In-Memory Computing with Self-Rectification and Dynamic Logical Reconfiguration of 12 Algorithms in a Single Halide Perovskites

Although memristor-based in-memory computing (IMC) prototypes demonstrate great progress and performance, integrating high flexibility and programmability into large-scale, high-density crossbar arrays remain a major hurdle for advanced computing systems. Herein, the execution of 12 distinct algorithms is successfully implemented in a single halide perovskite based IMC, leading to the construction of a halide perovskite memory with reconfigurable logic operation capabilities. Moreover, the device exhibits robust anti-crosstalk performance, paving the way for its potential application in crossbar integrated arrays. The work differs from common resistive switching, which needs electro-forming to shift from high-resistance state (HRS) to low-resistance state (LRS). Instead, it begins with LRS driven by ionic conduction, and the switching is controlled by reversible barriers due to ion migration and accumulation, enabling voltage magnitude and polarity to independently regulate various resistive behaviors. Additionally, mappings between environmental parameters and behavioral patterns are systematically established, providing an approach for adapting reconfigurable computing architectures to evolving conditions. This 1R-IMC device provides self-rectification and multiple reconfigurable functions, vital for flexible, programmable high-density crossbar arrays in advanced computing.