darpa mto mems technologies for communicationsctnguyen/research/plenarytalks/... · stress curled...

C. T.-C. Nguyen, Nanotech’03, 2/25/03 – Approved for Public Release, Distribution Unlimited

DARPADARPA

MTOMTO

MEMS Technologies for Communications

Clark T.-C. NguyenProgram Manager, MPG/CSAC/MX

Microsystems Technology Office (MTO)Defense Advanced Research Projects Agency

Nanotech’03Feb. 25, 2003


DARPADARPA

MTOMTOOutline

• Introduction: Miniaturization of Transceiversneed for high-Q

• MEMS Components for RF Front Endsmicromechanical RF switchestunable micromechancial C’s & L’svibrating micromechanical resonators

• Micromechanical Circuits• Chip-Scale Atomic Clocks• Conclusions


DARPADARPA

MTOMTOFrequency Division Multiplexing

• Information is transmitted in specific frequency channels within specific bands

TransmittedPower

Frequency

GSM Band DCS1800 BandAdj. BandBand


DARPADARPA


TransmittedPower

Frequency


Filter



DARPADARPA


TransmittedPower

Frequency


Filter • Need: high frequency selectivity• Need: high frequency stability



DARPADARPA


TransmittedPower

Frequency


Filter • Need: high frequency selectivity• Need: high frequency stability


need high Q

ivoi

ωωο

vi ioResonant Circuit

Higher QHigher Q


DARPADARPA

MTOMTO

• Problem: IC’s cannot achieve Q’s in the thousandstransistors consume too much power to get Qon-chip spiral inductors Q’s no higher than ~10off-chip inductors Q’s in the range of 100’s

• Observation: vibrating mechanical resonances Q > 1,000• Example: quartz crystal resonators (e.g., in wristwatches)

extremely high Q’s ~ 10,000 or higher (Q ~ 106 possible)mechanically vibrates at a distinct frequency in a thickness-shear mode

Attaining High-Q


DARPADARPA

MTOMTO

• The total area on a printed circuit board for a wireless phone is often dominated by passive components passives pose a bottleneck on the ultimate miniaturization of transceivers

TransistorChips

TransistorChips

QuartzCrystalQuartzCrystal

IF Filter(SAW)

IF Filter(SAW)

InductorsCapacitorsResistors


IF Filter(SAW)

IF Filter(SAW)

RF Filter(ceramic)RF Filter(ceramic)

So Many Passive Components!


DARPADARPA

MTOMTO

• Fabrication steps compatible with planar IC processing

Surface Micromachining


DARPADARPA

MTOMTO

• Completely monolithic, low phase noise, high-Q oscillator (effectively, an integrated crystal oscillator)

• To allow the use of >600oC processing temperatures, tungsten (instead of aluminum) is used for metallization

OscilloscopeOutput

Waveform

Single-Chip Ckt/MEMS Integration

[Nguyen, Howe 1993]


DARPADARPA

MTOMTOBenefits of MEMS: Size Reduction

TransistorChips

TransistorChips

QuartzCrystalQuartzCrystal

IF Filter(SAW)

IF Filter(SAW)


InductorsCapacitorsResistors IF Filter

(SAW)IF Filter(SAW)

RF Filter(ceramic)RF Filter(ceramic)

MEMSMEMSTechnologyTechnology

Single-ChipTransceiver


DARPADARPA

MTOMTOMEMS Replaceable Components

• Next generation cell phones need multi-band reconfigurabilityeven larger number of high-Q components needed

• Micromachined versions of off-chip components, including vibrating resonators, switches, capacitors, and inductors, could maintain or shrink the size of future wireless phones


DARPADARPA

MTOMTO

Micromechanical Switches


DARPADARPA

MTOMTOμMechanical RF Switch Uses

Switchable LCBandpass FilterSwitchable LC

Bandpass FilterSwitch in capacitors to program the filter

center frequency

Switch in capacitors to program the filter

center frequency

Again, switch in elements to program

center frequency

Again, switch in elements to program

center frequency


DARPADARPA

MTOMTOMicromechanical Switch

[C. Goldsmith, 1995]

• Operate the micromechanical beam in an up/down binary fashion

• Performance: I.L.~0.1dB, IIP3 ~ 66dBm (extremely linear)• Issues: switching voltage ~ 50V, switching time: 1-5μs

Electrode

OutputInput

Dielectric


DARPADARPA

MTOMTOZipper-Actuated RF MEMS Switch

• Tri-layer cantilever AC switchcomp-SiO2 / Al / tensile-SiO2zipper actuator allows low Vactuate ~35-40V, Vhold ~8-10Vtri-layer stress pull-up forcecan tailor actuation waveform for best performance

