More Moore or More Than Moore? - Faculdade de …moraes/prototip/artigos/moreMoore.pdf · To get...
Transcript of More Moore or More Than Moore? - Faculdade de …moraes/prototip/artigos/moreMoore.pdf · To get...
Accelerating the next technology revolution
Copyright ©2010 SEMATECH, Inc. SEMATECH, and the SEMATECH logo are registered servicemarks of SEMATECH, Inc. International SEMATECH Manufacturing Initiative, ISMI, Advanced Materials Research Center and AMRC are servicemarks of SEMATECH, Inc. All other servicemarks and trademarks are the property of their respective owners.
More Moore or More Than Moore?
Raj Jammy PhDVP, Materials and Emerging TechnologiesSEMATECH
SEMATECH Symposium TaiwanSeptember 7, 2010
Contents
•
Overview
•
Industry trends
•
CMOS (More Moore) options
•
Beyond CMOS (More than Moore) options
•
Conclusions
2
Overview•
More Moore: Scaling as per Moore’s Law–
Double number of transistors every ~two years–
Challenges: •
Unsustainable power dissipation•
Atomic scale dimensions-
physical limits•
Process and device variability•
No real performance increase with scaling•
Expensive R&D and manufacturing costs
•
More than Moore: Application driven components–
Analog, RF, M(N)EMS, sensors, opto, power devices,..–
Challenges:•
Bulky•
Incompatible with CMOS•
Often unsuitable for mobile platforms
3
• To get More from Moore, More than Moore is a must!
How is the industry changing? 1) Mobile trend –
performance AND power critical
• Mobile products have increased share of the Semiconductor TAM• IMPACT: Need low power device, sensing(analog), storage, display solutions
4
How is the industry changing? 2) Data center trend; performance AND power critical
http://www.emerson.com/neverbeendone/EnergyLogic.aspx
Dat
a ce
nter
s in
the
Uni
ted
Sta
tes
alon
e so
aked
up
abou
t 61
billi
on k
ilow
att-h
ours
(k
Wh)
, or $
4.5
billi
on w
orth
of e
lect
ricity
in 2
006.
If c
urre
nt e
nerg
y us
age
trend
s co
ntin
ue, a
U.S
. Env
ironm
enta
l Pro
tect
ion
Age
ncy
pred
icts
U.S
. dat
a ce
nter
s w
ill u
se
mor
e th
an 1
00 b
illio
n kW
h by
201
1, re
pres
entin
g $7
.4 b
illio
n in
ann
ual e
lect
ricity
cos
ts
and
2.5
perc
ent o
f the
nat
ion’
s to
tal e
lect
ricity
.
0
20
40
60
80
100
120
140
Power
Use
(Billion kW
‐hr)
Chem
ical
Meta
lDa
ta Ce
nterFo
odPa
per
Auto
Plasti
cs
Source: EIA, NREL May 2009
2011$7.4B Power Bill
• Cloud computing growing –
logic & memory for data centers important!• IMPACT: Dedicated power stations for data center ops and cooling (?)• IMPACT: Data centers need Both
Low Power & High Performance
Cloud Computing Growth
2006 Data
5
• SOC requires combining digital, analog, passives and lots of memory•
IMPACT: Mainstream chips are no longer restricted to discrete memory or logic• IMPACT: Memory needs to be low power too
Sou
rce:
Mar
k B
ohr,
Inte
l Dev
elop
er F
orum
200
9. a
vaila
ble
on w
ww
.inte
l.com
Source: Y. Nishi, S.M. Sze, EE319 Lecture Stanford 2009
6
How is the industry changing? 3) SOC trend; performance AND power critical
• Most “digital”
consumer systems have significant • Analog and Communication components• Memory
• Significant need for low power operation
SenseCompute
StoreTransmit
7
How is the industry changing? 4) It is not just a digital world!; performance AND power critical
Industry “must solve”
challenges
TODAY GOAL TODAY GOAL
Energy
(per OP or bit)1000pJ 10pJ 2pJ 0.06pJ
Power Cost $1B $1M $1B $1M
Enterprise Logic Enterprise Memory
Source: DARPA and Kryder, et. al. IEEE TRANS ON MAGNETICS, VOL. 45,, OCT 09
New materials, processes, tools, devices for next generation low power
technologies
Chip makers/design houses will drive multi-core designs,
circuit power optimization
• Must Solve: Power-Performance Trend
• Must Solve: ExaFLOP
/ Terabit Power, $$$
Adapted from: Y. Nishi, S.M. Sze, EE319 Stanford 2009
8
Common theme
• GOAL: Device with low power AND high performance • Avoid cost, complexity of dozens of devices on a chip
Universal, Single Device?
