You are on page 1of 9

290

IEEE JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 8, NO. 3, SEPTEMBER 1999

Speed-Energy Optimization of Electrostatic Actuators Based on Pull-In


Luis M. Casta er, Senior Member, IEEE, and Stephen D. Senturia, Fellow, IEEE n
AbstractThe speed and total energy required to accomplish pull-in switching of a generic electrostatic actuator is examined. It is found that the value of the source resistance of the voltage drive used for switching has a profound effect on both switching speed and energy requirements. The source resistance governs the charging time for the actuating capacitor. As long as this time is slower than the time required to accelerate the moving mass to maximum speed in the presence of damping, the total energy required for switching can be dramatically reduced without a signicant increase in switching time. Indeed, there exists a clear optimum source-resistance value that minimizes the product of switching time and switching energy. These ndings are demonstrated theoretically and then applied to specic examples from the literature. In addition, the limiting case of very large source resistance, essentially a current drive, is evaluated and compared to the voltage-driven case. It is found that for equivalent switching times, the current drive requires less total energy for a switching event. [374] Index TermsActuators, electrostatic devices, energy management, switching transients.

Fig. 1. Schematic of parallel-plate electrostatic actuator. The capacitor gap is . Squeeze-lm damping in this gap gives rise to the damper.

I. INTRODUCTION HERE is a wide range of actuators that depend on electrostatic pull-in for their operation. Pull-in is the well-known instability that results from the nonlinearity of parallel-plate electrostatic actuation working against a linear elastic restoring force and happens at 2/3 of the initial gap [1]. Pull-in devices include electrostatically actuated microrelays [2][12], optical ber network switches [13], [14], and electromechanical display devices [15], [16]. In many of these applications, it is desirable to optimize the energy consumed during the switching cycle; this is particularly important where it is desirable to have the devices operating in a portableelectronics environment. Surprisingly, there have been relatively few comprehensive studies of energy storage, transfer, and power dissipation in electrostatically actuated pull-in-dependent structures [17]. We have examined this issue using a lumped-model equivalent circuit for a parallel-plate electrostatic actuator working against a lumped linear spring with a damper added to account for squeeze-lm damping of the plate motion (Fig. 1). This is
Manuscript received August 5, 1998; revised May 18, 1999. This work was supported in part by the United States Defense Advanced Research Projects Agency under Contract J-FBI-95-215 and in part by the Spanish CICYT under Contract TIC-98-504. Subject Editor, N. de Rooij. L. Casta er was on leave at the Massachusetts Institute of Technology, n Cambridge, MA 02139 USA. He is with the Universidad Polit cnica de e Catalu a, Barcelona, Spain. n S. D. Senturia is with the Massachusetts Institute of Technology, Cambridge, MA 02139 USA. Publisher Item Identier S 1057-7157(99)07183-8.

a good rst-order representation of a wide variety of pullin devices, including both elastically supported rigid bodies and exible beams [18][21]. The usual assumption is that the actuation is voltage-controlled. What we have observed is that when considering the total energy transfer from the voltage source during the pull-in switching event, it is extremely important to pay attention to the value of the source resistance of the voltage supply. The reason is that as the pull-in collapse occurs, the - time constant of the supply circuit can become what limits the rate of charging of the actuation capacitor. When this occurs, the actuation shifts from being voltage-controlled to being charge-controlled. And under charge-controlled conditions, the power dissipation in the source resistor is vastly reduced. Therefore, by proper choice of the source resistance, it becomes possible to optimize a gure of merit, expressed as the product of pull-in time and total energy required to accomplish switching. Based on this observation, we have also modeled the totally chargecontrolled switching event, in which the voltage source plus source resistor is replaced by an ideal current source. It is seen that the charge-controlled system is a logical extension of the results obtained from the voltage-controlled system with increasing source resistance. The model in Fig. 1 does not include the parasitic capacitance. The effect this simplication has on the results is discussed in Section IV below. In the next section, nondimensionalized dynamical equations for the model of Fig. 1 are developed, along with normalized stored energy and power dissipation units for all elements. Then, the dynamics of the pull-in switching event are studied as functions of the normalized source resistance and normalized voltage drive, with particular emphasis on the development of stored potential energy in the capacitor and the spring, the kinetic energy in the inertial mass, and the power

