The Reverse Diode Behavior, Negative Photoconductivity Effect and Carrier Transport Mechanisms in PbZrO 3 Antiferroelectric Memristor for Neuromorphic Computing

Abstract Neuromorphic computing, which imitates biological neural networks, shows great potential in overcoming the limitations of the traditional Von Neumann architecture, with the advantages of parallel processing, low power consumption, and adaptive learning. In this work, the high‐quality PbZrO 3 (PZO) films are prepared by the sol–gel method, and the excellent antiferroelectric properties are demonstrated in the PZO memristor coupled with unique reverse diode behavior due to polarization effect. Furthermore, the Au/PZO/FTO memristor device exhibits analog‐type resistive switching behavior, and the formation and rupture mechanisms of oxygen vacancy conductive filaments are systematically elucidated. Moreover, the PZO antiferroelectric memristor shows a unique negative photoconductivity effect revealed by femtosecond laser transient absorption spectroscopy to demonstrate the carrier transport. By utilizing the nonlinear weight update characteristics of Au/PZO/FTO devices under photoelectric modulation, a convolutional neural network based on the LeNet‐5 model is constructed, achieving efficient recognition of MNIST handwritten digits and Fashion‐MNIST clothing images, with recognition accuracies reaching 96.66% and 81.33%, respectively. These research findings show that PZO has great potential in the field of optoelectronic neuromorphic computing.