9
How to address “must solve”
challenges? Pareto analysis of power issue
Sour
ce: N
EC (w
ww. N
ec.c
o.jp
) and
T. B
. Hoo
k et
al I
EDM
p 2
7.6
2006
#3: Gate LeakageHK/MG
#2: Subthreshold
LeakageJunctions, damage, TFET
20%
30%
10%
40%
Best Efficiency Vdd
~0.3V
For Vdd
Scaling, Transistors Need:• Good mobility [strain , Ge, III-V, etc…]• Robust electrostatics [scaled HK/MG, USJ, finFETs, nanowire
FETs, …]• Controlled stochastics
/variability [tsi
in SOI, low Rcon
materials , fin LER …]
Today
~1V#1: Dynamic “leakage”Vdd
Scaling (Vdd=0.3-0.6V)
10
Power performance trade-offs More Moore opportunities
12
Drive
Curr
ent
I d,s
at
Gate Voltage
On state currentMaximize for performance
Off-state currentMinimize for power
11
•
High Mobility materials•
SiGe, Ge, InGaAs
•
Better electrostatic control•
Multiple gates + more channel area
•
FinFETs, nanowire
FET•
Low Tinv: Scaled HKMG•
USJ•
Parasitics
(Rco, Cp..)
Gate stack plays critical role in roadmap
HK
MG MG
GeSi
HK
MG
Si
MG
InGaAs
•
Zero Interface Layer HK reduces Jg
by 107X•
2nd and 3rd generation HKMG will also require ZIL
0
20
40
60
80
100
120
140
2000 2002 2004 2006 2008 2010 2012 2014 2016Year of Introduction by Leading Edge Companies
Tech
nolo
gy S
calin
g
SiON/poly-Si
300 mm200 mm
1st Gen Hi-k/MGSiON/poly-Si LP
130nm
90nm
65nm
45nm32nm
22nm16nm2nd Gen Hi-k/metal
Multi-gate FETs
HOT, FDSOI, Novel Materials
12nm3rd Gen Hi-k/metalOn high channels
Tunable Vt
SEMATECH Confidential0
20
40
60
80
100
120
140
2000 2002 2004 2006 2008 2010 2012 2014 2016Year of Introduction by Leading Edge Companies
Tech
nolo
gy S
calin
g
SiON/poly-Si
300 mm200 mm
1st Gen Hi-k/MGSiON/poly-Si LP
130nm
90nm
65nm
45nm32nm
22nm16nm2nd Gen Hi-k/metal
Multi-gate FETs
HOT, FDSOI, Novel Materials
12nm3rd Gen Hi-k/metalOn high channels
Tunable Vt
SEMATECH Confidential
Zero InterfaceInterface
12
0.0 0.2 0.4 0.6 0.8 1.0 1.210-4
10-2
100
102
104
106
108
EOTIL=0.8nmSiOx/HfSiON
J g@V fb
-1 (-
A/c
m2 )
EOT (nm)
SiO2/poly-Si
ZIL/HfO2
EOTIL=0.45nmSiON/HfSiON
16nmHP
ZrO2
1020 C-1070°C spike
12nm HP
Ultra shallow junctions Scaled USJ: conventional and damage-free
•Standard implantation•Flash anneal•Rs~1050-1100/sq
Xj
~7.5nm
•“Monolayer”
doping•Low Rs
& damage free•Suitable for group IV & III-V 4 nmSiO2
InAs
S doped region
50 nm
Epoxy
SiO2
InAs
[110] Zone
111220
(a) (b)
13
Contacts: Key scaling challenge Schottky
Barrier height engineering to lower Rc
•
600 –
800 meV
tuning of Schottky
Barrier Height•
N and P-Si contact engineering for conventional Ni based dual silicides•
Potential dual silicide
solution for nodes well beyond 16nm•
