Total Dose Effects on Microelectromechanical Systems (MEMS): Accelerometers*

Transcription

1 EEE TRANSACTONS ON NUCLEAR SCENCE, VOL. 43, NO. 6, DECEMBER Total Dose Effects on Microelectromecanical Systems (MEMS): Accelerometers* C.. Lee, A. H. Jonston, W. C. Tang, and C. E. Bames Jet Propulsion Laboratory California nstitute of Tecnology Pasadena, CA J. Lyke Pilips Laboratory Kirtland Air Force Base, NM Abstract Microelectromecanical sensors, ADXL5O and XMMAS40G accelerometers wic are fabricated wit surface micromacining tecniques are caracterized for teir total dose radiation response. Different failure mecanisms were observed wen te sensor element or te wole device was irradiated.. NTRODUCTON Microelectromecanical systems (MEMS) are new tecnologies wit a wide range of applications; automotive, biomedical instrumentation, and scientific. t is a tecnology tat combines electronics and micromecanical devices on a single substrate. Te growt of MEMS as attracted a broad range of researc and industry interests. Te most important application of MEMS, so far, is in te microsensor area, suc as pressure sensors, cemical sensors, and accelerometers. MEMS tecnology is particularly attractive to space systems. t promises to play a crucial role in future miniaturized spacecraft, suc as tose planned by te New Millennium Program (NMP). Te NMP is a new generation of miniaturized spacecraft for cost reduction and performance improvement. Tis paper presents total dose test results for two MEMS accelerometers, Analog Devices ADXLSO and Motorola XMMAS40G. 11. DEVCE DESCRPTONS AND EXPERMENTAL APPROACH Te Analog Devices ADXL5O is a microelectromecanical sensor wic measures up to f50 g acceleration on a single monolitic integrated circuit. As sown in Figure 1, it contains an oscillator, a voltage reference, a polysilicon surface-micro macined sensor, and signal conditioning circuitry. Te sensor element and electronic circuitry were fabricated wit a BiCMOS process. External resistors were used to set te gain and offset of te intemal buffer amplifier to provide 1.5 V to 3.5 V full scale output range. Te device sensitivity is 19 mv/g and te full scale output swing is f0.95 volts for +SO g acceleration. Te zero g output level of te pre-amplifier (Vpr) is +1.8 volts. * Te work described in tis paper was carried out by te Jet Propulsion Laboratory, California nstitute of Tecnology, under contract wit te National Aeronautics and Space Administration Code Q. Work funded by te NASAMicroelectronics Space Radiation Effects Program (MSREP). Self Test (ST) ;. Reference 4d Figure 1. Functional block diagram of ADXLSO. Vref out vout Figure 2 is a simplified scematic diagram of te ADXL5O sensor element at rest (0 g). t uses a differential capacitor sensor consisting of independent fixed plates and a movable floating central plate wic deflects in response to canges in relative motion. t is not a scaled top-view of te sensor element. Te actual structure of te sensor contains 42 unit cells and a common beam. Te outer two capacitors are series connected, forming a capacitive divider wit a common movable central Direction of movement 4-b r , Center Plate Top View J Ancor Teter Figure 2. A simplified diagram of te ADXLSO sensor at rest /96$ EEE

2 ~~ 3128 plate. Te polysilicon sensor as a tickness of 2 pm and te air gap between afixed and amovable electrode is 1.3 pm [2]. Wen an acceleration is applied, te common central plate (beam) moves closer to one of te fixed plates wile moving furter from te oter. Te output amplitude increases wit te amount of acceleration applied to te sensor. Te Motorola XMMAS4OG is a silicon capacitive, micromacined accelerometer. t consists of a single polysilicon seismic mass wic is suspended between two xed polysilicon plates (G-cell). Te 6-cell is sealed at te die level, creating a particle-free environment. Te full-scale range is +40 g and te sensitivity ofte device is 35 mv/g. Te device is fabricated wit a commercial CMOS process. A special test set-up employing a variable speed rotating arm was fabricated using a slip-ring assembly, dc motor, and oter equipment to measure te linearity of te sensitivity of te accelerometer. Te device under test was located at one end of a 10" rotating arm. Te output