A multifunctional optoelectronic synaptic device for logic operation and neuromorphic computing

The evolution of energy-efficient neuromorphic hardware capable of mimicking biological synaptic functions is critical for next-generation artificial intelligence and edge-computing systems. However, the integration of optoelectronic synaptic functionality with chemically doped two-dimensional (2D) tungsten diselenide (WSe 2 ) remains insufficiently explored. Here, we report an optically driven neuromorphic memristor based on a tri-layer WSe 2 with engineered carrier polarity via benzyl viologen (BV) doping. The structural integrity and material quality were verified using atomic force microscopy (AFM) and Raman spectroscopy, while electrical characteristics are systematically analyzed using gate-dependent transfer curves ( I ds -V g ) and current-voltage measurements ( I ds -V ds ), supported by an energy band diagram illustrating the Fermi level shift after doping. The BV-doped WSe 2 devices demonstrate pronounced photoresponses under visible and deep-ultraviolet (DUV) illumination, with DUV-induced charge trapping enabling stable memory retention. Optical-pulse stimulation produces tunable conductance modulation, including paired-pulse facilitation (PPF) indicative of short-term plasticity, as well as long-term potentiation and depression. By controlling the DUV pulse duration, the BV-doped WSe 2 device response is increased from 25.2% to 1438.7%. The device further enables selective alphabet recognition and logic OR gate operation. Importantly, synaptic weight characteristics are implemented in an artificial neural network trained to identify handwritten digits on the MNIST dataset with an accuracy of 92%, validated through epoch-dependent learning and confusion matrix analysis. This work may provide a promising platform for low-power, optically programmable neuromorphic systems for intelligent sensing technologies.