• Reliability Performance:>100 B mechanical cycles~ 10 hr hold-down times7W max cold switched power

• Need: more indep. testing

[Keast 2002]


DARPADARPA

MTOMTOPhased Array Antenna


DARPADARPA

MTOMTO

Medium-Q Resonators


DARPADARPA

MTOMTOMedium-Q Resonator Needs

Switchable LCBandpass FilterSwitchable LC

Bandpass Filter

Problem: Switch losscompromises filter loss

• Medium-Q best achieved via tunable micromachined capacitors and inductors


DARPADARPA

MTOMTO

• Medium-Q best achieved via tunable micromachined capacitors and inductors

Medium-Q Resonator Needs

Tunable LCBandpass Filter

Tunable LCBandpass Filter

Mechanically tunable LC tank with higher Q than

conventional on-chip tanks

Mechanically tunable LC tank with higher Q than

conventional on-chip tanks

Eliminates switch loss better insertion loss

Eliminates switch loss better insertion loss


DARPADARPA

MTOMTO

• Micromachined, movable, aluminum plate-to-plate capacitors• Tuning range exceeding that of on-chip diode capacitors and

on par with off-chip varactor diode capacitors

• Challenges: microphonics, tuning range truncated by pull-in

Voltage-Tunable High-Q Capacitor


DARPADARPA

MTOMTO

Vtune

Larger Capacitive Tuning Range

• Use comb-transducers to actuate multiple plate capacitors

• Left: lateral comb-capacitor in deep RIE’ed silicon

• Nearly 250% tuning range with ~100V of actuation input

[Yao 1999][Rockwell]


DARPADARPA

MTOMTOSuspended, Stacked Spiral Inductor

• Strategies for maximizing Q:15μm-thick, electroplated Cu windings reduces series Rsuspended above the substrate reduces substrate loss


DARPADARPA

MTOMTOOut-of-Plane Micromachined Inductor

• Molybdenum-chromium metal solenoids perpendicular to the plane of the substrate

reduced substrate loss high Q• Assembled out-of-plane via curling

stresses, then locked into place• Record Q’s: ~70 on glass, ~40 on

20Ω-cm silicon

LockingMechanism

SolenoidInductor

Stress CurledMetal

Design/Performance:D=600μm, t=1μm

On Glass Substrate:L = 8nH, Q = 70 @ 1GHz

On 20Ω-cm Silicon:L = 6 nH, Q = 40 @ 1GHz

Design/Performance:D=600μm, t=1μm

On Glass Substrate:L = 8nH, Q = 70 @ 1GHz

On 20Ω-cm Silicon:L = 6 nH, Q = 40 @ 1GHz

D

[Chua, Hilton Head’02][PARC]


DARPADARPA

MTOMTO

High-Q Resonators


DARPADARPA

MTOMTOHigh-Q Resonator Needs

Best if Q >300Best if Q >300

Would likeQ’s >2,000




Best when highest Q used

Best when highest Q used

Would likeQ’s >10,000Would likeQ’s >10,000

• High-Q best achieved via vibrating micromechanical resonators


DARPADARPA

MTOMTO

Thin-Film Bulk Acoustic Resonator (FBAR)

• Piezoelectric membrane sandwiched by metal electrodesextensional mode vibration: 1.8 to 7 GHz, Q ~500dimensions on the order of 200μm for 1.6 GHzlink together in ladders to make filters

Agilent FBAR

• Limitations:Q only 500 (would like >2,000)difficult to achieve several different freqs. on a single-chip


DARPADARPA

MTOMTOCC-Beam μMechanical Resonator

• To date, most used design to achieve HF frequencies

• Smaller mass higher frequency range and lower series Rx


DARPADARPA

MTOMTOFabricated 8.5-MHz CC-μBeam

• Surface-micromachined, POCl3-doped polycrystalline silicon

• Extracted Q = 8,000 (vacuum)• Freq. and Q influenced by

dc-bias and anchor effects


DARPADARPA

MTOMTO92 MHz Free-Free Beam μResonator

• Free-free beam μmechanical resonator with non-intrusive supports reduce anchor dissipation higher Q


DARPADARPA

MTOMTONanomechanical Vibrating Resonator

• Constructed in SiC material w/ 50 nm Al metallization for magnetomotive pickup

192.0 192.5 193.0 193.5 194.00

200

400

600

800

1000

1200

Mag

neto

mot

ive

Res

p. [n

V]

