© imec 2002

Failure mechanisms and reliability issues of RF-MEMS switches

I. De Wolf, P. Czarnecki, A. Jourdain, S. Kalicinski, R. Modlinski, P. Muller, X. Rottenberg,

P. Soussan, H. Tilmans

IMECLeuven, Belgium

RF in

RF out

© imec 2005

Outline

Introduction

FMEA of RF-MEMS switches

Charging induced stiction

Packaging effect on switch lifetime

Conclusions

© imec 2005

RF MEMSRadio Frequency MEMS switches offer:– Low weight– Low volume– Lower insertion loss– High isolation– Large frequency range– Extremely high linearity– Low power consumption– Integration possibilities

Uses in space:Wireless personal communication, satellite communication, phasedarray for beam steering, smart antennas ...

BUT: reliability is a problem

© imec 2005

RF-MEMS capacitive switch

=0V

RF-in

shuntswitchV

g

Zo

RF signalgenerator

RFchokeDC block RF-out

Zo

DCblock

Vdc

GND GND

dielectric

Flexiblemetal bridge

ON

“small” C

RF+DC in

RF out

=0V=20V

RF+DC in

RF out

OFF

“large” C

Flexiblemetal bridge

RF in

RF out

Radio frequency MEMS

© imec 2005

ESA - ENDORFINS: synopsis

TitleEnabling deployment of RF MEMS technology in space

telecommunication.

ObjectivePerform an in-depth assessment of the reliability and related

failure modes of RF-MEMS, in view of their deployment in space and improve this reliability (for switches) through processing optimization.

Starting dateAug 15, 2005

Duration24 months

© imec 2005

FMEA: Failure Mode Effect Analysis

The physics, chemistry causing the failure.What is observed, the signature of the failure modeWhat is first measured indicating a failure

What is causing the failure mechanism

© imec 2005

FMEA: Failure Mode Effect Analysis

© imec 2005

FMEA: Failure Mode Effect Analysis

© imec 2005

Lifetime tests in dedicated chamber

Vacuum resistantRF and DC probes

Thermo-chuck-10 to +150 oC

Pressure control (down to 10-6 mbar)

X-Y-Z- Θ Chuck stage (150, 150,15 mm, + 7o

travel range)

Wafer load slotUp to 200 mm wafers

XYZ-microscope stage 12x zoom

Gases inlet

© imec 2005

Charging induced stictionLifetime test @ 100 HzVact = 25 V, unipolar actuationN2 environment

Cdown

Cup

0V

∆C=Cdown- Cup end of life (EOL)Failure due to charging induced stiction

ctr

failed

Stiction

© imec 2005

Fast C-V

200-20

200-20

200-20

Control:symmetric curve

Failed: curve shifts to right.

19 hours later: curve shifted partly back

van Spengen et al, IRPS 2005, P. Czarnecki et al., MEMSWAVE 2005

Stiction

1.8

1.4

1.0

0.6

0.2

Capacitance [pF]

100

101

102

103

104

105

106

# of cycles

Cclosed

Copen

1.8

1.4

1.0

0.6

0.2

Capacitance [pF]

100

101

102

103

104

105

106

# of cycles

Cclosed

Copen

© imec 2005

C/V characteristic shift without surface charge

-60 -40 -20 0 20 40 600

0.4

0.8

1.2

1.6 x 10-13

DC voltage - V [V]

Cap

acita

nce

-C [F

]

Pull-in Pull-in

Pull-out Pull-out

Pull-outwindow

X. Rottenberg et al, 34th EuMW conf., 2004

© imec 2005

C/V characteristic shift due to a positive surface charge

-60 -40 -20 0 20 40 600

0.4

0.8

1.2

1.6 x 10-13

DC voltage - V [V]

Cap

acita

nce

-C [F

]

C-V with chargingC-V without charging

Pull-outwindow

Vshift=Q/Cdown

X. Rottenberg et al, 34th EuMW conf., 2004

Stiction

© imec 2005

C/V characteristic shift due to a positive surface charge

-60 -40 -20 0 20 40 600

0.4

0.8

1.2

1.6 x 10-13

DC voltage - V [V]

