Screening of Ionic Liquids for Efficient CO2 Cycloaddition Catalysis under Mild Condition: A Combined Machine Learning and DFT Approach

Abstract

The industrial application of ionic liquid-catalyzed CO₂ cycloaddition reactions is impeded by harsh conditions. We propose a novel approach that utilizes machine learning and density functional theory (DFT) to overcome this challenge. By training regression algorithms on a data set of 10,174 experimental data points, we developed a predictive model for CO₂ solubility in ionic liquids. The random forest (RF) model exhibited exceptional accuracy, enabling the prediction of the CO₂ solubility in 1624 newly generated ionic liquids. Subsequent experimental validation confirmed the efficacy of the RF model. Moreover, employing the RF model and DFT calculation, we identified four ionic liquids with high CO₂ solubility and low energy barriers for catalytic reactions, presenting promising candidates for efficient CO₂ cycloaddition with epichlorohydrin under mild conditions. This study showcases a streamlined approach to catalyst discovery by integrating machine learning and DFT methods, offering a pathway toward sustainable CO₂ utilization.