Self-Consistent Reaction Field Model for Aqueous and Nonaqueous Solutions Based on Accurate Polarized Partial Charges

A new universal continuum solvation model (where "universal" denotes applicable to all solvents), called SM8, is presented. It is an implicit solvation model, also called a continuum solvation model, and it improves on earlier SMx universal solvation models by including free energies of solvation of ions in nonaqueous media in the parametrization. SM8 is applicable to any charged or uncharged solute composed of H, C, N, O, F, Si, P, S, Cl, and/or Br in any solvent or liquid medium for which a few key descriptors are known, in particular dielectric constant, refractive index, bulk surface tension, and acidity and basicity parameters. It does not require the user to assign molecular-mechanics types to an atom or group; all parameters are unique and continuous functions of geometry. It may be used with any level of electronic structure theory as long as accurate partial charges can be computed for that level of theory; we recommend using it with self-consistently polarized Charge Model 4 or other self-consistently polarized class IV charges, in which case analytic gradients are available. The model separates the observable solvation free energy into two components:  the long-range bulk electrostatic contribution arising from a self-consistent reaction field treatment using the generalized Born approximation for electrostatics is augmented by the non-bulk-electrostatic contribution arising from short-range interactions between the solute and solvent molecules in the first solvation shell. The cavities for the bulk electrostatics calculation are defined by superpositions of nuclear-centered spheres whose sizes are determined by intrinsic atomic Coulomb radii. The radii used for aqueous solution are the same as parametrized previously for the SM6 aqueous solvation model, and the radii for nonaqueous solution are parametrized by a training set of 220 bare ions and 21 clustered ions in acetonitrile, methanol, and dimethyl sulfoxide. The non-bulk-electrostatic terms are proportional to the solvent-accessible surface areas of the atoms of the solute and have been parametrized using solvation free energies for a training set of 2346 solvation free energies for 318 neutral solutes in 90 nonaqueous solvents and water and 143 transfer free energies for 93 neutral solutes between water and 15 organic solvents. The model is tested with three density functionals and with four basis sets:  6-31+G(d,p), 6-31+G(d), 6-31G(d), and MIDI!6D. The SM8 model achieves mean unsigned errors of 0.5-0.8 kcal/mol in the solvation free energies of tested neutrals and mean unsigned errors of 2.2-7.0 kcal/mol for ions. The model outperforms the earlier SM5.43R and SM7 universal solvation models as well as the default Polarizable Continuum Model (PCM) implemented in Gaussian 98/03, the Conductor-like PCM as implemented in GAMESS, Jaguar's continuum model based on numerical solution of the Poisson equation, and the GCOSMO model implemented in NWChem.