Screening\nof Ionic Liquids for Efficient CO<sub>2</sub> Cycloaddition Catalysis\nunder Mild Condition: A Combined Machine\nLearning and DFT Approach

The industrial application of ionic liquid-catalyzed\nCO<sub>2</sub> cycloaddition reactions is impeded by harsh conditions.\nWe propose\na novel approach that utilizes machine learning and density functional\ntheory (DFT) to overcome this challenge. By training regression algorithms\non a data set of 10,174 experimental data points, we developed a predictive\nmodel for CO<sub>2</sub> solubility in ionic liquids. The random forest\n(RF) model exhibited exceptional accuracy, enabling the prediction\nof the CO<sub>2</sub> solubility in 1624 newly generated ionic liquids.\nSubsequent experimental validation confirmed the efficacy of the RF\nmodel. Moreover, employing the RF model and DFT calculation, we identified\nfour ionic liquids with high CO<sub>2</sub> solubility and low energy\nbarriers for catalytic reactions, presenting promising candidates\nfor efficient CO<sub>2</sub> cycloaddition with epichlorohydrin under\nmild conditions. This study showcases a streamlined approach to catalyst\ndiscovery by integrating machine learning and DFT methods, offering\na pathway toward sustainable CO<sub>2</sub> utilization.