Designing Microelectromechanical Systems
The previous two chapters discussed the materials and fabrication techniques required for the production of MEMS. This chapter focuses on building an information and manufacturing infrastructure that will spur the development of MEMS, specifically metrology, modeling, CAD systems, and foundry facilities.
Metrology is an area that is just being established in MEMS. The IC industry has been well supported by an extensive, constantly expanding understanding of the behavior of silicon and related materials as they are scaled down. No comparable resource exists for MEMS, however. Many areas of research are required to understand the nature and properties of materials used in MEMS. For example, an extensive knowledge-base exists of the electrical properties of polysilicon thin-films, but knowledge about their micromechanical properties is limited as is detailed knowledge of the long-term reliability of mechanically stressed polysilicon or the surface mechanics related to friction, wear, corrosion, and stress-related failure. There is a similar lack of fundamental understanding about other thin-film materials borrowed from the electrical domain and now exercised mechanically (e.g., silicon nitride, silicon dioxide, and thin-film metals). The properties (e.g., strength and surface chemistry) of materials configured at the small scales of MEMS can influence the behavior of the devices of which they are integral parts, and material behavior crosses the boundary from volume or bulk effects to surface-driven effects. For example, frictional effects, in contrast to inertial effects, take on overwhelming importance in small mechanical systems.1 As MEMS enable the creation of fluid systems with smaller and smaller flow (signal) levels, understanding the nature of fluids in micrometer-sized channels and cavities becomes crucial. In general, a thorough understanding is needed of the mechanical properties of the materials to be used in MEMS at appropriate scales. Studies of the size effects associated
with physical elements that approach the size of one-to-several grains would be useful.
Measurements of MEMS mechanical elements are a challenge to the metrology equipment currently available. For example, optical measurement systems require trade-offs between feature size and depth of focus, and scanning electron microscope (SEM) measurement systems have limited working distances and also difficulty measuring nonconducting materials. It is often desirable in MEMS to measure physical dimensions (e.g., thickness and lateral extent) and other parameters (e.g., bow, warp, or surface roughness) simultaneously. There are also cases where the parameters are not directly visible. In these cases, the characterization of MEMS has been aided by high-resolution, time-resolved, visual inspection capabilities, such as the x-ray imaging system developed at the SRI Sarnoff Laboratories. This system makes the flow through chambers and valves in microfluidic systems visible (Lanzillotto et al., 1996) and is an important tool for characterizing MEMS.
In addition to methods of testing materials and measuring structural tolerances, devices are also required that can determine the mechanical properties of MEMS devices, demonstrate mechanical-device repeatability and reliability, and facilitate quality-control practices. Package-level testing is currently the most common way to measure MEMS performance, but the development of in-process wafer-level testing is clearly necessary for low-cost manufacture. Wafer-level testing of MEMS presents special challenges that are often product dependent.
Generic test-structures that indicate basic mechanical properties of MEMS materials at the wafer-level can (and should) be developed and characterized. In the IC industry, parametric wafer-screening methods have been developed based on knowledge of the effects of process variations on the performance of electronic devices and systems. Similar methods should be developed to support the growth of MEMS. Agreement will have to be reached on (1) the structures to monitor the mechanical and materials properties that are most critical to micromechanical performance and (2) the methods of testing and characterizing them.
Methods of testing and characterizing MEMS should be standardized. The development of standards, like the standards in the American Society for Testing and Materials (ASTM) or Institute of Electrical and Electronics Engineers
(IEEE), is necessary to support commercialization and foundry technologies. An effective fabrication facility needs to be able to provide a valid assessment of the characteristics and behavior of materials and processes that potential customers can use as a basis for comparing facilities. The same parameters can be used to facilitate reasonable modeling and simulation of designs prior to fabrication. The ability to deposit, characterize, and test a material or device must not depend on the skill or equipment of a single organization.
A particular challenge for MEMS is the establishment of a self-contained, complete, and integrated modeling and simulation suite appropriate to its computational analysis requirements. Most MEMS analyses to date have required numerical techniques based on methods using discrete data. Commercial mechanical-engineering and finite-element analysis (FEA) software has proven useful for modeling a variety of parameters (e.g., displacement, stress, electric field, magnetic field, temperature, and fluid velocity) under a wide variety of conditions. For example, FEA-based modal analysis has often been used to model the mechanical-vibration modes of structures.