10577157/99$10.00 1999 IEEE

CASTANER AND SENTURIA: SPEED-ENERGY OPTIMIZATION OF ELECTROSTATIC ACTUATORS BASED ON PULL-IN

291

dissipation in the source resistor and the assumed linear uid damper. Based on these results, the speed-energy optimization is studied, followed by the modeling of the charge-controlled case. Finally, the normalized results are scaled to physical device behavior using dimensions from examples drawn from the literature. II. VOLTAGE DRIVE MODEL Consider now the actuator model of Fig. 1. The capacitor involved in the actuation has capacitance given by (1) where is the permittivity of the dielectric (presumed to equal that of free space), is the area of the capacitor, and is the gap, which varies according to the displacement of the upper plate represented by . That is, (2) The voltage across this capacitor, by voltage , is related to the source

where (9) Additionally, the normalized voltage across the capacitor becomes (10) As can be seen, the parameter is a scaled damping parameter and is the product of the time constant and the mechanical resonance frequency . The dynamic behavior of the system for a given source voltage waveform is uniquely described by these two parameters. It is also useful to normalize the power and energy variables using the natural unit of energy (11) Using this energy unit in combination with the other normalized quantities, the various energy and power variables become: Electrostatic energy stored in the capacitor: (12)

(3) Elastic energy stored in the spring: is the charge on the capacitor (equal to ), and where is the time. The displacement of the inertial mass, which is also the upper plate of the capacitor, is governed by the following differential equation: (4) where is the damping coefcient, is the elastic constant of is the initial gap, and is the capacitance at the spring, is a measure of the initial gap. (Note that the product .) the capacitor area and equals In order to examine the energy storage and power dissipation in this system, it is useful to normalize these equations. Distance is normalized to the initial gap , time is normalized to the inverse of the undamped mechanical resonance frequency , voltage is normalized to the pull-in voltage , and charge which is the charge is normalized to the pull-in charge and gap is stored when the capacitor voltage is equal to reduced to 2/3 of the initial gap. The normalized variables are (5) where (6) leading to normalized dynamical equations (7) (8) Power dissipated in the damper: (15) Power dissipated in the resistor: (16) The normalized power and dividing each equation by The total energy supplied equal the sum of the energy where in the system. Thus, energy variables are found by . must by the voltage source either stored or dissipated else(17) is the total energy, and and are the time inwhere tegrals (with respect to normalized time) of the corresponding power dissipation. In the following studies of transient behavior, we use the normalized dynamic equations (7) and (8) to simulate the dynamics, then evaluate the various power and energy variables. What we will show, as stated earlier, is that the value of the source resistance, hence the parameter , has a large effect on the total energy dissipation through an actuation event. (13) Kinetic energy of the inertial mass: (14)

292

IEEE JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 8, NO. 3, SEPTEMBER 1999

