Event-Driven Spike Encoding via Asymmetric Schottky Junction Photodiodes for Bioinspired Vision Systems

Conventional frame-based vision systems inevitably generate large amounts of redundant data owing to the continuous capture of absolute light intensity. Here, we present a bioinspired vision system that converts dynamic self-powered photocurrents into sparse event-driven spikes using MoS2-based Schottky photodiodes with asymmetric electrodes integrated with a dual-branch differential circuit. These self-powered devices exhibit a high rectification ratio (>105), fast response time (<200 μs), and ultralow dark current (<5 pA). This photon-to-spike conversion pathway enables bioinspired bidirectional temporal encoding with positive and negative spikes, matching biological synaptic time scales. When directly interfaced with spiking neural networks, the system tracks only intensity changes of dynamic targets with no response to the static background. This significantly reduces computational overhead while maintaining 93.3% accuracy in dynamic gesture recognition. This work establishes a hardware foundation for direct photon-to-spike conversion at the sensor level, enabling ultralow-power neuromorphic vision systems for real-time edge applications.