Research and Teaching Plan
Michael P. Frank

Research Plan

I would of course be delighted to do this work in close cooperation with any other faculty in your department or your university that are interested in this area. My research philosophy is to cooperate and collaborate with many different people, so as to facilitate new interdisciplinary insights.

Advance publications. As part of my summer postdoc work at MIT, I am currently preparing an MIT Press book based on my thesis, together with that of a fellow graduate student. (The draft material is now undergoing review.) I will also write some new papers that I have been invited to submit by a journal and two conferences, and I will be seeking additional publications as well. I will have a student research assistant for part of the summer, on a project which should lead to another publication. (In the past, I have supervised several undergraduate assistants, with quite successful results.)

Funding. Once I have secured a position at an institution where I can pursue my research, I plan to submit a proposal to DARPA for the long-term continuance of my current research project, which was originally supported under DARPA contract #DABT63-95-C-0130, as part of the Scalable Computing Systems research program. My MIT advisor, Tom Knight, has good contacts at DARPA and can help guide me in this process.

Another likely source of funding is the Department of Energy, whose "Stockpile Stewardship Program" would benefit greatly from the development of fast, massively parallel reversible mesh processors, which would be an ideal platform for lattice physical simulations.

Other potential government sponsors include NASA, NSF, NSA, NIH. All these agencies have an interest in the development of future high-performance computing technologies. NASA also has an interest in low-power technology for use in space environments, where heat removal is difficult, and energy supplies may be limited. NIH also has an interest in low-power technology for wearable/implantable medical devices, hand-held digital instruments, etc.

After the initial research program is established, I will work with industry on technology-transfer efforts for the near-term low-power applications of reversible circuit technology; hopefully this will bring in additional sources of corporate funding. In the longer term, technology transfer of designs for high-performance computing will be pursued as well.

Throughout the life of the program, there will likely also be opportunities for income through the licensing of patents, and the founding of spin-off start-up companies to commercialize the research.

Planned Lines of Research. I have many ideas in mind for further lines of research that can be fruitfully pursued in connection with the overall project. At any given time, I expect that my research group will be actively pursuing several of these directions simultaneously.

1. Adiabatic circuit designs. Work with VLSI students to improve our present reversible CMOS circuit designs and build up a larger catalog of useful circuits. This summer, I have a project to find the simplest fully-adiabatic universal logic gate that can be constructed from CMOS transistors.
2. Resonant power supplies. Work with colleagues in power and AC engineering to design more efficient, more scalable power supplies capable of delivering the resonant waveforms needed for adiabatic circuit operation.
3. Near-term low-power applications. Work with companies and interested colleagues to design and test partially-adiabatic systems for use in portable or embedded systems in energy-limited environments, and in systems such as satellites where convective cooling is unavailable. My industry experience at places like IBM and NASA should help me work effectively at these technology transfer efforts.
4. Technology capability/cost scaling projections. Work with computing industry colleagues to continuously refine our estimates of the capabilities of future computing systems, and of the cost level of machines that would gain performance from reversible operation.
5. Cooling systems for high-performance computing. Any cooling technology has limits to its capabilities. But in collaboration with colleagues in thermal engineering, we would like to see how far various cooling techniques can be pushed before their cost becomes excessive. This affects the comparison of reversible and irreversible computing techniques.
6. Semiconductor device physics. Work with colleagues in semiconductor process design on ways to make lower-resistance switches using near-term feasible fabrication technologies. Lower resistance greatly improves the relative performance of reversible circuits in dissipation-limited contexts.
7. Design of reversible nano-scale devices. Work with applied physicists in various departments to help design advanced future (post-semiconductor) logic device technologies with the need for the option of reversible operation in mind. Includes study of superconductor-based technologies such as RSFQ (rapid single flux quantum) logic and its variants.
8. Physics of computation. Work with interested colleages in theoretical physics to fine-tune our understanding of the fundamental physical limits on computation. Includes study of quantum computing technologies and algorithms.
9. Reversible programming languages. Work with computer science students to develop more sophisticated and useful programming languages capable of compiling efficiently to fully reversible hardware. Includes investigation of languages for programming the reversible parallel architectures described in my thesis.
10. Reversible algorithm design. Work with computer science and logic-design theory students to enlarge our catalog of efficient reversible algorithms for various types of common computer science/engineering/scientific computing problems. Includes work on efficient parallel reversible algorithms.
11. High-performance applications. As time goes by, we expect an increasing interest from industry in the use of low-power techniques for enabling faster, more densely-packed parallel supercomputing systems. As traditional low-power techniques reach their limits, there should be a need for an increasing amount of adiabatic circuitry. I plan to work with industry and interested colleagues on a gradual technology transfer of our reversible circuits and architectural ideas for high-performance computing into commercial designs.

Teaching Plan

My primary model for working with students is my advisor Tom Knight. He inspires students, provides what they need, guides them as necessary, and does not stand in their way. My other mentor, Norm Margolus, also has served as an excellent example; he is always readily available for students to interact with, and always happy to entertain and critique their ideas, and tweak them in productive directions.

Lecuring. In addition to graduate advising, I am quite happy to also teach lecture-style courses. In my TA work I have enjoyed teaching recitation sections, and have received good reviews from my students - "clear, concise, correct" is how my teaching has been described by students in the MIT course guide.

I personally would most enjoy teaching subjects that are related to my research, for example:

EE subjects: In electrical engineering, interesting subjects for me at present would be digital circuit theory, VLSI design and device physics, low-power techniques, microarchitecture, and architectures and communication networks for massively parallel computing.

CS subjects: I am also well-prepared to teach various subjects in computer science such as the fundamental theory of computation, theory of algorithms, programming languages, and the CS side of computer architecture. If possible, I would enjoy having a joint appointment in both the EE and CS areas at your university, and having the opportunity to teach classes in both fields.

A new interdisciplinary course: I would also like to design my own novel, advanced graduate course to convey the state of what is currently known in the area of adiabatic circuits, reversible computing, and the physical limits of computation. Together with a selection of background readings, the book that I am currently writing could evolve into an excellent text for this course.

To help with the teaching load, I would of course benefit from a reasonable level of availability of intelligent graduate teaching assistants, as I'm sure can be found at an excellent institution such as yours.