UHF Micromechanical Compound-(2,4) Mode Ring Resonators With Solid-Gap Transducers

Transcription

1 L.-W. Hung, Y. Xie, Y.-W. Lin, S.-S. Li, Z. Ren, and C. T.-C. Nguyen, UHF micmechanical cmpund-(,4) mde ing esnats with lid-gap Tansduces, Pceedings, 006 IEEE Int. Fequency Cntl Symp., Geneva, Switzeland, May 9-June 1, 007, pp UHF Micmechanical Cmpund-(,4) Mde Ring Resnats With Slid-Gap Tansduces Li-Wen Hung and Clak T.-C. Nguyen Depatment f Electical Engineeing and Cmpute Science Univesity f Califnia at Bekeley Bekeley, CA 9470 USA Yuan Xie, Yu-Wei Lin, Sheng-Shian Li, and Zeying Ren Depatment f Electical Engineeing and Cmpute Science Univesity f Michigan Ann Ab, MI USA Abstact UHF vibating micmechanical ing esnats with slid-filled dielectic tansduce gaps (as ppsed t pevius ai gaps) peating in a cmpund-(,4) mde have been demnstated at MHz with Q s n the de f 3,100 and mtinal esistances effectively 4.7 smalle than ai gap cuntepats unde identical bias cnditins. Due t thei highe dielectic cnstant, substitutin f slid dielectic mateials f ai vacuum in the electde-t-esnat gap leads t inceased electmechanical cupling in capacitively tansduced micmechanical esnats, which in effect geneates me utput cuent, and thus, educes mtinal esistance. The advantages f using slid dielectic mateial as the gap ae multifld, as it (1) eliminates the need f the final gap elease etch in the MEMS fabicatin pcess, theeby enhancing yield; () lwes device impedance vesus ai gap enditins, allwing easie matching t the cmpnents; (3) eliminates the pssibility f paticles getting int an electde-t-esnat ai gap, which wuld thewise pse a ptential eliability issue; and (4) cnslidates the esnat stuctue, making it less susceptible t shck. This wk geatly extends the fequency f diect (as ppsed t indiect) slid-gap tansduced MEMS esnats, fm the pevius 60 MHz using a cmpund (,1) mde, t nw nealy 1 GHz, and in a ange desied f RF fnt ends, using a cmpund (,4) mde. B (a) W B Anch W v i Anch A R ut A Suppt Input Beam Electde A A R ut Silicn Nitide Output Electde B B i Output Electde i R L R L Index Tems micelectmechanical devices, micesnat, slid-gap tansduce, impedance, quality fact, high fequency. (b) v i Input Electde I. INTRODUCTION This wk was suppted by DARPA. Fig. 1: Pespective-view schematic f (a) an ai-gap micmechanical extensinal wine-glass ing esnat; and (b) a slid-dielectic vesin, bth in thei typical bias and excitatin cnfiguatins. Nte that in bth cases, each utput electde is cnnected t its cespnding inne electde. Electdes A and A, electdes B and B ae usually electnically cnnected t seve as input and utput electde, espectively. Plysilicn capacitively-tansduced vibating micmechanical esnats have s fa psted the highest GHz-ange n-chip Q s, in excess f 15,000 at 1.46 GHz [1]. Hweve, thei elatively high impedances still encumbe them in the eyes f RF designes, wh geneally lk t match t the ppula 50Ω f mainsteam cnventin. Given the lss advantages f systems based n highe matching impedances, whee matching impedances much lage than paasitic esistances allw lwe veall lsses than lwe impedance systems, it wuld nt be supising if 50Ω systems ae sn eplaced by highe impedance nes, especially as me cmpnents mve n-chip, whee impedance cnventins n lnge matte. Still, if impedances d ise, they likely will nt ise past the single-digit kω ange, beynd which paasitic capacitance can limit the system bandwidth if nt inductively cancelled thewise suppessed. Thus, althugh 50Ω will likely nt ultimately be necessay, a eductin f the impedances f GHz capacitively-tansduced micmechanical esnats dwn t the kω ange is desiable. One effective methd f lweing impedance entails the use f slid-dielectic mateial in the electde-t-esnat gap t incease the pemittivity, theeby incease bth the dive fce and the utput cuent f a given device, pvided the Q and amplitude f mtin ae nt educed t much in the pcess. Tw appaches t slid-gap capacitive tansductin have been demnstated peviusly: ne whee diect lateal tansductin is used, whee lateal fces induce lateal vibatins []; and ne whee an indiect vetical-t-lateal mechanism is emplyed, whee vetical fces squeeze a stuctue t induce a lateal esnance mtin [3]. T date, the latte has achieved the lwest fequency-impedance pducts, n the de f (809.1MHz) (59Ω) = MHz-Ω

