The following premises articulate our position on structured methods for design of MEMS.
| VLSI Systems | Macro-mechanical Systems | MEMS |
|---|---|---|
| Single energy domain | Multiple coupled energy domains | Multiple coupled energy domanins |
| Small set of primitives; decomposition is easy | Wide range of unstructured non-modular elements; hierarchical functional decomposition is not easy | The range of elements is not as broad as macro systems; scope for limited decomposition |
| Elements are clearly distinguishable functionally and topologically | Intrinsically shared topological boundaries. No direct mapping between function and form | Same as macro systems; topological segmentation is equally hard (e.g., a fluid volume bounded by moving parts) |
| Simple interconnection rules (KVL & KCL) among the decomposed elements | Interconnection rules are complex | Same as macro systems |
| Geometry of physical artifacts is not a big issue in design | Geometry of artifacts is intrinsically tied to the function they perform. Kinematics plays a big role | Same as macro systems, but kinematic issues are not as complex at present. Predominantly monolithic compliant structures |
| Manufactured with planar lithography | Wide range of manufacturing techniques including 3-D machining | Same as VLSI systems |
Table 1 contrasts different aspects of VLSI systems, macro (traditional) mechanical systems, and MEMS.
Structured design methods exist for VLSI systems, because they operate in a single energy domain, involve a small set of primitives that possess a direct mapping between function and topology, and have simple interconnection rules (KVL & KCL). In contrast, the macro mechanical field is so diverse in its designs and manufacturing techniques that it is not easily amenable to systematic design. It involves multiple energy domains and a wide range of unstructured primitives with intrinsically shared topological boundaries (e.g., a fluid volume bounded by moving parts). Furthermore, there is no clean correlation between the physical form of the artifacts and the function they perform. MEMS devices resemble macro systems in these aspects and retain some of the complexity and unstructured nature of macro mechanical systems. However, as can be seen in the last two items in the table, there are some features that make MEMS suitable for structured design. If limited to planar microfabrication processes and the types of MEMS devices available today (which require only a limited range of motion that is easily achieved with deformable structures), the geometry and kinematic issues are less complex than for macro mechanical systems. Another feature of MEMS that facilitates systematic design is the use of photo-lithography in microfabrication. The masks and the process sequence used in microfabrication provide a common interface between both the a description of the design and the recipe for manufacture. Therefore, there is an opportunity for the development of mask synthesis and/or process synthesis programs to assist structured design.
A close look at the MEMS devices available today reveals that there is a paradigm shift from jointed rigid mechanisms to compliant mechanisms which are essentially deformable structures. It is possible to obtain, from functional specifications, conceptual designs for rigid and compliant structures systematically to support loads, and for controlled force and motion transmission, respectively [1, 2]. This has been done by suitably extending the already developed structural optimization techniques. Such methods require nominal input from the user in the form of functional specifications that include input-output forces and displacements, and have the potential to generate complete designs with enough details to proceed with mask generation automatically. Therefore, this is an area where structured design is possible.
Analysis is an important component of design. Computer aided tools for analysis lead to systematic design methods. Due to the present lack of simulation tools, the only way MEMS designers can verify their designs is by building and testing, which is expensive and time consuming. For this reason, it is extremely important to be able to perform accurate simulations, accurate both in the geometric representation of the structure, and in the underlying constitutive properties and behavioral models for the MEMS device and the associated electronics. MEMS devices of today, most of which are transducers, operate in multiple energy domains, and there is a strong coupling among these domains (e.g., coupled electro-mechanics, fluid-structure interaction). This requires specialized analysis methods for accurate 3-D simulation [5]. Furthermore, the backgrounds of MEMS designers are very diverse because of the multi-disciplinary nature of the field. Consequently, it is important to develop the analysis tools in such a way that most MEMS designers can use them comfortably. For instance, if electrical engineers need to do finite element analysis of the mechanical structure or if mechanical engineers need to simulate an electronic circuit, there should be suitable interfacing tools.
A MEMS system typically contains a variety of devices including electronics. A common representation that encompasses multiple energy domains is useful in modeling the whole system. The bond-graph notation, which is based on energy transport (or power flow) may prove to be useful in representing the entire system at the highest level. Ultimately, one seeks the dynamical behavior of the entire system. But most transducers are nonlinear, involve at least two energy domains, and operate in the large signal regime. Direct numerical simulation of the dynamics of the fully meshed distributed model of the system is computationally difficult and is very expensive. Therefore, it is necessary to reduce the number of degrees of freedom from the hundreds or thousands of degrees of freedom of the meshed 3-D model to as few as possible. Such reduced order models can then be used for system level dynamic simulation [4]. These 'macromodels' should be developed in such a way that they agree with both 3-D numerical simulation and experimental results in describing the macro behavior of the system. Macromodels can also be used to represent the behavior of a subsystem within one energy domain, or the interaction from other domains. Hence, an important goal in MEMS design is to develop means to automatically generate macromodels and insert such models into a system- level dynamic simulator.
A cautious approach should be taken in trying to establish structured representation and design rules for micromechanical systems, as mechanical systems are known to be not amenable to either systematic design or clean separation of design and fabrication. In spite of difficulties, there are some avenues in the MEMS area for structured design such as mask and/or process synthesis for a selected class of geometries, synthesis of compliant structures, and computer aided tools for quick and accurate analysis of different classes of MEMS devices.