High on/off ratio SiO2-based memristors for neuromorphic computing: understanding the switching mechanisms through theoretical and electrochemical aspects

Abstract

Memristors have emerged as promising elements for brain-inspired computing applications, yet the understanding of their switching mechanisms, particularly in valence change memristors, remains a topic of ongoing debate. We report on the SiO2-based memristors, demonstrating a high on/off ratio (>105). Particularly, this study aims to enhance the fundamental understanding of switching behaviors and mechanisms. Our approach involved an extensive investigation using finite element analysis to provide visual insights into the conductive path evolution in these memristors over the set/reset bias cycle. Electrochemical impedance spectroscopy experimentally validated the theoretical investigations by interpreting the switching behavior through the lens of the equivalent circuit. In addition, we evaluated synaptic characteristics and incorporated them into neural networks for image recognition tasks with MNIST and fashion MNIST datasets. Our comprehensive exploration of both the underlying principles and potential applications is of practical relevance to studies that aim to realize and implement SiO2-based memristors in neuromorphic computing.