Shock Testing of MEMS Devices

Shock Testing of MEMS Devices Michelle A. Duesterhaus Vesta I. Bateman Darren A. Hoke Sandia National Laboratories, P.O. Box 5800 MS-1310, Albuquerque, NM 87185-1310 ABSTRACT Micro-Electro-Mechanical Systems (MEMS) are being considered for use in a wide range of devices that are subjected to mechanical shock environments. MEMS devices are of interest because they are expected to survive and function following severe shock environments due to their small relative size and mass. Tests have been conducted on surface micromachined MEMS devices, using a Hopkinson pressure bar, to quantify shock survivability and the results will be presented. MEMS die, consisting of both simple structures and complex actuators, were fabricated in the Sandia SUMMiT and Cronos MUMPS processes and subjected to compression, tension, and shear shocks up to 200,000 g s. During a field test, surface micromachined and bulk micromachined MEMS devices were subjected to the shock environment of a penetrator shot into a hard target. The unpackaged MEMS die were inspected before and after testing to quantify damage as a function of shock loading. Sensitivities to the direction and magnitude of the shock input, as well as associated failure mechanisms have been identified and will be presented. INTRODUCTION Micro-Electro-Mechanical Systems (MEMS) technology is producing devices that are expected to survive high-g shock environments. This work quantified shock levels at which MEMS structures and actuators fail, either by coming apart in a catastrophic manner or by failing to operate after a shock event. Field tests have exposed MEMS devices to the shock environment of a penetrator shot into a hard target. The field tests had shock levels of a few thousand g s with durations of milliseconds. Laboratory tests at high-g shock levels up to two hundred thousand g s with durations of tens of microseconds have been performed. The laboratory tests were focused on testing Surface micromachined (SMM) MEMS devices. These devices are produced by the Sandia SUMMiT and Cronos MUMPS processes which deposit alternating layers of polysilicon and sacrificial oxide from 1 to 3 microns thick. Average devices have lateral dimensions of to 1000 microns. Because of the small size of SMM MEMS, inertial forces on these devices are small. The field tests were performed on both SMM MEMS and Deep Reactive Ion Etched (DRIE) MEMS. DRIE is done by etching through a silicon wafer to create features with thickness of 0 microns and lateral dimensions from tens of microns to a few millimeters. The inertial forces on DRIE MEMS are an order of magnitude higher than SMM MEMS. By using silicon on insulator wafers the DRIE devices can have different features etched on the frontside and backside, allowing for more complexity in design. The DRIE devices tested were produced by the Sandia Compound Semiconductor Research Laboratory. SURFACE MICROMACHINED TEST STRUCTURES Several modules were designed for MEMS ing. Die #1 was a single module, fabricated using the SUMMiT process, and populated with actuators, cantilever beams, and fixed-fixed beams. The beams were 20 microns wide, with the length varying from 100 to 1000 microns. The square anchor cut attaching the beam to the Poly0 layer varied from 4 to 18 microns. The module also contained standard component library designs for a microengine, torsional ratcheting actuators (TRA), and thermal actuators. Die #2 was designed and fabricated using the MEMSCAP PolyMUMPS (formerly Cronos MUMPS ) process. This module contained cantilever beams, thermal actuators, resonators, and other standard components found in the MEMSCAP Consolidated Micromechanical Element Library (CaMEL). The beams were 20 microns wide, and the length varied from 100 to 1000 microns. The square anchor cut attaching the beam to the Poly0 layer varied from 2 to 8 microns. Die #1 and #2 were used for a series of laboratory tests which were previously summarized in [1] and these die were included in the field tests described herein. Die #3 was designed for further investigation of shock survivability. This module was a double module and fabricated using the SUMMiT process. A picture of Die #3 is shown in Figure 2. Cantilever beams with a width of 30 microns had length variations of 200, 400, 600, 800 and 1000 microns. The square anchor cut had side length variations of 4, 6, 8, 10, and 12 microns. The cantilever beams were fabricated with both the Poly1Poly2 and the Poly3 layer. The Poly1Poly2 beams have a 2.5 micron thickness and a 2.0 micron gap of between the beam and Poly0 base layer. The Poly3 layer beams have a 2. micron thickness and 6.8 micron gap. This set of beams was placed on the module twice, with one set oriented 90 degrees from the other. The module also contained 2 microengines, 1 TRA, 2 spiral springs, 4 resonators, 2 bent beam thermal

