Novel device technologies • Slides largely based on presentation Li Shang and I gave at Tsinghua University • Some of the images might be copyrighted by others: don't charge or distribute
Novel device technologies
• Slides largely based on presentation Li Shang and I gave at Tsinghua University
• Some of the images might be copyrighted by others: don't charge or distribute
Overview
• Historic overview of technology scaling
• CMOS: current status and challenges
• Nanotechnologies: An introduction• Nano circuit and architecture design
– Carbon nanotube technology– Single electron tunneling transistor
• Open questions
Evolution of electronics0
1
1850 1875 1900 1925 1950 1975 2000 2025
Mechanical
Electro-Mechanical
Electronic-VT
Bipolar
NMOS
CMOS……. ?
The first computer
Babbage difference Engine (1822)– 4000 components
– Three tons
– 31 digits
• Pros – automatic• Cons – slow, expensive, not
flexible
The first electronic computer
• Electrical numerical integrator and computer ENIAC (1946)– 18,000 vacuum tubes– 30 tons– 100,000 calculations per
second
• Pro – faster, flexible• Con – too big, not reliable
Semiconductors
The first transistor
The first integrated circuit 4004 Pentium® 4
1.E-21
1.E-18
1.E-15
1.E-12
1.E-09
1.E-06
1.E-03
1.E+00
1940 1960 1980 2000 2020
Cu
bic
Met
er
Vacuum tube
Transistor
NMOS
CMOS
Benefits of scaling
1.E-11
1.E-09
1.E-07
1.E-05
1.E-03
1.E-01
1.E+01
1940 1960 1980 2000 2020
Del
ay (
Sec
) Vacuum tube
Transistor
NMOS
CMOS
1.E-161.E-141.E-121.E-101.E-081.E-061.E-041.E-021.E+00
1940 1960 1980 2000 2020
Jou
les
Vacuum tube
Transistor
NMOS
CMOS
1.E-061.E-051.E-041.E-031.E-021.E-011.E+001.E+011.E+02
1940 1960 1980 2000 2020
Co
st (
$)
Vacuum tube
Transistor
NMOS
CMOS
ENIAC-on-a-chip
• 40 mm2
– 174,000 transistors
• 0.5 W• 20 MHz
Technology outlook
Medium High Very HighVariability
Energy scaling will slow down>0.5>0.5>0.35Energy/Logic Op scaling
0.5 to 1 layer per generation8-97-86-7Metal Layers
11111111RC Delay
Reduce slowly towards 2-2.5<3~3ILD (K)
Low Probability High ProbabilityAlternate, 3G etc
128
11
2016
High Probability Low ProbabilityBulk Planar CMOS
Delay scaling will slow down>0.7~0.70.7Delay = CV/I scaling
256643216842Integration Capacity (BT)
8162232456590Technology Node (nm)
2018201420122010200820062004High Volume Manufacturing
Si Substrate
Metal Gate
High-kTri-Gate
S
G
D
III-V
S
Carbon Nanotube FET
50 nm
35 nm
30 nm
SiGe S/D
Strained Silicon
SiGe S/D
Strained Silicon
90 nm65 nm
45 nm32 nm
20042006
20082010
2012+
Technology Generation
20 nm 10 nm
5 nm5 nm
5 nm
Nanowire
Manufacturing Development Research
CMOS research continues…
Nanotechnology outlook
• Why CMOS technology?– Performance
• Gain, low noise– Area
• Massive integration– Power– Reliability– Fabrication difficulty & cost
• Status & challenges– 65 nm Intel, IBM, TSMC, etc.– Power challenge
• Energy, thermal issues
– Fabrication cost• Photolithograph masks are expensive
– Reliability• Soft errors • Electromigration, dielectric breakdown, etc.
– Process variation
• Why nanotechnology?– To continue technology
scaling to further scale integration, reduce cost, improve performance, and minimize power consumption
• Candidates– Carbon nanotube– Nanowire– Single electron device
Emerging nano devices
Carbon nanotube
Forms of carbon…
Carbon atom can form several distinct types of valence bonds….
A bit of history….
Edison's original carbon-filament lamp
US Patent 223898
1880
1978
F/A-18 HornetThe first aircraft with carbon fiber wings
1985
Discovery of Fullerenes(Smalley)
Nanotubes discovered at NEC, by Japanese researcher
Dr. Sumio Iijima
1991
Carbon nanotube transistor based logic-performing ICs (IBM)
2001
Carbon nanotubes in interconnet applications
(Kreupl, Infineon)
2002
CNT Vias
What are carbon nanotubes?
Graphene Single-WalledNanotube(SWCNT)
Multi-WalledNanotube(MWCNT)
Courtesy: F. Kreupl, Infineon
Eg ~1/d: Thick (>5nm) MWCNTs have a vanishing band gap at 300K, hence metallic…..
What are carbon nanotubes?
