A voltage-controlled reconfigurable memristor with dual-mode synaptic plasticity for adaptive neuromorphic computing and mechanism analysis

ABSTRACT：In this study, we report a planar-integrated memristor fabricated with a CMOScompatible HfO₂/Al₂O₃ bilayer, in which voltage-controlled redistribution of oxygen vacancies at the HfO₂/Al₂O₃ interface enables reversible switching between analog and digital resistive states within a single device. Comprehensive surface and interface analyses-including X-ray photoelectron spectroscopy (XPS) profiling, atomic force microscopy (AFM), and temperaturedependent electrical transport-reveal that low-voltage operation (<1 V) induces gradual modulation of interfacial oxygen vacancy concentration, leading to continuous conductance tuning characteristic of analog switching. In contrast, high-voltage pulses (>2 V) trigger abrupt filamentary conduction via field-driven aggregation of oxygen vacancies, resulting in binary resistive behavior. The dual-mode functionality is directly correlated with the electric-fieldmediated evolution of interfacial defect profiles and energy barriers, as supported by conduction mechanism modeling (e.g., Poole-Frenkel emission and space-charge-limited conduction).Experimental validation through simulated array configurations empirically validates the device's capacity for achieving robust compatibility with diverse neural network architectures through its multi-model operational capabilities. This functional versatility demonstrates critical crossplatform adaptability essential for neuromorphic computing implementations. This dynamically mode-switchable device between dual resistance switching modes offers a scalable solution for energy-efficient reconfigurable neuromorphic systems, demonstrating promising potential in next-generation intelligent sensing-computing co-architectures.