Frontiers in MEMS Design

Kristofer S. J. Pister

Department of Electrical Engineering and Computer Sciences

University of California at Berkeley

From automobiles to operating rooms, microelectromechanical systems (MEMS) products have quietly become a part of our lives. As the capabilities of these devices become more widely recognized, the biggest obstacle to growth of MEMS applications is the design-cycle time. For example, meetings such as this one are both wonderful and frustrating. Tremendous opportunities are presented, with clear importance and immediate impact in other fields of engineering and science, and yet only a handful will be pursued. An hour's worth of technical discussion can generate several man-decades' worth of research ideas. MEMS engineers are in the enviable position of regularly having to choose from dozens of research or development projects, each of which is clearly worthy of effort and will clearly lead to new and useful products or tools. Unfortunately, choosing only some leaves the remainder—the vast majority—of projects unexplored.

The reasons for this are simple. There are not enough MEMS designers, and the design-cycle time is very long. Most major universities are working to rectify the former problem by educating a new wave of MEMS-capable students, and these are being snapped up by U.S. industry as quickly as they earn their degrees. Solving the latter problem, however, is not so simple.

To a large extent, MEMS design is still an art. Many of the products in production today were designed, with essentially no computer support, by intuition and "back-of-the-envelope" calculations. This, in itself, is not such a bad (or unusual) thing, for all design requires creativity. The problem in MEMS is that the validation or rejection of ideas typically must be done by fabrication rather than by simulation, with the result that one iteration of the design loop can take months.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 63
--> Frontiers in MEMS Design Kristofer S. J. Pister Department of Electrical Engineering and Computer Sciences University of California at Berkeley From automobiles to operating rooms, microelectromechanical systems (MEMS) products have quietly become a part of our lives. As the capabilities of these devices become more widely recognized, the biggest obstacle to growth of MEMS applications is the design-cycle time. For example, meetings such as this one are both wonderful and frustrating. Tremendous opportunities are presented, with clear importance and immediate impact in other fields of engineering and science, and yet only a handful will be pursued. An hour's worth of technical discussion can generate several man-decades' worth of research ideas. MEMS engineers are in the enviable position of regularly having to choose from dozens of research or development projects, each of which is clearly worthy of effort and will clearly lead to new and useful products or tools. Unfortunately, choosing only some leaves the remainder—the vast majority—of projects unexplored. The reasons for this are simple. There are not enough MEMS designers, and the design-cycle time is very long. Most major universities are working to rectify the former problem by educating a new wave of MEMS-capable students, and these are being snapped up by U.S. industry as quickly as they earn their degrees. Solving the latter problem, however, is not so simple. To a large extent, MEMS design is still an art. Many of the products in production today were designed, with essentially no computer support, by intuition and "back-of-the-envelope" calculations. This, in itself, is not such a bad (or unusual) thing, for all design requires creativity. The problem in MEMS is that the validation or rejection of ideas typically must be done by fabrication rather than by simulation, with the result that one iteration of the design loop can take months.

OCR for page 63
--> The goals in MEMS design, therefore, are clear: (1) shorten the MEMS design cycle, (2) make MEMS technology easily accessible to engineers in different disciplines, and (3) lower the barriers to entry. MEMS Design Because MEMS fabrication is based on integrated circuit (IC) fabrication, the design of MEMS devices ultimately comes down to two tightly coupled tasks. The designer must specify the light and dark shapes on a collection of glass plates, and he or she must specify the sequence of material deposition and removal to be used in conjunction with the transfer of these designs. The designs are transferred from the masks to the materials by photolithography and etching. This sequence of material deposition, photolithographic pattern transfer, and etching is applied repeatedly to build up the desired materials and shapes. The collection of glass plates is known as the mask set, and the specification of deposition and etching parameters is known as the process flow. Together, the mask set and process flow define the materials, geometry, and, ultimately, the performance of the device. Because there is such a variety of materials and etchants (a chemical or method for material removal) available, and since dozens of masks may be used in sequence in a given process flow, even the shape of the finished parts can be hard to predict, let alone the physical properties of the materials. For example, with three materials and three etchants to choose from in a 10-mask process, there are over three billion ways in which the process can be run. The geometry and material properties of a given layer are affected not only by the parameters of their own deposition and etching but also by the subsequent processing of other layers. Once the geometry and the material properties are known, the device physics then must be modeled. This task is complicated by the fact that MEMS devices store and exchange energy with their environment in many different ways, including electrical, mechanical, thermal, and fluidic. For a variety of reasons, the coupling among these energy domains typically is stronger for MEMS than for micro-systems, which makes it difficult either to predict device performance-based on simulation of a subset of the energy domains or to know which energy domains form the necessary set for simulation. As MEMS capabilities expand, single-device products have become less common, and collections of devices have been developed as integrated micro-systems. Integration of electromechanical elements with active electronics is a typical example of this. Such integration requires an additional stage of modeling, where the individual device models must be linked together and simulated to determine system performance.