Simple and manufacturable
approaches –
help low power operation
10 100
0
20
40
60
80
100
Cum
ulat
ive
Freq
uenc
y
Rcont()
W/TiN/Ti on NiSi
ControlNiSi
N-implantNiSi
Interface Engineered NiSi
NiSix
based
Target
Interface Engineering Co-
implantation
Dipole mediated
Rc
Reduction in NiSi: Three Methods
0 5 10 15 20 25-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
Reverse Bias IV Measurement Depletion CV Measurement
B(e
V)
tHigh- (A)
Si Bandgap
B.Coss
et al VLSI 2009 5BWY Loh
et al VLSI 2009 5B
14
15
•
24% enhancement in drive current for doped NiSi
DSS-nFETs
0.0 0.2 0.4 0.6 0.8 1.00
-100
-200
-300
-400
-500
-600
-700 Control N(3.5E15)
I s (
A/
m)
Vds
(V)
Vg-V
T = 1.0VDSS nFET 20%
B.Coss
et al VLSI 2009 5BWY Loh
et al VLSI 2009 5B
Transistor performance improvement Lower Rext
16
LDD
Deep Source/Drain
Silicide
IL
High‐K
MG
(Si)Ge
Si
High mobility channels: Quantum-well MOSFETs Short channel Planar Si/SiGe/Si QW pFETs
•
Best reported short Lg
pFET
(Ion/Ioff
~ 5X104
!) devices with high Ge%•
Defect-free thin SiGe, Ge
PMOS for high performance application •
Better than state-of-art Si on power / performance metric •
No uniaxial
strain in this work –
further boost possible•
Demonstrated concept in scaled non-planar scheme as well as in III-V SH
Lee
et a
l VLS
I 200
9JW
Oh
et a
l VLS
I 200
9
(Si)Ge
QW channelLow defect thin filmSelective
~3xSi
Value of III-V Performance @ Power
III-V demonstrated as superior to state-
of-art Si in :- Fundamental Energy to Switch - Energy Speed Benefits- Energy*Delay Benefit w/ scaling- Many issues to be resolved
17
S. Datta
[PennState]
Property/Material
Si Ge GaAs In0.53
Ga0.47
As InAs Graphene
Eg
(eV) 1.1 0.66 1.4 0.75 0.35 0*n
(cm2/v-sec) 1,350 3,900 4,600 >8,000 40,000 >100,000p
(cm2/v-sec) 480 1,900 500 350 <500 >100,000m*/mo 0.165 0.12 0.067 0.041 0.024 <0.01Lattice mismatch to Si
0 4% 4% 8% 12% n.A
Properties of promising high mobility materials [electrons and holes]
pMOSFET nMOSFET
2
,
)(2 tdd
invox
odsat VV
LW
TAI
* Tunable but at cost of degradation
18
III-V directly on Si
•
1st ever demonstrated flow for III-V on Si FETs
using industry standard tool set•
Systematic materials evaluation/process development for industry infrastructure
III-V on 200 mm Si wafer
III-V on 200 mm Si waferFully processed transistor flow
19
III-V/Si for VLSI using 200 mm line
20
SOI2 Gates
Tall Fins
Stacked wires
High Mobility Channel
Heterogeneous Fin Channel
High Mobility Channel
Platform
Heterogeneous Wire Channel
Single wire
3 Gates
Bulk
SOI2 Gates
Tall Fins
Stacked wires
High Mobility Channel
Heterogeneous Fin Channel
