Localized CO2 laser bonding process for MEMS packaging

Available online at www.sciencedirect.comsCIENCEs @oInEoT.Transactions of、Nonferrous MetalsSociety of ChinaScienceTrans. Nonferrous Met. Soc. China 16(2006)s577-s581Presswww.csu.edu.cn/ysxb/Localized CO2 laser bonding process for MEMS packagingSUN Li, A. P. MALSHE', S. CUNNINGHAM, A. MORRIS31. Department of Mechanical Engineering, University of Arkansas, Faytteville, AR 72701, USA; .2. wiSpry Inc., Irvine, CA 92618, USA; 3. wiSpry Inc., Cary, NC 27513, USAReceived 10 April 2006; accepted 25 April 2006Abstract: The packaging poses a critical challenge for commercialization of MEMS products. Major problems with the packagingprocess include degraded reliability caused by the excess stress due to thermal mismatch and altered performance of the MEMSdevice after packaging caused by thermal exposure. The localized laser bonding technique for ceramic MEMS packaging to addressabove-mentioned challenges was investigated. A continuous wave CO2 laser was used to locally heat sealing material for ceramicMEMS package lid to substrate bonding. To determine the laser power density and scanning speed, finite element analysis thermalmodels were constructed to simulate the localized laser bonding process. Further, the effect of external pressure at sealing ring on thebonding formation was studied. Pull testing results show that the scanning speed and external pressure have significant influence onthe pull strength at the bonding interface. Cross-sectional microscopy of the bonding interface indicates that the packages bondedwith relatively low scanning speed and external pressure conditions have higher bonding quality. This research demonstrates thepotential of localized laser bonding process for ceramic MEMS packaging.Key words: localized laser bonding; MEMS packaging; FEA; mechanical pull tests; cross-sectional microscopypackaging，most MEMS packaging accounts for1 Introductionexpensive roadblock for rapid commercialization ofMicro-Electro-Mechanical Systems (MEMS) arMEMS [2]. Another factor is the effect the package andmeso-to-submicron scale mechanical moving parts andpackaging process can have on the MEMS device. It isassemblies developed for sensing and actuation purposes.important to address the package and the MEMS devicesTheir diverse applications are in areas such as radioas a coupled design problem to address the thermalfrequency (RF) switches and antennas, biomedicalmismatch between package and MEMS, which can oftenfluidic drug delivery system and implants， opticalresult in reduced reliability [3, 4]. Another importantdisplays and switches, chemical reactors and fuel cells,consideration is the hermeticity capability of the packageetc [1]. According to its application requirements, everyrelative to an exchange of gas or moisture. A finalMEMS device needs an application specific interface,consideration is a change in the MEMS performancetypically called a package, to interact with applicationcaused by the thermal history of the packaging processenvironment. The package provides aspects as follows.[5].1) Physical support for mounting.To address the above-mentioned issues， local2) Interconnections from chip-to-physical supportheating and rapid heating approaches, such as resistiveand a window to the outside environment and datamicro-heaters [6], rapid thermal processing (RTP) [7],processing interface.and laser bonding [8- 10], have been developed for3) At the same time, protection from the outsideMEMS packaging. Particularly, laser processing is ofworld.interest due to [1] the following aspects.4) Waste management (heat, fluid, stray light).1) Tight control on the time and temperature win-These attributes clearly distinguish MEMS dow for processing.packaging from the traditional IC packaging .2) Non-con中国煤化工Due to the requirement of application specific3) Processit:YHCNM HGCorresponding author: A. P. MALSHE; Tel: +1-479-575-6561; Fax: +1-479-575-6982; E-mail: apm2@engr.urak edu.s578SUN Li, et al/Trans. Nonferrous Met. Soc. China 16(2006)4) Opportunity to select various suitable wavel-Table 1 Material properties used for FEA modelingengths, depending upon the wafer materialsDensity/SpecificThermalneat/conductivity/5) Region selectivity.