Improving solvation energy predictions using the SMD solvation method and semiempirical electronic structure methods.

Area of Science:

• Computational Chemistry
• Theoretical Chemistry
• Physical Chemistry

Background:

• The GAMESS program's PM6 implementation was extended to include d-integrals.
• This extension was interfaced with the conductor-like polarized continuum model of solvation (SMD), incorporating gradients.
• The accuracy of various semi-empirical methods (AM1, PM3, PM6, DFTB) with SMD for solvation energies was evaluated.

Purpose of the Study:

• To assess the accuracy of Neglect of Diatomic Differential Overlap (NDDO)-based methods combined with the Solvation Model Density (SMD) for computing aqueous solvation energies.
• To identify sources of error in NDDO/SMD predictions, particularly for ions.
• To improve the accuracy of solvation energy calculations using NDDO-based methods by re-parameterizing SMD.

Main Methods:

• Utilized the Minnesota Solvation Database for testing accuracy.
• Compared root mean square error (RMSE) values for different methods (AM1, PM3, PM6, DFTB, DFT, HF) against experimental data.
• Re-parameterized SMD by adjusting atomic radii for specific elements (H, C, O, N, S) to minimize errors for aqueous solvation energies.

Main Results:

• NDDO-based methods with SMD showed significantly larger errors (RMSE 3.4-5.9 for neutrals, 6-15 kcal/mol for ions) compared to DFT and HF methods (RMSE 2.4, ~5 kcal/mol).
• Errors were particularly pronounced for cations and higher than conductor-like screening model results.
• Re-parameterization of atomic radii led to RMSE values for PM3, PM6, and DFTB that were more comparable to HF/6-31G(d) results, especially for ions (~5 kcal/mol).

Conclusions:

• Re-parameterizing SMD atomic radii is crucial for improving the accuracy of NDDO-based solvation energy calculations, particularly for ionic species.
• The optimized radii improve predictions for aqueous solvation energies and also enhance accuracy for other polar solvents like DMSO, acetonitrile, and methanol.
• The enhanced NDDO/SMD approach offers a more accurate and computationally efficient alternative for solvation energy studies compared to higher-level methods.