Cap

acita

nce

-C [F

]

C-V with chargingC-V without charging

Pull-outwindow

Vshift=Q/Cdown -> stiction

X. Rottenberg et al, 34th EuMW conf., 2004, van Spengen et al., J. of Micromech. Microeng. 14, 2004

Stiction

© imec 2005

Solution: Alternative actuation

U

t

U

t

U

t

U

tReduced V across dielectric

Bipolar

© imec 2005

Unipolar vs bipolar

Ut

Ut

Unipolar actuation35V, 100Hz

Bipolar actuation35V, 100Hz

>2x1075x106

© imec 2005

3D problem description

The total charge (Q) can be zero, but there might be a charge distribution in the dielectric.

Assume a volume charge density Ψ(x,y,z) in the dielectric.-> Equivalent charge distribution Ψeq(x,y) (2D+ problem)

X. Rottenberg et al, 34th EuMW conf., 2004

Stiction

© imec 2005

2D+ problem description

( )( )

⎪⎭

⎪⎬⎫

⎪⎩

⎪⎨⎧

+−+

= )()(2

)()( 2

22

2222

21

221

eqeqelC

AreaCQVCdCd

CdCdF ψσ

= total equivalent charge = Area x mean of eqQ ),( yxeqψ

= variance of the equivalent charge distribution)(2eqψσ ),( yxeqψ

realizes a force offset (y-shift in the Fel vs. V curve))(2eqψσ

realizes a voltage offset (x-shift in the Fel vs. V curve) eqQ

© imec 2005

Evolution of Pull-in and Pull-outStiction

-60 -40 -20 0 20 40 600

0.5

1

1.5

2

2.5

3

3.5 x 10-7

DC Pull-in and Pull-out voltages - V [V]

Varia

nce

[(C

oulo

mbs

per

uni

t are

a)2 ]

2_

2 )( POnoeq σψσ =

2_

2 )( PInoeq σψσ = Evolution of VPI

Evolution of VPO

-60 -40 -20 0 20 40 600

0.5

1

1.5

2

2.5

3

3.5 x 10-7

DC Pull-in and Pull-out voltages - V [V]

Varia

nce

[(C

oulo

mbs

per

uni

t are

a)2 ]

-60 -40 -20 0 20 40 600

0.5

1

1.5

2

2.5

3

3.5 x 10-7

DC Pull-in and Pull-out voltages - V [V]

Varia

nce

[(C

oulo

mbs

per

uni

t are

a)2 ]

2_

2 )( POnoeq σψσ =

2_

2 )( PInoeq σψσ = Evolution of VPI

Evolution of VPO

2_

01

20002

)(22)( POnoeq ddCk

Areakd σεεψσ =

==≥

Increasing k, d0 and decreasing the Area decrease the sensitivity to the equivalent charge distribution

© imec 2005

Charging induced stictionTheory - Experiment

X. Rottenberg et al, 34th EuMW conf., 2004, P. Czarnecki et al., submitted for MEMSWAVE 2005

ctr

failed

Stiction

Unipolar actuation: Q = 0 -> shift of the C-V curvesBipolar actuation: Q = 0 but σ2(Ψeq) = 0 -> narrowing of Vpo and Vpi

© imec 2005

Package

Pressure, humidity, optical, chemicals, particles, …

MEMS package

Electrical Thermal Power

MEMS

-keep good things in(pressure, getters, …)

-throw excess things out(heath,…)

- allow easy in-out to VIPS(electrical, to-sense stuff,…)

- give mechanical support

- be reliable

- keep bad things out (particles, humidity, gasses,…)

Function: “Gate keeper”

© imec 2005

Some metals (ex. Al-alloys) change ‘stress’ when heated above T = Tc:

Deformation: T-effects

250 °C during 10 minutes

Before

After

Temperature:- during functioning - during packaging

Deformation

Al 600nm 4 inch Si wafer

0.00

0.02

0.04

0.06

0 50 100 150 200 250 300 350Temp [C]