(kgm 3)(Jkg-l.k)(W-m K)6) Fast manufacturing.KyoceraIn this research work, the localized laser bonding3.70x 1035024.0A476 (lid)process for ceramic MEMS packaging was investigated.This process has been demonstrated for CQFP packagingNCO-120RB 1.16x 1031 000).5in our previous research [12]. In this paper, the material(Sealant)system was extended, and finite element analysis (FEA)thermal models were constructed to understand the effectA44075014.0of laser processing parameters on the temperature(Substrate)distribution. Further, more processing parametersincluding laser power density, scanning speed, and2) Laser power density was applied as a heat flux ofexternal pressure were studied and discussed for thelocalized laser bonding process of ceramic MEMS18 MW/m2.3) Scanning speed was set to be 0.5 mm/s (This waspackages.used to simulate the moving of the sample stage. It was2 FEA thermal modelingrealized by applying the heat flux on a sequence ofelements according to the process time).4) The temperature of bottom surface of the packageIt is key to control temperature to form bonding andto minimize the effect of packaging process (heating) onwas fixed at 30A representative modeling result is shown in Fig.2.he MEMS deviceand change in the MEMSperformance caused by the thermal history of theIt can be seen that the temperature at the sealing layer ispackaging process. In this research, the temperaturearound 180 °C, a little higher than the glass transitiondistribution during bonding was mainly controlled by thepoint 150 C. The highest temperature on the lid islaser power density and the scanning speed. A FEAaround 360 C. The high temperature region is confinedthermal model was constructed withANSYS 9.0 toin a small area, which is favorable for minimal thermaleffect on the MEME device inside the package.understand the effect of laser power density and thescanning speed on the packaging temperature. Fig.IANillustrates the thermal model for heat transfer analysis of4 mmx4 mm ceramic packages. Packaging materialproperties used in the modeling are listed in Table 1.EpoxyAl2O3 lid3066578 1031515.133 121292.4628266 42259.919Fig.2 Temperature distribution on package during localizedlaser bonding processAl2O3 substrateTo understand the thermal history of the package, aspecifc node (node 1) was taken to record theFig. 1 FEA thermal model for heat transfer analysistemperature changed with time, as shown in Fig.3. It canbe seen that temperature changes fastly with time, on theIn the thermal analysis, the whole model was set toorder of a few seconds. This is desirable to reduce the22 C as an initial condition. The boundary conditionsinfluence of packaging process on the MEMS device.were defined as follows.The modeling results provide a guideline for1) Outer surfaces exposed to air were treated to beselection of the laser bonding parameters and assistinsulated (Based on the theoretical calculation, energyoptimization of中国煤化工: selection ofloss through convection and radiation was less than 3%the parameters (YHC N M H Gf temperatureof the energy input by the laser. Therefore, convectiondistribution at the sealing layer, t1me requrrement for theand radiation were ignored in the model).materials to form bonding, thermo-mechanical stress.SUN Li, et al/Trans. Nonferrous Met. Soc. China 162006)s579 .of the laser beam. The sample stage was mounted on aacomputer controlled speed rate screw driven X-Ytranslation stage (Velmex Inc., make MN10 bislideassembly).Node iMirrorLaser systems Lens positionig/micrometer300 rb)|盟Shround/ensprotection gas260Samplestage、NozzleStepperX-Y translation220 tmotor'talbe80 tFig. 4 Schematic diagram of experimental setup140 t4 Results and discussion100一2345Time/s4.1 Mechanical pull testingFig.3 Temperature history at node 1Pull tests using a material strength testinginstrument (The Quad Group Inc., SEBASTIAN FIVE-A)after bonding, and temperature distribution/history at the .vere