xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 22 Ka 4500
controlfield tag 007 cr-bnu---uuuuu
008 s2010 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0004631
Development of uhf micromechanical resonators and arrays based on silicon-on-insulator (soi) technology
h [electronic resource] /
by Mingke Xiong.
[Tampa, Fla] :
b University of South Florida,
Title from PDF of title page.
Document formatted into pages; contains X pages.
Thesis (MEE)--University of South Florida, 2010.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
ABSTRACT: A novel micromachining technology on SOI substrates is presented that is capable of producing on-chip high-Q resonators and resonator arrays equipped with high aspect-ratio (30:1) microstructures and nano-gap capacitive transducers filled with high-k dielectrics. The newly developed IC-compatible MEMS microfabrication process consists of merely three standard photolithography steps, which is much simpler than the other SOI-based resonator device technologies. In order to achieve the optimum performance and yield of the resonators and resonator arrays, this SOI-based fabrication process has been carefully designed and investigated step by step. For capacitively-transduced extensional mode (e.g., radial-contour and wine-glass mode) resonators, formation of nano-scale capacitive gaps and large resonator-to-electrode overlap area is essential for reducing the motional resistance Rx and DC bias voltage by strengthening the capacitive transduction. Atomic Layer Deposition (ALD) technology with superb conformability and uniformity as well as outstanding thickness controllability is used to deposit the ultra-thin layer (~10 nm) of high-k dielectric material that acts as the solid capacitive gaps, which allows the mass production of on-chip capacitively-transduced resonators and resonator arrays with greatly enhanced electromechancial coupling coefficient, and thus lower motional resistance and DC bias voltage. Using this technique, high-Q micromechanical resonators and resonator arrays on SOI substrates operating at ultra-high frequencies (UHF) have been developed. The ultimate goal of this project is to implement on-chip narrow-band micromechanical filters with unprecedented frequency selectivity and ultra-low insertion loss. By fine-tuning the nonlinear characteristics of the capacitive transducers enabled by the new SOI technology, novel on-chip mechanical signal processors for frequency manipulation, such as mixer and multiplier, will be investigated.
Advisor: Jing Wang, Ph. D
x Electrical Engineering
t USF Electronic Theses and Dissertations.
Development of UHF Micromechanical Res onators and Arrays Based on Silicon-OnInsulator (SOI) Technology by Mingke Xiong A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Electrical Engineering Department of Electrical Engineering College of Engineering University of South Florida Major Professor: Jing Wang, Ph.D. Andrew M. Hoff, Ph.D Thomas M. Weller, Ph.D Date of Approval: March 20, 2010 Keywords: ALD, capacitive transducer, DRIE, high-k, nano-gap Copyright 2010 Mingke Xiong
To my fiance, my parents, my friends, and my teachers.
ACKNOWLEDGEMENTS I would like to thank the members of my dissertation committee for their always helpful advice and suggestions: Prof. Jing Wang (Chair), Prof. Thomas Weller, and Prof. Andrew Hoff. Without their s upport, I could not ha ve finished this long journey. I would especially like to thank Prof. Wang for his motivation, guidance, and inspiration throughout the whole master prog ram. Working with him duri ng the past three years has been an extremely valuable and enjoyable experience. I would also like to thank the former and current students from the group, including Mian Wei, I-Tsang Wu, Cesar Augus to Morales, Julio Mario Dewdney, Kosol Son, Ivan Rivera, Tianpeng Wu, Paula Alga rin amaris, Kenza Mouttaki, and Chamila Siyamba. This thesis work would not be possible without the the constant help and support from Nanomaterial and Nanotechnology Research Center (NNRC): Robe rt Tufts, Richard Everly, Jay Bieber, Yusuf Emir ov, and Sclafani Louis-Jeune. Most importantly, I would like to thank my fianc, Ke Sun, my parents, Lan Yu and Chujiang Xiong for their love and support.
i TABLE OF CONTENTS LIST OF TABLES Â…Â… ... Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â….iii LIST OF FIGURES ........................................................................................................... iv ABSTRACT .................................................................................................................... vii CHAPTER 1 INTRODUCTION ........................................................................................ 1 1.1 Background of Wirele ss Transceiver Architecture .................................................. 3 1.1.1 Superheterodyne Architecture ......................................................................... 4 1.1.2 Direct Conversion Architecture ....................................................................... 5 1.1.3 Architecture Based on RF MEMS Resonator .................................................. 6 1.2 Macro-scale Vibrating Mechanical Resonator for Wireless Communication ......... 7 1.2.1 Ceramic Device ................................................................................................ 7 1.2.2 Quartz Crystal Device ...................................................................................... 8 1.3 Micromechanical Resonato rs for Wireless Communication ................................... 9 1.3.1 Piezoelectri cally-Actuated Re sonators .......................................................... 11 184.108.40.206 Surface Acoustic Wave (SAW) Resonators .......................................... 11 220.127.116.11 Bulk Acoustic Wave (BAW) Resonators .............................................. 12 1.3.2 Capacitively -Transduced MEMS Resonators ................................................ 16 18.104.22.168 Flexural Mode Beam Resonator ............................................................ 16 22.214.171.124 Radial-Contour Mode Disk Resonator................................................... 18 126.96.36.199 Wine-glass Disk Resonator .................................................................... 21 188.8.131.52 Wine-glass Mode Ring Resonator ......................................................... 22 184.108.40.206 Internal Di electrically-Transduced Bar Resonator ................................ 25 1.3.3 Resonator Based on Silicon-on-Insulator (SOI) Technology ........................ 26 220.127.116.11 Fabrication on SOI Substrate Utilizing Electron Beam Lithography .... 27 18.104.22.168 Fabrication on SOI Substrate Utilizing Fo cus Ion Beam (FIB) Technique .............................................................................................. 28 22.214.171.124 SOI Fabrication by Conventional Si Microm achining Technique ........ 30 1.4 Resonator Array ..................................................................................................... 34 1.5 Capacitively-Transduced Resona tors Using Materials Other than Si .................... 35 1.6 Overview ............................................................................................................. ... 36 CHAPTER 2 RESONATOR DESIGN ............................................................................. 39 2.1 Extensional Wine-glass Mode and Resonance Frequency Design ........................ 39 2.1.1 Wine-glass Mode of a Disk Resonator .......................................................... 39 2.1.2 Wine-glass Mode of a Ring Resonator .......................................................... 46 2.2 Equivalent Circuit Model ....................................................................................... 50
ii CHAPTER 3 MICROFABRICATION PROCESS OF THE WINE-GLASS MODE DISK RESONATOR AND ARRAY ON SOI SUBSTRATE ................... 56 3.1 Microfabrication Process on SOI Substrate ........................................................... 56 3.2 High-Aspect-Ratio Si DRIE .................................................................................. 60 3.2.1 Etch Rate ..................................................................................................... ... 64 3.2.2 Sidewall Smoothness ..................................................................................... 66 3.2.3 Aspect Ratio Dependent Etching (ARDE) .................................................... 67 3.3 Atomic Layer Deposition (ALD) ........................................................................... 68 CHAPTER 4 CONCLUSION........................................................................................... 73 LIST OF REFERENCESÂ…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â….................................................. 75 APPENDICESÂ….......Â…Â…Â…Â…Â…Â… Â…Â…Â…Â…Â…....Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…........ ... 83 Appendix A: Polysilicon Disk Resona tor on SOI Substrate Process Traveler............ 84
iii LIST OF TABLES Table 1.1 Properties for different material s and their relevant frequencies. ..................... 36 Table 2.1 The related paramete r for a disk resonator....................................................... 46 Table 3.1 Different Si DRIE recipe used in this work. ..................................................... 62 Table 3.2 Precursors, deposition condition and rate of HfO2 deposited by ALD. ............ 70
iv LIST OF FIGURES Figure 1.1 Simplified architecture of a supe rheterodyne receiver with single downconversion. ........................................................................................................ 4 Figure 1.2 Simplified architecture of direct conversion receiver ...................................... 5 Figure 1.3 Simplified architecture of RF MEMS resonator-based cha nnel select receiver. ............................................................................................................. 6 Figure 1.4 Mode shapes of quartz crystal uni t: (a) contour shear mode; (b) thickness shear mode. ........................................................................................................ 8 Figure 1.5 Schematic of a typical SAW resonator . ................................................... 11 Figure 1.6 Device implementation for the tw o types of bulk acoustic wave (BAW) resonators with two different types of acoustic isolation methods : (a) SMR employs BraggÂ’s reflector; and (b) FBAR sits on top of air cavity. ...... 13 Figure 1.7 Contour mode ring resonator : (a) one port circular ring resonator; (b) one port square-shape ring resonator. ........................................................ 14 Figure 1.8 (a) Photo of an AlN dual-mode filter; (b) measured transmission (S21) of the AlN dual-mode filter, along with simulated response based on its equivalent circuit model with differe nt termination impedance . ............. 15 Figure 1.9 Schematic view of different types of beam resonators: (a) cantilever beam resonator; (b) clamped-clamped beam resonator; (c) freefree beam resonator. ......................................................................................................... 17 Figure 1.10 SEM image and measured frequency re sponse in (a) vacuum and (b) air for a polysilicon capacitively-transduced radial-contour mode disk resonator ................................................................................................... 18 Figure 1.11 Comparison between (a) a pr evious disk resonator process; (b) the self-aligned disk resona tor process . ......................................................... 19 Figure 1.12 Fabrication process flow of th e self-aligned radial-contour mode disk resonator. ....................................................................................................... 21 Figure 1.13 SEM photo and frequency response spectrum in (a) air and (b) vacuum of a polysilicon wine-glass mode disk resonator . .................................. 22 Figure 1.14 Perspective-view schematic of th e wine-glass mode di sk resonator in a typical two-port bias and ex citation configuration. ....................................... 22 Figure 1.15 Perspective-view schematic of the extensional wine-glass ring (EWGR) resonator with typical driving and sensing configuration. ............................ 24
v Figure 1.16 SEM picture and measured frequenc y characteristics of a (a) un-notched and (b) notched Â“hollow diskÂ” EWGR device . ...................................... 25 Figure 1.17 Picture of a bulk-mode resona tor and the measured frequency characteristics at different vibrating mode . ............................................ 26 Figure 1.18 Fabrication proce ss using e-beam lithography fo r creating suspended NEMS device on SOI substrate. .................................................................... 28 Figure 1.19 Schematic of the process of MEMS re sonator with nano-gap utilizing FIB milling technique: (a) thermal oxide is grown on SOI wafer; (b) photolithography is used to open the rele ase hole; (c) nano-scale gaps are achieved by FIB milling; (d) patt erns are transferred to SiO2/Si/SiO2 layers by high aspect-ratio dr y etch; (e) metal contacts are patterned by lift-off process after striping the top SiO2 layer; (f) the resonators are released by etching away the buried oxide. ................................................... 29 Figure 1.20 Bulk lateral res onator with narrow air gap (<100 nm) fabricated by the proposed FIB-based process . ................................................................. 29 Figure 1.21 (a) SEM image and (b) frequency response of an 18 m-thick wine-glass disk resonator on SOI substrate . ............................................................ 30 Figure 1.22 Fabrication proce ss flow of HARPSS resonator on SOI substrates. ............. 31 Figure 1.23 SEM image of (a) the disk resonator and (b) a zoom-in on the 200 nm gap . ......................................................................................................... 32 Figure 1.24 Fabrication process flow of the TEOS and CMP based method. .................. 33 Figure 1.25 SEM image and frequency response spectra of a single and mechanically coupled square resonator arrays with three and five resonators, respectively . ........................................................................................... 34 Figure 1.26 Schematic of a 3-by-3 disk re sonator array on SOI substrate and its measuring configuration. ............................................................................... 38 Figure 2.1 Top view of a wine-gla ss mode disk resonator. .............................................. 40 Figure 2.2 Mode shapes and resonance frequency for a 20 m radius single crystal silicon <100> disk calculated from th e theoretical derivation using COMSOL Multiphysics 3.5a with (a) m = 2; (b) m = 3; (c) m = 4; (d) m = 5. ......................................................................................................... 45 Figure 2.3 Top view of a wine-gla ss mode ring resonator. .............................................. 46 Figure 2.4 The two-port electrical circuit model represented by Y -parameters. ............... 50 Figure 2.5 An infinitesimal element d along the circumferential direction ................ 52 Figure 2.6 Electrical equivalent circuit m odel for a two-port disk resonator. .................. 55 Figure 3.1 Cross-section view of a single wine-glass disk re sonator on SOI substrate.... 56 Figure 3.2 Cross-section view process flow of a wine-gla ss disk resonator on SOI substrate. .......................................................................................................... 57
vi Figure 3.3 SEM, cross-section, and 3-D schematic view after the 1st lithography step and Si DRIE to define the resonator body structure. ....................................... 58 Figure 3.4 SEM, cross-section and 3-D view after the 2nd lithography step and Si etch to define the electrodes. ................................................................................... 59 Figure 3.5 SEM image, cross-section and 3-D view of the final device. ......................... 60 Figure 3.6 Schematic view of Bo sch process principle. ................................................... 61 Figure 3.7 Schematic view graph of the two-step Bosch process. .................................... 62 Figure 3.8 A simplified model for the Bosch process: in one deposition/etch cycle, the deposition step lasts for a period of t1, and the etch step for t2, which includes both polymer removal and Si isotropic etch. .................................... 63 Figure 3.9 SEM photos of DRIE Si sidewalls (a) before and (b) after oxygen plasma treatment. ......................................................................................................... 64 Figure 3.10 Si etch time vs. (a) source power, (b) SF6 pulse time, and (c) C4F8/O2 pulse time....................................................................................................... 65 Figure 3.11 SEM photos of Si side wall scalloping formed by di fferent Si DRIE recipes: (a) original recipe A; (b) recipe B with reduced source power; (c) recipe C with reduced SF6 pulse time; (d) recipe D with increased C4F8/O2 pulse time. ........................................................................................ 66 Figure 3.12 Overall Si etch rate decrea ses with increasing etch time. ............................. 67 Figure 3.13 Si etch rate decreases dramati cally as aspect ratio increases. ....................... 68 Figure 3.14 Schematic concept of ALD process. ............................................................. 68 Figure 3.15 SEM image of a thin film of HfO2 of ~ 50 nm deposited by ALD on Si substrate. ........................................................................................................ 69 Figure 3.16 Etch rate in HF of HfO2 deposited at different temperature and with different post-deposition treatment. ............................................................... 70 Figure 3.17 XRD spectra of ALD HfO2 films on thermal oxide underlayer showing the effect of deposition te mperature and annealing. ...................................... 71 Figure 3.18 HF C-V measurement of a MOS capacitor with ALD HfO2 film as dielectric. ....................................................................................................... 72
vii Development of UHF Micromechanical Res onators and Arrays Based on Silicon-OnInsulator (SOI) Technology Mingke Xiong ABSTRACT A novel micromachining technology on SOI subs trates is presented that is capable of producing on-chip highQ resonators and resonator arra ys equipped with high aspectratio (30:1) microstructures and nano-gap ca pacitive transducers filled with high-k dielectrics. The newly developed IC-com patible MEMS microfabrication process consists of merely three standard photolithogr aphy steps, which is much simpler than the other SOI-based resonator de vice technologies. In orde r to achieve the optimum performance and yield of the resonators and resonator arrays, this SOI-based fabrication process has been carefu lly designed and inves tigated step by step. For capacitively-transduced extensional mode (e.g., radial-contour and wine-glass mode) resonators, formation of nano-scal e capacitive gaps and large resonator-toelectrode overlap area is essential for reducing the motional resistance Rx and DC bias voltage by strengthening the capacitive tr ansduction. Atomic Layer Deposition (ALD) technology with superb conformability and uni formity as well as outstanding thickness controllability is used to depos it the ultra-thin layer (~10 nm) of high-k dielectric material that acts as the solid capacitive gaps, wh ich allows the mass production of on-chip capacitively-transduced resonators and re sonator arrays with greatly enhanced
viii electromechancial coupling coefficient, and thus lower motional re sistance and DC bias voltage. Using this technique, highQ micromechanical resonators and resonator arrays on SOI substrates operating at ultra-high fre quencies (UHF) have been developed. The ultimate goal of this project is to implemen t on-chip narrow-band micromechanical filters with unprecedented frequency selectivity and u ltra-low insertion loss. By fine-tuning the nonlinear characteristics of the capacitiv e transducers enabled by the new SOI technology, novel on-chip mechanical signal pr ocessors for frequency manipulation, such as mixer and multiplier, will be investigated.
