Muller, R.S. “Microdynamic Systems in the Silicon Age” Handbook of Micro/Nanotribology. Ed. Bharat Bhushan Boca Raton: CRC Press LLC, 1999 © 1999 by CRC Press LLC © 1999 by CRC Press LLC 13 Microdynamic Systems in the Silicon Age Richard S. Muller 13.1 Introduction Origins 13.2 Micromachining Substrate Micromachining • Surface Micromachining • Polycrystalline Silicon Properties • Tribology in MEMS 13.3 MEMS Structures and Systems MEMS Actuation Forces • MEMS for Microphotonics • Coupling Efficiency 13.4 Berkeley Microphotonics Research Acknowledgments References ABSTRACT With the proven record of accomplishment in very large integrated circuits (VLSI) brought about by batch-fabrication technology for electronic devices of ever-decreasing size, there is widespread enthu- siasm about emerging opportunities to design mixed micromechanical and microelectronic systems. New development, based heavily on integrated-circuit-related technologies, have led to rapid progress in the development of microdynamic systems . These systems are based upon the science, technology, and design of moving micromechanical devices, and are a subclass of what is known in the U.S. as microelectromechanical systems (MEMS). The dimensions of the micromechanical devices in MEMS are typically smaller than 100 µm. Recent rapid progress gives promise for new designs of integrated sensors, actuators, and other devices that can be combined with on-chip microcircuits to make possible high-performing, compact, portable, low-cost engineering systems. Mechanical materials for some of the microdynamic systems that have thus far been demonstrated consist of deposited thin films of polycrystalline silicon, silicon nitride, aluminum, polyimide, and tungsten among other materials. To make mechanical elements using thin-film processing, microstructures are freed from the substrate by etching a sacrificial layer of silicon dioxide. First demonstrated as a means to produce electrostatically driven, doubly supported beam bridges, this sacrificial-layer technology has proved very versatile and has been used to make, among other structures, laterally vibrating doubly folded bridges, gears, springs, and impacting microvibromotors. Recently, micromirrors that consist of multiple hinged plates which Updated, expanded, and revised version of article first published in Micro/Nanotribology and Its Applications (B. Bhushan, ed.), NATO ASI Series, Kluwer Academic Publishers, Dordrecht, 1997. © 1999 by CRC Press LLC fold out of the surface plane in which they are formed (reaching vertical heights of millimeter dimen- sions) have been demonstrated. The polycrystalline silicon cross section for these mirrors is thinner than 2 µm. The mirrors can be moved using electrostatic comb drives or microvibromotors. Continued research on the mechanical properties of the electrical materials forming microdynamic structures (which previously had exclusively electrical uses), on the scaling of mechanical design, on tribological effects, on coatings, and on the effective uses of computer aids is now under way. This research promises to provide the engineering base that will exploit this promising technology. 13.1 Introduction Microdynamic systems, that is, systems in which micromechanical elements undergo controlled motion, are, from one very useful perspective, a logical consequence in the continued evolution of microelectronic systems. Microdynamics offer special opportunities for the production of extremely miniaturized, highly complex systems. The opportunities presented by microdynamics can be compared with those that enabled the field of electronics to be revolutionized by the achievement of large-scale electronic-device integration. The VLSI revolution has, of course, affected all of society and its multi-billion-dollar indus- trial component continues robust growth after three decades of development. The linkage between micromechanical methods and typical integrated-circuit processes has been developing strongly since it was first demonstrated in seminal work done in 1982 when R. T. Howe demonstrated means to make microbeams from polycrystalline silicon films (Howe and Muller, 1982). Following this demonstration, Howe built the prototype polysilicon MEMS to be used as a chemical vapor sensor (Howe and Muller, 1986); see Figure 13.1. Through the 15 ensuing years, engineers have focused on employing the key steps of batch processing and self-assembly that underlie modern microelectronics for the development of systems composed of micromechanical as well as microelectronic components to make microelectromechanical systems, now typically designated by the acronym MEMS. The impact of MEMS is very broad, potentially affecting engineering design in areas as diverse as sensing, biological, environmental, and process instrumentation, robotics, engine control, and guidance. MEMS techniques also make possible the introduction of new test specimens and new testing protocols that can uncover fundamental information about material properties. As specimen dimensions shrink toward lower and lower multiples of single-crystal and molecular sizes, their controlled fabrication and manipulation in MEMS begin to offer exciting prospects for proving fundamental theory. The MEMS field promises to impact a wide swath of research and industry and affect not only material scientists, but also engineers, physicists, chemists, and biologists. 