voltage, reference voltage, and Vpr were measured at various acceleration levels wit an HP345 SA precision DMM. Negative g-levels were measured by canging te device orientation 1 80" at te end of te 10" arm. Radiation tests of te entire device were performed using JPL CO-60 sources and ig energy (1 55 MeV) proton accelerator at ndiana University. Localized irradiation of only te sensor element was done wit a Scanning Electron Microscope (SEM) and low energy (5.5 MeV) proton accelerator at Caltec. a) Analog Devices ADXLSO Cobalt-60 Test Results 111. TEST RESULTS Tree devices were irradiated at a dose rate of 25 rad(si)/s wit te PL CO-60 room type irradiator in a static (nonrotating) condition (1 g was applied in tis static condition from te Eart's gravitational field). Normal electrical bias was applied. At levels between 5 and 20 krad(si) te device continued to operate wit a sligt sift in output voltage, causing small offset errors. At iger levels, an unusual failure mode developed, as sown in Figure 3. Te output voltage was initially stuck at a low voltage (50 mv). t remained stuck wen te device was accelerated to g-levels below 30 g. However, normal operation could be restored by accelerating te device to a sufficient ig g-level (approximately 30 g). Te devices remained functional from +50 g level to -50 g level as long as te device power supply was on. However, once te power supply voltage was removed, te output voltage reverted to te stuck mode. Normal operation could again be restored by accelerating te device to 30 g, but te device always returned to te stuck mode once power was removed. Tis suggests tat te device would no longer be functional in most applications unless very large accelerations were applied , krad(s1) after 30 g applied, maintaining 25 krad(si) Bias removed, ten applied, ; device started from rest...._... _....._. Figure 3. ADXLSO HDR [25 rad(si)/s] test result. As sown in Figure 1, te pin configuration of tis circuit allows separate measurements of te buffer amplifier. nput offset voltage, input bias current, and output voltage of te output buffer amplifier were measured, but insignificant degradation was observed after irradiation. Te 3.4 V reference voltage sowed no degradation at all. Electrical parameters of irradiated devices did not sow any indication of annealing after 24 ours at room temperature, after tey were irradiated to 25 krad(si). However, a device recovered fully after 144 ours at 100 "C. Tis device was irradiated furter to observe te failure mecanism at iger total dose level. Te device was functional until it was exposed to an additional 15 krad(si), 40 krad(si) total. t failed similarly as before, but in tis case it recovered witin 1 our of room temperature annealing. Te oter device was subjected to 168 our of 100 "C annealing period to observe any rebound effects. No rebound was observed. Te annealing test results imply tat tis BiCMOS accelerometer may witstand muc iger total dose irradiation at a low dose rate. A static functional test was performed using te self-test function. Tis test can be used to verify te functionality of te device statically wen te dynamic linearity test system is not available. Wen a logic ig voltage is applied to te self-test input, an electrostatic force is applied to te sensor so tat te sensor will be in a full-scale output state (50 g level). A correctly functioning accelerometer will respond by initiating about -1 volt output cange at Vpr. Te output will return to 0 g level wen te selftest input level goes low. Figure 4 sows test results of a control device and irradiated devices at 25 krad(si) using te self-test function. Te plot of Vpr and Vout of te control device (dotted line in Figure 4) indicates tat te device functions properly. Vpr canges from 1.8 V to 0.8 V andte outputjumps from 0 g-level voltage to +50 g-level voltage wen te self-test voltage is applied. Vpr of irradiated devices degraded significantly after 25 krad(si); 1.8 V of te initial value decreased to 1.5 V. However, te Vpr of bot irradiated devices still responded to te applied self-test input voltage. Te output of one device (solid line in Figure 4) sows a partial functionality wen te self-test input is ig (5 V), but it was non functional.'

