Predicting Adsorption Coefficients at Air−Water Interfaces Using Universal Solvation and Surface Area Models

Vapor-phase molecules are adsorbed at air−water interfaces to a much greater extent than can be explained by air−water partition coefficients, indicating that interface adsorption can play an important role, and this can be very important for environmental phenomena. On the basis of a statistical thermodynamic analysis, we separate the observable free energy of adsorption into a dimensionality change and a coupling part so that the modeling effort is correctly focused on the coupling part. On the basis of this analysis, we present two kinds of models for predicting partitioning between the vapor phase and the macroscopic surface of liquid water. The first model, called SM5.0R-Surf, involves atomic surface tensions developed previously for bulk solvation in organic liquids and a set of four solvent descriptors that characterize the properties of the water layer at the air−water interface. The latter descriptors are treated as parameters that are determined empirically by optimization for a set of 85 solutes for which the air−water surface adsorption coefficients (Ki/a) are known experimentally. The resulting descriptors indicate that interfacial water has increased hydrogen-bond acidity and increased hydrogen-bond basicity as compared to bulk water. A second kind of model involves an empirical correlation of the interfacial-water partition coefficient Ki/w with the calculated van der Waals surface area, and this kind of model can be based either on experimental data, yielding the semiempirical surface area (SESA) model, or on theoretical data, yielding the semitheoretical surface area (STSA) model. The SM5.0R-Surf and STSA models should be especially useful for environmental modeling because neither model requires any experimental data about the solute, other than its molecular structure. As an example, we use the above models to calculate air−water adsorption coefficients for 24 different pesticides, chlorinated arenes, and polyaromatic hydrocarbons (PAHs). We also show that several models in the literature can be used successfully even if we substitute calculated instead of experimental data for the solute parameters that they originally required. In related work reported here, the SM5.0R parametrization for predicting free energies of solvation in organic solvents is extended to include solutes containing phosphorus. This extension is based on the experimental free energies of 13 solutes in 9 organic solvents (37 data points). The SM5.0R model extended in this way and the new SM5.0R-Surf model can therefore be used to predict the free energy of solvation at air−water interfaces and in bulk organic liquids for any solute composed of H, C, N, O, F, S, Cl, Br, I, and/or P, whereas the STSA model does not contain parameters that depend on atomic number and can, in principle, be used for any molecule.