III. TRANSIENT DYNAMICS Extensive transient simulations have been made using MATand . The LAB for various values of the parameters values were selected based on practical examples found in the literature [3], [8], [11], [21] which dene an effective design space for this model. The ranges of selected values are described below. The quality factor QF of a basic mechanical resonator is as related to (18) corresponds to critical damping of Thus the value the primary resonance. Examining the literature, we found values ranging from 0.375, corresponding to an underdamped quality factor of 1.33, to an overdamped value of 4.85; in our simulations, we used values of 0.1, 1, and 10. And for the various actuators in the literature, actuation voltages typically exceeded the pull-in voltage by about 10%; in this work, we of 1.1, 1.5, and 2. Values of used normalized values for were varied over a wide range to explore fully this aspect of the design space. In order to illustrate the dynamical behavior of this system, and , and we rst select nominal typical values then look at the effects of varying . We then examine effects of variations in and . An important issue is how to model the end of the pull-in transient. We have elected simply to terminate the simulation , so as to avoid the modeling singularity when at . It can be argued that to carry the simulation this far is unrealistic because of the large electric elds that would be present across the vanishing small gap, leading to breakdown. However, the main point of this paper is that a nite source resistance will prevent the runaway increase in electric eld that accompanies the vanishing gap, because in the charge-limited regime, described below, the voltage across the capacitor at the end of the transient is actually decreasing as the gap decreases. Of course, the various values of the total energy at the end of the transient depend slightly on our . But our goal was to study the trend with use of varying source resistance, and that trend will be clear from our results shown below. Fig. 2 shows the dynamical behavior for an extremely small . This case, in effect, represents value of the voltage source as ideal, with a truly negligible source resistance. The charging time for the capacitor (all times are expressed in normalized time units) is then less than 1 10 ; therefore, the simulation can effectively begin with , but with the the capacitor in its equilibrium position capacitor charged to its initial normalized value. This value can be calculated as (19)

(a)

(b) Fig. 2. (a) Normalized displacement and velocity during the acceleration phase of a pull-in switching event. (b) Normalized displacement and velocity during the damping and pull-in phases of a pull-in switching event.

Assuming

Following this very short charging transient, there are three dynamic events. The rst is the acceleration phase, illustrated in Fig. 2(a), which lasts until a normalized time of about 0.1. During this phase, the velocity of the mass increases rapidly, but the position is still close to the initial rest position. The velocity reaches a maximum at a normalized time of about unity [see Fig. 2(b)], at which point the system is in the damping phase, with an initial decrease in velocity due to damping, followed by an increase in velocity as the nonlinear electrostatic force grows due to the continuing decrease in gap. The nal phase is the pull-in phase, characterized by a rapid increase in velocity with the mass moving to its nal . What is interesting here is the rapid rise position in the normalized charge at the end of the transient. This rise indicates that there is a sharp pulse of current at the moment of pull-in when the system is driven from a nearly ideal voltage source, and this turns out to have important consequences in the amount of energy consumed during the transient. The initial charging time is set by the value of , but it is the acceleration phase that determines the initial dynamics. Therefore, one would expect that one could increase until

CASTANER AND SENTURIA: SPEED-ENERGY OPTIMIZATION OF ELECTROSTATIC ACTUATORS BASED ON PULL-IN

293

(a)

Fig. 4. Normalized voltage transients for vs = 1:1; values of  .

= 1, and varied

(b)

Fig. 3. (a) Normalized pull-in time versus normalized time constant for = 1 and variable applied voltage. (b) Normalized pull-in time versus normalized time constant for vs = 1:1 and variable . Fig. 5. Normalized energy components during a switching transient.

the charging time is on the same order as the acceleration time (normalized time 1) and see only a modest increase in the total time required for pull-in. This is illustrated in Fig. 3(a), with three values of . As expected, the for the case pull-in time decreases with increasing voltage, but the onset of an increase due to the increase in source resistance is similar for all three voltages. If the damping is changed, the time at which the acceleration phase ends has a corresponding variation. Fig. 3(b) shows the , but with three pull-in time for a xed voltage of widely varied values of . At the lower values of , the behavior reects the strong effect of damping on the pullin time, a result that is well known from many experimental examples [21]. But for each value of , it is seen that the increase in pull-in time due to the source resistance begins . And for , the pull-in time shifts at a value of to a limit that is relatively independent of , which suggests a different dynamical mode. We call this the charge-limited mode. This charge-limited behavior is illustrated in Fig. 4, in which the capacitor voltage is plotted against time during a and for several values pull-in transient with

