A high linearity and energy-efficient artificial synaptic device based on scalable synthesized MoS2

Abstract

Synaptic devices based on 2D materials are being considered as potential solutions to mimic the behavior of synapses in neuromorphic computing. However, a scalable and CMOS complementary fabrication method of low-power-consumption 2D synaptic devices remains an important issue that hinders its actual use in neuromorphic computing applications. Here, we report a lateral memristor with high-linearity analog resistive switching behavior based on a large-scale atomic layer deposition (ALD) synthesized MoS₂ film. The stable analog resistive switching behavior of the device is proved to be modulated by lateral conductive filaments that are formed reproducibly by electric field-induced oxidation, which avoids the dependence of conventional lateral memristors on randomly occurring grain boundaries. The high linear weight update behavior (α = 1.07) in this device significantly improves the recognition accuracy in neural networks, and low power consumption (<0.3 μW) was also achieved, thus making it suitable for large-scale integrated neuromorphic circuit applications. This work demonstrates a new resistive switching phenomenon in lateral 2D material devices and shows that ALD-synthesized MoS₂ has considerable potential as a scalable fabrication method for neuromorphic computing devices.