MEMS-specific tools will be required, however, and these tools will have to be integrated into an environment where complete structural, as well as operational, analysis can be performed. Academic research systems (e.g., MEMCAD [Senturia et al., 1992] and CAEMEMS [Crary, Juma, and Zhang, 1991]) and small-business spin-offs from academe (e.g., IntelliSense and Microcosim)-many of which are supported by general and newly CAD-focused programs at DARPA-have begun to address the need for an easy-to-use interface between modeler and numerical tools. But much remains to be done.
The need for numerical tools that deal efficiently with cross-energy domain modeling is also beginning to be addressed (e.g., Wachutka, 1995). Commercial tools are coming to market that can treat some problems involving the coupled solutions of displacement, stress, electrostatic, and temperature fields (IntelliSense). But these tools are in early development. Accurate predictions of the energy dissipation and mechanical quality factors of MEMS structures are still elusive and also require further research.
COMPUTER-AIDED DESIGN SYSTEMS
MEMS devices have not yet been designed using CAD and computer-aided engineering (CAE) tools directly, in contrast to the more mature IC devices (Antonsson, 1996). Computer tools familiar in the IC-design world, such as schematic capture, schematic-to-layout generation, automatic routing, and design verification, need to be developed for MEMS. There is also a need for software tools designed for different tasks (e.g., layout, solid modeling, discretization, numerical computation, and visualization) that can function synergistically under a consistent user interface. Newer software techniques, such as object-oriented methods, will make it easier for solutions developed in one domain to be adopted in several others. For MEMS to flourish, computer descriptions will be needed for geometric, kinematic, and field views, as well as for layout and function. Other desirable features of an evolving system for CAD/CAE for MEMS include efficient interactive operation; modularity; flexibility to allow for changes; reliability; accuracy control, including error propagation from material-property and geometric uncertainties; and methods for discretization and for estimating numerical errors. Designers will eventually need to be able to determine such information as the cost of manufacture or the expected time to failure. A MEMS compiler that can start from a user specification and produce masking and processing information as outputs also needs to be developed.
Existing commercial CAD frameworks can provide a starting point for a MEMS CAD system (Broenink, Bekkink, and Breedveld, 1992; Gilbert et al., 1993; Beerschwinger et al., 1994; Senturia, 1995), but the great diversity of MEMS devices and implementations means that libraries of parameterized MEMS devices will be very large. Systems will have to be able to manipulate and gain access to very large data banks of hundreds of types of devices, with tens to hundreds of parameters each. Thus, an efficient means of library generation, organization, and accessibility is essential. Detailed process and materials information to model and simulate MEMS devices and systems accurately are also required. Unfortunately, material parameters change from fabrication facility to fabrication facility, so the CAD package will have to keep track of where the devices are to be made.
To assure industry and government users that they will be able to manufacture future MEMS products at competitive rates, the United States will have to develop a MEMS foundry-technology base similar to the base that supports the IC markets. Several elements of the infrastructure will have to be developed concurrently to create this technology base, the most important being the CAD infrastructure, which is the backbone of the foundry interface. A processing base must be developed so that foundries will be able to provide "technology files" to prospective users. These files must adequately describe the foundry technologies in terms of layout rules, modeling and simulation parameters, and behavioral characteristics (e.g., materials parameters). Qualified data regarding behavioral characteristics need to be made available to the user, who can then independently develop a system and
submit only a design-language file to the foundry. Ultimately, a compendium of data similar to the database available for ICs (Beadle, Tsai, and Plummer, 1985) will be necessary for MEMS. An important role for foundries will be to assist in the systematic development of computer-aided processing models, which could be integrated with CAD models and would aid in determining the interrelationships among requirements and in-process inspection parameters.
Many industries have also been strengthened by adopting flexible manufacturing, which are fabrication processes for which adding or removing individual process steps is an easy design option. The underlying reason for this trend is economic because the ability to adjust to changing market conditions and customer requirements in real time is often the key to success. Flexible manufacturing principles may also be valuable to the MEMS industry, especially in cases of relatively low volumes and numbers of specialty processes.
