3262
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 57, NO. 12, DECEMBER 2009
Robustness of RF MEMS Capacitive Switches With Molybdenum Membranes Cristiano Palego, Member, IEEE, Jie Deng, Student Member, IEEE, Zhen Peng, Student Member, IEEE, Subrata Halder, Member, IEEE, James C. M. Hwang, Fellow, IEEE, David I. Forehand, Member, IEEE, Derek Scarbrough, Member, IEEE, Charles L. Goldsmith, Senior Member, IEEE, Ian Johnston, Suresh K. Sampath, and Arindom Datta
Abstract—This paper compares the characteristics of an RF microelectromechanical systems (MEMS) capacitive switch with a molybdenum membrane versus that of a switch with similar construction but with an aluminum membrane. In comparison, the molybdenum switch exhibits a significantly reduced sensitivity to ambient temperature change so that its pull-in voltage varies by less than 0.035 V C. In addition, large-signal RF performance of the switches was compared under both continuous wave and pulse conditions. The results show that under large RF signals, the self-biasing effect is exacerbated by the self-heating effect and the self-heating effect is in turn amplified by nonuniform current and temperature distributions on the membrane. Measurements of both molybdenum and aluminum switches demonstrate a hot-switched power-handling capacity of approximately 600 mW. Since aluminum has been used as a membrane material for over a decade while molybdenum is new, the above results indicate that molybdenum is a promising membrane material for RF MEMS capacitive switches. Index Terms—Heating, microelectromechanical devices, microwave switches, molybdenum, temperature, thermal factors.
I. INTRODUCTION
A
S RF microelectromechanical systems (MEMS) switch technology continues to mature, much of the development has focused on packaging, reliability, and robustness. To date, many innovative ideas have been presented for MEMS packaging, and the reliability of the devices has seen much improvement [1]. Now, increasingly more attention is being focused on switch robustness.
Manuscript received April 15, 2009; revised July 24, 2009. First published November 10, 2009; current version published December 09, 2009. This work was supported in part by the U.S. Air Force Research Laboratory under Contract F33615-03-C-7003, which was funded by the U.S. Defense Advanced Research Projects Agency under the Harsh Environment, Robust Micromechanical Technology (HERMIT) Program, and in part by the Commonwealth of Pennsylvania, Department of Community and Economy Development, through the Pennsylvania Infrastructure Technology Alliance (PITA). C. Palego, J. Deng, Z. Peng, S. Halder, and J. C. M. Hwang are with the Department of Electrical and Computer Engineering, Lehigh University, Bethlehem, PA 18015 USA (e-mail:
[email protected];
[email protected];
[email protected];
[email protected];
[email protected]). D. I. Forehand, D. Scarbrough, and C. L. Goldsmith are with the MEMtronics Corporation, Plano, TX 75075 USA (e-mail:
[email protected];
[email protected];
[email protected]). I. Johnston, S. K. Sampath, and A. Datta are with the Innovative Micro Technology Company, Santa Barbara, CA 93117 USA (e-mail: ijohnston@imtmems. com;
[email protected];
[email protected]). 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/TMTT.2009.2033885
RF MEMS capacitive switches commonly include a thin flexible metal membrane, which is suspended across two support posts and actuated electrostatically via a stationary electrode beneath the membrane. One of the detractors to this arrangement is that changes in temperature modify the residual stress of the membrane, which can in turn change the pull-in voltage of the membrane by as much as 0.5 V C [2]. The root cause of the change in the pull-in voltage is the difference in thermal expansion coefficient between the membrane and the substrate [3]. Changes in temperature cause the two materials to expand at different rates, which induce a change in the residual stress of the membrane. This change in stress impacts the spring constant of the membrane, which in turn influences its pull-in voltage. Most MEMS capacitive switches are fabricated out of metal for low signal loss, but metals generally have thermal expansion much larger than that of commonly used substrate materials. Thus, the switches exhibit significant change in the pull-in voltage over temperature. One innovative method to mitigate this problem is to modify the geometry of the switch to compensate for the change in stress over temperature [4]. In this technique, the geometry of the membrane supports is designed to eliminate thermal stress changes at the anchor points of the membrane. The resulting structure shows stable operation over temperature, with the pull-in voltage changing less than 5% over 100 C temperature excursions. This technique works well for thicker membranes where the impact of stress gradients can be minimized. Further development has extended the concept of using switch geometry to mitigate stress change over temperature [5]. A more fundamental approach towards minimizing stress change is to pick a membrane material that has a thermal expansion coefficient much closer to that of the substrate material. In MEMS capacitive switches, the range of choices is not particularly broad, as other considerations such as electrical resistivity of the membrane directly impact the RF performance of the switch such as insertion loss or isolation. In addition, under large RF signals, self-biasing and self-heating effects become significant, which depend on not only the electrical resistivity but also the thermal conductivity of the membrane. Aluminum alloys have been used as a membrane material for MEMS capacitive switches for over a decade. While the electrical and thermal conductivities of aluminum are quite good, its expansion coefficient with temperature is very high. Compared to the common substrate material of silicon or glass, aluminum expands far more significantly over temperature. Refractory metals provide much less change over temperature.