High Mobility Channel
Platform
Heterogeneous Wire Channel
Single wire
3 Gates
Bulk
FinF
ETs
Nan
owire
sW
afer
s
0 .0 0 .2 0 .4 0 .6 0 .8 1 .0 1 .20 .0
0 .2
0 .4
0 .6
0 .8
1 .0
1 .2
V(s)
[V]
V (S B ) [V ]
SNM of 300mV-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
1E-71E-61E-51E-41E-30.010.1
110
1001000
10000
Lg~30nmVTH=-0.52VDIBL=60mV/VSS=67mV/decEOT~1nm
ID (
A/
m)
Vg (V)
open: Vds=|1V|closed: Vds=|0.05V|
Lg~30nmVTH=0.26VDIBL=49mV/VSS=70mV/decEOT~1nm
• 6-8 nm Fins
SEMATECH Si Nanowire
SiGeSi
SRAM
Nanowire
FETs
Dual Channel Fin FETFin FETs
Non-planar devices: FinFETS
and nanowires
21
Non-planar device scaling
•
Model based understanding using simulation + experimental data.
High w and w/o 3rd
gate ?
High Bulk vs
SOI
OR
OR
Scaling Pathways
Hom
gene
ous
Het
erog
eneo
us
Nanowire
10nm*10nmTrigate
40nm*10nm
InGaAs
QW Volume Inversion
45%
FinFET
with Heterogenous
Channel
FinFETs
with SiGe
and SiGe/Si Fins
22
20 nm20 nm
{110
}{100}
20 nm20 nm
{100
}
{100}
BOX
MGHK
BOX
Img: 'E1918 080622001 6081002 22 L86 SiGe FINS s dm3' MAG: 115kX
20 nm20 nm
{110
}
{100}
Img: 'E2092 080625015 6081002 22 2ndTry Fins SiGe 45deg s dm3' MAG: 115kX
20 nm20 nm
{100
}
{100}
SiGe SiGe
SiGeSi
SiGeSi
(110)
<110>
(110) fin sidewall surface.
Current in <110> direction.
(100)
<100>
(100) fin sidewall surface.
Current in <100> direction.
Si nanowire
arrays/stacks
450 nm
450 nmsource
drain
suspended wires
Single Si NW FETs
23
Stacked NWs
Si Nanowire
SiGeSi Si/SiGe
High
Nanowire
24
S. Datta
Voltage scaling
Pathways for low power CMOS & beyond…
Contact throughILD
Spacer
Si fin (~10nm
wide)
Img: ' E0985 080516013_ U T_ Tr ansi s tor _ 200_ TaN _ Zr O_InGaAs.Ed_r ot_ s.dm3' M AG: 590kX
5 nm5 nm5 nm
In0.53
GaAs
T
i
NZrO2
Gate-First nMOSFET700 0C anneal
Img: 'E2092 080625015_6081002_22 2ndTry_Fins SiGe 45deg_s.dm3' MAG: 115kX
20 nm20 nm
Si
SiGe
HKMG
10 nm10 nm
Gate
DrainSourceDamage-freeJunctions
High Channel Materials
Channel Structures
Low R Contacts
Disruptive Gate Stacks andDual WF Metal Gates for LP/HP
Numerous Options …breakthroughs needed, but exciting possibilities
HfOx
Si
TaNGraphene
Gate
N+ Drain P+ Source
BOX
TFETs
Metal
Si
Driving Electrode A
Driving Electrode B
SensingElectrode B
SensingElectrode A
Width = 90nmGap = <100nm
NEMS
25
Power performance trade-offs More than Moore opportunities
•
High Mobility materials•
SiGe, Ge, InGaAs•
Graphene
[e
~15000 cm2/V-s at RT]
•
Better electrostatic control•
Multiple gates + more channel area•
FinFETs, nanowire
FET
•
Improve on-off ratio •
Tunnel FET•
Very steep ΔSS
<< 60 mV/dec•
Low bias voltages (<< 1V) •
Nano
