New universal solvation model and comparison of the accuracy of the SM5.42R, SM5.43R, C-PCM, D-PCM, and IEF-PCM continuum solvation models for aqueous and organic solvation free energies and for vapor pressures

We present a new continuum solvation model, called Solvation Model 5.43R (SM5.43R). The model is based on the generalized Born approximation for electrostatics augmented by terms that are proportional to the solvent-accessible surface areas (SASAs) of the atoms of the solute, and it is parametrized to predict the free energy of solvation of solutes containing H, C, N, O, F, P, S, Cl, and Br in water and organic solvents. The new model is an improvement over our previous solvation model, SM5.42R, in that it is based on CM3 charges rather than CM2 charges, it was trained over a larger and more diverse training set, and the choice of the value of the solvent radius, which is used to compute the SASA of the atoms of the solute, was made on a different basis than was used for SM5.42R. This paper presents parametrizations of SM5.43R using HF/6-31G(d), B3LYP/6-31G(d), mPW1PW91/6-31G(d), and mPW1PW91/6-31+G(d) to describe the electronic structure of the solute. For a data set of neutral solutes with known experimental aqueous free energies of solvation containing at most H, C, N, O, F, P, S, Cl, and Br (257 data), the mean-unsigned error (MUE, in kcal/mol), as compared to experiment, calculated by SM5.43R is respectively 0.5.0, 0.49, 0.50, and 0.54 using these four solute wave functions. A similar MUE is obtained for SM5.42R using HF/6-31G(d). The corresponding MUEs calculated by several other generally available continuum solvation models are approximately a factor of 2 larger than those computed by SM5.43R and SM5.42R using the same electronic structure methods. For a data set of solutes with experimental free energies of solvation in 16 organic solvents (621 data), SM5.43R and SM5.42R yield MUEs 6.3 to 7.9 times smaller than the MUEs calculated by the other continuum solvation models. The SM5.43R model is, however, universal in that it can be used in any organic solvent, as well as water. Furthermore, it allows one to analyze solvation trends in terms of local properties, and this is illustrated for acetanilide in water and diethyl ether.