Abstract
Traditional image sensors face inherent trade-offs among accuracy, power consumption, and hardware complexity in dynamic recognition tasks. Drawing inspiration from the tetrachromatic vision of reindeer, we report a Ga2O3/TiO2/Ti3C2Tx neuromorphic optoelectronic sensor. The defect-rich TiO2 interlayer, formed via native oxidation of 2D Ti3C2Tx MXene, constructs a Type-II heterojunction with Ga2O3, dramatically boosting photogenerated carrier separation. The device exhibits multi-level photoresponses across four wavelengths (254-638 nm) and faithfully emulates key synaptic behaviors, including excitatory postsynaptic current (EPSC), paired-pulse facilitation (PPF) and spike-rate-dependent plasticity (SRDP). Based on a 20 × 20 array, we develop a dual-band (UV-visible) hybrid motion-perception system. Capitalizing on the pronounced conductance plasticity, this system effectively encodes spatial and temporal optical information. When coupled with a convolutional neural network, it achieves a remarkable 95.7% accuracy in recognizing complex trajectories, significantly outperforming conventional architectures. This work provides a versatile materials platform for broadband neuromorphic photodetection and paves the way for high-precision machine vision in complex environments.

京公网安备11010802044758号
Comments on this article