Analysis of Artificial Microstructure of Li-ion Battery Electrode Using Persistent Homology and Machine Learning

Abstract

In lithium-ion batteries (LIBs), the microstructure of porous electrodes is known to significantly affect various characteristics, including charge-discharge performance, safety, and degradation behavior. Although conventional microstructural parameters such as porosity, specific surface area, average pore size and tortuosity have been employed for evaluation, the microstructural features of these complex electrodes are not necessarily fully captured by them. In this study, artificial structures that simulate electrode porosity through sphere packing were constructed, and the applicability of topological data analysis using Persistent Homology was investigated. The relationship between the extracted microstructural features and effective properties, such as effective electronic conductivity, were evaluated using a linear machine learning algorithm, and the features with high contribution were discussed. Such information could be applicable to electrode design and structural optimization.