of . The charging transient is clearly evident in this gure. For , in this case for , the voltage charges quickly and then holds constant at 1.1 all the way to the moment of pull-in. However, for larger values of , the voltage begins to decrease before pull-in, and, in fact, never fully reaches the value of 1.1 anywhere during the transient. This is because during the pull-in event, the capacitor is changing so rapidly compared to the charging time that the voltage supply cannot keep the capacitor fully charged. Instead, the behavior is more like a xed-charge capacitor pulling in with constant force, and since the capacitance is increasing during the last phase of pull-in because of the collapsing gap, the voltage (at xed charge) must decrease. This means that there is no longer a spike of current at the end of the pull-in transient, and since the delivery of that current spike costs energy, it may be possible to save energy without compromising speed by an optimal choice of . This is examined in the next section. IV. SPEED-ENERGY OPTIMIZATION Fig. 5 shows a plot of the normalized total energy required from the source during a single pull-in transient, together with

294

IEEE JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 8, NO. 3, SEPTEMBER 1999

Fig. 6. Normalized energy components from Fig. 5 just before pull-in. (a)

Fig. 7. Normalized total energy versus normalized time constant for and for vs = 1:1; 1:5, and 2.

= 1,

(b) Fig. 8. (a) Product of normalized total energy and normalized pull-in time versus normalized time constant for vs = 1:1 and varied values of . (b) Product of normalized total energy and normalized pull-in time versus normalized time constant for = 1 and varied vs .

all the constituents [as dened in (11)(17)], for the case , and a value of , corresponding to the voltage-source driven case. Fig. 6 shows these same quantities just at pull-in. At the beginning of the transient in Fig. 5, before there is any signicant motion, the source energy is converted to electrostatic stored energy on the capacitor, together with an equal amount of energy dissipation in the resistor. Kinetic energy grows during the acceleration phase and then declines as the damping phase is entered. Signicant changes are observed near pull-in (Fig. 6): the kinetic energy climbs rapidly and becomes the major component followed by, in order, the resistor losses, damping losses, and elastic energy. The electrostatic energy declines sharply due to the onset of charge-limited behavior. Since the kinetic energy and resistor losses become the largest components just before pull-in, and since these rapid increases are both due to the small value of as the gap collapses, it is reasonable to expect that if the value of were increased, establishing charge-limited behavior earlier in the cycle, it should be possible to reduce the total energy required for a pull-in event. This is examined in Fig. 7, which plots , just before the mass the total required energy at and for increasing values collides with the substrate, for

of . Indeed, at each voltage level, the total energy decreases dramatically with increasing . If we recall from Figs. 3 and 4 that until becomes comparable with , there is only a very slight increase in pull-in time, it is clear that a tradeoff between pull-in time (i.e., switching speed) and total energy required per switching event will lead to an optimum choice of . The existence of an optimum choice for is illustrated in Fig. 8. In Fig. 8(a), the product of the normalized total energy and the normalized pull-in time has been plotted against for a xed voltage of and for three values of . It is seen that for each value of , this product is minimized , the precise optimum occurring for slightly less for than . Fig. 8(b) shows this same product plotted at a xed for three values of voltage. Here it is seen value of that the optimum, while shifting slightly to larger values at voltages well above pull-in, is well dened and is located in . the vicinity of In Section I, the model we have used to derive the results shown above did not include a parasitic capacitance. If there is a parasitic capacitance in parallel with the ac-

