MEMS:The maturing of a new technology
09/01/1997
Cover Article
MEMS: The maturing of a new technology
Michael Pottenger, Beverley Eyre, Ezekiel Kruglick, Gisela Lin University of California, Los Angeles, California
Microelectromechanical systems (MEMS) include sensors and actuators that are fabricated with processes similar to those used in mainstream IC production. Automotive, biomedical, aerospace, and robotic MEMS devices can be produced with combinations of bulk and surface micromachining.
The second half of the 20th century has seen information technology develop at an unprecedented rate. Beginning with the invention of the transistor in 1947, information technology has dramatically changed the manner in which modern society works and plays. Transistors led the way to ICs in the 1960s, and ICs enabled the development of virtually all commercial and consumer electronics products that are on the market today -including personal computers, cellular telephones, CD players, and home video games. The past 50 years can be characterized by the ability to move electronic information from one place to another, with continuous improvements in efficiency, reliability, and costs.
A similar revolution in information technology will occur in the first half of the 21st century, but the information conveyed will not be merely electronic. MEMS are the core technologies enabling the development of mechanical, chemical, or biological "smart systems."
At the heart of this revolution are two classes of instruments: sensors and actuators. Sensors are simply transducers that convert energy from one form to another (e.g., mechanical to electrical), and provide for passive measurement or monitoring. Actuators allow sensors to interact actively with the world. The ability to integrate sensors and actuators into efficient, reliable, and economic systems is fueling MEMS research in the US, Europe, and Japan.
In very large scale integration (VLSI) processing, photolithographic techniques control the patterning of thin films and the deposition of dopants used to make transistor gates and metal contacts. MEMS processing uses these same techniques to create structural components that are essentially sub-millimeter-sized machine parts. These parts usually require post-fabrication processing or assembly in order to become workable devices.
MEMS technology can generally be categorized into two groups: bulk and surface micromachining. These categories reflect not only different fabrication processes, but different post-fabrication techniques for finishing the mechanical subsystem. This article describes a handful of micromachined devices that are available, as well as some future prospects at the research stage.
Bulk and surface processing
Bulk micromachining, one of the two major categories of MEMS processing, involves etching away selected portions of the substrate much like a sculptor in marble will start with a solid block and remove material until a final shape is created. Bulk micromachining typically etches away most of a silicon chip, with the remaining single-crystal silicon as the final structure.
The two most common bulk etchants are wet-based potassium-hydroxide (KOH) and ethylene-diamine-pyrocatehol (EDP). Both etchants are anisotropic and thus remove material at different rates along different crystal planes to produce characteristic pyramidal pits and sloped sidewalls. KOH etching is incompatible with integrated active electronic devices "on-chip," since both potassium and hydroxyl ions contaminate the dielectric oxides that prevent conducting layers from shorting.
In surface micromachining, structures are built on top of the wafer using thin films deposited through various standard methods familiar to IC fabrication. Unfortunately, the standard processes used to create electronic devices are not optimal for the creation of moving parts, so mechanical and electrical integration is more difficult than in bulk micromachining.
For example, polysilicon-surface-micromachining processes typically use sacrificial oxide spacer layers that are etched away to render free polysilicon structures. This process flow is incompatible with standard IC processes that use oxides to isolate conductors. To create a complete electromechanical system, a hybrid arrangement of two separate processes is required, one for electronics and one for mechanical components.
One compromise uses a standard CMOS process to fabricate electrical and mechanical parts, with an EDP bulk-machining process to remove portions of the substrate. The structures on top of the substrate are not affected by the bulk etch, provided they are sufficiently masked from the etchant. The oxide that separates and protects the electronics can be used to create beams and membranes over "pits" in the substrate, and these structures are free to move after release. This hybrid process can create many types of sensors that can be integrated with on-chip electronics for complete MEMS devices.
Trade-offs between device function and manufacturing capability blur the division between MEMS device design and process development. Novel structures open the door to new applications and show that the advantages of MEMS are inherent in the approach and not unique to any specific process. However, any nonstandard IC process has all the electronic integration problems associated with surface micromachining.
Current commercial MEMS
The accelerometer market is perhaps the most dramatic example of a business changed by micromachining. Many companies first venture into micromachining with accelerometers because there is a large market for small, cheap accelerometers in automotive air bag systems.