13.1.1 Origins The modern integrated circuit (IC) is the direct descendant of the planar process which, when introduced in the 1950s, combined the arts of photolithography and silicon chemical technology to extraordinary advantage. The productivity of this combination was made evident very early by the rapidity with which circuits of ever-increasing complexity appeared. It was therefore easy to justify large and repeated capital expenditures for technology and equipment and to put new ideas into development shortly after their conception so that ICs advanced rapidly. Today, it is commonplace to find ICs built with millions of devices; in fact IC engineers speak of the immediate future as the Gigabit Era. There is no doubt that integrated microelectronics are keystone elements of information processing in the engineering systems of our modern silicon age. For the complete design of an engineering system, however, it is not only necessary to process infor- mation. The information itself must be exchanged between a largely nonelectrical world and the electrical processing medium of the IC. This is the job of the sensors and actuators or, taken as one, of the transducers . In transducers we find the formidable partnership between micromechanics and microelectronics taking shape to make possible the new level of integration into MEMS. © 1999 by CRC Press LLC Some basic MEMS ideas had surfaced very early in the IC era. In fact, among the creative ideas tried out within the first decade of IC history was the employment of microfabrication technologies to produce a microdynamic element by a group of researchers at the Pittsburgh Westinghouse Research Laboratories. This group employed the silicon planar process to fabricate a resonant-gate, field-effect transistor (FET) (Nathanson et al., 1967) which consisted of a conventional silicon FET having a metal gate that was cantilevered over a surface channel covered only by silicon dioxide. An electrostatic field pumped the cantilever and its mechanical resonant frequency had applications as a timing source. The Westinghouse team also designed a large-screen optical projector based upon electrostatically driven cantilever reflec- tors. This work was very early in the development cycle of silicon microfabrication which was then (in the 1960s) revolutionizing all aspects of circuit design. The technology for the Westinghouse micrody- namic elements was not refined sufficiently for widespread commercial adoption and the research can now be seen as an early harbinger rather than as a catalyst for new directions in microfabrication. In the early 1980s, with two decades of IC development completed, the sophistication of electronics had advanced markedly. Circuit performance that would have been astonishing in the 1960s had become commonplace. The interfaces of these circuits with the largely nonelectrical world had become the logical point of focus for engineering design. The advantages of silicon micromechanics had begun to show themselves in its applications to ink-jet printing at companies like IBM, Hewlett-Packard, Canon, and Texas Instruments and in miniature pressure sensors such as those made at Honeywell Corporation. The impressive review of applications of silicon as a mechanical material published by Kurt Petersen (1982) focused the interest of many engineers on this area. Petersen’s review paper established itself almost immediately as a prime reference for the practice of silicon micromechanics. FIGURE 13.1 Detail from the first polysilicon MEMS. SEM of an apertured microbridge that is resonated by electrostatic forces under the control of on-chip circuits. (From Howe. R. T. and Muller, R. S., 1982, Extended Abstracts of the 1982 Spring Meeting of the Electrochemical Society, Montreal, Canada, 82-1, May 9–14. With permission.) © 1999 by CRC Press LLC 13.2 Micromachining The earliest methods used to build structures from silicon making use of lithography and etch technology that was mastered as a part of solid-state device research was via substrate micromachining . 