Abstract

A Quantitative Structure-Property Relationship (QSPR) model for the prediction of surface tension of organic compounds was derived from a data set of 320 chemicals including N, O, F, Cl, Br, and/or S atoms and covering a range of about 14-45 dyn cm -1. The model, only involving six molecular descriptors obtained solely from the chemical structures, yielded an r 2 of 0.96. Its predictive capability was estimated from an external test set containing 55 structures not considered in the training set (r 2 = 0.94). It was shown that the selected molecular descriptors presented a physical meaning corresponding to the different intermolecular interactions occurring in the bulk solution. The model is applicable to a wider variety of compounds, includes less parameters and correlates better than other QSPR models reported in literature. © 2006 Taylor & Francis.