A single, low-cost accelerometer can replace a network of crash sensors joined with an expensive wire harness. Design and fabrication approaches differ among companies: the Analog Devices series is surface micromachined in tensile polysilicon, the EG&G IC Sensors line is bulk micromachined out of multiple stacked wafers (Fig. 1), and Motorola`s accelerometers are surface-micromachined polysilicon with a bulk-micromachined cap.
 |
Figure 1. Bulk-micromachined accelerometer with sensitivity in the vertical direction. Acceleration of the center mass results in support beam deflection, which produces a strain in diffused piezoresistive traces in the beams. Acceleration thus results in a detectable change in resistance. (Courtesy of EG&G IC Sensors)
Whatever fabrication approach is used, the goal is a small and inexpensive sensor that provides the desired functionality. Companies pursue micromachined accelerometers not because they are micromachined, but because they provide the desired function at a competitive cost. Siemens continues to build nonmicromachined accelerometers by applying advanced manufacturing techniques to produce a cheap, reliable sensor with only six parts, but the market continues to shift toward micromachined solutions to reduce cost and size.
The pressure sensor industry has been largely dominated by bulk-micromachined silicon sensors for many years (Fig. 2), representing the largest MEMS product market in terms of units sold/year. While the initial drivers for development were cost and size, final devices demonstrate additional benefits. When prices for micromechanical medical blood pressure sensors dropped sufficiently, replacing the sensor became cheaper than maintaining it. These devices are now used disposably, increasing both sales and utility. Micromachined pressure sensors are also common in automotive applications. Although micromechanical valves are available, they do not yet have the pressure tolerance and low leak rates that many applications require.
 |
Figure 2. Bulk-micromachined sensor capable of measuring differential pressure across the thin silicon membrane due to strain in piezoresistive traces. Other pressure sensor designs detect the change in capacitance between the membrane and a second parallel plate. (Courtesy of EG&G IC Sensors)
Atomic force and scanning tunneling microscopy (AFM and STM, respectively) tips are almost exclusively micromachined. The suspension material is typically silicon nitride, but tip fabrication techniques and materials vary widely. Micromachining is the only practical way to achieve the superlatively sharp tips (Fig. 3) and incredibly weak suspensions required for AFMs.
Micromachining can enable new technology by providing new solutions to existing problems or allowing for improvements to an existing idea. For example, Texas Instruments has a micromachining process for making large arrays of digitally controlled mirrors on a single silicon chip. Imaging applications of these mirror arrays include projection displays, rear projection televisions, and 600 dpi color printers. This unique micromachined device represents new competition in the more than $10 billion/year display market.
 |
Figure 3. Scanning electron microscope (SEM) image of a silicon nitride beam for AFM. The spring constant of the arm is extremely small in the vertical direction, resulting in high sensitivity. (Courtesy of Park Scientific Instruments)
MEMS devices in R&D
Optics. MEMS technology is hotly researched for other optical applications, with the possibility of producing low-cost, low-power, wireless devices. Many optical MEMS structures can be surface micromachined, with micro-hinged polysilicon plates as integral structural elements (Fig. 4). The hinges are defined by a first-layer polysilicon pin and a second-layer polysilicon staple separated by silicon dioxide sacrificial layers (that are eventually etched away in hydrofluoric acid) [1].
 |
Figure 4. a) Process flow for micro-hinge fabrication, and b) SEM image of a complete micro-hinge. (Courtesy of K.S.J. Pister)
One proposed surface-micromachined optical device that relies on micro-hinges is the corner cube reflector [2]. When hinged polysilicon plates are rotated into orthogonal positions, a laser beam is back-reflected; when a base plate is rotated, a laser beam is not back-reflected. This difference can be used to encode a binary optical data stream.
Miniature Fresnel lenses can be made by strategically patterning rotated polysilicon plates (Fig. 5) [3]. Rotated side plates are used to hold the lens vertically. The lens can be incorporated with laser components to form complete optical benches on a single silicon chip. Similar methods can be used to make vertical diffraction gratings and complete lens arrays with predetermined alignments [4]. With batch-fabrication economies of scale, each device can be relatively inexpensive.
 |
Figure 5. SEM of a micro-Fresnel lens coupled to a diode laser. Rotated guide plates hold an edge-emitting laser next to the lens. Since the guide plates can be defined to within 1 ?m, the structure is essentially self-aligned. (Courtesy of M.C. Wu)
Biomedical. MEMS are also being used in biomedical applications. A MEMS device to measure the contractile force of individual heart cells is under development. A single living heart cell is glued between two movable polysilicon clamps (Fig. 6) [5]. When the cell contracts, it pulls the clamps inward and the support beams are deflected. Using the known spring constant in the beams and the visually measured deflection, the amount of force can be estimated.
 |
Figure 6. SEM of one of two required MEMS polysilicon clamps used to measure contractile force in living heart cells. Micro-hinges rotate first, and then spring locks hold the clamps vertically.
