MEMS (microelectromechanical systems) is a set of technologies that make it possible to miniaturize and mass-produce integrated sensors, actuators, and computers, coupling computation with the physical world on a scale never before possible. From a computational perspective, this tight coupling will have two important long-term effects. First, MEMS has the potential to produce a dramatic improvement in the capabilities of computational systems. Second, large numbers of MEMS devices embedded within materials and distributed throughout the environment will create a new setting for distributed computation, posing an important new set of challenges and opportunities for information technology.
MEMS creates a new set of opportunities for mixed electrical/mechanical computing mechanisms. Today, MEMS is slowly finding its way into computer systems, with emerging applications in disk drives and optical storage applications [1,2]. Looking further down the road, it is clear that MEMS and computation will become closely intertwined. MEMS-based data storage mechanisms have already demonstrated 30 Gigabit per square inch storage densities (50 times that of CD-ROM) [3]. Similarly, arrays of micromachined STM/AFM tips enable a direct-write fabrication technology for microelectronics that has the potential to carry electronics technology well beyond the limits of optical or E-beam lithography [4]. Micromachines with electrical interfaces allow implementation of signal processing filters as microelectromechanical structures, allowing computation to be incorporated directly into sensing mechanisms [5,6]. Other potential applications range from MEMS virtual-reality displays with bio-sensor systems for creating immersive environments to on-chip MEMS optical modulation and beam-steering systems for creating parallel processing components that seek each other out and communicate without wires.
Although each of these applications of MEMS to computing is important in its own right, the real impact will come from integrating combinations of these technologies with microelectronics. An important direction for future computing research is to explore what impact this new set of capabilities will have on computer architecture and computer science. For example, the ability to integrate a mass-storage device that has an array of tens of thousands of access ports together with a processor node that includes optical, RF, and micro-machined finger-based 3D interconnection schemes, will enable very different types of parallel processor architectures than we have today -- systems will have tens of thousands of powerful nodes, each of which contains substantial data-storage and search activity. These highly integrated microelectromechanical computational systems will stress the limits of current programming paradigms so far that we may need to start thinking about computation in very different ways -- more as a problem of coordinating collections of (unreliable) interacting processes and less as a problem of specifying a set of (potentially parallel streams of) imperative programs.
Mass-production and miniaturization will make it possible to embed large numbers of MEMS devices within materials, to coat them on top of surfaces, and to distribute them throughout the environment, changing the way that we construct systems that interact with the physical world. This coupling with the physical world creates an exciting new set of opportunities to use computation to monitor and to modify the physical world. For instance, in the next 5-10 years it will become feasible to create MEMS-based airborne surveillance dust particles that float around, coordinating with each other to collect and correlate information. These distributed MEMS systems will pose new challenges for information technology that are very different from those we face today.
What is it about information technology for MEMS that's so different? After all, we have been working on the problem of coordinating large numbers of devices for a long time - in research on distributed computation. One of the things that makes it different is the coupling to the physical world, with constant interaction and adaptation to the environment. For instance, if we are going to couple computation to the physical world, the differential equations that govern the physical world are going to become part of our computer programs. So the challenge that MEMS places on information technology is not merely to coordinate lots of tiny computers, but rather to add a bit of computational behavior to materials and the environment. Other aspects of distributed MEMS systems that pose challenges for information technology range from time-varying spatial configurations to dynamic allocation of precious resources and real-time analysis of distributed data.
For further discussion of these new challenges and opportunities, see the ISAT Study on Distributed Information Systems for MEMS [7].
References
[1] Tang, W., et al., "Silicon micromachined electromagnetic microactuators for rigid disk drives", IEEE Transactions on Magnetics, Nov. 1995, vol.31, (no.6, pt.1):2964-6.
[2] Shen, J.L., et al., "A micromachined free-space integrated optical disk pickup head.", Proceedings of LEOS '95: IEEE Lasers and Electro-Optics Society 1995 Annual Meeting.
[3] Chui, B., "Improved Cantilevers for AFM Thermomechanical Data Storage", Proceedings of the 1996 Solid State Sensor and Actuator Workshop.
[4] Quate, Cal, Personal Communication
[5] Lin, L.; et al., "Microelectromechanical filters for signal processing", Proceedings of IEEE Micro Electro Mechanical Systems. An Investigation of Micro Structures, Sensors, Actuators, Machines and Robots. 1992.
[6] Nguyen, C., Micromechanical Signal Processors, Ph.D. Dissertation, University of California at Berkeley, 1994.
[7] Berlin, A., et al., Distributed Information Systems for MEMS, ISAT (Information Science and Technology Study Group) Study, 1995.
Last update: 6/11/96 by Andrew A. Berlin (berlin@parc.xerox.com)