OCR for page 63
--> Design Constraints Are a "Good" Thing? The biggest successes for computer-aided design (CAD) have been in the field of IC design, which would practically cease to exist without CAD. There is valid concern that CAD for "normal-scale" mechanical systems has not progressed as far or as fast as was originally hoped, and it is unrealistic to expect that micro-scale mechanical systems will be any different. This is an important point, and it indicates that we should be cautious in those areas where we hope to apply CAD for MEMS. Certainly, the MEMS design problem described above is a formidable one. Fortunately, from the perspective of the CAD system designer, the constraints of IC-like fabrication operate in our favor. Although there are many material, deposition, and removal options in MEMS process design, the way in which they are used is very structured and is virtually identical from process to process. This imposes a structure on the problem that can be exploited for process simulation and even process synthesis. An even more ideal constraint is the use of a single process, with the only design variable being the geometry on the masks. This is very close to the situation in IC design, where two or three classes of process have become dominant. The basic devices (transistors) fabricated in each process are qualitatively similar and differ only in the parameters needed to fit their performance with a standard ordinary differential equation (ODE) model. This imposes a serious constraint on circuit designers, who must work only with the devices available to them in a particular process. On the other hand, it imposes a structure on the IC design problem that has led to tremendous success in that field. The Frontier Currently, there are two main thrusts in CAD system design for MEMS. The first is a process-and device-oriented approach, and the second is a systems-design approach. Both approaches promise to greatly decrease the design-cycle time and to increase the probability of generating working devices in "first silicon," the first attempt at production, even for the novice designer. The Holy Grail for the process and device CAD is the ability to take a mask set and process flow, simulate the three-dimensional geometry and material properties, and solve the partial differential equations (PDEs) describing the device physics in order to determine the static and dynamic performance of the device. These tools will have an immediate impact in the MEMS industry, where much of this work currently is done by hand, and they will also open up MEMS design to a broader audience, enabling engineers to iterate through many generations of virtual prototypes in a period of hours instead of weeks or months.

OCR for page 63
--> The systems design approach is based on the existence of several commercially available standard MEMS processes offered by MCNC and MOSIS.1 Within a standard process, the PDEs that describe devices can be reduced to ODEs given parameters by key dimensions. These dimensions can be used to generate mask geometry automatically from parametric description. Networks of devices can be simulated by creating a graph with device dynamics at the nodes and energy flow through arcs connecting the nodes. This nodal analysis approach is the basis for some of the most successful electrical circuit simulation tools (SPICE, SABER). In this way, complex systems can be designed and simulated from previously tested parameter-given components. Design-cycle times can be reduced to minutes, and the probability of a functional part increased dramatically. The intent of this latter approach is to bring MEMS design to a point comparable to IC design, which will enable first-year graduate students or senior undergraduates to learn significant MEMS design skills in a single semester course. More than anything else, this wave of students and the products they design will make MEMS a ubiquitous part of our daily lives. 1   MCNC, located in Research Triangle Park, North Carolina, offers a polysilicon MEMS process, among many other services. MOSIS, located at the University of Southern California's Information Sciences Institute in Marina Del Ray, California, provides access to several IC processes that can be used for MEMS.