AND STlCTlON IN MEMS PACKAGING

HERMETICITY
.S. Joshua Jacobs, Seth A. Miller, Joshua J. Malone, William C . McDonald, Vincent C. Lopes, Lissa K. Magel

Texas Instruments, DLPTM Products

PO Box 655012, MS 914; Dallas, TX 15265
972-9274392; fax: 972.995-061 5 ; e-mail: [email protected]

., ABSTRACT The principal tool used to characterize the electrooptical

response of the D M D is the MirrorMaster [4,5], an intemally
Exposure of operating MEMS structures to atmospheric gasses developed optical inspection tool that couples DMD drive
may have deleterious effects on device performance. Here we electronics with a CCD-camera-equipped bptical microscope
consider from a phenomenological point of view the effects that on a motorized stage.
moisture has on both short- and long-term performance
characteristics of Texas Inshuments'. Digital Micromirror Device
(DMD). As an array ofup to 1.3 million mirrors, a single DMD can
provide a wealth of statistical information; compilations of device
statistics provide thorough descriptions of effects that may be
obtained through static or operational aging with a variety of package
environments. The detection sensitivity of our test methods provides
significant insight into the global and local effects of package
environment, including the effects of water on device operation.

INTRODUCTION

The DMD is an electrostatically addressed array -of between

500,000 and 1.3 million micioscopic mirrors fabricated on top of
standard CMOS S U M circuitry [I]. A DMDTMdevice.is shown in Figure 2. DMD Architecture.
Figure 1 and an illustration of a production pixel is shown in Figure
2. The pixels are currently produced with center-to-center spacings
In the test technique of Bias Adhesion Mapping (BAM), pixel-
of either 14 or 11 microns in SVGA, XGA, SXGA and HD formats
by-pixel response is measured versus net applied bias in order to
for commercial and consumer end equipment markets., Over
detennine the landing and release characteristics of the individual
1,000,000production devices have been shipped since they were first
mirrors of a particular DMD (Figure .3).
offered for sale. '.

.
For~wVbsl?~
capture m e "age hame lor each Mlfage chmqe

I F i g u r e 3. BAM.Biao Voltage Profile

BAM is measured for every device and yields a bevy of

characteristic data about each DMD; these data are recorded to a
central database for process monitoring and control. We have found
that BAM can highlight structural and process variations, including
Figure 1. Production DMDs (I to r) SVGA, XGA, s X G A packaging microenvironment variations, that give rise to very
different landing and release behavim of the micromirrors. Typical
Moisture has been reported to cause varying effects in MEMS BAM plots for landing and release are shown in Figure 4.
devices, from capillary condensation and post-release stiction [2] to
increased wear during operation in either low or high humidity
environments [3]. Here we discuss. the effects of water on the
operation and long-term reliability of the Digital Micromirror Device
(DMD) and the detection of various effects of moisture through
stiction meaburements.

DMD CHARACTERIZATION
The MirrorMoster and Bias Adhesion Mapping

0-7803-7352-91OZi$17 .OO WOO2 IEEE iEEE 02Cli37320. 40th Annual inkmationai Reliability
136 Physics Symposium, Dallas, Texas, 2002
 The combination of model and experiment demonstrates that
BAM is a sensitive technique that differentiates among levels of
stiction force across a population of DMDs. BAM thus provides a
production monitor to cull a small number of devices produced that
do not meet strict stiction specifications. Indeed, we have verified
empirically that devices failing these criteria would not survive in
field use.
t . .
DMD PACKAGING AND RELIABILITY

The reliability goals. of DMD packaging'are to eliminate or

mitigate the two primary failure modes that might he encountered
during extended use. These are stiction and hinge memory, the
accrual of a permanent torque of the tiinge element that prevents the
DMD from operating properly. .For detail on hinge memory testing
Figure 4. Typical B A M Plots, and reliability, see the accompanying paper by Sontheimer [71.

