Neuromorphic Color Vision Systems Enabled by Bandgap-Engineered Metal-Oxide Homostructure Synapses

Neuromorphic optoelectronic synapses, which integrate both visual recognition and memory functions, are key components for advanced machine vision systems that emulate human visual perception. Traditional approaches have focused on utilizing complex heterojunction structures or circuits, which often increase device complexity and power consumption. Here, we present a novel multispectral vision system based on homostructured optoelectronic synaptic transistors using a single-layer metal oxide semiconductor. By precisely doping alkali ion( lithium ions (5%) ) into the oxide layer, we effectively control subgap states, enabling the material to detect longer wavelengths of light. This subgap engineering allows the oxide semiconductor to function as a highly efficient artificial optical synapse with non-volatile memory. We demonstrate this approach with a 7x7 pixel photosensor array, achieving superior performance in full-color image recognition. Our method offers a simplified yet highly effective design for neuromorphic vision sensors, improving both efficiency and scalability