990 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 18, NO. 5, OCTOBER 2009 A Liquid–Solid Direct Contact Low-Loss RF Micro Switch Prosenjit Sen and Chang-Jin Kim Abstract—This paper reports the design, fabrication, and testing of a liquid-metal (LM) droplet-based radio-frequency microelectromechanical systems (RF MEMS) shunt switch with dc-40 GHz performance. The switch demonstrates better than 0.3 dB insertion loss and 20 dB isolation up to 40 GHz, achieving significant improvements over previous LM-based RF MEMS switches. The improvement is attributed to use of electrowetting on dielectric (EWOD) as a new actuation mechanism, which allows design optimized for RF switching. A two-droplet design is devised to solve the biasing problem of the actuation electrode that would otherwise limit the performance of a single-droplet design. The switch design uses a microframe structure to accurately position the liquid–solid contact line while also absorbing variations in deposited LM volumes. By sliding the liquid–solid contact line electrostatically through EWOD, the switch demonstrates bounceless switching, low switch-on time (60 μs), and low power consumption (10 nJ per cycle). [2008-0266] Index Terms—Electrowetting on dielectric (EWOD), liquidmetal droplet, microswitch, radio-frequency microelectromechanical systems (RF MEMS). I. I NTRODUCTION T HE CONTACT reliability of solid–solid contacts becomes more important for microelectromechanical systems (MEMS) switches as surface plays an increased role at microscale. The contact degradation [1] due to arcing, welding, and material transfer, which is aggravated by contact bounce, limits the operational life of all electric switches and in a greater degree for MEMS switches [2]. In order to solve the contact problems and enhance reliability, a liquid was placed at the contact point of a surface-micromachined switch as early as 1996 [3]. However, the high surface tension of liquid-metal (LM) becomes dominant in the reduced scale, making the development of microscale replicas of macroscale reed relays difficult. Surface-micromachined switches based on low actuation force mechanisms like electrostatic found their operational speed slowed down by the high surface forces between the liquid and the solid [4]. Better performance was possible for larger microswitches and through use of large force actuation mechanisms like electrothermal, which require more energy per actuation cycle [5]. Manuscript received October 29, 2008; revised July 9, 2009. First published September 11, 2009; current version published September 30, 2009. This work was supported by the DARPA HERMIT program. Subject Editor K. F. Bohringer. The authors are with Mechanical and Aerospace Engineering Department, University of California (UCLA), Los Angeles, CA 90095 USA (e-mail: senp@ ucla.edu). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JMEMS.2009.2029170 A more elegant approach in the development of an LMbased microswitch is through actuation of the LM droplet to achieve switching with no movable structures. LM droplets have been actuated by electrothermal [6], [7], electrostatic [8], [9], and electrowetting-on-dielectric (EWOD) [10]–[12] methods to achieve switching. Collectively, LM microswitches have demonstrated no contact bounce [6], [7], [12], low switchon time (60 μs) [12], fast signal rise/fall time (5 μs) [12], low contact resistance [5], long life [7], and the capability to handle large currents (1 A) [7]. A more comprehensive review of LM microswitches is presented in [13]. Although some of the aforementioned devices have been tested for radiofrequency (RF) performance, they have mostly been limited to 20 GHz. Most of these switches also suffered from slow actuation speeds, leading to slow switching with latencies on the order of 1 ms. The first known implementation of an LM-based microswitch handling RF signals reported an RF performance of 40 dB isolation and 0.1 dB insertion loss up to 2 GHz [6]. In this implementation, an LM droplet in a microchannel filled with deionized water was toggled by the fluidic pulse generated by thermal bubble expansion. However, the presence of water in the OFF-state would have severely degraded the isolation at higher frequencies. The switch-on latency was ∼10 ms, and required 100 mW of power. Another