Reconfigurable Ion-Migration Driven Memristor for Multistate Neuromorphic Associative Learning

Abstract

The development of adaptive, multistate neuromorphic and photonic-memristive devices is essential for advancing intelligent systems capable of complex learning and decision-making. However, conventional devices face limitations in achieving simultaneous, tunable electrical and optical responses required for such biomimetic functions, often necessitating complex circuitry or material-specific modifications that hinder scalability and integration. Here, we present a reconfigurable, ion-migration driven WSe₂-based memristor that addresses these challenges by enabling multistate, reversible switching between photoconductive and photovoltaic states. First-principles calculations were employed to investigate the Pd migration mechanism, revealing how controlled Pd ion movement dynamically modulates the device’s band structure and contributes to its multistate functionality. Ultimately, the device achieves notable rectification ratios─up to 3 orders of magnitude─and a photovoltage modulation range of approximately ±0.5 V. These capabilities allow the device to emulate associative learning and multicondition decision-making in response to both external stimuli and internal states, directly supporting neuromorphic applications. These advancements, combined with an integration-friendly fabrication process, underscore the device’s potential for secure communication, adaptive signal processing, and scalable neuromorphic systems.