DEVELOPMENT OF SURFACE MICROMACHINED MAGNETIC ACTUATORS USING ELECTROPLATED PERMALLOY

Transcription

1 DEVELOPMENT OF SURFACE MICROMACHINED MAGNETIC ACTUATORS USING ELECTROPLATED PERMALLOY CHANG LIU Microelectronics Laboratory University of Illinois at Urbana-Champaign Urbana, IL Mailing address: 313 Microelectronics Laboratory 208 North Wright Street Urbana, IL Phone: (217) Fax: (217) ABSTRACT Results on design, fabrication and testing of silicon micromachined magnetic actuators are presented. Magnetic actuators are capable of providing large force (on the order of mn) and large displacement (micrometer to millimeter range) within micro electromechanical devices and systems. Electroplated Permalloy (Ni 80 Fe 20 ) material is the media for magnetic interaction and force generation. The Permalloy piece is supported by a structural plate, which consists of polycrystalline silicon thin-film. Applications of such magnetic actuators in massive parallel assembly and high-yield surface-structure release are discussed.

2 I. INTRODUCTION Micromachining technology and micro electromechanical systems (MEMS) have been undergoing dynamic development in the past 15 years [1,2]. MEMS offers unique advantages including miniaturization, mass fabrication and monolithic integration with microelectronics. It has enabled successful demonstration of novel sensors, actuators and systems in many diverse application areas such as optics [3], fluid mechanics [19], biomedical engineering, communication and information storage. The characteristic length scale of micromachined devices ranges from micrometer to millimeter. Actuators transform various forms of energy into the domain of mechanical energy. Force, torque, displacement or strain can be generated using a number of energy-transformation mechanisms. These include electrostatic interaction, magnetostatic interaction, fluidic momentum transfer, thermal expansion, thermal-induced phase change and piezoelectric effects. The relative merits of these mechanisms are summarized in the following. (1) Electrostatic interaction is widely used in micromachined sensors and actuators, including accelerometers [4], rotation-rate gyro and micro optical components. Comprehensive understanding toward materials, processing and mechanisms have been established. Implementation of electrostatic mechanisms in MEMS is generally compatible with integrated circuit processes. However, the magnitude of electrostatic forces is known to decrease rapidly as the spacing between electrodes increases. This can be illustrated with the example of a parallel-plate capacitor, which represents a fundamental configuration for electrostatic sensors and actuators. For a parallel-plate capacitor with an overlapping area of A and a distance of d, the magnitude of the attractive electrostatic force is linearly proportional to A and inversely proportional to d 2. As a consequence, an actuator based on the parallel-plate capacitor configuration can not simultaneously satisfy requirements for large force (on the order of mn) and large displacement (on the order of tens of µm). Likewise, most electrostatic-actuation mechanisms with various electrode configurations suffer from the same limitations. (2) Actuation based on thermal processes, on the other hand, typically has the following disadvantages: high level of energy consumption and a slow time response. (3) With regard to piezoelectric actuation, the piezoelectric co-efficient of typical active materials is low and large material dimension is required if a large displacement is required. The compatibility of high-quality piezoelectric films with conventional IC and MEMS processes remains an active and challenging research topic. Magnetic actuation is potentially capable of realizing both large force and large displacement in an energy-efficient manner. This performance advantage is derived from fundamental differences of static magnetic and electrical fields. Many materials (e.g. silicon and silicon dioxide) typically encountered in MEMS have low magnetic susceptibility and do not develop appreciable internal magnetization. These materials are therefore transparent to a static magnetic field; it is feasible to active integrated micro devices using a global, external magnetic field. On the other hand, the electrical susceptibility (x e ) of most materials are greater than zero (e.g. x e of silicon is 10.7); therefore, globally-applied electric field lines usually can not penetrate material layers easily. High electrical field is also known to create material damages (e.g. dielectric breakdown), whereas high magnetic field is often not associated with material defects. Magnetic force is a body force; the magnitude of magnetic interaction can be increased by simply increasing the volume of the magnetic material. Such a process involves minimum process complexity. Processes required for accomplishing the volume expansion is rather straightforward. In contrast, the electrostatic force is a surface force and a large electrode area must be provided if large forces 2