thermally actuated switch used thermal expansion of air to break and move LM droplets [7]. Use of air as a working fluid improved the RF performance significantly. Better than 1 dB insertion loss and 20 dB isolation were reported up to 18 GHz. In addition, reported were ∼0.92 ms switching latency and 10 μJ energy required per cycle. Recently, an electrostatically actuated LM capacitive switch was reported with a 0.6 dB insertion loss of up to 20 GHz [14]. The isolation, however, degraded to 10 dB at 15 GHz and was attributed to the slotline mode arising due to the asymmetrical switch design. The switch-on time was reported to be ∼1 ms. To avoid the toxicity, nonmercury liquids have been explored, although, so far, with little success. An LM alloy Galinstan was immersed in a Teflon solution for a reflective switch [15], and water was used for a reflective and absorptive switch [16]. However, the LM alloy or water was manually pumped to achieve switching, thus restricting its use as a microswitch. Nevertheless, better than a 1.3 dB insertion loss and a 20 dB isolation were demonstrated up to 100 GHz. Pumping of fluid also limited the expected switching time to 10 ms [16]. In this paper, we design, fabricate, and evaluate an LM-based low-loss RF switch for dc-40 GHz operation range. First, the feasibility of LM to function in a microswitch at RF frequencies 1057-7157/$26.00 © 2009 IEEE SEN AND KIM: LIQUID–SOLID DIRECT CONTACT LOW-LOSS RF MICROSWITCH Fig. 1. Schematic of the experiment to study LM (mercury) behavior at high frequency. Manually placed LM droplet acts as a capacitive RF short. Contact angle and droplet curvature lead to a liquid–solid contact area smaller than the SU-8 frame dimension. The switch width is 600 μm. is assessed. Then, the LM actuation mechanism adopted is described. After illustrating the problems associated with a single-droplet switch design, a two-droplet design is proposed. After discussing the device fabrication, we report experimental and simulation results. 991 Fig. 2. Measured insertion loss (without LM) and isolation (with LM) from test devices shown in Fig. 1. Device capacitance is calculated to be 25 pF from curve fitting. Series resistance is 0.225 Ω. A good low-frequency fit is obtained even when inductance is ignored. the liquid–solid contact line in Fig. 1 than the inner area of the microframe. III. EWOD-BASED D ROPLET ACTUATION FOR S WITCHING II. T EST OF LM D ROPLET FOR H IGH -F REQUENCY RF For an LM droplet to perform as the RF shunt-switching element, it should be able to provide sufficient isolation when shorting the signal plane to the ground plane. Despite the new possibilities (e.g., Galinstan), mercury is the only viable option for an actuated submillimeter LM droplet at room temperatures. Even though mercury is not among the best conductors, with its resistivity ∼40 times that of gold, its conductivity is not considered a main issue. For example, a 100 μm wide droplet should provide ∼40 dB isolation at 40 GHz. Since most of the previous LM-based RF switch work has been limited to 20 GHz, we conducted an experimental study to reveal any unexpected behavior and assess the suitability of mercury as an RF switching element up to 40 GHz. As schematically shown in Fig. 1, the test device consisted of a 60–460–60 μm coplanar waveguide (CPW), fabricated using 8000 Å gold liftoff. A thin dielectric layer (1500 Å silicon nitride) over the patterned electrodes protected the gold CPW from mercury. A 100 μm tall SU-8 microframe defined the switch width (600 μm) and held the LM droplet in position, while the experiment was conducted. An HP 8510C network analyzer was used to measure the scattering parameters. The device insertion loss was measured without the LM droplet. An LM droplet was placed manually, and the device isolation was measured, as shown in Fig. 2. Isolation of better than 25 dB over the range of 5–40 GHz proves the suitability of LM at high frequencies. Since the isolation profile solely depends on switch capacitance at low frequencies [17], a good low-frequency fit was obtained even when switch inductance was ignored. The switch capacitance and resistance extracted from the curve fitting were 25 pF and 0.225 Ω, respectively, while the switch capacitance calculated using the microframe dimensions was 75 pF. The