0018-9480/$26.00 © 2009 IEEE Authorized licensed use limited to: LEHIGH UNIVERSITY. Downloaded on December 14, 2009 at 14:10 from IEEE Xplore. Restrictions apply.
PALEGO et al.: ROBUSTNESS OF RF MEMS CAPACITIVE SWITCHES WITH MOLYBDENUM MEMBRANES
However, this reduced thermal expansion comes at the expense of increased electrical and thermal resistances. For many applications, this increase is justified by more robust operation over ambient temperature changes. However, this increase may limit the isolation or power-handling capacity of capacitive shunt switches. In order to investigate the potential improvements in switch robustness using refractory metals, molybdenum was chosen to replace aluminum as the membrane material in MEMS capacitive shunt switches. Some data already exist demonstrating that molybdenum is suitable for MEMS applications [7], and molybdenum provides a reasonable balance between thermal expansion coefficient and electrical and thermal resistivities. If molybdenum shows promise, then other refractory metals such as tungsten may also be investigated. The initial results obtained on MEMS capacitive shunt switches fabricated with molybdenum membranes were very encouraging, with their pull-in voltages varying by less than 0.035 V C up to 150 C and their lifetimes exceeding 20 billion cycles [8]. This paper expands on [8] by assessing the impact of different membrane materials on RF power handling of the switches, as evidenced by the reduction in pull-in voltage during hot switching [9]. The experimental assessment is complemented by a simple theoretical analysis in the following. II. THEORY
3263
Fig. 1. (a) Micrograph of an MEMS capacitive switch with a bowtie-shaped molybdenum membrane straddling an 80-m-wide center conductor of a coplanar transmission line. The equivalent circuit of the switch is overlaid on top of the micrograph to illustrate its correspondence to the MEMS switch configuration relative to the transmission line. (b) Measurement setup for testing the large-signal performance of the switch at 15 GHz.
A. Ambient Temperature Effects The membrane’s residual stress temperature according to
changes with the ambient
(1) where is the Young’s modulus of the membrane, is the difference in thermal expansion coefficient between the memis the change in temperature. brane and the substrate, and This change in stress impacts the spring constant of the membrane according to [10] (2) where is a geometric factor to account for the nonrectangular shape of the membrane, is the Poisson’s ratio of the is the residual stress at the reference membrane material, temperature, and , , and are the thickness, width and length of the membrane. The change in the spring constant in turn in(subscript “0” denotes small RF fluences the pull-in voltage signal) (3) where is the air gap between the membrane and the stationary is the vacuum electrode without any applied bias voltage, is the width of the stationary electrode. The permittivity, and defines the contact area between the membrane product and the stationary electrode when the membrane is pulled in.
B. Small-Signal Effects 0 dBm RF signals, both self heating and self Under small biasing can be ignored. Further, at frequencies much lower than resonant frequency of a MEMS capacitive switch, its the electrical characteristics can be represented by a series circuit [11] as shown in Fig. 1(a) (4) where accounts for current crowding in the membrane [12] as well as the difference between thin-film and bulk properties and is the electrical resistivity of the bulk material (5) where and are the dielectric constant and thickness of the is a geoinsulator covering the stationary electrode, metric factor to account for the incomplete contact due to suris the fringing capacitance that is usually a face roughness, , and is the applied bias voltage. Typically, fraction of , 0.05 pF and 1 pF. For well designed switches, the RF power dissipated in the membrane is [11] (6) where is the RF power, is the system impedance, and is the angular frequency. Notice that the dissipated power is directly proportional to and, in turn, .
Authorized licensed use limited to: LEHIGH UNIVERSITY. Downloaded on December 14, 2009 at 14:10 from IEEE Xplore. Restrictions apply.
