Modeling of superconducting nanowire single-photon detectors typically requires custom simulations or finite-element analysis in one or two dimensions. Here, we demonstrate two simplified one-dimensional SPICE models of a superconducting nanowire that can quickly and efficiently describe the electrical characteristics of a superconducting nanowire. These models may be of particular use in understanding alternative architectures for nanowire detectors and readouts.

We describe a superconducting three-terminal device that uses a simple geometric effect known as current crowding to sense the flow of current and actuate a readout signal. The device consists of a "Y"-shaped current combiner, with two currents (sense and bias) entering separately through the top arms of the "Y", intersecting, and then exiting together through the bottom leg of the "Y". When current is added to or removed from one of the arms (e.g., the sense arm), the superconducting critical current in the other arm (i.e., the bias arm) is modulated. The current in the sense arm can thus be determined by measuring the critical current of the bias arm, or inversely, the sense current can be used to modulate the state of the bias arm. The dependence of the bias critical current on the sense current occurs due to the geometric current crowding effect, which causes the sense current to interact locally with the bias arm. Measurement of the critical current in the bias arm does not break the superconducting state of the sense arm or of the bottom leg, and thus, quantized currents trapped in a superconducting loop were able to be repeatedly measured without changing the state of the loop. Current crowding is a universal effect in nanoscale superconductors, and so this device has potential for applicability across a broad range of superconducting technologies and materials. More generally, any technology in which geometrically induced flow crowding exists in the presence of a strong nonlinearity might make use of this type of device.

Integrated photonic neural networks provide a promising platform for energy-efficient, high-throughput machine learning with extensive scientific and commercial applications. Photonic neural networks efficiently transform optically encoded inputs using Mach-Zehnder interferometer mesh networks interleaved with nonlinearities. We experimentally trained a three-layer, four-port silicon photonic neural network with programmable phase shifters and optical power monitoring to solve classification tasks using “in situ backpropagation,” a photonic analog of the most popular method to train conventional neural networks. We measured backpropagated gradients for phase-shifter voltages by interfering forward- and backward-propagating light and simulated in situ backpropagation for 64-port photonic neural networks trained on MNIST image recognition given errors. All experiments performed comparably to digital simulations ( > 94% test accuracy), and energy scaling analysis indicated a route to scalable machine learning.

We experimentally demonstrate in situ backpropagation in a programmable nanophotonic interferometer network, achieving inference accuracies matching digital implementations. Error gradients are computed by simultaneously measuring optical interference at intermediate network components, eliminating expensive digital computations.

We propose LightHash, the first feasible photonic cryptographic hash function for blockchain technology using programmable photonic networks. We experimentally evaluate LightHash and assess whether photonic circuits can outperform digital competitors in latency and energy efficiency.