MEMS Reliability Methods and Results

1
MEMS Reliability Methods and Results

• Danelle M. Tanner, Ph.D.
• Principal Member of the Technical Staff
• Sandia National Laboratories
• Reliability Physics Department
• tannerdm_at_sandia.gov
• http\\www.mems.sandia.gov

Sandia is a multiprogram laboratory operated by
Sandia Corporation, a Lockheed Martin
Company,for the United States Department of
Energy under contract DE-AC04-94AL85000.
2
Outline

• Introduction
• MEMS Taxonomy
• Microengine video
• General Reliability Method
• Test Structures and Models
• Reliability Triad
• Case Histories
• Shock and Vibration Results

3
Taxonomy of MEMS Devices
Class III Moving Parts, Impacting Surfaces
Class IV Moving Parts, Impacting and Rubbing
Surfaces

• Class II
• Moving Parts, No Rubbing or Impacting Surfaces

• Class I
• No Moving parts

Optical Switches Corner Cube Refl. Shutters Scanne
rs Locks Discriminators
Gyros Comb Drives Resonators Filters
Accelerometers Pressure Sensors Ink Jet Print
Heads Strain Gauge
TI DMD ( 1B) Relays Valves Pumps Optical Switches
4
An MEMS example - Microengines are driven by
electrostatic actuation
150 mm
5
The MEMS Reliability Process assures
Qualification of MEMS devices
Identify Failure Modes
WEAR
Science-based Understanding
ESD
Statistical Characterization
Testing and Test Structure Design
Develop Predictive Reliability Models
6
Test structures determine material properties and
provide data for models
Adhesive properties
Cantilever properties
Frictional properties
Stress properties
FR
7
Cantilevers can be used to measure adhesion, G
t
h
s
(de Boer and Michalske, Journal of Applied
Physics, 1999)
8
Adhesion is exponentially dependentupon RH for
rough hydrophilic surfaces
20
Assuming nAsp
de Boer et. al., MRS vol. 518
µm2
9
Friction tester in a controlled environment
provides model data
Critical for performance aging models of
packaged parts!
2993-2 dry air

• agreement with other measurements of friction for
  FDTS-coated polysilicon

m 0.06
3010-1 air1.5 vol. H2O
device stuck
m 0.02

• debris generation begins before sticking -
  monolayer coating gone?

Mike Dugger, SNL
10
Modeling the dominant failure mechanism of wear
Force
Force
Asperities on both surfaces
Material transfers to top surface or breaks,
generating particles
DV - wear volume DL - length moved F - normal
force applied c - wear variable, ? wear
coefficient and ?-1 hardness
11
A physical model describes failure and
acceleration factors
Rubbing Surfaces
Before
Pin
Hub
Accumulated Debris
10 mm
Tanner et al., 1998 IEEE IRPS Proc, Reno, NV,
March 30 - April 2, pp.26-35 Tanner et al,
Microelectronics Reliability, 39, 1999,
pp.401-414.
12
Improvements in reliability and performance are
typically achieved through changes in design,
operation, and materials
13
Case History Design - change shuttle guide to
prevent lateral clamping
solution
problem motion
14
Case History Operation - Distributions become
bimodal with lower frequency showing orders of
magnitude improvement
Resonant frequency was 1500 Hz
3000 Hz
1720 Hz
500 Hz
860 Hz
All data was from microengine failures
15
Case History Operation At higher frequencies,
microengine motion degrades from circular
Gearless microengine
Up-Down Actuator
Desired motion lt 500 Hz
Measured motion gt 500 Hz
Dynamics of motion are not modeled properly for
the higher frequencies
Left-Right Actuator
16
Case History Materials - Selective,
self-limiting, conformal tungsten hard coatings
1 million cycles
1 billion cycles
W-coated
pin joint
3 mm
W-coated
(Sita Mani et al., MRS vol. 605)
pin joint
17
Micromachines are extremely robust in shock
environments
shock levelG
40000
10000
20000
500
4000
die attach failure
debris movement
ceramic pkg fracture
structural damage
D. Tanner et.al., IRPS 2000

• working devices found out to 40 kG
• design/process modifications can allow survival
  beyond 20,000 G

18
Vibration produced an adhesion failure
D. Tanner et.al., IRPS 2000
adhesion
debris movement
Shuttle

• Observed debris movement
• failure was due to adhesion between the guides
  and shuttle

10 mm
Guides
19
Summary

• Polysilicon MEMS in the SUMMiT process are
  robust through many environments
• Reliability in MEMS will be the key to use in
  space and defense applications
• While many reliability issues are being actively
  studied and understood, dormancy is just
  beginning to be explored
• Need development of aging models with
  acceleration factors
• Government labs should drive science-based
  reliability and qualification
• Industry will develop reliability data, but not
  the science to understand effects
• Industry will not share commercially valuable
  information