3264
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 57, NO. 12, DECEMBER 2009
C. Large-Signal Effects Under large RF signals, significant power is dissipated in the membrane even when it is up, which changes the membrane stress through self heating. Assuming heat loss by conduction only [12], the temperature rise due to self heating is
TABLE I BULK PROPERTIES OF CANDIDATE MEMBRANE METALS
(7) where the thermal resistance of the membrane
is (8)
where accounts for nonuniform temperature distribution in the membrane as well as the difference between thin-film and is the thermal conductivity of the bulk bulk properties and material. Since large-signal effects often depend on the product , the fitting factors of electrical and thermal resistances and are often lumped together as well. However, small-signal effects depend on only. By extracting from small-signal effects first, can then be extracted from large-signal effects. Similar to (1), the change in membrane stress due to self heating is (9) where according to (6) and (7) (10) In addition to self heating, self biasing occurs under large RF signals through the square-law voltage dependence of the electrostatic force between the membrane and the stationary electrode, which reduces the applied bias required to pull in the membrane. Ignoring the small loss of the transmission line as , the root-mean-square voltage well as the small reflection by resulting from the combination of the applied bias and the RF voltage is [13] (11) Therefore, the large-signal pull-in voltage is (12) Notice that both terms on the right side of (12) depend linearly on the RF power. The first term accounts for self heating, while the second term accounts for self biasing. The self-heating term depends on the properties of the membrane, whereas the selfbiasing term does not. Also, for moderate RF power, (12) can be approximated by a linear relationship (13) To separate the self-biasing effect from the self-heating effect, pulse RF power can be applied for a period much shorter than
the thermal time constant of the switch. In such an isothermal case (14) For robustness under both small and large RF signals, there are three important requirements of the membrane material: 1) small thermal expansion over temperature; 2) low electrical and thermal resistivities; and 3) mechanical reliability. The membrane must have a low electrical resistivity to minimize RF loss and a low thermal resistivity to minimize self heating. The residual stress must be sufficiently low to enable the switches to operate with low voltages, the lower the better with respect to dielectric charging [14]. Last, the mechanical reliability of the switch should be such that the switch can cycle for long periods without mechanical degradation. Table I summarizes the bulk properties of a number of metals commonly used in microelectronics fabrication, including Young’s modulus, coefficient of thermal expansion, electrical resistivity, thermal conductivity, and two figures of merit and . According to (1), is a figure of merit for robustness against ambient temperature change and molybdenum appears to be a better choice than aluminum. However, according to (10) for robustness against self heating, should be considered and, in this case, molybdenum may be slightly inferior to aluminum. Depending on the deposition conditions, thin-film properties can differ significantly from bulk properties and different metals may require different levels of deposition control in order to achieve consistent thin-film properties and manufacturing yields. The bulk properties of Table I serve only as rough guidelines. The real proof is through experiments as described next. III. EXPERIMENTAL A. Device Fabrication The present RF MEMS capacitive switches are similar in construction to those previously reported [15]. The molybdenum membrane is constructed with 0.28 m, 100 m, and 320 m. The molybdenum was deposited using dc magnetron sputtering. Argon flow, plasma power, plasma pressure, and substrate bias were adjusted to achieve stress in the range
Authorized licensed use limited to: LEHIGH UNIVERSITY. Downloaded on December 14, 2009 at 14:10 from IEEE Xplore. Restrictions apply.
PALEGO et al.: ROBUSTNESS OF RF MEMS CAPACITIVE SWITCHES WITH MOLYBDENUM MEMBRANES
of 50-150 MPa tensile. The maximum temperature during deposition was 90 C. The resulting molybdenum membrane is supported by plated copper posts approximately 2.5 m above a Pyrex 7740 glass substrate. The stationary electrode consists of 0.4 m of gold, which is overcoated with 0.25 m of sputtered silicon dioxide. Fig. 1(a) shows a micrograph of the molybdenum switch. Any new metal deposition process has challenges, especially if one is concerned about both mechanical and electrical properties; molybdenum deposition is no exception. One challenge encountered in using molybdenum for the membrane material is achieving the desired residual stress. Although parameterized characterization of the deposition system has been completed, achieving consistency of the desired stress in molybdenum is difficult. The stress tends to be close to zero or heavily tensile, which resulted in switches with very different pull-in voltages. Though it is generally desirable to have switches with low residual stress, switches with almost zero stress are much more subject to curling and buckling, which can be problematic. Achieving a moderate tensile stress on a repeatable basis will require further development. Notice that the residual stress of gold has also been difficult to control, although it has attractive figures of merits as listed in Table I. During fabrication and testing of multiple wafers using molybdenum it was observed that, if the temperature of the oxygen plasma release is too high, significant oxidation occurs on the surface of the molybdenum. The oxidation does not appear to degrade switch operation in any significant way. However, the oxidation does prevent measuring the electrical resistivity of the deposited films. The resistivity of sputtered metal films is typically two to three times of the bulk values. Measurement of film resistivity can be accomplished with four-point probes or suitable test structures. In this case, long meandering lines of molybdenum were measured and resistivity computed. The resulting values . With a 0.28- m-thick of sheet resistivity were cm, just membrane, this is equivalent to approximately 11 over twice the bulk resistivity. All wafers in this lot had similar cm to 11 cm. results, with the resistivity averaging 10 This sheet resistivity is consistent with the end-to-end resistance of the membrane that was determined by dc current-voltage . Assuming in (4), this measurements to be or greater than 40 dB isolation at corresponds to 35 GHz [8]. B. Device Testing The fabricated switches were evaluated for their small- and large-signal RF performances under both continuous wave (CW) and pulse conditions in room ambient. Small-signal parameters were measured from 2 to 50 GHz by using an Agilent 8510C vector network analyzer. Hot-switched life tests were conducted with a 35 GHz CW signal at 1 mW using a custom setup [15]. The bias waveform consisted of a square wave with a peak voltage of 39 V and a repetition frequency of 50 kHz. The up and down times of the membrane were equal at 5–7 s, with switching transients accounting for the remainder of the 20 s period. Large-signal tests were conducted with a 15 GHz signal up to 1 W by using a Maury MT980M
3265
Fig. 2. Data from [8] showing little temperature dependence of pull-in voltages measured on four representative molybdenum switches.