1 CHAPTER 1 INTRODUCTION Ever since David E. Hughes introduced the concept of electromagnetic waves in a signal system eight years before Hertz' s experiments, wireless communication has gradually become one of the most exciting ar eas and has started to play an increasingly significant role in our everyday lives. T oday, wireless communication technology makes impact on our lives in all as pects, such as cellular tele graphy, satellite communication, broadcasting, wireless sensor network, a nd a lot more . Particularly, these achievements have led to a low-cost, powe r-efficient wireless communication system as well as the overwhelming boost of personal wi reless communication de vices like cellular phone, Global Positioning System (GPS), Pers onal Digital Assistan t (PDA), portable PC, so on so forth. According to statistics released by Global System for Mobil Communications (GSM), the number of globa l GSM subscribers has surpassed 3 billion in 2008 and will continue to grow by over 10% annually. However, the overly crowded spectrum, whic h is fully occupied by a wide variety of wireless communication standa rds, has imposed a big cha llenge in reception of the desired signal among substantial amount of inte rferences in adjacent frequencies. Hence, on-chip and fully integrated devices with be tter band selection are needed in order to catch up with the fast development of wireless communi cation system. In order to satisfy the stringent speci fications for communication standards, especially those based on traditional supe rheterodyne architecture, a number of high quality ( Q )-factor mechanical components are re quired for precise frequency generation
2 and selection. So far, tremendous efforts have been devoted to using alternative transceiver architectures, such as direct c onversion (zero-IF) , low-IF [3, 4], and RF sampling down-conversion , which rely on hi gher levels of transistor integration to minimize the need for highQ passives at RF front-ends. Un fortunately, performances of them are not ready yet to compete with th at of their counterparts in traditional superheterodyne architecture, thus trumpeti ng a need for on-chip replacements of the highQ RF passives (e.g., filters, resonators, etc.). Recent development in the radio frequency micro-electro-mechanical systems (RF MEMS) technology has attracted a great deal of attention from both academia and industries, which holds great promises to potentially revolutionize the entire regime of the wireless technology by bringing togeth er microelectronics and micromechanical elements. As such, the implementation of comp lete wireless transcei vers on a single chip could lead to a viable solution to many cu rrent issues and challenges in present-day wireless communications. Particularly, due to their orders-of-magnitude smaller size as compared to traditional off-chip passives (e.g., quartz crystal, ceramic s, etc.), the next generation of wireless trans ceivers equipped with RF MEMS components can be realized with greatly enhanced performance. A CMOS-compatible MEMS technology has been demonstrated lately that enables alternat ive communication architecture by facilitating the integration of highQ passive devices with active transistor electronics, allowing great size reduction, lower power consumpti on and enhanced performance . Among the various types of MEMS devi ces and applications, capacitivelytransduced micromechanical resonators have ob tained most interests due to their UHF to SHF operation frequencies and ultra-high Q -factors exceeding 10,000 [7-11]. In addition
3 to their superb frequency selectivity, the use of capaci tive transduction makes the resonators operating under electric charge poten tials with no dc current flow thus no dc power consumption. Moreover, due to the fact that the resonance frequency is defined by the lateral dimensions, capaci tively-transduced resonator offers CAD layout-definable frequencies. Other benefits include better thermal stability , higher frequency stability , better voltage-controlled tunability , better CMOS-compati bility, and selfswitching capability . 1.1 Background of Wireless Transceiver Architecture A transceiver is a device that consists of both a transmitter and a receiver sharing the same electronic circuitry. A transmitter modulates the baseband data and up-converts it into a carrier frequency with sufficient power amplification. Key parameters of the transmitter performance are the modulation accuracy, signal purity, and RF output power level. The main function of a receiver is to demodulate the desired signal from the presence of undesired interf erence and noise. Therefore, comparing to a transmitter which processes a locally avai lable strong signal, a receiver is much more challenging because of the requirement of high dynami c range and high out-of-band attenuation. Due to the increasing data traffic, the a ssociated signal bandwi dth is limited. As a result, selectivity becomes the most important characteristic of a receiver, which is defined by the capability of picking up the wanted signal while re jecting adjacent frequency interferers. Filters with high off-band attenuation ar e used to select the narrow band channels. The selectivity of a filter is determined by its quality factor, Q given by: BW f Q0 1.1 where f0 is the center frequency and BW the 3 dB bandwidth of the filter.
4 In the following sections, several most popular receiver architectures and a novel MEMS-based architecture proposed this work are discussed. 1.1.1 Superheterodyne Architecture Since invented by Edwin Armstrong in 1917, superheterodyne architecture has still been used within a majority of wirele ss systems. As illustra ted in Figure 1.1, the desired signal received by an antenna passes through a pre-select ba ndpass filter, a low noise amplifier (LNA), and then an image -reject filter to remove the out-of-band interference as well as the imag e frequency. The selected RF signal is then converted to an intermediate frequency (IF) signal by mi xing with a local oscillator (LO) signal generated by a voltage controlle d oscillator (VCO). A channel-select filter is used to assign the desired channel and reject all the in-band interference. This is followed by an analog-to-digital converter (ADC ) and a digital signal proce ssor (DSP) that perform the demodulation and data decoding to provide the output baseband data, respectively. Figure 1.1 Simplified architecture of a supe rheterodyne receiver with single downconversion. As shown in Figure 1.1, highQ vibrating mechanical components, such as ceramics, quartz crystal, and surface acoustic wa ve (SAW) resonator are used to integrate with bandpass filters and oscillators. Filters utilizing such technologies successfully
5 extinguish themselves by outstanding quality factor, low inserti on loss, high percent bandwidth, and high out-of-band rejection . Oscillators also benefit from highQ because the phase noise decreases as Q increases. However, as the demand for highselectivity devices keeps on in creasing, quartz and SAW devices have gradually failed to satisfy the stringent highQ requirement. More importantly, current highQ devices are bulky, where off-chip components make the ultimate miniaturization of the wireless communication systems difficult. 1.1.2 Direct Conversion Architecture Figure 1.2 Simplified architecture of direct conversion receiver The motivation of ever increasing integrati on level has led to the invention of the direct conversion receiver, which is also known as Â“homodyneÂ” or Â“zero-IFÂ” [3, 4]. Figure 1.2 shows a simplified architecture of direct conversion receiver. As the local oscillator frequency in di rect conversion receiver is set equal to the RF frequency, the IF frequency becomes zero and the image freque ncy could be successfully eliminated. Hence, eliminating image-reject filter that is required in superhet erodyne architecture enables the reduction of the number of off-chip components. Nevertheless, several issues remain in todayÂ’s communication system. The LO -leakage due to the settings of the same RF and LO frequency results in a time-varying dc offset, and can lead to degradation of
6 the upper boundary of the dynamic range. T hus, complicated offset cancellation techniques are needed in practical implementations. 1.1.3 Architecture Based on RF MEMS Resonator Figure 1.3 Simplified architecture of RF MEMS resonator-based channe l select receiver. Considering the size and extra cost of the large quantity discrete highQ components, a technology that can realize the multi-channel selection on a single chip monolithic implementation will be highly desirable. Recent advances of CMOScompatible micro-electro-mechanical-system (MEMS) technology has made it possible to implement on-chip RF MEMS elements, which are able to not only reduce the size, cost and power consumption, but also achieve be tter performances (high frequency, highQ high dynamic range, sharp cut-off, etc.). Aside from direct replacement of the off-chip highQ passive devices, an RF channel-select ar chitecture has been demonstrated . Figure 1.3 presents the system block diagram for a newly-invented RF channel select receiver that takes full advantages of achievable complexity utilizing MEMS elements. CMOS-compatible micromechanical devices with highQ (>10,000) and high frequencies (>1 GHz) have been reporte d recently [7-10, 17, 18], pr oviding the potential of integration of wireless communication system. In addition to miniat urization, if channel selection is possible at RF carrier frequencies, succeeding electronic components such as
7 LNA and mixer are no longer needed to handle the power of alternate channel interference. Therefore, dynamic range can be greatly relaxed, allowing significant reduction in power consumpti on as well as the cost. 1.2 Macro-scale Vibrating Mechanical Resonator for Wireless Communication Nowadays, frequency-selective mechanic al components, such as ceramic and quartz crystal resonators, are needed for RF and IF bandpass filtering and local oscillator reference frequency generation. Outstanding perfo rmances have been achieved including low insertion loss, small percent bandwidth, sharp cut-off, and dynamic range. However, the most popular RF passives are all off-ch ip components that must interface with transistor at the board level, thus impos ing two main technical challenges in nextgeneration multiband wireless transceivers: size and cost. The following sections will focus on the status of current off-chip co mponents used in wireless communications and then explain the requirement for their replacement. 1.2.1 Ceramic Device Piezoelectric materials, such as barium titanate and Lead-Zirconate-Titanate (PZT), have been used in many fields since the discovery of piezoel ectric effect in 1880. Barium titanate ceramic is a good candidate fo r electromechanical transducers because of their high electromechanical coupling coeffi cient, ease of fabrication, and non water solubility. However, two key weakness of this material have greatly limited its further development, namely bad temperature coefficient and low Curie point. On the other since PZT ceramics are discovered in 1954, they have rapidly taken the places of barium titanate in most piezoelectric applications due to their high electromechanical coupling factor, good frequency-temperat ure characteristics, and su itable quality factor.
8 Various types of ceramics with high di electric constant and good temperature stability have been investigated and implement ed into practical dielectric filters over the past decades [19-22]. The dielec tric filters at frequencies up to 75 GHz have been widely used in wireless communication applications. However, ceramic filters have been facing the issue of Â“bulkyÂ” size as the increasing re quirement of miniaturization of wireless communication systems. 1.2.2 Quartz Crystal Device Figure 1.4 Mode shapes of quartz crystal unit: (a) contour shear mode; (b) thickness shear mode. Quartz crystal is made of single crystal silica that has piezoelectric properties. Quartz crystal has various vibration m odes according to its crystallinity and piezoelectricity. Parameters that determin e the resonance frequency are different depending on the vibration mode. As shown in Figure 1.4 frequencies of a contour shear mode resonator (CT cut, DT cut) and a thickne ss shear mode resonator (AT cut, BT cut) are determined by the length of one side of the square and the thickness, respectively. Some of the vibration modes, such as AT cut and GT cut crystal units, have zero temperature coefficient over a broad temperat ure range, and therefore these two crystal units have excellent frequencytemperature characteristics. Moreover, quartz crystal is extremely stable both physically and chemica lly Â– no significant fr equency change after
9 aging, which permits the widely use of quart z crystals with accurate frequency control , timing[24, 25] and filtration [26, 27]. In particular, among the off-chip components in wireless communication transducers, quartz crystal used in reference oscillators is the most difficult to be miniaturized and integrated on chip, since the its Q -factor is too high to be matched by current on-chip devices. The Q Â’s of on-chip elements such as on-chip spiral inductor for the LC tank, on conventional CMOS silicon are limited to less than 10, nowhere near the required high quality factor for reference oscillators . Hence, lots of efforts are focused on finding replacements for the highQ but bulky quartz crystals that dominate the market of reference oscillators. An ultrahighQ oscillator based on 60 MHz wineglass mode disk resonator using a hybrid co mbination of on-chip components (see section 126.96.36.199 ), has exhibited an oscillator phase noise of -110 dBc/Hz at 1-kHz offset from the carrier, and -132 dBc/Hz at far-f rom-carrier offsets. Astonish ing performances make this technique a very promising alternative to quartz crystal to meet the GSM reference oscillator phase noise performance specificati ons (of -130 dBc/Hz at 1-kHz offset from a 13-MHz carrier and -150 dBc/Hz at far-from-carrier offsets). 1.3 Micromechanical Resonators for Wireless Communication Due to the ever increasing demand for multi-band and multi-functional wireless handsets, on-chip highQ resonators have become the only viable choice for numerous future wireless applications. Although nanomech anical resonators have been proven with their ability to operate at GHz frequencies even in their f undamental mode [29, 30], they are also more susceptible to scaling-i nduced performance limitations, such as the adsorption/desorption noise and temperatur e fluctuation noise, than their MEMS
10 counterparts. For instance, noise sources such as Brownian motion and Johnson noise, which are considered negligible for most of MEMS devices, become significant as the sizes of nanomechanical resonators continue to shrink . Additionally, insufficient power handling is another issu e that may hinder the rapid deployment of NEMS devices in wireless systems to satisfy todayÂ’s co mmunication standards. Even though packaging devices under proper pressure and temperature conditions may mitigate the noise problems, the more serious constraints in power handling ability would remain unresolved. Moreover, both MEMS and NEMS resonators operate with higher motional impedance than their macro-s cale counterparts. Hence, in order to seamlessly interface the MEMS/NEMS resonators with the macr o-scale electronics such as antennas, strategies for reducing their equivalent mo tional impedance are urgently needed. This paper will review both device -level and system-level me thodologies for lowering the motional impedance of the resonators. As such, the implementation of capacitive transducers filled with high-k dielectrics to improve the electromechanical coupling coefficient would be considered as one of the device-level methods; whereas parallel combination of a large array of resonators enables reductio n of the motional impedance while improving the overall linearity and pow er handling ability would be treated as a system-level approach. In terms of the readiness to be inserted into 50 -matched wireless subsystems, MEMS resonators wo uld be a much more practical choice comparing to NEMS. Therefore, development of MEMS resonators becomes essential for realization of the next-gen eration wireless systems.