2 Oiginal Ring Shape Quasi-Ndal-Pints h E f = (1) π ρ(1 σ ) Fig. : ANSYS-simulated esnance vibatin mde shape f an extensinal wine-glass mde ing esnat. when using HfO as the dielectic mateial. The fme has s fa nly been demnstated using silicn nitide dielectic mateial, s has nt achieved as lw an impedance, but has achieved the highest Q s, in excess f 0,00 at 160 MHz. T futhe exple the efficacy f the lateal slid-gap appach at nea GHz fequency, this wk applies lateal nitide slid gaps t a UHF vibating micmechanical ing esnat peating in a cmpund-(,4) mde, peviusly dubbed the extensinal wine-glass mde in [4], and cmpaed with an ai-gap vesin in Fig. 1. The demnstated device vibates at MHz with Q s n the de f 3,100 n pa with pevius ai-gap vesins, and with mtinal esistances R x effectively 4.7 smalle. Althugh this eductin fact is smewhat sht f expectatins, and the demnstated device culd benefit fm the use f highe-k dielectic mateials t incease the eductin fact, this wk still geatly extends the fequency ange f diect (as ppsed t indiect) slidgap capacitively tansduced MEMS esnats fm the pevius 160 MHz using a cmpund (,1) mde [] t nw nealy 1 GHz, and in a ange desied f RF fnt ends, using a cmpund (,4) mde. II. STRUCTURE AND OPERATION OF THE EXTENSIONAL WINE GLASS RING RESONATOR The cmpund (,4) mde ing esnat f this wk, shwn in Fig. 1 and fist demnstated in [4], is cmpsed f a μm-thick ing stuctue with an inne adius f 36.8μm and an ute adius R ut f 50μm. Fu pais f electdes suund the inside and utside f the ing, with each electde pai ccupying ne f fu ing quates defined by suppt beams attached at the quasi-ndal pints f the cmpund-(,4) mde, shwn in Fig.. In this mde shape, the widths f tw diagnal ing quates expand alng ne axis, while the the tw cmpess alng the thgnal axis. T effect a slid-gap capacitive tansduce, the electde-tesnat gaps ae filled with slid silicn nitide, depsited 8-nm thick via lw-pessue chemical vap depsitin (LPCVD) at 800 C. The esnance fequency f the extensinal wine-glass ing with slid electde-t-esnat gaps geneally ends up being vey clse t that f its ai-gap cuntepat. Igning the effect f -induced electical stiffness [5], the esnance fequency f f an extensinal wine-glass mde ing esnat is given by whee ρ, σ, and E ae the density, Pissn ati, and Yung s mdulus, espectively, f the ing stuctual mateial, and h is a paamete that depends (amng the things) upn the inne and ute adius, and R ut, espectively, f the ing, shwn in Fig. 1, and n the de f the vibatin mde shape. The esnance fequency f an extensinal wine-glass ing is mst accuately specified by (1). But cnsideing that the extensinal wine-glass mde is cmpised lagely f expansin and cntactin abut the ing width, which is simila t the lngitudinal vibatin f a ba, the esnance fequency can be appximately specified by f appx m E =, m = 1,3,5, L,. () W ρ whee W = (R ut ) is the ing width, shwn in Fig. 1, and m is the de f the vibatin mde. The stng dependence in () n the lateal dimensin W means the fequency f the device can be specified via mee CAD layut, which in tun means that many diffeent fequency esnats can be fabicated nt a single chip in ne stuctual film a maj advantage ve many UHF esnats used tday, f which fequencies ae detemined pimaily by thickness, which cannt be specified via CAD layut. T excite the device, a dc-bias vltage is applied t the cnductive ing and an ac vltage v i t input electdes. Tgethe, these vltages geneate an AC electical