actuators, 2 differential leg thermal actuators, and a rotary table. presented herein. The laboratory results for Die #3 will be DRIE TEST STRUCTURES Other MEMS devices produced for various projects were included in the field test. The g-sensor is a mass on a folded spring suspension and is used as an inertial sensor. The g-sensor is created by using Deep Reactive Ion Etching to pattern silicon on insulator wafers. The dual environment sensor is also a mass on folded spring suspension, but it is created by using advanced DRIE to create features on two levels. A thin clamp must be moved out of engagement by an inertial force before a perpendicular inertial force can move the mass. LAB TEST SETUP The die were shocked in three directions, and the conventions of compression, tension and shear have been assigned. referred to the direction of shock in which a MEMS structure moved closer to the substrate due to its inertia and anchors were put into compression. referred to the direction of shock in which the MEMS structures moved away from the substrate and anchors were put into tension. The direction of both tension and compression shocks were normal to the plane of the MEMS die. Shear referred to a shock direction parallel to the plane of the MEMS die. All s were performed on a Hopkinson pressure bar. For the tests done in compression and tension, each die was bonded into a cavity of an aluminum fixture. The die was protected from dust and handling by covering the cavity with a machined cover. The fixture was held by vacuum to the end of a ¾ in. aluminum Hopkinson pressure bar. For the compression tests the fixture cavity faced away from the bar, as shown in Figure 1. For tension tests the cavity faced towards the bar. Previous work by Bateman [2] determined there was no significant impedance mismatch between the aluminum and silicon substrate; so the pulse was not attenuated at the silicon and aluminum interface. The reference measurement for all tests was a strain gage bridge located in the center of the aluminum bar. The uncertainty was +6% for these strain gage measurements [3]. Pulse durations were measured at 10% of the positive peak acceleration magnitudes. For tests conducted at -65F or 165F the end of the Hopkinson bar was placed in an environmental chamber. Each die was visually and electrically inspected before and after ing. Actuators have been observed to fail in two ways, by a noticeable physical failure or by not functioning even though the actuator is intact. A beam was considered to have failed if any part of it was broken off. The method of visual inspection used before and after s did not determine if the beams were free-standing or stuck down to the substrate. Figure 1. Shock fixture on end of Al Hopkinson bar Figure 2. Die #3 layout LAB TEST RESULTS In Figure 3 the results of the Lab ing for Die #3 are shown as the percentage of structures that survived. Each device or structure is represented by a column. The data are ordered in rows first by orientation as this is the most significant factor for survivability and secondly by shock level, and finally by the temperature at which the test was run. The percentage is calculated by dividing the number of working actuators or intact beams after the test by the number that were working or intact before the test. This accounts for devices and structures that were damaged during release, bonding or handling. The color coding divides the survivability into three categories. Red is shown when the survivability is between 0% and 33%, gray for 34% to 66%, and green for 67% to 100%. The columns start on the left with the most complicated actuator, with decreasing complexity toward the right. Then the beams are listed by increasing length. One would expect the complicated actuators and longest beams to fail at lower shock