3 ..""" SEA4 rradiation Result 3129 A SEM irradiation tecnique was used to irradiate only te sensor portion of te ADXL5O using a metod previously developed [3-51. Te dose rate of 1 krad(si)/s was calculated based on te area of te sensor wic was to be irradiated and beam current was set accordingly wit te 30 KeV beam voltage. 9 ov 5v ov 5v ov Voltage at SelGTest Pin Figure 4. Self-test functional test of control unit (dotted lines) and two failed devices; one device was partially functional (solid lines); te oter device was stuck low (dased lines). wen te selftest input is low (0 V). Te output of anoter device stuck at low (dased line in Figure 4) and did not responded to te self-test input at all at 25 krad(si). Te internal reference voltage and te buffer amplifier sowed insignificant degradation during total dose irradiation. However, te output voltage of te pre-amplifier, Vpr, degraded significantly wit radiation and sowed rapid annealing after ig-temperature annealing, altoug it did not anneal at room temperature witbias. Tese results are sown infigure 5. Note tat te pre-amplifier and te output buffer amplifier are identical op-amps. However, te sensor output is directly connected to te input of te preamplifier, wereas te input of te buffer amplifier is a user input to set te gain and offset of te transfer curve of te device. t is clear tat wen Vpr degrades and reaces a certain value, below 1.5 V, ten te device is non functional. Te 3.4 V reference voltage output canged less tan 1 mv after irradiation, as sown in Figure 5. After ig temperature annealing, Vpr values recovered to te initial values and devices functioned properly. Tis annealing beavior indicates tat electronic circuit degradation caused te Vpr-related failure after irradiation rater tan te electromecanical element \Vpr L i \ -v ~ 1.5- Higtemp '- rradiation annealing i T r 144r Total Dose [krad(si)] Figure 5. ADXL50 Vpr and Vref degradation (U 5 Te linearity curve of te device did not sow any nonlinearities up to te dose level of 50 krad(si). However, it sowed offset errors, in oter words, te curve sifted up along te y-axis as in Figure 6. Note tat te ysteresis like response wen te wole device was irradiated as sown in Figure 3 was not observed wen te only sensor was irradiated. Te linearity curve sifted up to te final dose level of 50 krad(si). L c s TD irradiation Acceleration (g) Figure 6. Results wit SEM localized irradiation [50 krad(si)] at sensor. Vpr also sowed significant degradation during SEM irradiation and it is sown in Figure 7. Vpr canged from V initial value to V at 50 krad(si) and te device was still ctional at te low value of Vpr unlike te total dose test results wen te wole device was irradiated. After 50 krad(si) of SEM irradiation, te Vpr was less tan 1 V wic means tat te self test function would fail at tis level of radiation. Vpr of 0.83 V canged to 0.13 V wen a logic ig was applied at te self test input. However, tis is less tan te electrical specification, wic requires tat Vpr be reduced by one volt. n tis case te Vpr reaced te saturation limitation and failed te self-test function test. However, te overall transfer curve siftedupward from te initial curve and te device remained functional as sown in Figure 6 even toug te self-test function test failed. Low Energy Proton Test Results (Localized rradiation) Te sensor element was locally irradiated to 100 krad(si) wit 5.5 MeV protons using te Pelletron accelerator at Caltec. Te lid of te device was removed and an alumum collimator wit a 20 mil diameter ole directly above te sensor was used to cover te device. Te accelerometer was statically biased and Vpr and Vout were monitored during proton irradiation. Unlike previous total dose test results, Vpr increased severely at 20

4 3130 A 9 W $ , e sensor was ir Device nonfunctional 0.9 ~~ Device functional at tis low Total Dose Frad(Si)] Figure 7. Comparison of results wit wole-cip and localized irradiation of te sensor. krad(si) as sown in Figure 8. Note tat Vpr decreased wit cobalt-60 irradiation wen te sensor was irradiated. Te reference voltage was measured as a calibration ceck and sowed insignificant degradation up to te final total dose level of 100 krad(si). Even after te Vpr increased to 2.46 V at 100 krad(si), te device remained functional. Te Vout vs. g-level transfer curve sifted down about 0.5V as sown in Figure 9. Tis is an opposite result to te total dose test result and SEM localized test results wic sowed te Vpr decreased and Vout transfer curve sifted upward. Te ysteresis effect tat was observed during total dose irradiation wit wole device was not observed wit tis localized sensor only irradiation wit protons ::.: rradiation TDE r Total Dose, krad(si) > Sensor only irradiatio 0000, Acceleration (8) Figure 9. ADXLSO Proton Test Results: Sensor only was irradiated to 100 krad(si). MeV proton up to 100 krad(si). t was irradiated to an additional 100 ad(%) wit ig energy protons, but now te entire device was irradiated, not just te sensor. For all tree devices, Vpr increased about 400 mv from te initial value and output voltages decreased to V from 2.49 V initial value after irradiation. Previous tests ad sown tat Vpr degraded in a similar way wen te sensor was locally irradiated wit 5.5 MeV protons. However, wen te entire device was irradiated te reference voltage degraded significantly, increasing to 4.0 V from 3.4 V initial value. Tis is te first test result tat sowed significant degradation in te reference voltage. Tis degradation is due to te proton displacement damage response of te bipolar reference voltage circuit of te device; no significant canges were observed in te reference voltage wen devices were irradiated wit cobalt-60 gamma rays. Te oter device tat was initially unirradiated beaved differently, as sown in Figure 10. Te output of sn 2 sifted upward, wit ysteresis-like effects tat were similar to results after gamma irradiation. However, after proton irradiation te device would only function at accelerations above 30 g, unlike te gamma results were te device would function at lower g Figure 8. Vpr and Vref degradation &er 100 krad(si) proton irradiation of te sensor element. Hig Eneray Proton Test Results Hig energy proton irradiation tests were conducted wit 155 MeV protons at ndiana University. Only one radiation level [ 100 krad(si)] was used. Two devices [serial numbers (sn) 1 and 21 were irradiated to 100 krad(si) wit 155 MeV protons and ad not been previously irradiated. Te sensor element of anoter device, sn 3, ad been previously irradiated wit t Acceleration (g) Figure 10. Proton irradiation results: Wole device was irradiated wit 155 MeV protons. Te sensor of sn 3 ad been previously irradiated to 100 krad(si) wit 5.5 MeV protons.

