Stability of surface tension self-assembled 3D MOEMS

Transcription

1 Sensors and Actuators A 127 (2006) Stability of surface tension self-assembled 3D MOEMS Y.K. Hong, R.R.A. Syms Optical and Semiconductor Devices Group, Electrical and Electronic Engineering Department, Imperial College London, Exhibition Road, London SW7 2BT, UK Received 28 September 2005; received in revised form 2 December 2005; accepted 9 December 2005 Available online 23 January 2006 Abstract The thermal and mechanical stability of self-assembled MOEMS components consisting of 3D mirrors have been evaluated. The thermal mechanical constants of the hinge material (AZ4562 photoresist) used to power the surface tension self-assembly process have been measured and used to predict the stability of assemblies. The Young s moduli and thermal expansion coefficients showed constant values of average N/m 2 and C 1 under various thermal conditions. The angle variation of all the components tested under thermal cycling was less than C 1 up to 110 C. The use of mechanical limiters enhanced the thermal stability to near 0 up to 150 C. During dynamic mechanical stability measurement, for a maximum external acceleration of >50 g, no device failure was observed Elsevier B.V. All rights reserved. Keywords: Thermal stability; Dynamic stability; MOEMS; Surface tension; Self-assembly 1. Introduction 3D micro-opto-electro-mechanical-systems (MOEMS) components including moving mirror switches [1 3], scanners [4 6], beamsplitters [7], and corner cubes [8,9] are used to create small optical sub-systems that process quasi-free space beams travelling above the surface of a chip. These 3D structures are formed by out-of-plane rotation of parts hinged to the substrate [10]. So far, the assembly has mainly been manual [9,11] or by using additional actuators. Methods used to power rotation include electromagnetic force [12], surface micromachined vibromotors [13], and microengines [14]. In particular, 3D microstructures actuated by a microengine have been demonstrated at Sandia National Laboratories, using a microtransmission to obtain sufficient torque with a low input voltage [15,16]. The efficiency of the microengine has also been improved by design optimization [16]. Although these methods allow dynamic repositioning, they are complex and require large chip area. Some failure mechanisms including wear and adhesion remain as challenges [16]. An alternative process uses surface tension force. Rotation is powered by melting small pads of material (for example, solder [3,17], glass [18], or photoresist [19 21]) linking movable parts Corresponding author. Tel.: ; fax: address: r.syms@ic.ac.uk (R.R.A. Syms). to the substrate. This process provides a mass-parallel fabrication method for 3D microstructures. Recent improvement using thick resist, and single crystal parts formed in bonded siliconon-insulator (BSOI) provides a simple, compact, precise, and IC-standard compatible process [21]. Using this method, a number of MOEMS components including mirrors [21], lenses [22], scanners [20], and corner cubes [23] have been demonstrated. Although surface tension powered self-assembly is a compact batch fabrication method, there are possible concerns about the structural rigidity of hinges made from polymer materials. For example, relatively short life-times have been reported for a conjugated polymer actuator [24]. On the other hand, Ebefors et al. have demonstrated a robust hinge structure made by polyimide [25]. Since polymer materials usually have high thermal expansion coefficients, the thermal stability of self-assembled MOEMS components requires evaluation. Similarly, dynamic testing is also required to evaluate mechanical stability. Although some dynamic testing of MOEMS devices [26,27] and of RF- MEMS devices [28] has been reported, full evaluation of surface tension self-assembled devices has not been carried out. In this paper, we have investigated both the thermal and the dynamic stability of self-assembled MOEMS components formed by photoresist powered self-assembly. In Section 2, we briefly describe the fabrication process. In Section 3, we determine the thermal mechanical constants of the hinge material. In Section 4, experimental measurements of the thermal stabil /$ see front matter 2006 Elsevier B.V. All rights reserved. doi: /j.sna

2 382 Y.K. Hong, R.R.A. Syms / Sensors and Actuators A 127 (2006) ity of self-assembled MOEMS components are presented. Both metallized and non-metallized components, and components constrained using mechanical limiter mechanisms are evaluated. The results are compared with the predictions of finite element analysis. In Section 5, dynamic testing of MOEMS components is described. Conclusions are presented in Section Fabrication Fabrication of surface tension self-assembled MOEMS is carried out using a simple process involving just two masks. The process is based on BSOI material obtained commercially and consisting of 100 mm diameter Si substrates carrying 3.5 m thick bonded Si layers on 2 m of thermal oxide. Details of the process have been given elsewhere [19 21], and involve the following