Frequency [MHz]

Design/Performance:Lr =3 μm, Wr =150 nm, h= 259 nm

fo=732.9MHz, Q=7,330

Lr

Wr

[Roukes, Zorman 2000]

h


DARPADARPA

MTOMTO

Scaling-Induced Performance Limitations

Mass Loading Noise

• Differences in rates of adsorption and desorptionof contaminant molecules

mass fluctuationsfrequency fluctuations

Temperature Fluctuation Noise

• Absorption/emission of photons

temperature fluctuationsfrequency fluctuations

ContaminantMolecules

mass ~10-13 kg

mk

2π1

of =

Photons

volume ~10-15 m3

• Problem: If dimensions too small phase noise significant!• Solution: operate under optimum pressure and temperature

[J. R. Vig, 1999]


DARPADARPA

MTOMTO

156 MHz Radial Contour-Mode Disk Micromechanical Resonator

• Below: Balanced radial-mode disk polysilicon μmechanicalresonator (34 μm diameter)

μmechanical DiskResonator

MetalElectrode

MetalElectrode

R

Anchor

Design/Performance:R=17μm, t=2μm

d=1,000Å, VP=35Vfo=156.23MHz, Q=9,400

fo=156MHzQ=9,400

[Clark, Hsu, Nguyen IEDM’00]


DARPADARPA

MTOMTO

733 MHz Self-Aligned Radial Contour-Mode Disk μMechanical Resonator

• Self-aligned stem for reduced anchor dissipation• Polysilicon electrodes for better gap stability• Q > 6,000 seen even in air (i.e., atmospheric pressure)!• Below: 20 μm diameter disk

PolysiliconElectrode


R

Self-Aligned Stem

GroundPlane

fo=433MHzQ=4,066

μMechanical DiskResonator

-105

-100

-95

-90

-85

732 732.5 733 733.5 734

Frequency (MHz)

Tran

smis

sion

(dB

)

[Wang, Nguyen 2002]

fo = 732.9MHzQ = 7,330 (vac)

Design/Performance:R=10μm, t=2.1μm, d=800Å, VP=6.2Vfo=732.9MHz (2nd mode), Q=7,330

Q = 6,100 (air)


DARPADARPA

MTOMTO

1.14-GHz Self-Aligned Radial Contour-Mode Disk μMechanical Resonator

• Self-aligned stem for reduced anchor dissipation• Operated in the 3rd radial-contour mode• Q > 1,500 seen even in air (i.e., atmospheric pressure)!• Below: 20 μm diameter disk



R

Self-Aligned Stem

GroundPlane


[Wang, Nguyen 2003]

Design/Performance:R=10μm, t=2.1μm, d=800Å, VP=6.2V

fo=1.14 GHz (3rd mode), Q=1,595

-84-83-82-81-80-79-78-77

1138.5 1139.5 1140.5 1141.5

fo = 1.14 GHzQ = 1,595 (vac)Q = 1,583 (air)

Tran

smis

sion

[dB

]

Frequency [MHz]


DARPADARPA

MTOMTOPolysilicon Resonator fo-Q Product

• Freq.-Q product rising exponentially over the past years• In the process: sizes of resonators have not changed much

1990 1995 2000 2005 2010

1015

109

1010

1011

1012

1013

1014

Year

Freq

uenc

y-Q

Prod

uct

Intrinsic material Q limit nowhere

in sight?

Intrinsic material Q limit nowhere

in sight?


DARPADARPA

MTOMTO

733 MHz Self-Aligned Radial Contour-Mode Disk μMechanical Resonator

• Self-aligned stem for reduced anchor dissipation• Polysilicon electrodes for better gap stability• Q > 6,000 seen even in air (i.e., atmospheric pressure)!• Below: 20 μm diameter disk



R

Self-Aligned Stem

GroundPlane

fo=433MHzQ=4,066


-105

-100

-95

-90

-85

732 732.5 733 733.5 734

Frequency (MHz)

Tran

smis

sion

(dB

)

[Wang, Nguyen 2002]

fo = 732.9MHzQ = 7,330 (vac)Q = 6,100 (air)

Design/Performance:R=10μm, t=2.1μm, d=800Å, VP=6.2Vfo=732.9MHz (2nd mode), Q=7,330


DARPADARPA

MTOMTOResearch Issue: Impedance Matching

• Impedance matching needednot a new problem …

( ) 22

4

221

hVd

Qk

rR

Po

rx ⋅=

ωπ

μMechanical Disk Resonator

Antenna

50Ω

ResonanceResonance

kr = 7,350,000 N/m

5kΩ

Mismatch!

