Neuromorphic Computing Based on Superconductive Quantum Phase-Slip Junctions
Metadata Field | Value | Language |
---|---|---|
dc.contributor.advisor | Hamilton, Michael | |
dc.contributor.author | Cheng, Ran | |
dc.date.accessioned | 2021-11-29T14:57:14Z | |
dc.date.available | 2021-11-29T14:57:14Z | |
dc.date.issued | 2021-11-29 | |
dc.identifier.uri | https://etd.auburn.edu//handle/10415/7981 | |
dc.description.abstract | Superconductive electronics that exhibit ultra-low power consumption and high speed are good candidates for neuromorphic computing, which aims to emulate the human brain, as CMOS-based systems are approaching the bottleneck of Moore’s law. Quantum phase-slip junctions (QPSJs) are 1-D superconducting nanowires that have been identified as exact duals to Josephson junctions (JJs), based on charge-flux duality in Maxwell’s equations. These superconductive circuits that operate by the propagation of small voltage or current pulses, corresponding to propagation of single flux or charge quantum, are naturally suited for implementing spiking neuron circuits. Superconductive circuits based on QPSJs can conduct quantized charge pulses, which naturally resemble action potentials generated by biological neurons. Synaptic circuits, which work as dynamic weighted connections between two neurons, can also be realized by circuits comprised of QPSJs and magnetic Josephson junctions (MJJs) or only using QPSJs as a means of charge modulation for quantized charge propagation. A fan-out circuit uses charge-flux converters to emulate dendrites that allow a neuron to connect with many other neurons. Unlike a JJ splitter circuit that provides very limited fan-out, charge-flux converters, along with the corresponding circuitry, can provide significantly more fan-out. We present basic neuromorphic computing circuitry components, including neuron, synaptic, fan-out, and axon circuits. In addition to that, a learning circuit is introduced to explore learning functions in these systems. We use a simplified spike timing dependent plasticity (STDP) learning rule to automatically update the synaptic weight between a presynaptic and postsynaptic neuron according to their relative spike timings. Using a SPICE model developed for QPSJs, circuits were simulated in WRspice to demonstrate corresponding functionalities. An important step for the experimental realization of QPSJ-based circuits is to fabricate superconducting nanowires that exhibit coherent quantum phase-slip. Since niobium nitride (NbN) was identified as an appropriate material for QPSJ, we optimized an NbN deposition process in our available equipment and fabricated ultra-narrow NbN nanowires in search of quantum phase-slips. We investigated low-temperature transport behavior in our NbN superconducting nanowires. NbN nanowires with different dimensions on different substrates were fabricated and tested at temperatures down to 1.5 K. Resistive tails were observed for ultrathin and narrow nanowires below the superconducting critical temperature TC. These results suggest that phase-slips may exist in these test structures. This project and work presented in this dissertation provide not only multiple neuromorphic circuits using QPSJs and other superconducting technologies, such as JJs, for high-speed, low power dissipation neuromorphic computing, but also provide experimental results of NbN nanowire fabrication and characterization that show potential evidence of quantum phase-slips. These results are critical for the future development of the fabrication process of reproducible and controllable QPSJs and physical implementations of QPSJ-based neuromorphic circuits. | en_US |
dc.subject | Electrical and Computer Engineering | en_US |
dc.title | Neuromorphic Computing Based on Superconductive Quantum Phase-Slip Junctions | en_US |
dc.type | PhD Dissertation | en_US |
dc.embargo.status | NOT_EMBARGOED | en_US |
dc.embargo.enddate | 2021-11-29 | en_US |
dc.creator.orcid | 0000-0001-7917-7922 | en_US |