steps: Patterning of fixed and movable parts on a BSOI substrates by deep reactive ion etching. Patterning of thick pads of photoresist linking fixed and movable parts. Sacrificial etching of the underlying oxide. Freeze-drying to remove wash water without sacrificial layer collapse. Out-of-plane rotation of the movable parts by melting the photoresist. Metallizing the devices by sputter-coating with 500 Å Au. The devices used for testing are corner cubes, consisting of three mutually perpendicular mirrors engaged with a mechanical limiter that is also assembled by out-of-plane rotation. Fig. 1 shows schematic diagrams of 3D self-assembled corner cubes. Fig. 1(a) shows the self-assembly sequence for corner cube reflectors, Fig. 1(b) shows the limiter mechanism for selfassembled 45 structures and Fig. 1(c) shows a fabricated passive corner cube reflector. The overall dimensions vary depending on the device type, but mirror widths are typically 1 mm. The dimensions of the meltable pad are 240 m 40 m with 11 m thickness. Etching is carried out using a Surface Technology Systems inductively coupled plasma etcher equipped with a dual frequency RF source. The sacrificial oxide layer is removed by wet etching for 3 h in 7:1 buffered hydrofluoric acid. The remaining rinse water is then removed by freeze drying using a water/methanol mixture. 3. Thermal mechanical constants of photoresist To evaluate the thermal stability of self-assembled devices, it is essential to know the thermal mechanical constants of the hinge material (e.g., its Young s Modulus, Poisson s ratio and linear thermal expansion coefficient). In our experiment, Shipley AZ4562 photoresist is used as the hinge material. Unfortunately suitable data are not reported in the literature. To determine these constants, a resonant frequency measurement technique suggested by Petersen and Guarnieri [29] and further developed by Osterberg and Senturia was used [30]. This technique required the fabrication of suspended parts using photoresist as a structural material. Arrays of suspended cantilever beams and fixed fixed beams formed in AZ4562 photoresist with 40 m width, 11 m thickness and different lengths were prepared on a silicon substrate with an oxide sacrificial layer. First, the AZ4562 photoresist was spin coated at 1350 rpm for 30 s to form an 11 m thick layer on a 100 mm diameter substrate using a Karl Süss Gyrset spinner equipped with a rotating wafer cover to control the local air flow. To define beams, the photoresist was patterned using a structural feature mask by UV lithography, using a Quintel Q4000-IR aligner. The wafers were then hard-baked in an oven for 30 min to enhance adhesion of the photoresist to the substrate. Three hard-bake temperatures of 120, 140 and 160 C were used, to represent the actual re-melting temperature used during surface tension self-assembly. To free the beams, the oxide sacrificial layer beneath the photoresist was removed by wet etching for 3 h in 7:1 buffered hydrofluoric acid. The remaining washing water was then removed by freeze-drying. A Cr layer of 100 Å thickness was then sputter-coated on the beams to allow an ac voltage to be applied between the beams and the substrate. Fig. 2 shows the two different kinds of beam structure used in subsequent experiments. Fig. 2(a) shows cantilever beams, which have an intrinsic upward bending caused by a stress gradient within the resist and an intrinsic tensile stress in the Cr-layer. Fig. 2(b) shows fixed fixed beams, where much less bending can be seen Measurement of Young s modulus The Young s modulus of the photoresist can be calculated from the resonant frequency of cantilever beams. Since the thickness of the Cr layer is negligible compared to the thickness of the photoresist, the un-damped free vibrations of a beam can be estimated using Euler bending theory [31]. The governing equation for undamped vibration is EI 4 y x 4 + ρa 2 y t 2 = 0 (1) where A is the cross-sectional area of the cantilever, I its second moment of area, E the Young s modulus, and ρ is the density of the beam material, y is the transverse deflection and t is the time. Using this equation, with the boundary conditions y = y/ x =0atx = 0 and 2 y/ x 2 = 3 y/ x 3 =0atx = L, the resonant frequency of a cantilever may be found as ω i = (β il) 2 L 2 EI ρa Similar solutions may be found for different beam configurations. The values of (β i L) 2 thus obtained are shown in Table 1. Fig. 3 shows the measured electrical resonant frequencies plotted against reciprocal length squared. These data represent averages taken from five sets of cantilevers, and the resonant frequencies were very consistent from set to set. The graph shows a linear relation, in agreement with Eq. (2). Knowing (2)

3 Y.K. Hong, R.R.A. Syms / Sensors and Actuators A 127 (2006) Fig. 1. Schematic diagram for (a) self-assembly sequence for corner cube reflectors, (b) limiter mechanism for self-assembled 45 structures, and (c) fabricated passive corner cube reflector. Fig. 2. SEM view of (a) cantilever beam array and (b) fixed fixed beam array used for thermal mechanical constant measurement.