fitted capacitance was smaller due to the convex curvature of the nonwetting droplet, leading to a smaller actual contact area as outlined by Microswitches based on LM actuation have suffered mostly due to their slow actuation speeds [13]. As the first bottleneck, switching speed has been limited by the accuracy with which the droplets have been deposited and positioned. We have previously reported that the EWOD actuation of the contact line of an LM droplet confined by a microframe, as shown in Fig. 3, allowed a high-speed operation and demonstrated the mechanism for dc applications [12]. While a microframe structure, in this case made of SU-8, held the droplet in position with a lithographic accuracy, high surface tension of the LM droplet ensured that the interface position was defined accurately within the microframe. Furthermore, any variation in the deposited LM volume was absorbed by the meniscus at the large back opening rather than by the meniscus at the small front opening (see Fig. 3). Any variation in the LM droplet volume due to variation in ambient temperature will also be absorbed by the back opening keeping the front meniscus practically unaffected. By keeping the droplet meniscus at the important front opening unaffected by the volume variation of the droplet, the droplet contact line was positioned always accurately from the signal electrode. This means high droplet placement accuracy is not required and manual placement is sufficient. These features allowed a switch design with very small switching gaps (e.g., 10 μm for 600 μm diameter droplet), resulting in a fast switching. Device failure by escape of LM droplet from the microframe due to vibrational and horizontal shock has been evaluated in [12]. For similar device design, ∼16 G stability was reported. Vibrational stability will be improved once device fabrication is advanced to further miniaturize the devices. The grounding electrode grounded the LM droplet. When a potential (100 V) was applied to the actuation electrode, the liquid–solid contact area spread and made contact with the signal electrode. The area of the signal electrode which the LM 992 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 18, NO. 5, OCTOBER 2009 Fig. 3. Schematic of an EWOD-actuated confined LM droplet switch. (Left and center) The switch is in OFF-state showing the liquid–solid contact line accurately positioned (∼10 μm) away from the signal electrode. (Right) Application of an electrical potential to the actuation electrode causes the liquid–solid interface to spread by EWOD, and the LM makes contact with the signal electrode (switch-on). Exemplary switch dimensions: back opening 400 μm, front opening 250 μm, droplet height (i.e., gap between substrate and top plate) 400 μm. Fig. 4. Schematic of the two-droplet LM RF switch. Switch design consists of two mirror-imaged microframe structures. Note that the LM droplets sit on the ground planes instead of the signal electrode, eliminating signal leak path. covers and makes contact is hereafter known as the contact region. Using the initial regime of a fast (0.5 m/s) contact-line motion of EWOD, 60 μs of switch-on time was demonstrated. Signal rise and fall time was better than 5 μs with no contact bounce, and the switch used 10 nJ of energy per cycle. It was also demonstrated that dielectric charging had little ill effect on the actuation mechanism for 105 switching cycles. We will use the same actuation mechanism to design and demonstrate a low-loss RF microswitch. IV. T WO -D ROPLET RF S WITCH D ESIGN A natural extension of the EWOD-based dc switch design demonstrated in [12] for an RF switch would be to place a single droplet on the signal line of a CPW. When actuated, the droplet would spread and make contact with the ground plane, shunting the RF signal. However, the required overlap between the actuation electrode and the droplet contact area for EWOD mechanism to work, as shown in Fig. 3, will create a signal leak path at high frequencies as described in detail in [18]. Using high-resistivity SiCr with surface resistivity of 1200 Ω/ has been demonstrated as a solution for this kind of problems [19]. In order to simplify fabrication, having more freedom in selecting the cap material and achieve lower insertion loss, a twodroplet design was developed, as