performed to measure the bonding strength.devices. More parameter studies for laser bonding ofAccording to the FEA modeling results, a few trials ofMEMS packages are in progress.laser bonding were conducted to determine the range oflaser bonding parameters. Then specific parameters were3 Experimentalapplied for bonding the ceramic packages. The results ofpull tests are summarized in Table 2.The ceramic packages used in this research workwere from wiSpry Inc. with material properties specifiedTable 2 Pull strength of laser bonded packages with variedin Table 1. The lids were 4.00 mmX 4.00 mmX0.25 mmprocessing parametersin dimension with a 30 um thick epoxy layer. TheLaserScanningPassesExternal Bondingpowerofsubstrates were 4.00 mm x 4.00 mmX 1.10 mm in dimens-densityspeedpressure/ strength/laserion._(MW.m3) (mms )bondingMPaA continuous wave CO2 laser was used for the laser.01.31bonding process. The laser (model 3 080) was from54.00.79Preco Inc. with a wavelength of 10.6 μum. The theoretical1.27focused spot size of the laser beam is calculated to be0.11.050.530.34 mm in diameter. A piece of plexiglass was exposed0.130.56to the laser beam at the focal distance for a short20.66exposure duration to confirm the calculation. The0.2.55diameter of the burmned region on the plexiglass was2.30approximated to be the focused laser beam spot size.Using this method, the laser beam focused spot size is .It can be seen that the pull strength of laser bondedmeasured to be around 0.5 mm in diameter at the focalpackages form three distinct groups. When no extemaldistance.pressure is applied on the sealing ring, the bondingFig.4 shows the schematic diagram of thestrength is around 1 MPa. When external pressure isexperimental setup. The output from the laser wasapplied (by putting potassium bromide, which is nearlydelivered to the sample using an optical beam delivery100% transparent to the 10.6 μm wavelength infaredsystem. The beam was focused on the package using alight, on the pa中国煤化工ed is 2 mm/s,25.4 mm diameter planoconvex lens with an efecivethe bonding streYH ICNMH(Ethe last group,focal length of 254 mm. A power meter (Ophir Ine,the bonding stvhen externalmodel Laser Star) were used to check the incident powerpressure is applied and scanning speed is 0.3 mm/s. Thiss580SUN Li, et a/Trans. Nonferrous Met. Soc. China 16(2006)observation shows that the external pressure andscanning speed have significant effects on the bondingstrength of the packages. Comparing the bondingLidstrength, it is evident that applied external pressure andlow scanning speed are preferable in the curent setup toobtain higher lid to substrate bonding strength of theEpoxyceramic packages.Cavity ?4.2 Cross-sectional microscopyTo understand the reason for varied pull strength oflaser bonded packages with different processingSubstrate20μmconditions，the packages were cross-sectioned andpolished with a polisher (Buehler,Inc，modelFig.6 Bonding interface of packages processed at 22 MW/m2, 260-1950-160) to reveal the bonding interface. Themm/s, 4 passes, and 0.13 MPa external pressure applied oncross-section was then observed under a microscopesealing ring(Nikon Inc., model EPIPHOT" 200). Fig.5 shows arepresentative cross-sectional picture of packages bondedspeed/acceleration further， low scanning speed wasapplied. Fig.7 shows a representative cross-sectionalwithout external pressure applied on the sealing ring.picture of packages processed with an external pressureof0.13 MPa and scanning speed of 0.3 mm/s.Cavity IFig.5 Bonding itrface of packages processed at 12 MW/m, 320 ummm/s, 4 passes, and no external pressure applied on sealingFig.7 Bonding interface of packages processed at 15 MW/m,ring0.3 mm/s, 1 passes, and 0. 