I. Introduction

Microelectromechanical systems (MEMS) are small integrated devices or systems that combine electrical and mechanical components. They range in size from the sub micrometer (or sub micron) level to the millimeter level, and there can be any number, from a few to millions, in a particular system. MEMS extend the fabrication techniques developed for the integrated circuit industry to add mechanical elements such as beams, gears, diaphragms, and springs to devices.

Examples of MEMS device applications include inkjet-printer cartridges, accelerometers, miniature robots, microengines, locks, inertial sensors, microtransmissions, micromirrors, micro actuators, optical scanners, fluid pumps, transducers, and chemical, pressure and flow sensors. New applications are emerging as the existing technology is applied to the miniaturization and integration of conventional devices.

These systems can sense, control, and activate mechanical processes on the micro scale, and function individually or in arrays to generate effects on the macro scale. The micro fabrication technology enables fabrication of large arrays of devices, which individually perform simple tasks, but in combination can accomplish complicated functions.

MEMS are not about any one application or device, nor are they defined by a single fabrication process or limited to a few materials. They are a fabrication approach that conveys the advantages of miniaturization, multiple components, and microelectronics to the design and construction of integrated electromechanical systems. MEMS are not only about miniaturization of mechanical systems; they are also a new paradigm for designing mechanical devices and systems.

The MEMS industry has an estimated $10 billion market, and with a projected 10-20% annual growth rate, it is estimated to have a $34 billion market in 2002 [1]. Because of the significant impact that MEMS can have on the commercial and defense markets, industry and the federal government have both taken a special interest in their development.

The invention of the transistor at Bell Telephone Laboratories in 1947 sparked a fast-growing microelectronic technology. Jack Kilby of Texas Instruments built the first integrated circuit (IC) in 1958 using germanium (Ge) devices. It consisted of one transistor, three resistors, and one capacitor. The IC was implemented on a sliver of Ge that was glued on a glass slide. Later that same year Robert Noyce of Fairchild Semiconductor announced the development of a planar double-diffused Si IC. The complete transition from the original Ge transistors with grown and alloyed junctions to silicon (Si) planar double-diffused devices took about 10 years. The success of Si as an electronic material was due partly to its wide availability from silicon dioxide (SiO2) (sand), resulting in potentially lower material costs relative to other semiconductors.

Since 1970, the complexity of ICs has doubled every two to three years. The minimum dimension of manufactured devices and ICs has decreased from 20 microns to the sub micron levels of today. Current ultra-large-scale-integration (ULSI) technology enables the fabrication of more than 10 million transistors and capacitors on a typical chip.

IC fabrication is dependent upon sensors to provide input from the surrounding environment, just as control systems need actuators (also referred to as transducers) in order to carry out their desired functions. Due to the availability of sand as a material, much effort was put into developing Si processing and characterization tools. These tools are now being used to advance transducer technology. Today's IC technology far outstrips the original sensors and actuators in performance, size, and cost.

Attention in this area was first focused on microsensor (i.e., microfabricated sensor) development. The first microsensor, which has also been the most successful, was the Si pressure sensor. In 1954 it was discovered that the piezoresistive effect in Ge and Si had the potential to produce Ge and Si strain gauges with a gauge factor (i.e., instrument sensitivity) 10 to 20 times greater than those based on metal films. As a result, Si strain gauges began to be developed commercially in 1958. The first high-volume pressure sensor was marketed by National Semiconductor in 1974. This sensor included a temperature controller for constant-temperature operation. Improvements in this technology since then have included the utilization of ion implantation for improved control of the piezoresistor fabrication. Si pressure sensors are now a billion-dollar industry [2].

Around 1982, the term micromachining came into use to designate the fabrication of micromechanical parts (such as pressure-sensor diaphragms or accelerometer suspension beams) for Si microsensors. The micromechanical parts were fabricated by selectively etching areas of the Si substrate away in order to leave behind the desired geometries. Isotropic etching of Si was developed in the early 1960s for transistor fabrication. Anisotropic etching of Si then came about in 1967. Various etch-stop techniques were subsequently developed to provide further process flexibility.

These techniques also form the basis of the bulk micromachining processing techniques. Bulk micromachining designates the point at which the bulk of the Si substrate is etched away to leave behind the desired micromechanical elements [3]. Bulk micromachining has remained a powerful technique for the fabrication of micromechanical elements. However, the need for flexibility in device design and performance improvement has motivated the development of new concepts and techniques for micromachining.