4 384 Y.K. Hong, R.R.A. Syms / Sensors and Actuators A 127 (2006) Table 1 Eigenvalues for a simple beam with different beam configurations Beam configuration First mode Second mode Third mode (β 1 L) 2 (β 2 L) 2 (β 3 L) 2 Cantilever Fixed fixed From Eq. (3) and using the boundary conditions y = y/ x =0at x = 0 and x = L, we can obtain the resonant frequency as ω i = β 4 i ( ) EI + βi 2 ρa ( ) σ ρ When the temperature of the fixed fixed beam is raised, the beam experiences further thermal stress if differential thermal expansion occurs relative to the substrate. By measuring the change of resonant frequency during a heating cycle, the thermal stress can be calculated. Together with any intrinsic stress, the thermal expansion coefficient is then deduced from the thermal stress equation as (4) σ = E PR (α PR α Si ) T (5) Fig. 3. Electrical resonant frequencies of AZ4562 cantilevers as a function of reciprocal length squared. I = m 4, A = m 2 and ρ = 1268 kg/m 3, the Young s modulus of the resist can be found from the slope of the graph. Samples prepared with three different hard baking temperatures were measured to investigate the effect of process temperature. Table 2 shows the calculated Young s modulus. The value shows little variation over the temperature range from 120 to 160 C, which corresponds to the melt temperature used in surface tension self-assembly. The melting temperature therefore shows a relatively stable process window Measurement of thermal expansion coefficient The thermal expansion coefficient (TCE) of the photoresist can be estimated from the vibration of fixed fixed beams. Since the beam is constrained at both ends, the system is now governed by the equation for a beam subject to an axial stress σ. In case of a tensile axial stress, the beam bending equation is [32] EI 4 y x 4 σa 2 y x 2 = ρa 2 y t 2 (3) Table 2 Young s modulus of AZ4562 photoresist with different hard bake conditions Hard bake temperature ( C) Young s modulus ( 10 9 N/m 2 ) where E PR and α PR are the Young s modulus and TCE of the photoresist, α Si the TCE of the substrate silicon and T is the temperature increase. The first resonant frequency of an unstressed fixed fixed beam is 22.4/3.52 = 6.36 times larger than that of a cantilever of the same length, as can be seen from Table 1. Using the first resonant frequency of a cantilever as described in Section 3.1, and measuring the first resonant frequency of a fixed fixed beam, the built-in stress (σ 1 ) of the beam can be determined using Eq. (4). The temperature was then raised and the frequency shift was measured. From the results, the thermal stress (σ 2 ) and the TCE of the resist can be obtained using Eqs. (4) and (5). Measured mechanical resonant frequencies of cantilever beams and fixed fixed beams with a length of 3150 m at room temperature, calculated intrinsic stresses and thermal expansion coefficients (TCEs) are shown in Table 3. The TCE values of the photoresist are smaller and more stable, compared to the values of other common polymers, as shown in Table 4. However, since the process-induced stress increased rapidly as original melting temperature rose, the melting temperature should be carefully selected Measurement of Poisson s ratio Finally, a thermal expansion measurement was performed on a resist-coated silicon wafer to obtain Poisson s ratio. Fig. 4 shows a schematic diagram of the experiment. A 100 mm diameter single crystal silicon wafer coated with AZ4562 photoresist was placed on a hot plate. Differential thermal expansion between the silicon and the photoresist caused the wafer to bow into a spherical convex surface. Laser reflection was used to determine the curvature. The incidence angle of the laser was set to 30, and the angle variation of laser beam reflected from the wafer during the heating was measured at two points, A and B. If we assume the initial substrate to be flat, the change in the radius of curvature (R) after thermal load can be determined from the equation, r AB = Rθ, where r AB is approximately the distance between points A and B for large R, and θ is the angle between points A, O and B. From the geometry, θ can be determined as (θ A + θ B )/2, where θ A, θ B are the angle variations obtained during the thermal excursion at the points A and B, respectively.

