Bio-Inspired Optical Synapse Based on MoS 2 for Neuromorphic Computing

The traditional von Neumann machine vision system has hindered the development of data-intensive technologies. Therefore, neuromorphic architectures that integrate perception, memory and processing have attracted much attention. Here, we fabricate an ITO/ZnO/MoS₂/ITO/p-Si optical synaptic device for detecting and processing optical signals. It exhibits wavelength-dependent responses in the visible range (405-650 nm) and successfully emulates short-term/long-term plasticity, learning-forgetting-relearning, and associative learning by tuning optical pulses. Further, according to the wavelength selection and memory characteristics of the device, the 3 × 5 pixel unit is constructed to achieve color extraction and perception functions. More importantly, an optical reservoir computing system is constructed by utilizing the nonlinear optical response and attenuation characteristics of the device. It still maintained a high accuracy rate of 90.91% in face recognition tasks with up to 48% noise interference. These results provide a highly promising solution for the development of a neuromorphic vision system that integrates sensory storage and computation.