13.2.1 Substrate Micromachining The technology for substrate micromachining is based upon orientation-dependent chemical etching of the silicon substrate. When used inventively together with etch-stopping techniques and masking films, this directed-etch technique can produce surprisingly complex structures. Some significant applications have been to a number of different sensing mechanics, such as silicon diaphragms (for pressure sensors) and cantilever beams (for accelerometers). Details about surface micromachining are not included in this chapter. 13.2.2 Surface Micromachining In the early 1980s, research being carried out at Berkeley showed that polycrystalline silicon had use as a mechanical material with very good characteristics for compatible IC processing. The design and fabrication by Howe of polysilicon resonant beams together with on-chip MOS circuitry demonstrated the practicality of what has become known as surface micromachining (Howe, 1985). Howe designed his beams to be driven by electrostatic forces like those at Westinghouse, mentioned earlier. The polysilicon beams were fabricated from patterned thin films by etching an underlying sacrificial silicon dioxide layer. The sacrificial layer was heavily doped with phosphorus to enhance its etch rate in hydrofluoric acid. To carry out surface micromachining with polysilicon, it is necessary that the sacrificial silicon dioxide be etched considerably faster than is the polysilicon mechanical material, itself. Laboratory studies have shown that polysilicon is etched in hydrofluoric acid (HF) at a negligible rate while silicon dioxide is etched at rates of 100 nm/min to 1 µm/min, depending on composition. This high ratio of etch rates is the key to successful surface micromachining. In contrast to substrate micromachining in which mechan- ical parts are sculpted from the wafer itself, surface micromachining makes mechanical elements out of materials deposited on the wafer surface. 13.2.3 Polycrystalline Silicon Properties For surface micromachining with polycrystalline silicon (polysilicon) deposited by low-pressure chemical vapor deposition (LPCVD) techniques, the polysilicon is often also used as a mechanoelectric transducer through its piezoresistivity. The use of CVD polysilicon strain gauges has advantages because the trans- ducing layer is dielectrically isolated from the substrate by a silicon nitride layer. Gauges made of polysilicon layers typically exhibit higher coupling factors and superior temperature characteristics to those made of single-crystal silicon. Electrical Properties . The electrical properties of polysilicon have been thoroughly studied and can be reviewed in several reference works (Kamins, 1988). Piezoresistance in polysilicon is a consequence of strain effects on the passage of carriers across the barriers between the crystalline grain boundaries, as well as the contributions of the bulk piezoresistivity of the silicon crystallites (Seto, 1976; French and Evans, 1985). Mechanical Properties . To use polysilicon effectively as a mechanical material, it is necessary to know well its mechanical properties. For most applications, the most important of these are the Young’s modulus, Poisson’s ratio, residual strain, and ultimate strength. In work over the past 10 years, consid- erable dependence has been found in these properties dependent on the details of the polysilicon fabri- cation process. At typical IC-LPCVD conditions, polysilicon films are in a residual compressive state after cool down to room temperature. This is generally undesirable and can cause multiply constrained structures, such as diaphragms, plates, or doubly fixed bridges to buckle. The compressive strain © 1999 by CRC Press LLC can have several possible sources including crystallites impinging during growth (Guckel and Burns, 1986) or entrapped gases such as oxygen (Muraka and Retajczyk, 1983). Studies of residual strain have been carried out using measurements of substrate curvature. The strain can be reduced signif- icantly by annealing (typically, at 850°C) (Choi and Hearn, 1984). Residual strain can be reduced significantly and even brought to a tensile state if LPCVD polysilicon is grown at temperatures very near the amorphous boundary (580°C). By annealing for 3 h at 1150°C, very low residual strains can be achieved (a typical value is –1.4 × 10 –4) (Guckel et al., 1987). Young’s modulus E and Poisson’s ratio ν for polysilicon have been measured using the curvature of different substrates coated with a given thin film over a range of temperatures. From these measurements, the ratio E /(1 – ν ) and the thermal expansion coefficient α are found. Some reported results from polysilicon heavily doped with phosphorus are E /(1 – ν ) ≈ 140 Gpa, about 70% as large as is found in single-crystal silicon. 