5 levels after it was accelerated to 30 g or more under bias (see Figure 3). Tis difference may be due to te different radiation levels [ 100 ad(%) wit protons, and a maximum level of 25 krad(si) wit gamma rays] rater tan inerent differences in te response of te device to te two different environments. Figure 10 also sows te results for sn 3, wic ad been previously irradiated wit 5.5 MeV protons. After te initial (localized) irradiation to 100 krad(si) te transfer curve of sn 3 sifted up, similar to te result after wole-cip irradiation for sn 2, but witout any ysteresis. Wen an additional 100 krad(si) was applied to sn 3 (irradiating te entire device) resulting in 200 krad(si) on te sensor), te transfer curve of sn 3 sifted down and te ysteresis-like response occurred. Low Dose Rate (LDR) Test Results Devices were irradiated at rad(si)/s and statically biased during irradiation. Te JPL LDR Co-60 room type irradiator was used for irradiation. Vpr increased about 58 mv and te reference voltage remained uncanged up to te total dose level of 25 krad(si) and did not display te ysteresis-like beavior tat occurred for devices irradiated at ig dose rate. Two different date codes were evaluated for LDR testing. Bot date codes, 9403 and 9602 remained functional up to 25 krad(si). Figure 11 compares transfer functions at 25 krad(si) at low and ig dose rate , "3 3 E \ 25krad(Si)w/ / 0.5, HDR(25 r/s) ', : ol ; 4 4 b t 9 - B Acceleration (g) Figure 11. Comparison LDR, rad(si)/s and HDR, 25 rad(si)/s tests results ofadxl5o (LDC9602) ' > 3 1 v 2 krad(si) 3131 Hig-temperature annealing caused / te output to be stuck at 27 mv. o! k,, no o O Acceleration (8) Figure 12. Motorola XMMAS40G HDR test result. canged from 4.86 V to 27 mv and remained nonfunctional after 168 ours 100 "C annealing. After te ig temperature annealing test, a peculiar postirradiation lockup condition was observedtor tis device. Wen a device was accelerated to about 40 g level and stayed about 20 seconds, te output jumped back to V wic indicates tat te device is nonfunctional. Te supply current increased from 3.4 ma to 112 ma. Te nominal supply current was about 3.5 ma trougout full-scale-output g-levels. Ten, te power supply was removed and reapplied for several seconds. Te output stayed at 27 mv and te supply current was 3.03 ma. Te power supply was turned on and off several times to make sure te output was in a stable mode. Again, te device was spun up to about 40 g level, ten, te device output increased to V and te supply current jumped up to ma. To clear tis mode, te power ad to be recycled. Te output retumed to 27 mv, but te device remained nonctional. Wen te device is in ig g-level mode, te device starts drawing a tremendous amount of current and te output is stuck at ig level. Tis appears to be an effect ofte micromecanical sensor and CMOS electronic components interaction. Devices were delidded for SEM analysis. None ofte devices functioned after delidding. Te die was too delicate to touc and none ofte delidding attempts seemed to work successfully. Terefore, localized irradiation of te sensor element was impossible for tis device, and it was not possible to perform additional diagnostic tests. j b) Motorola XMMS40G Accelerometer LDR Test Result HDR Test Result Tis CMOS fabricated device exibited small canges in te output g-level response at low total dose levels, and failed functionally at 4 krad(si) wit a dose rate of 25 rad(si)/s. Te output of te device stuck at ig level (4.86 V) as sown in Figure 12. Te JPL HDR CO-60 room type irradiator was used for irradiation. Devices did not sow any indication of annealing during 24 ours at room temperature. However, te output LDR test at rad(si)/s was performed wit te JPL Co- 60 room type irradiator. Results sowed tat te output of te sensor stuck at ig 4.86 V after 4 krad(si) of irradiation. Tis result is very similar to te ig dose rate result. However, LDR irradiated devices did not recover at 168 our room temperature and 100 OC annealing. Te results at LDR and HDR indicate tat extremely sensitive CMOS circuitry is responsible for lowlevel total dose failures compared to te ADXL5O BiCMOS device.

