Analyzing the efficacy of different machine learning models for property prediction of solid polymer electrolytes

In this work, we use machine learning to discover new polymer electrolytes for lithium-ion batteries. Polymer electrolytes are solid materials that can conduct ions and are safer than liquid electrolytes. However, they have lower ionic conductivity, which means they cannot transport ions as fast or as far. This limits their performance in batteries. Developing high ionic conductivity polymer electrolytes will result in high energy density and high-power density energy storage devices with rapid charging ability. Furthermore, polymer electrolytes need to have high mechanical stability to make safer solid-state batteries. However, developing new polymer electrolytes using multiple experiments is a tedious process. As such, machine learning can help in identifying the important parameters responsible for the discovery of high ionic conductivity polymer electrolytes. As part of this project, we study different copolymer electrolytes in context to their ionic conductivities and build a framework to identify the parameters governing the ionic conductivity of these electrolytes. We use different machine learning models including random forest, XGboost, KNN, linear regression, and chemprop model to predict the ionic conductivity of polymer electrolytes based on their chemical composition. The chemprop model uses a message-passing neural network, which is a type of deep learning that can learn from graph data, such as molecular structures. The model was trained on data from experimental publications, particularly, from Bradford et al. [1] that measured the ionic conductivity of different polymer electrolytes. The results showed that XG boost outperformed other models in predicting the ionic conductivity of polymer electrolytes. The significance of discovering new polymer electrolytes lies in addressing the current limitations of ionic conductivity. By identifying materials that exhibit improved conductivity, this work contributes to the development of high-performance lithium-ion batteries. Enhanced ionic conductivity translates to batteries with faster and more efficient ion transport, leading to improved battery performance and durability. This advancement is crucial for meeting the growing energy demands and ensuring the continued evolution of safe and reliable energy storage solutions.