0.8–18 GHz automated tuner system with both input and output [see Fig. 1(b)]. To control self heating, a tuners set at 50 pulse generator was used to ensure both the RF power and the bias voltages were applied only during the pulse. For each RF power level, the pulse test was repeated with the bias voltage gradually increased in 0.25 V steps until the output power abruptly dropped. The bias voltage at this point was deemed . To avoid degrading the large-signal pull-in voltage the switch by dielectric charging [13], large-signal test were stopped as soon as the membrane is pulled in. Usually, after measurements was within 0.1 V of the pristine value. For these large-signal pulse tests, the duty cycle was varied from 0.1% to 100% with a pulse repetition frequency of 25 Hz. 1 mW is on Notice that in small-signal tests, the RF power is applied intermittently. constantly while the dc bias By contrast, in large-signal tests, both the RF power 1W and the dc bias are applied together intermittently, so the duty cycle refers to the fraction of time the RF power and dc bias are applied and the membrane is always up at all duty cycles. IV. RESULTS AND DISCUSSION A. Performance Over Temperature To evaluate the change in pull-in voltage over temperature, a representative sampling of switches was chosen with pull-in voltages over the 30 to 45 V range. The capacitance-voltage characteristics of these devices were measured from 20 C to 150 C by using a Boonton 4200 capacitance meter at 1 MHz with fF resolution. The resulting change in pull-in voltage is shown in Fig. 2. Over the 130 C temperature change, the pull-in voltage varied between 1.2 and 4.5 V, or equivalently 0.009 V C and 0.035 V C. This change is an order of magnitude smaller than the typical 0.1 V C to 0.3 V C experienced with aluminum membranes [3]. The voltage variation over temperature is also dependent on the 3-D shape of the membrane. Some of the variation in these results may be due to curling of the membrane caused by vertical stress gradient, as evidenced by the dark contour on the left side of the switch micrograph in Fig. 1. To ensure that the switches were stable over temperature, the up-capacitance of the switches without any bias was monitored over the same temperature range. In all cases, varied by less than 5 fF, varied by which, according to (5), implies that the air gap
Authorized licensed use limited to: LEHIGH UNIVERSITY. Downloaded on December 14, 2009 at 14:10 from IEEE Xplore. Restrictions apply.
3266
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 57, NO. 12, DECEMBER 2009
Fig. 4. Data from [8] showing switching waveforms of transmitted power at 35 GHz and bias voltage measured on a molybdenum switch with little difference ) and after (—) 20 billion cycles of life test. The transmitted (a) before ( power of the switch is detected by a diode, whose output voltage decreases with increasing power. The bias voltage of the switch is scaled by a factor of 100 so that it can be plotted on the same scale as the diode output voltage.
000
Fig. 3. (a) S 1 and (b) S 1 parameters measured on a molybdenum switch with and without 40 V bias (––). With bias, the S parameters are comparable to that of another molybdenum switch (– –) fabricated with its membrane permanently down. Without bias, the S parameters are comparable to that of a through line (– –).
