Superconducting optoelectronic single-photon synapses

Monolithically integrated superconducting single-photon detectors and Josephson junctions can be used to create superconducting optoelectronic synapses with analogue weighting and temporal leaky integration of single-photon presynaptic signals Superconducting optoelectronic hardware could be used to create large-scale and computationally powerful artificial spiking neural networks. The approach combines integrated photonic components that offer few-photon, light-speed communication with superconducting circuits that offer fast, energy-efficient computation. However, the monolithic integration of photonic and superconducting devices is needed to scale this technology. Here we report superconducting optoelectronic synapses that are created by monolithically integrating superconducting nanowire single-photon detectors with Josephson junctions. The circuits perform analogue weighting and the temporal leaky integration of single-photon presynaptic signals. Synaptic weighting is implemented in the electronic domain allowing binary, single-photon communication to be maintained. Records of recent synaptic activity are locally stored as current in superconducting loops, and dendritic and neuronal nonlinearities are implemented with a second stage of Josephson circuitry. This hardware offers synaptic time constants spanning four orders of magnitude (hundreds of nanoseconds to milliseconds). The synapses are responsive to presynaptic spike rates exceeding 10 MHz and consume approximately 33 aJ of dynamic power per synapse event before accounting for cooling. This demonstration also introduces new avenues for realizing large-scale single-photon detector arrays.

Automated summary

Superconducting optoelectronic synapses with analogue weighting and temporal leaky integration can be created by integrating superconducting nanowire single-photon detectors with Josephson junctions. This hardware offers a wide range of time constants and is responsive to high spike rates.