13.2.4 Tribology in MEMS Tribology takes center stage when one considers the motions of very small bodies in which micrograms are proper units of mass. At this scale, as we know from observing the insect world, inertial effects play far smaller (in many cases negligible) roles in the relationships between driving force and motion. The nature of friction on this tiny scale is in need of fundamental study as systems for a multitude of applications are explored. A major challenge for MEMS designers employing surface micromachining is to understand and overcome the effects of stiction. The term stiction is used to describe two circumstances: (1) sticking of a “freed” member together with the newly exposed surface underneath the sacrificial layer after the final “freeing” etch step of the micromachining process, and (2) adherence of two mechanical microparts that approach or touch one another when the microsystem is in operation. MEMS engineers employing surface micromachining frequently encounter fatal stiction effects when they attempt to release structures in the final etch step. Research into this problem has identified the major role of surface tension during the drying stage after a final liquid-etch release step (Mulhern et al., 1993). As MEMS designs have incorporated larger-area, very thin, low-stiffness members (lateral surfaces tens to hundreds of a micrometer on a side and 1 or 2 µm thick), the release-stiction problem becomes of first order. When sacrificial layers are dissolved by the etch, adjacent surfaces within micrometer dimensions result in menisci that lead to attractive forces. These forces can, in turn, buckle the “freed” member, and possibly cement it to the underlying surface. Mechanical analysis of release-stiction effects has been performed (Mastrangelo, 1997). By using mechanisms to sublimate the final etch, release stiction can be avoided, and large surface-micromachined structures processed successfully (Guckel et al., 1990; Lebouitz et al., 1995). Already with the first microdynamic devices in which surfaces slide past one another (Fan et al., 1998), stiction was observed to account for the major retarding forces, and techniques, such as the production of “dimples” in moving polycrystalline silicon elements, were demonstrated. Recently, methods that enable quantitative study of stiction effects have been described. Mastrangelo and co-workers have shown a method for quantifying the post-release stiction of structures using an array of released cantilever beams having graduated lengths (Mastrangelo, 1997). By applying electrostatic forces, the beams are attracted to a surface. When the force is removed, beams longer than a critical “detachment” length remain adhered to the surface while shorter beams are freed because of their internal stiffness (Mastrangelo and Hsu, 1992). The detachment length can be related to an adhesion energy. By observing the influences of different surface coatings on adhesion-energy values, one can form comparative evaluations about the coatings (Houston et al., 1995, 1996). The MEMS field has developed sufficiently to produce micromechanisms as complex as gear boxes capable of two-and-three levels of speed reduction (Sniegowski, 1997). For such structures to enter commercial applications, surface-lubricating films are a necessity. At Texas Instruments, Inc., surface © 1999 by CRC Press LLC micromachining has been used to produce a micromirror array composed of movable aluminum struc- tures. Results with lubricants used on this digital micromirror array are reported by Henck (1997). Research results on films for polysilicon MEMS mechanisms are discussed by Maboudian and Howe (1997). 13.3 MEMS Structures and Systems Both surface and bulk micromachining have been employed to make devices and systems having very surprising complexity. Some of these are now incorporated into commercial products, while others are still the subject of research. Because of the particular interests of the author and the need to limit the length of this discussion, the following examples will be drawn exclusively from surface-micromachined MEMS. An impressive example of a commercial MEMS that makes use of surface micromachining to build a fully integrated accelerometer is the AD-XL50 produced by Analog Devices, Inc., for use in automobile airbag-deployment systems. This accelerometer integrates a tiny seismic mass in the midst of formidable bipolar-CMOS circuitry. The system employs force feedback using coulombic force to hold the seismic mass fixed and it senses position making use of a differential capacitance measurement. The AD-XL50 has been widely described in the popular literature (Goodenough, 1991) and it now accounts for a large and growing share in a very competitive marketplace. Figure 13.2 lists some of the surprising properties of this monolithic MEMS. A coulombic actuating force (as employed in the AD-XL50) is used frequently in the MEMS field, especially for surface-micromachined embodiments because coulombic force (alternatively, electrostatic force) is easily incorporated into the system and employs only an