SBIR Phase I:Energy Efficient Superconducting Neuromorphic Computing Circuits

The broader impact of this Small Business Innovation Research (SBIR) Phase I project is potential commercial development of superconducting neuromorphic computing (NC) circuits with the ability to enable true biological brain-inspired deep neural network circuit algorithms and to improve efficiency, speed, and scalability of NC by orders of magnitude. The knowledge and approaches developed through this effort may help advance the foundational development of next generation computing hardware, helping NC continue its advance toward broad market adoption, and helping the US maintain its position as a leader in processor development and production. Additionally, the integrated synthesis-characterization-application approach can be extended to a range of applications, including sensors, metamaterials, catalysis, and renewables, which require atomic-scale control of materials and interfaces. Finally, through a partnership with University of Kansas, the project will facilitate university technology transfer and will serve to educate the next generation of materials and advanced electronics scientists and engineers. The atomic-to-nanoscale design, fabrication, characterization, and application experience will not only assist in recruiting top-quality students and provide them opportunities for entrepreneurship.This Small Business Innovation Research (SBIR) Phase I project seeks to develop novel superconducting neuromorphic computing (NC) circuits consisting of atomically tunable memristors (synapses) with superconductor interconnects and superconducting quantum interference devices (SQUIDs, neurons). This superconducting NC circuit aims to enable true biological brain-inspired deep network circuit algorithms and to achieve currently unattainable levels of energy efficiency, switching speed, and scalability in NC.The proposed research will design, fabricate, and characterize superconducting memristor-SQUID NC circuit hardware including development of the corresponding algorithms for pattern recognition, with machine learning capabilities, using the Modified National Institute of Standards and Technology database to prove viability. The intellectual merit of the proposed research is illustrated in: (1) novel, atomically-tunable memristors with 3-4 orders of magnitude dynamic range in the on/off ratio and switching frequency that can enable spikes of different amplitudes and frequencies as demanded for emerging deep circuits, (2) SQUID neurons with very high sensitivity and low noise, and (3) neurons and interconnects that can significantly reduce power consumption by eliminating the parasitic wire resistance that, in current NC circuits, increases substantially with circuit scale.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.