5 Table 3 Thermal expansion coefficients of AZ4562 photoresist with different hard bake conditions Y.K. Hong, R.R.A. Syms / Sensors and Actuators A 127 (2006) Hard bake temperature ( C) Resonant frequency (cantilever) (Hz) Resonant frequency (fixed fixed beam) (Hz) Processing stress ( 10 5 N/m 2 ) Frequency shift ratio (Hz/ C) Thermal expansion coefficient ( 10 6 C 1 ) where σ PR is the stress of the photoresist, E Si and ν Si the biaxial Young s modulus and Poisson s ratio of the substrate, and d Si and d PR are thicknesses of the substrate and photoresist, respectively. For the thermal expansion relation, Eq. (5) can be used. In this case, the Young s modulus of the photoresist must be changed to the biaxial value, E PR /(1 ν PR ). Using Eqs. (6) and (5), the thermal mechanical constants relation can then be obtained as [33] E PR (α PR α Si ) = θ A + θ B 1 E Si (d Si ) 2 1 ν PR 2 d PR r AB 6(1 ν Si ) T (7) Fig. 4. Schematic diagram of experiment used for thermal constants measurement. To evaluate the mechanical constants, Stoney s formula [33] for the curvature of a strained wafer was then used: σ PR = E Si (d Si ) 2 (6) 6R (1 ν Si )d PR Knowing α Si and ν Si, the value of ν PR can be calculated from Eq. (7) and the results obtained in the previous section. For the bowing test, an AZ4562-coated silicon wafer with a thickness of 546 m and a resist thickness of 10 m was used. The resist was hard baked at 140 C. A Young s modulus of N/m 2 and a Poisson s ratio of 0.28 for [1 0 0] oriented silicon were used in the calculation. The Poisson s ratio of the AZ4562 photoresist was then found as This value is reasonable by comparison with data for other polymer materials (see Table 4). Table 4 Mechanical constants of materials used in MEMS [34] Material Young s modulus ( 10 9 N/m) Thermal expansion coefficient ( 10 6 C 1 ) Poisson s ratio Dielectric SiO Semiconductor Si Polymer PMMA Epoxy Polyimide PVC Polycarbonate Nylon Polystyrene AZ Metal Al Au Zn

6 386 Y.K. Hong, R.R.A. Syms / Sensors and Actuators A 127 (2006) Fig. 5. A 45 un-engaged mirror used for the thermal stability measurement. 4. Thermal stability of self-assembled MOEMS Thermal stability measurements were performed on three types of self-assembled components, each of which is a subcomponent of a corner cube. These included 45 mirrors (Fig. 5), 90 mirrors, and 45 mirrors with a mechanical limiter. The components were placed on a heater, and the temperature was varied from 25 to 145 C. A laser beam reflection method was used to measure the resulting angular variation of the mirrors. Fig. 6 shows a schematic of the thermal stability measurement. A beam from a He Ne laser mounted on a rotation platform with a Vernier angle gauge was reflected on the mirror surface. The laser was rotated from perpendicular to the substrate plane to perpendicular to the mirror surface. The mirror angle was then measured with a resolution of As the temperature rose, the angle was measured at 10 C intervals. Similar measurements were performed on uncoated and Au-coated samples. Fig. 7 shows the variation of mirror angle with temperature fora49 -rotated mirror, for different temperature excursions. Fig. 7(a) shows a small excursion and (b) a larger one. In Fig. 7(a) Fig. 6. Schematic diagram for thermal stability measurement of self-assembled components. the angle gradually decreases as the temperature rises. The angle variation with temperature shows a roughly linear relation with a slope of C 1, and reversible behaviour up to a temperature of 110 C. However, when the temperature rises beyond 120 C, the angle alters rapidly and shows irreversible characteristics as shown in Fig. 7(b). This result suggests that the melting point of the AZ4562 photoresist lies between 110 and 120 C. The hinge therefore re-melts and further rotates to reduce the surface energy. This behaviour was confirmed for other 45 -rotated mirrors. For Au-coated mirrors, the angle also decreases with the temperature, as shown in Fig. 8 fora52 -rotated mirror. When the temperature rises above 120 C, melting of the photoresist can again be seen, as shown in Fig. 8(b). However, in contrast to the un-metallized mirror, the angular variation of the metallized mirror alters, depending on the position of the optical beam with respect to the mirror. This can be explained by the residual stresses and the difference in thermal expansion coefficients (TCEs) between the gold layer and the silicon layer, as Fig. 7. Variation of angle with temperature for a un-metallized mirror, with temperature excursion (a) below and (b) above the melting point.