Electro Mechanical switch (NEMS)
•
Hybrid: Ion
by CMOS + Ioff
by NEMS
•
Zero Leakage Power
12
3Drive
Curr
ent
I d,s
at
Gate Voltage
On state currentMaximize for performance
Off-state currentMinimize for power
26
Schottky-source tunneling FET Sub-60mV/dec subthreshold
swing Conventional Si CMOS limit: 62mV/dec
•
New device concept: Schottky
Source TFET with SS ~ 46mV/dec
With UC Berkeley
NEMS: Nano
Electro-Mechanical Switch
MEMS NEMS GainFrequency kHz GHz Ultra low power
Power mWW pW Faster operationQuality factor 107 104 Challenge
Sensitivity High Highest High sensitivity
Dadgour
et al, Session 1.1.1, ISLPED 2010
M. Hussain
et al, NATO Nano
DSA, 2009
Gate
Cantilever
28
MEMS/NEMS “zero leakage”
devices Disruptive memory/logic switch
[1] W
. Y. C
hoi,
and
T.-J
. Kin
g Li
u IE
DM
p. 6
03 (2
007)
.[2
] A. D
riski
ll-S
mith
, Y. H
uaiF
utur
e Fa
bp.
28
(200
7).
SEMATECH sub 100nm NEMS
0 10 20 30 40 50100
101
102
103
104
105
DRAM
PCRAMFeRAM
NEMS SRAMMRAM
NOR
low power mid power hi power
Log
Spee
d (r
ead+
writ
e) [n
s]
Density AF2
NAND
SSTRAM
• NEMS may get to 6F2 [1,2]
Why NEMS Memory?
Drive A
Drive B DrainB
BeamDrain
A
Source
0 2 4 6 80.00
0.04
0.08
0.12
Cur
rent
(A
)
Voltage (V)
Vpi = 4.25V
Hysteresiswindow ~5V
gap g width w
sacrificial thicknessdevice thickness t
29
NEMS NANDflo
atin
g e
lect
rode
out = AB
A
VDD
VDD
GN
D
B
out = 1
A = 1
VDD
VDD
GN
D
B = 0
out = 0
A = 1
VDD
VDD
GN
D
B = 1
Bef
ore
initi
aliz
atio
n
VGS
1
32
0VpiVpo
VDD
0
Once the beam is pulled-in, VDD
> Vpi
is no more needed to hold the beam, VDD
> Vpo
is sufficientComparison for the same delay
(10.5 ns) and realistic dimensions
Time (ns)
Nor
mal
ized
Bea
m P
ositi
on
Conventional
Energy-Reversible
VDD = 2.24 V
VDD = 0.59 V
Drive A
Drive B Drain B
Beam
Drain A
Source
30
NEM switch fabrication
•
CMOS Compatible NEMS•
Can be integrated easily with conventional CMOS for logic/memory/sensing
Beam thickness = 25nm
Air gap = 60nm
Amorphous metalDriving
Electrode ADriving
Electrode B
SensingElectrode B
SensingElectrode ALength =
2~10m
Width = 90nmGap = <100nm
Source
Sacrificial layer
Cantilever
Metallic contact
0 1 0 2 0 3 0 4 0
-1 .0 x 1 0 -1
-5 .0 x 1 0 -2
0 .0
5 .0 x 1 0 -2
1 .0 x 1 0 -1
Cur
rent
(mA
)
V o lta g e (v o lts )
C o lle c tio n T e rm in a l C a n tile v e r
Z e ro Io ff
In fin ite Io n /Io ff
Lswitch = 2m Wswitch = 130nmGelectrode = 120nm Gcollection = 110nm
31
0 2 4 6 80.00
0.04
0.08
0.12
Cur
rent
(A
)
Voltage (V)
Vpi = 4.25V
Hysteresiswindow ~5V
NEMS-based logic circuitsInverter Gate
GN
DVD
D
XOR/XNOR Gates
Alternative implementations of XOR/XNOR gates
BAA B