11 1.3.1 Piezoelectrically-Actuated Resonators Piezoelectric materials, such as zinc oxide, aluminum n itride, barium titanate and lead-zirconate-titanate (PZT), have been widely studied for numerous device applications since its discovery in 1880. Piezo electric material deforms wh en an electric field is applied, whereas any strain induced in piezoelectric material generates charges within the material. A simple piezoelectri cally-actuated resonato r consists of a piezoelectric material and a mechanical structure along with st rategically placed electrodes, which are employed to facilitate the coupling between th e mechanical and electrical domains. When the applied ac signal matches the resonance fre quency with a particular mode shape of interest, the MEMS resonator will vibrate at its resonance. Different types of piezoelectrically-actuated res onators are discussed and co mpared in this section. 188.8.131.52 Surface Acoustic Wave (SAW) Resonators Figure 1.5 Schematic of a typical SAW resonator . A Surface Acoustic Wave (SAW) resonator co nsists of three parts, as illustrated in Figure 1.5: piezoelectric substrate, interdigital finger transducers (IDT) locating on the crystal surface, and reflectors disposed near the opposite ends of IDT. The SAW resonator utilize surface acousti c wave vibrating between two reflectors and the resonant frequency is determined by both the width of IDT and spacing between two fingers. The equation can be expressed as
12 sv f 1.2 where is the velocity of the surface acoustic wave in the piezoelectric substrate, and is the pattern period of the IDT, as shown in Figure 1.5. Unlike the bulk-acoustic-wave device, ac oustic energy propagates along and is confined to a single surface of the substrate. Hence, SAW devices are not as sensitive as their bulk-acoustic-wave counterparts to the su bstrate shape or scale, resulting in a much easier design and fabrication. In addition, due to the lower power density that occurs in the distributed geometry, SAW devices have superior power ha ndling capability as compared to bulk-wave devices. As a result of these advant ages, SAW devices are widely used in todayÂ’s wireless co mmunication systems . 184.108.40.206 Bulk Acoustic Wave (BAW) Resonators Another type of acoustic wave is bulk acoustic wave (BAW). Unlike surface acoustic wave, the bulk acoustic wave travel s from one surface through the bulk material to the other surface to form so called bulk acoustic wave. BAW devices usually operate with resonance frequencies of 1~20 GHz [ 34-37]. BAW resonators have many features superior to mostly used devices such as SA W devices and ceramic devices. Comparing to ceramic and SAW devices, it has a high Q -factor that leads to low insertion loss and sharp cut-off characteristics and low power consumption with reduced size. Figure 1.6 shows the Agilent process using free-standing membranes that are anchored at edges to the silicon substrate.
13 Figure 1.6 Device implementation for the tw o types of bulk acoustic wave (BAW) resonators with two different types of acous tic isolation methods : (a) SMR employs BraggÂ’s reflector; and (b) FBAR sits on top of air cavity. In a typical BAW device, the acoustic la yer is stacked in between the top and bottom electrodes, where the acoustic wave is confined. The performance of a BAW device is heavily dependent upon the im pedance mismatch on the boundary of the piezoelectric body which helps trapping the as -generated acoustic wave within. Two different kinds of implementation are mostly used, as illustrated in Figure 1.6: solid mounted resonator (SMR) and film bulk acoustic resonator (FBAR). In SMR, the BAW structure is sitting on top of multiple reflective layers (BraggÂ’s reflector)  to reflect/retain the acoustic wave back into the piezoelectric film. On the other hand, FBAR simply uses an air cavity to create a huge impedance mism atch. The operation principles of SMR and FBAR are the same except the technology employed to provide the acoustic isolation. Compared to SMR, FBAR is much eas ier to fabricate as it does not require the perfect quarter wave length reflective la yers. Since 2002, FBAR has been widely employed for wireless telecommunications. However, in order to achieve precise fr equencies and a better yield, the thickness of each thin film must be accu rately controlled from device to device, which may cause a serious problem in FBAR device fabricati on. Nevertheless, as wireless communication
14 technology pushing towards to the ultimate mini aturization, multi-function transceivers that operate at different frequencies instead of a set of several discrete components are soon going to be needed, which is a serious bottleneck against the current FBAR technology. Recently, due to the advance of microm achining technology, a new type of BAR device has attracted large attention. Unlike conventional FBARs, the novel contour-mode FBAR device (see Figure 1.7) us es its lateral dimension (i.e., radius of the disk or the width of the ring) to cont rol resonance frequency, thus allowing implementation of multiple frequency circuitries on a single chip It has been proven that piezoelectricallytransduced contour-mode resonators have un ique characteristics, such as a CAD-layout definable resonance frequency, high quality factor up to 4,300 at 230 MHz and low motional impedance (50~700 ), thus making it a perfect candidate as the key building block for the next generation wireless transceivers [39, 40]. Figure 1.7 Contour mode ring resonator : (a ) one port circular ring resonator; (b) one port square-shape ring resonator. Figure 1.8 depicts the top-view image and measured frequency characteristics of post-CMOS compatible AlN-based dual-m ode filter developed by Sandia National Laboratory, which employs an unique molded tungsten (W) techni que along with the
15 AlN process to realize potentially higher Q . In addition, this technology allows the scaling of AlN resonators into the GHz ra nge without introducing spurious modes or reducing the quality factor, but offering accep table power handling for both the transmit and receive paths in full-duplex radios. Figure 1.8 (a) Photo of an AlN dual-mode filte r; (b) measured transmission (S21) of the AlN dual-mode filter, along with simulated response based on its equivalent circuit model with different term ination impedance . As good as it seems to be, these newly-em erged piezoelectric MEMS resonators still have some remaining issues to be addressed. On one hand, the FBAR resonator, whose resonance frequency is controlled by its thickness, is not amenable to realization of multiple frequencies on a single-chip. On the other hand, piezoelectrically-transduced contour-mode resonators suffer from its rela tively low frequencies. In addition, the relatively large temperature co efficient in the range of 25 ppm/C  as opposed to 2 ppm/C  of less for capacitively-transduced resonators is yet another concern. In order to overcome these issues, there ar e ongoing research studies which focus on optimization of piezoelectrically -transduced MEMS resonators.
16 1.3.2 Capacitively-Transduced MEMS Resonators Capacitively-transduced resonator offe rs in general the highest frequencyQ product among micromechanical resonators, due to the employment of high quality structural material which is much less susceptible to acoustic losses to the substrate as opposed to piezoelectrically-act uated resonators. Besides th e high frequency selectivity, the use of capacitive transduction also allows the resonators to ope rate under a dc bias voltage without dc current flow, thus consuming ultra-low dc power. Moreover, capacitive transduction is also employed in larg e part to simplify future integration with transistors. Additionally, the employment of extensional-mode resonators with CADlayout definable resonance frequencies provide a solution to resolve the key limitations of the conventional film bulk acoustic resonators (FBARs), in which the thickness of the piezoelectric film determines the resonance frequency. Unfortunately, the majority of capacitively-transduced MEMS resonators deve loped so far suffer from their excessive motional impedance, which hinders their direct implementation into the wireless systems. On the contrary, much lower moti onal impedance in the range of 50 can be easily obtained in MEMS resonators equipped with piezoelectric transducers. This section reviews recent progress in the research and development of MEMS resonators in silicon and CVD diamond materi als for wireless communications, with a particular focus on existing and possible solutions of the aforementioned issues. 220.127.116.11 Flexural Mode Beam Resonator Vibrating beam micromechanical resonators have attained a great deal of attention over the past few years, due to thei r achievable VHF and UHF ranges, high Q Â’s, tiny sizes, and virtually zero dc pow er consumption. Three different types of beam resonators
17 have been widely investigated and utilized in various areas, includ ing sensing, actuation and communication . Categorized by th eir boundary conditions, they are clampedfree beam resonators (i.e., cantilevers)  clamped-clamped beam resonators [45, 46], and free-free beam resonators , shown by Figure 1.9 (a), (b), and (c), respectively. Figure 1.9 Schematic view of different types of beam resonators: (a) cantilever beam resonator; (b) clamped-clamped beam res onator; (c) free-free beam resonator. Cantilevers have much lower resonant frequencies and worse dynamic range when compared to free-free beam and clamped-clamped beam resonators because of their relatively low stiffness. However, for the case of clamped-clamped beam resonators, larger stiffness is gained at the cost of higher anchor dissipation, which makes it much harder to achieve highQ at high frequencies Q reduces from 3,000 to less than 300 when resonant frequency increase s from 10 MHz to 70 MHz. Although high Q -factor can be achieved by shrinking dimensions (masses) of clamped-clamped b eam resonators , inadequate dynamic range and power handling capability of such devices are hard to satisfy most communication applications. On the other hand, free-free beam resona tor has been demonstrated with center frequency of 30 ~ 90 MHz and Q of ~8000 . As shown in Figure 1.9 (c), the free-free beam resonator is supported at its flexural node points by four torsional beams which are
18 anchored to the substrate. The supporting torsional beams are de signed with quarterwavelength dimensions, which acoustically is olate the free-free beam from the rigid anchors. Thus, ideally, the free-fr ee beam operates as if it is levitated without any anchor. The ideally eliminated anchor dissipation enables the free-free beam design greatly surpasses previous clamped-clamped b eam resonators, allowing much higher Q -factor. 18.104.22.168 Radial-Contour Mode Disk Resonator Polysilicon micromechanical radial-contour mode disk resonators have been firstly demonstrated with Q Â’s over 1500 at frequency of 1.14 GHz in vacuum and air, respectively . One year later, the simila r resonator with resonance frequency up to 1.156 GHz along with measured Q Â’s > 2650 has been reported. In addition, a 734.6-MHz version has been demonstrated with Q Â’s of 7,890 and 5,160 in vacuum and air, respectively . Figure 1.10 shows the SEM pict ure and frequency response spectrum of a polysilicon self-aligned radial-c ontour mode disk resonator. Figure 1.10 SEM image and measured frequency response in (a) vacuum and (b) air for a polysilicon capacitively-transduced radial-c ontour mode disk resonator . Self-alignment of the supporting stem to th e exact center of the resonator disk allowing the superior symmetri cal modal vibration of the re sonator is the key to obtain highQ at gigahertz frequency range while reta ining the similar micro-scale dimensions and adequate power handling. Self-alignment a llows the resonator disk being supported
19 by the stem at the motionless nodal point at th e center of the disk during its pure radial vibration, and minimizes the energy loss to th e substrate from the center anchor, thus allowing highQ in spite of the high resonator stiffness that is essential for maintaining decent power handling capabilit y. As illustrated in Figure 1.11, the self-aligned center anchor is achieved by defining both the stem position and the disk edges during a single lithography step (i.e. in one mask), which e ffectively eliminate the misalignment between two different masks (the first one defines th e stem and the second mask aligned to the previously patterned stem to pattern the disk around the center stem), as opposed to previous resonator designs . Another difference between this self-aligned disk resonator and the previous one is that an electrically accessible substrate contact is included in the present design, which successf ully eliminates feedthrough current and allows much cleaner measurement of fre quency characteristics (see Figure 1.12). Figure 1.11 Comparison between (a) a previous disk resonator process; (b) the selfaligned disk resonator process .
20 The complete process flow of the self-a ligned radial-contour mode disk resonator is illustrated in Figure 1.12: (a) a layer of high-temperature oxide (HTO) is deposited on n+ Si substrate by LPCVD, followed by a layer of LPCVD Si3N4; (b) substrate contact trenches are dug through the SiO2/Si3N4 layer by wet/dry etch, th en the first layer of polysilicon is deposited via LPCVD and POCl3 doped, and finally patterned to form ground planes, interconnects and substrate contact pads; (c) a field HTO is deposited via LPCVD, followed by the structural POCl3 doped LPCVD polysilicon layer, and then capped with a film of HTO serv ed as the hard mask for stru ctural polysilicon dry etch and as the spacer layer to separate the disk and the overhanging electrode s; (d) after annealed in N2 atmosphere for an hour, the spacer HTO film and the structural polysilicon layer are patterned, defining not only the shape of the di sk but also an opening at the center of the disk, which is eventually the location of the stem; (e) an LPCVD sidewall sacrificial HTO layer is conformally deposited with a specific thickness that equals to the desired electrode-to-disk capacitive gap spacing; (f) the sidewall sacrificial HTO layer is removed in the stem opening, after which the stem and electrode via are opened down to the polysilicon substrate contact s; (g) a third layer of POCl3 doped polysilicon is deposited and then patterned not only to define the side electrodes but also to refill the center via, forming a self-aligned stem; (f) the structure is released in HF to yield the final cross section view.
21 Figure 1.12 Fabrication process flow of th e self-aligned radial-contour mode disk resonator. 22.214.171.124 Wine-glass Disk Resonator En route to pursuing higherQ polysilicon wine-glass mode disk resonator using a stem-less, side-supporting suspension structure has been demonstrated at frequency of 74 MHz with Q Â’s as high as 98,000 in vacuum and 8,600 in atmosphere  (see Figure 1.13). The lack of a center stem allows this device to minimize anchor losses, thus achieving higher Q. However, the device operates at a relatively lower frequency range due to the native of wine-glass vibration mode.
22 Figure 1.13 SEM photo and frequency response sp ectrum in (a) air and (b) vacuum of a polysilicon wine-glass mode disk resonator . A perspective-view schematic of the wine-g lass mode disk res onator in a typical two-port bias and excitation configura tion is illustrated in Figure 1.14. Figure 1.14 Perspective-view schematic of th e wine-glass mode di sk resonator in a typical two-port bias and excitation configuration. 126.96.36.199 Wine-glass Mode Ring Resonator Despite the sufficient highQ and high frequency for applications of wireless communication achieved by the two aforementioned types of capacitively-transduced resonators (i.e. radial-contour mode disk re sonator and wine-glass mode disk resonator), their large motional impedance (> 1 M ) is so far too high to compete with todayÂ’s conventional RF components, which are desi gned to match to the front-end system impedance of 50
23 An extensional wine-glass ring (EWGR) re sonator structure , shown in Figure 1.15, has been designed so that the res onance frequency primarily depends upon the width of the ring other than the average radius thus the ringÂ’s perimeter sidewall surface area is independent of its frequency. B ecause the motional impedance of a capactivelytransduced resonator is proportional to the ov erlap area between the resonance structure and the electrodes, the impedance of the devi ce can be strategically designed simply by choosing an appropriate radius without affecting its frequenc y. In order to achieve lower impedance than previous micromechanical reso nators, inner and outer electrodes are used for a side-supported EWGR to increase the el ectrode-to-disk overl ap area exhibiting a moderate impedance of 282 k with DC-bias of 10 V at 1.2 GHz. In addition, the mode shape of the ring resonator combines the aspects of previously demonstrated modes, radial-contour mode  and wine-glass mode , to achieve the best of each design. Specifically, the EWGR structure allows ultra high resonance frequenc y due to the use of the radial-contour mode, and higher Q because of stemless side-supporting structure resembling the wine-glass mode. As demonstrat ed in , frequenc ies as high as 1.2 GHz with a Q of 3,700, and 1.52 GHz with a Q of 2,800, have been successfully achieved, with the motional resistance 2.2 times lower than the measured resistance of the radialcontour mode disk counterparts.
24 Figure 1.15 Perspective-view schematic of the extensional wine-glass ring (EWGR) resonator with typical drivi ng and sensing configuration. Moreover, by properly designing impeda nce-mismatched resonator-to-anchor transition, Q Â’s of 14,603 at 1.2 GHz has been achie ved, as demonstrated by the spokesupported Â“hollow diskÂ” EWGR resonator  The device is supported by a center stem with four supporting beams attached to th e ring at notched nodal locations, achieving minimal anchor losses, thus raising Q to 14,603 from 5,846 at 1.2 GHz. In addition, the supporting beams are designed to match the quarter wavelength of the resonance frequency to further prohibit energy dissipation from the cen ter anchor to the resonance ring. SEM picture and measured frequency ch aracteristics are show n in Figure 1.16 for the Â“hollow diskÂ” EWGR device.
