Electrostatic Actuators with Intrinsic Stress Gradient

Transcription

1 /2002/149 8 /H139/7/$7.00 The Electrochemical Society, Inc. Electrostatic Actuators with Intrinsic Stress Gradient I. Materials and Structures Anil K. Chinthakindi, a, * Dhananjay Bhusari, a Brian P. Dusch, a Jürgen Musolf, b Balam A. Willemsen, b Eric Prophet, b Mark Roberson, c and Paul A. Kohl a, **,z a School of Chemical Engineering, Georgia Institute of Technology, Atlanta, Georgia , USA b Superconductor Technologies, Incorporated, Santa Barbara, California , USA c MCNC, Research Triangle Park, North Carolina 27709, USA H139 Electrostatic actuators based on microelectromechanical systems MEMS have many attractive features for use as variable capacitors in high-frequency applications. The devices consist of two electrodes, one fixed and the other movable. In this study, a curved, cantilever beam was used as the movable electrode. A novel process has been developed for fabricating an all-gold, curved beam. The cantilever beams were curved due to an intrinsic stress gradient in the metal. Electroplating and conventional lithography were used to metallize the cantilever beam electrodes. The internal stress gradient in the gold was obtained by changing the electroplating conditions during fabrication. Stiction during release and operation of the variable capacitors was alleviated by treating the gold with an alkane thiol self-assembled monolayer. The intrinsic stress gradient and the stress-induced bending moment were calculated using a generalized model for the stress gradient in the films. Compared to bimetallic, cantilever beams, the curvature of the all-gold beam was found to be independent of temperature. This implies that the operation of the single-metal variable capacitor will be more reproducible and stable with temperature than a comparable bimetallic device The Electrochemical Society. DOI: / All rights reserved. Manuscript submitted July 19, 2001; revised manuscript received February 10, Available electronically June 14, Electrostatically actuated beams are one of the fundamental building blocks in microelectromechanical systems MEMS and find applications in a variety of fields. They are used in a number of areas such as communications, sensing, optics, microfluidics, and measurement of materials properties. 1-4 The principle of electrostatic actuation is used in the fabrication of optical switches for optical networking circuits. 3 In the field of communications, radio frequency rf switches and voltage-controlled oscillators VCOs using electrostatic actuation form an integral part of rf filter circuits. 5 The electrostatically actuated rf capacitors provide greater isolation and more linear response compared to solid-state varactor diodes. The MEMS rf switches and variable capacitors also have higher quality factors and provide higher on/off capacitance ratio compared to traditional p-n junction varactors. A MEMS variable capacitor with a movable electrode and a fixed electrode is studied here. A curved cantilever beam design is considered for the movable electrode. A schematic diagram of the electrostatic actuator is shown in Fig. 1. The change in position of the movable electrode upon electrostatic actuation changes the capacitance of the device. The initial position of the movable electrode establishes the low or off capacitance value of the device. A common approach to setting the initial curvature of the movable electrode is to use two metals with different coefficient of thermal expansion CTE. There are a number of material issues in the use of bimetal films, such as inter-metallic compound formation and recrystallization of the metal films during processing. The mechanical properties of bimetallic films have been shown to change with time and temperature aging due to intermetallic diffusion and recrystallization. 