Optically Controlled Memristor Enabling Synergistic Sensing‐Memory‐Computing for Neuromorphic Vision Systems
Jianhui Zhao, Dingxin Liu, Kangbo Zhao, Jianning Wang, Yufei Shang, Fengyuan Wang, Shuang Ma, Yifei Pei, Jianzhong Lou, Xiaobing Yan
Abstract
Abstract Neuromorphic Visual Devices hold considerable promise for integration into neuromorphic vision systems that combine sensing, memory, and computing. This potential arises from their synergistic benefits in optical signal detection and neuro‐inspired computational processes. However, current devices face challenges such as insufficient light/dark resistance ratios, mismatched transient photo‐response, and volatile retention characteristics, limiting their adaptability to complex artificial vision systems. Here, a novel optically controlled memristor is developed by leveraging the unique properties of KNbO 3 , where resistive switching originates from the dynamic reconstruction of conductive filaments controlled by light‐modulated oxygen vacancy charges. Experimental results reveal exceptional reconfigurable electrical characteristics across the 405–650 nm visible spectrum: optically triggered nonvolatile resistive switching, dual‐modal dark‐state operation, switching ratio >10 2 , endurance 6 × 10 6 cycles, retention >10 4 s, photo‐to‐dark current ratio >10 2 , and self‐powered capability. The enhanced light/dark resistance contrast significantly improves visual perception in complex illumination environments, whereas the synergistic interplay between the dynamic photo‐response and nonvolatile storage balances transient signal processing with persistent information retention. In integrated sensing‐memory‐computing tasks, the devices achieved recognition accuracies of 96% and 85% on the standard Jellyfish dataset and its low‐illumination variant, demonstrating technical feasibility and providing a theoretically grounded solution with practical implications for the development of bioinspired vision chips.