Micromachining technology began to develop rapidly over fifteen years ago, using the materials and processes developed for the integrated circuit industry to form miniature structures in silicon and related materials for purposes other than electronic devices. The fabrication of miniature solid state sensors has always been the main thrust of micromachining technology. What is today called MEMS is in most ways only an evolution and expansion of this technology thrust. The improvements possible today with MEMS technology involve the expanded range of materials and processes that can be employed, and the precise dimensional scale and the mechanical complexity that can be achieved in devices. Sensors are still the major application thrust of MEMS, particularly in commercial products.
As the term "microelectromechanical systems" suggests, MEMS technology relates most directly and has the greatest impact on sensors for mechanical variables. It must be noted that, because of the evolutionary development of micromachining technology for sensors, it is often not possible to specify that certain mechanical sensors employ MEMS while others do not. In contrast, most of the solid-state sensor developments for chemical sensors, gas sensors, and biosensors do not employ MEMS concepts and technology and are not discussed here.
There are major MEMS sensor efforts in the United States focusing on the development of advanced silicon pressure sensors and silicon accelerometers. These efforts are primarily driven by existing and potential high-volume applications in automotive, medical, and consumer products. The most important requirements are typically for moderate performance at very low cost. The batch-fabricated nature of MEMS sensors addresses these needs. Other advantages of MEMS sensors for many applications include small size and low power. Development work is under way in industry as well as at many universities. Development efforts also address high-performance needs in military and industrial systems. Some advanced pressure sensors and accelerometers are commercially available, while others may be expected in the marketplace in the next three to five years.
While silicon diaphragm pressure sensors have been commonplace for many years, more recent developments in the United States resulting from MEMS technology are resulting in significantly smaller and potentially less expensive devices. These devices are most suitable for low to moderate performance applications where cost and perhaps size are critical, such as disposable medical sensors and consumer products. One technique developed at the University of Wisconsin (Guckel et al. 1987) uses thin film polysilicon for the sensor diaphragm. Because the polysilicon diaphragm is only about one-tenth the thickness of conventional silicon pressure sensor diaphragms, the polysilicon devices are correspondingly smaller. Thus, there are many times more sensor die per wafer, and the cost is much lower. Another technique originated by NovaSensor (Petersen et al. 1988) employs high-temperature fusion bonding of silicon wafers to form inward tapering cavities under single-crystal silicon diaphragms. This technique can also result in much smaller sensor die than standard techniques. As a result, these devices are being used in medical catheters where both size and cost are critical.
Many pressure sensing applications may require measurement of low pressures in the range of 1,000 Pa, but adequate sensitivity to such low pressure is not easily accomplished with conventional silicon diaphragm sensors. However, by using advanced MEMS micromachining technology to add corrugations or bosses to the diaphragm, areas of stress concentration can be formed that facilitate these low- pressure measurements (Mallon et al. 1990). Several such low-pressure micromachined sensors are commercially available.
Technology for miniature resonant strain sensors has been demonstrated (Guckel et al. 1992; Petersen et al. 1991) that may replace capacitive or piezoresistive readout of silicon diaphragm pressure sensors for improved performance. These devices employ a miniature beam driven into mechanical resonance by feedback electronics. The resonant frequency is a sensitive measure of strain, but is relatively insensitive to temperature or to the electrical properties of the device. The devices are located on a conventional silicon pressure sensor diaphragm, where they function as frequency output strain gauges. There are no critical analog gain stages, and the frequency output can be easily interfaced to digital signal processing circuitry. A vacuum shell can be integrally formed around the beam to allow operation in any media. The mechanical complexity and very small size required for these devices is made possible only by advances in MEMS technology.
Because of the large potential markets in automotive applications for sensors for air bag deployment and active suspensions, there are many efforts under way in the United States to develop miniature silicon accelerometers. MEMS technology is key to achieving the required performance combined with the low sensor costs demanded by these applications. Several accelerometers using bulk silicon micromachining to form a single-crystal silicon proof mass and supporting flexures have been developed (Bart et al. 1988; Terry 1988) and are commercially available. Either piezoresistive or capacitive readout of proof mass displacement can be employed. These devices also use multiple wafer bonding technology to fabricate sandwich structures to provide overrange protection and damping. More recently, microminiature accelerometers fabricated in polysilicon by surface micromachining were reported (Ristic et al. 1992; Payne and Dinsmore 1991). These accelerometers may also be integrated with CMOS electronics. Several commercial products are available.
