High-performance memristor for energy-efficient artificial optoelectronic synapse based on BiVO4 nanosheets

Abstract

Artificial optoelectronic synapses have drawn extensive attention owing to their ability to combine the optical and electrical responses. Those photosensitive materials that has been reported in memristors have the potential to be adopted in such optoelectronic synapses. Herein, optoelectronic synapses can be achieved based on two-terminal memristors by using two-dimensional (2D) BiVO4 with monoclinic scheelite (m-BiVO4) structure as the memristive layer. The fabricated memristors exhibit bipolar resistive switching (RS) characteristics with stable retention and reliable endurance performance. Furthermore, both electrical and optical synaptic functions can be achieved based on the memristors, such as long-term potentiation (LTP) and depression (LTD) stimulated by the voltages pulses as well as short-term and long-term memory under the light stimulation. More importantly, owing to the LTP and LTD functions with high symmetry, linearity and good repeatability of the LTP/LTD cycles, the artificial neural network can be simulated to achieve image recognition with a high accuracy for handwritten digits. Moreover, both the electrical and optical synapses can be used to emulate the “learning-forgetting” behaviors of human brain. Remarkably, the energy consumption per synaptic event in the optical and electrical modes can be calculated, indicating potential applications in energy-efficient synaptic devices. The m-BiVO4 nanosheets not only demonstrate good non-volatile RS behaviors but also implement energy-efficient optoelectronic synaptic functions, providing a promising strategy for applications in future neuromorphic computing.