less than 0.3 m. This verifies that there was no buckling of the membranes over the investigated temperature range. By contrast, zero residual stress is reached in aluminum switches between 80 C and 90 C, at which point the membrane buckles and increases by more than 5 fF. B. Small-Signal Performance Fig. 3 shows the measured small-signal performance in the 2–50 GHz band of a representative molybdenum switch with 38 V. Without any bias, the performance is comparable 40 V, the performance is to that of a through line. With comparable to that of another molybdenum switch fabricated with the membrane permanently down. Without bias, the insertion loss is 0.3 dB at 35 GHz and is better than 0.4 dB across the according band. Such an insertion loss corresponds to to (4) and (6). The return loss is 23 dB at 35 GHz and better than 18 dB across the band. With the membrane pulled in, the 0.7 pF. isolation is 12 dB at 35 GHz, corresponding to The isolation can be improved by increasing the bias voltage beyond the pull-in voltage, thereby increasing or the actual contact area between the membrane and the stationary electrode. As mentioned in Section IV-A, the molybdenum membrane tends to be curled at the edge and does not conform to the stationary electrode as well as the aluminum membrane does. These differences were observed under the microscope and have yet to be quantified by using an optical interferometer or a laser Doppler vibrometer. Nevertheless, the small-signal performance of the and values molybdenum switch is consistent with the given in Section II-B and is similar to that of switches made with aluminum membranes.
Detailed qualification of mechanical reliability is an extensive undertaking. To demonstrate the initial viability of molybdenum as a membrane material, a switch was actuated for 20 billion cycles without failure. Fig. 4 shows that the switching waveforms are very similar before and after the life test. The on/off contrast of the transmitted power actually improves after the initial burn in. Full characterization of more switches over longer cycle times and ambient conditions is ongoing, but the present results give an optimistic indication that molybdenum is a suitable material for mechanical operation of MEMS capacitive switches. C. Large-Signal Performance By measuring the transient response in terms of either transmitted or reflected power of a switch, its thermal time constant can be extracted. In general, the reflected power is more sensitive to the temperature transient and is more reliable for parameter extraction. Fig. 5 shows the transient response of a molybdenum switch with relatively low residual stress 84 MPa and pull-in voltage 23 V , so that it self actuates under high RF power even without any applied bias. It can be seen 650 mW , the memthat upon applying moderate power brane gradually heats up, which exacerbates self biasing and reduces the transmitted power while increasing the reflected power. From the increase in the reflected power, the thermal time constant was extracted to be 150 s. With an RF power between 650 and 850 mW, the membrane becomes mechanically unstable and self heating is perturbed by the onset of actuation. Beyond 850 mW, self actuation occurs in 400 s and the transmitted power abruptly drops as shown in the top curve of Fig. 5. However, the transmitted power does not drop by more than 100 mW because, at this point, the self heating is highly nonuniform and the membrane is highly deformed so that upon actuation it makes only a partial contact to the stationary electrode. Further, after self actuation, although heat can be more easily dissipated through the partial contact with the stationary electrode, the membrane continues to heat up and the contact area continues to enlarge, so that the transmitted power continues to decrease. Fig. 6 shows the pull-in voltages measured under different CW RF powers on: (a) the molybdenum switch of Fig. 5; (b) an-
Authorized licensed use limited to: LEHIGH UNIVERSITY. Downloaded on December 14, 2009 at 14:10 from IEEE Xplore. Restrictions apply.
PALEGO et al.: ROBUSTNESS OF RF MEMS CAPACITIVE SWITCHES WITH MOLYBDENUM MEMBRANES
3267
Fig. 5. Transient (a) transmitted and (b) reflected powers measured on a molyb), 890 (– – –), and 1000 denum switch under RF powers of 630 (…), 790 ( mW (—) at 15 GHz, which indicate the temperature rise through self heating. The switch has a relatively low small-signal pull-in voltage of 23 V.
000
Fig. 6. Measured (symbol) versus modeled (solid curve) pull-in voltage as a function of CW RF power at 15 GHz for an aluminum switch ( ) and two molybdenum switches (5), which shows how isothermal self biasing (dashed curve) is exacerbated by self heating. Although the switches have different small-signal pull-in voltages due to different residual stresses, their large-signal pull-in voltages have comparable power dependence due to comparable self heating.
other molybdenum switch with 120 MPa and 28 V; 150 MPa and (c) a representative aluminum switch with and 30 V. It can be seen that for all three switches the pull-in voltage decreases linearly with increasing RF power up to 600 mW, which is consistent with the predictions by using and . This shows that the self-heating (13) with effect can be amplified as much as 50 times by nonuniform current and temperature distributions. The slope of power dependence is comparable for all three switches, which suggests that molybdenum and aluminum switches can have comparable self-heating characteristics. The isothermal prediction by using (14) is also included to show that the self-biasing effect is exacerbated by self heating, because the latter is greatly amplified by nonuniform current and temperature distributions. Similar to the transient tests of Fig. 5, beyond 600 mW in the present CW
Fig. 7. Pull-in voltages (0 0 0) measured on: (a) molybdenum switch with V = 23 V; (b) molybdenum switch with V = 28 V; and (c) aluminum switch with V = 30 V under pulse RF power of 15 GHz with pulse repetition frequency of 25 Hz and duty cycles of 1%, 10%, 40%, and 100% (top down). The data measured with 0.1% duty cycle are exactly the same as that with 1% duty cycle. Under 1% duty cycle, the measured values approach that calculated for the isothermal case (. . .). Under 100% duty cycle, the measured values approach that measured for the CW case (—).