schematically shown in Fig. 4. Based on 20–180–20 μm CPW on fused silica substrates, the design uses two mirror-imaged microframes, each enclosing an LM droplet placed on the ground planes of the CPW. The droplets are grounded through a window etched in the dielectric layer covering the CPW, as shown in Fig. 4. Placing the droplets on the ground plane, as opposed to the case of one droplet on the signal line, solves the problem of a signal leak through the bias lines, because there is no LM above the signal electrode when the droplet is not actuated. In the mean time, the actuation electrode, which is capacitively coupled to the LM droplet, is also grounded at RF frequencies. When the droplet is actuated, the interface spreads over to the signal electrode to make contact, and the signal is shunt to the ground plane via the LM across the two direct contacts, i.e., signal-to-LM and LMto-ground. Grounding the LM droplets also provides a greater freedom in selecting the package cap material and allows use of grounded cap. SEN AND KIM: LIQUID–SOLID DIRECT CONTACT LOW-LOSS RF MICROSWITCH Insertion loss of a switch is due to a variation of the switch impedance from the characteristic impedance. The presence of actuation electrodes causes a change in the gap between the signal line and the ground planes. At first sight, it may seem that this discontinuity will lead to a large insertion loss. However, the actuation electrode, which is capacitively coupled to the ground plane through the LM droplet, acts as an extension of the ground plane at RF frequencies. To reduce losses, the signal line of the CPW is tapered in accord with the actuation electrode to maintain 50 Ω impedance. This tapering minimizes the impedance-mismatched section, as shown in Fig. 4, to the pointed extension of signal line, where the LM droplet makes contact with the signal line when actuated. Thus, the insertion loss in this design is from the capacitance due to the impedance-mismatched contact region and the LM droplet. To avoid formation of dielectric bridges and hence minimize their effect on insertion loss, microframe structures were positioned away from the gaps of the CPW. The design of the microframe and its positioning with respect to the contact region was described in [12]. In the current device (Fig. 4), the microframe is designed to be 400 μm high with the opening at the back and front of 400 and 250 μm, respectively. The actuation electrode length “w” is 250 μm. At the ends of the actuation electrode (location B, as shown in Fig. 4), the gap between the actuation electrode and the signal line is 20 μm, forming a 20–180–20 μm CPW. Prior to the contact region (location C, as shown in Fig. 4), the gap between the actuation electrode and the signal line is 17 μm, and the signal linewidth is tapered to 146 μm to obtain a 50 Ω 17–146–17 μm CPW. The contact region is 100 μm long. However, the gap between the actuation electrode and the signal line is 5 μm. Using Maxwell SV from Ansoft Corporation, the capacitance due to each contact region is approximately calculated as 6.2 fF. The 3-D shape of the droplet interface makes the calculation of the capacitance between the signal line and the droplet difficult. The minimum gap between the signal line and the LM interface is 15 μm. Although the surface area of the 3-D LM droplet is larger than the contact region, most of the LM surface is significantly farther away from the signal line. Thus, a smaller contribution to the overall switch capacitance is expected from the LM droplets. V. D EVICE FABRICATION Device fabrication, as shown in Fig. 5, starts with a 700 μmthick fused silica substrate. One-thousand-ångström chromium is evaporated on the substrate and patterned lithographically using a wet etchant. Eight-thousand-ångström silicon oxide, which isolates the bias lines (same as actuation electrode) from the CPW, is deposited using plasma-enhanced chemical vapor deposition (PECVD) and etched using reactive ion etching (RIE). CPW is formed by liftoff of 8000 Å-thick gold using 200 Å Cr as adhesion layer. LOR-20B from MicroChem is used with AZ5214 from Clariant to obtain a clean liftoff of such a thick metal. For the dielectric separating the actuation electrode from the LM, 3500 Å nitride is deposited using PECVD and etched by RIE. Since most LMs including mercury attack most metals, a protective layer is required at the contact regions. 