Among these is the sacrificial layer technique, first demonstrated in 1965 by Nathanson and Wickstrom [15], in which a layer of material is deposited between structural layers for mechanical separation and isolation. This layer is removed during the release etch to free the structural layers and to allow mechanical devices to move relative to the substrate. A layer is releasable when a sacrificial layer separates it from the substrate. The application of the sacrificial layer technique to micromachining in 1985 gave rise to surface micromachining, in which the Si substrate is primarily used as a mechanical support upon which the micromechanical elements are fabricated.

Prior to 1987, these micromechanical structures were limited in motion. During 1987-1988, a turning point was reached in micromachining when, for the first time, techniques for integrated fabrication of mechanisms (i.e. rigid bodies connected by joints for transmitting, controlling, or constraining relative movement) on Si were demonstrated. During a series of three separate workshops on microdynamics held in 1987, the term MEMS was coined. Equivalent terms for MEMS are microsystems (preferred in Europe) and micromachines (preferred in Japan).

III. Fabrication Technologies

The three characteristic features of MEMS fabrication technologies are miniaturization, multiplicity, and microelectronics. Miniaturization enables the production of compact, quick-response devices. Multiplicity refers to the batch fabrication inherent in semiconductor processing, which allows thousands or millions of components to be easily and concurrently fabricated. Microelectronics provides the intelligence to MEMS and allows the monolithic merger of sensors, actuators, and logic to build closed-loop feedback components and systems. The successful miniaturization and multiplicity of traditional electronics systems would not have been possible without IC fabrication technology. Therefore, IC fabrication technology, or microfabrication, has so far been the primary enabling technology for the development of MEMS. Microfabrication provides a powerful tool for batch processing and miniaturization of mechanical systems into a dimensional domain not accessible by conventional (machining) techniques. Furthermore, microfabrication provides an opportunity for integration of mechanical systems with electronics to develop high-performance closed-loop-controlled MEMS.

Advances in IC technology in the last decade have brought about corresponding progress in MEMS fabrication processes. Manufacturing processes allow for the monolithic integration of microelectromechanical structures with driving, controlling, and signal-processing electronics. This integration promises to improve the performance of micromechanical devices as well as reduce the cost of manufacturing, packaging, and instrumenting these devices [7].

Any discussion of MEMS requires a basic understanding of IC fabrication technology, or microfabrication, the primary enabling technology for the development of MEMS. The major steps in IC fabrication technology are film growth, doping, lithography, etching, dicing, and packaging.

Film growth: Usually, a polished Si wafer is used as the substrate, on which a thin film is grown. The film, which may be epitaxial Si, SiO2, silicon nitride (Si3N4), polycrystalline Si (polysilicon), or metal, is used to build both active or passive components and interconnections between circuits.

Doping: To modulate the properties of the device layer, a low and controllable level of an atomic impurity may be introduced into the layer by thermal diffusion or ion implantation.

Lithography: A pattern on a mask is then transferred to the film by means of a photosensitive (i.e., light sensitive) chemical known as a photoresist. The process of pattern generation and transfer is called photolithography. A typical mask consists of a glass plate coated with a patterned chromium (Cr) film.

Etching: Next is the selective removal of unwanted regions of a film or substrate for pattern delineation. Wet chemical etching or dry etching may be used. Etch-mask materials are used at various stages in the removal process to selectively prevent those portions of the material from being etched. These materials include SiO2, Si3N4, and hard-baked photoresist.

Dicing: The finished wafer is sawed or machined into small squares, or dice, from which electronic components can be made.

Packaging: The individual sections are then packaged, a process that involves physically locating, connecting, and protecting a device or component. MEMS design is strongly coupled to the packaging requirements, which in turn are dictated by the application environment.

Bulk micromachining is an extension of IC technology for the fabrication of 3D structures. Bulk micromachining of Si uses wet- and dry-etching techniques in conjunction with etch masks and etch stops to sculpt micromechanical devices from the Si substrate. The two key capabilities that make bulk micromachining a viable technology are:

1) Anisotropic etchants of Si, such as ethylene-diamine and pyrocatechol (EDP), potassium hydroxide (KOH), and hydrazine (N2H4). These preferentially etch single crystal Si along given crystal planes.

2) Etch masks and etch-stop techniques that can be used with Si anisotropic etchants to selectively prevent regions of Si from being etched. Good etch masks are provided by SiO2 and Si3N4, and some metallic thin films such as Cr and Au (gold).

