Lead-Free FABi3I10 Self-Rectifying Memristors for In-Memory Security and Reservoir Computing

Abstract

Sneak-path leakage blocks large-scale passive memristor crossbar integration, making rectifying memristors essential to address this bottleneck. Herein, we develop a self-rectifying memristor (SRM) with an Ag/FABi3I10/ITO device architecture based on FABi3I10 active layers. Because of the regulatory effect of the spontaneously formed heterojunction interfacial barrier, the device achieves resistive switching behaviors with a rectification ratio over 105 and an ON/OFF ratio of approximately 104, as well as outstanding cycling stability exceeding 300 switching cycles. On this basis, in-situ arithmetic logic units, including half-adders and full-adders, are successfully constructed based on the proposed SRM. Leveraging the full-adder units integrated with XOR logic operations, we achieve hardware-oriented encryption and decryption for color images. Furthermore, a memristive reservoir computing system is established utilizing the SRM device, which delivers a high recognition accuracy of ∼91% on the MNIST handwritten digit dataset. This work offers a strategy for lead-free halide perovskite (LFHP) SRMs toward high-performance neuromorphic computing.