaling
EXTENDING SILICON
• Investigate solutions to extend the scaling of FinFETs from 14nm 7nm
Implant damage & Advanced doping
Atom Probe
Tomography
Dense FinFET modules & Process Integration
Performance Elements
(SiGe, Si:P. Si:C)
Replacement
Metal Gate stack
& WF Engineering
62nm pitch
FinFET Variability & Reliability
Device Design, Self-Heating etc.
26
© IMEC 2014
FW
FH
1.0 1.5 2.0 2.5 3.0
Fin Height (FH) / Fin Width (FW)
60
65
70
75
80
85
90
Su
b-t
hre
sho
ld S
win
g (
mV
/dec)
LIMITATIONS OF FINFET ELECTROSTATICS
NEW STRUCTURAL ENHANCEMENTS
N10:
Lg=20nm
FW =7nm
Strong
Electrostatics
C. Auth VLSI2012
(22nm FinFET)
• Simple Si FinFET are hitting limits to continual Lg scaling for N7 and below.
N7:
Lg=14nm
FW =7nm
Poor
Electrostatics
• New FinFET “Structural Enhancements” available to extend scaling
QW
QW Fin
Gate-All-
Around Fin
EXTENDING MULTI-GATE STRUCTURES
Fin 2 wires 3 wires
STI
Fin
30n
m
NW
spacing
5nm
SiO2
0.5nm
HfO2
1.5nm Rounding
radius
2.5nm
Fin
10n
m
10n
m
5n
m
5n
m
5n
m
30n
m
30n
m
•New studies in progress to understand the trade-off between electrostatics,
Resistance & current efficiency
1 wire
25n
m
30n
m
Spacing
5nm
STI STI STI
Decrease Sswing & DIBL
Increase Rseries
Fully-depleted
Channel for Improved
Electrostatics
Multi-gate FETs
Band-Engineered
Channel for
Enhanced Transport
High-Mobility
Channels
(SiGe, Ge IIIV)
Novel Materials/
New Transport/
Extreme
Electrostatics
2D Materials
(Bi-layer
Graphene)
Quantum/
Spin Devices
• Si-Channel Scaling & Beyond-Si Channel Scaling
Vdd 1.0-1.1V 0.9-1.0V 0.8-0.9V 0.7-0.8V 0.6-0.7V 0.5-0.6V < 0.5V
Strain &
Advanced Gate Stack
Engineering
SD/stressors Metal Gate
+High-k
MGHigh-k
Si substrate
MGHigh-k
Si substrate
45nm
32/28nm
22/20nm
14nm
10nm
7nm
5nm
Tech Node ...
• 2 Scaling paths for FinFETs down to 7nm
IMEC LOGIC DEVICES R&D ACTIVITIES
• “Beyond Si”
•New Materials for FinFETs
& other non-planar FETs
BEYOND SI FINFET:
IIIV/GE HETERO-FIN PROCESS @ IMEC
Epitaxial Defect Trapping &
Replacement Fin Process
Si
Ge
IIIV & Ge Gate stack & Channel Passivation
IIIV-IV Heterogeneous Material
Integration
Device integration
(RMG, SD Epi, Contact)
III-V/Ge FinFET Design
• Heterogeneous integration: Activities spanning new materials, integration ,
Device design. Fundamentals to Engineering. ..Lab to Fab.
ADV. METROLOGY : IN-LINE FAB –TO- LAB-SCALE
1E+15
1E+16
1E+17
1E+18
1E+19
1E+20
10 100 1000 10000
Co
nce
ntr
ati
on
(/c
m3
)
Mobility (cm2/Vs)
Scilab fitted spectrum
InGaAs HEMT
In
P
Ge
Si10 nm
10 nm
Thermal wave probe
Defects in Nanostructures
1 2
3 4
Defect detection &
review (Pat.Wafer) Crystal quality Gate stack
Atom Probe
Tomography
(Atomic) SSRM
QMSA( Nano-scale)
InP InGaAs
Second Harmonic Gen.
