Programmable superconducting neuron with dual-timescale plasticity for ultra-efficient neuromorphic computing

The escalating energy demands of artificial intelligence pose a critical challenge to conventional computing. Combining the efficiency of event-driven, in-memory neuromorphic architectures with the ultra-high speed and low power dissipation of superconducting circuits offers a promising solution to energy-efficient computing. However, such a solution remains elusive due to the lack of a superconducting neuron that simultaneously supports programmability, local memory, and multi-timescale plasticity. Here, we introduce a programmable Josephson-junction-based leaky integrate-and-fire (LIF) neuron that features intrinsic static memory and precise programmability by encoding neuronal parameters directly in the bias current. This neuron is also capable of dual-timescale plasticity: picosecond-scale short-term modulation of spike transmission and long-term weight retention exceeding 10 4 seconds, facilitating both rapid temporal adaptation and robust weight storage. It can operate up to 45 GHz with femtojoule-level energy dissipation per spike, and supports 10 somatic threshold levels and 20 accessible synaptic states. Furthermore, we demonstrate a crossbar-based spiking neural network (SNN) circuit utilizing this neuron, which achieves outstanding performance across multiple tasks. System-level evaluations demonstrate a 95.07% classification accuracy on the MNIST dataset and a projected energy efficiency of 93,184 GSOPS/W excluding cooling overhead, outperforming state-of-the-art CMOS neuromorphic systems by at least 144 times. By fusing computation, memory and plasticity into a single superconducting unit, our work paves the way for the next generation of ultrafast, energy-efficient neuromorphic computing.