6 3132 V. DSCUSSON Te two microelectromecanical sensors tat are discussed in te paper sow unusual failure modes wen tey were subjected to ionizing radiation. Te BiCMOS device functioned at muc iger radiation levels tan te sensor tat is fabricated wit CMOS tecnology, altoug te two devices also use different sensor designs and device arcitectures. Localized irradiation of te sensor element in te ADZ50 sowed tat te sensor element is affected by radiation, as well as te periperal electronics. Tis may be caused by carge build up in te oxide under te polysilicon sensor structure. Results of total dose irradiation and SEM localized irradiation of te sensor indicate tat if only te sensor is irradiated te output transfer curve sifts up. However, wen te wole device was irradia~ed~ ten not only did te transfer curve sift up, but ysteresis effects were seen tat resulted in te sensor losing functionality unless it was accelerated to levels above 30 g wit bias applied. Te proton test results also indicates similar responses, but te transfer curve sifted down wen only te sensor element was irradiated. A ysteresis-like effectwas also seen after proton i~adiation, but te device only functional at levels above 30 g. Te failure mode of te CMOS device was also complex, altoug it failed at a far lower level tan te BiCMOS device at a ig dose rate. Low dose-rate tests didnot cange te failure level of te device. t seems tat te CMOS output circuitry of te device dominates te radiation failure tresold. Unfortunately, devices could not be delidded to test locally different sections of te device due to te extreme sensitivity of te internal sensor element. Te key issue for tese devices is to understand ow te EM sensor tecnology and device arcitecture affect te results. Tese tests suggest tat te interplay of te sensor and electronic devices affects bot part types. Because te sensor is loaded wit different dc voltages on te upper and lower capacitors, tis causes electrostatic forces wic are several times larger tan te mecanical restoring forces [6]. Te carging of te oxide under te sensor element can cange te capacitance wic affects Vpr and te output voltage of te device. Tis is consistent wit te upward or downward sifting of te transfer curve after irradiation. COTS devices tat do not contain MEMS elements. Localized irradiation of te sensor element sowed tat bot te sensor and te supporting electronics were affected by ionizing radiation, altoug tis may not be te case for all types of sensors tat can be used in MEMS devices. Te small size, low cost, and versatility of MEMS devices will likely be of great interest in future space systems. Furter studies and increased understanding of teir radiation failure mecanisms will be needed in order to successfully apply tem in space. However, te results in tis paper sow tat existing accelerometer tecnologies can be used at sufficiently ig levels to meet many space applications. V. ACKNOWLEDGMENT Te autors wis to tank M. Wiedeman for fabricating te caracterization test system and C. Swift for proton testing wit te Pelletron accelerator at Caltec. V. REFERENCES [l]. Analog Devices DataBook, "Monolitic Accelerometerwit Signal Conditioning: ADXL50". [2].W. Kuenel, ' Modellingoftemecanical beavior ofadifferential capacitor accelerationsensor," Sensors andactuatorsa48 (1995) [3] S Coen, H. Huges, "SEMrradiation for Hardness Assurance Screening and ProcessDefinition," EEETrans Nucl. Sci. U, 387, Dec [4]. C. G. Tomas, K. 6. Evans, "Te SEM as ADiagnosticTool for Radiation Hardness Design ofmicrocircuits," Scanning Electron Microscopy, 735,1978, Vol. 1. [5]. K. F. Galloway, P. Roitman, "mportant Considerations for SEMTotal Dose Testing," EEETrans. Nucl. Sci -,2066 (1977). [6]. W. Kuenel, S, Serman, "Asurfacemicromaciued silicon accelerometerwit on-cip detectioncircuitry," Sensors andactuators.445 (1994) V. CONCLUSONS Two different MEMS accelerometers were tested for sensitivity to ionizing radiation. Even toug te design of te sensor components was similar, te sensitivity to total dose radiation was quite different due to te tecnology difference of te supporting electronic circuits. Te sensor wit a BiCMOS structure integrated circuit (C) was more tolerant to total dose irradiation compared to te sensor fabricated wit a CMOS supporting C. Tis is similar to te results obtained for oter