Since the mass of this system is so small compared to nonmicromachined systems currently used, a higher bandwidth and, thus, higher fidelity measurements on the cellular level may be attainable. Typical measurements are 8-10 microNewtons for rat cardiac cells tested, a force equal to the weight of 10 crystals of ordinary table salt. CMOS electronics can be integrated with the clamps to produce a submersible heart-cell-force transducer system.
Tiny, sharp probes lined with electrodes and on-board signal processing electronics can measure electrical signals in the brain [6]. In the future, tiny MEMS hands that can grip and move biological tissue and/or cells may be used in minimally invasive microsurgery, reducing patient discomfort and healing time.
Chemical analysis may be transformed by developments in micromachining. Miniaturized chemical analysis systems may soon replace tabletop mass spectrometers, gas chromatographs, and electrophoretic separation systems. The benefits of miniaturization include the ability to work with minute chemical samples, fast response time of the instruments, and a dramatic overall reduction in instrument volume (Fig. 7).
 |
Figure 7. SEM of micromachined channels, reagent reservoirs (circular), and testing chambers (hexagonal) on a "lab-chip." Electrokinetic forces move picoliters of liquids to perform separations and analyses. (Courtesy of Caliper Technologies)
For example, electrophoresis is the procedure of separating various ionic species according to their response to an applied electric field. Larger ions move more slowly through a solution than smaller ions, producing separation according to their size. Several groups around the world have proven this to be an effective method of separation using anisotropically etched channels in silicon wafers as the separation medium. When driven by micromachined pumps, such a system becomes a fully independent instrument capable of introducing, separating, and detecting the chemical samples. The small scale of these instruments results in separation times of several seconds for sample sizes of interest.
Aerospace. Designers of spacecraft and satellites take advantage of the small, low-power, and high-reliability components that can be realized using MEMS technology. Traditional spacecraft designs are single crafts with payloads in the hundreds of kilograms. However, micromachining makes possible the development of microsatellites with payloads of only a few kilograms. Work on microsatellites is currently underway at Jet Propulsion Laboratory, The Aerospace Corporation, and companies in Germany.
Regardless of their size, these proposed miniature spacecraft require the same basic functions as their full-sized counterparts. Guidance, navigation, and attitude control are accomplished through sensors such as accelerometers and gyroscopes, coupled to actuators and/or a nonmicromachined propulsion system. Accelerometers for these applications use the same operating principles as those for automotive applications.
Angular position is typically determined by sensors that measure the rotational velocity of a body, and then integrate this signal over time to determine position. A common angular-rate sensor design utilizes the Coriolis effect to couple the driven vibrations of tuning fork tines into a secondary vibration whose amplitude is proportional to the rate of rotation. A variety of designs and fabrication techniques have been demonstrated for both the oscillating masses and the actuator which drives them, using surface and bulk micromachining, electroforming, and piezoelectric materials. Such sensors must have minimal drift over long time periods for navigation purposes, or high fidelity over short periods for attitude control.
Robots. The crossover of microfabrication techniques from ICs to mechanical components has brought the idea of microrobots out of the realm of speculation and into the laboratory. Mechanical links and couplings, joints, and motors (for drive and control) are the building blocks of articulate microstructures developed at the University of California, Los Angeles [7]. All components were demonstrated in surface micromachining, with the mechanical elements assembled after fabrication into fully 3-D structures. Micromachined hinges (described earlier in this article) are the key structures that allow for such assembly.
Depending on the hinge design, two plates can either be rigidly joined or connected to allow for rotation of one relative to the other. Three plates can be folded into a hollow triangular beam and kept in place through snap locks to make a mechanical link. Mechanical links can then be joined through a series of hinges that allow one link to rotate with respect to the next, producing a functional robotic arm (Fig. 8). Articulation of these arms is achieved by coupling the arms to drive and control motors fabricated in the same-surface micromachined process.
 |
Figure 8. SEM of a microrobotic arm with 3? freedom of motion. The triangular beams are made rigid through snap locks, which resemble arrowheads passed though narrow slots. (Courtesy of R. Yeh)
Conclusion
As the systems we use in our lives become more complex and more intelligent, the number of sensors employed in those systems will have to multiply. As the number of sensors goes up, the size and the cost will have to go down in order to maintain economical and usable systems. MEMS are the inevitable and necessary next step in the evolution of complex high-tech systems.