Slop

e [1

/m]

1st sample 200C 2nd sample 250C 3rd sample 300C

Heating up

Cooling downTc

© imec 2005

Tc is alloy dependent: Tc AlCuMgMn > Tc Al

Deformation: T-effectsDeformation

AlCuMgMn200100

0-100-200-300

300

-400

50 150 250 350

Stress (MPa)

Temp (oC)200100

0-100-200-300

300

-400

50 150 250 350

Stress (MPa)

Temp (oC)200100

0-100-200-300

300

-400

50 150 250 350

Stress (MPa)

Temp (oC)

Al 600nm 4 inch Si wafer

0.00

0.02

0.04

0.06

0 50 100 150 200 250 300 350Temp [C]

Slop

e [1

/m]

1st sample 200C 2nd sample 250C 3rd sample 300C

• Use metal with high Tc

• Or do a pre-anneal (but different stress)

• Optimize design to minimize the impact of stress changes on the shape of the bridge.

© imec 2005

Pressure in cavity

2x10-4 mbar1 bar

1.00.80.60.40.20.0

∆(C) (a.u.)

1000600200Time (µs) 10006002000

- faster switching in vacuum- vibrations

© imec 2005

Pressure in cavity

∆C (a.u.)

Time (µs)10080604020

1 bar

0.5

0.250.125

Time (µs)10080604020

1 bar

0.5

0.250.125

Stiff Floppy

The ‘floppy’ switch shows overshoot already at 0.125 bar.

© imec 2005

Pressure in cavity

bar0.090.0750.0560.0520.0450.030.020.01vac

∆C (a.u.)

Time (µs)80604020

The ‘STIFF’ switch shows overshoot at ~ 0.075 bar: clear dependence of the ‘overshoot point’ on the design

Stiff

© imec 2005

Pressure in cavityLower lifetime at lower pressure (larger Cdown, better charging).

Details and more results will be presented at MEMS2006 by P. Czarnecki et al., IMEC

© imec 2005

Gasses in cavity: N2 vs air

Nitrogen vs air: longer lifetime + larger Cdown

1.2

1.0

0.8

0.6

0.4

0.2

0.0

∆C (a.u.)

100 101 102 103 104 105 106 107

# of cycles

air

nitrogen

1 10 102 103 104 105 106 107 108 1090

0.1

0.2

0.3

0.4

∆C [a.u.]

# of cycles

nitrogen

air

1 10 102 103 104 105 106 107 108 1090

0.1

0.2

0.3

0.4

∆C [a.u.]

# of cycles

nitrogen

air

SiO2Ta2O5

0.40

0.35

0.30

0.25

0.20

0.15

∆C (a.u.)

100 101 102 103 104 105 106

# cycles

air

nitrogen

SiNx

Different technologies, different designs, different dielectrics, different electrical test conditions (Vact, freq.)

-> packaging in N2 gives a better reliability(different damping, dielectric constant, gap breakdown V, humidity,…)

© imec 2005

Conclusions- FMEA: main failure mechanism in capacitive RF-MEMS = charging of the dielectric leading to stiction of the bridge- Possible solutions:

- Design for low Vpi, but high Vpo- Design for flat bridge (uniform charging + low charge distribution)- Make the insulator area as small as possible (lower sensitivity of Vpo)- Use bipolar actuation waveform- Package the switch in a nitrogen environment- Be careful with vacuum (bouncing + lower lifetime possible)

- FMEA: main packaging induced failure is deformation of the bridge - Possible solutions:

- Use metal with high Tc- Try to reduce the packaging T

What else can be done?- alternative ways of bipolar actuation- better dielectric (less charging sensitive)- worse dielectric (such that the charges disappear faster)

© imec 2005

Acknowledgments

IST 28231IST 28231

IST 28276M ed i n aM ed i n a14627/00/NL/KW14627/00/NL/KW

IWT MISTRAIWT MISTRA

IMEC: MEMS, packaging, reliability and RF team

Endorfins