Prediction of Henry's law constants of CO2 in imidazole ionic liquids using machine learning methods based on empirical descriptors

In this study, a total of 160 experimental data points of Henry's law constant of CO2 in 32 imidazole ionic liquids (ILs) were collected, with the temperatures range from 283 to 350 K. Herein intuitive and explanatory descriptors related to Henry's law constant (HLC) were suggested from the 2D structural features of the ILs according to experimental experience and laws. Temperature was used as another variable due to its significant effect on Henry's law constant. Three machine learning methods were used to construct models to fast predict the HLC based on suggested descriptors. Multi-layer Perceptrowas mainly used to build the model and compared with the results of Random forest and Multiple Linear Regression after investigating the outliers and variable selection. In addition, if only one data point was left at a similar temperature and the reduced dataset was also used to build models in the same procedure, the results were not as good as those of the full dataset but still satisfactory.

Automated summary

In this study, 160 experimental data points of Henry's law constant of CO2 in 32 imidazole ionic liquids were collected, and intuitive and explanatory descriptors related to HLC were suggested based on the 2D structural features of the ILs. The significant effect of temperature on HLC was also highlighted, and three machine learning methods were used to construct models for fast prediction of HLC. The results showed that using Multi-layer Perceptron to build the model was satisfactory compared to Random Forest and Multiple Linear Regression methods.