tests, the on/off contrast of the switch output diminishes due to membrane deformation so that the pull-in voltage can no longer be reliably determined. Typically, the linear power dependence covers the power range before the pull-in voltage is halved. This power range was then used to test all switches under pulse RF excitation. The same three switches of Fig. 6 were tested under pulse RF power with a duty cycle of 0.1%, 1%, 10%, 40%, and 100%. The pull-in voltage was sampled near the end of each pulse. Fig. 7 shows that in all cases, the pull-in voltage decreases linearly with increasing power, but the slope is increasingly steeper with increasing duty cycle. With 100% duty cycle, the measured pull-in voltage is very close to that measured under CW power as in the case of Fig. 6. On the other hand, with decreasing duty cycle, the pull-in voltage converges to a value somewhat lower than the isothermal prediction of (14). This difference may be attributed to nonuniform voltage distribution unlike the simple assumption that led to (11).
Authorized licensed use limited to: LEHIGH UNIVERSITY. Downloaded on December 14, 2009 at 14:10 from IEEE Xplore. Restrictions apply.
3268
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 57, NO. 12, DECEMBER 2009
As mentioned earlier, the most salient feature of Figs. 6 and 7 is that the pull-in voltage decreases linearly with the RF power. While this feature has been observed [9] in CW power tests, to our knowledge, this is the first time it is observed in pulse power tests. Although the simple theory of Section II correctly predicts such a linear dependence in both CW and pulse power tests, there is insufficient evidence to support that the theory is correct or complete. It has been challenging to model the highly nonuniform voltage, current and temperature distributions in a compact manner and the resulted lumped model contains large fitting parameters such as and in the electrical and thermal resistances and , respectively. As discussed in Section II-A, for strictly fitting purpose, and can be lumped together to arrive at the best agreement with the experimentally observed . However, by analyzing the physical origins of and separately, they can be separately extracted and compared to that predicted by a more detailed analysis such as finite-element analysis. To this end, since and account for not only the nonuniform distribution, but also the difference between thin-film and bulk properties, unless the thin-film properties are better known, a more detailed analysis may not be more accurate. The state of the art may only be advanced through more detailed theoretical analysis and experimental characterization together. Although the pulse power tests provide additional means to experimentally validate the theoretical analysis, a more direct validation would be to measure the temperature distribution on the membrane. However, the temperature nonuniformity is likely beyond the resolution of typical infrared cameras. Further, without detailed analysis, it is not obvious what kind of “average” temperature would correspond to the lumped value predicted by (7). V. CONCLUSION Two figures of merit were proposed for selecting the memfor brane material of RF MEMS capacitive switches: robustness against ambient temperature change and for robustness against self heating. In terms of , molybdenum is superior to aluminum mainly due to the smaller thermal expansion of molybdenum. This was experimentally verified by the much reduced temperature dependence of the pull-in voltage of the molybdenum switch and by being able to operate it up to at least 150 C without any evidence of buckling. Initial life tests up to 20 billion cycles also showed that the molybdenum membrane could be mechanically reliable. , molybdenum is comparable However, in terms of to aluminum mainly due to the higher electrical and thermal resistivities of molybdenum, which compensate for its ad. This was experimentally verified by the vantage in comparable power dependence of the pull-in voltages between molybdenum and aluminum switches. When both ambient temperature dependence and power-handling capacity of the and should be switch are important, both carefully considered. REFERENCES [1] C. L. Goldsmith, J. Maciel, and J. McKillop, “Demonstrating reliability,” IEEE Microw. Mag., vol. 8, pp. 56–60, Dec. 2007.
