Gate-Tunable Resistive Switching and Negative Differential Resistance in Monolayer MoS2 for Neuromorphic Computing

A monolayer MoS 2 memristor with bridge- and bisecting-domain boundaries (DBs) exhibits gate-tunable memristive behavior due to the influence of DBs on the migration kinetics of sulfur vacancy. With increasing the drain-to-source voltage bias, the rectification effect as well as the rectification ratio remain almost unchanged in bridge-DBs. However, for bisecting-DBs, an obvious bipolar resistive switching (RS) accompanying negative differential resistance (NDR) appears when the drain-to-source electric field strength is above 267 V/μm. Furthermore, a gating is effective in inducing bipolar RS and NDR in bisecting-DBs even at a lower drain-to-source electric field strength of 50 V/μm. The above behaviors can be fully understood by the generation and drift of sulfur vacancies coupled with electron trapping/detrapping. The RS and NDR in monolayer MoS 2 are analyzed in detail in a mechanism with a doubly ionized sulfur vacancy and further used for neuromorphic computing. Based on the highly linear conductance weight updates in the bisecting-DB memristor, an artificial neural network (ANN) was established with high recognition accuracies of 88.7% and 96% for the Fashion-MNIST and MNIST data sets, respectively. This lays a solid foundation for the application of two-dimensional (2D) materials in neuromorphic computing.