7 Y.K. Hong, R.R.A. Syms / Sensors and Actuators A 127 (2006) Fig. 8. Variation of angle with temperature for a metallized mirror, with temperature excursion (a) below and (b) above the melting point. well as the photoresist hinge expansion. Since the metallized mirror is a film-substrate structure, there is initially internal residual stress, and the mirror is curved downward. However, because the TCE of gold ( C 1 ) is larger than the TCE of silicon ( C 1 ), the Au-film is under more compressive stress as the temperature is raised and the mirror curvature increases. The angle variation therefore slightly varies from to C 1 depending on the measuring point. However, the amount of variation is similar to that of the non-metallized mirror. The thermal behaviour can be compared with other results obtained using polymer actuators. For example, Ebefors et al. have reported a degradation of polyimide v-groove actuators above the melting point [25]. Finally, to compensate for melting of the photoresist pads, we measured the temperature stability of 45 mirrors with mechanical limiters, as shown in Fig. 9. For three mirrors measured, the devices show almost zero temperature sensitivity and the angle Fig. 9. Variation of angle with temperature for a metallized mirror with a mechanical limiter. does not change even when the temperature is increased above 150 C. After three repeated temperature cycles up to 150 C, the mirror angle was 44.88, very slightly different from the initial angle of The thermal stability of self-assembled 45 mirrors (Fig. 5) was then estimated numerically using the data obtained in the previous sections. Finite element simulation software (ANSYS) was used for the simulation and the thermal mechanical constants of silicon and AZ4562 photoresist are summarized in Table 4. The mirror dimensions are 976 m 3.5 m with a length of 956 m. The hinge dimensions are 240 m 20 m, following a slight shrinkage during the melting process. There are also 230 m 6 m rectangular holes over the hinge structure to enhance sacrificial layer etching. Since the mirror has a symmetric geometry, only half of the structure was simulated, using symmetric boundary conditions. Fig. 10(a) shows the simulated rotation of the mirror, one end of which is fixed to the substrate by the hinges, for a temperature rise of 100 C. The mirror angle alters at a rate of C 1. This result is of the same order but slightly smaller than the experimental result. This difference may result from the inaccurate three dimensional hinge shape used in the simulation and the different TCE values of the resists used in the cantilever experiment and the self-assembled mirror structure. For design optimization, simulation shows that if the rectangular holes over the hinge structure are removed, the thermal stability is further improved to C 1. A simulation was also performed for the case of a mirror whose free end is engaged by a mechanical limiter. In this case, the mirror bends rather than rotates (Fig. 10(b)). Therefore, we conclude that the thermal stability of surface tension self-assembled devices is greatly improved by using mechanical limiters, and has sufficient tolerance for most applications. Moreover, we know from the simulation results that the thermal stability of the devices can be further enhanced by optimizing the design (for example, reducing the size of etch access holes near hinges) without changing the materials.