[2] B. Schawwecker, J. Mehner, K. Strohm, H. Haspeklo, and J.-F. Lu, “Investigations of RF shunt airbridges among different environmental conditions,” Sens. Actuators A, vol. 114, pp. 49–58, May 2004. [3] C. L. Goldsmith and D. I. Forehand, “Temperature variation of actuation voltage in capacitive MEMS switches,” IEEE Microw. Wireless Components Lett., vol. 15, no. 10, pp. 718–720, Oct. 2005. [4] H. Nieminem, V. Ermolov, S. Silanto, K. Nybergh, and T. Ryhanen, “Design of a temperature stable RF MEM capacitor,” J. Microelectromech. Syst., vol. 13, pp. 705–714, Oct. 2004. [5] I. Reines, B. Pillans, and G. M. Rebeiz, “A stress-tolerant temperature-stable RF-MEMS switched capacitor,” in Proc. IEEE Int. Conf. Microelectromech. Syst., Jan. 2009, pp. 880–883. [6] D. S. Gardner and P. A. Flinn, “Mechanical stress as a function of temperature in aluminum films,” IEEE Trans. Electron Devices, vol. 35, no. 12, pp. 2160–2169, Dec. 1988. [7] R. B. Brown, M. L. Ger, and T. Nguyen, “Characterization of molybdenum thin films for micromechanical structures,” in Proc. IEEE Int. Conf. Microelectromechanical Syst., Feb. 1990, pp. 77–81. [8] C. Goldsmith, D. Forehand, D. Scarbrough, I. Johnston, S. Sampath, A. Datta, Z. Peng, C. Palego, and J. C. M. Hwang, “Performance of molybdenum as a mechanical membrane for RF MEMS switches,” in IEEE MTT-S Int. Microw. Symp. Dig., Jun. 2009, pp. 1229–1232. [9] B. Pillans, J. Kleber, C. Goldsmith, and M. Eberly, “RF power handling of capacitive RF MEMS devices,” in IEEE MTT-S Int. Microw. Symp. Dig., Jun. 2002, pp. 329–332. [10] P. M. Osterberg and S. D. Senturia, “M-TEST: A test chip for MEMS material property measurement using electrostatically actuated test structures,” J. Microelectromech. Syst., vol. 6, pp. 107–118, Jun. 1997. [11] G. M. Rebeiz, RF MEMS Theory, Design Technology. Hoboken, NJ: Wiley, 2003. [12] J. B. Rizk, E. Chaiban, and G. M. Rebeiz, “Steady state thermal analysis and high-power reliability considerations of RF MEMS capacitive switches,” in IEEE MTT-S Int. Microw. Symp. Dig., Jun. 2002, pp. 239–242. [13] J. R. Reid, L. A. Starman, and R. T. Webster, “RF actuation of capacitive MEMS switches,” in IEEE MTT-S Int. Microw. Symp. Dig., Jun. 2003, pp. 1919–1922. [14] C. L. Goldsmith, D. I. Forehand, Z. Peng, J. C. M. Hwang, and J. L. Ebel, “High-cycle life testing of RF MEMS switches,” in IEEE MTT-S Int. Microw. Symp. Dig., Jun. 2007, pp. 1805–1808. [15] Z. J. Yao, S. Chen, S. Eschelman, D. Denniston, and C. Goldsmith, “Micromachined low-loss microwave switches,” J. Microelectromech. Syst., vol. 8, pp. 129–134, Jun. 1999. [16] C. L. Goldsmith, D. I. Forehand, X.-B. Yuan, and J. C. M. Hwang, “Tailoring capacitive switch technology for reliable operation,” presented at the Government Microcircuits Appl. Critical Technol. Conf. Dig., Mar. 2006, paper 2.01.
Cristiano Palego (S’03–M’07) received the B.S. and M.S. degrees in electrical engineering from the University of Perugia, Perugia, Italy, in 2003, and the Ph.D. degree from the University of Limoges, Limoges, France, in 2007. Since 2007, he has been a Research Associate with Lehigh University, Bethlehem, PA. His research interests include design, technology, and characterization of RF-MEMS devices for reconfigurable frontends, phased antenna arrays, and high-power applications.
Jie Deng (S’04) received the B.S. degree in materials science from Fudan University, Shanghai, China, in 2003, and the Ph.D. degree in electrical engineering from Lehigh University, Bethlehem, PA, in 2009, respectively. He was a Process Engineer with Philips Mobile Display System in 2003. He was an Intern Reliability Engineer with Velox Semiconductor, Somerset, NJ, in 2006 and an Intern Modeling Engineer with IBM, Fishkill, NY, in 2009. He is currently a Staff Device Engineer with Peregrine Semiconductor, San Diego, CA. His research interests include modeling and characterization of RF transistors based on III-N compound, Si CMOS, and MEMS technologies.
Authorized licensed use limited to: LEHIGH UNIVERSITY. Downloaded on December 14, 2009 at 14:10 from IEEE Xplore. Restrictions apply.
PALEGO et al.: ROBUSTNESS OF RF MEMS CAPACITIVE SWITCHES WITH MOLYBDENUM MEMBRANES
Zhen Peng (S’06) was born in Shanghai, China, in 1980. He received the B.E. degree in electrical engineering from Shanghai Jiao Tong University, Shanghai, China, in 2003. He is currently pursuing the Ph.D. degree in electrical and computer engineering from Lehigh University, Bethlehem, PA. From 2003 to 2005, he was an IC design engineer with Ricoh Electronics (Shanghai) Company focusing on low-voltage dc-dc converter circuit design and reliability test. He is currently with Lehigh University, where he is involved in the research and development of RF-MEMS capacitive switches, focusing on modeling and characterization of charging of different dielectric materials under different electrical stresses.
