Predictive Modeling of Ionic Conductivity in Solid Polymer Electrolytes

Abstract

Solid polymer electrolytes (SPEs) hold significant potential for energy storage systems, particularly in next-generation solid-state lithium-ion batteries. However, their poor ionic conductivity remains a critical challenge. This thesis investigates the relationship between the microstructural characteristics of SPEs and their ionic conductivity using machine learning techniques. Our goal is to utilize existing literature data to develop predictive models that forecast ionic conductivity based on structural attributes.We begin with a review of the fundamental properties of SPEs, focusing on ionic transport mechanisms from both theoretical and experimental perspectives. Our methodology involves extracting and synthesizing data from numerous studies to create a comprehensive dataset. Machine learning models, including neural networks, are then trained on this dataset to predict ionic conductivity. The results highlight significant correlations between specific structural features and ionic conductivity. However, the performance of our predictive models was limited by the small dataset size, leading to potential overfitting and reduced generalizability. Despite these limitations, the study demonstrates the feasibility of using machine learning to gain insights into SPE design. This research advances our understanding of the factors influencing ionic conductivity in SPEs and underscores the need for larger datasets. It also highlights the potential of combining traditional scientific approaches with modern data science techniques to drive materials innovation.