13 MPa external pressure applied onIt can be seen that the bonding between epoxy andsubstrate is partially formed after laser processing.It is evident that bonding is formed between epoxyHowever, unbonded areas still presente forming cavitiesat the bonding interface. These unbonded areas areand substrate, which results in a higher bonding strengthstrength limiting points and thus need to be removed. To(around 2.5 MPa) in this case. This observation alsoimprove the bonding quality， external pressure wasconfirms that the low scanning speed is favorable toapplied. Fig.6 ilustrates a cross-sectional view of aimprove the bonding quality.package processed with an external pressure of0.13 MPa5 Conclusionsand scanning speed of 2 mm/s.Compared to the case in Fig.5, the gap betweenepoxy and substrate is much larger after the externalThe localized laser bonding process for ceramicpressure is applied, which is consistent with the lowerMEMS packaging was established. A FEA thermalpull strength (around 0.6 MPa) in this case. It is believedmodel was constructed to simulate the laser bondingthat the reason for formation of the big gap is theprocess and provide guideline for experimental work.acceleration and deceleration of the sample stage. WhenCeramic packages were sealed using a CO2 laser with apotassium bromide is placed on the lid, relativeset of combinations of processing parameters includingmovement between lid and substrate or uneven load mayform when the package is moving with a relatively highexternal pressu中国煤化工processingacceleration, which resultes in the undesirable big gap atparametersMHC NMHG tests andbonding interface. To investigate the effect of scanningcross- sectional microscopy. It is concluded that scanning.SUN Li, et al/Trans. Nonferrous Met. Soc. China 16(2006)s581Performance[A]. Proceedings of SPIE-The International Society forspeed and external pressure have significant influence onOpticalEngineering International Symposium onthe bonding strength at the bonding interface. LowMicroelectronics[C]. Boshton: The International Society for Opticalscanning speed and applied external pressure arEngineering, 2003, 5288: 407- -411.preferable for high quality bonding.REBEIZ G M, RF MEMS Theory, Design, and Technology[M].Hoboken: John Wiley & Sons Inc, 2003.[6] LIN L. MEMS post-packaging by localized heating and bonding[J].AcknowledgmentsIEEE Transactions on Advanced Packaging, 2000, 23(4): 608- 616.This work is sponsored by WiSpry Inc. and[7] CHIAO M, LIN L. Hermetic wafer bonding based on rapid thermalECS-NSF (0501597). The experiments were conductedprocessing[]. Sensors and Actuators, A: Physical, 2001, 90(3):398- 402.n Material and Manufacture Research Laboratories8] WILD M J, GILLNER A, POPRAWE R. Locally selective bonding(MRL), University of Arkansas.of silicon and glass with laser[J] Sensors and Actuators A: Physical,2001, 93: 63-69.] LUO C, LIN L The application of nanosecond-pulsed laser weldingReferencestechnology in MEMS packaging with a shadow mask[]. Sensors andActuators A: Physical, 2002, 97- 98: 398- -404.[lDRESSENDORFER P V, PETERSON D A, REBER C A. MEMS10] MESCHEDER U M, ALAVI M, HILTMANN K, LIETZAU Ch,packaging-current issues and approaches[A]. Proceedings 2000 HDNACHTIGALL Ch, SANDMAIER H. Local laser bonding for lowIntemationgl Conference on Hiah.Densitv Interconneet and sustemstemperaturebudetrn. PackaingICLRetstn:IAs2000 4217.:208-2135temperatre budgetJ]. Sensos and Actuators A: Physica[2] MAMALSHECP, O"NEALC,sB,BROWNWD,EATONW.[] TAOMALSHE A R, BROWN w D. Seletive bonding andp, MILLER w M. Challenges in the packaging of MEMS[].encapsulation for wafer-level vacuum packaging of MEMS andIntermational Journal of Microcircuits and Electronic Packaging,related micro systems[]. Microelectronics Reliability, 2004, 44(2):1999,9 2(3):233 -241.251- -258.[3] MERCADO L L, LEE T I, KUO s M, LEE R. Process-induced[12] TAO Y, MALSHE A p, BROWN w D, DEREUS D R,thermal effect on packaging yield of RF MEMS switches[A].CUNNINGHAM s. Laserasted sealing and testing for ceramicAmerican Society of Mechanical Engineers, EEP, v 2, Electronic andpackaging of MEMS devices[J]. IEEE Transactions on AdvancedPhotonic Packaging, Electrical Systems and Photonics Design andPackaging, 2003, 26(3): 283- -288.Nanotechnology[C]. New Orleans: ASME, 2002, 25- -32.(Edited by ZHAO Jun)[4] DE SILVA A. P, LIU L, HUGHES H G Impact of Thermal CyclesDuring the Packaging Assembly Process on RF-MEMS Switch中国煤化工MYHCNMH G.