levels and the results confirm this is true. A microengine consists of 2 combdrives that are linked at 90 degree angles to a drive gear. When the microengine fails due to the connection to the drive gear, the combdrives may still be functional. Thus, both the survival percentage of the microengine as a system and the combdrives individually are shown. The beams of a given length are grouped together without regard to the variable size anchor cuts. Orientation Shock Level (1000 g's) Temp (deg F) Micro engine TRA Comb drive Resonator Thermal Actuator 200 400 600 800 1000 165 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 72.5 % 0% 63% 100% 100% 100% 100% 100% 100% 100% -65 % % % 100% 100% 100% 100% 100% 100% 100% 165 75% 100% 100% 88% 100% 95% 98% 95% 83% 73% 72.5 % 0% % 100% 100% 100% 100% 100% 100% 100% -65 % % 75% 75% % 88% 90% 93% 75% 75% 165 100% 100% 100% 100% 100% 100% 98% 98% 100% 100% 72.5 % 100% % 100% 100% 100% 98% 93% 90% 90% -65 % % 75% 100% 100% 97% 98% 100% 100% 100% 165 75% % 75% 75% 100% 100% 90% 83% 98% 98% 72.5 0% % 75% 100% 100% 88% 95% 95% 89% 90% -65 0% 0% 0% 100% 100% 100% 98% 98% 100% 98% 165 % 100% % 100% 100% 95% 80% 75% 68% 80% 72.5 % 0% % 100% 100% 48% 93% 58% 65% 70% -65 0% % % 63% 100% 93% 95% 88% 80% 78% 165 0% % 0% 75% 100% 98% 70% 38% % 5% 72.5 0% % 0% 75% 100% 93% 53% 20% 3% 0% -65 0% % % % % 28% 28% 30% 20% 8% 165 0% 0% 0% 0% % 33% 3% 0% 0% 0% 72.5 0% 0% 0% % 100% 70% 0% 0% 0% 0% -65 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 165 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 72.5 0% 0% 0% % % 28% 0% 0% 0% 0% -65 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% Figure 3. Results from laboratory ing. The percentage of devices that survived the are shown. Figures 4 and 5 are SEM images of broken beams showing the effect of anchor size on the beam failure. Figure 6 summarizes the results with respect to the anchor cut size of the beams. The columns are labeled by the size of the anchor cut in microns, with various beam length lumped together. A smaller anchor has a smaller area of attachment between layers, a smaller perimeter of conformal material linking the beam to the anchor, and higher stress concentrations at the beam to anchor transition. The larger the anchor, the more robust the beam is. In this test, 30 micron wide beams with anchors of 10 microns by 10 microns were as robust as 12 by 12 anchors, indicating that above this size the anchor is no longer the weak spot. In previous tests with 20 micron wide beams this cutoff point was included 8 by 8 anchors. Therefore, anchors should be maximized based on the size of the feature they attach to the substrate and optimized to reduce stress concentrations. MEMS are much more robust in compression shock and are susceptible to tension shock. This indicates that MEMS should be designed to minimize displacement away from the substrate. On Die #3, some keepers were incorporated over 1000 micron long beams, resonators and one microengine. For example a Poly4 keeper acted as a bridge over some Poly2 and Poly3 beams. However, it was evident after ing that the keeper was too close to the free end of the beam and the large deflection of the beam allowed the free end of the beam to slip past the keeper. Proper orientation of a MEMS device in an application should ensure that the largest shock will cause the MEMS features to move toward the substrate, as in the compression orientation.

Orientat ion Shock Level (1000 g's) Temp (deg F) 4 by 4 6 by 6 8 by 8 10 by 10 12 by 12 165 100% 100% 100% 100% 100% 72.5 100% 100% 100% 100% 100% -65 100% 100% 100% 100% 100% 165 75% 88% 90% 95% 98% 72.5 100% 100% 100% 100% 100% -65 75% 80% 88% 88% 90% 165 95% 100% 100% 100% 100% 72.5 80% 75% 95% 98% 100% -65 85% 85% 91% 94% 94% 165 85% 90% 98% 98% 98% 72.5 64% 97% 98% 98% 100% -65 93% 100% 100% 100% 100% 165 38% 85% 93% 88% 95% 72.5 30% 65% 90% 98% 98% -65 58% 93% 93% 95% 95% 165 18% 36% 59% 68% 73% 72.5 18% 28% 38% 43% 40% -65 0% 10% 30% 38% 35% 165 0% 5% 10% 13% 8% 72.5 8% 11% 16% 15% 18% -65 0% 0% 0% 0% 0% 165 0% 0% 0% 0% 0% 72.5 0% 3% 5% 10% 10% -65 0% 0% 0% 0% 0% Figure 6. Survival percentages of cantilever beams based on anchor cut size Figure 4. SEM of broken Poly1Poly2 beams at anchors Figure 5. SEM of broken Poly3 beams at anchors FIELD TEST SETUP The field test was conducted at a hard target facility. A penetrator was shot into a hard target using a Davis gun as part of a demonstration of USAF Next Generation Penetrator Technologies. The penetrator was positioned for an 88 degree impact angle with an impact velocity of approximately 1000 ft/s. The shock environment included a launch acceleration of approximately 1200 g s for 20 milliseconds, a peak penetration load of 3000 g s for 5 milliseconds and a final rest acceleration of 2000 g s for 20 milliseconds. Shear 88 Penetrator Nose Test Item Inserts Figure 7. Schematic of MEMS carriers assembled into penetrator