SiO2
IG
A/
Ga
As
SiO2
36°
IG
A
/G
a
1E+00
1E+03
1E+06
-40 10 60 110 160
Inte
nsi
ty
(Co
un
ts/s
)
Time (min)
Nano-Trench SIMS
• Adv. materials & structures activities drove new metrology R&D
GE OR SIGE PFET BENCHMARK
imec
sSiGe Planar
(2011)
imec
sGe QW FinFET (STI last)
(Unpublished 2013)
imec
sGe QW FinFET (ART)
(2013 To be published)
IBM
sSiGe on Insulator
VLSI 2013
[IEDM’10 INTEL]
sGe/SiGe (75%)
Planar
TSMC
Ge Bulk FinFET (ART)
(IEDM 2012)
• Significant improvement achieved in strained strained-Ge QW FinFET
TSMC Ge Bulk FinFET (ART)
(IEDM 2013)
III-V NFET BENCHMARK
imec (lab) (Unpublished, 2013)
InGaAs planar
Intel IEDM 2009
InGaAs planar
imec (300mm Fab)
InGaAs FinFET
Kim VLSI 2011
InAs channel
Egard IEDM 2011
InGaAs planar
Egard IEDM 2011
InGaAs planar
Ko, Nature2010
InAs-O-I
Lee, VLSI 2013
InAs channel
• We achieved one of the higher performance InGaAs planar devices (IIIV Lab)
Reported strained-
Si ~1300 mS/mm
InAs channel TSMC IEDM 2013
Fully-depleted
Channel for Improved
Electrostatics
Multi-gate FETs Ultra-Thin
SOI
Band-Engineered
Channel for
Enhanced Transport
Gate-All-
Around,
Nanowires/
Tunnel FETs
Novel Materials/
New Transport/
Extreme
Electrostatics
2D Materials
(Bi-layer
Graphene)
Quantum/
Spin Devices
• Material & Device Architecture Innovations Enablers of continual scaling
Vdd 1.0-1.1V 0.9-1.0V 0.8-0.9V 0.7-0.8V 0.6-0.7V 0.5-0.6V < 0.5V
Strain &
Advanced Gate Stack
Engineering
SD/stressors Metal Gate
+High-k
MGHigh-k
Si substrate
MGHigh-k
Si substrate
45nm
32/28nm
22/20nm
14nm
10nm
7nm
5nm
Tech Node ...
• Feature Dimension & Voltage Scaling are concurrent drivers
IMEC LOGIC DEVICES R&D ACTIVITIES
• Beyond FinFETs
• New Electrostatics & Transport
34
© IMEC 2014
FinFET GAA
No Room for
Lateral
Contacted Pitch
Vertical?
?
• Eventually disruptive architectures like Vertical NWs can extend density scaling
20 10 7 5 3
10
40
60
90 P
hysi
cal D
imen
sio
n (
nm
)
CMOS Technology Node (nm)
• Continual gate pitch (density) scaling will be limited by space for Contact & Gate
LIMITS TO DENSITY/LGATE SCALING
N-well P-well
P+ junction N+ junction
Gate overlap
Lg
-SiMetal
High-k
c-SiGe(P+)
i-Si
c-III/V (N+)
Si shell Tunnel diode rooting
N-well P-well
P+ junction N+ junction
Gate overlap
Lg
Gate overlap
Lg
-SiMetal
High-k
c-SiGe(P+)
i-Si
c-III/V (N+)
Si shell Tunnel diode rooting
H -Vs.- V Layout Efficiency
Horizontal
Vertical
Channel electrostatics & Device Design
Channel Material – Si
Vs. Ge Vs. III-V
Electrostatics
- Quantum simulations
V-NWFET Circuit Layout & Design
Directed InAs NW on
Si(111)
TFET Physics
Band Engineering for tunneling
NANOWIRE / TUNNEL FET
VERTICAL NANOWIRES PROCESS & CIRCUITS
Parasitic Capacitances Layout & Parasitic
Analysis
Selective
area
growth
{1-10}
(111)
SiO2
InAs
Si (111)
Si (111)
InAs Nanowires
on Silicon
Template Formation D~30-40nm D~100’s nm
PMOS
NMOS
NW pitch
Lateral
Conventional S
RA
M
Vertical
NW
NA
ND