BAA B
Gat
e A
Gat
e B
Pull-
up
NEM
SPu
ll-do
wn
NEM
S
VDD
GND
A B
VDD
GND
A B
XNOR (Design #2)XNOR (Design #1)
AND Gates
The implementation of a two-input AND gate
• NEMS relay outperforms conventional relay in terms of energy, voltage and reliability
VDrive
Vpull-in
t1 t2
Vhold
t3 t4
Vhold
t5 t6
Vhold
Hamed
Dadgour
et al, ISLPED 2010
100mV switching is possiblewith few nanometers gap
NEMS: Robust, CMOS compatible, nanoscale
switch
•
Potential on-the fly impedance matching for analog/mixed signal applications
•
Versatile device platform: the “transistor” of the MEMS world
Capacitance SensorDielectric coating
To power harvester
To transmitter
Temperature Sensor
To power harvester
Bimetallic strip
To transmitter
Resistance SensorMetallic coating
To power harvester
To transmitter
Pressure Sensor
To power harvester
To transmitter
tBeam
Substrate
Sacrificial Material
Beam
tGap
LBeamwBeam tBeam
Substrate
Sacrificial Material
Beam
tGap
LBeamwBeam
+ Well studied + Highly manufacturable
+ Extremely versatile + Scaling friendly higher sensitivity, lower cost, powerful arrays
Basic Building Block
CMOS
Analog
Z!NEMS NEMS
Example: in-situ process monitoring
33
34
-2 -1 0 1 24.0
4.5
5.0
5.5
6.0
6.5
7.0
7.5
40
45
Single ZrO2 (6 nm) Single SiO2 (6 nm) ZrO2(6 nm)/SiO2(3 nm)/ZrO2(6 nm)
Cap
acita
nce
Den
sity
(fF
/um
2 )
Voltage (V)
-6 -4 -2 0 2 4 610-11
10-9
10-7
10-5
10-3
10-1
101
Cur
rent
Den
sity
(A
/cm
2 )Voltage (V)
Single ZrO2 (6 nm) Single SiO2 (6 nm) ZrO2(6 nm)/SiO2(3 nm)/ZrO2(6 nm)
VCC=25ppm/V2
VCC=25ppm/V2SEMATECH Stack (6 nm)
SEMATECH Stack (6 nm)
Optimized low VCC MIM stack for Analog/MS
•
Meets ITRS specs to 2013 for analog and mixed signal needs•
Robust manufacturing ready process
Varactor Resistor Inductor CapacitorStructure
Issue(s)
•Linearity•Area scaling(cap/um2)
•Linearity•Heat Dissipation•FINFET, MG
•
Linearity•
Area Scaling(induct/µm2)
•
Capacitive coupling
•
Linearity•
Area scaling(cap/um2)
Approach
•Abrupt Junction•Barrier Height Reduction @ M-S contact•
Parasitic R
•Thin metal•Materials screening for TCR•
Parasitic R
•
3-D/ serpentine coils
•
Remove Si between coils
•
Parasitic R
•
SEMATECH solution to 2013
•
Higher k
SEMATECH Contribution
• Dipole tuning SBH•N Implant tuning SBH•
Damage free monolayer doping • Plasma doping •
Anneal: ms (flash, laser, spike
• Tune Rco
NiSi-Si• O, Si, N implant • TaN
optimization(C, N, O) for TCR• W system• I-V measurements
•
3-D test structure
•
Series Resistance
•
High freq capability to 40GHz
•
VCC<<100•
Vg<10-7
A/cm2
•
Cap den=7fF/um2
•
Work with tool supplier
•
Etch / etch stop development.