CASTANER AND SENTURIA: SPEED-ENERGY OPTIMIZATION OF ELECTROSTATIC ACTUATORS BASED ON PULL-IN

295

tuation capacitance, as will be the case to some degree in all practical devices, some of the results presented here will be affected. Including parasitic capacitance introduces another nondimensional parameter into the model, namely, the ratio of the parasitic capacitance to the actuation capacitance. Full quantitative study of this more complex model and its parameter space will be reported separately. However, the qualitative effects of a parasitic capacitance can be readily understood, as discussed below. The major idea of this paper, namely that the source resistance can be increased until the charging time approaches the characteristic damping time without an increase in pull-in time, continues to be correct, but the charging time is now determined by the combination of the parasitic and actuation capacitances. Hence, it will not be possible to use as large a value of source resistance without compromising pull-in time, and this, in turn, will reduce the energy savings at the speedenergy optimum. And if the parasitic capacitance is much larger than the actuation capacitance, it will be necessary to place as much as possible of the total source resistance between the parasitic and actuation capacitances; only then will the benets of the transition to charge-limited behavior be realizable in practice. Finally, the total energy reported here must be increased to include the additional energy required to charge the parasitic capacitance. As this parasitic capacitance increases, the actuation energy savings achieved by increasing the source resistance will be a less signicant fraction of the total energy required to accomplish switching. The results shown above describe the transient up to the pull-in of the actuator. If the voltage supply is left on after pull-in has occurred, then the pulled-in capacitor (if it still is a capacitor and not simply a resistive contact as in [20]) gradually charges to the full supply voltage at a rate determined by the source resistance. If the source resistance is large, as is suggested for optimum speed-energy performance, this recharging will be slow, but it will nevertheless take place. If it is necessary to hold the switch in the pulled-in position, then the results of this paper suggest that the most energyefcient way to control this actuator is to use a large source voltage and correspondingly large source resistance to drive the pull-in event, taking maximum advantage of the chargelimited regime at the end of the pull-in transient, and then to reduce the source voltage to as low a value as possible, but above the release voltage of the device. V. CURRENT DRIVE The results shown above clearly demonstrate the importance of the source resistance in the dynamics of electrostatic actuators. The energy consumed can be dramatically reduced by increasing the internal resistance of the drive without signicantly reducing the speed of the device, at least until the source resistance gets very large. This suggests that, at least for theoretical purposes, it would be useful to examine the case of an ideal current-source drive. Based on our original model, we can identify the normalized Norton equivalent source current , and using this quantity in the limit, becomes as innite (corresponding to an ideal current source drive), (8)

(a)

(b)

(c)

(d)

Fig. 9. Comparison of state dynamics for voltage-driven (solid) and current-driven (dashed) switching transients. (a) Normalized displacement. (b) Normalized velocity. (c) Normalized charge. (d) Normalized voltage.

becomes (20) For a physical current source with a high enough source resistance to be effectively innite, the appropriate normalization that is consistent with earlier denitions is (21) For a constant-current source, (20) has the trivial solution grows proportional to . Equation (7) can then be that solved to yield an analytical solution consisting of a seconddegree polynomial in (forced response) plus two naturalresponse terms. However, for comparison purposes with the general voltage-driven case, we continue to use straightforward numerical integration of the state equations. The important differences in the dynamics of voltage drive versus current drive for these electrostatic actuators are shown in Figs. 9 and 10. For the idealized voltage-drive case, , and were used. Then, the value of the idealized current source was adjusted to yield the same ; this value is . A practical pull-in time with current source, built, for example, with a transistor circuit, that exceeds the pull-in would require a supply voltage voltage in order for this model to apply over the entire range of voltages that appear during the transient, but as long as this condition is met, the current source model is realizable. Fig. 9(a) and (b) shows the normalized deection and velocity for the two cases. As can be seen, with current drive, the beam approaches pull-in more gradually, and has a velocity at the end of the transient that is two orders of magnitude

296

IEEE JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 8, NO. 3, SEPTEMBER 1999

MODEL PARAMETERS

FOR

TABLE I ACTUATORS

FROM THE

LITERATURE

(a)

(b) Fig. 10. Comparison of energy for voltage-driven (solid) and current-driven (dashed) switching transients. (a) Normalized kinetic energy. (b) Normalized electrostatic energy.