imposed voltage between two elements. The force increases linearly with the capacitance between the electrodes and quadratically with the applied voltage. For typical surface-micromachined parts and spacings, the force is quite small (of the order of micronewtons) but still adequate for many applications. To increase the range of achievable forces in MEMS, a number of alternative actuation mechanisms have been used or are under study at this time. Most of these alternatives to coulombic force introduce complication into the fabrication technology for the MEMS. Nonetheless, to obtain millinewton forces and higher, it is likely that these other forcing FIGURE 13.2 Facts about the Analog Devices AD-XL50 accelerometer. (Courtesy of Richard Payne, Analog Devices, Inc.) © 1999 by CRC Press LLC techniques will be necessary. A listing of other actuation means that have already been applied is given below. 13.3.1 MEMS Actuation Forces 13.3.1.1 Electrostatic Comb Drive As already discussed, actuation in MEMS using coulombic force has distinct advantages for compatibly processed MEMS. Howe, for example, made use of coulombic force to power his resonating beam in the first polysilicon micromechanical structure (Howe and Muller, 1982). In further work with electrostatic forcing, Howe and student W. Tang first constructed a polysilicon comb-drive resonant actuator which employs coulombic force in a flexure structure to cause it to move parallel to the substrate surface (Tang et al., 1989). The layout of this comb drive, which is supported by only two pedestals attached to the substrate, is shown in Figure 13.3. The comb drive has become a very frequently used microactuator both for resonant and nonresonant systems. Because the actuated part is a flexing structure, there is no surface friction in its motion, and it can be driven in resonance with a very high mechanical Q value (100,000 in vacuum). Much work is continuing on comb-drive resonators and their applications and they have been made by substrate as well as by surface micromachining. Figure 13.4 compares aspects of the behavior of the comb-drive to that of a parallel-plate actuator. A photograph of one of the original moving comb structures undergoing oscillation is shown in Figure 13.5. 13.3.1.2 Vertical vs. Lateral Oscillation Vibrating micromechanical structures are useful for a variety of sensors and actuators. For sensing, one can make use of the dependence of the frequency of a mechanical resonator to physical or chemical parameters which affect the vibrational energy. Microfabricated resonant structures for sensing pressure, acceleration, and vapor concentration have been demonstrated. An elegant example of resonant drive for actuators is provided by the successful Bulova accutron wristwatch movement in which an electronic tuning fork is coupled mechanically to a rotating mechanism. Pisano has pointed out the possible applications of this design to microdynamics (Pisano, 1989). Recently, some very spectacular mechanisms have been made using polysilicon surface micromachining at the Sandia Laboratories in New Mexico. The Sandia process makes use of chemical-mechanical polishing in order to planarize wafers after the FIGURE 13.3 Comb-drive resonator. (Courtesy of William Tang, Jet Propulsion Laboratory.) © 1999 by CRC Press LLC micromechanisms have been made in a recessed area of the chip. Then, in a subsequent CMOS process, very high quality electronic circuits can be fabricated in a continuous-batch process. Figure 13.6 shows a gear assemblage driven by resonant comb drives that was made at Sandia. The tiny rotor is 55 µ m in diameter and it has achieved rotation rates as high as 300,000 rpm. It can be rotated either counterclockwise or clockwise. To gain a measure of the size of these structures, Figure 13.7 FIGURE 13.4 Comparative behavior of a vertical resonator with a comb-driven substrate-parallel resonator. (Cour- tesy of William Tang, Jet Propulsion Laboratory.) FIGURE 13.5 Resonating comb drive. (Courtesy of William Tang, Jet Propulusion Laboratory.) © 1999 by CRC Press LLC shows a photograph of the tiny driven rotor of Figure 13.6. Next to the rotor in Figure 13.7 are clumps of red blood cells and, adjacent to the side arms, a grain of sand. The MEMS field has begun to expand in several directions where full miniature “systems on a chip” are being realized. Important commercial MEMS applications to accelerometry, to environmental sensing, and to display have received fairly wide coverage in the technical and industrial product literature. An area in which the author has worked that is just now developing into fully engineered systems is the area of miniature integrated optical systems for communications, instrumentation, and sensing applications. The remainder of this chapter will focus on developments in this area called microphotonics . 