Halide Mixing-Driven Modulation of Synaptic Plasticity in Lead-Free Double Perovskite Memristors for Advanced Neuromorphic Computing

Abstract

Recently, silver–bismuth double halide perovskites, Cs2AgBiCl6, have been considered promising alternatives owing to their ambient stability and lower toxicity than that of lead-based perovskites. In this study, we report the synthesis of an inorganic Cs2AgBiCl6 perovskite with and without halide mixing. Devices fabricated with pure Cs2AgBiCl6 did not exhibit satisfactory switching behavior, whereas bromide (Br–) substitution at chloride (Cl–) sites enabled pronounced analog switching. The optimized Ag/Cs2AgBiCl4.8Br1.2/ITO memristor successfully emulates key synaptic functionalities, as well as demonstrating typical bipolar resistive switching behavior with low operating voltages (±1.5 V), an on/off ratio of 102, an endurance over 2000 cycles and a retention time of 104 s. Moreover, the device exhibited stable long-term learning and forgetting behaviors, with environmental stability lasting over six months. Multifunctionality was validated through edge computation, associative learning using Pavlov’s dog experiment, and decimal arithmetic operations. Moreover, an artificial neural network was trained on LTP/LTD data, achieving an impressive 94% accuracy in handwritten-digit recognition. These results highlight a promising route for lead-free halide perovskite memristors in future neuromorphic computing and neural hardware accelerators.