With traditional computing systems struggling to meet the demands of modern technology, new approaches to both hardware and architecture are becoming increasingly critical. In this work, I develop the foundation of a power-efficient alternative computing system using superconducting nanowires. Although traditionally operated as single photon detectors, superconducting nanowires host a suite of attractive characteristics that have recently inspired their use in digital circuit applications for amplification, addressing, and memory. Here, I take advantage of the electrothermal feedback that occurs in resistively shunted nanowires to develop two new technologies: (1) A multilevel memory cell made by incorporating a shunted nanowire into a superconducting loop, allowing flux to be controllably added and stored; and (2) An artificial neuron for use in spiking neural networks, consisting of two nanowire-based relaxation oscillators acting analogously to the two ion channels in a biological neuron. By harnessing the intrinsic dynamics of superconducting nanowires, these devices offer competitive energy performance and a step towards bringing memory and processing closer together on the same platform.
Karl K. Berggren, Professor of Electrical Engineering and Computer Science, MIT
Qing Hu, Professor of Electrical Engineering and Computer Science, MIT
Nancy Lynch, Professor of Computer Science and Computer Science, MIT
Thomas Ohki, Group Leader of Quantum Engineering and Computing group, Raytheon BBN Technologies
Please email the candidate for the URL and password to the Zoom defense:
Contact: etoomey at mit dot edu