Brain‐Inspired Computing Based on Large‐Scale Memristor Crossbar Arrays
Kaikai Gao, Bai Sun, Zelin Cao, Mengna Wang, Junchao Zhang, Kun Wang, Guangdong Zhou, Zhenjiang Lu, Jinyou Shao
Abstract
ABSTRACT Inspired by the efficient information processing of the human brain, the construction of biomimetic neural networks for brain‐inspired computing (BIC) has become a research hotspot. Currently, memristors are considered one of the most promising passive electronic components in the post‐Moore era due to their structural and functional resemblance to biological synapses, as well as their potential for achieving ultrahigh energy efficiency and powerful computing performance. While significant progress has been made in using memristor crossbars for neural network implementation in artificial intelligence (AI), challenges remain in fabricating large‐scale arrays from diverse resistive switching (RS) materials for various applications. The ultimate commercialization of memristors, necessitating a paradigm shift from traditional transistor dominance, also faces substantial hurdles. This review first outlines the development history of neuromorphic memristors and their typical crossbar structures, especially 3D integration. Subsequently, the structure, working principle, and key performance parameters of memristor crossbars based on different RS materials are compared in detail. Furthermore, this review systematically elaborates on the key obstacles faced by large memristor crossbars in different application scenarios, offering valuable insights into their future commercialization prospects.