Nanoelectromechanical systems, paths for co …Co-integration of CMOS and NEMS on 40 nm FD-SOI J. Philippe et al, IEEE NEMS 2014, E. Ollieret. al., IEEE MEMS 2012 G. Arndt et al.,

Nano electromechanical systems,

paths for co integration with CMOS

Thomas Ernst

IEEE WIMNAC 45 – Tokyo Institute of Technology

19/2/2015

© CEA. All rights reserved

| 2

Motivations – some NEMS applications

Scaling laws and convergences with CMOS

From mass to gas sensing

Toward VLSI and multi-physics models

Conclusions & perspectives

Outline

T. Ernst, IEEE WIMNAC 45


| 3

The conjunction of the cost per function decrease and

the emergence of “killer applications” which demanded a

large volume of leading edge chip has been the revenue

generator for the µε industry and its sustainability

� 1980s � Analog for TVs and VCRs

� 1990s � Digital for PCs

� 2000s � Analog and Digital for Cell Phones

� 2010s � Analog and Digital for Mobile Internet

� 2020s � Smart interacting (sensors) devices,

IoT ???

Moore's law needs « killer applications »

newtechnologies

newmarket

moreR&D

morerevenue



| 4

Toward multi-physics signals

Information capture

Processing Storage

Restitution

Integrated Circuit

Analog or logic electrical signalElectro-magnetic

light sound Chemical motion

heat

Analog or logic electrical signal Electro-magnetic

light sound Chemical motionheat T. Ernst, IEEE WIMNAC 45


| 5

Nano Switches

30 µmUCBerkeley

From MEMS to NEMS – Typical sizes

MEMS accelerometer

500 µm

1 µm1 µm

Nano cantilever

80 nm

10 µm

NEMS

accelerometer

1 µm

16nm

Suspended silicon nanowire

Todayapplications

Emergingapplications

Stanford U.

Clocking



| 6

First large arrays of NEMS

~60.000 NEMS / mm²

Ability to process NEMS at VLSI scale



| 7

Mechanical switches electronics

� Motivations� For high ION/IOFF ratio

� Clear need of VLSI

� Huge but existing market (already in-place technology)

� Challenges (at nm-scale)� Mechanical/contact properties ?

� Scalability (size, supply voltage)

� Speed

� Reliability/cycling & variabilityS DG

gap Dimplegap

G DS



| 8

State of the art –M/NEM Relay

� Main features

� Good leakage~10-14 A

� Switching time: 3 ns (expected)

� First logic circuit demonstrated

� Variability of R~40-400

� Co-integration with CMOS for power reduction

Stanford

K. Akarvardar et al., IEEE Elec Device Lett 30, p. 1143-1145 (2009)

M. Spencer et al., IEEE J. of Solid-State Circuits, 46, p. 308-320 (2011)

H. Kam et al., IEEE TED, VOL. 58, NO. 1, (2011)

S. Chong et al., ICCAD’09,

V. Beiu et al. ,IEEE Nano 2014

MEM relay logic mapping

UC Berkeley

30 µm



| 9

Ultra-small & dense sensing devices

� For new "chemical imaging" application

� Need of VLSI may depend on the application !

� Looking for "generic" sensing technology for high volume

� Driving market not (yet) clear … but …



| 10






Outline



| 11

Electromechanical features of NEMS

t

� High frequency f 0, low mass m

� Very sensitive to mass loading (ag – zg)

� Very sensitive to very weak forces (fN-aN) / fields / charges (e-)

20l

tEf

ρ∝

0

2f

fmm

δδ =

E = Young modulus

l

Modeled as damped simple harmonic oscillator



| 12

MEMS resonator bode diagram

Bandwidth :f0Q

12Q

f0 (1 - ) 12Q

f0 (1 + )f0 f0

The resonance can be measured through the max gain or the phase condition



| 13

Scaling laws

Law Scaling Typical values

Effective Mass 1 pg – 10 fg

Stiffness 1 N.m-1 – 10-2 N.m-1

Frequency 10 MHz- 1GHz

Quality factor

total/dissipated

energy per cycle?

