SystemX: HIEA · 2014. 11. 11. · HIEA Focus Area Vision Heterogeneous Integration of Everything onto Anything (HIEA) ... Printed Electrochemical Transistors on Paper Substrates

SystemX: HIEA

E. Pop, H.-S.P. Wong (co-leads), J. Fan, K. Goodson, S. Mitra, Y. Nishi, J. Plummer, K. Saraswat, D. Senesky, X. Zheng

H E T E R O G E N E O U S I N T E G R AT I O N O F E V E RY T H I N G O N TO A N Y T H I N G

Computing Is Everywhere

(not just here)2

Heterogeneous Integration is Driving Research

Societal “pull”: every day 2.5 exabytes are created

Existing hardware not fast enough, consumes too much energy Companies often do not have the breadth to tackle full range from

nanomaterials to systems

Approaches:› In-memory computing› Monolithic 3D interleaving of memory & logic› Address dense-system thermal challenges› Process data in energy-efficient way at the source› Integrated sensors

“… a single advanced brain laboratory would produce 3 petabytesof data annually, roughly as much data as the world's largest and most complex science projects [Large Hadron Collider (LHC)].”

‐ Kavli Foundation (2013)

3

HIEA Focus Area Vision

Heterogeneous Integration of Everything onto Anything (HIEA) Integration of “beyond-Si” platforms for “beyond Moore” applications

› Monolithic integration of logic, memory, sensors, thermal management, flexible substrates› Energy-efficient and brain-inspired design opportunities› Autonomous electronics and energy-harvesting opportunities

Core faculty: E. Pop, H.-S.P. Wong (co-leads), J. Fan, K. Goodson, S. Mitra, Y. Nishi, J. Plummer, K. Saraswat, D. Senesky, X. Zheng

4

Example: Device Scaling Challenges

10

100

1000

90 65 45 32 22 16 10 7 5 3G

ate

Pitc

h (n

m)

Tech. Node (nm)source: IEEE Spectrum, Intel

Gate Pitch is key metric of device scaling

Must scale both LG and LC

May be possible ONLY with atomically thin channels

old pitch scaling relying on LG

LG

today

LG + LC scaling? ? 20 nm

5

Project Areas and Topics

Mid- & near-field wireless power with

stretchable electronics (Fan)

Chip-integrated thermoelectric power for wireless sensor

networks (Goodson)

Body-Heat-Powered Autonomous Wearable

Electronics (Pop)

autonomous sensors & energy

Monolithic 3D integration of logic & memory for energy-efficient computation

(Mitra, Wong)

Architectural studies of monolithic 3D

integration benefits for energy-efficient

computation (Mitra)

Fully Integrated Thermal Management of 2D/3D Chips

(Goodson)

3D monolithic integration

Heterogeneous integration of Ge and III-Vs on Si

(Saraswat and Plummer)

Monolithic integration of AlGaN/GaN high

electron mobility N/MEMS (Senesky)

X‐on‐Si

Autonomous wireless sensor nodes for harsh

environments via AlGaN/GaN

heterostructures (Senesky)

X‐on‐flexSkin-like electrodes for long-term

electrophysiological monitoring (Fan)

Integration of Atomically Thin 2D

with 3D Materials and Devices (Pop)

Transfer Printing Methods for

Integrating Thin Film Electronics onto Any Substrates (Zheng)

Electronic synaptic devices for brain-inspired computing (Wong)

Applications of Ge and III-Vs on Si for MOSFETs, photonics and

photovoltaics (Saraswat)

Flexible electronic devices(Nishi)

6

Project Areas and Topics: 3D Monolithic Integration




networks (Goodson)


Electronics (Pop)






X‐on‐Si












(Mitra, Wong)



computation (Mitra)


(Goodson)





7

Monolithic 3D Integration (Wong, Mitra)

local ILVs, no TSVs process T < 250 oC

1D/2D FETs

Cache (L2, L3), main memory

STTRAM

1D/2D FETs

3D RRAMMain memory,

storage

On-chip heat spreading

On-chip heat spreading

ALU & flip-flops, register files,

cache(L1)