25 Figure 1.16 SEM picture and measured frequenc y characteristics of a (a) un-notched and (b) notched Â“hollow disk Â” EWGR device . 188.8.131.52 Internal Dielectrically -Transduced Bar Resonator However, the cost of extending frequencies by scaling dimensions of MEMS resonators is greatly increased motional impedance Rx, which is a function of electrodeto-disk gap spacing, d overlap area, A and dielectric constant of the material using for the capacitive gap, respectively, written as: 2 2 0 2 4 0 2 0 2 2 0A d QV k x C QV k Rr P r P r x 1.3 Thereby, replacing air-gap that is employed by most capacitively-transduced resonators with solid gap filled by high-k dielectrics has several benefits. It is desirable to achieve smaller gap spacing, to prevent stiction symptoms of air-gap transducers, and to enhance capacitive sensing and reduce moti onal impedance due to higher dielectric constant, which will become even promin ent as frequency goes higher and as the dielectric thickness approaches acoustic half-w ave length in silicon [ 10]. A dielectricallytransducer silicon bar resonato r with 15 nm nitride solid gap has been demonstrated with the so far highest resonance frequency of 6.2 GHz and Q of 4,277 as shown in Figure 1.17. A frequencyQ product of 3.11013 at 4.7 GHz was also repo rted as the highest in
26 polysilicon reported to date [ 10]. A silicon bar resonator employing 10 nm air + 90 nm HfO2 and 100 nm air capacitive gap exhibited an impedance of 1,256 and 40,356 respectively, with Q > 66,000 at 223 MHz, showing that high-k dielectric solid gap can effectively reduce the motional impedance . Figure 1.17 Picture of a bulk-mode resonator an d the measured frequency characteristics at different vibra ting mode . 1.3.3 Resonator Based on Silicon-on-Insulator (SOI) Technology A lot of efforts have been made on VHF and UHF micromechanical resonators using polycrystalline silicon as the structural material. As demonstrated in the previous sections, astonishing performan ce of the resonators (i.e., highQ high frequency, etc.) have been achieved, however, higher-than-no rmal motional impedance is still the main issue that hinders its further developmen t. Silicon-on-insulator (SOI) technology based on wafer bonding technique provides the possibil ity of greatly increased overlap area between resonator body and electrodes, which is able to reduce the motional impedance of the resonator while improving the power handling ability. Motional resistance as low as 43.3 k has been measured for a 18 m-thick disk resonator operating in its wineglass mode at 149.3 MHz with Q of 45,700 in vacuum and 25,900 in atmosphere, which is orders of magnitude smaller than the 883 k of a 3 m thick device .
27 In addition, comparing to conventional silicon substrate and surface micromachining, SOI technology distinguishes its elf in several aspects: (1) the use of single crystal silicon as the structural laye r other than commonly used polycrystalline silicon provides superior mech anical properties, such as lower energy dissipation due to less crystallographic defects [ 56] and lower internal stress ; (2) specific thickness, dopant type and concen tration of the device silicon laye r are available in commercial manufacturers, thus minimizing the need of the structural polysilicon deposition and doping during the fabrication process, which requires precisely controlled equipment and processing environment; (3) buried oxide layer offers superior dielectric isolation which protects the device Si layer from parasitic effects indu ced by the substrate . Various techniques have been develope d to fabricate su spended single Si structures on SOI substrate, which will be di scussed and compared with each other in the following section, with a focus on the aspe ct of the fabrica tion techniques. 184.108.40.206 Fabrication on SOI Substrate Util izing Electron Beam Lithography Electron beam lithography has been used fo r nanostructure patterning for years with resolution limit as low as several nm . Combination of electron beam lithography and conventional silicon micromach ining technique provi des the possibility of fabricating MEMS devices with nano-scale structures. A typical e-beam lithography-enabled MEMS device process includes four steps  as illustrated in Figure 1.18: (1) resist (e.g., PMMA) patterning by e-beam lithography; (2) metal lift-off; (3 ) Si anisotropic etching using metal as the hard mask; (4) underlying SiO2 isotropic etching to suspended resonatorÂ’s body.
28 Figure 1.18 Fabrication proce ss using e-beam lithography fo r creating suspended NEMS device on SOI substrate. 220.127.116.11 Fabrication on SOI Substrate Utiliz ing Focus Ion Beam (FIB) Technique Focus Ion Beam (FIB), which uses Ga+ ion to scan over the surface of a sample in a similar way as the electron beam in a scanning electron microscope (SEM), can mill a very narrow trench on Si substrate by accurate ly positioning on the sample at high current density , offering a promising alternat ive of conventional UV lithography technology to fabricate nano-scale electrode-to-disk gaps in a capacitively-transduced resonator. MEMS resonators with capacitiv e gaps less than 100 nm on thin SOI substrate has been demonstrated . The fabrication process (s ee Figure 1.19) is based on a combination of standard UV lithography and a subsequent FIB milling with only two levels of photolithography, yielding the fina l SEM view in Figure 1.20.
29 Figure 1.19 Schematic of the process of MEMS resonator with nano-gap utilizing FIB milling technique: (a) thermal oxide is grow n on SOI wafer; (b) photolithography is used to open the release hole; (c) nano-scale gaps are achieved by FIB milling; (d) patterns are transferred to SiO2/Si/SiO2 layers by high aspect-ratio dr y etch; (e) metal contacts are patterned by lift-off proce ss after striping the top SiO2 layer; (f) the resonators are released by etching aw ay the buried oxide. Figure 1.20 Bulk lateral res onator with narrow air gap (<100 nm) fabricated by the proposed FIB-based process .
30 18.104.22.168 SOI Fabrication by Convention al Si Micromachining Technique Although nano-scale devices and narrow capacitive gaps can be achieved by ebeam lithography and FIB technique, low prod uctivity and high cost have hindered the further employment in mass production of micr omechanical resonator fabrication. On the contrary, the conventional Si surface and bul k micromachining technique remains as the best choice to fabricate VHF/UHF and highQ capacitively-transduced resonators. Single crystal silicon capacitively-trans duced micromechanical resonators with sub-100 nm electrode-to-disk gaps based on high aspect-ratio poly and single crystal silicon (HARPSS) fabrication te chnique have been successfully demonstrated at VHF range with Q as high as 45,700 in vacuum (see Fi gure 1.21) and 25,900 in atmosphere. Motional resistance as low as 43.3 k has been measured for a 18 m-thick disk resonator operating in its wine-glass mode at 149.3 MHz, which is orders of magnitude smaller than the 883 k of a 3 m thick device [55, 63, 64]. Figure 1.21 (a) SEM image and (b) frequency response of an 18 m-thick wine-glass disk resonator on SOI substrate . The HARPSS fabrication pro cess is shown in Figure 1.22: (a) thermal oxide is grown on SOI wafer and patterned, then Si3N4 is deposited by LPC VD and patterned by lithography; (b) trenches on device Si layer is etched by DRIE with SiO2/Si3N4 as the
31 hard mask; (c) a thin layer of sacrificial oxi de is deposited by HTO and is blanket etched on the surface, leaving only on the resonator sidewalls; (d) doped polysilicon is deposited by LPCVD and patterned on the surface to form the wire-bonding pads, then a thin layer of metal is deposited on the wire-bonding pa ds by e-beam evaporator; (e) releasing openings are etched in the device Si laye r and the polysilicon inside trenches using photoresist mask; (f) buried ox ide layer is removed by SiO2 wet etch to suspend the device. Figure 1.22 Fabrication process flow of HARPSS resonator on SOI substrates. However as described above, the HARPSS process on SOI substrate consists of 6 lithography steps as well as several etchi ng and deposition steps, which makes the fabrication process very complicated and time-consuming. A much simpler ICcompatible process on SOI substrate has b een reported utilizing Chemical Mechanical Polishing (CMP), achieving 100 Â– 200 nm cap acitive gaps with expected operating
32 frequency of the highQ resonator extend to MHz and GHz range . Utilizing a similar fabrication process, study done by D. Grogg  has demonstr ated a laterally vibrating bulk-mode resonators based on connected parallel beam resonators (PBRs) with Q of 100,000 and motional impedance of 55 k at 24.58 MHz. Figure 1.23 SEM image of (a) the disk res onator and (b) a zoom -in on the 200 nm gap . The cross-section view of the CMP based fabrication process is illustrated in Figure 1.24: (a) starting with SOI su bstrate, a layer of LPCVD SiO2 (TEOS) is deposited and then patterned, which defines the resona tor disk structure, followed by a conformal coating of un-doped polysilicon thin layer, serv ing as the electrode-to-disk capacitive gap spacing; (b) a second layer of SiO2 (TEOS) is deposited via LPCVD, followed by a combination of CMP and wet SiO2 etch to expose the polysilicon layer; SiO2 dry etch is then performed to etch the second SiO2 (TEOS) layer; (c) 1st and 2nd SiO2 (TEOS) layers as hard mask, resonator disk and side elect rodes are then patterned by a Si dry etch, forming an air gap with a nano -scale width determined by the previous polysilicon layer; (d) a back-side SiO2 wet etch is used to remove the remaining SiO2 (TEOS) layer and at the same time to transfer the Si patterns down to the buried SiO2 layer; (e) a thick layer of
33 photoresist is spun, exposed and developed, not only to remove the undoped polysilicon layer but also to serve as a lift-off mask for the metal film in the fo llowing step; (f) a film of metal is plated and patte rned by lift-off process; af ter striping the photoresist and releasing the structure in HF, a fi nal cross section view is yield. Figure 1.24 Fabrication process flow of the TEOS and CMP based method. However, currently demonstrated fabr ication processes either are very complicated (e.g., HARPSS process) or re quire nonstandard techniques (e.g., CMP). Moreover, the thickness of the capacitive air gap is limited by the thermal oxidation process and the small dielectric constant of air is not ideal for a capacitive transducer as well. A novel high aspect-ratio SOI micromachin ing technique is presented in this work that is capable of producing vert ical disk resonators and resonator arrays with ultra-thin high-k dielectric solid gap by a simplified proc ess consisting of mere ly three lithography steps. The detailed processing steps will be discussed in Chapter 3.
34 1.4 Resonator Array In the ideal case an array of N resonators if the resonators are excited at exactly the same resonance frequency, the output current will increase by N times for the same input voltage, thus lowering the motional im pedance by N times. Unfortunately, even a tiny frequency mismatch can dramatically affect the combined output. Mechanical coupling provides a superb solution to the frequency-matching problem, which mechanically forces all coupled resonators vi brating at the same frequency as the signal of a given modal resonance frequency applied to the overall resonator array. By means of mechanical coupling and excitation of a pa rallel array of corner-coupled polysilicon transverse-mode square plate resonators (see Figure 1.25), motional impedance of the Â“compositeÂ” resonator array has been successfully reduced down to 4.4 k at 64 MHz, 4.8 times smaller than the 21.3 k measured by a single square resonator, with Q > 9000 . Figure 1.25 SEM image and frequency response spectra of a single and mechanically coupled square resonator arrays with thr ee and five resonators respectively .
35 1.5 Capacitively-Transduced Resonators Using Materials Other than Si Among the currently availabl e thin-film-depositable mate rials, diamond offers the best mechanical properties, which has potentials to realize the best the frequencyQ performance. Chemical vapor deposition (CVD) polycrystalline diamond was employed as the material for the resonance disk and pol ysilicon for the anchoring stem, successfully raising the Q of radial-contour mode disk resonato rs to 11,555 in vacuum at frequencies of 1.51 GHz . An acoustic impedance mism atch between the two different material suppresses energy transfer from the disk to the stem, thus eliminates anchor losses and leading to a higher Q A MEMS cantilever type resonator exhibited a resonance frequency of 318.2 KHz with Q > 116,000, the highest reported value for a polycrystalline cantilever re sonator . Ultra nanocryst alline diamond (UNCD) thin film deposited by hot filament chemi cal vapor deposition (HFCVD) has been demonstrated, with YoungÂ’s modulus up to 920 GPa , showing a great potential for MEMS resonators to achieve even higher opera ting frequencies. For applications such as sensing and wireless comm unication, mechanical re sonators with a high Q and high resonance frequency are desirable, as they w ould lead to the covete d high resolution and high frequency selectivity for sensors and wi reless transceivers, respectively. Of course, less energy losses also implies lower power consumption, which is crucial for the portable sensors and wireless transceivers. In addition to the outstanding mechanical strength, diamond as a promising alternative to Si or SiC offers a lot more outstanding properties, such as exceptiona lly low friction coefficient, excellent thermal stability, great chemical inertness, and very good plasma etching selectivity to Si and SiO2 making micromachining much easier .
36 A comparison of mechanical properties and relevant freque ncies among several common MEMS materials is summarized in Table 1.1. Table 1.1 Properties for different material s and their relevant frequencies. Material YoungÂ’s Modulus E (GPa) Density (kg/m3) Acoustic Velocity (m/s) Frequency Scaling Factor Polycrystalline Silicon 150 2.33 8024 1 Silicon or SOI Substrate 165.7 2.33 8433 1.05 Silicon Carbide 415 3.12 11500 1.433 Polycrystalline Diamond (800 C1000 C) 1144 3.5 18076 2.256 Ultrananocrystalline Diamond (UNCD) (<400 C) 920 3.5 16210 2.02 Material other than commonly used polys ilicon, diamond, or single crystalline silicon has also been investig ated [72-74]. Some metals ha ve been applied as a main structural material of MEMS resonator by taking the advantage of its low deposition temperature. Particularly, electroplat ed nickel can be deposited at 40 C-60 C to obtain potentially high aspect radio along with its low cost. With a frequency range from 18 MHz to 426 MHz, electroplated-Ni-based MEMS resonators have been built with Q Â’s of 6,405 and 2,467 in vacuum and in air, respective ly . Another attr active benefit from this type of devices is their compatibility to simple fabrication over CMOS circuitry, since the fabrication temperat ure is lower than 100 C, thus making it amenable to postCMOS MEMS device fabrication . 1.6 Overview Despite many efforts being made in pa st studies for the purpose of reducing motional impedance, using employment of na no-scale high-k solid gap, usage of SOI
37 substrate and coupling resonators in an arra y, there has never been sufficient effort in putting all possible techniques together to achieve a optimally low impedance and ultimately to realize a 50 match to the front-end ci rcuitry. A newly-developed fabrication process based on SOI technology u tilizes atomic layer deposition (ALD) for formation of nano-scale solid capacitive gap with high-k dielectric material. Through ALD technology, which is capable of providing superb conformability and uniformity as well as outstanding thickness controllability, an ultra-thin layer (~10 nm) is deposited on a high-aspect-ratio feature formed by deep reactive ion etch (DRIE) on SOI, thus allowing the mass production of on-chip capacitively-t ransduced resonators and resonator arrays with greatly enhanced electromech anical coupling coefficient. The newly developed IC-compatible MEMS microfabrica tion process consisting of merely three standard photolithography step s, thus showing a great ad vantage as opposed to other SOI-based resonator device technologies. In addition, the newly developed SOI process allows DC-bias voltage to be directly appl ied on to the substrate, which ensures the resonator body to be grounded perfectly, th us having a great promise to achieve minimized feed-through capacitance by steer ing feed-through current to ground. Figure 1.26 illustrates the schematic-vie w of a 3-by-3 disk resonato r array on SOI substrate and the corresponding measurement configuration.
