Multifunctional Synaptic and Reliable Resistive Switching Behavior in Al/Ga₂O₃ Based Memristors for Neuromorphic Computing

Neuromorphic devices with extremely low energy consumption are greatly demanded for brain-like computing and artificial intelligence (AI). In this work, the Al-Ga₂O₃ nanofilm as the active layer is used to create artificial synaptic memristor devices with memory functions, including high ON/OFF ratio, stable and filamentary resistive switching behavior, long-term/short-term plasticity (LT/STP), and "learning-experience" response. These qualities closely resemble biological learning and memory activities. Controlled production and rupture of Ag filaments result in resistive switching with a switching ratio of ~10⁴, making them ideal for nonvolatile memory demands. Before electroforming, the progressive conductance modulation of an Ag/Al-Ga₂O₃/Pt/Ti/SiO₂ memristor may be observed, and the working mechanism can be described by the subsequent development and contraction of Ag filaments induced by a redox reaction. Furthermore, the nanocomposite memristors demonstrated an exponential decay curve with a 2.40 μs decay time constant and an artificial neural network (ANN) outstanding identification accuracy of 90.5% for handwritten digits. This work suggests that proposed memristors might enable efficient neuromorphic designs.