Phenomenological Modeling on the Nonideal Factors of Memristor Based on Single-Crystalline LiNbO₃ Thin Film

As a novel device, memristors attracted great attention because of its potential in neural network computing. However, the nonideal factors of memristors, such as conductance drift and programming errors, limit their performance in practical applications. Single-crystalline LiNbO3 thin film memristor (LN memristor) exhibited good characteristics for neural network computing, but few work about the nonideal factors of the memristor has been reported. This work aims to model these nonideal factors of the LN memristor and explore the influence of these nonideal factors on the memristor-based neural network computing. We extracted key nonideal parameters from the fabricated LN memristor and established the phenomenological model. The model results agree with the measured results, which proves the validity of the model. We embedded these models into the device simulation platform to evaluate the effects of different nonideal factors on memristor-based neural network. This study provides an efficient way to model the nonideal factors of the LN memristor, which can accurately capture the complex behavior of the LN memristor in practical applications. In addition, through the modelling and analysis, researchers can better understand the mechanism of the LN memristor, so as to optimize memristor design and improve memristor performance for the neural network computing.