Self-rectifying dynamic memristor based on 2D-graphene/V-doped Ga2O3 for neuromorphic processing

Abstract

Memristors adjust their internal resistance in response to external stimuli, enabling random-access memory, neuromorphic computing, sensing, and in-memory processing functions that are difficult to achieve with conventional devices. These capabilities arise from diverse switching mechanisms, making mechanistic understanding and the discovery of new memristive operation modes crucial for reliable implementation and advanced applications. In this work, we present a flexible Au/V-doped a-Ga₂O₃/single-layer graphene (SL-Gr)/Cu memristor platform, benchmarked against a reference device without graphene, in which mechanical strain directly modulates the switching mechanism. The SL-Gr interlayer regulates Cu⁺ injection and diffusion so that Cu-rich nanoclusters nucleate predominantly at graphene defect and edge sites (ECM contribution), while concurrent oxygen-vacancy (V_o) enrichment and field-driven redox in the V-doped a-Ga₂O₃ contribute to a valence-change pathway (VCM contribution), yielding a controllable hybrid-filament system. Notably, under stable and flexible operation, the device exhibits key neuromorphic functions, including Pavlovian associative learning, image memorization in a 3 × 3 array of letter “C”, “F”, “I”, and “L”, 4-bit state encoding, and digit recognition within a reservoir computing (RC) framework, underscoring its potential as a promising candidate for next-generation artificial intelligent (AI) hardware.