A drawback of wet anisotropic etching is that the microstructure geometry is defined by the internal crystalline structure of the substrate. Consequently, fabricating multiple, interconnected micromechanical structures of free-form geometry is often difficult or impossible. Two additional processing techniques have extended the range of traditional bulk micromachining technology: deep anisotropic dry etching and wafer bonding. Reactive gas plasmas can perform deep anisotropic dry etching of Si wafers, up to a depth of a few hundred microns, while maintaining smooth vertical sidewall profiles. The other technology, wafer bonding, permits a Si substrate to be attached to another substrate, typically Si or glass. Used in combination, anisotropic etching and wafer bonding techniques can construct 3D complex microstructures such as microvalves and micropumps [3].

Surface micromachining enables the fabrication of complex multicomponent integrated micromechanical structures that would not be possible with traditional bulk micromachining. This technique encases specific structural parts of a device in layers of a sacrificial material during the fabrication process. The substrate wafer is used primarily as a mechanical support on which multiple alternating layers of structural and sacrificial material are deposited and patterned to realize micromechanical structures. The sacrificial material is then dissolved in a chemical etchant that does not attack the structural parts. The most widely used surface micromachining technique, polysilicon surface micromachining, uses SiO2 as the sacrificial material and polysilicon as the structural material.

At the University of Wisconsin at Madison, polysilicon surface micromachining research started in the early 1980s in an effort to create high-precision micro pressure sensors. The control of the internal stresses of a thin film is important for the fabrication of microelectromechanical structures. The microelectronic fabrication industry typically grows polysilicon, silicon nitride, and silicon dioxide films using recipes that minimize time. Unfortunately, a deposition process that is optimized to speed does not always create a low internal stress film. In fact, most of these films have internal stresses that are highly compressive (tending to contract). A freestanding plate of highly compressive polysilicon that is held at all its edges will buckle (i.e., collapse or give way). This is highly undesirable. The solution is to modify the film deposition process to control the internal stress by making it stress-free or slightly tensile.

One way to do this is to dope the film with boron, phosphorus, or arsenic. However, a doped polysilicon film is conductive, and this property may interfere with the mechanical devices incorporated electronics. Another problem with doped polysilicon is that it is roughened by hydrofluoric acid (HF), which is commonly used to free sections of the final mechanical device from the substrate. Rough polysilicon has different mechanical properties than smooth polysilicon. Therefore, the amount of roughening must be taken into account when designing the mechanical parts of the micro device.

A better way to control the stress in polysilicon is through post annealing, which involves the deposition of pure, fine-grained, compressive (i.e., can be compressed) polysilicon. Annealing the polysilicon after deposition at elevated temperatures can change the film to be stress-free or tensile. The annealing temperature sets the film's final stress. After this, electronics can then be incorporated into polysilicon films through selective doping, and hydrofluoric acid will not change the mechanical properties of the material [16].

Deposition temperature and the film's silicon to nitride ratio can control the stress of a silicon nitride (Si3N4) film. The films can be deposited in compression, stress-free, or in tension [6].

Deposition temperature and post annealing can control silicon dioxide (SiO2) film stress. Because it is difficult to control the stress of SiO2 accurately, SiO2 is typically not used as a mechanical material by itself, but as electronic isolation or as a sacrificial layer under polysilicon.

In the micromolding process, microstructures are fabricated using molds to define the deposition of the structural layer. The structural material is deposited only in those areas constituting the microdevice structure, in contrast to bulk and surface micromachining, which feature blanket deposition of the structural material followed by etching to realize the final device geometry. After the structural layer deposition, the mold is dissolved in a chemical etchant that does not attack the structural material. One of the most prominent micromolding processes is the LIGA process. LIGA is a German acronym standing for lithographie, galvanoformung, und abformung (lithography, electroplating, and molding). This process can be used for the manufacture of high-aspect-ratio 3D microstructures in a wide variety of materials, such as metals, polymers, ceramics, and glasses. Photosensitive polyimides are also used for fabricating plating molds. The photolithography process is similar to conventional photolithography, except that polyimide works as a negative resist [3].

IV. Applications of MEMS

Here are some examples of MEMS technology:

MEMS pressure microsensors typically have a flexible diaphragm that deforms in the presence of a pressure difference. The deformation is converted to an electrical signal appearing at the sensor output. A pressure sensor can be used to sense the absolute air pressure within the intake manifold of an automobile engine, so that the amount of fuel required for each engine cylinder can be computed. In this example, piezoresistors are patterned across the edges of a region where a silicon diaphragm will be micromachined. The substrate is etched to create the diaphragm. The sensor die is then bonded to a glass substrate, creating a sealed vacuum cavity under the diaphragm. The die is mounted on a package, where the topside of the diaphragm is exposed to the environment. The change in ambient pressure forces the downward deformation of the diaphragm, resulting in a change of resistance of the piezoresistors. On-chip electronics measure the resistance change, which causes a corresponding voltage signal to appear at the output pin of the sensor package [3].