3 are desired. The process for increasing the electrode areas, in both vertical and horizontal configurations, requires more complex processing procedures. Several authors have published results on micromachined magnetic actuators in the past. Wagner et al. [5] manually attached precision-machined permanent magnet pieces on suspended plates. Integrated in-plane coils on the same chip generate an external magnetic field. Due to the relatively large volume of the permanent magnet, large force can be demonstrated. However, such a permanent-magnet piece can only be achieved through precision assembly of discrete units and is not suited for integrated fabrication and packaging. Liu et al. developed an integrated coiltype magnetic actuator capable of achieving out-of-plane vertical displacement of several hundred micrometers and magnetic forces of 10's of µn [6, 7]. A torque is generated via the interaction between an external magnetic field and the magnetic moment of a planar electric-coil. Unfortunately, coil-type actuators typically require large biasing electric current (i50 ma). This current level, coupled with a long wire length and large number of turns, potentially causes significant thermal heating problems. Judy et al. demonstrated in-plane motion of a suspended polycrystalline silicon structure with an electroplated magnetic piece [9]. The plate was driven by an external magnetic field. The actuator achieved large deflection angle (over ) under a torque of approximately nnm. Recently, Miller et al. [8] and Judy et al. [10] have demonstrated Permalloy magnetic actuators capable of individual addressing. In addition to unit actuators, complex electromagnetic sub-systems, including a planar electromagnet [11] and magnetic micromotors [12], have also been developed. In this paper, I will present results on the design, fabrication and testing of surface micromachined magnetic actuators based on electroplated Permalloy materials. Two applications of these actuators will also be reviewed. II. THEORIES OF OPERATION Schematic diagrams of three types of developed actuators are shown in Fig. 1. A common component of these actuators is a thin-film structure plate that supports an electroplated Permalloy piece. The Permalloy piece generates mechanical force and torque when it is placed within a magnetic field. These actuators are distinguished by the nature of their mechanical supports, which are based on cantilever beams (Type-1 actuator, Fig. 1a), torsion beams (Type-2 actuator, Fig. 1b) or mechanical hinges (Type-3 actuator, Fig. 1c&d). The structure plate and support beams all use polycrystalline silicon thin films. The mechanism of actuation is illustrated using the example of a Type-1 actuator (Fig 1a and Fig. 2). Key physical dimensions of the actuator are identified in Fig. 2. When the external magnetic field is zero, the structural plate is parallel to the substrate plane (Fig. 3a). When an external magnetic field, H ext, is applied normal to the plane of the structure plate, a magnetization M s develops within the Permalloy piece and subsequently interacts with H ext (Fig. 3b). The interaction creates a torque (M mag ) and a small force (F), which cause the beam to bend out-ofplane (Fig. 3c). An analysis of the quasi-static characteristics of these actuators is provided in the following two sections. The torque M mag and force F due to magnetic interaction will first be analyzed. The overall displacement of the actuator is then derived from the balance of torque and force. 3

4 II. 1. Torques and forces due to magnetic interaction The magnetization vector (M) within the Permalloy material is influenced both by the externally applied magnetic field and by the angular position of the magnetic piece. The direction of M is governed by the geometric shape of the Permalloy piece. In the actuator shown in Fig. 2, the lateral dimension of the Permalloy piece is much greater compared with the thickness; the direction of M is within the plane and perpendicular to the direction of the thickness. The magnitude of the magnetization can be obtained from the B-H curve (Fig. 7) of the Permalloy material. (For high-permeability magnetic material such as Permalloy, the magnetization, M, and the internal magnetic flux density, B, are almost identical in magnitude.) Since the remnant magnetization is low, the Permalloy piece is considered non-magnetized when the external magnetic field, H, is zero. As H increases, xmx grows linearly until the saturation magnetization M s is reached. Since the permeability of the material is high (µ r i4500), saturation of magnetization occurs at a relatively low H, which is denoted as H k. The current actuator design focuses on the large-displacement regime and does not involve comprehensive modeling of actuator behavior at low field levels (below saturation). When an external bias is applied, the Permalloy material is treated as having a fixed in-plane magnetization with its magnitude equal to the saturation magnetization, M s. The force and torque acting on the Permalloy piece is estimated using models of effective magnetic charges. It is assumed that two magnetic charges of opposite polarities emerge along the upper and lower edges of the Permalloy plate when the material is magnetized. These two charges generate F 1 (acting at the upper edge) and F 2 (acting at the lower edge) (Fig. 3 b). The magnitudes of these two force components are F1 = M s W T H1 (1) F2 = M s W T H 2 where H 1 and H 2 are the magnetic field strengths at the top and bottom edges of the plate (H 2 > H 1 in the current configuration). The magnitudes of H 1 and H 2 are linearly dependent on the respective distance to the surface of the electromagnet core. The structure plate, along with the Permalloy piece, has a thickness of t+t. Its moment of inertia, I, is proportional to (t+t) 3 and is much greater compared with that of the cantilever beam, which has a thickness of t. The combined structure plate and the Permalloy piece is thus considered as a rigid body. Based on this assumption, the force system is simplified by translating F 1 to coincide with F 2. The result is a counter-clockwise torque M mag and a point force F (diagramed in Fig. 3) acting on the bottom edge of the structural plate. These are expressed as M mag = 1L cosθ (2) F = F2 F F1 The torque always tends to minimize the overall energy in an actuator system by aligning the magnetization with the magnetic field lines. II. 2. Displacement analysis The profile of the actuator is solved by coupling the magnetic torque and force to the supports. The analysis for three types of actuators is presented in the following. Type-1 actuator with cantilever-beam supports 4

