(169da) Predicting Ionic Conductivity of Ionic Liquid and Solvent Mixtures Using Machine Learning

Ionic liquids (ILs) are a special type of molten salts that are molecular in nature and therefore can be designed to exist as liquids at low temperatures. Their high thermal and electrochemical stability, negligible volatility, low melting point and the capability to be tuned for various target properties make them advantageous to be used as solvents in industrial processes. However, low ionic conductivity, high viscosity, and high market price of many ionic liquids are a hindrance to their widespread usage. One of the ways these limitations can be addressed is by mixing ionic liquid with an organic solvent. Although there exists some work with specific IL-solvent combinations, the lack of a general predictive model encompassing various solvents and ionic liquids implies that a large number of models need to be developed for such combinations. In this work, we showcase our attempt at developing a unified ionic liquid-solvent machine learning model to predict ionic conductivity of ionic liquid-solvent mixtures. Specifically, we apply three machine learning models Random Forest, XGBoost, and Graph Neural Network coupled with molecular and graph-based structural features. We assess the effect of featurization techniques and identify important features governing ionic conductivity of ionic liquid-solvent mixtures. We will also show how these models enable high-throughput screen to identify ionic liquid-solvent mixtures exhibiting high ionic conductivity.