5NM & BELOW VERT-NW
PHYSICAL DESIGN STUDY
0
0.01
0.02
0.03
0.04
0.05
0.06
16.91513.411.9
Are
a (m
m2)
M1 1/2-pitch (nm)
NAND with VFET for N5 SRAM with VFET for N5
NAND(ITRS)
SRAM (ITRS)
NW
7nm 5nm
• Vertical Nanowire Circuits can
promise further logic density
scaling for 5nm & below
38
© IMEC 2014
VARIABILITY: “NATURE VS. NURTURE”
(Various collected sources)
DIFFERENT VARIABILITY = DIFFERENT SOLUTIONS
DIE-DIE,
WAFER-
WAFER
ACV,
LAYOUT, CD
OVERLAY
RANDOM
DOPANTS,
LER, LWR
WEAR-OUT
RELIABILITY
TDDB,
NBTI,PBTI
ELECTROMIGRATION Eg. EM ~ 1ns, HCI ~ 10ns,
BTI ~1us
RANDOM TELEGRAPH /
FLICKER NOISE
SUPPLY
VOLTAGE
& PACKAGE
NOISE
TEMPERA-
TURE
VARIATION Self-Heating ~ 1e-4s
THERMAL
NOISE
PLL
JITTER
INTER-
CONNECT
CROSS-TALK
NOISE
Time Scale 100GHz-100MHz 1MHz-100kHz 10kHz-100Hz
Days/Months/Yrs 10ps-10ns Static 100ms-10ms 100ns-10ms
<1
0n
m
100n
m-1m
m
10m
m-1
00m
m
Tra
nsis
tor
Fe
atu
res Inte
rcon
ne
cts
&
Sm
all
Ckts
C
om
ple
x C
kt B
locks
Ch
ip
Featu
re S
cale
(Various collected sources)
DIFFERENT VARIABILITY = DIFFERENT SOLUTIONS
DIE-DIE,
WAFER-
WAFER
ACV,
LAYOUT, CD
OVERLAY
RANDOM
DOPANTS,
LER, LWR
WEAR-OUT
RELIABILITY
TDDB,
NBTI,PBTI
ELECTROMIGRATION Eg. EM ~ 1ns, HCI ~ 10ns,
BTI ~1us
RANDOM TELEGRAPH /
FLICKER NOISE
SUPPLY
VOLTAGE
& PACKAGE
NOISE
TEMPERA-
TURE
VARIATION Self-Heating ~ 1e-4s
THERMAL
NOISE
PLL
JITTER
INTER-
CONNECT
CROSS-TALK
NOISE
Fast/Transistor-level Variations
Materials & Device Process
Solution
Slow Die/Wafer-Level Variations
Design/Layout Optimizations
&
Smart System Adaptation
<1
0n
m
100n
m-1m
m
10m
m-1
00m
m
Tra
nsis
tor
Fe
atu
res Inte
rcon
ne
cts
&
Sm
all
Ckts
C
om
ple
x C
kt B
locks
Ch
ip
Featu
re S
cale
Time Scale 100GHz-100MHz 1MHz-100kHz 10kHz-100Hz
Days/Months/Yrs 10ps-10ns Static 100ms-10ms 100ns-10ms
(Various collected sources)
DIFFERENT VARIABILITY = DIFFERENT SOLUTIONS
DIE-DIE,
WAFER-
WAFER
ACV,
LAYOUT, CD
OVERLAY
RANDOM
DOPANTS,
LER, LWR
WEAR-OUT
RELIABILITY
TDDB,
NBTI,PBTI
ELECTROMIGRATION Eg. EM ~ 1ns, HCI ~ 10ns,
BTI ~1us
RANDOM TELEGRAPH /
FLICKER NOISE
SUPPLY
VOLTAGE
& PACKAGE
NOISE
TEMPERA-
TURE
VARIATION Self-Heating ~ 1e-4s
THERMAL
NOISE
PLL
JITTER
INTER-
CONNECT
CROSS-TALK
NOISE
<1
0n
m
100n
m-1m
m
10m
m-1
00m
m
Tra
nsis
tor
Fe
atu
res Inte
rcon
ne
cts
&
Sm
all
Ckts
C
om
ple
x C
kt B
locks
Ch
ip
Featu
re S
cale
Time Scale 100GHz-100MHz 1MHz-100kHz 10kHz-100Hz
Days/Months/Yrs 10ps-10ns Static 100ms-10ms 100ns-10ms
Fast/Transistor-level Variations
Materials & Device Process
Solution
Slow Die/Wafer-Level Variations
Design/Layout Optimizations
&
Smart System Adaptation
• Pattern Loading Effect: Layout control (RDRs) & Process Optimization
N10 (10nm) Layout -Dependent
Fin Variation
(Various collected sources)
DIFFERENT VARIABILITY = DIFFERENT SOLUTIONS
DIE-DIE,
WAFER-
WAFER
ACV,
LAYOUT, CD
OVERLAY
RANDOM
DOPANTS,
LER, LWR
WEAR-OUT
RELIABILITY
TDDB,
NBTI,PBTI
ELECTROMIGRATION Eg. EM ~ 1ns, HCI ~ 10ns,
BTI ~1us