fce that dives the device int esnance vibatin if the fequency f v i matches the esnance fequency f. Once vibating, the ensuing dc-biased time vaying electde-t-esnat capacits geneate utput cuents that, when pltted vesus the fequency f the input, tace ut the desied high-q bandpass biquad tansfe functin. III. EFFICACY OF A SOLID-GAP TRANSDUCER The mtinal esistance f a capacitively tansduced micmechanical esnat is geneally given by the expessin R x k = ω QV P C x k ω QV P 4 d ε ε A whee n the left hand side k is the effective stiffness f the disk; ω =πf is the adian esnance fequency; is the DC-bias; and C/ x is the change in esnat-t-electde capacitance pe unit displacement, which due t the nnunifm extensinal wine-glass mde shape in electde velap egins, equies integatin t accuately specify [6]. An appximatin t its value, hweve, can be btained by igning the mde shape, which then yields the ightmst expessin f (3), whee A and d ae the electde-t-esnat velap and gap spacing, espectively, f the wine-glass disk esnat; and ε and ε ae the pemittivity in vacuum and the elative pemittivity f the gap mateial, espectively. Equatin (3) indicates that thee ae multiple ways t educe R x. As cveed in [4], making the ing lage wuld in- (3)

3 Oxide Plysilicn Oxide Silicn Nitide Plysilicn Silicn Nitide Plysilicn n+ n+ n+ (a) Plysilicn Oxide (b) Silicn Nitide Plysilicn Oxide (c) Plysilicn n+ n+ n+ n+ (a ) (b ) (c ) Fig. 3: Css-sectins f the fabicatin pcess flw used f the slid-gap micmechanical extensinal wine-glass ings f this wk. cease the velap aea A, hence lwe R x, but the degee t which this can be dne is limited by the fact that lage ings have a highe tendency t esnate in unwanted spuius mdes. Inceasing the dc-bias vltage effects a squae-law eductin in R x, but the value f is eventually bunded by pull-in effects [5] dc supply ails, which ae ften limited by system cnsideatins. Reducing the disk-t-electde gap d, is appealing because R x depends n gap distance d t the futh pwe. The minimum achievable gap distance, hweve, is limited by the fabicatin pcess technlgy. In additin, a smalle gap spacing ften esults in a smalle pull-in vltage, whee the esnat pulls in and shts t an electde if the applied dc-bias vltage is t lage. In the face f these limitatins, the use f a dielectic mateial t aise the gap pemittivity ε ve that f ai ( vacuum) pesents an attactive methd f educing R x. Fm (3), R x clealy deceases via ε, which means it can decease faste than the electde-t-esnat velap capacitance C inceases as lng as n the paametes in the R x expessin cause an ffsetting incease in its value. The paametes mst likely t cunteact the deceasing actin f ε and aise R x ae: the Q, which can degade due t inceased dissipatin in the slid dielectic laye its intefaces with the adjacent disk and electde sidewalls; and the effective stiffness k tt, which inceases if intductin f the slid-gap attenuates the vibatin amplitude f the esnat t a value belw the ai-gap case. Fm (3), if the decease in mtinal esistance R x attained via intductin f a slid gap des nt utpace the incease in velap capacitance ( me imptantly, the decease in capacitive impedance (ω C ) -1 ), the design f sme applicatins becmes difficult. F example, single-ended filtes cnstucted in all cnductive mateials (e.g., all dped plysilicn) suffe fm passband disttin if paasitic cuents feeding thugh the C s fm input t utput ise t magnitudes simila t that f the mtinal cuent. Thus, f such all-cnductive single-ended capacitively-tansduced micmechanical filtes, a useful metic that gauges whethe nt feedthugh-deived passband disttin is a pblem can be defined as the ati between the mtinal cuent magnitude and that f the feedthugh cuent when thee is a diect capacitive path fm the input t the utput f the filte, which in its simplest fm can be witten as 1 FOM = (4) ω C R F a slid-gap t be beneficial ve