A Multi‐Mode Optoelectronic Device with Bionic Visual Synaptic Functionality and Self‐Powered Capability

Advancing multimodal perception in artificial intelligence remains a challenge, requiring photodetectors with intelligent multi‐mode switching. Here, we present an innovative multifunctional photodetector based on a type‐II band‐aligned In 2 S 3 /InSe van der Waals heterojunction. It exhibits conventional photoconductive detection capabilities, achieving a light on/off current ratio of 1.75 × 10 5 under 380 nm relatively strong light illumination at a 5 V bias. It switches to a photonic synapse mode under weak light illumination, emulating human learning and forgetting behaviors, and exhibits the characteristics of visual neural synapses. Inspired by this, we simulated a biomimetic visual neural network capable of waste recognition and classification, achieving 97.77% identification accuracy. Without bias, the device exhibited self‐powered photodetection with an ultrafast response, high‐resolution imaging, and encrypted optical communication. A “waveform analysis” software was developed to decode the encrypted current signals, enabling automated data processing and secure communication. This study provides a foundation for innovative optoelectronic devices and efficient artificial vision systems, demonstrating significant potential for next‐generation multifunctional components.