13.3.2 MEMS for Microphotonics The incorporation of micromechanical structures into fiber-optic systems holds promise of reducing costs and providing new opportunities for systems applications. Research groups around the world are FIGURE 13.6 Batch-fabricated polysilicon assemblage made at Sandia Laboratories. (Courtesy of P. McWhorter.) FIGURE 13.7 Rotor driver on the wheel assemblage of Figure 13.6 made at Sandia Laboratories. (Courtesy of P. McWhorter.) [...]... Maboudian, R., and Howe, R.T (1996), Technical Digest, Solid-State Sensor and Actuator Workshop, Hilton Head Island, SC, pp 42–47, June Howe, R.T (1985), “Surface Micromachining,” in Micromachining and Micropackaging of Transducers, (C.D Fung, ed.), Elsevier Science Publishers, New York, 169 Howe, R.T and Muller, R.S (1982), “Polycrystalline Silicon Micromechanical Beams,” Extended Abstracts of the 1982... (Courtesy of M Daneman, O Solgaard, and N Tien, UC Berkeley.) FIGURE 13.14 Folded micromirror driven by four microvibromotor drivers (Courtesy of M Daneman, O Solgaard, and N Tien, UC Berkeley.) in steps that are only a fraction of a micrometer per blow In the BSAC design, pairs of vibromotors strike the carriage at 45° relative to its guiding keyway slot Figure 13.13, shows the concept for a microvibromotor... (ultimately the mount for a micromirror) over about 100 µm displacement Pairs of these vibromotors are used to position the support vane of a folded mirror in the structure shown in Figure 13.14 Some details of the microvibromotor-driven micromirror system are indicated in Figure 13.15 © 1999 by CRC Press LLC FIGURE 13.15 Features of an actuated microreflector system (Courtesy of M Daneman, UC Berkeley.)... Head Island, SC, 2–6 June 1996 17 M.-H Kiang, O Solgaard, R.S Muller, and K.Y Lau, “High-Precision Silicon Micromachined Micromirrors for Laser-Beam Scanning and Positioning,” 1996 Sensor and Actuator Workshop, (late news) Transducer Research Foundation, Hilton Head Island, SC, 2–6 June 1996 Acknowledgments Many colleagues have contributed their thoughts, suggestions, and work to this brief review of the... evaporation of 50 nm of gold to increase mirror reflectivity—resulting in measured reflectivities of 85% at 1.35 µm The reduction from the theoretical 96% reflectivity of a perfect gold mirror is due to scattering caused by polysilicon surface roughness and by etch holes and dimples in the mirror Movable Micromirror Applications of these folded mirrors are for beam steering, optical alignment, scanning, and switching... exploited Work at the Berkeley Sensor & Actuator Center (BSAC) at UC Berkeley and at UCLA has concentrated on surface micromachining with polysilicon and has developed actuation techniques appropriate both for beam switching and for scanning using micromirrors made by folding micromachined polysilicon structures out of the surface plane and covering them with a gold reflecting surface Figure 13.8 shows such... Optical Components,” MEMS ‘95, 1995 Int Conference on Microelectromechanical Systems, Amsterdam, The Netherlands, January, 1995 8 O Solgaard, M Daneman, N.C Tien, A Friedberger, R.S Muller, and K.Y Lau, “Precision and Performance of Polysilicon Micromirrors for Hybrid Integrated Optics,” 1995 International Symposium on Optoelectronic, Microphotonics, and Laser Technologies, Conference 2383A, SPIE, The... 1996, San Jose, CA 13 M.-H Kiang, O Solgaard, R.S Muller, and K.Y Lau, “Surface Micromachined Electrostatic-CombDriven Scanning Micromirrors for Barcode Scanners,” IEEE MEMS ‘96 Workshop, 11–15 February, 1996, San Diego, CA 14 M.-H Kiang, O Solgaard, R.S Muller, and K.Y Lau, “Design and Fabrication of High-Performance Silicon Micromachined Resonant Microscanners for Optical Scanning Applications,” Integrated... N.C Tien, O Solgaard, K.Y Lau, and R.S Muller, “Actuated Micromachined Microreflector with Two Degrees of Freedom for Integrated Optical Systems,” Integrated Photonics Research Conference, Boston, MA, May, 1996 16 M.J Daneman, N.C Tien, O Solgaard, K.Y Lau, and R.S Muller, “Linear Vibromotor-Actuated Micromachined Microreflector for Integrated Optical Systems,” 1996 Sensor and Actuator Workshop, 109–112,... motor and mirror carriage are built using only two layers of polysilicon © 1999 by CRC Press LLC FIGURE 13.11 Measured performance of folded micromirrors used as scanners (Courtesy of M Daneman, O Solgaard, and M H Kiang, UC Berkeley.) FIGURE 13.12 Scanning micromirror with comb drives for scanning around three axes (Courtesy of M H Kiang, and O Solgaard, UC Berkeley.) The impact force in a vibromotor . Recently, micromirrors that consist of multiple hinged plates which Updated, expanded, and revised version of article first published in Micro/ Nanotribology. Muller, R.S. “Microdynamic Systems in the Silicon Age” Handbook of Micro/ Nanotribology. Ed. Bharat Bhushan Boca Raton: