NER: Designing Reliable Computers Using Molecular Nanotechnology

University Of Minnesota-Twin Cities, Minneapolis MN

Investigators

Abstract

David Lilja U Minnesota 0210197 Recent work in physics, chemistry, and materials science has produced nanometer-scale structures out of exotic materials using sophisticated fabrication techniques. However, very little work has been conducted in computer engineering to investigate how to build full-scale computer systems out of these new devices. The goal of this exploratory project is to begin to develop new techniques for constructing finite state machines (FSMs) out of molecular nanodevices. FSMs are one of the fundamental building blocks of any digital computing system. In this project, we introduce NanoBoxes as a possible abstraction for constructing reliable finite state machines for use in molecular computer systems. The reliability and error characteristics of molecular nanodevices are substantially different from the corresponding characteristics of traditional silicon-based CMOS transistors. These nanodevice characteristics present new challenges to computer designers which will require an entirely new approach for designing finite state machines. The techniques we develop eventually could be used to build entire computer systems out of molecular nanodevices. This is a high-risk/high-reward project. The risk is that we need to develop new approaches for designing computers out of a large collection of devices which are still under development themselves. By observing common trends in newly developed molecular devices, we make assumptions about their weak drive capabilities and their unstable nature. While we expect that we can adapt and extend traditional space, time, and information redundancy techniques for fault-tolerance into this new domain of molecular computers, new ideas will be necessary to develop appropriate solutions. This project is high-reward, however, since by conducting computer architecture research in tandem with research on the nanodevices themselves, we will be streamlining the development process to be able to have fully-functional molecular computers more quickly than if we wait for the nanodevice research to solidify. Furthermore, while we are tailoring our techniques for molecular nanodevices, our techniques also will be applicable to error detection and correction in quantum nanodevices and in nanometer-scale conventional CMOS devices, which are becoming more fault-prone as transistor sizes shrink. This project will make substantial contributions to educational and human resource development. It will initiate the dissertation research of a Ph.D. student in electrical and computer engineering to focus their research in this new area. We also expect to involve a few M.S. students in this work and possibly provide research opportunities for undergraduate students through existing internship programs at the University of Minnesota.