PNNL

Abstract

In recent years, computer processing capabilities have improved, allowing computationally demanding solvation models to become practically implemented. Since the determination of experimental solvation free energies for short-lived intermediates such as radicals may be difficult, expensive, and time consuming, developing computational solvation models has become an effective alternative approach. In this work, the COSMO (COnductor like Screening MOdel) continuum solvation model was used to predict solvation free energies of a set of solutes with oxo, hydroxyl, amino, and methyl functional groups in the solvent water (e=78.39). The continuum solvation model requires a cavity to be defined around a solute to predict solvation free energy (?Gs*). To reproduce experimental values, the size and shape of the cavity must reflect the strength of solute/solvent interactions. The cavity is defined as interlocking spheres around atoms or groups of atoms in the solute. The sphere radii are defined by linear functions parameterized in terms of potential derived charges known as CHELPG charges. A training set of neutral and ionic solutes was created, and coefficients in the radii definitions were fitted to reproduce experimental ?Gs* values by minimizing residuals between experimental and calculated ?Gs* values for compounds in the training set. Data used in fitting the radii definitions’ coefficients was generated by performing electronic structure and solvation computations, using varying cavity sizes. The calculations were done using Density Functional Theory with the B3LYP functional and 6-311+G** basis set in the Gaussian98 package of programs. Radii definitions reproduce ?Gs* of neutrals and singly-charged ions in training set to within experimental uncertainty and accurately predict ?Gs* of some compounds outside the training set. The mean unsigned error of this training set is 0.18 kcal/mol. These findings suggest that the protocol described here for developing potential derived charge dependent cavity definitions may successfully be extended to more functional groups to increase the applicability of the scheme and obtain greater accuracy in using continuum solvation models in predicting equilibrium properties of aqueous solutes.