Section III. Fig. 10 shows the comparison of the normalized kinetic energy and electrostatic energy for the two cases. As expected, the kinetic-energy term is smaller for the currentsource drive, and while the electrostatic energy climbs sharply at pull-in for the voltage-source drive, it drops during the pullin phase for the current-source drive, demonstrating that in this latter case, the stored electrostatic energy is being used as the energy source during the nal pull-in phase. In order to model the source-resistance losses for the case of a current-source drive, it becomes necessary to consider the detailed circuit implementation. That is, a practical currentsource drive would require some kind of transistor circuit, and while it is possible to analyze all of the energy transfer in the elements external to the drive circuit, it is not possible to evaluate the source-resistance losses without specifying the drive circuit in detail (for example, simply using a Nortonequivalent drive circuit with a large parallel source resistance fails, because the Norton circuit dissipates internal power under open-circuit conditions, while the corresponding Th venin e circuit does not). Detailed analysis of transistor-circuit currentsource congurations will be the subject of a separate paper. VI. SCALING
TO

DEVICE DIMENSIONS

lower compared to the voltage-drive case. This means that the impact with the substrate is more gentle. Fig. 9(c) and (d) shows the comparison of normalized charge and voltage for the two cases. It is clear that the charge builds gradually (in fact, linearly with time) when current drive is used, whereas in the voltage-drive case, the charge has a very large derivative at the beginning and at the end of the transient. The different dynamics are also observed looking at the voltage pattern [Fig. 9(d)], where a square shape is observed in voltage drive and a gradual evolution with a maximum is observed in current drive. This is consistent with the observations made in

The model described in Section II can be scaled to real device geometries taken from the literature. From the key references cited in Section I, we have analyzed four cases taken from [3], [8], [11] and [21] and proceeded to analyze the evolution of the pull-in time/energy product as described in Section III and in Fig. 8. In all cases, geometrical data were found in the references: thickness, initial gap, and area, as well as pull-in times and pull-in voltages. In some cases, the values for the elastic constant and of the damping parameter were also provided. In the cases where the value for the elastic constant was not given, it was inferred from the reported pullin voltage and from device dimensions. In cases where the linear damping coefcient was not given, it was estimated

CASTANER AND SENTURIA: SPEED-ENERGY OPTIMIZATION OF ELECTROSTATIC ACTUATORS BASED ON PULL-IN

297

by doing several simulations and adjusting in order to get agreement with the reported values of the pull-in times. Hand . Once all calculations provided values for the mass and the values for the parameters of our model are known, we proceeded to systematic MATLAB simulations for several values of and the minimum value of the pull-in-time/energy product calculated. The results are shown in Table I. The last four rows of Table I are the values generated by our simulations. The were typically selected to values for the drive voltage be 10% higher than the pull-in voltage. As can be seen, the optimum product is obtained for rather high values of the drive source resistance for all cases considered, typically in the 1 to 300 M range. The resulting optimum speed-energy product is in the range of 1 to 500 fJ s. VII. CONCLUSION We have demonstrated that the specic value of the source resistance used in voltage-driven electrostatic actuators that exploit pull-in has an enormous effect on the total power dissipation during a switching transient. In particular, we have identied four phases during switching: charging, acceleration, damping, and pull-in. The key idea is that total power can be reduced if the charging time is increased to be on the same order as the characteristic damping time, thus combining the charging and acceleration phase into a single start-up phase. It is also shown that when the source resistance is increased to reduce power, a charge-limited mode of operation occurs during the pull-in phase, when the capacitor voltage actually drops as the gap collapses. This has the effect of greatly reducing both the power dissipation in the source resistance and the kinetic energy transfer in the nal switching phase. Both contribute signicantly to a reduction in total energy required for a switching event. Because there is a tradeoff between switching speed and total switching energy as the source resistance is varied, there is also an optimum choice of source resistance that will minimize the pull-in-time/energy product. This has been examined both using normalized general quantities and specic quantities based on studies from the literature. We nd that for all the cases in the literature, the optimum pull-in time/energy product requires relatively large source resistances, in the megohm range. Finally, as the limiting case of an innite source resistance, we examine the purely current-driven case and show that for the same switching time as the voltage-driven case, current drive can be energetically favorable. However, detailed analysis of the source losses requires specication of the exact circuit used to achieve current-source drive. This subject will be studied separately. ACKNOWLEDGMENT The authors are indebted to the Programa de Becas en el Extrajero de la Secretaria de Estado de Universidades e Investigaci n del Ministerio de Educaci n for support during o o the 1997 sabbatical leave of L. Casta er at the Massachusetts n Institute of Technology.