6,7 Intermetallic diffusion between the two films hinders the long-term reliability and reproducible behavior of these devices. Other material properties that must be considered in the choice of metals in bimetallic beams are ductility, resistance to fatigue or work hardening, electrical conductivity, the temperature of alloy or compound formation, and the ease of processing. During the operation of bimetallic actuators, temperature has to be controlled precisely, as any variation in temperature leads to a change in the deflection of the actuator, which in turn changes the capacitance of the device. In addition, the mechanical and electrical * Electrochemical Society Student Member. ** Electrochemical Society Active Member. z paul.kohl@che.gatech.edu properties, such as resistive loss and the electrical conductivity of the movable electrode, can be compromised due to the presence of a different material in the bimetallic film stack. The initial curvature of the movable electrode can also be obtained by using a single material, stress gradient approach. Deposited films with intrinsic residual stresses have been used previously to create nonplanar beams. 4 High growth rates and low surface mobility of the added material introduces strain in the film. 2,4 Polysilicon beams using the Multi-User MEMS Processes MUMPs have been reported to have residual compressive stress and stress gradients leading to deflection of the beam. 8 The residual stress within the cantilever beam was relieved upon release, resulting in an initial deflection of the beam. Conventional metal deposition techniques such as dc magnetron sputtering have also been used to develop novel chip to module compliant interconnections. 9 The bias voltage and deposition pressure were varied during the sputtering process to obtain controlled deflection of the cantilever beams for the fabrication of compliant interconnections. 9 The single material, stress gradient approach for fabricating the movable electrode has considerable advantages over the conventional bimetallic film approach. Precisely controlled stress gradients in the films can be used to achieve a desired level of deflection in the devices. The deflection of the movable electrode deposited with an intrinsic stress gradient is expected to be largely independent of time and temperature. The mechanical and electrical properties of the material are not compromised, as this approach uses a single material. Furthermore, the material issues such as intermetallic reaction between the two materials are also avoided. The long-term reliability and reproducibility of the device could be greatly improved using this approach. The goal of this study is to select a materials set, which is highly reproducible and stable, and to develop a simple process technique to obtain a controllable stress gradient in the movable electrode. Gold has been considered as the material for the movable electrode due to its high electrical conductivity, low loss, ease of deposition, and excellent flexibility. The fixed electrode was made of aluminum. The deposition conditions during metallization were varied to investigate the effect on deflection of the movable electrode. The devices were then subjected to time and temperature aging. Finally, using the generalized model for the stress gradient in the films, the intrinsic stress gradient and the stress-induced bending moment were calculated.

2 H140 Journal of The Electrochemical Society, H139-H The uniform stress, ( 1 2 )/2 accounts for the in-plane elongation of the beam whereas, the ( 1 2 )/b (y) the stress gradient component causes out-of-plane deflection. The net force on the element of area w(dy) is given by Eq. 2 Figure 1. Schematic diagram of the electrostatic actuator x direction is in-plane length of beam, y direction is in-plane width, and z direction is through-plane height. Generalized Model for Calculation of Internal Stress Gradient in the Film A cantilever beam with dimensions of length, L, width, w, and thickness, b, is used in the derivation of the initial deflection of the beam with an intrinsic stress gradient. The beam is assumed to be fully clamped at one end and free at the other end. The stress in the top and bottom layer of the beam is assumed to be 1 and 2, respectively. It is assumed that the in-plane deformation of the beam is negligible compared to its out-of-plane deformation. Assuming a linear stress profile within the beam, the stress can be divided into two components, as given in Eq. 1 Stress b y 1 df 1 2 b yw dy 2 The moment about the axis through the center of the plane is given by Eq. 3 dm df y 1 2 b yw dy y 3 The net moment acting on the beam cross section can be obtained by integrating Eq. 3 over the thickness of the beam, as shown by Eq. 4 b/2 b/2 dm 1 2 b/2 b/2 b y 2 w dy 4 Using symmetry within the beam cross section, the net moment acting on the beam, M is given by Eq. 5 M wb 2 5 Figure 2. Generic process sequence for the fabrication of electrostatic actuators.

3 H141 where z is the deflection of the cantilever beam, E is elastic modulus of the material, and I is the moment of inertia (wb 3 /12. For the beam fixed at one end, the boundary conditions used are x 0, z 0, and x 0, (dz/dx) 0. Solving the differential equation, Eq. 6 with these boundary conditions, and solving for the deflection, z gives Eq. 7 z M x 2 EI b2 w x 2 EI b2 w x 2 EI 2 7 Figure 3. SEM image of a cantilever beam deposited using electron-beam evaporator at 0.8 nm/s. For a cantilever beam fixed at one end, with moment M acting on the other end, the governing equation for static deflection from continuum mechanics is given by Eq. 6 EI d2 z M 6 2 dx Experimental A generic process sequence for the fabrication of the electrostatic actuators is given in Fig. 2 steps 1-9. A n-si 100 wafer with 400 nm of thermally grown oxide layer was used as the base substrate. Aluminum was deposited 0.4 m thick using dc magnetron sputterer CVC products, step 1 in Fig. 2. The Al film was patterned to form the bottom electrode and a 1.6 m thick dielectric polymer was used to isolate the Al from the movable, top electrode, steps 2 and 3, respectively. The polymer film was patterned to form the via holes for contact to the top electrode. Photoresist Shipley Chemical Co series was spun on the wafer and then patterned to expose the regions for the via contact or cantilever beam anchor, step 4. Titanium 30 nm thick, followed by gold 1 m thick were deposited using electron-beam evaporation CVC products at a rate of 0.3 nm/s. This gold layer was patterned to form the cantilever beam anchor regions, step 5. Potassium iodide etchant potassium iodide 100 g/l and iodine 25 g/l was used to etch the gold layer. For etching the bottom titanium layer, EDTA etchant 0.1 M ethylenediaminetetraacetate, disodium salt dihydrate solution, adjusted to a ph of 11 with 5% H 2 O 2 was used. The photoresist was stripped using acetone. The Shipley 1800 series photoresist was spun at 3000 rpm to form a 1.5 m film and patterned to expose the beam anchor regions. This photoresist served as the release layer for the cantilever, step 6. Three different deposition techniques, namely, evaporation, sputtering, and electroplating, were investigated to form the movable, actuator electrode in an effort to control the residual stress in the film. Figure 4. SEM images of the full hinged cantilever beams mm deposited using electroplating a without deposition of 0.3 m of hard gold and b with deposition of 0.3 m of hard gold.

4 H142 Journal of The Electrochemical Society, H139-H Figure 5. Figure showing the intrinsically stressed cantilever beams with 2.3 m of soft gold and 0.2 m of hard gold after release. In the first approach, an electron-beam evaporator was used to deposit 20 nm of titanium followed by 2.5 m of gold. The deposition rate of gold was varied from 0.1 to 1 nm/s from run to run. In the second approach, gold films were deposited using dc sputterer at 300 W power. The sputtering pressure was varied between 0.8 and 2.4 Pa in order to introduce different magnitudes of intrinsic stress in the films. In the third approach, titanium and gold 20 and 200 nm, respectively were deposited at a rate of 0.2 nm/s using e-beam evaporation to form the seed layer for electroplating. The electroplating bath composition, current density, and temperature of gold electrodeposition were varied to investigate the effect of residual stress in the films. Buffered phosphate gold plating bath KAu(CN) 2 20 g/l, K 2 HPO 4 40 g/l, KH 2 PO 4 10 g/l adjusted to aphof7 was used to deposit soft gold at current densities varying from2to10ma/cm 2. Electroplating of the soft gold was done at 60 C. The deposited soft gold has been previously found to have low residual stress. 10,11 An acid plating bath KAu(CN) 2 15 g/l, citric acid 50 g/l, cobalt added as acetate 0.07 g/l, adjusted to a ph of 3.5 was used to deposit hard gold films over the soft Au films at room temperature at a current density of 5 ma/cm 2. The two deposited gold films hard and soft gold exhibit different physical properties hardness, wear resistance, brightness, and residual stress. 10,12 The cobalt-doped gold films were harder and had higher residual stress due to small grain size. 13 After metal deposition, the gold and titanium films were patterned to form the cantilever beam. The photoresist release layer was removed using acetone. The titanium adhesion layer below the evaporated gold was then etched using the EDTA solution. The released cantilever structures were rinsed using isopropanol and methanol, and dried in the nitrogen-purged oven at 90 C. Several methods were used to mitigate the effects of stiction and release the cantilever tips. After dissolving the sacrificial photoresist layer in the acetone, the cantilever beams were rinsed in methanol. If the beams were immediately dried, they experienced stiction due to the surface tension of the liquid. The surface tension of the draining liquid methanol draws the microstructure into contact with the underlying substrate. 