Within a short time, microsensors and micro-actuators will be part of even the most simple electronic devices -enhancing their ability to perform their functions intelligently. Arrays of invisible microsensors will be in rooms and automobiles for control of the environment. MEMS sensors will monitor the many systems in automobiles to provide optimal functioning.
Miniature spacecraft will take over many of the tasks now performed by large and very costly manned missions. Nonintrusive microtools will be used by surgeons for procedures such as the retrieval of tumor samples for analysis. MEMS devices will be used in research to probe the secrets of cells and microorganisms on their own scale. Opto-electronics will benefit from the use of MEMS devices to provide low-power communication.
If recent trends continue, it will be very difficult to define the MEMS field clearly in 20 years; instead, a wide variety of micromachined devices and components will be embedded in other nonmicromachined systems. Automobile air bag triggers and blood pressure sensors are current examples of embedded MEMS. Researchers in this rapidly growing field share an excitement that comes from exploring a new frontier.n
References
1. K.S.J. Pister, M.W. Judy, S.R. Burgett, R.S. Fearing, "Microfabricated Hinges," Sensors and Actuators A, Vol. 33, pp. 249-256, 1992.
2. D.S. Gunawan, L.Y. Lin, K.S.J. Pister, "Micromachined Corner Cube Reflectors as a Communication Link," Sensors and Actuators A, Vol. 46-47, pp. 580-583, 1995.
3. L.Y. Lin, S.S. Lee, K.S.J. Pister, M.C. Wu, "Micro-Machined Three-Dimensional Micro-Optics for Integrated Free-Space Optical System," IEEE Photonics Technology Letters, Vol. 6, No. 12, pp. 1445-1447, December 1994.
4. M.C. Wu, S.S. Lee, L.Y. Lin, K.S.J. Pister, "Micromachined Micro-Optical Bench for Free-Space Integrated Optics," Government Microcircuit Application Conference (GOMAC), San Diego, CA, pp. 203-206, Nov. 8-10, 1994.
5. G. Lin, K.S.J. Pister, K.P. Roos, "Heart Cell Contractions Measured Using a Micromachined Polysilicon Force Transducer," Proc. of the SPIE 1995 Symposium on Micromaching and Microfabrication, Vol. 2642, pp. 130-137, Austin, TX, Oct. 23-24, 1995.
6. K. Najafi, "Solid-State Microsensors for Cortical Nerve Recordings," IEEE Engineering in Medicine and Biology, Vol. 13, No. 3, pp. 375-387, June/July 1994.
7. R.Yeh, K.S.J. Pister, "Measurement of Static Friction in Mechanical Couplings of Articulated Microrobots," Proc. of the SPIE 1995 Symposium on Micromachining and Microfabrication, Vol. 2642, pp. 40-50, Austin, TX, Oct. 23-24, 1995.
MICHAEL POTTENGER received his BS degree from the California Institute of Technology in 1991, and his MS degree from the University of Southern California in 1995, both in mechanical engineering. He is currently a PhD candidate in the electrical engineering department at the University of California, Los Angeles (UCLA), where his research is focused on micromachined inertial sensors. UCLA Electrical Engineering Dept., Engineering IV 56-125B, 420 Westwood Plaza, Los Angeles, CA 90095-1594; ph 310/206-3995, fax 310/825-7928, e-mail [email protected], www.janet.ucla.edu/~mikep.
BEVERLEY EYRE received his BS degree in electrical engineering from UCLA in 1994, and is currently pursuing his MS and PhD degrees at UCLA. His research interests are in magnetometry from a MEMS perspective and microrobotics.
EZEKIEL KRUGLICK received his BS degree in electrical engineering in 1995, and his MS degree in 1996, both from UCLA. He is in the PhD program at the University of California, Berkeley. Research interests include microrobotics, electron tunnel sensors, integrated CMOS microstructures, and distributed sensor networks.
GISELA LIN received her BS degree in electrical engineering and material science engineering from the University of California, Berkeley, in 1990, and her MS degree in electrical engineering from the University of California, Santa Barbara, in 1992. She expects to receive her PhD degree in electrical engineering from UCLA this year. Her research interests are in the biomedical applications of MEMS sensors.