5 The angular as well as vertical deflections due to M mag and F are solved independently and the results are linearly super-imposed. This simplification is justified because the magnitude of deflection due to F is estimated to be at least one order of magnitude smaller compared with the deflection caused by M mag. Beam displacement under the M mag is solved first. The magnitude of the vertical displacement at the free-end of the beam is much more than the thickness of the beam; linearized, smalldisplacement assumptions are no longer valid. The deflection of the cantilever beam is determined by a universal governing equation [14] that relates the radius of curvature with M mag (Fig. 4) " 1 y M x mag = = (3) 3 r ' 2 2 (1 + y EI x ) Here, x and y are the horizontal and vertical coordinates of a point along the cantilever beam at an arc length of s, Ε andι are the Young s modulus and the moment of inertia of the cantilever beam, respectively. The cantilever beam assumes the shape of a circular arc, with the radius of curvature being r. The angular deflection is the greatest at the free-end of the cantilever beam (s=l) and can be written as θ = l torque r (4) The x and y coordinates at s=l are therefore l x( s= l) = r sin( ) r (5) l y( s= l) = r[1 cos( )] r Beam bending due to the force is solved by applying F at the free end of a pre-curved beam under the influence of M mag. The maximum angular and vertical deflections occur at the end of the cantilever beam. These deflections can be expressed as [14] 2 ( / 2 1) FR θ force = π (6) EI 2 ( 3 / 4 2) FR y force = π (7) EI The overall angular deflection of the beam (at s=l) is obtained by combining of Eq. 4 and Eq. 6, θ( s= l ) = θtorque θ force. (8) The maximum vertical deflection at the end of the rigid structural plate is = y y + L θ. (9) y max ( s= l) force sin ( s= l ) Type-2 actuators with torsion-beam supports The above analysis for a Type-1 actuator considers only the pure bending mode under ideal loading conditions. In the pure bending mode, cantilever beams can withstand 180-degree bending without fracture. However, non-ideal external loading conditions occur frequently in certain applications such as fluid-dynamical control. Non-ideal loading generates non-desirable mode of displacements. One example of an un-desired mode occurring in Type-1 actuators is twisting along an axis that is parallel with cantilever beams. It has been experimentally observed 5

6 that this twisting motion introduces a prevalent failure mode that causes a large percentage of devices to fracture. Type-2 actuators with torsion-beam supports successfully suppress the twisting motion. These are more robust compared with Type-1 actuators. For a Type-2 actuator, the angular displacement is related to the torque by the following expression, 2θ M mag = KG (10) l where h is the angular displacement experienced by each torsion beam, l is the length of each torsion beam and G is the torsion modulus of elasticity of the material. K is a constant determined by the specific cross-sectional geometry of the beams; for a torsion beam with a rectangular cross-section and an area of w % t, a b k = ab [ 3.36 (1 )] (11) 4 3 b 12a where a=w/2 and b=t/2. The force F also creates bending within the support beams that contribute to an out-of-plane translation. However, this displacement is small (due to the fixedfixed beam boundary conditions) and is typically ignored in our analysis. Type-3 actuators with hinge supports There are two variations within Type-3 actuators: one is supported by a plain hinge and another is supported by a hinge and an add-on spring-loading mechanism. A plain hinge does not provide any restoring torque that can balance M mag. A Type-3 actuator with a plain hinge (Fig. 1c) would experience a net magnetic torque until a 90 o displacement is reached, at which point M mag is equal to zero. Such an actuator has only two possible positions, with the angular displacement being either 0 or 90 degrees. However, if the objective of a hinged magnetic actuator is to provide controllable displacement, a cantilever or a torsion beam must be used to provide a counteracting torque (Fig. 1d). A spring-loading mechanism will generate such a torque and allow the hinged actuator to achieve continuously variable positions. The analysis can be pursued in a similar fashion as Type-1 actuators. II. 3. Summary of actuator design A number of geometric parameters are fixed within our design. The rigid structure plate has a fixed area of 1 by 1 mm 2. The thickness of the magnetic material is 5 µm. For Type-1 actuators, the typical length and width of the cantilever beam are 400 and 100 µm, respectively. For Type-2 actuators, the width and length of the torsion beam are 2 and 50 µm, respectively. III. FABRICATION The fabrication process for a typical magnetic actuator is summarized in Fig. 5. The entire process is divided into five steps and discussed accordingly. Step 1 (Fig. 5a) A 3-µm-thick phosphosilicate glass (PSG) thin-film is first deposited on top of the silicon substrate at 450 C o. The thin film functions as a sacrificial material. The PSG layer is patterned using photolithography and then etched using buffered hydrofluoric acid (BHF) to form separated 6

7 mesas on top of which individual actuators will be located. These mesas isolate individual actuators and limit the total amount of lateral dimension expansion that will result from undercut during the sacrificial-layer etching processing. This feature therefore provides robust process control and results in high structural yield even when over-etching is applied. It also increases the potential area density of actuators by allowing actuators to be placed closer to one another. After removal of the photoresist layer, the wafer is annealed in a nitrogen ambient at 1000 C o for one hour. This step serves two purposes. First it activates the phosphorus dopant (6 wt.%) within the PSG layer and increases its etch rate by BHF. Secondly, the PSG material reflows slightly at the temperature of the oxidation, creating rounded, smooth profile along the perimeter of PSG mesas. The wafer is then covered by a highly conformal deposition of thin LPCVD polycrystalline silicon. The improved profile of the mesas translates directly into rounded corners in structural layers. The rounding alleviates stress concentration and enhances the reliability of actuators. A 0.5-µm-thick PSG layer is then deposited on top of the polysilicon. It serves as a complimentary doping source. During a 1-hour, 950 C o stress-relief anneal in nitrogen ambient, the polysilicon is doped symmetrically from both sides. The symmetric doping reduces the intrinsic-stress gradient across the thickness of the polysilicon and minimizes residue beam bending. In contrast, a doping from only one side would generate a non-symmetric doping concentration and a stress gradient. The top PSG layer is later removed by using BHF. Step 2 (Fig. 5b) Before performing the electroplating, an electrically conductive seedlayer must be applied to the front surface of the wafer. The seedlayer contains 200 -thick Cr and thick Cu thin films. Both layers are thermally evaporated. The Cr layer enhances adhesion between the Cu and the polysilicon layers. For seedlayer evaporation, it is critical to maintain electrical continuity throughout the entire wafer. The continuity can potentially be broken because coverage of thermally evaporated metal is not conformal (compared with LPCVD polysilicon) and the deposition thickness on vertical sidewalls is much reduced compared to the nominal value. If an insufficient amount of seedlayer material is deposited, the active region on the top of the mesa could be electrically isolated from the seed-layer region on the bottom. Electroplating will not occur on top of each mesa. The front-surface of the wafer must be protected after the wafer is removed from the vacuum ambient of the evaporation chamber. Copper thin film oxidizes readily in air and the resulted copper oxide layer will hinder the electroplating process. Care must be taken to ensure that the copper layer is not exposed to air and water for long periods of time. In the current development, a 5-µm-thick photoresist layer is immediately applied to the wafer to insulate the copper film. The photoresist is patterned and developed only immediately before electroplating. Patterned photoresist form narrow frames. In regions that are not covered by the photoresist, the seedlayer is exposed and Permalloy (Ni 80 Fe 20 ) electroplating will take place (Fig. 5). Step 3 (Fig. 5c) The recipe and technique for Permalloy electroplating was originally developed in the thin-film read/write-head industry. During the plating process, the wafer is affixed to the cathode and a pure Ni piece serves as the anode. An external biasing magnet (450 Oe) is applied with the field lines being parallel to the wafer substrate. This bias establishes directions of preferred magnetization (easy axis) within the Permalloy piece. Electroplating takes place at a rate of 5 7

8 µm/hour under a bias-current density of 8 to 12 ma/cm 2. Two different plating techniques are available: mold plating and frame plating. In the mode-plating technique, photoresist covers all area of the wafer except where Permalloy is intended. In the frame-plating technique, which is applied in this study, plating occurs over a majority portion of the wafer area. Narrow photoresist frames isolate regions on a wafer. In certain regions, the plated magnetic material is not used and it is selectively after the plating is completely. Frame plating technique allows more uniform electroplating and the plating parameter is not varied when the geometry is changed. Step 4 (Fig. 5d) After electroplating, the wafer is flood-exposed with ultraviolet radiation and the photoresist is removed with a standard photoresist developer. The wafer is further cleaned using acetone and then isopropanal alcohol solutions. Step 5 (Fig. 5e) The seedlayer that is currently exposed will be removed by using Cu etchant (100:5:5 wt. water:acetic acid: hydrogen peroxide) and then a Cr-mask etchant. The Cr-layer removal can be accomplished using either a commercial etchant [17] or diluted HCl (Cr etchant: 10 water: 1 HCL). It typically requires 10 seconds to remove the Cr layer using diluted HCl. Occasionally, Cr etch does not occur spontaneously; namely, the Cr layer stays intact even when the wafers are immersed in Cr etchants for upto 100 times the nominal etching time. This phenomenon is believed to result from non-favorable electro-chemical potential on the wafer. To solve this issue, the electrochemical potential of the seedlayer is modified by contacting the Cr layer with a piece of pure aluminum. Actuators are then released by 49% HF within 20 minutes. To facilitate the sacrificial release process, etch holes (30 µm by 30 µm in area, and 250 µm apart) are opened on the plate. The Permalloy material sustains HF etching without any structural or chemical damage. Since the structure plates have large surface areas and the supporting beams are soft (spring constant i100 µn / 1 mm=0.1 N/m for cantilever beams), these can be easily pulled down by surface tension to the substrate and form permanent bonds [24, 26] if conventional drying techniques are used. To ensure high yield, the structural plate is levitated away from the substrate surface through magnetic interactions. This method effectively prevents the actuators from coming into contact with the substrate, therefore guaranteeing that 100% yield is routinely achieved. More details of the release technique will be reviewed in section V.1. Shown in Fig. 6 are top and perspective views of fabricated actuators, which exhibit no intrinsic bending. Magnetic properties of the Permalloy material The electroplated Permalloy has a composition of 80% Nickel and 20% Iron [16]. Thin film Permalloy is a preferred soft magnetic material for two main reasons: first, the material exhibits a near-zero magnetostriction effect and stress-free films can be realized; secondly, the magnetic switching speed is fast (on the order of femto-second to micro-second). The Permalloy material has a poly-crystalline structure and contains a large number of magnetic domains. Each magnetic domain has to atoms and is spontaneously magnetized in one direction at room temperature. The directions of magnetization of different domains are 8

9 randomly organized. Despite this, there are directions of easy and hard magnetization, a phenomena called crystalline an-isotropy. H k is defined as the magnetic field intensity needed to saturate a soft magnetic material in a specific direction (Fig. 7). In the direction of easy magnetization, the easy axis, H k, easy of the hysteresis is small. In the direction of difficult magnetization, the hard axis, H k, hard is greater. Experimental B-H hysteresis curves along the easy axis and the hard axis are shown in Fig. 7. During the NiFe electroplating, the direction of the biasing magnetic field dictates the orientation of the easy axis [22, 23]; domain walls move in such a manner that domains favorably oriented with the magnetic field grow at the expense of unfavorably oriented domains. (It should be noted that the shape of the Permalloy plate also favors shape-anisotropy with the plate plane.) Therefore, the direction parallel with the external magnetic field during the plating process is the easy axis while the orthogonal in-plane direction is the hard axis. This phenomenon has been studied by Takahashi [22]; it has been found that an external field of Hm30 Oe is sufficient to induce the direction order. Other methods for establishing the easy axis include critical cold working and annealing. The Permalloy material used in the current study show only a minor difference in magnetic behavior between easy and hard axes. The properties of the electroplated magnetic material are summarized in Table 1. IV. TESTING Actuation is quantitatively characterized by using a microscope-monitoring system (Fig. 8). An electro-magnet provides an external magnetic field. The magnetic core has a large cross-sectional area (3%3cm 2 ) and therefore provides uniform field density. The variation of H with respect to the vertical spacing (d) to surface of the magnetic core is calibrated experimentally. In a region near the surface of the core, H is nearly linear with respect to the spacing d; it can be expressed as H=14x %10 4 d (12) The units of d and H are millimeter and A/m, respectively. In their resting positions, actuators are separated from the electromagnet by a distance of 0.5 mm, which is the thickness of the silicon substrate. Fig. 9 contains a sequence of video images showing the position of a Type-1 actuator at the resting position and activated positions. The angular and vertical displacement of the actuator is measured directly from the side profile of the actuator. Fig. 10 shows the measured deflection magnitude, together with theoretical predictions obtained using Eqs. 8 ad 9. Results from theory and experiments match well, especially in the saturation regime. The behavior of magnetic actuators in low field situations is also experimentally characterized (Fig. 11). An actuator is first activated with the magnetic field increasing in one polarity. The polarity is then reversed and the magnitude of the magnetic bias is increased again. The angulardeflection curves for both cases highlight a peculiar bend at the low field level (H below 12x103 A/m). This lowered angular displacement is contributed by the switching of the magnetic domain. The maximum response speed of a magnetic actuator is studied. The time constant is currently not limited by the domain switching process, which dominate the establishment of magnetization and has a typically time constant on the order of one pico-second. Rather, the maximum response speed is limited by the electromagnet that is modulated using mechanical switches. The relatively large size and inductance of the electromagnet increases the time, which has been measured to be between 1 to 10 ms. The time-constant measurement was conducted by using a 9

10 miniature secondary coil placed on the front surface of the core. If higher response speed is intended, magnets with smaller inductance must be developed. To measure the vertical force-loading capacity of actuators under the current experimental conditions, a 5%5 array of actuators is used to hold controlled weights, in the form of precision cut silicon chips. The strongest magnetic field achievable using the current setup configuration is applied to activate the entire array. Silicon chips of known weights are sequentially stacked on top of raised actuators, until the actuator array can no longer hold against the stack weight. The amount of weight that can be held by an array of actuators in their fully raised position is defined as the maximum vertical loading capacity (MVLC). The measured MVLC weighs 222 mg, or 2.18 mn. This translate into roughly a maximum loading force of 87 µn (or 8.88 mg) for each actuator, which has a mass of only 44.5 µg itself. It should be noted that the loading capacity is a function of the angular position. At lower displacement angles, a same weight would produce more torque. The MVLC can not be lifted by the actuators from their rest positions. In pure bending modes, actuators with cantilever-beam and torsion beam supports can achieve more than 180 o displacement without fracture. Fracture strain equals to 0.93% according to an early report by Y.C. Tai et al. [15]. This unique characteristic is due to the reduced thickness of the cantilever beam. V. APPLICATIONS Developed magnetic actuators have been used in magnetic-assisted levitation (for surface structure release), massively parallel assembly of array MEMS, fluid dynamic control [19] and active robotics surfaces. The first two applications are described in detail in sections V.1. and V.2. V.1. Magnetic assisted levitation Successful use of array MEMS devices demands a robust, efficient and high-yield fabrication process. Surface micromachined devices are typically developed using sacrificial-layer etching techniques. Devices are freed by removing the underlying sacrificial layer in a wet chemical ambient (e.g. concentrated HF solution for removing PSG). A drying process follows the wet chemical etching. This step is known to cause significant yield losses due to stiction (sticking and friction). The mechanism for stiction is the following: during the drying process, the liquid on top of a free-standing structure will evaporate first; liquid that is trapped underneath structures remains. The trapped liquid exerts a full-down force due to surface tension. Because surface microstructures are located close to the substrate (with spacing of only several µm) and are typically compliant (with spring constant <1 N/m), the pull-down force is capable of drawing the structure into intimate contact with the substrate. In many cases, this contact produces permanently bonding and irreversible sticking [24-27]. The probability of stiction damage increases with decreasing structural stiffness and increasing surface-contact area. Previously published results on anti-sticion techniques focus on the following methodologies. First, certain post-release chemicals modify the surface-layer composition [26, 28] and prevent sticking by chemical bonds. Secondly, the liquid-vapor phase transformation, which is the cause of the surface-tension force, can be replaced by a freeze-sublimation procedure [28, 31]. Third, microstructures can be kept away from the substrate during release/drying by solid organic polymer columns, which are then removed by plasma dry etching [31-32]. Other novel techniques include using special anti-stiction geometry [27], applying pulsed magnetic forces to 10

