Micromechanics uses construction tools that fall into three loose categories: bulk machining, surface machining, and processing sequences that are neither bulk nor surface machining procedures. Bulk micromachining, which typically uses single crystal material with orientation and dopant-controlled etches, is the oldest micromechanical processing tool. This technology, when coupled with wafer-to-wafer bonding and wafer thinning techniques, has produced economic successes in the United States, as illustrated by Lucas NovaSensor, Inc., and the company's many products that use this technology. Research in both this processing tool and its applications continues in the United States, and is exemplified by the efforts and progress that Professor M. Schmidt of MIT has demonstrated (Parameswaran et al. 1993; Huff, Gilbert, and Schmidt 1993).
Surface micromachining is dominated by low-pressure chemical vapor deposited polysilicon and silicon nitride films with variable compositions. Deposition techniques are now sufficiently refined to allow production of films with adequate mechanical performance at many U.S. facilities. This situation has been taken advantage of by a foundry approach to micromechanics. The MCNC Electronic Technologies Division now offers a double-layer polysilicon process to the U.S. micromechanics community with user-defined geometries. A second indication of maturity in surface micromachining involves Analog Devices, Inc. This company has integrated a surface-micromachined accelerometer with electronics for automotive applications.
Difficulties in surface micromachining persist in undesired surface adhesion during processing for mechanically weak structures such as large area diaphragms. Some of these problems have been solved by using freeze-sublimation cycles based on cyclohexane, as demonstrated by the University of Wisconsin Micromechanics Group (U.S. Patent 1991). A very elegant solution has been reported by R.T. Howe and associates at the University of California at Berkeley (Mulhern, Sloane, and Howe 1993). They use liquid CO(¯2) at 25°C and 1,200 psi as their starting ambient and form a supercritical fluid at 35°C, which exits in gaseous form. Both processes avoid surface tension-induced deflections and lead to free-standing structures.
On the positive side in surface micromachining is the increasing confidence in polysilicon as a mechanical construction material. Long-term test data for 500 kHz clamped-clamped beam resonators has been reported (Guckel et al. 1993d), and indicates that material-induced drift is less than 1 ppm/month with quality factors above 100,000. These results point at excellent future sensors with direct optical to mechanical interfaces.
A very elaborate set of processing procedures that employ surface micromachining in various forms has been developed at Cornell University by Professor N.C. MacDonald's group (MacDonald 1993). The work is motivated by the interest that the Cornell group has in tunneling structures. The Cornell process has progressed from simple tunneling tips to triple-tip structures to multiple tips on moveable surface-micromachined three-axis actuators.
Very encouraging efforts in surface micromachining that use metal films instead of polysilicon as a construction material are the digital micromirror devices that Texas Instruments, Inc., is fabricating (Sampsell 1993b). These devices have been produced in array form with sizes to 768 x 576 pixels. The mechanical structure is that of a torsional mirror that is driven by a static random access memory cell (SRAM) circuit. The entire mirror array can therefore be addressed on a pixel-by-pixel basis and forms a high definition display system.
There are several efforts in the United States that do not fall into the traditional micromechanics construction tool categories. These programs deal with at least two major issues: thick photoresist processing and special processes for packaging.
In the United States, thick photoresist processing has been pursued from two perspectives: optical exposure and LIGA-like X-ray exposures. The two approaches differ in their intent: modest photoresist thicknesses of a few hundred microns, and very large height processing of several millimeters. The first category, optically- defined thick photoresist processing, is exemplified by the work of Professor M.G. Allen, Georgia Institute of Technology, who uses spin-coated, light-sensitive polyimide as the photoresist of choice (Allen 1993). The chemical stability of the polyimide is exploited by applying it over a plating base and filling the recesses via electroplating. Ferromagnetic metals have been used in this work to produce magnetic micromechanical devices.
The Micromechanics Group at the University of Wisconsin-Madison has pursued LIGA-like processing for some time. The group's photoresist of choice is polymethyl methacrylate (PMMA), which the group's mechanical engineers apply by casting and in situ polymerization with varying degrees of cross-linking. This procedure, which was first used in Germany, can produce acceptable results to 500 mm or so, and in fact has been used in Madison to produce devices with heights to 700 mm. Larger thicknesses are limited by photoresist strain, which is a consequence of this processing procedure. This difficulty has recently been overcome by using cell-casted annealed sheets of PMMA and solvent bonding them to the substrate (Guckel et al. 1993c). Figure D.3 illustrates the result. The thickness of the PMMA layer may be adjusted by mechanical machining. The process may be used for other photoresist and can produce strain-free photopolymer layers of arbitrary thickness. The technique has been used to produce LIGA-like structures with heights above 1 mm.
Figure D.3. Three-inch silicon wafer with two-inch PMMA photoresist layer of 3.2 mm thickness.
Thick photoresist processes can be exploited by additive processing, which fills the photoresist recesses. This is typically (but not always) done by electroplating. Research in this area has centered on two major problems: plating into small geometries with large heights, and repeatable properties of deposited metals and alloys. Of particular significance are magnetic properties because this type of technology lends itself to electromagnetic device construction. The fabrication of 78% Ni, 22% Fe alloys with permeabilities of 10,000 or so, and saturation flux densities of 1.0 Tesla is noteworthy and fills the need for a soft magnetic material (Guckel et al. 1993a). Materials that can be used for permanent magnets are also of concern. However, these are more difficult to process and progress is slower.
Magnetics offers many device possibilities. Emphasis at Wisconsin has been on magnetic micromotors and, in particular, machines with low friction via magnetic levitation for rotational speeds above 1 x 10(6) rpm. Figure D.4 indicates recent results.
Packaging of micromechanical devices forms a major challenge to fabrication tools. Normally these problems are left to commercial laboratories. An exception is the work of Professor K. Najafi, University of Michigan, who works on biomedical devices that must be packaged for functional testing (Ziaie et al. 1993). His recently reported hermetic packaging technology with feedthroughs demonstrates that micromechanics can be used to contribute to the solution of highly complex packaging problems.
Figure D.4. A planar magnetic micromotor with rotor diameter of 150 mm and a central shaft diameter of 35 mm. The construction material is 78% Ni, 22% Fe permalloy. The machine has been operated to 55,000 rpm.