an ai-gap capacitive tansduce, (4) shuld incease upn intductin f the slidgap, which means that the mtinal cuent inceases faste than the feedthugh cuent. It shuld be nted that the chice f (4) as a metic gauging the efficacy f a given capacitive tansduce, althugh cnvenient and lgical f sme applicatins, is nt univesally elevant. In paticula, as shwn in [7], a diffeential mechanical filte is much less susceptible t velap capacitance C than its single-ended cuntepat, since it allws f cmplete cancellatin f C via inductive esnance. In additin, the feedthugh issue is eally nly pblematic f single-ended filtes cnstucted in all cnductive mateials, since the feedthugh path can be bken by simply using a nncnductive mateial smewhee in the filte stuctue, such as dne in [10], whee the cupling links wee made nncnductive. Futheme, sme scillat designs, such as the ppula Piece cnfiguatin, ely n velap capacitance t pvide phase shifts f ppe peatin. Thus, althugh (4) is suggested hee as an inteesting metic f sme singleended filtes, the R x eductin fact is pehaps still the best metic by which t gauge slid-gap efficacy. IV. FABRICATION PROCESS AND PRACTICAL ADVANTAGES OF SOLID-GAP DEVICES The slid-gap extensinal wine-glass ing esnats f this wk wee fabicated using the pcess flw f [8], which is summaized via the css-sectins f Fig. 3. The fabicatin pcess stats with n+ dping f a silicn wafe t fm a gund plane that helps t suppess feedthugh duing testing [8]. This is fllwed by subsequent lw-pessue chemical vap depsitins (LPCVD s) f μm xide and 300 nm nitide as electical islatin layes. Next, vias f electical cntacts t the n+ laye ae dy etched thugh the nitide and xide in places nt shwn in the css sectin. A fist laye f plysilicn is then depsited, n+ POCl 3 -dped, and pattened, t fm electical intecnnects and device-specific gund x

4 Anch h = thickness Oute Electde Netwk Analyze Signal Geneat V DC = R ut f RF Bias-T V DC =0 Bias-T f LO Inne Electde Suppt Beam f = f RF -f LO Fig. 4: SEM f a fabicated extensinal wine-glass ing esnat. planes. Successive depsitins f 750 nm f a sacificial xide, μm f stuctual plysilicn t late fm the esnat stuctue, and 1. μm f xide t seve as an etch had mask, ae then depsited via LPCVD. (The plysilicn is POCl 3 - dped t n+ befe depsiting the xide had mask.) The xide had mask is pattened t the gemety f the plysilicn ing via lithgaphy and dy etching. Vias f the stems that anch the ing suppt bases t the substate ae als etched duing this step, using the same mask, which effectively selfaligns the anch stems t the ing. The ing and stem via pattens ae then tansfeed t the plysilicn ing via etching, using the xide as a had mask. At this pint in the pcess, the css-sectin lks as in Fig. 3 (a) and (a ). T define the electde-t-esnat gap, a thin film is next depsited cnfmally ve the whle wafe, and mst imptantly alng the sidewalls f the ing, t a thickness equal t the desied electde-t-esnat gap spacing. F an ai-gap device, this laye is LPCVD high tempeatue xide (HTO) depsited at 90 C, which can be emved alng with all the the sacificial xides duing the hydfluic acid (HF) elease step at the end f the pcess. F the pesent slid-gap devices, the depsited film is silicn nitide, which stays intact duing the HF elease step. Afte fmatin f sidewall spaces, a thid laye f plysilicn is depsited μm-thick, n+ POCl 3 -dped, and pattened t fm inne and ute electdes, afte which the pcess css-sectins lk as in Fig. 3 (b) and (b ). The stuctues ae finally eleased in the afementined HF elease step, whee again, the sidewall space is emved f ai-gap devices, but stays intact f slid-gap devices, leaving the final css-sectin shwn in Fig. 3 (c) and (c ). The main diffeence in the fabicatin sequence between ai-gap and slid-gap devices is simply that the electde-tesnat gap mateial must be emved f ai-gap, but nt f slid-gap. This tuns ut t be a maj