Subrata Halder (M’96–S’04–M’07) received the B.Tech. degree from the Indian Institute of Technology, Kharagpur, India, in 1988, the M.Eng. degree from Nanyang Technological University, Singapore, in 2001, and the Ph.D. degree from Lehigh University, Bethlehem, PA, in 2007, all in electrical engineering. In 1988, he was with SAMEER, Mumbai, India. From 2001 to 2003, he was with DenseLight, Singapore. He was a Device Engineer with Anadigics Inc., Warren, NJ. Currently, he is Senior Research Scientist with Lehigh University, Bethlehem, PA. He has contributed over 30 technical papers and he holds one U.S. patent. His research interests include MEMS, HBTs, HEMTs, and other microwave devices and circuits.
3269
Derek Scarbrough (S’04–M’06) received the B.S. and M.S. degrees in electrical engineering from the University of Texas at Dallas, in 2004 and 2006, respectively. Since 2007, he has been an employee of MEMtronics Corporation, Dallas, TX. His recent work has included the development of MEMS-based phase shifters and tunable filters.
Charles L. Goldsmith (S’79–M’80–SM’94) received the B.S. and M.S. degrees from the University of Arizona, Tucson, and the Ph.D. degree from the University of Texas at Arlington, all in electrical engineering. Since 1982, he has been affiliated with M/A COM, Texas Instruments (TI) Incorporated, and Raytheon. In 2001, he founded the MEMtronics Corporation, Plano, TX, where he pursues business opportunities for RF MEMS. Since 1993, he has developed RF MEMS devices and circuits. He is the inventor of the capacitive membrane RF MEMS switch. He has authored or coauthored over 45 technical papers. He holds nine patents.
Ian Johnston received the B.S. degree in electronics and Ph.D. degree in microelectronics from the University of Southampton, Southampton, U.K. He is a Project Manager with Innovative Micro Technology, Santa Barbara, CA, where he has been working on capacitive RF switches and related components for five years.
James C. M. Hwang (M’81–SM’82–F’94) received the B.S. degree in physics from National Taiwan University, Taipei, Taiwan, in 1970, and the M.S. and Ph.D. degrees in materials science from Cornell University, Ithaca, NY, in 1976 and 1978, respectively. After working with IBM, AT&T, GE, and GAIN, he joined Lehigh University, Bethlehem, PA, in 1988. He has authored or coauthored over 200 technical papers. He holds four U.S. patents. His current research interests include MEMS, microwave transistors and integrated circuits, lasers, and photodetectors. Dr. Hwang was a recipient of the 2007 IBM Faculty Award.
Suresh Sampath received the M.Sc. degree from University of Hyderabad, Andhra Pradesh, India, in 1992, and the Ph.D. degree from Michigan Technological University, Houghton, in 1998, both in physics. He is currently a Program Manager with Innovative Micro Technology (IMT), Santa Barbara, CA, where he is responsible for managing multiple MEMS projects as well as managing vacuum engineering department that includes thin film depositions, dry etching, and wafer bonding. He has successfully guided multiple MEMS programs from initial R&D phase to manufacturing/production phase.
David I. Forehand (M’84) received the B.S. and M.S. degrees in chemical engineering from the University of New Mexico, Albuquerque. From 1989 to 2001, he was with Texas Instruments (TI) Incorporated, Dallas, TX, where he was involved with infrared focal-plane infrared detector arrays and micromirror-based digital light processors. In 2001, he joined Raytheon’s RF MEMS team to lead process development. In 2002, he helped found the MEMtronics Corporation, Plano, TX, to continue the development and manufacturing of RF MEMS. He holds four patents. Mr. Forehand was the recipient of the 1995 TI Incorporated DSEG Technical Award for Excellence.
Arindom Datta received the Ph.D. degree in materials science from the Indian Institute of Technology (IIT), Kharagpur, India, in 2001. He worked as a Research Scientist and Visiting faculty in different academic institutions including University of Wisconsin-Madison, where he worked on metal embedded micro/nano sensors. He is currently working as a Process Development Engineer with Innovative Micro Technology (IMT), Santa Barbara, CA. His area of research is thin films deposition and characterization, micro/nanofabrication, and MEMS/NEMS. He has over 35 publications in refereed journals, conference proceedings including technical presentations in international conference and symposia. He has two U.S. patents, one issued and one pending.
Authorized licensed use limited to: LEHIGH UNIVERSITY. Downloaded on December 14, 2009 at 14:10 from IEEE Xplore. Restrictions apply.