993 Fig. 5. Process flow for device fabrication. Section AA from Fig. 4 is depicted. Several metals (e.g., chromium, iron, platinum, nickel, and tungsten) are known to be compatible with mercury. A layer of 2000 Å Cr/Ni is deposited at the contact regions using liftoff. To reduce hysteresis (i.e., static friction which restricts contactline motion) and have a reasonable actuation voltage, a thin hydrophobic coating of Teflon is used. Teflon is spin coated to obtain a 2000 Å film and baked at 320 ◦ C for 3 h. Further processing of Teflon-coated wafer is difficult due to the low surface energy of Teflon, which results in poor adhesion of any film coated on it. To successfully coat photoresist (PR), we add surfactant to the PR. The Teflon layer is patterned lithographically and etched in oxygen plasma. After the PR is removed in acetone, the patterned Teflon layer is baked again at 320 ◦ C for 3 h. To allow the building of the microframe in subsequent steps, the Teflon-covered area is minimized. To 994 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 18, NO. 5, OCTOBER 2009 Fig. 6. (Top) Optical micrograph showing hydrophobic layer was processes (i.e., coated and patterned) before microframe fabrication. CPW is 20–180–20 μm, (green) actuation electrode is 260 μm long, and signal electrode is 100 μm long and 50 μm wide. (Bottom) SEM of a fabricated device before LM droplet placement. Fig. 7. Measured profile of a current generation device with very small difference of height between the actuation electrode and the contact region. obtain 400 μm-thick microframe structures, SU-8 2150 from MicroChem is deposited by a single spin and soft baked at 95 ◦ C for 3 h. Temperature is always ramped up or down with a rate of 60 ◦ C/hr from 50 ◦ C. A 300 s step exposure with each step consisting of a 30 s exposure and a 20 s delay is used to help reduce surface hardening due to heat. After a 45 min postexposure bake, the features are developed with agitation. The Teflon layer is again baked in a nitrogen environment, but this time at 200 ◦ C for 3 h. A lower temperature and nitrogen environment are used to prevent SU-8 burning. Finally, a ∼ 400 μm-diameter mercury droplet is placed in each microframe manually. SEM micrographs of a fabricated device before LM placement are shown in Fig. 6. This process deviates from the usual processing practice of coating and patterning a Teflon layer as the last step. Such a conventional practice protects the Teflon from any further chemical processing, which may degrade its quality. Following this convention, however, is not possible for our case due to the presence of the tall SU-8 microstructures. If Teflon is coated over tall microstructures, the surfactant-mixed PR film is destabilized during spin coating, leading to the dewetting of the PR on the Teflon. Furthermore, the capillary force causes PR to accumulate in the small spaces between the tall SU-8 microstructures. Teflon is baked after every step to recover the film quality from any degradation during processing. Fig. 6 shows the final two steps, where the Teflon is processed first and then the SU-8 microstructures are fabricated. Another aspect of designing the process flow is to minimize the surface topography; the moving contact line needs to slide across. When an electric potential is applied to the actuation electrode, the contact line spreads over it (see Fig. 3). The section of contact line over the signal electrode, however, does not see the electric field from the actuation electrode and hence feels no actuation force. Instead, the contact line over SEN AND KIM: LIQUID–SOLID DIRECT CONTACT LOW-LOSS RF MICROSWITCH 995 Fig. 8. Optical micrograph of the experimental setup. (Left) Bias tees connection to the GSG probe tips and the dc actuation probe tip. (Right) Device under test with two LM (mercury) droplets. the actuation electrode pulls this forceless section toward the signal electrode. With the local loss of actuation force over the signal electrode, the contact line is prone to pinning by any small surface bump as explained in detail in [18]. This problem was identified when the previous generation devices gave a capacitive shunt response instead of the expected dc shunt. The