Metal oxide-based resistive switching memristors for neuromorphic computing

Abstract

Brain-inspired neuromorphic computing has gained great interest considering its advantages of massive parallelism and high energy efficiency. Metal oxide-based resistive switching memristors are emerging as promising nanodevices for the hardware implementation of neuromorphic computing owing to their structural and functional resemblance to their biological counterparts and excellent compatibility with the advanced CMOS technique. In this article, we review the recent progress in metal oxide-based memristors and their application in neuromorphic computing. First, metal oxide-based memristive devices with different structural features are summarized, namely, two-terminal and three-terminal devices. Furthermore, various memristive mechanisms are systematically discussed, such as the formation of conductive filaments, Mott transition, and ferroelectric polarization. We examine the applications of different forms of memristive devices in artificial intelligence, such as electrical and optoelectronic synapses and neurons, and neuromorphic perception systems. Finally, we discuss the challenges and prospects of materials, devices, and integrations in this rapidly progressing field of research.