The AMSOL models in AMPAC allow for the calculation of free energies of solvation for compounds containing H, C, N, O, F, P, S, Cl, Br, and I in water and organic solvents. Geometry optimization within a solvated environment is also supported.
The AMSOL Model Module (AMM) in AMPAC contains two universal solvation models, SM5.2 and SM5C. Universal solvation models are parameterized for aqueous solution or any organic solvent as well as other solvents or media for which certain solvent descriptors are known. SM5.2 is based on the generalized Born approximation for bulk electrostatics augmented by geometry-dependent atomic surface tensions for cavitation, dispersion, and solvent structure and is parameterized for AM1, PM3, MNDO, and MNDO/d. SM5C is based on the COSMO algorithm for solving the nonhomogeneous Poisson equation for bulk electrostatics augmented by geometry-dependent atomic surface tensions for cavitation, dispersion, and solvent structure and is parameterized for AM1, PM3, and MNDO/d. When the models are used with gas-phase geometries, they are called SM5.2R and SM5CR. The solvation models in AMM are the only SMx solvation models that have been parameterized for MNDO/d, and the implementation for semiempirical molecular orbital theory with d functions is not present in any other program. Each model is discussed briefly below.
AMPAC's advanced CI (configuration interaction) capability has also been integrated with AMSOL. This allows efficient treatment of open-shell systems in a solvated environment.
AMPAC has also implemented analytical gradients algorithms for all of the AMSOL models. As a result, calculations run much more rapidly reliably than in any other program.
The original version of the AMM was contributed to Semichem by its authors, namely, D. A. Liotard, G. D. Hawkins, D. M. Dolney, D. Rinaldi, C. J. Cramer, and D. G. Truhlar. The following reference should be cited in any research publications using the methods:
SM5.2R or SM5.2
SM5CR or SM5C
"A Universal Solvation Model Based on the Conductorlike Screening Model," D. M. Dolney, G. D. Hawkins, P. Winget, D. A. Liotard, C. J. Cramer, and D.G. Truhlar, Journal of Computational Chemistry 21, 340-366 (2000).
SM5.2 and SM5.2R
The SM5.2 models were parameterized using the AM1, PM3, MNDO, and MNDO/d methods for water using a training set containing 248 neutral solutes with a variety of functional groups. The SM5.2R model was parameterized using gas-phase geometries calculated at the Hartree-Fock level with a heteroatom-polarized valence-double-zeta basis set (HF/MIDI!), and it achieves a mean unsigned error of 0.47 kcal/mol when the model is applied using HF/MIDI! gas-phase geometries, and a mean unsigned error of 0.66 kcal/mol when it is applied using gas-phase geometries from MNDO/d. SM5.2 uses the same parameters as SM5.2R, but geometries are optimized in the liquid-phase rather than gas-phase.
SM5CR and SM5C
The SM5CR models were parameterized using the AM1, PM3, and MNDO/d methods using a training set containing neutral and ionic solutes with a variety of functional groups for which either experimental solvation free energies or experimental water/solvent partition coefficients were available. The training set contains experimental solvation free energies for 327 solutes in water and 90 organic solvents and partition coefficients for 54 solutesbetween water and one of 12 organic solvents. A total of 2141 experimental solvation free energies for 243 neutral solutes and 76 experimental water/solvent partition coefficients for 54 neutral soluteswere used in the parameterization. The SM5CR model was parameterized using gas-phase geometries calculated at the Hartree-Fock level with a heteroatom-polarized valence-double-zeta basis set (HF/MIDI!). The SM5CR model achieves a mean unsigned error of 0.55 kcal/mol when the model is applied using gas-phase HF/MIDI! geometries. The SM5C solvation model uses the same parameters as the SM5CR model, but geometries are optimized in the liquid phase.