14 It has been reported in the literature that the strong adhesion for large surface area MEMS structures is generally caused by capillary, electrostatic, and van der Waals forces and in some cases, by chemical forces such as hydrogen bonding and solid bridging. Rinsing with solvents, such as isopropanol and methanol, was not adequate in releasing the cantilever beams. Self-assembled monolayers have been reported 15,16 to create a hydrophobic surface and eliminate the in-use stiction problem. Selfassembled monolayer SAM was formed by soaking the released beams in a 1 mm dodecyl thiol solution. Contact angle measurements of water on an untreated gold surface was less than 80, whereas the contact angle of water on SAM-treated gold surface was 110. The increase in contact angle indicates that the SAM-treated gold surface is hydrophobic in nature. Results and Discussion The goal of this study was to develop a deposition process for fabrication of cantilever beams wherein the intrinsic residual stress can be controlled. The precise control of the deposition parameters and reproducibility of the processes are the governing factors. As described in the Experimental section, electron-beam evaporation, dc sputtering, and electroplating were investigated to metallize the cantilever beam. Au films were deposited by electron-beam evaporation at deposition rates between 0.1 and 0.5 nm/s. The films deposited at low deposition rates did not contain an adequate stress gradient and therefore the beams did not curl. The films deposited at high deposition rates nm/s, on the other hand, possessed high intrinsic

5 H143 Figure 6. SEM images and tip angles of the full hinged cantilever beams deposited using sequential electroplating of 2.3 m of soft gold and varying thickness m of hard gold. stress and the cantilever beams exhibited a significant amount of curl. Figure 3 shows the scanning electron microscopy SEM image of the released beam deposited at 0.8 nm/s. However, the magnitude and direction of the stress i.e., tensile or compressive, was not consistent from run to run, or within each sample. The stress incorporated in the beam could not be easily controlled using the electron-beam current. Hence, this process was not reproducible enough for fabrication of gold microbeams with controllable deflection. Au films were deposited to thicknesses between 2 and 3 m ina dc sputtering chamber at 300 W power. The sputtering pressure was varied between 0.8 and 2.4 Pa in order to introduce different levels of intrinsic stress in the films. Films deposited at 0.8, 1.6, and 2.4 Pa exhibited compressive stress gradient. It was observed that the magnitude of the compressive stress gradient was greater at low sputtering pressure. Nevertheless, since none of the films exhibited tensile stress gradients, this process was unsuitable for fabrication of Au microbeams under the conditions tested in this study. The gold films electrodeposited from a buffered cyanide plating bath are known to possess intrinsic stress which is dependent on the bath temperature and the current density. 10,11 Increasing the current density or lowering the temperature increases the hardness and stress in the deposited films due to the smaller grain sizes. 17 Cantilever beams were fabricated by electroplating at current densities between 2 and 10 ma/cm 2 at room temperature. An SEM image of the cantilever beam deposited using soft gold plating bath is shown in Fig. 4a. The cantilever beams did not show deflection after release, indicating the lack of an adequate intrinsic stress gradient in the beams. The electroplating of gold cantilever beams was done in a two-step process, soft gold followed by hard gold. 2.5 m of soft, stress-free gold was deposited using the buffered cyanide plating bath at current density of 3 ma/cm 2 at 60 C. Then, 0.3 m of hard gold was deposited over the soft gold using the acid cyanide plating bath at room temperature and at a current density of 5 ma/cm 2. Figure 4b shows the SEM image of the curled cantilever after the deposition of hard gold. Figure 5a-d shows four different shapes of cantilever beams fabricated by the previously described two-step plating process. The soft gold thickness was 2.3 m and the hard gold was 0.2 m. Figure 4a shows the SEM image of a gold cantilever beam with two anchor points at the bottom of the image. Figure 5b is the