•
Conformal process
Analog mixed signal Passive components development
35
Moore’s Law & MoreMore than Moore: Diversification
Mor
e M
oore
: M
inia
turiz
atio
nM
ore
Moo
re:
Min
iatu
rizat
ion
Combining SoC and SiP: Higher Value SystemsBas
elin
e C
MO
S: C
PU, M
emor
y, L
ogic
BiochipsSensorsActuators
HVPowerAnalog/RF Passives
130nm
90nm
65nm
45nm
32nm
22nm...V
130nm
90nm
65nm
45nm
32nm
22nm...V
Information Processing
Digital contentSystem-on-chip
(SoC)
Interacting with people and environment
Non-digital contentSystem-in-package
(SiP)
Beyond CMOS
Traditional Models
[Geo
met
rical
& E
quiv
alen
t sca
ling]
Scal
ing
(Mor
e M
oore
)Functional Diversification (More than Moore)
HVPower Passives
SIP “White Paper”A&P ITWGwww.itrs.net/papers.html
Continuing SoC and SiP: Higher Value Systems
36
2009 ITRS: Overview of scaling trends
Logic and memory roadmap Technologies converge for higher value solutions
2009
Advanced Materials Advanced Structures
2010 2011 2012 2013 2014
Bulk/SOI Si Bulk/SOI SiHigh
(b)
r ans i stor _200_TaN _Zr O_InGaAs .Ed_r ot_s .dm3' M AG:
MetalHigh-k
InGaAs
SEMATECH Si Nanowire
HfOx
Si
50nm
WOx
TiN
W
TiN
W
WOx
TiN
W
TiN
W
CT Flash ReRAM1T DRAM
SiG
eSi
2009 2011 2013 2015 2017 2019
Logic
Memory
Beyond CMOS Materials/Structures
Driving Electrode A
Driving Electrode B
SensingElectrode B
SensingElectrode A
Width = 90nmGap = <100nm
Bit Line
Word Line
Common
Drain
Source
Gate Dielectric
GateElectrode
Bit Line
Word Line
Common
Drain
Source
Gate Dielectric
GateElectrode
STTRAM
ReRAM/Nanowire
3DArray
I d,s
at
Vg
ControlSTEEP
NEMS
TFET
Graphene
High mobilityFins/Nanowires
BoxGate-2 (substrate)
N+ N+
N+ Poly-SiGate Ox
Ge (SiGe)Box
Gate-2 (substrate)
N+ N+
N+ Poly-SiGate Ox
Ge (SiGe)Box
Gate-2 (substrate)
N+ N+
N+ Poly-SiGate Ox
Ge (SiGe)
<20nm ReRAM
Semi-conducting filament
BE
MOx
TESemi-conducting filament
BE
MOx
TE
Gate stack Channels, contacts, USJ
High Fin/NW
SiSi
20 nm20 nm
{100
}
{100}
SiGe
37
Summary
•
More Moore or More than Moore is NOT
the question–
Sense, compute, store, transmit: Integrated functionality
•
Convergence of technologies is the future–
System level power reduction with performance gain is key
•
Functional diversity vs
device density–
Focus of miniaturization shifts to adding diverse components–
CMOS+ : Hybrid integration of beyond-CMOS with CMOS–
Challenge: How can we fabricate them and make them work seamlessly
•
Fundamental shift in memory to non-charge-based memory–
Embedded memory for level 3/4 cache –
Reconfigurable memory-logic devices in the horizon
•
3D Interconnects is a game changer•
Process-Integration-Design co-optimization is a must
38
Vision of Future
Source: Brian Cristie
Design
Accelerating the next technology revolution
Research Development Manufacturing