Neuromorphic Computing Using a Floating Gate Modulated Photoelectric Synaptic Transistor

Abstract Confronting the limitations of traditional von Neumann architectures—chiefly their inability to process unstructured data efficiently, this work proposes a new in‐memory sensing and computing paradigm exemplified by introducing a pioneering tin oxide nanowire‐based photoelectric synaptic transistor (PST) with a floating gate replicating biological synapses' functionality. Capitalizing on the nanowire structure's high surface area to volume ratio, the PST overcomes the challenges of photo response in metal oxide semiconductors. Integrating metal oxides' Photoelectric‐Phenomenon‐Coupling effect with the charge storage capacity of field‐effect transistors enables effective optical detection and weight storage. The PST performs exceptional optoelectrical and synaptic properties, including capabilities for long‐term memory, short‐term memory, paired‐pulse facilitate, and synaptic plasticity. Furthermore, by incorporating a learning mechanism, the PST achieves an impressive 94.65% accuracy in recognizing patterns in the Modified National Institute of Standards and Technology's(MNIST) handwritten digit dataset within 50 epochs. This research indicates a significant step toward intelligent computing systems closely mimicking the human brain's computational prowess.