Prediction of CO2 and N2 solubility in ionic liquids using a combination of ionic fragments contribution and machine learning methods

Abstract

Ionic liquids (ILs) with many unique features can act as green solvents to dissolve some gases. In this study, two databases are collected to predict the CO2 and N2 solubility in various kinds of ILs with different temperature and pressure ranges. Firstly, 13,055 CO2 solubility data in 164 kinds of ILs and 415 N2 solubility data in 38 kinds of ILs are established. The hundreds of ILs are divided into dozens of ionic fragments (IFs). Then, the quantitative structure–property relationship (QSPR) model is built by combining ionic fragments contribution (IFC) with support vector machine (SVM) and artificial neural network (ANN) to establish the relationship between gas solubility and ILs structure. As a result, for CO2 solubility prediction, the determination of coefficient (R2) is 0.9855 and 0.9732 for training sets by IFC-SVM and IFC-ANN, respectively, while for N2 solubility prediction, the R2 is 0.9966 and 0.9909 for training sets by IFC-SVM and IFC-ANN, respectively. The result indicates that both IFC-SVM and IFC-ANN models can accurately and reliably predict CO2 and N2 solubility in ILs, so as to guide the screening of ILs.