Ultra‐High Switching Ratio Memtransistor Based on Van Der Waals Heterostructures Toward Neuromorphic Computing
Wen Deng, Yimeng Yu, Xinping Yan, Yifei Li, Lisheng Wang, Jinsong Wu, Jean‐Jacques Gaumet, Wen Luo
Abstract
The exceptional resistive switching characteristics and neuromorphic computational potential of memristors are crucial for advancing information processing in both traditional and non‐traditional computing paradigms. However, the non‐ideal resistive switching behavior of conventional oxide‐based memristors hardly meets the performance requirements for neuromorphic computing applications. Besides, the two‐terminal memristors are restricted by their configuration limitations toward multi‐field/multi‐functional modulation. Herein, this article presents a 2D GaSe/MoS 2 heterojunction thin‐film transistor with four‐terminal (4‐T) tuning capability and flexible programming/erasing operations for non‐volatile storage. The heterojunction transistor demonstrates an exceptional resistance switching ratio exceeding 10 7 , an ultra‐wide modulation range of 10–10 6 , highly reliable stability, and cyclic durability. The in situ Kelvin probe force microscope and dynamic characterization reveal the conduction mediated by defect‐induced space charge limitations, as well as the tuning filling process of trap states within the channel by dual‐gate terminals. This device functions as a 4‐T artificial synapse, capable of achieving basic optoelectronic synaptic operations. The self‐denoising and pattern recognition capabilities exhibited by artificial neural networks based on this device serve as excellent examples for developing efficient and energy‐saving neuromorphic computing architectures.