image of double-anchored cantilever with an elliptical edge to the beam. Figure 5c and d shows the double-anchored cantilever beams with a square edge. During release, the beams were treated with the alkanethiol solution resulting in formation of self-assembled monolayers, thereby reducing the stiction. It can be seen that the beams deposited at these conditions produced spatially uniform deflections. Figure 6a-d shows the SEM images of the full hinged cantilever beams. The soft gold was 2.2 m thick and the hard gold thickness was 0, 0.1, 0.2, and 0.3 m in Fig. 6a-d, respectively deposited. The tip angle for each case was measured using SEM. The tip angle of the cantilever beam gives a direct measure of the deflection of the cantilever beam. It can be observed that the beams with only soft gold Fig. 6a did not show any deflection. This indicates that the residual stress in the films deposited from a buffered cyanide bath was not sufficient to overcome the van der Waals and surface forces acting on the cantilever beam. Figure 6b shows the SEM image of a 1 mm long cantilever beam with hard gold thickness of 0.1 m, showing a tip angle of 18. The cantilever beam upon release has a radius of curvature of 3.2 mm. Upon deposition of 0.2 m of hard gold, the tip angle of the cantilever beam increased to 35 Fig. 6c and the radius of curvature decreased to 1.64 mm. Finally, the tip angle of the beam with 0.3 m of hard gold was 72 and the radius of curvature of the cantilever beam was 795 m Fig. 6d. From the

6 H144 Journal of The Electrochemical Society, H139-H Figure 7. Stress-induced bending moment N m, calculated from Eq. 5 using the measured tip deflection as a function of thickness of hard gold m. tip angle values, it can be seen that the initial deflection of the cantilever beam is directly proportional to the thickness of the hard gold. The stress-induced bending moment N m and residual stress gradient in the deposited films were calculated using Eq. 5 and 7, respectively. Figure 7 shows the estimated values of the stressinduced bending moment for the full-hinged beams as a function of hard gold thickness in the beams. The bending moment for the beams deposited with 0.1 m of hard gold is N m. The induced bending moment gives a measure of the amount of out-ofplane force exerted on the beam due to the deposition of the hard gold. It should be noted that the bending moment acting on the beam is linearly proportional to the amount of hard gold deposited on the soft gold. Figure 8 shows the calculated value of residual stress gradient in the composite film. The effective residual stress gradient in the film deposited with 0.1 m of hard gold is 15.5 MPa/ m. It can also be seen that the stress gradient linearly increases with the amount of hard gold deposited. Assuming that the residual stress in the soft gold is zero and the modulus of the film stack is dominated by the modulus of soft gold soft gold thickness is 2.3 m compared to hard gold thickness of 0.2 m, Stoney s equation can be used to calculate the stress in the hard gold. 17 A cantilever beam of 1000 m in length provided a tip deflection of 366 m corresponding to radius of curvature of 1360 m. Using these values, the calculated value of residual stress in the hard gold was 220 MPa. Previously, it was reported that the residual stress in hard gold deposited using the acid cyanide plating bath was 210 MPa. 13,18 The calculated values of stress are within 5% of the reported values of the stress. The bulk elastic modulus of gold of 74 GPa 19 was used in these calculations. Table I. Results of temperature annealing at 100 C for 48 h on gold microbeams. Average tip deflection, m L w mm Before annealing After annealing Figure 8. Stress gradient MPa/ m, calculated from Eq. 7 using the measured tip deflection as a function of thickness of hard gold m. The effect of aging on the deflection of the electrostatic actuators was also studied. The effect of aging can be accelerated by subjecting the actuators to annealing temperatures for extended periods of time. The effect of the annealing temperature on the released cantilever beams was tested by measuring the tip deflection of the cantilever beams before and after annealing. The sample was placed in a vacuum oven at 100 C for 2 days. The tip deflection of the cantilever beams was measured using an optical microscope. The effect of annealing temperature is given in Table I. From the average values of tip deflection before and after annealing, it can be concluded that there is approximately a 10% increase in the deflection of the beams. Recrystallization of gold in the beams and plastic deformation of the gold at the anchor region due to CTE mismatch 6 between the gold and silicon CTE of gold 14.7 ppm/ C, CTE of silicon 2.7 ppm/ C accounts for the change in tip deflection of the cantilever beams. The error in the measurement of the tip deflection using this technique was 5 m. The deflection of the movable electrode was measured as a function of temperature. The room temperature tip deflection of the cantilever beam of 500 m in length was 120 m. There was negligible change in tip deflection when the temperature was lowered to 50 C and when the temperature was raised to 125 C. The error in measurement was 5 m. The single metal film, stress gradient approach provides considerable advantages over the conventional bimetallic film approach. 