advantage f slid-gap devices, and pssibly the mst imptant ne. Specifically, the mst difficult step in a small lateal ai-gap pcess is emval f the sacificial sidewall space mateial that defines the gap, since HF might have difficulty getting int the tiny gaps, and etch by-pducts might blck it fm getting all the way thugh. The smalle the ai-gap, the me difficult is the elease pcess, hence the lwe the yield f functinal devices. In cntast, the descibed slid-gap pcess has the distinct advantage f nt equiing a gap elease, Spectum Analyze V DC = 0 Bias-T Cmmn Gund Stage s it nt nly allws a highe yield, but can als achieve much thinne slid-filled gaps, n the de f 8nm vesus the 65nm ften used f ai-gap devices. In additin, finished slid-gap devices ae me bust than ai gap devices, since they ae bette stabilized against shck, and they d nt suffe fm gap cntaminatin issues, whee paticles mistue might get int the electde-t-esnat gap. V. EXPERIMENTAL RESULTS Vacuum Chambe Fig. 5: Pictial desciptin f the mixing methd used f measuement f slid-gap wine-glass ing esnats. Bth slid-gap and ai-gap (f cmpaisn) devices wee fabicated using apppiate vesins f the pcess flw in Fig. 3. Fig. 4 pesents the SEM f a cmpleted 980-MHz extensinal wine-glass ing. A. Measuement Stategy As mentined abve, the inceased pemittivity f a slidgap capacitively tansduced esnat geneates significantly lage electde-t-esnat velap capacitance C than typically pesent f ai-gap devices. Althugh the C s f the diect lateal devices f this wk ae geneally much smalle than thse f vetical-t-lateal types [3], they might still encumbe measuement f devices if the mtinal esistance R x eductin attained via use f slid-gaps is smalle than the decease in (ω C ) -1. In diffeential devices [7], this capacitance can be esnated ut via an inductive me functinal LC-based fequency shaping cicuit. F such diffeential mechanical cicuits, lage values f C can even be beneficial, since this wuld pemit the use f smalle esnating inductance values, such as the 3-5 nh cmmnly available n-chip in cnventinal CMOS technlgy. F single-ended devices, hweve, inductive esnance cancellatin is nt as simple t implement, since finite intecnnect esistance pevents an added induct fm esnating ut all the C capacitance [7]. Thus, f the pesent efft, athe than esnate ut ptentially lage slid-gap C s, it is simple and me cnvenient t just sepaate the mtinal cuents f main inteest fm the paasitic feedthugh cuents induced by C. In this wk, the mixing methd f [9], shwn in the

5 Pwe[dBm] Ai-gap (,V LO ) =(7,8) V 1000 (,V LO ) =(5,4) V Fequency [MHz] 100 Data = 3 μm R ut = 50 μm h = μm d = 65 nm = 7 V V RF = 0.63 V V LO = 8 V f = MHz Fig. 7: Measued fequency espnse f a MHz ai-gap micmechanical wine-glass ing esnat. schematic f Fig. 5, is used t sepaate mtinal cuents fm feedthugh cuents in the fequency dmain. Hee, the applicatin f a lcal scillat (LO) t the esnat stuctue allws the esnat t be diven via a adi fequency (RF) signal that mixes with the LO via nnlineaity in the capacitive tansduce, theeby geneating a fce cmpnent at the diffeence fequency, ω RF -ω LO. When ω RF -ω LO =ω, the esnance fequency f the ing, the ing vibates at ω, geneating a mtinal cuent at ω widely sepaated fm feedthugh at the RF and LO fequencies. This then allws a spectum analyze peating in MAX HOLD mde t detect esnance mtinal cuents all by thei lnesme, withut intefeence fm feedthugh. Resnat fequency chaacteistics btained in this manne ae thus quite clean, allwing a vey accuate deteminatin f the degee f R x eductin pvided by slid-gap tansduces, plus a vey accuate measuement f the actual mechanical Q f the device unde test. It shuld be nted that mixing measuement is nly a chaacteizatin methd. In an actual applicatin, unless the device in questin is a mixle [10], devices ae nt peated in this way. In paticula, a filte utilizing esnats like these