Temperature dependent surface tension estimation using COSMO-RS sigma moments

• QSPR model was developed for the prediction of temperature dependent surface tension.
• Regression was done with nonlinear artificial neural network.
• COSMO-RS sigma-moments were used as molecular descriptors.

In this work, a nonlinear multivariate QSPR model based on the COSMO-RS sigma moments was presented for the estimation of the temperature dependent surface tension of various organic compounds in wide surface tension (0.07–45.08 mN m−1) and temperature (283–373 K) range. 1500 data points were used to establish, validate and test the model. An artificial neural network was developed, optimized and used as regression model. The prediction power of the new model was validated with an external data set, with a squared correlation coefficient of R2 = 0.963 and a mean absolute error of MAE = 0.81 mN m−1. The factor sensitivity and importance analyses show that all of the proposed five COSMO-RS sigma moments and the temperature are significant input variables of the ANN model and the kind of skewness of the σ-profile, the electrostatic interaction energy and the hydrogen bonding acceptor function are the most sensitive and important molecular descriptors used in the new nonlinear multivariate QSPR model.

Figure optionsDownload high-quality image (96 K)Download as PowerPoint slide