Accelerometers are acceleration sensors. An inertial mass suspended by springs is acted upon by acceleration forces that cause the mass to be deflected from its initial position. This deflection is converted to an electrical signal, which appears at the sensor output. The application of MEMS technology to accelerometers is a relatively new development.

One such accelerometer design is discussed by DeVoe and Pisano (2001) [8]. It is a surface micromachined piezoelectric accelerometer employing a zinc oxide (ZnO) active piezoelectric film. The design is a simple cantilever structure, in which the cantilever beam serves simultaneously as proof mass and sensing element. One of the fabrication approaches developed is a sacrificial oxide process based on polysilicon surface micromachining, with the addition of a piezoelectric layer atop the polysilicon film. In the sacrificial oxide process, a passivation layer of silicon dioxide and low-stress silicon nitride is deposited on a bare silicon wafer, followed by 0.5 micron of liquid phase chemical vapor deposited (LPCVD) phosphorous-doped polysilicon. Then, a 2.0-micron layer of phosphosilicate glass (PSG) is deposited by LPCVD and patterned to define regions where the accelerometer structure will be anchored to the substrate. The PSG film acts as a sacrificial layer that is selectively etched at the end to free the mechanical structures. A second layer of 2.0-micron-thick phosphorus-doped polysilicon is deposited via LPCVD on top of the PSG, and patterned by plasma etching to define the mechanical accelerometer structure. This layer also acts as the lower electrode for the sensing film. A thin layer of silicon nitride is next deposited by LPCVD, and acts as a stress-compensation layer for balancing the highly compressive residual stresses in the ZnO film. By varying the thickness of the Si3N4 layer, the accelerometer structure may be tuned to control bending effects resulting from the stress gradient through the device thickness. A ZnO layer is then deposited on the order of 0.5 micron, followed by sputtering of a 0.2-micron layer of platinum (Pt) deposited to form the upper electrode. A rapid thermal anneal is performed to reduce residual stresses in the sensing film. Afterwards, the Pt, Si3N4, and ZnO layers are patterned in a single ion milling etch step, and the devices are then released by passivating the ZnO film with photoresist, and immersing the wafer in buffered hydrofluoric acid, which removes the sacrificial PSG layer [8].

Inertial sensors are a type of accelerometer and are one of the principal commercial products that utilize surface micromachining. They are used as airbag-deployment sensors in automobiles, and as tilt or shock sensors. The application of these accelerometers to inertial measurement units (IMUs) is limited by the need to manually align and assemble them into three-axis systems, and by the resulting alignment tolerances, their lack of in-chip analog-to-digital conversion circuitry, and their lower limit of sensitivity. A three-axis force-balanced accelerometer has been designed at the University of California, Berkeley, [8] to overcome some of these limitations. The accelerometer was designed for the integrated MEMS/CMOS technology. This technology involves a manufacturing technique where a single-level (plus a second electrical interconnect level) polysilicon micromachining process is integrated with 1.25-micron CMOS.

A three-level polysilicon micromachining process [10,11] has enabled the fabrication of devices with increased degrees of complexity. The process includes three movable levels of polysilicon, each separated by a sacrificial oxide layer, plus a stationary level. Operation of the small gears at rotational speeds greater than 300,000 rpm has been demonstrated. Microengines can be used to drive the wheels of microcombination locks. They can also be used in combination with a microtransmission to drive a pop-up mirror out of a plane. This device is known as a micromirror.

MEMS IC fabrication technologies have also allowed the manufacture of microtransmissions using sets of small and large gears interlocking with other sets of gears to transfer power.

A recently developed MicroStar cross-connect fabric developed by Bell Labs [12], a micro-optoelectromechanical system device, is based on MEMS technology. The most pervasive bottlenecks for communications carriers are the switching and cross-connect fabrics that switch, route, multiplex, demultiplex, and restore traffic in optical networks. The optical transmission systems move information as photons, but switching and cross-connect fabrics until now have been largely electronic, requiring costly and time-consuming bandwidth-limiting optical-to-electronic-to-optical conversions at every network connection and cross point. MicroStar is composed of 256 mirrors, each one 0.5 mm in diameter, spaced 1 mm apart, and covering less than 1 square inch of silicon. The mirrors sit within the router so that only one wavelength can illuminate any one mirror. Each mirror can tilt independently to pass its wavelength to any of 256 input and output fibers. The mirror arrays are made using a self-assembly process that causes a frame around each mirror to lift from the silicon surface and lock in place, positioning the mirrors high enough to allow a range of movement. MicroStar is part of Lucent Technology's Lambda Router cross-connect system aimed at helping carriers deliver vast amounts of data unimpeded by conventional bottlenecks.