Armchair Nanotube (metallic) Zigzag Nanotube (semi-conducting)
SWCNTs
Various roll-ups possible depending on chirality
Chiral nanotube
Properties
• Carbon nanotube – Single-wall or multi-wall– Diameter: 0.4-100nm– Length: up to millimeters– Ballistic transport– Excellent thermal
conductivity– Very high current density– High chemical stability– Robust to environment– Tensile strength: 45 TPa! (high
strength steel ~ 2TPa)– Temperature Stability: up to
2800 oC in vacuum and up to 700 oC in air
3855800
Hone, et al., Phys. Rev.
B, 1999
Thermal conductivity
(W/mK)
40>1000
McEuen, et al., Trans.
Nano., 2002
Mean free path (nm)
@ room temp
<1x107>1x109
Wei, et al., APL, 2001
Max current density (A/cm2)
CuCNT
NRAM
• Non-volatile nanotube random-access memory (NRAM)– Mechanically bent or not: determines bistable on/off states– Fully CMOS-compatible manufacturing process– Prototype chip: 10 Gbit NRAM
– Will be ready for the market in the near future
Source: Nantero
NRAM
• Properties of NRAM– Non-volatile– Similar speed to SRAM
– Similar density to DRAM– Chemically and mechanically stable
Single-electron tunneling transistor (SET)
SET structure
SET background
• Device structure– A nanometer-scale conductive island embedded in an
insulating material– Electrons travel between the island and electrodes
through tunneling junctions
S
CG :gate capacitance CD :drain tunnel junction capacitanceCG2 :optional 2 nd gate capacitance RS :source tunnel junction resistanceCS :source tunnel junction capacitance R D :drain tunnel junction resistance
G
D
G2
CG
CG2CS,RS CD,RD
Source
drain
Gate (G)island
i(t)
optional 2nd gate (G2)
tunneljunction
(S) (D)
SET background
• Coulomb blockade effect– Electrostatic charge: e2/C– Island has a minimal-energy number of electrons
– When integer, tunneling in and out blocked– When ½ integer, tunneling permitted
– Can tune with applied gate voltage, VG
SET characteristics
• Periodic I-V characteristics• This is not due to de Broglie wavelength!
SET characteristics
• Ultra-low power consumption• Strong temperature dependency
• Fabrication challenge• Reliability concern
– Random background charge noise
• Low drive strength
• Low circuit gain
e2 /C≥K BT
R≥h/e2≈26K
g=CG/ C SC D
ICE-FLEX: hybrid SET/CMOS reconfigurable architecture design
• Design space– Power consumption– Performance– Fabrication challenge– Room-temperature operation– Reliability concerns
SET configuration memory
SET local interconnect Hybrid SET/CMOS globalinterconnect
Majority voting logic
SET multi-gate lookup table
SET input switch fabric SET registers
• Reconfigurable Architecture– SET logic block– SET/CMOS interconnect– SET configuration memory– SET/CMOS majority voting logic
SET reconfigurable logic
• SET multi-gate look-up table– Multi-gate structure– Strong temperature
dependence
• SET configuration memory– Phase shifting
controlled by Coulomb charge
SET configuration cache
Curre
nt configura
tion mem
ory
VCG
charge
VGVoutD SIDS
VG
VCG
VGVoutD SIDS
VG
VGStore 1
Store 0
SET configuration cache
Configuration setsset k-1 set0set1
Config. Bit0
Config. Bitm-1
config
ura
tion
m-to-1 multi-gate multiplexer SET tree
Vdd
VG2
Vss
-VG2
Config. Bit0
Config. Bit1
Config. Bitmc-1
s0 s1 snc
s0 s1 snc
s0 s1 snc
mc-to-1 multi-gate SET multiplexer
SET interconnect
• SET local interconnect– Static power dominant
• Hybrid SET/CMOS global interconnect– Dynamic power dominant
VG2
-VG2
VG2
-VG2
HLB input
SINV1 SINV2CINV1 CINV2
Inter-HLB metal wire
Majority voting & arithmetic logic
• SET MVL– Efficient implementation– Suffer from random
background charge effect itself
• SET arithmetic logic– Threshold logic– Non-unate functions
VG2
-VG2
Out
Vss
Vdd
IN1
INk-1
IN1
INk-1
Characterization
Characterization
Characterization
• Ultra-low power sensor applications– AVR (4 MHz) can run for 9.7 years on one AA battery– Scavenging indoor solar energy is sufficient to power
AVR at 1.6 MHz
• Power-efficient high-performance computing– LEON2 SPARC: 1.6 Terra IPS at 100 W power
budget, 16 Terra IPS under 1 kW power budget
Open questions
• Will nanotechnologies replace CMOS completely? – CMOS: high-performance, low cost, mature
fabrication, reliable, relatively low power
• When will nanotechnologies be ready? – Nanotechnologies: fabrication challenge (10–15 years
to mature), reliability challenge, performance concern, etc.
• What can we do?– CAD support, circuit design, architecture research
Open questions
• Characterizing impact of new technologies on applications: start with technology library
• Changes to classical problems: buffer insertion• Nano modeling: Compact fast models validated
against reported results and Monte-Carlo
• Reliability models: Correlations, distributions, rates
• Reliability solutions: Markov random fields or modular redundancy