current process, shown in Fig. 5, solves the problem of contactline pinning by lowering the contact region to the same level as the actuation region, as shown in Fig. 7. VI. S IMULATION AND T EST R ESULTS Device simulations were carried out using high frequency structure simulator (HFSS) from Ansoft Corporation. Resistivity of mercury was defined as 961 nΩ · m. SU-8 was defined as a dielectric with dielectric constant of 3.25. The shape of the LM droplet remains spherical and is easily defined when not actuated. Determining the interface shape when actuated is, however, not trivial. To solve this issue, we assume a simplified polygonal shape approximately matching the actuation electrode shape for the droplet interface. Although an approximate interface shape is used, important parameters (e.g., the overlap area at the contact region and width of the bridge between the signal line and the ground plane) will vary insignificantly from the true situation. An experimental setup to test the RF switching performance is shown in Fig. 8. HP 8510C or Agilent E8361A network analyzer is used to measure the device performance. A dc signal from a National Instrument multifunctional data acquisition amplified using a Trek amplifier is used to actuate the switch. Ground-signal-ground (GSG) probe tips from Picoprobe are used to contact the CPW. To calibrate the setup on a wafer, thru-reflect-line calibration was performed. The measured insertion loss is better than 0.3 dB up to 40 GHz, as shown in Fig. 9 along with the simulation results, showing a good match. The return loss is given by |S11 | = ωCu Z0 2 (1) where Cu is the switch capacitance. The switch capacitance calculated by curve fitting the return loss is 14 fF, while switch capacitance calculated from HFSS simulation is 18 fF. This difference is due to uncertainty in the contact-line position. Contact-angle hysteresis leads to an uncertainty in the static contact angle, resulting in a variation of the contact line position. A 5◦ variation in the contact angle will lead to a variation of ∼ 6 μm in the contact-line position. This variation of the LM Fig. 9. Device insertion loss and return loss. Fig. 10. Device isolation. interface position will lead to a variation in the contribution of the LM to the switch capacitance. With approximately 12.4 fF contribution from the contact regions (calculated above), the LM contribution is approximately 1.6 fF measured or 5.6 fF simulated. The switch is actuated using 100 V, and isolation is measured, as shown in Fig. 10. The switch isolation is given by √ R2 + ω 2 L2 (2) |S21 | = (R + 0.5Z0 )2 + ω 2 L2 where R and L are the switch resistance and inductance, respectively. The isolation measured is better than 20 dB up to 996 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 18, NO. 5, OCTOBER 2009 40 GHz. The 5.2 pH switch inductance is obtained from the curve fitting the simulation data. It is not possible to obtain a good match for the resistance from the simulation, as the software does not account for the contact resistance, which is the major source of the switch resistance. The switch resistance is obtained as 1.32 Ω by fitting to the measured isolation. The extracted resistance of two contacts in parallel is in good accord with the previously reported 2.35 Ω for a single 50 μm × 50 μm contact [12]. VII. S UMMARY AND C ONCLUSION In this paper, a low-loss RF MEMS switch based on liquid–solid direct contact using two LM droplets has been presented. With ∼ 600 μm in size (not including the CPW extension required for testing), this switch is comparable to other beam-based RF MEMS switches, which are usually hundreds of micrometers in length [2]. The basic design was based on the previously reported fast EWOD actuation of an LM droplet in the microframe, which led to bounceless operation with 60 μs switch-on time, less than 5 μs signal rise/fall time, and nanojoules of energy consumption per actuation cycle. After identifying a problem of signal leakage at RF frequencies, a two-droplet design has been developed to solve the biasing problem that otherwise would degrade the RF performance and require deposition of a high-resistivity SiCr layer. Switch design was optimized with the aim of improving RF performance. Device fabrication required special care to prevent contact-line