must be ppely teminated and measued diectly, nt via mixing, since this is hw it wuld be used in the actual applicatin. F this wk, hweve, mixing is quite apppiate f slid-gap esnat chaacteizatin, since it yields much me accuate esults. B. Measuement Results Each measued sample was pepaed by fist attaching a die cntaining devices nt a custm-built test bad and bnd wiing device pads t apppiate metal intecnnect taces n this bad. T enable vacuum measuements, the test bad was then inseted int a custm-built vacuum chambe that allwed cnnectin f test bad I/O pts thugh the vacuum inteface t measuement instumentatin n the utside. Fig. 6 and Fig. 7 pesent fequency chaacteistics measued in vacuum using the mixing set-up f Fig. 5 f ai-gap and nitide-gap vesins, espectively, f extensinal wineglass ings, like that shwn in Fig. 4. Because the utput in the measuement setup is sensed by a spectum analyze in MAX HOLD mde, these fequency plts epesent nly the aw utput f the esnat pwe attenuated by the impedance mismatch between the esnat and the 50Ω input f Pwe [dbm] Slid-gap (,V LO ) =(5,4) V Fequency [MHz] Data = 3 μm R ut = 50 μm h = μm d = 8 nm = 5 V V RF = 0.63 V V LO = 4 V f = MHz Fig. 6: Measued fequency espnse f a slid-gap MHz micmechanical wine-glass ing esnat. the spectum analyze. This explains the dbm units n the y- axis. The smewhat vey lw values f dbm ae caused by a athe lage impedance mismatch between the measued devices and 50Ω. The tw esnats measued hee ae identical, except the ne in Fig. 6 has a 65 nm ai gap, and the ne in Fig. 7 has a 8 nm nitide gap. Immediately appaent is that thei esnance fequencies, at MHz and MHz, espectively, ae quite clse, and this facilitates the pedictin f fequency f the slid-gap device. Inteestingly, the 3,160 Q f the slid-gap device is actually highe than the,00 f the ai-gap device smething that pesently eludes igus explanatin, but which might be attibutable t simple pcess-elated vaiance in device Q s. Me data n these devices ae needed t detemine statistical chaacteistics, such as mean and standad deviatin, befe any cncete cnclusins can be made, hee. Hweve, the fact that the Q f the slid-gap device is n pa with that f the ai gap, is at least encuaging. Hweve, the Q less than 10,000 f bth, althugh sufficient f pe-select filtes, is nt sufficient f RF channel-selectin [11], s is smewhat disappinting. Futhe study is needed t islate the cause f the lwe Q, which quite pssibly might just be a peculiaity f this specific pcess un. F fai cmpaisn f mtinal esistance, the data f the ai-gap device, iginally taken with =7V and V LO =8V, was adjusted using knwn (and pven) dependencies n dcbias and lcal scillat amplitude t allw a diect cmpaisn unde the same bias cnditins between the tw devices, with =5V and V LO =4V. Afte the adjustment, a decease in slid-gap R x (vesus ai-gap) f 4.7 is seen, which is gd, but disappinting, since this is smalle than the pemittivity incease f 7.8, meaning that f this paticula 980-MHz device, the C capacitance inceased faste than the mtinal esistance R x deceased when the nitide gap was intduced, making an all-plysilicn single-ended vesin f this device me susceptible t feedthugh than its ai gap cuntepat. Even wse, the FOM value is f the ai gap device, and f the nitide-gap device, futhe indicating that the intductin f the nitide-gap was nt beneficial fm a pefmance pespective, at least f all-cnductive singleended devices. Thus, at least in this instance, the intductin f slidgap capacitive tansduces did nt impve the FOM. In fact,