Acceptable sticlion Stiction can be a primary culprit in premature failure of MEMS

devices; manufacturing and packaging strategies must' effectively
No MEMS device with contacting surfaces can escape the effects deal with this phenomenon to provide a viable product. In the
of stiction, hut the design and execution must manage stiction well to manufacture of the DMD, dry etch technology is used to eliminate
ensure reliable operation. 'In many cases, exact strategies for stiction post-release .stiction, while a passivation layer is employed to
management remain closely guarded trade secrets, and ours is no mitigate in-use stiction. The production DMD package [SI
exception. The very low device failure threshold for the DMD (one comprises a multilayer Land Grid Array ceramic substrate With a
mirror) demands that we pay attention to the last pixels to release. brazed seal ring and a metal lid with a fused, optical quality window.
The maximum stiction force that may be overcome depends on pixel Further, a heat sink is anached externally, while proprietary genering
geometry, hut the acceptable range of forces spans at least two orders systems are used internally (Figure 6). This hermetic technology,
of magnitudc for all DMD designs. Under normal operation, the adapted from military 'systems, facilitates hut does not ensure
DMD provides additional energy to the pixels to overcome stiction, suitable conditions for reliable operation. This combination of
providing substantial operating margin. However, the introduction technologies allows reliable operation of DMDs for many years
of moisture under conditions described below can severely under standard field conditions.
compromise or even eliminate the latitude inherent in the design and 4. Hermetic
operation of the DMD.

BAMDislributions

BAM distributions may be constructed by plotting frequency vs.

Vbias for voltages at which a specific number of mirrors remain
landed, e&, 90%, SO%, 10% (Figure 5 ) . Inspection of these
distributions over large numbers of parts provides substantial
information ahout the general control of stiction within a population
strips
of devices. We have previously presented a computer-based
6. Heat sink header
integrated numerical model for the behavior of the DMD pixel [ 6 ] .
The Gaussian shape of these curves is consistent with our treatment stud
of the stiction interaction as a small. normallv distributed retarding
force at each contact point.
Figure 6. Illustration of the features of the DMD package.

Release Histogram ' THEEFFECTS

OF W A T E R
7nn
A continuing theme of our research is to understand the effects
that determine reliability in both hermetic and nonhermetic package
environments. We have previously :reported a model for the
penetration of moisture through a nonherhetic seal counterbalanced
by the action-of a moisture genering material [9], as well as a
mechanical model of such a package [lo]. These' studies have
demonstrated that particular attention must be paid to material
selection, component design, and system 'conditions in order to
ensure reliable operation throughout .the intended service period.
The constraints of optical and mechanical alignment, over and above
normal electrical interface considerations, provide a .significant
Figure 5. Release Histogram showing distribution of design and execution challenge; a holistic approach is necessary to
voltages at which IO%, 50%, and 90% of the pixels are success. ..
. . .
released.
Stiction effects of added water

137
 Continuing studies will elucidate the effects of added water and
In a series of studies, we have focused on the effects of water the competition among various surfai:es in the package to adsorb this
introduced to the package by accelerated aging in humid water in a nonhermetic environment.
environments at temperatures up to 100 "C. The primary metric in
these studies was stiction accumulation as measured. by BAM. The Getter as Relative Humidity Contrder
effects of can be pronounced depending on the mode of moisture
introduction and the net amount of moisture that is introduced. During the course of our investigations, a potentially useful
Figure 7 shows progressive degradation of the BAM curves of property emerged of a reversible mcisture getter, that is, one whose
DMDs exposed to an 85 "C/85% relative humidity environment. As capacity is dependent on the local relative humidity. Examples of
DMDs typically are subjected to a substantial thermal gradient these getters include the various families of zeolites [Ill. In this
during normal operation, the effects of thermal gradient on the BAM case, for a getter and package of typical dimension, the conditioning
performance are also studied. The significant increase in stiction of the getter prior to final package sealing determines the internal
under these conditions is also found to occur only afler significant conditions of the package. Furthermore, as the temperature is
moisture introduction (Figure 8). changed, the relative humidity in the package varies only over a
small range, causing a significant increase in the partial pressure of
water with temperature (Figure 10). This result implies that the use
or test conditions of the componr:nt, including thermal gradient
performance, are strong determiners of the occurrence of stiction; a
system that is useful over one temperature range may not serve at all
well in another. On the other hand, achieving a particular package
RH range, as may be required for certain devices [3], is
straightforward. These results are derived from the fact that the
amount of moisture contained in package headspace is negligible
compared with the moisture containf d in the gettering composition.
Figure I. BAM of devices exposed to accelerated moisture
aging Package WV!I Temperature
Zeolite Getter Condilioned at 25 "C/40% RH