38 Figure 1.26 Schematic of a 3-by-3 disk resonator array on SOI substrate and its measuring configuration. Chapter 2 presents a theoretical derivati on of wine-glass mode shape of a disk resonator and a ring resonator as well as equivalent circuit models. Chapter 3 details the novel fabrication methodology on SOI substrate. Finally, Chapter 4 conc ludes this thesis and discusses possible future research directions.
39 CHAPTER 2 RESONATOR DESIGN 2.1 Extensional Wine-glass Mode and Resonance Frequency Design 2.1.1 Wine-glass Mode of a Disk Resonator Figure 2.1 illustrates key dimensional parame ters of a capacitive disk resonator in polar coordinate ( r ) with origin located at the center of the circular plane, which is of radius R anchored by two side-support beams of width b and length L A pair of input/output electrode located on each side of the disk, span an equal angle of e and separate from the disk by a capacitive gap of ds and dd for the sensing (input) and driving (output) electrodes, respectively. The device operates in a two-port bias, excitation and measurement configuration. In order to excite the device into resonance vibration, a direct-current DC-bias Vp is applied to the resonant stru cture and an AC voltage signal vd to the driving electrode, genera ting an electrostatic input force pointing outward from the disk. When the frequency of the input signal vd matches the wine-glass mode resonance frequency of the disk, the result ing force drives the disk into a elliptic vibrating mode as illustrated by the dot line in Figure 2.1. The elliptic mode (i.e. wine-glass mode) involves both radial and circumferential displacements, with four nodal po ints locating at the disk pe riphery, 90 apart from each other, where the supporting beams are aligned to in order to mitigate the anchor losses. The wine-glass mode generates a DC-biased ca pacitance between the disk and the output electrode which sources an output current is. The resonator disk is made of low-resistivity
40 single crystal silicon, while th e input/output electrodes ar e built with p-type doped polysilicon. Figure 2.1 Top view of a wine-g lass mode disk resonator. A comprehensive derivation of the in-plane vibration of a disk resonator and the mathematical expression of its mode shape and resonance frequency is provided by this section. Due to the orders of magnitude sm aller vertical dimension (i.e. thickness, h ) of the resonance structure than that of its lateral dimension (i.e. radius, R ), it may be assumed that the vibration variables are inde pendent of the thickness. In addition, effects of supporting beams on the in-plane vibration can be negligible, if narrow beams are located on the nodal points on the edge of the vibrational disk. Thus the resonator disk can be modeled approximately as an ideal 2-D circular thin plane with free edges, which results in a plane stress scenario. The differential equation of a 2-D disk in -plane vibration, as derived by Love , may be expressed as
41 2 2 2) ( t u u u 2.1 where 2 1 1 E and 1 2 E are LameÂ’s constants, E , and are YoungÂ’s modulus, PoissonÂ’s ratio, and densit y of the resonance structure material, respectively. The displacement vector u may be defined in term s of pressure-wave (Pwave) scalar potential, and shear-wave (S-wave) vector potential, as  u 2.2 Combining equation 2.1 and 2.2, the scalar potential, and the vector potential, can be written as 2 2 2 2t 2.3 2 2 2 2t 2.4 where 2 2 2 21 1 r r r r 2.5 21 E 2.6 1 2 E 2.7 are propagation velocities of th e P-wave and S-wave, respectively. Equation 2.3 and 2.4 may be solved, in terms of the trigonometric and Bessel function. For time-harmonic ex citation with time dependencet jme, the mode shape may be expressed as t j m m m mme m R r k J A cos / 2.8
42 t j m m m mme m R r h J B sin / 2.9 where m mf 2 is the angular resona nce frequency of the m -th order, Am and Bm are constants determined by th e excitation amplitude, and km and hm are / R km m 2.10 / R hm m 2.11 In equation 2.8 and 2.9, Jm is the BesselÂ’s function of first order of the mth order. The mode order, m represents numbers of the nodal diamet er in the free vibrational response. When m = 0, the disk resonator entails an axisymmetric uncoupled mode, the displacement is only in the radial direction (r adial) or in the circumferential direction (torsional) with no nodal rings. While m = 1 implies a nonzero response at the center point of the disk. Finally, when m is equal to or larger th an 2, nodal points on the disk periphery will occur, which are well-suited for attaching the supporting beams to obtain low-loss suspension. By substituting equation 2.8 and 2.9 in to 2.2, the time-independent radial ( U ) and circumferential ( V ) components of the displacement, u may be expressed as m R r h J B r m R r k J dr d A Um m m m m m mcos / / 2.12 sin / /m R r h J dr d B R r k J r m A Vm m m m m m m 2.13 Let r and r be the normal and tangential stre ss at any boundary point of the disk. For a disk with free edge, the stre ss at the boundary must vanish, namely, 0 12 V U r r U ER r r 2.14
43 0 1 1 2 V U r r V ER r r 2.15 which leads to the mode freque ncy equation given by  1 1 2 1 2 ) 1 ( 22 2 2 2 2 1 2 2 1m h m m h m h J h J h m h m k J k J km m m m m m m m m m m m m. 2.16 Substituting Equation 2.12 and 2.13 into 2.14 and 2.15 leads to the following matrix: 022 21 12 11 m mB A a a a a 2.17 which is associated only with km, hm, and Thus, the ratio betw een the constants of m and m is calculated as: 2 1 1 2 21 2 1m m h J h J h m m h k J k J k h J k J A Bm m m m m m m m m m m m m m m m m m 2.18 From equation 2.12, 2.13, and 2.18, the radial displacement at ( r ) can be written, instead of 2.12 and 2.13, as r mU R A U 2 cos 2.19 where 2 21 R r h J r R R r k J r R R r k J k Um m m m m m m r 2.20 When m = 2, the disk resonator operates in its wine-glass mode, the dimensionless maximum radial displacement at th e disk edge can be expressed as
44 2 22 2 2 1h J k J k J k UR 2.21 Since the capacitive gap is extremely small compared to the size of the disk, the disk-to-electrode configuration can be treated as a parallel plate capacitor. Hence, the electrostatic excitation force for driving a nd sensing electrode, can be given by  2 22 cos 2 2 2 1p d R d p d dV d U R A v V d R h F 2.22 2 22 cos 2 2 1p s R s sV d U R A d R h F 2.23 where Fd, Fs, dd, ds are the electrostatic force and cap acitive gaps for the driving and sensing electrodes, respectively, is the permittivity of air, and A denotes Am ( m = 2). For a disk and its side elec trodes spanning an angle of e, the electrostatic excitation force for driving and sensi ng electrodes, can be expressed as 2 2 2 1 2 1 2 2sin 2 2 2 1 2 cos 2 2 2 1p d e R e d p d p d R d p d dV d U R A v V d R h V d U R A v V d R h Fe e 2.24 2 2 2 1 2 1 2 2sin 2 2 1 2 cos 2 2 1p s e R s p s R s sV d U R A d R h V d U R A d R h Fe e 2.25 It should be noted that the equation 2.16 is sole ly a function of the PoissonÂ’s ratio of the resonance structure material. Given km determined by 2.16, the resonance frequency can be calculated by equation 2.10 or 2.11, shown as
45 21 2 E R k fm m 2.26 where m 2. Figure 2.2 Mode shapes and resonance frequency for a 20 m radius single crystal silicon <100> disk calculated from the theoretical derivation using COMSOL Multiphysics 3.5a with (a) m = 2; (b) m = 3; (c) m = 4; (d) m = 5. Mode shapes and the corresponding re sonance frequencies are obtained by solving equation 2.16 using COMSOL, some of which are shown in Figure 2.2. As illustrated in Figure 2.2, the wine-glass mode shape that we study in this work appears when m =2. The related mechanical parameters of single crystal silicon (SCS) along <100> orientations used in the theoretic al derivation are listed in Table 2.1.
46 Table 2.1 The related parameter for a disk resonator. 2.1.2 Wine-glass Mode of a Ring Resonator A compound resonance mode design, named Â“extensional wine-glass resonatorÂ”, or Â“EWGRÂ” in , which combines two prev iously demonstrated modes, radial-contour mode  and wine-glass mode , togeth er with a geometric advantage of a ring structure, is descri bed in this session. Figure 2.3 Top view of a wine-g lass mode ring resonator. Figure 2.3 illustrates the schematic of a wine-glass mode ring resonator, identifying key dimensions and an excitation configuration. Similar to wine-glass mode disk resonator, the ring resonator is anchor ed by supporting beams se t to the quasinodal points at the ring outer periphe ry, which minimizes the anchor loss and thus retain the highest Q Electrodes are designed ar ound the ring structure, bot h inside and outside, Material Single crystal silicon <100> Polysilicon Young's modulus ( E ) [Mpa]170160 Possion's ratio ( )0.280.22 Thermal expansion parameter ( ) [1/K] 0.00000260.0000026 Density ( ) [kg/m3] 23302320
47 connected by bridges hanging over the res onator body, for the purpose of obtaining the largest transducer overlap cap acitance. Excitation mechanis m of the ring resonator is similar to the disk resonator. The devi ce firstly converts the input ac signal, vd, to a mechanical force to excite the ring resonator into a modal vibration at the resonance frequency, thus generating a mechanical displa cement between the ring structure and side electrodes. The mechanical displacement is th en converted back to an electrical signal, sensed by the output electrode. It should be no ted that output current is only generated if the dc-bias voltage Vp is finite. When Vp = 0 V, the device is e ffectively Â“offÂ”. Thus, the capacitively-transduced micromechanical resonato r with dc bias volta ge essentially acts as a switch. The Â“EWGRÂ” mode shape consists of four quarter cycles to finish an entire modal vibration cycle. In the first qu arter cycle, starting from the original ring shape, the two ring quarters along x-axis gradually expand while the other two quarters along y-axis gradually contract, finally reaching the maxi mum expansion of the x-axis quarters and maximum contraction of y-axis quarters. Th en the motions of the x-axis and y-axis quarters become contraction and expansi on, respectively, until reaching the maximum displacement at the end of the first half cycle, as illustrated by the dotted line in Figure 2.3. In the second half cycle, the motion is re versed with the continuing contraction in xaxis quarters and expansion in y-axis quarters. After reac hing the maximum displacement, the ring starts to restore towa rds the original state, finishing the entire cycle. As a result, expansion and contrac tion is similar to the radial-contour mode vibration, indicating that a high frequency can be achieved, while the displacement of the
48 inner and outer perimeters of the ring resemb les the mode shape of a wine-glass disk, which minimizes anchor-related dissipation to allow highQ performances. Figure 2.3 indicates a 2-D ring plate in the cylindrical coordinates ( r ) with origin point at the center of the ring, where the thickness of the ring is much lesser than the inner and outer radius of the ring, whose effect on vibra tion is negligible while still offering sufficient precision. The time-inde pendent radial and tangential displacement Ur and U may be expressed as  n kr DY kr CJ r n dr hr dY B dr hr dJ A h Un n n n rcos 1 2.27 n dr kr dY D dr kr dJ C hr BY hr AJ r n h Un n n nsin 1 2.28 where E hn 2 2 21 2.29 E kn 2 22 2 2.30 and E and are the density, the YoungÂ’s modulus, a nd the PoissonÂ’s ratio of the ring structure material, respectively. n is the n -th order angular resonance frequency. A B C D are the constants determined by excitation amplitude. Boundary conditions at r = Rout and r = Rin are as follows, due to the zero normal ( r) or tangential ( r ) stress at the free outer and inner ring edges : 0 1 1 12 2 2 U r E r U E r U Er r r 2.31 0 1 2 2 r U r U r r U Er r 2.32
49 Substituting equation 2.20 and 2.21 into 2.22 and 2.23 yields the resonance frequency given by 2 01 2 E h f 2.33 where h is a ring geometry-related pa rameter that satisfies  det [Hij] = 0 2.34 where the matrix elements are in n in in n in in n in in n out in n in n in in n in n in out n out out n out out n out out n out out n out n out out n out n out in n in in n in n in in n in n in in n in in n in in n in out n out out n out n out out n out n out out n out out n out out n outkR Y kR n n kR Y kR H kR J kR n n kR J kR H hR Y n n hR Y nkR H hR J n n hR J nkR H kR Y kR n n kR Y kR H kR J kR n n kR J kR H hR Y n n hR Y nkR H hR J n n hR J nkR H kR Y nkR kR Y n n H kR J nkR kR J n n H hR Y hR hR Y n n kR H hR J hR hR J n n kRH kR Y nkR kR Y n n H kR J nkR kR J n n H hR Y hR hR Y n n kR H hR J hR hR J n n kR H 2 1 2 1 1 1 2 1 2 1 1 1 1 1 1 2 1 2 1 1 1 2 1 22 1 44 2 1 43 1 42 1 41 2 1 34 2 1 33 1 32 1 31 1 24 1 23 1 2 22 1 2 21 1 14 1 13 1 2 12 1 2 11 2.35 where Rin and Rout are the inner and outer radius of the ring, respectively, Jn and Yn are the Bessel function of the fi rst and second kind, and n is the circumferential order of the mode shape.
50 Considering the extensional wine-glass mode of a ring resonator is basically the expansion and contraction in the ring width, wh ich is similar to the longitudinal vibration of a bar, the resonance frequency, thus, can be approximately expressed as  5 3 1 20 m E W n fapprox 2.36 where in outR R W is the width of the ring, and n is the order of the mode shape. 2.2 Equivalent Circuit Model Figure 2.4 The two-port electrical circuit model represented by Y -parameters. A model describing the admittance parameters ( Y -parameters) of a micromechanical disk resonator is developed to aid in analysis and design of the device. As shown in Figure 2.4, four Y -parameters in the two-port equivalent circuit model are defined as the ratio of the current measured at one port to the voltage at the driving port while short-circuiting the undriv en port , expressed as 0 22 0 21 0 12 0 11) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( d s d sv s s v d s v s d v d dj v j i j Y j v j i j Y j v j i j Y j v j i j Y 2.37 Equation 2.37 can be expressed as the product of the mechanical forcedisplacement transfer function, j F j Z /, the electromechanical coupling at the input and output ports, 1, 2, 1Â’ and 2Â’ where 1 and 2 denotes the electromechanical coupling from the driving electrode to the sensing electrode, while 1Â’ and 2Â’ denote the
51 coupling from the sensing electrode to th e driving electrode. The input and output coupling terms can be expressed as  j v j F jd1 2.38 j Z j i j j Z j Q js s 12 2.39 j Z j i j j Z j Q jd d 11 2.40 j v j F js 2 2.41 where Qs and Qd are the charge going through th e driving and sensing electrode, respectively, and the displacement Z here denotes the vibration amplitude A/R By combining the above equations, Y11 and Y21 may be rewritten as  1 1 11 j F j Z j j Z j i j v j F j F j Z j Yd d 2.42 .2 1 21 j F j Z j j Z j i j v j F j F j Z j Ys d 2.43 To determine the resonator admittance, three terms on the right side of equation 2.42 and 2.43 must be expressed in terms of the electromechanical properties of the resonator.