8 388 Y.K. Hong, R.R.A. Syms / Sensors and Actuators A 127 (2006) Fig. 10. Thermal stability simulation results for (a) a mirror attached only by one edge and (b) a mirror equipped with a mechanical limiter. 5. Vibration stability of self-assembled MOEMS The mechanical stability of MOEMS components is another concern for real-world applications. The response to external vibration was therefore evaluated for the components previously shown in Fig. 5. Mirrors with and without a mechanical limiter were assessed. Fig. 11 shows the arrangement for measurement of in-plane dynamics. To provide external excitation, the component was mounted on a small stage attached to a circular piezoelectric diaphragm. In-plane displacements of the mirror were measured using an optical microscope equipped with 20 and 50 objectives, a video camera and an on-screen cursor measurement system. The maximum resolution of this system is 1 m. The piezoelectric transducer showed an approximately linear response with a voltage, and 1.4 m/v displacement was obtained at the frequency of 1 khz. Its frequency response was flat at low frequencies, with a primary resonance at 390 Hz. Above this frequency, the response reduced quickly. However, measurements could still be made at high frequency and measurements were made from 50 Hz

9 Y.K. Hong, R.R.A. Syms / Sensors and Actuators A 127 (2006) Fig. 11. Arrangement for measurement of in-plane dynamics. Fig. 12. Variation of angle with frequency for different mirrors under inertial load. to 10 khz. The maximum input power of the transducer is 150 mw. Fig. 12 shows the measured frequency responses of surface tension self-assembled mirrors. Two un-engaged mirrors (one rotated 90 and the other rotated 45 ), and one 45 mirror engaged with a mechanical limiter were measured. Un-engaged mirrors had resonant frequencies of 2830 and 2863 Hz for mirror angles of 90 and 45, respectively. In case of the mechanically limited mirror, no resonance peak appeared within the frequency range. The acceleration externally applied to the mirror can be calculated as follows. Since the transducer was driven by a sinusoidal voltage, its displacement can be written as y = A sin ωt (8) The acceleration of the transducer is then: ÿ = Aω 2 sin ωt (9) The maximum acceleration applied to the mirrors is then ÿ = Aω 2. Since the response of transducer is 1.4 m/v at 1 khz, the displacement of the transducer at the applied voltage of 20 V p p is 28 m, which corresponds to 2A. The maximum acceleration is therefore m/s 2 or 56.4g. Under this acceleration, none of the mirrors showed any failure. Life cycles of the devices were then measured. Three 90 -rotated un-engaged mirrors were mounted on the transducer, which was driven at 20 V and 2840 Hz, the resonant frequency of the device. The mirror angle was measured every 30 min. The angle of all three mirrors was stable within 0.2 until cycles, and the measurement was stopped. This result can be compared with the performance of other actuators based on polymer materials. Smela et al. reported a life cycle of 10,000 for a device formed from conjugated polymers [24] and Ebefors et al. reported more than cycles for polyimide v-groove joints [25]. The life cycle of the AZ4562 pads appears sufficient compared to other polymer actuators. The AZ4562 photoresist pads therefore Fig. 13. Shape of first mechanical resonant mode for (a) a mirror attached only by one edge and (b) a mirror equipped with a mechanical limiter.

10 390 Y.K. Hong, R.R.A. Syms / Sensors and Actuators A 127 (2006) endured both high external acceleration and internal resonant frequency conditions, demonstrating the mechanical robustness of the hinges. The resonant vibration modes of the mirrors can also be estimated using ANSYS. To do so, we assumed a density of 2330 kg/m 3, Young s modulus of N/m 2 and Poisson s ratio of 0.28 for silicon. For the photoresist hinges, the values obtained in Section 3 were used, as shown in Table 4. Fig. 13 shows the first mechanical resonant modes obtained from the simulation. For a mirror simply hinged by the photoresist pads, the predicted first mechanical resonant frequency is 3054 Hz, as shown in Fig. 13(a). For a mirror with a limiter, both ends of the mirror are fixed. In this case, the first mechanical resonant frequency rises considerably to 24,430 Hz, as shown in Fig. 13(b), which is far above the capability of the transducer. The difference between the measured frequencies and the simulated values is mainly due to the variation of the device s dimension during process. For