102 (air) – 104

(vacuum)

Dissipated power 100 aW – 10 fW

Mechanical time

(resonator)

0.1 µs – 10 µs

Limit of Detection

(mass)

1 zg – 1 ag

20l

tEf

ρ∝

twlmeff ∝

3α

3

3

l

Ewtk ∝ α

Q

TkfP B

th02π

∝ 1−α

00 2 f

Q

πτ =

α

0

2f

fmLODm

δ=

3α

1−α

Q

l’=αlt'= αtw'=αw

T. Ernst et al., IEEE DRC 2012



| 14

Convergence with CMOS ?

T. Ernst et al. IEDM 2006, 2008, 2009

SIZE CMOS Nanowires / FinFet / SOI technologies …

… can be "adapted" for NEMS !!!

Si

HfO2/TiN



| 15

� NEMS-CMOS integration means

(i) enhanced signal detection

(ii) increased Signal/Noise Ratio

→ ultra-high resolution

� Ultra-dense NEMS arrays

(i) individually addressed resonators

(ii) increased robustness (redundancy)

(iii) larger capture area

(iv) further noise reduction by signal averaging

Why embedding NEMS on CMOS ?



| 16

0100110…0000110…

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

E-Cell

Signal amplification and multiplexing

A sensing array scheme



| 17

Emerging NEMS - which path to follow ?

WW II: Simple machines and manual

laborers fill a room.

1955: ENIAC, the first electronic

computer, fills a room.

2000s: The integrated circuit has

made computation ubiquitous.

The microelectronic revolution… inspired by silicon and VLSI

� Complex functions integrated - High performances

� Existing manufacturing or design tools, processes, design rules

� Existing micro sensors… (Accelerometers, gyros, imagers, TCD, inkjets

etc…)

� A lot of opportunities (assembly, costs, architecture etc…)T. Ernst, IEEE WIMNAC 45


| 18

NEMS / CMOS Co integration on bulk

� AMS 0.35µm (Post-CMOS processing)� Aluminum metallic NEMS (Metal 4) or PolySilicon (Poly1 / Poly2)� Capacitive actuation/detection limits resolution

Metallic resonator

Poly Si resonator

Electrode

Poly22Electrode

Poly2N-Well

PAD

ANCHORANCHORPoly1 ANCHORANCHOR

VIA

N-Well

Electrode1

Electrode2

ANCHOR

MET4

Electrode1

Electrode2

ANCHOR

PAD Layer

J. Verd, IEEE Electron Device Letters, vol. 27 (6), pp. 495497, 2006.



| 19

NEMS on O,35 um CMOS

� Mono-crystalline Si� Bulk CMOS ST 0.35µm (MSCMOS) on thick SOI� NEMS integrated before CMOS, released after CMOS

Layout

CMOS Amplifier + buffer

NEMS

10µm

Vue MEB

CMOS

NEMS

NEMS (cantilever) HF

J. Philippe et al, IEEE NEMS 2014 T. Ernst, IEEE WIMNAC 45


| 20

NEMS 2D co-integrated in FD SOI technology

Co-integration of CMOS and NEMS on 40 nm FD-SOI

J. Philippe et al, IEEE NEMS 2014, E. Ollier et. al., IEEE MEMS 2012

G. Arndt et al., IEEE International Solid-State Circuits Conference 2012

T. Ernst et al., IEEE DRC 2012

Typical NEMS dimensions

L_beam=1µm

w_beam=100nm

t=40nm

w_gauge=40nm



| 21

NEMS-CMOS measurements

� 1V dc biased gauges / actuation: ac (10mV pp) + dc (8.5V)� First demonstration in the literature of:

� direct (homodyne) detection of such small VHF NEMS� CMOS integration of PZR NEMS

� unsuccessful measurement on a stand-alone device

VacuumQ=5000


| 22






Outline



| 23

Frequency-shift based mass sensing

23

doubly-clampedbeam resonator

Resonant response

1 µm

16nm

Frequency shift (a.u)