pinning, which led to poor liquid–solid contact. The insertion loss was measured to be better than 0.3 dB and isolation better than 20 dB both up to 40 GHz. The fitted switch characteristics included a 14 fF up-state capacitance and a 5.2 pH down-state inductance. The switch contact resistance extracted from curve fitting was 1.32 Ω, similar to the value reported previously for a dc switch. With a liquid–solid contact, the reported switching technology is expected to lead to high-reliability RF MEMS switches including contact switches. It is further expected that the use of LM droplets will allow the development of high-power hot switching devices. There are, however, several key issues that need to be solved before the technology matures to deploy the devices. With contact reliability not a primary issue for this switch, the major expected failure mechanism is through dielectric charging. Even though dielectric charging has been demonstrated to have little effect on the actuation mechanism up to 105 cycles [12], reliability demonstrations up to 109 cycles, or more would require hermetic packaging. Hermetic packaging of these devices in an inert environment is a significant challenge due to the low boiling point of mercury. Gallium and its alloys (e.g., Galinstan) would allow packaging at higher temperatures, although instead they are more susceptible to surface oxidation. Detailed results related to hermetic packaging of liquid metal droplets and long-term device reliability will be presented elsewhere after completion of the ongoing study. Significant switch-to-switch contact resistance variation was observed, which is attributed to oxidation of the LM in air. It is also expected that cycle-to-cycle contact resistance degradation will happen in air due to continuous oxidation of LM. This problem will be solved once hermetic packaging in an inert environment is developed for these switches. Even with switch contact resistance of 1.32 Ω which is comparably higher than the best MEMS switches [20], the LM-based switches are expected to fare better due to no contact degradation from arcing or welding. Simplified 2-D FEM simulation shows for designs based on silicon substrates with adequate heat sinks our devices should be able to handle ∼1 A before mercury boiling at 357 ◦ C. These tests were not performed considering the toxic nature of mercury vapor. Galinstan which boils at 1300 ◦ C should be able to handle much larger currents. ACKNOWLEDGMENT The authors would like to thank J. Jenkins and T. Wu for their discussions about the project, Dr. S. Mathai and the staff of the Center for High Frequency Electronics, University of California, Los Angeles, for their help with the RF measurements, and A. Lee for her help with this paper. R EFERENCES [1] D. Hyman and M. Mehregany, “Contact physics of gold microcontacts for MEMS switches,” IEEE Trans. Compon. Packag. Technol., vol. 22, no. 3, pp. 357–364, Sep. 1993. [2] G. M. Rebeiz, RF MEMS: Theory, Design, and Technology., 1st ed. Hoboken, NJ: Wiley-Interscience, 2003. [3] S. Saffer, J. Simon, and C.-J. Kim, “Mercury-contact switching with gap-closing microcantilever,” in Proc. Micromach. Devices Compon. II, Austin, TX, Oct. 1996, pp. 204–209. [4] J. Simon, S. Saffer, F. Sherman, and C.-J. Kim, “Lateral polysilicon microrelays with a mercury microdrop contact,” IEEE Trans. Ind. Electron., vol. 45, no. 6, pp. 854–860, Dec. 1998. [5] A. Cao, P. Yuen, and L. Lin, “Microrelays with bidirectional electrothermal electromagnetic actuators and liquid metal wetted contacts,” J. Microelectromech. Syst., vol. 16, no. 3, pp. 700–708, Jun. 2007. [6] J. Simon, S. Saffer, and C.-J. Kim, “A liquid-filled microrelay with a moving mercury microdrop,” J. Microelectromech. Syst., vol. 6, no. 3, pp. 208–216, Sep. 1997. [7] Y. Kondoh, T. Takenaka, T. Hidaka, G. Tejima, Y. Kaneko, and M. Saitoh, “High-reliability, high-performance RF micromachined switch using liquid metal,” J. Microelectromech. Syst., vol. 14, no. 2, pp. 214– 220, Apr. 2005. [8] J. Kim, W. Shen, L. Latorre, and C.-J. Kim, “A micromechanical switch with electrostatically driven liquid-metal droplet,” Sens. Actuators A, Phys., vol. 97/98, pp. 672–679, Apr. 2002. [9] W. Shen, R. T. Edwards, and C.-J. Kim, “Electrostatically actuated metaldroplet microswitches integrated on CMOS chip,” J. Microelectromech. Syst., vol. 15, no. 4, pp. 879–889, Aug. 2006. [10] P. Sen and C.