Tri-terminal IGZO optoelectronic synapse with photoelectric co-modulation for neuromorphic computing

Abstract

Optoelectronic synaptic devices show great promise for neuromorphic computing (NC), offering high-speed and energy-efficient information processing. This study explores the potential of indium gallium zinc oxide (IGZO) based optoelectronic synapses for applications in neuromorphic computing. A systematic analysis of excitatory postsynaptic currents and inhibitory postsynaptic currents was conducted, examining various optical parameters and electrical parameters, which integrates short-term memory and long-term memory mechanisms and highlights the device's capabilities for information storage and learning. The device mimics biological synaptic plasticity, as demonstrated by experimental data. An evaluation of image recognition on the Fashion-MNIST dataset, using a confusion matrix, resulted in a recognition rate of approximately 93% after 100 training epochs. This indicates the effectiveness of pattern recognition, showcasing the potential of IGZO optoelectronic synapses for high-precision neuromorphic computing.