Mag

nitu

de (

a.u)

actuationdetection

• Electrostatic detection

• Piezoresistive transduction

• Measurement of adsorption/desorption of particles

molecules



| 24

NEMS-CMOS closed-loop oscillators

Key building block for

• real-time monitoring of resonance frequency

• individual addressing of dense arrays

NEMS resonator

A DPLL

NEMS resonator

A D

Amplifier & phase-shifter circuit

Phase

comparator

PLL

correctorVCO

| 24

PLL Self-oscillators



| 25| 25

NEMS-CMOS self-oscillator

8MHz - Si capacitive NEMS

Compact NEMS-CMOS pixel (50x70µm²)

J. Philippe et al., IEEE MEMS conference 2014 J. Philippe et al., MEMS 2014


| 26

Why using resonant NEMS for mass sensing ?

40

2−∝=−=ℜ l

M

f

m

f

effδδ 3

0

2 lf

fMm eff ∝−= δδ

Sensitivity Resolution

T. Ernst et al., IEDM 2008


| 27

Limit of detection?

� LOD depends on noise processes

� Electronic noise (Johnson noise, 1/f…)

� Transducer itself (Johnson noise…)

� Thermomechanical noise related to Q

� Others…

E. Colinet et al., J Appl. Phys. 105, 124908 (2009)

1.968 1.97 1.972 1.974 1.976 1.978 1.98

x 107

0

0.5

1

1.5

2

2.5x 10

-3 amp

freqam

p

Q=5433

A=2.14mV

1.968 1.97 1.972 1.974 1.976 1.978 1.98

x 107

0

0.5

1

1.5

2

2.5x 10

-3 amp

freqam

p

Q=5433

A=2.14mV

Frequency noise spectral density

in Hz²/Hz

2

02

00

)(

2)(

signal

i

inoise

v

S

QS

∑

==ω

ωωωω

(1/signal to noise ratio (SNR))-1

Figure of merit:

Large detection gain

Low noise system

Large Q

0

0

0

)(2

ωωω

ωδωδ ω =

∝−=S

Mm


| 28

Mass sensing demonstration

Yang et al., Nano Lett., Vol. 6, No. 4, 2006

best mass resolution corresponds to 7 zg (30 xenon atoms)


Atomic mass unity = 1Da = 1 u ≈ 1.66053886 x 10-27kg1zg = 10-21g = 602 Da ≈ a nucleotides pair (DNA)

Hemoglobine A molecule

66.2 kDa

Parvoviridae viruses: Hepatitis B

1.1 MDa

E. Coli bacteria

4.2×1011 Da

Protein PrP (Prion)

150 kDa

G-C

616.4 Da

A-T

613.4 Da

Nanowires

NEMS

Mass units in biology


| 30

Nanowire used for mass detection

Capacitive actuation & detection Capacitive actuation & piezo-resistive detection with nanowires Thermo-elastic actuation

& piezo-resistive detection.

15 nm oscillator



| 31

Mass resolution with nanowire

Released nanowireBending oscillation

Mass resolution according to the diameter

10 nm

80 nm

R. He, M. Roukes et al. Nanoletters 12/08This work

He et al.



| 32

A new design example

E. Mile et al., Nanotechnology 21 (2010) 165504

� Electrostatic actuation

� Piezo-resistive detection (down mixing scheme)

� Excellent Signal to background ratio



| 33

Mass Sensing (toward mass spectrometry)

Modes 1 & 2 continuously tracked (with 2

home made PLL)

electro spray of gold nano particles

Focusing of nano particles1 gold nano particle of 400 kDa was detected

in 100 ms – 1 Dalton=1.6x10-24 g

(~ 100 zg in 1s)

M.S. Hanay et al., Nature Nanotech (2012)



| 34

Technologies for gas sensing

Multi-gas selectivity,

Sensibility

Affordability, Compactness

Poor Correct High Best-in-Class

Poor

Correct

High

Best-in-Class

Bench-top GC/MS or

GC/XX systems

Handheld single gas

detectors (NDIR or

electrochem.) (industrial

safety, …)

