We present eight new parameterizations of the SM5.42R solvation model: in particular we present parameterizations for HF/MIDI!, HF/6-31G*, HF/6-31+G*, HF/cc-pVDZ, AM1, PM3, BPW91/MIDI!, and B3LYP/MIDI!. Two of the new cases are parameterized using the reaction-®eld operator presented previously, and six of the new cases are parameterized with a simpli®ed reaction-®eld operator; results obtained by the two methods are compared for selected examples. For a training set of 2135 data for 275 neutral solutes containing H, C, N, O, F, S, P, Cl, Br, and I in 91 solvents (water and 90 nonaqueous solvents), seven of the eight new parameterizations give mean unsigned errors in the range 0.43±0.46 kcal/mol, and the eighth ± for a basis set containing diuse functions ± gives a mean unsigned error of 0.53 kcal/mol. The mean unsigned error for 49 ionic solutes (containing the same elements) in water is 3.5±3.9 kcal/mol for the Hartree±Fock, Becke±Perdew± Wang-1991 and Becke three-parameter Lee±Yang±Parr cases and 4.1 and 4.0 kcal/mol for parameterized model 3 and Austin model 1, respectively. The methods are tested for sensitivity of solvation free energies to geometry and for predicting partition coecients of carbonates, which were not included in the training set.
The SM5.0R model for predicting solvation energies using only geometry-dependent atomic surface tensions was developed previously for aqueous solution. Here we extend it to organic solvents. The method is based on gas-phase geometries and exposed atomic surface areas; electrostatics are treated only implicitly so a wave function or charge model is not required (which speeds up the calculations by about 2 orders of magnitude). The SM5.0R model has been parametrized for solvation free energies of solutes containing H, C, N, O, F, S, Cl, Br, and I. The training set for organic solvents consists of 227 neutral solutes in 90 organic solvents for a total of 1836 data points. The method achieves a mean unsigned error of about 0.4 kcal/mol when applied using gas-phase geometries calculated at either the Hartree-Fock level with a heteroatom-polarized valence-double-ζ basis set (HF/MIDI!) or when applied using semiempirical molecular orbital gas-phase geometries. In related work reported here, the parametrization for predicting aqueous solvation free energies is also extended to include organic solutes containing iodine. This extension is based on eight solutes and yields a mean unsigned error of 0.25 kcal/mol. The resulting SM5.0R model for solvation energies in aqueous and organic solvents can therefore be used to predict partition coefficients (including log P for octanol/water) for any solute containing H, C, N, O, F, S, Cl, Br, and/or I.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.