As a final example, MEMS technology has been used in fabricating vaporization microchambers for vaporizing liquid microthrusters for nanosatellites [13]. The chamber is part of a microchannel with a height of 2-10 microns, made using silicon and glass substrates. The nozzle is fabricated in the silicon substrate just above a thin-film indium tin oxide heater deposited on glass.

V. The Future

Each of the three basic microsystems technology processes we have seen, bulk micromachining, sacrificial surface micromachining, and micromolding/LIGA, employs a different set of capital and intellectual resources. MEMS manufacturing firms must choose which specific microsystems manufacturing techniques to invest in [14].

MEMS technology has the potential to change our daily lives as much as the computer has. However, the material needs of the MEMS field are at a preliminary stage. A thorough understanding of the properties of existing MEMS materials is just as important as the development of new MEMS materials.

Future MEMS applications will be driven by processes enabling greater functionality through higher levels of electronic-mechanical integration and greater numbers of mechanical components working alone or together to enable a complex action. Future MEMS products will demand higher levels of electrical-mechanical integration and more intimate interaction with the physical world. The high up-front investment costs for large-volume commercialization of MEMS will likely limit the initial involvement to larger companies in the IC industry. Advancing from their success as sensors, MEMS products will be embedded in larger non-MEMS systems, such as printers, automobiles, and biomedical diagnostic equipment, and will enable new and improved systems [3].

References

1. Micromachine Devices, European Study Sees MEMS Market at More Than $34 billion by 02, May 1997, p. 1.

2. Micromachine Devices I(2) (1996)

3. M. Mehregany and S. Roy, Introduction to MEMS, 2000, Microengineering Aerospace Systems, El Segundo, CA, Aerospace Press, AIAA, Inc., 1999.

4. M. Mehregany, K. J. Gabriel, and W. S. N. Trimmer, Integrated fabrication of polysilicon mechanisms, IEEE Transactions on Electron Devices ED-35, 719-723 (June 1988).

5. L. S. Fan, Y. C. Tai, and R. S. Muller, Integrated movable micromechanical structures for sensors and actuators, IEEE Transactions on Electron Devices ED-35, 724-730 (June 1988).

6. Sekioto, Yoshihara, and Ohkubo, Silicon nitride single layer X-ray mask, Journal of Vacuum Science and Technology, 21(4), Nov./Dec., 1982, pp. 1017-1021.

7. http://www.sandia.gov/mems/micromachine/overview.html. Sandia National Laboratories, Albuquerque, New Mexico 87185.

8. Don L. DeVoe and Albert P. Pisano, Surface micromachined piezoelectric accelerometers (PiXLs), Journal of Microelectromechanical Systems, vol. 10, no. 2, June 2001, pp. 180-186).

9. M. Lemkin, M. Ortiz, N. Wongkomet, B. Boser, and J. Smith, A 3-axis surface micromachined sigma-delta accelerometer, Proc. ISSCC 97, pp. 202-203, 1997.

10. R. Nasby, J. Sniegowski, J. Smith, S. Montague, C. Barron, W. Eaton, P. Mc Whorter, D. Hetherington, C. Apblett, and J. Fleming, Application of chemical-mechanical polishing to planarization of surface-micromachined devices, Proc. Solid-State Sensor and Actuator Workshop, pp. 48-53, 1996.

11. J. Sniegowski, Multi-level polysilicon surface micromachining technology: Applications and issues, ASME International Mechanical Engineering Congress and Exposition, 1996.

12. David Bishop, Randy Giles, and Charles Roxlo, Micromirrors relieve communications bottlenecks, Photonics Spectra, vol. 34, no. 3, Mar. 2000, pp. 167-169.

13. Koji Takahashi, Hirofumi Sakai, Kunihito Nagayama, Akihito Baba, Kenji Makihara, Tanemesa Asano, Vaporization and nucleation on microheater in microchannel with nozzle, AIAA Thermophysics Conference, 35th, Anaheim, CA, June 11-14, 2001, AIAA Paper 2001-3075).

14. James H. Smith and Steven T. Walsh, Selecting a process paradigm for an emergent disruptive technology: Evidence from the emerging microsystems technology base, 1997.

15. H. C. Nathanson and R. A. Wickstrom, A resonant-gate silicon surface transistor with high Q bandpass properties Applied Physics Letters 7, 84 (1965)

16. http://mems.engr.wisc.edu/polysilicon.html. University of Wisconsin, Madison, Wisconsin.