VDD

VSS

Off-ChipCapacitor

(2 pF)

MinimumSize

Inverter

2fFProgressively Larger Inverters

r h

Gap = d

VP


DARPADARPA

MTOMTOMicromechanical Circuits

• A single mechanical beam can’t really do much on its own• But use many mechanical beams attached together in a

circuit, and attain a more complex, more useful function

InputForce

Fi

OutputDisplacement

xo

t

xo

t

Fi

Key Design Property: High Q


DARPADARPA

MTOMTODesired Filter Characteristics

• Small shape factor generally preferred


DARPADARPA

MTOMTOMicromechanical Filter Circuit


DARPADARPA

MTOMTO

• Below: 7.81 MHz, 2-resonator, CC-beam μmechanical filter in POCl3-doped polysilicon

Spring-Coupled μMechanical Filter

Design:Lr=40.8μm, Wr=8μm, h=2μm,

Ls12=20.35μm, VP=35V

Design:Lr=40.8μm, Wr=8μm, h=2μm,

Ls12=20.35μm, VP=35V

Performance: (70 mTorr vacuum)fo=7.81 MHz, BW=18 kHz, Qres=6,000

PBW=0.23%, Ins. Loss < 1 dB

Performance: (70 mTorr vacuum)fo=7.81 MHz, BW=18 kHz, Qres=6,000

PBW=0.23%, Ins. Loss < 1 dB

Mechanical, not electrical, interconnect reduced

susceptibility to EMI

Mechanical, not electrical, interconnect reduced

susceptibility to EMI


DARPADARPA

MTOMTO

Provides robustness against jammers

and extends battery lifetime

Provides robustness against jammers

and extends battery lifetime

Reduces loss and removes power consumption by active devices

Reduces loss and removes power consumption by active devices

Miniaturization Miniaturization Eliminates active phase-locking ckt.

power consumption

Eliminates active phase-locking ckt.

power consumption

MEMS-Based Receiver Front-End

LowLoss


DARPADARPA

MTOMTO

Chip-Scale Atomic Clocks


DARPADARPA

MTOMTOChip-Scale Atomic Clocks

State-of-Practice

State-of-Research

HP 5071AVol: 29,700 cm3

Power: 50 WAcc: 5×10–13

$50K Datum R2000Vol: 9,050 cm3

Power: 60 WAcc: 5×10–11

$7K Temex RMOVol: 230 cm3

Power: 10 WAcc: 1×10–11

$2K

NG MC250Vol: 27 cm3+

Power: 220 mW+Acc: 1.5×10–11

KerncoVol: 20 cm3+

Power: 300 mW+Acc: 1×10–11

NIST CPT cellVol: 30 cm3+

Power: 300 mW+Acc: 1×10–11

NISTNIST-- F1F1

VolVol: ~3.7 m: ~3.7 m33

Power: ~500 WPower: ~500 WAcc: Acc: 3.83.8××1010––1515

CSACCSACVolVol: 1 cm: 1 cm33

Power: 30 Power: 30 mWmWStab: Stab: 11××1010––1111


DARPADARPA

MTOMTOChip-Scale Atomic Clock

• Key Challenges:cell design for maximum QGHz high-Q reference resonatorthermal isolation for low power

ModulatedLaser

PhotoDetector

133Cs vapor at 10–7 torrMod f

Atomic Clock Concept Cs or Cs or RbRbGlassGlassDetectorDetector

VCSELVCSEL

SubstrateSubstrate

GHzGHzResonatorResonatorin Vacuumin Vacuum

MEMS andMEMS andPhotonic Photonic

TechnologiesTechnologies

Chip-ScaleAtomic Clock


DARPADARPA

MTOMTOConclusions

• Integrated mechanical technologies, whether micro or nano, possess high-Q and low loss characteristics capable of revolutionizing the performance of wireless communications

• MEMS or NEMS offer the same scaling advantages that IC technology offers (e.g., speed, low power, complexity, cost), but they do so for domains beyond electronics:

• Debate over micro vs. nano: a waste of timeopportunities abound in both domains

• More important: MEMS or NEMS have brought together people from diverse disciplines this is the key to growth!

resonant frequency (faster speed)actuation force (lower power)

# mechanical elements (higher complexity)integration level (lower cost)

Size

darpa mto mems technologies for communicationsctnguyen/research/plenarytalks/... · stress curled...

Documents