L
2.5
___-
5 7.5 10 12.5 15 17.5
-____I
20

Figure 8. Thermal gradient B A M of devices exposed to

accelerated moisture aging. I 0 20 40

'C
60 80 100

Hinge memory effects ofsurface water

In a complementary effort, we have studied the effects of surface

water adsorbed to the metal surfaces of the DMD. A primary effect Figure 10. Modulation of pac:kage RH by zeolite getter.
that we have discovered is the role of surface water in the accrual of
hinge memory; effective removal of this water prior to packaging
substantially increases hinge memory resistance in a hermetic SmMmRy
environment (Figure 9). .
Control of stiction will continue to he a critical determiner of
Hinge memory accumuldion rates success or failure of MEMS packaging strategies. The management
of water vapor, even within arc hermetic environment, is a
requirement. An array of MEMS si.ructures, such as the DMD, can
serve as a very sensitive monitor ofstiction conditions and provide
feedback for experiments designed to understand packaging process
effects on stiction. A computational model of the behavior of the
MEMS device provides a useful complement to experiments, and
provides further insight into stictiori effects. These tools should he
used in concert to design a MEMS device with substantial margin
against stiction in order to ensure acceptable manufacturing yield and
reliable field performance. Finally, production monitors must be
established to ensure performance and reliability of the finished
component.

F i g u r e 9. Hinge Memory accumulation w i t h high

temperature treatment.

138
 REFERENCES
[I] L.J. Hombeck, “Digital Light ProcessingTM and MEMS:
Timely Convergence for a Bright Future”, Plenary Session,
SPIE Micromachining and Microfabrication ‘95, Austin, TX,
October 24, 1995, p. 2.
[Z] N. Tas, T. Sonnenberg, H. Jansen, R. Legtenberg, M.
Elwenspoek, “Stiction in Surface Micromachining”, Journal
of Micromechanics and Microengineering, 1996, 6(4), 385-
397.
[31 S.T. Patton, W.D. Cowan, K.C. Eapen, J.S. Zabinski, ‘*Effect
of Surface Chemistry on the Tribological Performance of a
MEMS Electrostatic Lateral Output Motor” Tribologv Letters,
2000,9(3,4), 2001, pp. 199-209.
.~ D.L. Doane, R. Roy. C.K. Hams. J.D. Woodall. andT.J. Daw,
141
U.S. Patent 5,987,isi.
.. D.V. Dudasko. G.L. S t r e c k ” . D.L. Doane. A S . Leonard.
151
P.G. Barker, D.E. Russell, U.S:Patent6,088,474.
[6] P.F. Van Kessel, L.J. Homheck, R.E. Meier, M.R. Douglas,
“A MEMS-Based Projection Display”, Proceedings of the
IEEE, Vol. 86, No. 8, August 1998, pp. 1687-1704.
[7] A.B. Sontheimer, “Digital Micromirror Device (DMD) Hinge
Memory Lifetime Reliability Modeling,” Proceedings of the
IEEE, this volume.
[XI I. Faris, T. Kocian, “DMD Packages--Evolutionand Strategy,”
Texas Instruments Technical Journal, 1998, 15 (3), pp. 87-94.
[9] J. Jacobs, J. Malone “Suitability of an Epoxy Seal for a
MEMS Package,” in Proceedings of the Intemafional
Microelectronics and Packaging Sociefy 2000 Symposium,
Baltimore, MD, September 11-15,2000.
[IO] M.N. Variyam “Behavior of the DMD Package under
Mechanical and Thermal Loads,” Proceedings, ASME
International Mechunical Engineering Congress and
Exposition, A4EMS Symposium. November 2000, Vol. 2, pp.
601-608.
[ I l l K: Nagel, R. Von h m o n “Modeling of Adsorption of
‘%ton on Silica Gel and a Molecular Sieve,” Chem.-Ing:
Tech., 1985,57(11),pp. 996-7.

139