8

Monolithic 3D Integration (Wong, Mitra)

0%

20%

40%

60%

80%

100%

0%

20%

40%

60%

80%

100%Energy

Architecture Energy benefit Delay benefit Energy‐Delay‐Product benefit

2D 1 1 1

Monolithic 3D(CNFET, MRAM, RRAM)

25X 60X 1,500X

InterconnectMem. access

Stall due to contentionIdleInstructions

Stall due to contention

Instructions Mem. Access

Delay

Cache

0.0%

1.0%

2.0%

3.0%

4.0%

0.0%

0.5%

1.0%

1.5%

2D Mono‐3D 2D Mono‐3D

Application: PageRank (Google Datasets)

9

Imperfection-Immune Design Approach (Mitra, Wong)

[PI: Mitra, P. Wong, DAC ’13, Nature ’13, IEDM ‘14]10

Fully Integrated Thermal Management of Monolithic 3D

hydrophilic

hydrophobic

hot spot

2D heat spreaderVapor chamber handles hot spot

Coating for phase separation

Monolithic Thermal Platform (MTP)

Copper meshCopper nanomesh heat spreader & phase change

500 nm Electrical / thermal co-design

K. Goodson Group

11

Saraswat Group Research

Motivation:

• Si CMOS device performance increase commensurate with size scaling is beginning to saturate.

• Relentless scaling paradigm is also threatened by fundamental limits including power dissipation, communication bandwidth, and signal latency for both off-chip and on-chip interconnects.

• High efficiency solar cell technology is not economical

Objective

• Heterogeneous integration of alternate materials and structures for future ICs with scalability to future nodes

• Develop alternative technology for low cost high efficiency solar cells

12

Saraswat Group Research Projects

Nanoscale Ge & III-V MOSFET Technology

Ge Optical Detector

Capacitorless 1T-DRAM

Optical Interconnects

n+-Ge

Si

Stressor

Etch Mask

p-Ge LTO

Metal Metal

SiN

Metal gate/high‐k dielectric with improved scalability

Integra on on Si

High mobility strained Ge or III‐V channel high Ion

Non‐planar device structure for improved electrosta c integrity low Ioff

Low resistance contacts

Si oxide

S D

G

Junctionless Solar CellGe/Si Electro-Absorption

Modulator

GaP GaP Silicon SiGe

BOX

Ge Laser

13

Synaptic Devices for Brain-Inspired Computing (Wong)

14

Brain-Inspired Computing (Wong)

Crossbar architecture inspired from grid-like connectivity of the brain

Pattern recognition on hardware using a 10x10 synaptic grid:

Synapses are implemented by PCM cells and neurons by software

S. B Eryilmaz, p 25.5 IEDM 2013.

Phase change memory (PCM)

PCM array PCM device

N1

N2

N10

N1

. . .

. . .

. . .

.

.

.

WL #1

WL #2

WL #10Output of neuron

Input to neuron

N10

BL #1 BL #2BL #10

Firing neurons

No signal

time

Recall phaseUpdate phase

Voltage

BL

WL

BL

WL

Non‐firing neurons

15




networks (Goodson)


Electronics (Pop)





Project Areas and Topics: X-on-Si and X-on-Flex


(Mitra, Wong)



computation (Mitra)


(Goodson)









X‐on‐Si








16

Heterogeneous Integration on Si with Rapid Melt Growth (1)

Deposited amorphous material contacts the Si substrate at a “seed hole” A microcrucible (Typically SiO2) contains the target material The target material is melted in an RTA machine; crystallization occurs

as it cools below the melting temperature Perfect single crystal material results suitable for device fabrication. The

example chip shows a CMOS circuit with NMOS in the Si substrate and PMOS in recrystallized Ge

Target material

Seed regionSupporting substrate

Micro-crucible

J. Plummer Group

17

Technique is applicable to binary compounds on Si Columns in blue have been recrystallized using RMG

Heterogeneous Integration on Si with Rapid Melt Growth (2)J. Plummer Group

GaSb

Si

nitride

Material Si Ge InSb GaSb InAs AlSb InP GaAsMelt Point (C) 1412 937 524 712 942 1065 1070 1238

