Functional interface layer for a high-performance self-rectifying memristive device using hafnium-zirconia thin film

Abstract

With the accelerated development of artificial intelligence-oriented hardware components, research on low-power, high-density memory devices is actively being conducted. Among various memory devices, resistive switching devices with crossbar structures have been extensively researched owing to their many advantages. To address the sneak current issue that is inherent in memory devices with crossbar structures, additional selection devices have been considered. However, self-rectifying resistive switching devices are known to be advantageous for harnessing structural benefits. Although significant research has been conducted in this area and remarkable results have been published, further research is required to improve the electrical characteristics for low-power, high-density memory applications This paper introduces self-rectifying devices with low power consumption, high rectification ratios, and high reliability. By combining HfZrOx resistance-change layers and SiOx interlayers, the characteristics of self-rectifying devices were confirmed, achieving a rectification ratio of 106 and 100 % operational yield in 1 kb crossbar devices. The essential multiply-and-accumulate operations in artificial intelligence-oriented hardware components were verified, and the applicability of the device as an artificial neural network was explored through simulations.

Keywords

• Self-rectifying memristor
• Passive crossbar array
• Process-in-memory architecture
• Artificial neural network