Whereas variations in IC devices and performance can often be made through changes in layout, the physical and mechanical natures of MEMS will also require that variability in processes be available to produce families of devices. In contrast to IC manufacturing, MEMS production can be expected to be in relatively small lots, at least initially. Economies of scale can more easily be achieved if common processes are standard to a facility or foundry, and process steps unique to each design can be efficiently added or omitted. In the flexible manufacturing environment, it should be possible to include higher-capability steps and advanced materials when and where they are needed.
The advancement of flexible manufacturing for MEMS foundries presents formidable challenges. Meeting them successfully will strongly affect the commercialization of MEMS designs. One area that will particularly benefit from flexible manufacturing is higher force and displacement actuation. In many cases, achieving higher forces and displacements in MEMS requires using materials not used in the conventional IC industry (see Chapter 3). Using these materials in commercial systems will require the introduction of appropriate processing methods, and incorporating these methods into flexible manufacturing foundries would be a significant advance for MEMS.
Work on developing a foundry infrastructure includes the DARPA-supported MEMS Infrastructure programs at MCNC, Analog Devices, and Sandia and the custom prototyping and manufacturing capabilities at organizations such as IC Sensors, NovaSensor, Silicon Microstructures, and others.
The establishment of standard processes and foundries for MEMS depends on a complex set of issues but is essential in the near future to support the growth of MEMS from the prototype and low-volume commercial level to the volume-driven, low-cost commercial regime. Several standard process MEMS foundries need to be available to enable companies with no wafer-processing capabilities to enter the field. MEMS applications exist for consumer, automotive, aerospace, and medical products so access to foundries will be essential. Custom, flexible fabrication facilities must also be available, however, for users who require access and manipulation of the process to produce and optimize their products. In the IC industry, these facilities have been best organized at universities and national laboratories, where not-for-profit development is more acceptable.
Rapid development in the IC industry has been aided by the establishment of a foundry infrastructure that ensures that industry and government users will be able to manufacture IC products at competitive rates and enables companies that do not have wafer-processing capabilities to enter the field. One of the key factors in the development of the IC foundry infrastructure was the development of a CAD infrastructure that became the backbone of the foundry interface. Design methods were implemented that allowed IC designers to develop systems independently and then have them manufactured by submitting only a design-language file. The MEMS field is more complicated because of the broad range of electrical and mechanical applications, including consumer, automotive, aerospace, and medical products. Thus, several standard process MEMS foundries would have to be available and accessible, as well as custom, flexible fabrication facilities for users who require access to and manipulation of the process to produce and optimize their products.
The development of the MEMS field will be greatly aided by the development of an extensive information infrastructure that includes the adoption of a generally accepted set of standards and metrics, the establishment of advanced modeling and computer-design tools, and the establishment of foundries that can support the most promising generic MEMS processes.
The committee recognizes that realizing MEMS foundries may be difficult because many commercial companies have difficulty seeing "what's in it for them." Besides the danger of compromising proprietary know-how, companies offering a foundry service will have to commit to specific processes and reasonable turnaround schedules. In the instances where small industries have tried to accommodate MEMS foundry runs so far, the results have not been warmly appreciated. A more feasible road to at least moderate success at the present juncture appears to be using academic and government laboratories to provide foundry services. The recent expansion of the National Nanofabrication Laboratory to sites at several universities and the capabilities of national laboratories, like Sandia and Livermore, may provide opportunities for MEMS foundries of a different nature, where direct hands-on work can be done by the MEMS researcher. This kind of operation could not be as widely extended as the more traditional
foundry approach of MCNC, which interacts with users only through exchanges of software, but it may provide an interim avenue until specific areas in the MEMS field are further developed.
Conclusion. Establishing standard CAD and foundry infrastructure for MEMS is essential in the near future to support the growth of MEMS from the prototype and low-volume commercial level to the volume-driven, low-cost commercial level. The development of a MEMS foundry-technology base, similar to the base that supports ICs, would assure users that MEMS products can be manufactured at competitive rates and would enable small companies and research organizations to enter the field.
Recommendation. A MEMS CAD-infrastructure that extends from processing and basic modeling to full system design capabilities should be established. A process technology infrastructure (e.g., electrical, mechanical, fluid, chemical, etc. and their integration to form complete systems) that is widely available to MEMS designers and product engineers should be developed.