Bandgap (eV) 1.12 0.74 0.235 0.81 0.43 1.57 1.42 1.51% Mismatch 0.0 4.0 19.7 12.5 11.7 12.9 8.3 4.0

Electron Mobility 1,500 3,900 80,000 5,000 33,000 200 4,600 8,500

Hole Mobility 450 1,900 1,250 850 460 420 150 400

Breakdown Field x105V/cm

3.0 0.86 .027 1.1 0.17 8.2 7.5 3.5

18

Scaling Challenges of Transistors (Pop, Wong)

Challenge: FETs “carved” out of bulk 3D materials (Si) surface roughness restricts mobility, band gap, scaling of dimensions

Possible Solution: FETs with atomically thin channel 1D and 2D materials (

Flexible Electronics (J. Fan)

21

Printed Electrochemical Transistors on Paper Substrates

Cellulose – most abundant polymer in the world Good mechanical properties, renewable, biodegradable

Y. Nishi Group (with Y. Cui)

22

inkjet printed NWs

inkjet printedP3HT or CNTs

AlGaN/GaN High Electron Mobility M/NEMS (D. Senesky)

Principal Investigator:› Prof. Debbie G. Senesky› EXtreme Environment

Microsystems Laboratory (XLab) Research Summary:

› New breed of micro/nano-scale sensors and electronics for harsh environments• Temperatures > 600°C• Radiation• Chemically corrosive

› Increase scientific knowledge of the physical mechanisms of electronic materials within unique environments

Extreme Environment Gallium Nitride (GaN) Sensing Platform

Radiation Exposure

Integration of Nanomaterials &

Catalysts

Wide Bandgap GaN HEMT

Sensor

-- -

---

V

High-Temperature Operation

Integrated μCooling

23

Transfer Printing Methods for X-on-Flex (X. Zheng)

Thin film electronics

Cheap and Light Solar CellsC. H. Lee, et al., Scientific Reports, 1000 (2012)

Attachable ElectronicsC.H. Lee, et al., Nano Letters, 11 (2011) Ultrathin Bio-sensors

Wearable Memory DevicesJ. Choi, et al., Phys. Status Solidi, 1-6 (2013)

24





X‐on‐Si









(Mitra, Wong)



computation (Mitra)


(Goodson)





Project Areas and Topics: Sensors and Energy




networks (Goodson)


Electronics (Pop)





25

AlGaN/GaN Sensors (D. Senesky)

SEM image of AlGaN/GaN high electron mobility strain sensors.

Monolithic integration of AlGaN/GaN Embedded Sensors for Hot‐spot Detection

Metal-organic chemical vapor deposition system for GaN heterostructure growth.

NEW!

26

Energy Scavenging Sensors for Smart Buildings (Goodson)

Optimization Contours

• Smart Buildings bring multitudes of sensors and wiring.• TE energy scavenging from heat sources (HVAC, water,

windows) allows “truly wireless” sensor networks.• Low power, low temperature TE conversion demands

revolutionary new materials and design.

Lifelink.com Students:Marc Dunham, Michael Barako

27

Thermoelectrics and “Self-Cooled” Electronics

Huge thermal anisotropy, good thermoelectric properties of nanomaterials

Exploit thermal anisotropy for low-power electronics (e.g. phase-change memory) Separate thermal and electrical flow (thermal transistor, thermal diode) Electronics with built-in thermoelectric cooling Flexible thermoelectrics for energy harvesting from human body

E. Pop Group (with Goodson and Wong)

28

HIEA Summary

Moore-Like Scaling ~10x (??)

Heterogeneous Integration ~10,000x

Heterogeneous integration of:– Low power logic and memory– Sensors and energy harvesters– Novel nanomaterials

Exploit heterostructures, thermal engineering, unconventional substrates

29

Documents

SystemX: HIEA · 2014. 11. 11. · HIEA Focus Area Vision Heterogeneous Integration of Everything onto Anything (HIEA) ... Printed Electrochemical Transistors on Paper Substrates