Unraveling Dual Operational Mechanisms in an Air-Stable All Inorganic Perovskite for Nonvolatile Memory and Neuromorphic Computing
Zhiqiang Xie, Difei Zhang, Long Cheng, Chaohui Li, Jack Elia, Jianchang Wu, Jingjing Tian, Lijun Chen, Maria Antonietta Loi, Andres Osvet, Christoph J. Brabec
Abstract
Two-terminal drift memristors (nonvolatile) are widely employed to emulate biological synaptic functionalities in neuromorphic architectures. However, reliable emulations of synaptic dynamics can only be achieved through the integration of their counterparts, diffusive memristors. Moreover, the combination of drift and diffusive memristors represents a desirable approach to address the escalating demands posed by the increasing complexity of neuromorphic computing frameworks, which are still in their nascent stages. Accordingly, an air-stable inorganic perovskite memristor (RbPbI 3 ) is demonstrated with adjustable drift and diffusive modes. By employing an electroforming process, the drift-type devices demonstrate bipolar resistive switching with a large ON/OFF ratio (10 2 ), stable endurance (2000 cycles), long retention (1.2 × 10 5 s), and robust air stability. In contrast, diffusive-type devices, without an electroforming process, effectively emulate synaptic behaviors, including paired-pulse facilitation, long-term potentiation/depression, and spike-timing-dependent plasticity. Additionally, experimental data are utilized to train neural networks constructed with perovskite artificial synapses on image classification tasks. The results demonstrate accuracies of 89.24% (MNIST) and 79.10% (Fashion-MNIST) under supervised learning, closely approximating their theoretical values.