In order to achieve wider dynamic range and address higher performance applications such as inertial navigation, conventional accelerometers frequently use closed-loop, force-rebalance techniques. Several closed-loop silicon accelerometers have been reported in the United States. One uses a bulk silicon structure (Henrion et al. 1990), while another employs a surface micromachined single-crystal structure (Boxenhorn and Greiff 1990) and consists of a polysilicon accelerometer integrated with CMOS detection circuitry (Yun, Howe, and Gray 1992).
For critical applications such as air bag deployment, efforts are under way to develop self-test capability in accelerometers. In one such device (Allen, Terry, and DeBruin 1990), an input signal heats an actuation beam that applies a known force to the device structure. Another concept uses electrostatic forces to apply the test signal to the accelerometer (Pourahmadi, Christel, and Petersen 1992). If proper response is obtained, the functionality of the accelerometer is confirmed. This rather complex device structure is another example of the added sensor capabilities made possible by MEMS technology.
Several other types of sensors for mechanical variables are being developed. A vibratory silicon gyroscope has been demonstrated (Greiff et al. 1991). This device is a doubly gimbaled structure supported by torsional flexures. The gimbaled mass is driven into torsional vibration. Coriolis forces during rotation transfer energy into vibration in the orthogonal torsional mode. The amplitude of this second vibration is a measure of input rotation rate. A micromachined silicon condenser hydrophone has also been reported (Bernstein 1992). This device uses a surface-micromachined silicon plate that is deflected by the acoustic wave, causing a capacitance change with respect to an overlying micromachined electrode. High acoustic sensitivity is obtained in a device only 1 mm on a side, comparable to ferroelectric hydrophones many times larger.
Complex, micromachined structures can be designed and fabricated to have unique thermal properties that can be useful for sensors. Microminiature versions of thermal mass flowmeters have been demonstrated to be more sensitive, faster responding, and lower in power than macroscopic devices (Ohnstein et al. 1990; Tai, Muller, and Howe 1985). While the basic thermal principles of operation are similar, the miniature devices have significantly better performance due solely to their small size. A fully-integrated, miniature Pirani pressure/vacuum gauge has been reported (Mastrangelo and Muller 1991). This device measures absolute pressure between 10(1) to 10(4) Pa by the pressure dependent change in thermal conductivity. The MEMS device is very low in power, and is fully integrated with NMOS circuitry giving a digital output. Microstructure bolometer devices have useful sensitivity to infrared radiation and can be easily fabricated as arrays of miniature devices. Thermal infrared detectors (Choi and Wise 1986) and very large imaging arrays (Guckel et al. 1993b) are also under development.
MEMS technology may also be advantageously combined with optics for sensing. One example is the fabrication of a miniature Fabry-Perot interferometer for optical measurements in the near infrared spectral region (Jerman, Clift, and Mallinson 1990). Another demonstration is the use of light beams to both excite a microminiature resonant beam strain sensor and to sense the beam's vibrational motion (Guckel et al. 1993b).
In an advanced concept, micromachined structures have been fabricated as extremely sensitive displacement transducers using electron tunneling concepts. The devices employ a microfabricated tunneling tip in close proximity to a surface. Resolution of 10(-2) Å displacements between tip and surface by variation in the tunnel current have been demonstrated. Applications to Golay cell infrared detectors, accelerometers, and seismometers are being explored. Such extremely high displacement sensitivity raises questions of dynamic range and stability that must be addressed.
As described above, there are many significant developments under way in the United States for sensors for new variables, new structures and operating principles, with improved performance, and reduced cost. However, many issues remain in addressing potential practical applications. The specialized nature of many MEMS fabrication processes makes cost-effective fabrication an issue for low to moderate volume applications. High reliability and stability in real operating environments are critical requirements for many sensors. These requirements are difficult to achieve, and continued development is required for many applications and products.
The important questions of cost, reliability, and performance go beyond the basic sensor element itself. The sensor package can have a dominant effect. New packaging and assembly schemes, including integrated wafer-level packaging, are being developed to address these issues, and are described in another section of this report. Finally, integration of sensors and electronic circuitry into more sophisticated instrumentation subsystems is an increasing trend. These smart sensors can provide compensation, linearization, output ranging, two-way communication, and many other functions.