REFERENCES
[1] H. C. Nathanson, W. E. Nevel, R. A. Wickstrom, and J. R. Davis Jr., The resonant gate transistor, IEEE Trans. Electron Devices, vol. ED-14, pp. 117133, 1967. [2] K. E. Petersen, Micromechanical membrane switches on silicon, IBM J. Res. Develop., vol. 23-4, pp. 376385, 1979. [3] J. Schimkat, L. Kiesewetter, H. G. Gevatter, F. Arndt, A. Steckenborn, and H. F. Schlaak, Moving wedge actuator: An electrostatic actuator for use in a microrelay, in Proc. 4th Int. Conf. Micro Electro Opto, Mechanic Systems and Components, Microsystem Technologies94, Berlin, Oct. 1994, pp. 989996. [4] S. Roy and M. Meheregany, Fabrication of electrostatic nickel microrelays by nickel surface micromachining, in Proc. 8th Workshop on Microelectromechanical Systems MEMS95, Amsterdam, 1995, pp. 353357. [5] J. Drake, H. Jerman, B. Lutze, and M. Stuber, An electrostatically actuated microrelay, in Dig. Tech. Papers, 8th Int. Conf. Solid State Sensors and Actuators, Transducers95, Stockholm, 1995, vol. 2, pp. 380383. [6] J. J. Yao and M. F. Chang, A surface micromachined miniature switch for telecommunications applications with signal frequencies from dc up to 4 GHz, in Dig. Tech. Papers 8th Int. Conf. Solid State Sensors and Actuators, Transducers95, Stockholm, 1995, vol. 2, pp. 384387. [7] M. A. Gr tillat, P. Thi baud, C. Linder, and N. F. de Rooij, Integrated e e circuit compatible electrostatic polysilicon microrelays, J. Micromech. Microeng., vol. 5, pp. 155160, 1995. [8] C. Goldsmith, T. Lin, B. Powers, W. Wu, and B. Norvel, Micromechanical membrane switches for microwave applications, in Tech. Dig. MTT-S, 1995, pp. 9194. [9] D. J. Young and B. Boser, A micromachined variable capacitor for monolithic low noise VCOs, in Tech. Dig. Solid State Sensor and Actuator Workshop, Hilton Head Island, SC, June 36, 1996, pp. 8689. [10] P. Zavracky, S. Majumder, and N. E. McGruer, Micromechanical switches fabricated using nickel surface micromachining, IEEE J. Microelectromech. Syst., vol. 6, no. 1, pp. 39, 1997. [11] I. Schiele, J. Huber, C. Evers, B. Hillerich, and F. Kozlowski, Micromechanical relay with electrostatic actuation, in Proc. 9th Int. Conf. Solid-State Sensors and Actuators (Transducers97), Chicago, June 1997, pp. 11651168. [12] M. Sakata, Y. Komura, T. Seki, K. Kobayashi, K. Sano, and S. Horiike, Micromachined relay which utilizes single crystal silicon electrostatic actuator, in Proc. 12th IEEE Int. Conf. Microelectromechanical Systems MEMS99, Orlando, FL, 1999, pp. 2124. [13] E. Ollier and P. Mottier, Integrated electrostatic microswitch for optical ber networks driven by low voltage, Electron. Lett., vol. 32, pp. 20072009, 1996. [14] C. Marxer, M. A. Gr tillat, O. Anthamatten, R. Baettig, B. Valk, P. e Vogel, and N. F. de Rooij, Vertical mirrors fabricated by reactive ion etching for ber optical switching applications, in Proc. 10th Workshop on Microelectromechanical Systems MEMS97, Nagoya, Japan, 1997, pp. 4954. [15] P. F. van Kessel, L. J. Hornbeck, R. E. Meier, and M. R. Douglass, A MEMS-based projection display, Proc. IEEE, vol. 86, pp. 16871704, 1998. [16] O. Solgaard, F. Sandejas, and D. Bloom, A deformable grating optical modulator, Opt. Lett., vol. 17, p. 688, 1992. [17] H. Fujita and A. Omodaka, Electrostatic actuators for micromechatronics, in Proc. IEEE Int. Workshop on Microelectromechanical Systems (MEMS87), Micromechanics and MEMS, 1996, pp. 276285. [18] V. P. Jaecklin, C. Linder, N. F. de Rooij, J. Moret, and R. Vuilleumier, Optical microshutters and torsional mirrors for light modulated arrays, in Proc. IEEE Int. Workshop on Microelectromechanical Systems (MEMS93), Ft. Lauderdale, FL, Feb. 710, 1993, pp. 124127. [19] R. K. Gupta, E. S. Hung, Y.-J. Yang, G. K. Ananthasuresh, and S. D. Senturia, Pull-in dynamics of electrostatically actuated beams, in Tech. Dig. Solid State Sensor and Actuator Workshop, Late-News- Paper Suppl., Hilton Head, SC, June 1996, pp. 12. [20] R. Puers, J. Bienstman, and J. Vandewalle, The autonomous impact resonator: A new operation principle for a silicon strain gauge, in Proc. Int. Conf. Solid-State Sensors and Actuators (Transducers97), Chicago, June 1619, 1997, pp. 11051108. [21] R. K. Gupta and S. D. Senturia, Pull-in time dynamics as a measure of absolute pressure, in Proc. IEEE Int. Workshop on Microelectromechanical Systems (MEMS97), Nagoya, Japan, Jan. 2630, 1997, pp. 290294.

298

IEEE JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 8, NO. 3, SEPTEMBER 1999

Luis M. Castaner (SM92) received the degree of Ingeniero Superior de Telecomunicaci n from the Universidad Polit cnica de Madrid in 1971 and o e Dipl me dEtudes Approfondies (D.E.A.) en Physique Spatiale and Docteuro Ingenieur by the Universit Paul Sabatier, Toulouse, France, in 1973. He e received the Doctor Ingeniero de Telecomunicaci n from the Departamento de o Ingeniera Electr nica de la Escuela T cnica Superior de Ingenieros (E.T.S.I.) o e de Telecomunicaci n at the Universidad Polit cnica de Catalu a, Barcelona, o e n Spain. He is a Professor with the Departamento de Ingeniera Electr nica de o la Escuela T cnica Superior de Ingenieros de Telecomunicaci n at the e o Universidad Polit cnica de Catalu a. He has also been Dean of E.T.S.I. e n Telecomunicaci n and Head of Departamento de Ingeniera Electr nica and o o has held several positions in research agencies and committees: representative in the Comite de Gestion y Coordinacion on nonnuclear energies, DGXII Comission of the EU, Head of the Programa Nacional de Tecnologas de la Informaci n y Comunicaciones of the CICYT (19921994), and Coordinator o of the Technology Foresight at the Agencia Nacional de Evaluaci n y o Prospectiva. He has contributed to semiconductor device research, covering solar cells in various aspects: technology of CIS, space degradation of silicon and GaAs devices, and has contributed to the theory and technology of advanced bipolar transistors with polysilicon emitters and their application to high-efciency silicon solar cells, in particular, to the emitter resistance of these devices. He has also contributed to the design, simulation, and monitoring of photovoltaic power plants and systems and is active in the microsystems technology area, working on ow sensors, power devices for MEMS, and electrostatic actuators. Dr. Casta er is a member of the Association and Charter of Telecomn munication Engineers in Spain and a member of the Spanish Engineering Academy.

Stephen D. Senturia (M77SM91F93), for a photograph and biography, see this issue, p. 289.

You might also like