example, if the thickness of the mirror changes from 3.5 to 3 m, the resonance frequency changes from 3054 to 2629 Hz. 6. Conclusion The thermal mechanical constants of the hinge material (AZ4562 photoresist) used for surface tension self-assembled 3D MOEMS components have been measured. Young s moduli and thermal expansion coefficients showed relatively constant values under various hard bake conditions. The re-melting process window therefore ranges from 120 to 160 C, which is sufficient for surface tension self-assembly. The thermal and mechanical stability of self-assembled MOEMS components has been evaluated and compared to simulation results. The components are surprisingly stable. The angle variation of all the components tested under the thermal stresses was less than C 1 up to 110 C. However, melting of AZ4562 photoresist occurs between 110 and 120 C, which changes the mirror angle significantly for mirrors without mechanical limiters. A limiter enhances the thermal stability to near zero, and the angle appears stable up to 150 C. The finite element simulation result also shows that the stability of the devices can be further improved by optimizing the device design without changing materials. The dynamic mechanical stability of the components has also been evaluated using vibration testing. No failure was observed in any of the devices tested, for external accelerations greater than 50g. The lifetime of devices has been measured to more than cycles of the internal resonance, which is enough for practical applications. Surface tension self-assembled mirrors based on AZ4562 photoresist hinges therefore show relatively stable thermal response and considerable resistance to external mechanical vibrations. References [1] H. Toshiyoshi, H. Fujita, Electrostatic micro torsion mirrors for an optical switch matrix, J. Microelectromech. Syst. 5 (1996) [2] L.Y. Lin, E.L. Goldstein, R.W. Tkach, On the expandability of freespace micromachined optical cross connects, J. Lightwave Technol. 18 (2000) [3] B. McCarthy, V.M. Bright, J.A. Neff, A solder self-assembled torsional micromirror array, in: Proceedings of the 16th Annual International Conference on Micro Electro Mechanical Systems, IEEE, Kyoto, Japan, January 19 23, 2003, pp [4] D.A. Francis, M.H. Kiang, O. Solgaard, K.Y. Lau, R.S. Muller, C.J. Chang-Hasnain, Compact 2D laser beam scanner with fan laser array and Si micromachined microscanner, Electron. Lett. 33 (1997) [5] M.H. Kiang, O. Solgaard, K.Y. Lau, R.S. Muller, Electrostatic combdrive-actuated micromirrors for laser-beam scanning and positioning, J. Microelectromech. Syst. 7 (1998) [6] V. Milanovic, G.A. Matus, T. Cheng, B. Cagdaser, Monolithic high aspect ratio two-axis optical scanners in SOI, in: Proceedings of the 16th Annual International Conference on Micro Electro Mechanical Systems, IEEE, Kyoto, Japan, January 19 23, 2003, pp [7] C. Pu, Z. Zhu, Y.H. Lo, Surface micromachined integrated optical polarization beam splitter, Photon. Tech. Lett. 10 (1998) [8] D.S. Gunawan, L.Y. Lin, K.S.J. Pister, Micromachined corner cube reflectors as a communication link, Sens. Actuators A 46/47 (1995) [9] L. Zhou, J.M. Kahn, K.S.J. Pister, Corner-cube retroreflectors based on structure-assisted assembly for free-space optical communication, J. Microelectromech. Syst. 12 (2003) [10] A. Friedgerger, R.S. Muller, Improved surface-micromachined hinges for fold-out structures, J. Microelectromech. Syst. 7 (1998) [11] A.Q. Liu, X.M. Zhang, V.M. Murukeshan, C. Lu, T.H. Cheng, Micromachined wavelength tunable laser with an extended feedback model, J. Select. Top. Quant. Electron. 8 (2002) [12] Y.W. Yi, C. Liu, Assembly of micro-optical devices using magnetic actuation, Sens. Actuators A 78 (1999) [13] M.J. Daneman, N.C. Tien, O. Solgaard, A.P. Pisano, K.Y. Lau, R.S. Muller, Linear microvibromotor for positioning optical components, J. Microelectromech. Syst. 5 (1996) [14] E.J. Garcia, Micro-flex mirror and instability actuation technique, in: Proceedings of the 11th Annual International Workshop on Micro Electro Mechanical Systems, Heidelberg, Germany, January 25 29, 1998, pp [15] M.S. Rodgers, J.J. Sniegowski, S.L. Miller, G.F. LaVigne, Designing and operating electrostatically driven microengines, in: Proceedings of the 44th International Instrumentation Symposium, Reno, NY, May 3 7, 1998, pp [16] D.M. Tanner, J.A. Walraven, S.S. Mani, S.E. Swanson, Pin-joint effect on the reliability of a polysilicon microengine, in: Proceedings of the 40th Annual Reliability Physics Symposium, April 7 11, 2002, pp [17] B. McCarthy, V.M. Bright, J.A. Neff, A multi-component solder selfassembled micromirror, Sens. Actuators A 103 (2003) [18] R.R.A. Syms, Rotational self-assembly of complex microstructures by the surface tension of glass, Sens. Actuators A 65 (1998) [19] R.R.A. Syms, Surface tension powered self-assembly of 3-D microoptomechanical structures, J. Microelectromech. Syst. 8 (1999) [20] R.R.A. Syms, Self-assembled 3-D silicon microscanners with selfassembled electrostatic drives, Photon. Tech. Lett. 12 (2000) [21] R.R.A. Syms, C. Gormley, S. Blackstone, Improving yield, accuracy and complexity in surface tension self-assembled MOEMS, Sens. Actuators A 88 (2001) [22] R.R.A. Syms, Refractive collimating microlens arrays by surface tension self-assembly, Photon. Tech. Lett. 12 (2000) [23] Y.K. Hong, R.R.A. Syms, K.S.J. Pister, L.X. Zhou, Design, fabrication and test of self-assembled optical corner cube reflectors, J. Micromech. Microeng. 15 (2005) [24] E. Smela, M. Kallenbach, J. Holdenried, Electrochemically driven polypyrrole bilayers for moving and positioning bulk micromachined silicon plates, J. Microelectromech. Syst. 8 (1999)

11 Y.K. Hong, R.R.A. Syms / Sensors and Actuators A 127 (2006) [25] T. Ebefors, J.U. Mattsson, E. Kalvesten, G. Stemme, A robust micro conveyer realized by arrayed polyimide joint actuators, J. Micromech. Microeng. 10 (2000) [26] A. Bosseboeuf, S. Petitgrand, Characterization of the static and dynamic behaviour of MOEMS by optical techniques: status and trends, J. Micromech. Microeng. 13 (2003) S23 S33. [27] R.R.A. Syms, H. Zou, P. Boyle, Mechanical stability of a latching MEMS variable optical attenuator, J. Microelectromech. Syst. 14 (2005) [28] G.W. Dahlmann, E.M. Yeatman, P. Young, I.D. Robertson, S. Lucyszyn, Fabrication, RF characteristics and mechanical stability of selfassembled 3D microwave inductors, Sens. Actuators A 97/98 (2002) [29] K.E. Petersen, C.R. Guarnieri, Young s modulus measurements of thin films using micromechanics, J. Appl. Phys. 50 (1979) [30] P.M. Osterberg, S.D. Senturia, M-TEST: a test chip for MEMS material property measurement using electrostatically actuated test structures, J. Microelectromech. Syst. 6 (1997) [31] W.T. Thomson, M.D. Dahleh, Theory of Vibration with Applications, 5th ed., Prentice-Hall, New Jersey, 1998, pp [32] E. Volterra, E.C. Zachmanoglou, Dynamics of Vibration, Charles Merrill Book Co., Ohio, 1965, p [33] M. Ohring, The Materials Science of Thin Films, Academic Press, Inc., San Diego, 1992, pp [34] G.W.C. Kaye, T.H. Laby, Tables of Physical and Chemical Constants, 16th ed., Longman, Harlow, 1995 (Chapters 2 and 3). Biographies Since then, he has worked at Samsung Semiconductor R&D institution for 8 years. He is currently studying towards a PhD in the optical and semiconductor devices group at Imperial College London. His research interests include optical MEMS devices, surface micromachining and material characterization. Richard R.A. Syms was born in Norfolk, VA, in He received his BA degree in engineering science in 1979, and the DPhil degree (on volume holographic optical elements) in 1982, both from Worcester College, Oxford University, U.K. He has been Head of the Optical and Semiconductor Devices Group in the Department of Electrical and Electronic Engineering, Imperial College London, U.K., since 1992 and professor of Microsystems Technology since Prior to that, he was reader in Electro-optics, senior lecturer and senior research fellow at Imperial College London, and prior to that, Atlas Research Fellow at Oxford University. He currently lectures on guided-wave optics and MEMS at undergraduate and postgraduate level, at Imperial and elsewhere. He has published over 100 journal papers and two books on holography, integrated optics, laser and amplifier devices and microengineering. Most recently, he has been developing electrical MEMS such as microconnectors, RF probes for magnetic resonance imaging, and miniature quadrupole mass spectrometers, optical MEMS devices such as alignment devices, variable optical attenuators and tunable lasers, and three-dimensional self-assembling microstructures. He is Co-Founder and Research Director of the MEMS spin-out company Microsaic Systems since He currently acts as an associate editor for the IEEE/ASME Journal of Microelectromechanical Systems. He is a Fellow of the Institution of Electrical Engineers (IEE-U.K.) and a Fellow of the Institute of Physics. Young Ki Hong was born in Seoul, South Korea, in He received his master degree in physics from Yonsei University, Seoul, South Korea in