RANDOM TELEGRAPH /
FLICKER NOISE
SUPPLY
VOLTAGE
& PACKAGE
NOISE
TEMPERA-
TURE
VARIATION Self-Heating ~ 1e-4s
THERMAL
NOISE
PLL
JITTER
INTER-
CONNECT
CROSS-TALK
NOISE
Fast/Transistor-level Variations
Materials & Device Process
Solution
Slow Die/Wafer-Level Variations
Design/Layout Optimizations
&
Smart System Adaptation
<1
0n
m
100n
m-1m
m
10m
m-1
00m
m
Tra
nsis
tor
Fe
atu
res Inte
rcon
ne
cts
&
Sm
all
Ckts
C
om
ple
x C
kt B
locks
Ch
ip
Featu
re S
cale
Time Scale 100GHz-100MHz 1MHz-100kHz 10kHz-100Hz
Days/Months/Yrs 10ps-10ns Static 100ms-10ms 100ns-10ms
(Various collected sources)
DIFFERENT VARIABILITY = DIFFERENT SOLUTIONS
DIE-DIE,
WAFER-
WAFER
ACV,
LAYOUT, CD
OVERLAY
RANDOM
DOPANTS,
LER, LWR
WEAR-OUT
RELIABILITY
TDDB,
NBTI,PBTI
ELECTROMIGRATION Eg. EM ~ 1ns, HCI ~ 10ns,
BTI ~1us
RANDOM TELEGRAPH /
FLICKER NOISE
SUPPLY
VOLTAGE
& PACKAGE
NOISE
TEMPERA-
TURE
VARIATION Self-Heating ~ 1e-4s
THERMAL
NOISE
PLL
JITTER
INTER-
CONNECT
CROSS-TALK
NOISE
<1
0n
m
100n
m-1m
m
10m
m-1
00m
m
Tra
nsis
tor
Fe
atu
res Inte
rcon
ne
cts
&
Sm
all
Ckts
C
om
ple
x C
kt B
locks
Ch
ip
Featu
re S
cale
Time Scale 100GHz-100MHz 1MHz-100kHz 10kHz-100Hz
Days/Months/Yrs 10ps-10ns Static 100ms-10ms 100ns-10ms
Slow Die/Wafer-Level Variations
Design/Layout Optimizations
&
Smart System Adaptation
• Device Structure and Materials are important factors
Fast/Transistor-level Variations
Materials & Device Process
Solution
Bulk FinFET Vs. SOI Vs. Planar
Vt Mismatch sensitive to device
architecture & design
Vt fluctuations due to Traps/defects – Fundamental link to
BTI – Need to engineer Trap-Charge Kinetics
EOT (nm)
Vo
ltage O
verd
rive @
10 Y
ears
(V
)
RELIABILITY BENCHMARK
NBTI
Planar Si
channel
+ HKMG
FinFET NBTI
worsens due
to (110) surface
(110)
40% higher
atomic density
• Fundamental material properties influences reliability
• Quantum Well to engineer the Trap-Charge Kinetics
SiGe QW
planar
SiGe QW
FinFET(N)
SiGe
Si
EMERGING NOISE LIMITERS AT LOW VDD
• RTN Impact is enhanced under linear operation and worsens with
lowering Vdd
• Expects to be major impact on Near-VT operations
• Need to engineer trap-carrier interactions (Eg. QW devices)
1.E-03
1.E-02
1.E-01
1.E+00
0 0.05 0.1
1-C
DF
-DVth (V)
-0.5V -0.2V -0.1V 0V
VG-Vth0 =
Broaden distribution
with decreasing |VG - VT|
Measurement
RTN-influenced VT shift
Fully-depleted
Channel for Improved
Electrostatics
Multi-gate FETs Ultra-Thin
SOI
Band-Engineered
Channel for
Enhanced Transport
Gate-All-
Around,
Nanowires/
Tunnel FETs
Novel Materials/
New Transport/
Extreme
Electrostatics
2D Materials
(Bi-layer
Graphene)
Quantum/
Spin Devices
• Material & Device Architecture Innovations Enablers of continual scaling
Vdd 1.0-1.1V 0.9-1.0V 0.8-0.9V 0.7-0.8V 0.6-0.7V 0.5-0.6V < 0.5V
Strain &
Advanced Gate Stack
Engineering
SD/stressors Metal Gate
+High-k
MGHigh-k
Si substrate
MGHigh-k
Si substrate
45nm
32/28nm
22/20nm
14nm
10nm
7nm
5nm
Tech Node ...
