Extensive Databases and Group Contribution QSPRs of Ionic Liquid Properties. 3: Surface Tension

Quantitative structure–property relationships (QSPR) for
calculating temperature dependence of surface tension (σ) of
ionic liquids (ILs) in terms of group contributions (GCs) is proposed
and broadly presented. A statistical learning method including stepwise
multiple linear regression and two machine learning methods including
feed-forward artificial neural network and least-squares support vector
machine was applied to express σ as a function of GCs. The models
were developed using the largest experimental data compilation reported
thus far (570 ILs, 1008 datasets, 6114 data points). The GC assignments,
the “reference + correction” modeling scheme, as well
as the model validation protocol were adopted from the previous contributions
of the series [Paduszyński, K. Ind. Eng. Chem. Res. 2019, 58, 5322−5338; Paduszyński, K. Ind. Eng. Chem. Res. 2019, 58, 17049–17066]. The influence of the chemical family of both cation
and anion on the quality of predictions is discussed. The potential
applications of the proposed model in estimating the critical temperature
of ILs are discussed. Finally, the obtained model is confronted with
other methods reported in the literature. In particular, an extensive
comparative analysis is presented in the case of the selected QSPRs
accounting for atomic contributions and topological descriptors.