11 relieve stuck structures [33], roughening contact surfaces and reducing contact areas [34] and using gas phase etchant for sacrificial-layer removal [35]. In practice, many microstructures have relatively large surface areas (e.g. >1x1 mm 2 ). The probability of stiction-induced damages is higher, as demonstrated by an initial low yield (10%) for Type-1 actuators when no specific drying technique was applied. To achieve high-yield drying for large-area structures, a novel drying process has been developed. During the liquid removal process, freestanding structures are actively levitated out-of-plane; the liquid is removed, after which suspended structures are returned to the substrate plane. This method prevents stiction because levitation force counteracts the surface tension force so microstructures and the substrate will never come into contact. This process requires that an additional patch of Permalloy material be integrated with individual micro devices. Effectiveness of this method is demonstrated using developed Type-1 magnetic actuators. Fabricated test structures (so called dies) were first immersed in HF solution (40%) for 10 minutes to remove the sacrificial layer. These wet dies are then transferred directly to de-ionized (DI) water and immersed for a period of time to completely remove the HF contents. Dies are then immersed in a final rinse solution, which is one of the following chemicals: isopropyl alcohol, methyl alcohol, acetone or water. Wet dies are placed within the magnetic field of an electromagnet and air-dried. An infrared lamp is used to provide heat and accelerate the evaporation process. The result shows that 100% yield is routinely achieved. V.2. Parallel assembly of MEMS MEMS technology is inherently characterized by mass production and three-dimensional structures. Arrayed MEMS devices will no doubtedly offer unique advantages un-available in macro-scopic, conventional systems and singular micro devices. Very-large-scale-integrated circuits (VLSI) exemplifies the tremendous benefits offered by array operation of modular components (e.g. transistors). Nowadays, many MEMS devices are developed based on fundamentally two-dimensional fabrication techniques. In order to realize true 3-D structures, a fabrication process for developing three-dimensional devices from as-fabricated, two-dimensional layers is required. Such a process must offer high yield as well as global (instead of local) addressability. One notable example is a hinged surface microstructure that has been applied for many applications. However, the wide use of this technology depends on an efficient, robust and highyield parallel assembly technology. In this study, Type-3 actuators are used to demonstrate the massive parallel assembly of these hinged surface microstructures. Type-3 actuators with hinges and spring loading provide a means of sequentially levitation, which allows for inter-locking mechanism and assembled microstructures (Fig. 12b). Array devices can be assembled simultaneously by a globally-applied magnetic field. The mechanism is demonstrated in Fig. 12a. Two actuators, Structure-1 and -2, are attached with different spring loading. Both structures are in their rest position when the magnetic field is zero. As the external magnetic field reaches H 1, Structure-1 is fully activated (deflection angle=90 o ) first while Structure-2 has a much smaller angular displacement. Structure-2 is then actuated and locks Structure 1 in place when the external magnetic field is increased to H 2. VI. CONCLUSIONS 11

12 Design, fabrication and testing results of surface micromachined magnetic actuator has been presented. Electroplated Permalloy material is used to provide magnetic interaction and generate force/torque. The magnetic actuators are mechanically supported by flexural cantilever beams, torsion beams or hinges. The advantage of magnetic actuators is to satisfy requirements for large force and large displacement simultaneously. This aspect has been demonstrated in the developed actuators. Applications of magnetic actuators in high-yield release/drying of surface microstructures and parallel assembly of array MEMS has been discussed in detail. VII. ACKNOWLEDGEMENTS The author wishes to thank Prof. Yu-Chong Tai, Prof. Chih-Ming Ho, Thomas Tsao, Dr. Weilong Tang and Dr. Denny K. Miu for their helpful discussions. VIII. REFERENCES 1. H. Fujita, A decade of MEMS and its future, Proc., 10 th IEEE workshop on MEMS, pp. 1-8, Nagoya, Japan, K. Petersen, Silicon as a mechanical material, Proc. IEEE, 70, p , S.S. Lee, E. Motamedi and M.C. Wu, Surface microamchined free-space fiber optics switches with integrated microactuators for optical fiber communication systems, Tech. Digest., 1997 International conference on solid-state sensors and actuators, p. 85, Chicago, IL, W. Clark, R.T. Howe and R. Horowitz, Surface micromachined angular accelerometer with force feedback, Proc., Solid-state sensor and actuator workshop, pp , B. Wagner, W. Benecke, G. Engelmann and J. Simon, Micro actuators with moving magnets for linear, torsional or multi-axial motion, Sensors and Actuators, A(32), pp , C. Liu, T. Tsao, Y.C. Tai and C.M. Ho, Surface micromachined magnetic actuators, Tech. Digest, 1994 IEEE Workshop on Micro-Electro-Mechanical-Systems, pp , T. Tsao, C. Liu, Y.C. Tai and C.M. Ho, Micromachined Magnetic Actuator for Active Fluid Control, Application of Microfabrication to Fluid Mechanics, FED-Vol. 197, pp , R. A. Miller, Y.C. Tai, G. Xu, J. Bartha and F. Lin, An electromagnetic MEMS 2x2 fiber optic bypass switch, Proc., 1997 International Conference on Solid-state sensors and actuators, Chicago, IL, Vol. 1, pp , J.W. Judy, R.S. Muller and H.H. Zappe, Magnetic microactuation of polysilicon flexure structures, Journal of MEMS, Vol. 4, pp , J. Judy and R.S. Muller, Magnetically Actuated, addressable microstructures, Journal of MEMS, Vol. 6, No. 3, p. 257, C.H. Ahn and M.G. Allen, A fully integrated surface micromachined magnetic microactuator with a multilevel meander magnetic core, Journal of MEMS, Vol. 2(1), pp , H. Guckel. T.R. Christenson, K.J. Skrobis, T.S. Jung, J. Klein, K.V. Hartojo and I. Widjaja, A first functional current excited planar rotational magnetic micromotor, Tech. Digest, 1993 IEEE Workshop on Micro Electro Mechanical-Systems, pp. 7-11, W.C. Young, Roark's Formulas for Stress and Strain, 6th Ed., McGraw-Hill, R. Frisch-Fay, Flexible Bars, Butterworth,