-J. Kim, “Electrostatic fringe-field actuation for liquid-metal droplets,” in Proc. 13th Int. Conf. Solid-State Sens., Actuators Microsyst., Seoul, Korea, Jun. 2005, pp. 705–708. [11] Z. Wan, H. Zeng, and A. Feinerman, “Area-tunable micromirror based on electrowetting actuation of liquid-metal droplets,” Appl. Phys. Lett., vol. 89, no. 20, p. 201 107, Nov. 2006. [12] P. Sen and C.-J. Kim, “A fast liquid-metal droplet microswitch using EWOD-driven contact-line sliding,” J. Microelectromech. Syst., vol. 18, no. 1, pp. 174–185, Feb. 2009. [13] P. Sen and C.-J. Kim, “Microscale liquid-metal switches—A review,” IEEE Trans. Ind. Electron., vol. 56, no. 4, pp. 1314–1330, 2009. [14] C.-H. Chen and D. Peroulis, “Electrostatic liquid-metal capacitive shunt MEMS switch,” in IEEE MTT-S Int. Microw. Symp. Tech. Dig., Jun. 2006, pp. 263–266. [15] C.-H. Chen, J. Whalen, and D. Peroulis, “Non-toxic liquid-metal 2–100 GHz MEMS switch,” in IEEE MTT-S Int. Microw. Symp. Tech. Dig., Jun. 2007, pp. 363–366. [16] C.-H. Chen and D. Peroulis, “Liquid RF MEMS wideband reflective and absorptive switches,” IEEE Trans. Microw. Theory Tech., vol. 55, no. 2, pp. 2919–2929, Dec. 2007. SEN AND KIM: LIQUID–SOLID DIRECT CONTACT LOW-LOSS RF MICROSWITCH [17] J. B. Muldavin and G. M. Rebeiz, “High-isolation CPW MEMS shunt switches—Part 1: Modeling,” IEEE Trans. Microw. Theory Tech., vol. 48, no. 6, pp. 1045–1052, Jun. 2000. [18] P. Sen, “Driving liquid-metal droplets for RF microswitching,” Ph.D. dissertation, Univ. California, Los Angeles, CA, 2007. [19] G.-L. Tan and G. M. Rebeiz, “A DC-contact MEMS shunt switch,” IEEE Microw. Wireless Compon. Lett., vol. 12, no. 6, pp. 212–214, Jun. 2002. [20] L. L. W. Chow, J. L. Volakis, K. Saitou, and K. Kurabayashi, “Lifetime extension of RF MEMS direct contact switches in hot switching operations by ball grid array dimple design,” IEEE Electron Device Lett., vol. 28, no. 6, pp. 478–481, Jun. 2007. Prosenjit Sen was born in Calcutta, India, in 1978. He received the B.Tech. degree in manufacturing science and engineering from Indian Institute of Technology, Kharagpur, India, in 2000 and the Ph.D. degree in mechanical engineering from the University of California, Los Angeles (UCLA), in 2007. He is currently with the Micro and Nano Manufacturing Laboratory, Department of Mechanical and Aerospace Engineering, UCLA. His research interest includes microfludic systems, droplet dynamics, liquid-metal-based RF MEMS, and reliability of electrowetting-on-dielectric devices. Dr. Sen was the recipient of the Institute Silver Medal from the Indian Institute of Technology. 997 Chang-Jin “CJ” Kim received the B.S. degree from Seoul National University, Seoul, Korea, the M.S. degree from Iowa State University, Ames, with Graduate Research Excellence Award, and the Ph.D. degree in mechanical engineering from the University of California, Berkeley, in 1991. In 1993, he joined the faculty of the University of California, Los Angeles (UCLA), where he has developed several MEMS courses and established a MEMS Ph.D. major field in the Department of Mechanical and Aerospace Engineering. Directing the Micro and Nano Manufacturing Laboratory, UCLA, he is also active in the commercial sector as board member, scientific advisor, and consultant. He currently serves as a Subject Editor for the IEEE/ASME J OURNAL OF M ICROELECTROMECHANICAL S YSTEMS and on the Editorial Advisory Board of the IEEJ Transactions on Electrical and Electronic Engineering. His research interests include MEMS and nanotechnology, including the design and fabrication of micro/nanostructures, actuators, and systems, with a focus on the use of surface tension. Dr. Kim has served on numerous technical program committees, including Transducers and the IEEE MEMS Conference, and on the U.S. Army Science Board as a Consultant. He is currently chairing the Devices and Systems Committee of the ASME Nanotechnology Institute and serving on the National Academies Panel on Benchmarking the Research Competitiveness of the U.S. in Mechanical Engineering. He was the recipient of a TRW Outstanding Young Teacher Award, National Science Foundation CAREER Award, Association for Laboratory Automation Achievement Award, Samueli Outstanding Teaching Award.