Portable multi-gas FTIR

analyzers

Portable GC systems

Colorimetric tubes

Mono-gas analyzers for

O2 / H2S / NO / …

monitoring (air quality,

process, emissions, …)DOAS spectrometers (air quality,

process, emissions, …)

Handheld multigas

detectors

Indoor air quality

sensors

Detectors

AnalyzersAPIX

Integrated

NEMS

Sensor



| 35

Multi-gas sensing with NEMS+GC

� Gas Chromatography (GC) is a well-known separation method

Injection of the

gas mixtureCarrier gas flow

GC column

NEMS chip

� GC column provides selectivity by separating in time and space the gas mixture components



| 36

Gas recognition



| 37






Outline



| 38

For any new devices, a sufficient maturity is

needed for:

24/05/2011 38

� Compact models

� Introduction in a Design Kit

� Evaluation of performance in the system environment

"Multi physics" simulation & technology platforms

High volume applications !!!



| 39

Global simulation of a multi-physics system

( ) ( )( )

( )( )

++−

−

−−−

=uttuD

ttuzerf

uttuD

ttuzerf

ctzc

FGi

FG

FGi

FG

/2/4

2/

/2/4

2/

2, 0

O. Martin et al., Sensors et Actuators B, Volume 194, April 2014, Pages 220–228


| 40

Library, design kit & design rules for NEMS

Suspended SNW

Cross-beam

Electrostatic

force

F (VG1, VG1, VD, X)

Mechanical

ODE

NEMS

Current

computation at

every nodes

VG1

VS2

VD

VS1

VS2

F X

ID

IS1

IS2

IG1

IG2

� Electrical libraries of NEMS were developed (VHDL-AMS) and

are used for designing circuit



| 41

Optimization thanks to compact modeling

� Introduction of sources noises, parasitic effects

� Optimization of the whole system

0 0.5 1 1.5-200

-150

-100

-50

0

50

100

150

200

250

300

time in s

df in

Hz

0 0.5 1 1.5-100

-50

0

50

100

150

200

time in s

df in

Hz

0 0.5 1 1.5-20

0

20

40

60

80

100

120

time in s

df in

Hz

1 NEMS Sigma(df) = 200 Hz �dm = 1.1 fg

10 NEMSSigma(df) = 100 Hz �

dm = 0.55 fg

1000 NEMSSigma(df) = 8 Hz �dm = 44 ag



| 42

Conclusions & Perspectives

� We reported ultra-scaled single-crystal Si nanowire NEMS

resonators operating in the 100MHz frequency range

� Their first monolithic integration at the front-end level with

CMOS enables to extract the signal from the background

thanks to reduced losses and local amplification leading to a

possible implementation of a direct measurement, for high-

sensitivity sensing applications and portable systems

� The VLSI environment provides moreover an opportunity to

build large dense NEMS array for ultra fast and high

performance mass sensing for complex gas detection,

mechanical relays, and others …

� 3D integration gives opportunities to optimize NEMS on exiting

CMOS technologies.T. Ernst, IEEE WIMNAC 45


| 43

Acknowledgements

� Co-authors: W. Ludurczak, J Arcamone, J. Philippe, O.

Martin, E. Ollier, C. Marcoux, F. Ricoul, C. Dupré, G. Billot, P.

Villard and L. Duraffourg

� Dr P. Andreucci, E. Colinet and P. Puget from APIX

� Prof. M. Roukes & team, CALTECH for collaboration

within CEA-LETI/CALTECH VLSI alliance on NEMS

� European Research Council for funding

� European project NEMSIC


Documents

Nanoelectromechanical systems, paths for co …Co-integration of CMOS and NEMS on 40 nm FD-SOI J. Philippe et al, IEEE NEMS 2014, E. Ollieret. al., IEEE MEMS 2012 G. Arndt et al.,