52 Figure 2.5 An infinitesimal element d along the circumferential direction As illustrated in Figure 2.5, along the circumferential direction, as the reference point, the effective mass for an infinitesimal element, d may be expressed as  d U R h d mR 2 2 2.44 where r d R r Ur 1 0 2is the integral for the kinetic energy. The dynamic behavior of the infinitesimal element can be described by the second-order equation of motion as e R R d Rf U R A K U R A dt d c U R A dt d m 2 cos2 2 2.45 where cd is the damping-related coefficient for the infinitesimal element and fe( ) is the radial electrostatic force per unit ra dian. Multiplying 2.45 by the mode shape, RU 2 cosand integrating from 0 to 2 gives rise to
53 d f U R A U R h R A dt d C R A dt d U R he R R d R 2 0 2 2 2 2 2 2 22 cos 1 2.46 where Cd is the damping-related coefficient of th e disk resonator. Thus, the equivalent mass and equivalent stiffness, re spectively, can be expressed as 2 2RU R h M 2.47 2 M K 2.48 The equivalent electrostatic stiffness and the equivalent force for wine-glass mode disk resonator, by substituting the electrost atic excitation force calculated in 2.24 and 2.25 to 2.46, may be expressed as 3 3 21 1 4 2 sin 2s d p e e ed d V R h K 2.49 sin2e d p R dv V U d R h F 2.50 Hence, the equivalent mechanical model fo r the disk resonator can be written by .2 2F R A K K R A dt d C R A dt d Me d 2.51 From 2.51, the first term of the admittance parameter equation (i.e. the forcedisplacement transfer function of the resonator), can be de termined via modal analysis, which is j K K Q j j K j F j Ze1 1 1 12 2.52 where Q is the quality factor of the disk resonator.
54 Substituting 2.51 to 2.38 and 2.41, the coupling of the input and output electrodes can be written as : e P R dV U d R h j sin2 1 2.53 e P R dV U d R h j sin2 2 2.54 sin2e P R sV U d R h j 2.55 sin2 2e P R sV U d R h j 2.56 Similarly, Y12 and Y22 can be derived, expressed as .2 1 22 1 2 11 21 12 Y Y Y Y 2.57 Substituting 2.522.56 to 2.42 and 2.43 gives rise to the transfer function in the form of Y -parameters of a series RLC tanks with the equivalent inductance, capacitance, and resistance expressed as, respectively  2 2 1 21 K L 2.58 K K K Ce 12 1 21 2.59 2 1 21 Q M K R 2.60
55 Figure 2.6 Electrical equivalent circuit m odel for a two-port disk resonator. Note that R21 is commonly referred to the motional resistance. Therefore, the electrical equivalent circuit model of a two-port disk resona tor may be derived as shown in Figure 2.6 where C0 represents the electro-to-r esonator overlap capacitance.
56 CHAPTER 3 MICROFABRICATION PROCESS OF THE WINE-GLASS MODE DISK RESONATOR AND ARRAY ON SOI SUBSTRATE 3.1 Microfabrication Process on SOI Substrate Given the advantages of utilizing SOI substrate described in Chapter 1, the fabrication process is designed to reduce th e complicity associated with SOI substrate while producing the best performance of the resonator device. Particularly, instead of solely applying dc-bias to the resonator body, the whole substrate is grounded in this process, allowing the possibility of mini mized feed-through capacitance. A layer of boron-doped poly-Si is deposited serving as the electrode s for ac input and output, isolated from dc by a composite film of thermally grown SiO2 and ALD high-k dielectric as shown in Figure 3.1. Figure 3.1 Cross-section view of a single wi ne-glass disk resonator on SOI substrate. The fabrication process of the wine-gla ss mode resonator on SOI substrate is based on bulk micromachining and surface micr omachining, including substrate etching,
57 several thin film deposition steps and selective removal of thin films. The entire process flow is illustrated in Figure 3.2 in cross-sectional view. Figure 3.2 Cross-section view process flow of a wine-gla ss disk resonator on SOI substrate. Starting with a SOI wafer, a 1.5 m thick layer of SiO2 is grown by wet oxidation on the SOI wafer. AZP4620 photoresist is used for the 1st photolithography st ep due to its high resistibility to plasma etch. The resist patterns are then transferred to the SiO2 layer by deep reactive ion etch (DRIE) (AMS 100, Alcatel Micro Machining Systems) using the CH4/C4F8 chemistry. With the patterned SiO2 layer as hard mask, the device Si layer
58 is etched by Si DRIE that employs Bosch process  with buried SiO2 as a stopping layer, yielding the schematic view and SEM image in Figure 3.3. Figure 3.3 SEM, cross-section, and 3-D schematic view after the 1st lithography step and Si DRIE to define the resonator body structure. After patterning the resonator structure by DRIE process, a layer of 10 nm-20 nm HfO2 is deposited to serve as the high-k dielec tric solid gap between the disk resonator body and surrounding electrodes. ALD deposition using SavannahTM S100 system from Cambridge Nanotech was used to deposit HfO2 at 200C with Hf(NMe2)4 (Strem Chemicals Inc., 99.99%) and H2O used as the precursors. Th e substrate was alternatively exposed to Hf(NMe2)4 and water vapor that were carried by nitrogen flow for a desired cycle number with 0.9 /cycle deposition rate. The HfO2 film is then annealed in N2 atmosphere at 800C for 5 minutes for the purpose of reducing its etch rate in HF solution used to remove the buried SiO2 layer and release the resonator structure.
59 Following the ALD process, 2 m of poly-silicon is deposited via LPCVD at 605C and then doped by boronÂ–nitride so lid source at 1050C, achieving Rs of 20~30 /sq. To define the electrodes, a layer of SU-8 3025 photoresist is next spun 30 m-thick to completely submerge the poly-silicon topography deep under the quasi-planarized photoresist film. After exposing and deve loping the photoresist, the poly-silicon electrodes are patterned via a combined dry an d wet Si etch to yield the cross-section shown in Figure 3.4. Figure 3.4 SEM, cross-section and 3-D view after the 2nd lithography step and Si etch to define the electrodes. A subsequent (3rd mask) layer of photoresist is spun, exposed and developed, for the purpose of releasing the resonator structur e as well as opening via to the dc bias substrate. After soaking in 49% concentrated HF, the final cross-se ction of Figure 3.5 is formed. During the releasing step, the highk dielectric capacitive solid gap is well protected because of two main reasons: (1) it is very difficult for the HF to enter the nm-
60 scale gap spacing while a large area of SiO2 is exposed directly to wards the solution; (2) N2 annealing makes the etch rate of HfO2 much lower than that of SiO2. Figure 3.5 SEM image, cross-section and 3-D view of the final device. 3.2 High-Aspect-Ratio Si DRIE DRIE is one of the most important a nd popular techniques among the options for fabricating high aspect ratio st ructures (HARS) in silicon. Th e Bosch process, also called time multiplexed deep etching (TMDE), successf ully fulfils the requirements of HARS: high etch rate, good selectivity to masking material, anisot ropy, and compatibility with other processes. The basic etching mechanism is based on etch/deposition cycle to allow silicon anisotropic etching by cycling SF6 and CF4 gases, as illustra ted in Figure 3.6.
61 Figure 3.6 Schematic view of Bosch process principle. Single crystal silicon can be etched by any ha logen atoms, such as F, Cl, Br, I, but only can react with F atoms spontaneously: 0 44 G SiF F Sig s SF6 is used in Bosch process to create F ra dicals while being dissociated in plasma: F SF SF 5 6 Silicon etch rate depends on the available F partial pressure and on the area of silicon to be removed. However, the etching of silicon in F chemistry is pure isotropic, hence another gas is essential to provide protection to the side walls. C4F8 is dissociated to -(CF2)species able to react with Si to crea te the polymer protecting the Si side walls. In the meantime, the polymer deposition on th e mask can improve the mask selectivity. To sum up, SF6/C4F8 cycling reactions in Bosch process can be expressed as:
62 Etching: 4 *4 SiF F Si (volatile molecule) Passivation: n CF CF 2 2(Teflon like material), as shown in Figure 3.7. Figure 3.7 Schematic view graph of the two-step Bosch process. The Si DRIE recipe used in this work is listed in Table 3.1. A small portion of O2 is added to C4F8 in order to achieve higher aspect ratio. During the passivation step, very weak O2 plasma is formed and a small amount of deposited polymer can be removed, thus increasing Si etch rate a nd aspect ratio while keeping the selectivity in a good range. Table 3.1 Different Si DRIE recipe used in this work. A simplified physical model can be used to simulate the Bosch process. As illustrated in Figure 3.8, for a complete deposi tion/etch cycle, within the deposition time of t1, a thin layer of polymer with a thickness of a1t1 is deposited, where a1 is the polymer deposition rate. In the etch step with a time of t2, the first portion of t2 is used to SF6 flow rate (sccm) SF6 pulse time (s) C4F8/O2 flow rate (sccm) C4F8/O2 pulse time (s) A3003200/201.42400 B3003200/211.4 2000 C300 2.6 200/221.42400 D3003200/23 1.8 2400 EtchingPassivation Cycle Source power (w) Recipe
63 remove the polymer deposited on the botto m of the feature and the rest of t2 is spent on silicon etch, which can be expressed as 3.1 where a2 is the polymer removal rate. Given the isotropic Si etch rate of a3, the etch depth achieved during one deposition/etch cycle can be written as 3.2 Figure 3.8 A simplified model for the Bosch pro cess: in one depositi on/etch cycle, the deposition step lasts for a period of t1, and the etch step for t2, which includes both polymer removal and Si isotropic etch. Therefore, an oxygen plasma cleaning proce ss is required at the end of each etch to remove the passivation layer deposited on the sidewalls of the DRIE patterned SEM image of Figure 3.9(a) shows the eviden ce of polymer accumulation on Si sidewalls after DRIE, and Figure 3.9(b) is taken from the same sample after 10 min oxygen plasma cleaning process. As seen, residues on th e sidewall are removed in oxygen plasma.
64 Figure 3.9 SEM photos of DRIE Si sidewalls (a) before and (b) after oxygen plasma treatment. High aspect-ratio DRIE is an essential step that enables fabrication of many MEMS devices such as high precision motion sensors [85, 86] and high performance low motional capacitively-transduced resonators [55, 87]. In these applications, trench aspectratio, sidewall smoothness and trench profile are amongst the critical process parameters. Optimization of these parameters will be discussed in detail in the following sections. 3.2.1 Etch Rate From equation 3.2 the etch rate can be incr eased in several ways. First, the overall Si etch rate can be increased by in creasing the isotropic Si etch rate a3, which can be realized by increasing the ICP source power, SF6 gas flow rate and pulse time. Second, if step time is fixed, the overall etch rate can also be increased by reducing the polymer deposition rate a1, which can be realize by decrease C4F8 flow rate and duration time.
65 For the purpose of obtaining better side wall smoothness and better control of the trench thickness, the etch rate of 8~9 m/min of the original recipe (see Table 3.1) needs to be reduced. The eff ect of source power, SF6 pulse time and C4F8/O2 pulse time on Si etch rate have been studied independently. Th e Si etch rate decreased as the source power or SF6 pulse time reduced and C4F8/O2 pulse time increased, which shows a perfect match to equation 3.2 as illustrated in Figure 3.10. Figure 3.10 Si etch time vs. (a) source power, (b) SF6 pulse time, and (c) C4F8/O2 pulse time.
66 3.2.2 Sidewall Smoothness One limitation of Bosch process is the sidewall Â“scallopingÂ” formed due to etch/deposition cycle. The scallop size, i.e. peak-to-peak dimension, is the etch depth achieved during etching step in one cycle, L suggesting that the scallop size is directly linked to the Si etch rate. T hus high etch rate is often ach ieved at the expense of rougher sidewalls in Bosch process. As demonstrated in section 3.2.1, Si etch rate decreases when source power, SF6 pulse time increases and C4F8/O2 pulse time decreases. SEM images in Figure 3.11 shows a reduction of scalloping size when the Si DRIE recipes (i.e. recipe B, C, and D shown in Table 3.1) is modified to realize lower etch rate. Figure 3.11 SEM photos of Si side wall scalloping formed by di fferent Si DRIE recipes: (a) original recipe A; (b) recipe B with re duced source power; (c) r ecipe C with reduced SF6 pulse time; (d) recipe D with increased C4F8/O2 pulse time.
67 3.2.3 Aspect Ratio Dependent Etching (ARDE) MEMS devices usually have structures with different dimensions and aspect ratios that coexist on a single microchip. However, Si etching is aspect ratio dependent, which can be manifested in two ways: first, fo r a specific feature, et ch rate decreases as the aspect ratio increases over time; second, fo r features with different dimensions, etch rate is higher of wider feat ures than narrower features. The Si etch rate decreases when etch time increases, which dropped from ~ 10 m/min in a 1 min etch to ~ 7.5 m/min in a 9 min one, as de picted in Figure 3.12. The etch rate reduction could be caused by insu fficient passivation layer etching during a deposition/etch cycle. Therefore, the polym er deposited at the bottom of the etched trenches accumulates as the etch time increases, leading to a longer polymer removal step time and a shorter Si isotropic etch step time, thus a reduced Si etch rate. Figure 3.12 Overall Si etch rate decr eases with increasing etch time. In addition, as the aspect ratio of the trench increases, effective removal of the passivation layer becomes more crucial mainly because of the decayed ion flux down to
68 the bottom of the trench (see Figure 3.13). Hence, the Si etch rate decreases as the trench dimension decreases. Figure 3.13 Si etch rate decreases dramati cally as aspect ratio increases. 3.3 Atomic Layer Deposition (ALD) Figure 3.14 Schematic concept of ALD process. Atomic Layer Deposition (ALD) is a gas phase deposition method for ultra-thin films . Comparing to other deposition techniques, ALD has many advantages, such as excellent conformal and uniform coating on a very large area, as well as thickness and composition control at atomic level. ALD is based on self-limiting surface reactions.
69 During ALD process, two (or more) precursors are pulsed onto the substrate alternatively, and the pulses are separated by purges with an inert gas. Generally, a b c d is referred as one cycle, where a is the pulse time of the first precursor, c is that of the second precursor, b and d indicate the purging period of inert gas after each pulse of precursor. In each cycle, a series of saturative reactions happen on the substrate surface and an atomic layer of target material is formed, describe d in Figure 3.14. This self-limiting growth is the key feature for ALD. During each cycle, gaseous precursor reacts with the solid substrate surface. Atoms which are included in the target material are absorbed to the substrate surface until the amount of atoms sa turate. Simultaneously, atoms which are not included in the target materi al are removed as reaction byproducts . Thus, as the number of cycles is determined, the target material with accura te thickness should be expected. Figure 3.15 shows a SEM phot o of a ~ 50 nm layer of HfO2 deposited by ALD on Si substrate. Figure 3.15 SEM image of a thin film of HfO2 of ~ 50 nm deposited by ALD on Si substrate. Table 3.2 summarizes the precursors and depos ition rate at different temperatures of HfO2, showing that deposition rate increases as deposition temperature decreases.