13 15. Y.C. Tai, R. S. Muller, Fracture strain of LPCVD polysilicon, Technical Digest, IEEE Solid-State Sensors and Actuators Workshop, Hilton Head Island, SC, USA, pp , V. Temesvary, S. Wu, W.H. Hsieh, Y.C. Tai and D.K. Miu, Design, fabrication and testing of micromachined electromagnetic microactuators for rigid disk drives, Journal of MEMS, Vol. 4, No.1, pp , Cr mask etchant, Transene Co., USA. 18. H. Guckel, J.J. Sniegowaki, T.R. Christenson, The application of fine-grained, tensile polysilicon to mechanically resonant transducers, Sensors and Actuators, Vol. 20, pp , C.M. Ho. and Y.C. Tai, MEMS and Its Applications for Flow Control, Journal of Fluid Engineering, Sep R.T. Howe and R.S. Muller, Polycrystalline Silicon Micromechanical Beams, J. Electrochemical Soc., Vol. 130, pp , C. Liu, T. Tsao and Y.C. Tai, A high yield drying/release process for surface micromachined structures, Proc., 1997 International Conference on solid-state sensors and actuators, pp , Vol. 1, Chicago, IL, M. Takahashi, Induced magnetic anisotropy of evaporated films formed in a magnetic field, Journal of Applied Physics, Supplement to Vol. 33, No. 3, pp , S. Abel, H. Freimuth, H. Lehr and H. Mensinger, Defined crystal orientation of Nickel by controlled microelectroplatings, J. Micromechanics and Microengineering, Vol. 4, pp , C.H. Mastrangelo and C.H. Hsu, A Simple Experimental Technique for the Measurement of the Work of Adhesion of Microstructures, Proc., IEEE Solid-State Sensor and Actuator Workshop, Hilton Head Island, SC, pp , R. Legtenberg, J. Elders and M. Elwenspoek, Stiction of Surface Micromachined Structures After Rinsing and Drying: Model and Investigation of Adhesion Mechanism, Proc., the 1993 International Conference on Solid-State Sensors and Actuators, Transducer'93, pp , R.L. Alley, G.J. Cuan, R.T. Howe, et. al., The Effect of Release-Etch Processing on Surface Microstructure Stiction, Proc., IEEE Solid-State Sensor and Actuator Workshop, pp , T. Abe, W.C. Messner and M.L. Reed, Effective Methods to Prevent Stiction During Post- Release-Etch Processing, Proc., IEEE Workshop on Micro Electro Mechanical Systems 1995, MEMS'95, pp G.T. Mulhern, D.S. Soane and R.T. Howe, Supercritical Carbon Dioxide Drying of Microstructures, Proc., the 1993 International Conference on Solid-State Sensors and Actuators, Transducer '93, p. 296, N. Takeshimo, et. al., Electrostatic parallelogram Actuators, Proc., the 1991 International Conference on Solid-State Sensors and Actuators, Transducers '91, pp , M.R. Houston, R. Maboudian and R.T. Howe, Ammonium Fluoride Antistiction Treatments for Polysilicon Microstructures, Proc., the 1995 International Conference on Solid-State Sensors and Actuators, Transducer'95, Vol. I, pp , Stockholm, Sweden, C.H. Mastrangelo and G.S. Saloka, A Dry-Release Method Based on Polymer Columns for Microstructure Fabrication, Proc., IEEE Workshop on Micro Electro Mechanical Systems, MEMS'93, pp , M. Orpana and A.O. Korhonen, Control of Residual Stress of Polysilicon Thin Films by Heavy Doping in Surface Micromachining, Proc., the 1991 International Conference on Solid-State Sensors and Actuators, Transducers '91, pp ,

14 33. B.P. Gogoi and C.H. Mastrangelo, Post-processing Release of Microstructures By Electromagnetic Pulses, Proc., the 1991 International Conference on Solid-State Sensors and Actuators, Transducer'95, Vol. I, pp , Y. Yee, K. Chun and J.D. Lee, Polysilicon Surface Modification Technique to Reduce Sticking of Microstructures, Proc., the 1993 International Conference on Solid-State Sensors and Actuators, Transducer '95, p. 206, J.H. Lee, Y.I. Lee, W.I. Jang, et al., Gas Phase Etching of Sacrificial Oxides Using Anhydrous HF and CH3OH, IEEE Workshop on Micro Electro Mechanical Systems, MEMS 97, Nagoya, Japan, pp ,