• Feature Dimension & Voltage Scaling are concurrent drivers
IMEC LOGIC DEVICES R&D ACTIVITIES
• “Beyond CMOS”
• Charge Vs. Non-Charge Based
• Novel Materials (E.g. Graphene, MX2)
• Fundamental Investigation (Material + Components)
ENERGY EFFICIENCY BENCHMARK
Graphene Logic
Ultimate
Logic
Benchmark: Praveen Raghavan (Logic-Insite)
After Nikonov & Young (Intel 2012)
InAs TFET
• Active investigations on (selected) New competitive devices
All Spin Logic
MTJ Logic
Spin Wave
Piezoelectric Element Transistor
• Non-charge based devices being surveyed
CMOSFET continues to
more efficient with
scaling
2-D MATERIAL DEVICE + INTERCONNECT
Graphene
electrodes
MX2/Multi-layer
Graphene Channels
BLG Tunnel FET Simulations
Lab-Fab
Processing
Gate
Contact
F2030
Contact Resistance Molecular Doping
MX2 MOSFET Experiments
2-D MATERIAL DEVICES
• Demonstrated electrical devices with Molybdenum disulfide (MoS2) nFET &
Tungsten Di-selenide (WSe2) pFET
Si (Back Gate)
SiO2
W(S) W(D)
MX2
WSe2
(Pd)
MoS2
(Ti)
(To be published)
MoS2
Early fundamental work on Flakes
50
© IMEC 2014
CONCLUSION
51
© IMEC 2014
1 µm 0.8 µm 0.5 µm 0.35 µm 250 180 130 90 65 45 32 22 14 10 7 5
imec founded
What ’s next ?
SOI
FinFET
Bulk FinFET
FUSI MG
HKMG RMG-HK
SiGe Epi rSi Epi
Ge
MOSFET
GaAs on Ge
Si Epi GAA-SOI
PD/FD-SOI TFET
SiGe pFET
SiGe channel
Buried Spacer-defined Fins
Vertical SiGe FET
today
IMEC: INNOVATIVE PROCESS OPTIONS
Semiconductor Industry Technology nodes
52
© IMEC 2014
1 µm 0.8 µm 0.5 µm 0.35 µm 250 180 130 90 65 45 32 22 14 10 7 5
imec founded
SOI
FinFET
Bulk FinFET
FUSI MG
HKMG RMG-HK
SiGe Epi rSi Epi
Ge
MOSFET
GaAs on Ge
Si Epi GAA-SOI
PD/FD-SOI TFET
SiGe pFET
SiGe channel
Buried Spacer-defined Fins
Vertical SiGe FET
today
IMEC: INNOVATIVE PROCESS OPTIONS
Semiconductor Industry Technology nodes
Vertical circuit
Arch.
Spin for Logic
Heterogeneous
FinFET
Heterogeneous
Nanowires
53
© IMEC 2014
MESSAGES
• Innovative solutions may be disruptive – early R&D needed to
build the infrastructure
• New drivers & challenges continue to demand Innovative
options at all levels – Material, Tools, Device, Circuits, & Systems
• IMEC continues to Explore the Solutions
• Co-operation across disciplines essential
• System & Process Specializations key for CMOS scaling in
the next decade to meet Power, Performance, Area, Cost needs.