Regulating oxygen vacancies to achieve stable non-volatile switching characteristics and neuromorphic computing in a-Ga2O3 based memristors

Abstract

Gallium oxide has attracted considerable interest as a promising resistive switching (RS) layer for memristors, owing to its wide bandgap, high breakdown electric field, and excellent thermal stability. However, practical adoption remains hindered by challenges such as relatively low switching ratio and insufficient cycling stability, which limit device performance and reliability. Herein, amorphous Ga₂O₃ (a-Ga₂O₃) films serve as the RS layer to construct the memristors, where oxygen vacancies are precisely regulated to enhance the switching performance. The W/a-Ga₂O₃/Pt device fabricated with an argon-to-oxygen ratio of 40:10 outperforms those made with other ratios in RS behaviors, demonstrating a larger switching window (∼10³), superior retention (10⁴ s), faster response times (90 μs), high stability, and concentrated distributions of switching voltages. Furthermore, we elucidate the electrical transport mechanisms and conductive filaments (CFs) models responsible for the enhanced switching performance. More encouragingly, the a-Ga₂O₃ device achieved up to 98.67 % accuracy on the Mixed National Institute of Standards and Technology image recognition task. This work provides an in-depth exploration of a-Ga₂O₃-based memristors, offering a promising avenue for switching materials in storage and neuromorphic computing applications.