The MEMS die were attached with epoxy to carriers with milled recesses. The carriers had recesses on the side for shear loading, on the top for compression loading, and the bottom for tension loading as shown in figure 7. The MEMS die were covered with Kapton tape, the carriers assembled into a housing, and the housing assembled into the penetrator nose. FIELD TEST RESULTS All devices were inspected before and after the tests. The process of bonding die to the carriers caused some damage which was noted. Figure 8 shows the results for the Die #1 and Figure 9 shows the results for Die #2. Two die were tested in each orientation of tension, compression, and shear. Cronos parts had nearly a 100% survival rate. However die C1 cleaved and half of the die detached from the fixture. The detached half contacted the tape and all parts were stripped. The other half, which remained adhered to the fixture, had minimal damage and operable actuators. Therefore the failure shown here is not so much a function of the shock, but a function of an inadequate bond between the die and the fixture. No beams were damaged due to the shock environment. Figure 10 shows a summary of results from SMM MEMS and DRIE MEMS as survival percentages. Shock Orientation Die number S1 Micro engine Thermal Actuator TRA curved teeth TRA straight teeth Comb drives fuctional, but no rotation S2 Shear Shear S3 S4 S5 S6 Inoperable, broken flexure Figure 8. Results of SUMMiT devices in field test Shock Orientation Die number C1 Thermal Actuator 1 Inoperable Thermal Actuator 2 Destroyed Worked for 20 seconds, then quit Worked for 20 seconds, then quit Initial movement, then quit Initial movement, then quit Bond pad damaged, but TRA operational Comb Resonator C2 C3 C4 Shear C5 Shear C6 Figure 9. Results of Cronos MUMPS devices after field test Inoperable Inoperable

SUMMiT V Microengine SUMMiT V Thermal Actuator MUMPS Thermal Actuator 1 Level DRIE 8 Kg Switch 2 Level DRIE Dual Environment Sensor Shear 100 0 100 100 100 100 100 100 0 0 Figure 10. Summary of DRIE and SMM MEMS devices after field test. 0 CONCLUSIONS These tests show that MEMS will fail at very high g levels due to a short, single axis shock pulse or at lower g levels due to a long duration, complex shock event. These tests have shown that MEMS devices are most susceptible to shock in tension and that orienting a MEMS device properly should be a primary design consideration. Also, a MEMS device should be limited in its displacement away from the substrate, which can be accomplished by robust keepers or innovative packaging. The DRIE devices were more sensitive to the field test environment than SMM MEMS. Large spring mass devices such as the microengine and inertial sensors are less robust in shock environments and during handling. Data from Figure 3 and 6 can be used as a guide for designing MEMS which will survive shock events. ACKNOWLEDGEMENTS The authors would like to acknowledge Ed Vernon for design and layout of the test die and the staff of the Microelectronics Development Laboratory for fabricating and releasing the SUMMiT test die and staff of the Compound Semiconductor Research Laboratory for fabricating and releasing the DRIE test die. Thanks to Fred Brown of the Mechanical Shock Lab for conducting the laboratory s, Mike Nusser for fixture design, and Joyce Zamora, Rosemarie Renn, and Darin Graf for inspecting and operating die before and after shock. We would like to acknowledge Robb Lee for designing the field test fixtures and David Koehler for documenting the results of the field tests. Also, we give credit to M. Barry Ritchey for the SEM images. Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy s National Nuclear Security Administration under Contract DE-AC04-94AL800. References 1. Duesterhaus, M. A., Bateman, V. I., Hoke, D. A. Shock Testing of Surface Micromachined MEMS Devices, Workshop on Nano and MIcrosystems Technology and Metrology, Redstone Arsenal, AL, December, 2002. 2. Bateman, V. I., Brown, F.A., Hoke, D. A., New Concepts in Mechanical Shock Characterization of MEMS Components, 72nd Shock and Vibration Symposium, Destin, FL, November 2001. 3. Bateman, V. I., Leisher, W. B., Brown, F. A., and Davie, N. T., "Calibration of a Hopkinson Bar With a Transfer Standard," Journal of Shock and Vibration, Vol. 1, No. 2, November-December, 1993, pp. 145-152.