6 if an ai-gap device with the same 0 nm gap as the slid-gap device wee achievable, they pedicts that the ai-gap device wuld fa utpefm the measued slid-gap ne, in bth R x and FOM. It appeas that the need t cmpess and expand the slid nitide gap laye has a me encumbeing effect n the pesent devices than n pevius wine-glass disks [], as might be expected when cnsideing that the ing fequencies ae quite diffeent fm the wine-glass disks, s the wavelength-ptimized gap-electde thickness needed t maximize the edge-t-edge displacement f the slid-gap is diffeent. Althugh an FOM impvement was nt bseved, hee, the the benefits f slid-gap tansductin wee still seen, such as highe fabicatin yield and geate mtinal cuent (than the achievable 65 nm ai-gap vesin). F diffeential patially nn-cnductive mechanical filtes [7], and f scillats, these ae me imptant than the FOM, s the use f nitide-gaps des pve quite useful f these applicatins. T attain even geate eductins in R x, gap mateials with much highe pemittivity than silicn nitide ae attactive, such as HfO Baium Stntium Titanate (BST). Wk t incpate these is in pgess. VI. CONCLUSIONS UHF vibating micmechanical ing esnats with slid-filled dielectic tansduce gaps (as ppsed t pevius ai gaps) peating in a cmpund-(,4) mde have been demnstated at MHz with Q s n the de f 3,100 and mtinal esistances 4.7 smalle than ai gap cuntepats unde identical bias cnditins. This wk geatly extends the fequency f diect (as ppsed t indiect) slid gap tansduced MEMS esnats, fm the pevius 60 MHz using a cmpund (,1) mde, t nw nealy 1 GHz, and in a ange desied f RF fnt ends, using a cmpund (,4) mde. Hweve, the pefmance enhancement affded by slid-gap capacitive tansduces in this wk was smewhat less than desied, since the eductin fact attained f mtinal esistance was smalle than the fact by which the electde-t-esnat velap capacitance inceased, and this can be pblematic f sme single-ended (as ppsed t diffeential) applicatins f micmechanical esnats. T emedy this, wk twads ptimizing the slid-gap mateial type and thickness t allw maximum gap mateial expansin and cntactin is nging. Acknwledgment: This wk was funded by DARPA. REFERENCES [1] S.-S. Li, Y.-W. Lin, Y. Xie, Z. Ren, and C. T.-C. Nguyen, Micmechanical hllw-disk ing esnats, Pceedings, IEEE Int. MEMS Cnf., Maasticht, The Nethelands, Jan. 5-9, 004, pp [] Y.-W. Lin, S.-S. Li, Y. Xie, Z. Ren, and C. T.-C. Nguyen, Vibating micmechanical esnats with slid dielectic capacitive tansduce gaps, Pceedings, Jint IEEE Int. Fequency Cntl/Peciisn time & time Inteval Sympsium, Vancuve, Canada, Aug. 9-31, 005, pp [3] H. Chandahalim, d. Weinstein, L.-F. Chew, and S. A. Bhave, Channel-select micmechanical filtes using high-k dielectically tansduced MEMS esnats, Pceedings, IEEE Int. Cnf. Mic-Elect- Mechanical Systems, Istanbul, Tukey, Jan. -6, 006, pp [4] Y. Xie, S.-S. Li, Y.-W. Lin, Z. Ren, and C. T.-C. Nguyen, UHF micmechanical extensinal wine-glass mde ing esnats, Technical Digest, IEEE Int. Electn Devices Meeting, Washingtn, DC, Dec. 8-10, 003, pp [5] H. Nathansn, W. E. Newell, R. A. Wickstm, and J. R. Davis, J., The esnant gate tansist, IEEE Tans. Electn Devices, vl. ED- 14, N. 3, pp , Mach [6] Y.-W. Lin, S. Lee, S.-S. Li, Y. Xie, Z. Ren, and C. T.-C. Nguyen, Seies-esnant VHF micmechanical esnat efeence scillats, IEEE J. Slid-State Cicuits, vl. 39, n. 1, pp , Dec [7] S.-S. Li, Y.-W. Lin, Z. Ren, and C. T.-C. Nguyen, An MSI micmechanical diffeential disk-aay filte, Tech. Digest, Dig. f Tech. Papes, the 1 4h Int. Cnf. n Slid-State Senss & Actuats (Tansduces 07), Lyn, Fance, 007, pp [8] J. Wang, Z. Ren, and C. T.-C. Nguyen, GHz self-aligned vibating micmechanical disk esnat, IEEE Tans. Ultasn., Feelect., Feq. Cnt., vl. 51, n. 1, pp , Dec [9] J. Wang, J. E. Butle, T. Feygelsn, and C. T.-C. Nguyen, 1.51-GHz plydiamnd micmechanical disk esnat with impedancemismatched islating suppt, Pceedings, MEMS 04, Maasticht, The Nethelands, Jan. 004, pp [10] A.-C. Wng and C. T.-C. Nguyen, Micmechanical mixe-filtes ( mixles ), IEEE/ASME J. Micelectmech. Syst., vl. 13, n. 1, pp , Feb [11] C. T.-C. Nguyen, MEMS technlgy f timing and fequency cntl, IEEE Tans. Ultasn., Feelect., Feq. Cnt., vl. 54, n., pp , Feb. 007.