20 Using this approach, the deflection of the movable electrode can be controlled precisely. The initial deflection of the movable electrode was found to be nonvariant with time and temperature aging. Since this deposition technique involves a single metal film, the long-term reliability problems due to intermetallic diffusion are avoided. The deflection of the movable electrode is independent of the temperature, and hence, a precise control of temperature during operation is not required. Furthermore, since the residual stress is incorporated in the metal using electroplating, this technique can be extended to other metals that can be deposited using electroplating. Conclusion The possibility of using beams formed of single metal with a uniform and controlled gradient in residual stress incorporated during deposition was explored. Au films deposited by electron-beam evaporation did not show reproducible curl. Neither the magnitude of stress nor its gradient could be controlled in the E-beam process. dc sputtered Au films exhibited compressive stress rather than ten-

7 H145 sile and were found to be unsuitable. Electrochemically deposited pure Au films soft Au were stress free, whereas those hardened with 0.5% Co hard Au contained significant tensile stress. The sequential electrodeposition of soft and hard Au films was found to provide adequate control over the tip deflection of the beams. 0.2 m thick hard gold deposited on top of 2.2 m thick pure Au film, both electrochemically deposited, was found to produce the desired magnitude of curl uniformly across the whole wafer and was found to be reproducible from run to run. The stress gradient incorporated in the film was found to be linear with the thickness of hard gold. The induced bending moment acting on the beam was found to be linear with the thickness of hard gold. Temperature annealing at 100 C of the cantilever beams resulted in approximately 10% change in the tip deflection. Finally, the initial position of the movable electrode was found to be independent of the temperature. Acknowledgments This material is based upon work supported by the Defense Advanced Research Projects Agency, Defense Sciences Office, DARPA order no. J607 Totally Agile RF Sensor Systems TASS, issued by DARPA/CMD under contract no. MDA C Georgia Institute of Technology assisted in meeting the publication costs of this article. References 1. C. Goldsmith, J. Randall, S. Eshelman, T. H. Lin, D. Denniston, S. Chen, and B. Norvell, IEEE MTT-S Int. Microwave Symp., Vol.2,p J. D. Zook, D. W. Burns, H. Guckel, J. J. Sniegowski, R. L. Engelstad, and Z. Feng, Sens. Actuators A, 35, C. T. C. Nguyen, in Proceedings of the IEEE Ultrasonics Symposium, p W. Fang and J. A. Wickert, J. Micromech. Microeng., 6, A. Dec and K. Suyama, IEEE Trans. Microwave Theory Tech., 46, K. Chinthakindi, D. M. Bhusari, B. P. Dusch, and P. A. Kohl, J. Electron. Mater., In press. 7. T. C. Hodge, S. A. Bidstrup-Allen, and P. A. Kohl, IEEE Trans. Compon., Packag. Manuf. Technol., Part A, 20, Y. Min and Y. Kim, Sens. Actuators A, 78, D. L. Smith and A. S. Alimonda, IEEE 46th Electronic Components and Technology Conference Proceedings, p M. Schlesinger and M. Paunovic, Modern Electroplating, 4th ed., John Wiley & Sons, New York W. Chu, M. L. Schattenburg, and H. I. Smith, Microelectron. Eng., 17, H. Angerer and N. Ibl, J. Appl. Electrochem., 9, W. H. Cleghorn, J. A. Crossley, K. J. Lodge, and K. S. A. Gnanasekaran, Trans. IMF 50, R. Maboudian and R. T. Howe, J. Vac. Sci. Technol. B, 15, R. Maboudian, W. R. Ashurst, and C. Carraro, Sens. Actuators A, 82, P. E. Laibinis, G. M. Whitesides, D. L. Allara, Y. Tao, A. N. Parikh, and R. G. Nuzzo, J. Am. Chem. Soc., 113, G. G. Stoney, Proc. R. Soc. London, Ser. A, A82, C. C. Lo, J. A. Augis, and M. R. Pinnel, J. Appl. Phys., 50, Center for Information and Numerical Data Analysis and Synthesis CINDAS, Microelectronic Material Database, Purdue University, West Lafayette, IN. 20. M. Hoffmann, P. Kopka, and E. Voges, Sens. Actuators A, 78,