Research of non-volatile memory and neuromorphic computing based on 2D ferroelectric perovskite

Abstract

With the advancement of artificial intelligence and machine learning, electronic devices have become a burgeoning research field and a cornerstone of modern Internet science progress. Prior to the advent of computers, manual calculations were the sole reliance for task completion, albeit with no guarantee of accuracy. In the 1950s, von Neumann, recognized as the "father of modern computers," delineated the fundamental components of computers, encompassing input devices, output devices, controllers, arithmetic units, and memory. Among these, the arithmetic unit and controller collectively form the Central Processing Unit (CPU). Nonetheless, despite decades of evolution, substantial breakthroughs in computer performance based on the von Neumann architecture have proven elusive. Confronted with the technological landscape of today's intelligent era, scientists are increasingly focused on enhancing energy efficiency. Consequently, a paradigm shift in architecture is imperative to surmount the existing bottleneck, addressing issues such as the performance cost associated with frequent data transmission between memory and arithmetic units, sluggish transmission speeds, and other performance-related challenges. Diverging from the conventional von Neumann architecture, artificial intelligence chips typically engage in the collection, transmission, processing, and storage of information by emulating the neural network of the human brain for information perception and decision-making. This approach is better suited for integrating information distributed computing and storage on hardware platforms, particularly in scenarios involving multi-sensory cross-modal data and intelligent task processing applications (e.g., image and speech recognition). The pronounced advantages of artificial intelligence chips in power consumption, energy efficiency, and hardware overhead are substantial, motivating numerous scholars to further the development of such devices. The material synthesis part of this thesis focuses on the utilization of two-dimensional perovskite. Perovskite, being a novel type of optoelectronic material, possesses the advantageous characteristics of two-dimensional materials like solution processing and easy preparation, while also retaining the inherent traits of perovskite materials such as light responsiveness and low energy loss. Consequently, it has emerged as a prominent subject within the realm of material research, garnering considerable attention. In this study, Dion-Jacobson (DJ) phase 2D perovskites have been selected as the active layer material, employing the spin-coating method to facilitate the fabrication of large-area devices. The device fabrication process involves the use of a metal-perovskite-metal sandwich structure. Regarding applications, the focus lies on the advancement of new neuromorphic device technology, involving the creation of intelligent chips founded on the neuromorphic transformation of traditional CMOS, as well as the development of neuromorphic chips reliant on novel devices. Moreover, this thesis delves into the future developmental trajectory and highlights areas necessitating enhancement within the experimental process.