70 Table 3.2 Precursors, deposition condition and rate of HfO2 deposited by ALD. In order to protect the high-k dielectric material deposited by ALD in the SiO2 releasing process, thus enabling the optimized di electric transducer, the etch rate in HF of HfO2 film at different deposition temperatur e is then investigated. As-deposit HfO2 film is etched in 50:1 HF and the film deposited at lower temperature showed lower etch rate, thus higher resistivity to HF, as illustrated in Figure 3.16. HfO2 film is then annealed at 800oC for 5 minutes in a tube furnace with an N2 ambient at atmospheric pressure and etched in 49% pure HF. Etch rate of annealed HfO2 film in HF drops dramatically, which is almost zero in pure HF. Thickness of HfO2 films is measured by Rudolph ellipsometer with reflective index of 2.05. Figure 3.16 Etch rate in HF of HfO2 deposited at different temp erature and with different post-deposition treatment. Flow rate (sccm)Pulse H2O (s)Wait (s)Pulse Hf(NMe2)4 (s)Wait (s) Deposition rate ( /cycle) 150200.015600.15601.0523 200200.015250.1250.9012 250200.0350.350.8444 Temperature (oC) Deposition condition
71 In addition, grazing incidence x-ray di ffraction (GI-XRD) measurement obtained by Philips PW3040 XÂ’pert PRO system indicates that not only the deposition rate, crystal structure of these materials also depends on the deposition temperature and annealing treatments, as demonstrated in Figure 3.17. XRD patterns show that all peaks grow dramatically when ALD HfO2 films are annealed, suggestin g a much better crystallized film is obtained after annealing. Moreover, as deposition temperature decreases, all monoclinic peaks (i.e., oriented in (110), (200 ), and (201)) grow in intensity while the peak near 2 = 30.4o does not, which corresponds to eith er a tetragonal or orthorhombic phase. This observation suggested that the tetragonal or orthorhombic phase is preferred in higher deposition temperature. Figure 3.17 XRD spectra of ALD HfO2 films on thermal oxide underlayer showing the effect of deposition temper ature and annealing. ALD HfO2 film is served as a high-k diel ectric solid gap in the capacitivelytransduced resonator, thus ma king the dielectric constant a crucial property of the film.
72 High frequency (HF) Capacitance-Voltage (C -V) measurements are performed on the asdeposit and N2 annealed ALD HfO2 films with a deposition temperature of 150oC. Aluminum top circular electrodes with di ameter of 1 mm are deposited by an e-beam evaporator through a shadow mask and HP 4145B was used for C-V measurements. Figure 3.18 HF C-V measurement of a MOS capacitor with ALD HfO2 film as dielectric. As shown in Figure 3.18, the capacitance of the ALD HfO2 film increases from 800 pF to 2.5 nF after annealing. With an as sumption that the capacitance of the HfO2 film equals to accumulation capacitance, the dielectric constant can be calculated as 3.3 where Cox is the capacitance of the dielectric layer, d is the thickness of the HfO2 film, A is the area of the aluminum electrode, and 0 is the dielectric constant of a free space. Hence, the calculated dielectric constant of the ALD HfO2 film deposited at 150oC is 6.06 and 15.66 for as-deposit and annealed film, respectively, suggesting that annealed HfO2 is preferred in this process not only because of its high resis tivity to HF solution but also because of its much higher dielectric constant.
73 CHAPTER 4 CONCLUSION Due to the highQ and CAD-layout definable ultr ahigh resonance frequencies, MEMS resonators offer a promising on-chip replacement to eliminat e the board-level RF passives, thus leading to ultimate miniat urization and performance improvement. In particular, the future transceiver architec ture can benefit from large arrays of micromechanical resonator filters, becau se of their tiny si zes and low power consumption, and possibility for multiband sel ection. It is beyond doubt that researches on MEMS devices for wireless communication on device and system level still have a boarder space to develop and will make influence in the near future. However, the excessive motional impeda nce has impeded further development of the capacitively-transduced MEMS resonato rs. Motional impedance is a function of several parameters, including the disk-to-el ectrode overlap area, gap distance of the dielectric, and dielectric cons tant of the gap material. SOI technique provides a promising approach to greatly enhance the overlap area thus reduce the motional impedance, while offering better mechanical and thermal charact eristics of single crystal silicon comparing to conventionally used polycrys talline silicon. Therefore, this work is preceded for the purpose of designing a novel fa brication process of capac itively-transduced MEMS resonators and resonator arrays on SOI subs trate, which can be CMOS compatible and much easier to realize than curr ently available processes. Bosch process is used to etch the device Si layer of a SOI wafer in order to define the resonator body structure. In sight investigation has been performed for the purposes of
74 optimizing Si etch rate, Si trench si dewall smoothness and profile. The optimized parameters, including gas pulse time, fl ow rate, power, and post-etching surface treatment, are experimentally demonstrated in this work. The solid capacitive gap is deposited by AL D, and high-k dielectric materials are chosen. The dielectric constant of these high-k materials under different deposition conditions and different post-deposit ion treatments has been studied.
75 LIST OF REFERENCES  P. V. David Tse, Fundamentals of Wireless Communication : Cambridge University Press, 2005.  T. P. Liu and E. Westerwick, "5-GHz CMOS radio transceiver front-end chipset," Ieee Journal of Solid-State Circuits, vol. 35, pp. 1927-1933, 2000.  S.-J. C. a. Y.-H. Hsieh, "Tra nsceiver Arechitecture Design," in IQ Calibration Techniques for CMOS Radio Transceivers : Springer Netherlands, 2006, pp. 1142.  M. Brandolini, P. Rossi, D. Manstretta, and F. Svelto, "Toward multistandard mobile terminals Fully integrated recei vers requirements and architectures," Ieee Transactions on Microwave Theory and Techniques, vol. 53, pp. 1026-1038, 2005.  D. Jakonis, K. Folkesson, J. Dabrowski, P. Eriksson, and C. Svensson, "A 2.4GHz RF sampling receiver fron t-end in 0.18-mu m CMOS," Ieee Journal of Solid-State Circuits, vol. 40, pp. 1265-1277, 2005.  C. T. C. Nguyen, "Transceiver fr ont-end architectures using vibrating micromechanical signal processors," Rf Technologies for Low Power Wireless Communications, pp. 411-461, 2001.  J. Wang, J. E. Butler, T. Feygelson, and C. T. C. Nguyen, "1.51-GHz nanocrystalline diamond micromechanical disk resonator with materialmismatched isolating support," in IEEE International Conference on. (MEMS) 2004, pp. 641-644.  Y. Xie, S. S. Li, Y. W. Lin, Z. Y. Ren, and C. T. C. Nguyen, "1.52-GHz micromechanical extensional wine -glass-mode ring resonators," IEEE Transactions on Ultrasonics Ferroele ctrics and Frequency Control, vol. 55, pp. 890-907, 2008.  J. Wang, Z. Y. Ren, and C. T. C. Nguyen, "1.156-GHz self-aligned vibrating micromechanical disk resonator," IEEE Transactions on Ultrasonics Ferroelectrics and Frequency Control, vol. 51, pp. 1607-1628, 2004.
76  D. Weinstein, S. A. Bhave, S. Mori ta, S. Mitarai, and K. Ikeda, "Frequency scaling and transducer efficiency in inte rnal dielectrically transduced silicon bar resonators," in Solid-State Sensors, Actuator s and Microsystems Conference, 2009. TRANSDUCERS 2009. International 2009, pp. 708-711.  D. Weinstein and S. A. Bhave, "Int ernal Dielectric Tran sduction of a 4.5 GHz Silicon Bar Resonator," in IEEE International Elect ron Devices Meeting 2007, pp. 415-418.  H. Wan-Thai and C. T. C. N guyen, "Stiffness-compensated temperatureinsensitive micromechanical resonators," in Micro Electro Mec hanical Systems, 2002. The Fifteenth IEEE Inte rnational Conference on 2002, pp. 731-734.  M. A. Hopcroft, H. K. Lee, B. Kim, R. Melamud, S. Chandorkar, M. Agarwal, C. Jha, J. Salvia, G. Bahl, H. Mehta, and T. W. Kenny, "A High-Stability MEMS Frequency Reference," in Solid-State Sensors, Ac tuators and Microsystems Conference, 2007. TRANSDUCERS 2007. International 2007, pp. 1307-1309.  C. Kuan-Lin, S. Wang, J. Salvia, R. T. Howe, and T. W. Kenny, "Encapsulated out-of-plane differential square-plate resonator with integrated actuation electrodes," in Solid-State Sensors, Actuato rs and Microsystems Conference, 2009. TRANSDUCERS 2009. International 2009, pp. 1421-1424.  S.-S. Li, Y.-W. Lin, Z. Ren, and C. T. C. Nguyen, "Self-switching vibrating micromechanical filter bank," in Frequency Control Symposium and Exposition, 2005. Proceedings of the 2005 IEEE International 2005, p. 7 pp.  H. Nakahata, H. Kitabayashi, T. Ue mura, A. Hachigo, K. Higaki, S. Fujii, Y. Seki, K. Yoshida, and S. Shikata, "St udy on surface acoustic wave characteristics of SiO2/interdigital-transducer/ZnO/diam ond structure and fabrication of 2.5 GHz narrow band filter," Japanese Journal of Applied Physics Part 1-Regular Papers Brief Communications & Review Papers, vol. 37, pp. 2918-2922, 1998.  A. Gaidarzhy, M. Imboden, P. Moha nty, J. Rankin, and B. W. Sheldon, "High quality factor gigahertz frequencies in nanomechanical diamond resonators," Applied Physics Letters, vol. 91, 2007.  N. Kacem, S. Hentz, D. Pinto, B. Reig, and V. Nguyen, "Nonlinear dynamics of nanomechanical beam resonators: impr oving the performance of NEMS-based sensors," Nanotechnology, vol. 20, 2009.  S. B. Cohn, "Microwave Bandpass Filters Containing High-Q Dielectric Resonators," IEEE Transactions on Microwav e Theory and Techniques, vol. 16, pp. 218-227, 1968.  S. J. Fiedziuszko, I. C. Hunter, T. Ito h, Y. Kobayashi, T. Nishikawa, S. N. Stitzer, and K. Wakino, "Dielectric materi als, devices, and circuits," IEEE Transactions on Microwave Theory and Techniques, vol. 50, pp. 706-720, 2002.
77  J. K. Plourde and C. L. Ren, "Applica tion of dielectric resonators in microwave components," IEEE Transactions on Microwave Theory and Techniques, vol. 29, pp. 754-770, 1981.  R. D. Richtmyer, "Dielectric resonators," Journal of Applied Physics, vol. 10, pp. 391-398, 1939.  J. Viennet, M. Jardino, R. Barillet, and M. Desaintfuscien, "Frequency control loop with digital integrator," IEEE Transactions on Instrumentation and Measurement, vol. 32, pp. 322-326, 1983.  A. S. Matistic, "Quartz-crysta l timing accuracy is hard to beat," Electronic Design|Electronic Design, vol. 24, pp. 74-9, 1976.  T. Ormond, "Crystal-oscillators-s ignal sources handle tough timing jobs," Edn, vol. 35, p. 93, 1990.  W. P. Mason, "Electrical wave filters employing quartz crystals as elements," Bell System Technical Journal, vol. 13, pp. 405-452, 1934.  P. Lloyd, "Monolithic crystal filters," Proceedings of the 7th international congress on acoustics|Proceedings of the 7th international congress on acoustics, pp. 309-12|xvi+568, 1971.  C. P. Yue and S. S. Wong, "On-chip Spiral Inductors With Patterned Ground Shields For Si-based RF IC's," in VLSI Circuits, 1997. Di gest of Technical Papers., 1997 Symposium on 1997, pp. 85-86.  X. M. H. Huang, C. A. Zorman, M. Mehregany, and M. L. Roukes, "Nanodevice motion at microwave frequencies," Nature, vol. 421, pp. 496-496, 2003.  D. Weinstein and S. A. Bhave, "The Resonant Body Transistor," Nano Letters, 2010.  J. R. Vig and Y. Kim, "Noise in mi croelectromechanical system resonators," IEEE Transactions on Ultrasonics Ferro electrics and Frequency Control, vol. 46, pp. 1558-1565, 1999.  Y. Satoh, T. Nishihara, T. Yokoyama M. Ueda, and T. Miyashita, "Development of piezoelectric thin film resonato r and its impact on future wireless communication systems," Japanese Journal of Applie d Physics Part 1-Regular Papers Brief Communications & Review Papers, vol. 44, pp. 2883-2894, 2005.  D. P. Morgan, "History of saw devices," Proceedings of the 1998 Ieee International Frequen cy Control Symposium, pp. 439-460, 1998.
78  Z. Hao and K. Eun Sok, "Air-backe d Al/ZnO/Al film bulk acoustic resonator without any support layer," in Frequency Control Symposium and PDA Exhibition, 2002. IEEE International 2002, pp. 20-26.  K. Umeda, H. Kawamura, M. Takeuchi and Y. Yoshino, "Characteristics of an AlN-based bulk acoustic wave resonator in the super high frequency range," Vacuum, vol. 83, pp. 672-674, 2008.  J. B. Lee, H. J. Kim, S. G. Kim, C. S. Hwang, S. H. Hong, Y. H. Shin, and N. H. Lee, "Deposition of ZnO thin films by magnetron sputtering for a film bulk acoustic resonator," Thin Solid Films, vol. 435, pp. 179-185, 2003.  M. A. Dubois and P. Muralt, "Propert ies of aluminum nitr ide thin films for piezoelectric transducers and microwave filter applications," Applied Physics Letters, vol. 74, pp. 3032-3034, 1999.  K. M. Lakin, G. R. Kline, and K. T. McCarron, "Development of miniature filters for wireless applications," IEEE Transactions on Microwave Theory and Techniques, vol. 43, pp. 2933-2939, 1995.  G. Piazza, P. J. Stephanou, and A. P. Pisano, "One and two port piezoelectric higher order contour-mode MEMS resonators for mechanical signal processing," Solid-State Electronics, vol. 51, pp. 1596-1608, 2007.  G. Piazza, P. J. Stephanou, and A. P. Pisano, "Single-chip multiple-frequency ALN MEMS filters based on contour-mode piezoelectric resonators," Journal of Microelectromechanical Systems, vol. 16, pp. 319-328, 2007.  R. H. Olsson, K. E. Wojciechowski, M. S. Baker, M. R. Tuck, and J. G. Fleming, "Post-CMOS-Compatible Aluminum Nitr ide Resonant MEMS Accelerometers," Journal of Microelect romechanical Systems, vol. 18, pp. 671-678, 2009.  G. Piazza, P. J. Stephanou, and A. P. Pisano, "Piezoelectric aluminum nitride vibrating contour-mode MEMS resonators," Journal of Microelectromechanical Systems, vol. 15, pp. 1406-1418, 2006.  C. T. C. Nguyen, "MEMS technology for timing and frequency control," Ieee Transactions on Ultrasonics Ferroele ctrics and Frequency Control, vol. 54, pp. 251-270, 2007.  R. Raiteri, M. Grattarola, H.-J. Butt, and P. Skldal, "Micromechanical cantileverbased biosensors," Sensors and Actuators B: Chemical, vol. 79, pp. 115-126, 2001.  Y. W. Lin, S. Lee, S. S. Li, Y. Xie, Z. Y. Ren, and C. T. C. Nguyen, "Seriesresonant VHF micromechanical resonator reference oscillators," Ieee Journal of Solid-State Circuits, vol. 39, pp. 2477-2491, 2004.
