Abstract High‐density memory arrays are essential for neuromorphic computing, providing the massive parallelism and connectivity needed to emulate complex brain‐like functions with low energy consumption. Self‐rectifying memristors (SRMs), which combine rectification and resistive switching, effectively address the sneak‐path issue—a key challenge in large‐scale, high‐density 3D integration. However, lead‐free halide perovskite (LFHP)‐based SRMs remain largely unexplored, despite their mixed ionic–electronic conductivity and potential for ultralow‐power operation. Here, the first LFHP CsBi 3 I 10 ‐based SRM with a self‐organizing heterostructure enabled by halide ion migration is reported. The device achieves ultralow SET power (≈16.8 fJ), a high rectification ratio (≈6.2 × 10⁴), and fast switching (30 ns), arising from Ag/iodide vacancy conductive channels and a spontaneously formed pn junction‐like heterojunction. Functionally, the device leverages its characteristics to enable versatile applications. It realizes basic logic gates (OR, AND, XOR), achieves image encryption and reconstruction using keys generated by voltage‐driven stochastic switching, and further exhibits excellent neuromorphic computing capabilities in array configuration, achieving a recognition accuracy of ≈92.08% and supporting real‐time image edge detection. This work establishes a new device/material paradigm for low‐power, multifunctional SRMs that unify storage, logic, encryption, and neuromorphic computing—paving the way for next‐generation AI‐oriented information technologies.
Chai et al. (2025) studied this question.