79  F. D. Bannon, J. R. Clark, a nd C. T. C. Nguyen, "High-Q HF microelectromechanical filters," IEEE Journal of Solid-State Circuits, vol. 35, pp. 512-526, 2000.  K. Wang, A. C. Wong, and C. T. C. Nguyen, "VHF free-free beam high-Q micromechanical resonators," Journal of Microelectromechanical Systems, vol. 9, pp. 347-360, 2000.  A. N. Cleland and M. L. Roukes, "F abrication of high fre quency nanometer scale mechanical resonators from bulk Si crystals," Applied Physics Letters, vol. 69, pp. 2653-2655, 1996.  W. Jing, R. Zeying, and C. T. C. Nguyen, "Self-aligne d 1.14-GHz vibrating radial-mode disk resonators," in TRANSDUCERS, Solid-State Sensors, Actuators and Microsystems, 12th International Conference on, 2003 2003, pp. 947-950 vol.2.  J. R. Clark, H. Wan-Thai, and C. T. C. Nguyen, "High-Q VHF micromechanical contour-mode disk resonators," in Electron Devices Meeting, 2000. IEDM Technical Digest. International 2000, pp. 493-496.  M. A. Abdelmoneum, M. U. Demirci, and C. T. C. Nguyen, "Stemless wineglass-mode disk micromechanical resonators," in Micro Electro Mechanical Systems, 2003. MEMS-03 Kyoto. IEEE Th e Sixteenth Annual International Conference on 2003, pp. 698-701.  W. Jing, Z. Y. Ren, and C. T. C. Nguyen, "1.156-GHz self-aligned vibrating micromechanical disk resonator," Ieee Transactions on Ultr asonics Ferroelectrics and Frequency Control, vol. 51, pp. 1607-1628, 2004.  L. Sheng-Shian, L. Yu-Wei, X. Yu an, R. Zeying, and C. T. C. Nguyen, "Micromechanical "hollow-disk" ring resonators," in Micro Electro Mechanical Systems, 2004. 17th IEEE Intern ational Conference on. (MEMS) 2004, pp. 821824.  T. J. Cheng, Bhave, S. A. "HighQ, low impedance polysilicon resonators with 10 nm air gaps," in IEEE International Conferen ce on Micro Electro Mechanical Systems 2010, pp. 695-698.  S. Pourkamali, Z. Hao, and F. Ay azi, "VHF single crysta l silicon capacitive elliptic bulk-mode disk resonators Part II: Implementation a nd characterization," Journal of Microelect romechanical Systems, vol. 13, pp. 1054-1062, 2004.  V. T. Srikar and S. D. Senturia, "Thermoelastic damping in fine-grained polysilicon flexural beam resonators," Journal of Microelectromechanical Systems, vol. 11, pp. 499-504, 2002.
80  K. Y. Yasumura, T. D. Stowe, E. M. Chow, T. Pfafman, T. W. Kenny, B. C. Stipe, and D. Rugar, "Quali ty factors in micronand s ubmicron-thick cantilevers," Journal of Microelect romechanical Systems, vol. 9, pp. 117-125, 2000.  M. Bruel, "Silicon on insulator material technology," Electronics Letters, vol. 31, pp. 1201-1202, 1995.  Y. Taur, D. A. Buchanan, W. Chen, D. J. Frank, K. E. Ismail, S. H. Lo, G. A. SaiHalasz, R. G. Viswanathan, H. J. C. Wann, S. J. Wi nd, and H. S. Wong, "CMOS scaling into the nanometer regime," Proceedings of the Ieee, vol. 85, pp. 486-504, 1997.  D. W. Carr and H. G. Craighead, "Fab rication of nanoelectromechanical systems in single crystal silicon using silicon on insulator substrates and electron beam lithography," Journal of Vacuum Science & Technology B, vol. 15, pp. 27602763, 1997.  L. A. Giannuzzi and F. A. Stevie "A review of focused ion beam milling techniques for TEM specimen preparation," Micron, vol. 30, pp. 197-204, 1999.  D. Grogg, N. D. Badila-Ciressan, and A. M. Ionescu, "Focussed ion beam based fabrication of micro-electro-mechanical resonators," Microsystem TechnologiesMicro-and Nanosystems-Information St orage and Processing Systems, vol. 14, pp. 1049-1053, 2008.  S. Pourkamali and F. Ayazi, "SOI-b ased HF and VHF si ngle-crystal silicon resonators with SUB-100 nanometer vertical capacitive gaps," in TRANSDUCERS, Solid-State Sensors, Actuators and Microsystems, 12th International Conference on, 2003 2003, pp. 837-840 vol.1.  Z. Hao, S. Pourkamali, and F. Ayazi, "VHF single-crystal s ilicon elliptic bulkmode capacitive disk resonators Part I: Design and modeling," Journal of Microelectromechanical Systems, vol. 13, pp. 1043-1053, 2004.  C. Nicoleta Diana Badila, H. Cyrille, M. Marco, and M. I. Adrian, "Fabrication of silicon-on-insulator MEM resonators w ith deep sub-micron transduction gaps," Microsyst. Technol., vol. 13, pp. 1489-1493, 2007.  D. Grogg, H. C. Tekin, N. D. CiressanBadila, D. Tsamados, M. Mazza, and A. M. Ionescu, "Bulk Latera l MEM Resonator on Thin SOI With High Q-Factor," Microelectromechanical Systems, Journal of, vol. 18, pp. 466-479, 2009.  M. U. Demirci, M. A. Abdelmoneu m, and C. T. C. Nguyen, "Mechanically corner-coupled square microresonator array for reduced series motional resistance," in TRANSDUCERS, Solid-State Sensors, Actuators and Microsystems, 12th International Conference on, 2003 2003, pp. 955-958 vol.2.
81  W. Jing, J. E. Butler, T. Fe ygelson, and C. T. C. Nguyen, "1.51-GHz nanocrystalline diamond micromechanical disk resonator with materialmismatched isolating support," in IEEE International Conference on. (MEMS) 2004, pp. 641-644.  N. Sepulveda, L. Jing, D. M. Aslam, and J. P. Sullivan, "High-Performance Polycrystalline Diamond Mi croand Nanoresonators," Microelectromechanical Systems, Journal of, vol. 17, pp. 473-482, 2008.  V. P. Adiga, A. V. Sumant, S. Sures h, C. Gudeman, O. Auciel lo, J. A. Carlisle, and R. W. Carpick, "Mechanical stiffne ss and dissipation in ultrananocrystalline diamond microresonators," Physical Review B, vol. 79, 2009.  A. R. Krauss, O. Auciello, D. M. Grue n, A. Jayatissa, A. Sumant, J. Tucek, D. C. Mancini, N. Moldovan, A. Erdemir, D. Er soy, M. N. Gardos, H. G. Busmann, E. M. Meyer, and M. Q. Ding, "Ultrananocry stalline diamond thin films for MEMS and moving mechanical assembly devices," Diamond and Related Materials, vol. 10, pp. 1952-1961, 2001.  C. Frster, V. Cimalla, V. Lebedev, J. Pezoldt, K. Brueckner, R. Stephan, M. Hein, E. Aperathitis, and O. Ambacher, "Group III-nitride a nd SiC based microand nanoelectromechanical resonato rs for sensor applications," physica status solidi (a), vol. 203, pp. 1829-1833, 2006.  J. R. Clark, W. T. Hsu, M. A. Abdelmoneum, and C. T. C. Nguyen, "High-Q UHF micromechanical radial-cont our mode disk resonators," Microelectromechanical Systems, Journal of, vol. 14, pp. 1298-1310, 2005.  D. J. Young, I. E. Pehlivanoglu, and C. A. Zorman, "Silicon carbide MEMSresonator-based oscillator," Journal of Micromechanics and Microengineering, vol. 19, 2009.  H. Wen-Lung, L. Sheng-Shian, R. Zeying, and C. T. C. Nguyen, "UHF Nickelmicromechanical Spoke-Supported Ring Resonators," in Solid-State Sensors, Actuators and Microsyste ms Conference, 2007. TRANSDUCERS 2007. International 2007, pp. 323-326.  H. Wen-Lung, R. Zeying, L. Yu-Wei, C. Hsien-Yeh, J. Lahann, and C. T. C. Nguyen, "Fully monolithic CMOS nickel mi cromechanical resonator oscillator," in Micro Electro Mechanical Syst ems, 2008. MEMS 2008. IEEE 21st International Conference on 2008, pp. 10-13.  A. E. H. Love, A treatise on the mathematical theory of elasticity : University Press (Cambridge), 1906.  S. H. C. Stanley and L. Ta Ming, "Exten sional vibration of thin plates of various shapes," The Journal of the Acoustic al Society of America, vol. 58, pp. 828-831, 1975.
82  N. H. Farag and J. Pan, "Modal characteristics of in-p lane vibration of circular plates clamped at the outer edge," The Journal of the Acoustical Society of America, vol. 113, pp. 1935-1946, 2003.  M. Onoe, "Contour Vibrations of Isotropic Circular Plates," The Journal of the Acoustical Society of America, vol. 28, pp. 1158-1162, 1956.  T. Takano, H. Hirata, and Y. Tomika wa, "Analysis of nonaxisymmetric vibration mode piezoelectric annular plate and tis application to an ultrasonic motor," IEEE Transactions on Ultrasonics Ferroele ctrics and Frequency Control, vol. 37, pp. 558-565, 1990.  D. M. Pozar, Microwave Engineering 2nd ed.: WILEY, 1998.  D. L. DeVoe, "Piezoelectric thin film micromechanical beam resonators," Sensors and Actuators A: Physical, vol. 88, pp. 263-272, 2001.  C. L. Chang, Y. F. Wang, Y. Kanamori, J. J. Shih, Y. Kawai, C. K. Lee, K. C. Wu, and M. Esashi, "Etching submicromete r trenches by using the Bosch process and its application to the fabrication of antireflection structures," Journal of Micromechanics and Microengineering, vol. 15, pp. 580-585, 2005.  A. Sharma, F. M. Zaman, B. V. Amin i, and F. Ayazi, "A high-Q in-plane SOI tuning fork gyroscope," in Sensors, 2004. Proceedings of IEEE 2004, pp. 467470 vol.1.  B. V. Amini, R. Abdolvand, and F. Ayazi, "Sub-micro-gravity capacitive SOI microaccelerometers," in Solid-State Sensors, Actuators and Microsystems, 2005. Digest of Technical Papers. TRAN SDUCERS '05. The 13th International Conference on 2005, pp. 515-518 Vol. 1.  S. Pourkamali, G. K. Ho, and F. Ayazi, "Vertical capacitive SiBARs," in Micro Electro Mechanical Systems, 2005. MEMS 2005. 18th IEEE International Conference on 2005, pp. 211-214.  M. Leskel and M. Ritala, "Atomic la yer deposition (ALD): from precursors to thin film structures," Thin Solid Films, vol. 409, pp. 138-146, 2002.  R. L. Puurunen, "Surface chemistry of atomic layer deposition: A case study for the trimethylaluminum/water process," Journal of Applied Physics, vol. 97, pp. 121301-52, 2005.
84 Appendix A: Polysilicon Disk Resonator on SOI Substrate Process Traveler A.1 Deposit Isolation Layers A.1.1 Thermally grow 2 um SiO2 on SOI substrate Equipment: FNB2 A.2 Pattern Resonator Body Structure (Mask 1) A.2.1 Lithography AZ4620 ~5.0 um Dehydration bake: 5 min @ 150C Spin: Laura Spinner HMDS: 20 sec @ 1500 RPM AZ4620: 20 sec @ 1500 RPM 50 sec @ 4000 RPM 10 sec @ 6000 RPM Softbake: 5 min @ 100C Exposure: 5 sec @ 25 mW/cm2, vacuum contact Develop: 4 min in diluted (4:1) AZ 400K A.2.2 Descum Equipment: Plasma Therm O2: 50 sccm Pressure: 300 mTorr Power: 100 watts Time: 10 min A.2.3 SiO2 DRIE Equipment: Alcatel AMS 100
85 Appendix A (Continued) C4F8: 17 sccm He: 150 sccm CH4: 13 sccm Power: 2800 watts Time: 1.5 min O2 chamber clean: 1.5 min A.2.4 Si DRIE Equipment: AMS 100, Alcatel Vacuum Technology, France SF6: 300 sccm, 3 sec C4F8: 200 sccm; O2: 20 sccm, 1.4 sec Power: 2400 watts Pulsed power: 25 ms @ 100 watts; 75 ms @ 0 watts Substrate temperature: -15 C Time: 4 min Etch rate: ~0.8 um/min O2 chamber clean: 5 min A.3 Deposit and Anneal Dielectric layer (HfO2) A.3.1 Deposit HfO2 by Atomic Layer Deposition (ALD) ~ 20 nm Equipment: Savannah, Cambridge NanoTech Inc. USA Temperature: I=O=200 C, T=V=B=150 C Flow rate: 20 sccm Recipe: pulse H2O, 0.015 sec
86 Appendix A (Continued) wait 25 sec pulse Hf(NMe2)4, 0.1 sec wait 25 sec cycle 220 Deposition rate: 0.9 /cycle A.3.2 HfO2 Anneal Equipment: tube furnace Temperature: 800 C Gas: N2 Time: 5 min A.4 Deposit and Dope Poly-Si Layer A.4.1 Deposit poly-Si by Low Pressure Chem ical Vapor Deposition (LPCVD) ~2 um Equipment: FNB 4 Program: Z9641 Temperature: 605 C Time: 290 min Deposition rate: ~70 /min A.4.1 Dope poly-Si using solid source Equipment: FNB 3 Solid source: BN1050, Saint-Gobain Ceramics & Plastics, Inc., USA A.4.1.1 Predeposition
87 Appendix A (Continued) Program: M4007 Temperature: 1050 C Time: 30 min A.4.1.2 Drive-in Program: M4008 Temperature: 1050 C Time: 30 min A.4.1.3 Deglaze 20 min in 6:1 BOE A.4.1.4 Low temperature dry oxidation Program: M4005 Temperature: 900 C Time: 20 min A.4.1.5 Deglaze 20 ~ 50 min in 6:1 BOE until the surface become hydrophilic A.4.1.6 Sheet resistance measurement ~ 20 Ohm/sq Equipment: four-point probe A.5 Pattern Poly-Si Electrodes (Mask 2) A.5.1 Lithography SU-8 3025 ~ 50 um Dehydration bake: 5 min @ 150 C Spin: Laura Spinner 10 sec @ 500 RPM
88 Appendix A (Continued) 40 sec @ 2000 RPM Soft bake: 12 min @ 95 C Exposure: 20 sec @ 25 mW/cm2, hard contact, with high-wavelength pass filter Post exposure bake: 1 min @ 65 C; 4 min @ 95 C Develop: 8 min in SU-8 developer A.5.2 Descum Equipment: Plasma Therm O2: 50 sccm Pressure: 300 mTorr Power: 100 watts Time: 10 min A.5.3 Poly-Si RIE Equipment: Plasma Therm SF6: 50 sccm Pressure: 100 mTorr Power: 100 watts Time: 4 min A.6 Pattern Opens to Substr ate Contacts (Mask 3) A.6.1 Lithography S1827 ~2.8 um Equipment: Laura Spinner Spin time: 30 sec @ 3500 RPM
89 Appendix A (Continued) Softbake time: 1 min @ 115C Exposure time: 14 sec @ 25 mW Develop time: 70 sec