Implementation of the SM12 Solvation Model into ADF and Comparison with COSMO

Abstract

In this article, an implementation of the newest iteration of the Minnesota solvation model, SM12, into the Amsterdam density functional (ADF) computational package is presented. ADF makes exclusive use of Slater-type orbitals (STO), which correctly represent the true atomic orbitals for atoms, whereas SM12 and the underlying charge model 5 (CM5) have previously only been tested on Gaussian-type orbitals (GTO). This new implementation is used to prove the basis set independence of both CM5 and SM12. A detailed comparison of the SM12 and COSMO solvation models, as implemented in ADF, is also presented. We show that this new implementation of SM12 has a mean unsigned error (MUE) of 0.68 kcal/mol for 272 molecules in water solvent, 4.10 kcal/mol MUE for 112 charged ions in water, and 0.92 kcal/mol MUE for 197 solvent calculations of various molecules. SM12 outperforms COSMO for all neutral molecules and performs as well as COSMO for cationic molecules, only falling short when anionic molecules are taken into consideration, likely due to CM5's use of Hirshfeld charges and their poor description of anionic molecules, though CM5 seems to improve upon this discrepancy.

A Computational Protocol Combining DFT and Cheminformatics for Prediction of pH-Dependent Redox Potentials

“…We then compared two other solvation methods with COSMO. In Figure 1 c, we show that with SM12 [ 38 ] single-point calculations (optimization is not possible in ADF) at COSMO geometries, we obtained a systematic error of 0.1 V. When empirically correcting for this error by subtracting the MSE from the computed potentials, the MAE became comparable with COSMO. In Figure 1 d, we assess the performance of the COSMO-RS method, [ 39 , 40 ] which combines quantum chemistry with statistical mechanics.…”

Acidity of M(VI)O2(OH)2 for M = Group 6, 16, and U as Central Atoms

“…The SM12 p K a ’s are significantly more negative than any of the other methods, and the SMD values show a somewhat different variation, being more comparable to the COSMO values. As noted by Schreckenbach, the COSMO approach provides better values for anions than does SM12 due to the treatment of the more diffuse charge in the anions. Even though the SM12 model is better for neutrals than COSMO, the anion solvation energies are much larger; therefore, we choose to employ the COSMO values in our discussion below.…”

“…60−63 We also used Cramer and Truhlar's SMD solvation approach 64 as implemented in Gaussian 09 65 and the SM12 approach 66 implemented in ADF. 67 For the COSMO [B3LYP 68,69 /aug-cc-pVDZ(-PP)] calculations in Gaussian, the radii developed by Klamt and co-workers were used to define the cavity. 57,70 For the COSMO calculations at the BLYP/ TZ2P level 71 in ADF, the Klamt radii were used to define the cavity as well.…”

Evaluation of the performances of different atomic charge and nonelectrostatic models in the finite‐difference Poisson–Boltzmann approach

“…The first term here accounts for how each solute atom affects the solvent, while the second encompasses how the solvent interacts with the full solute [73]. Surface tensions are functions of both optimizable parameters and experimental solvent and solute macroscopic descriptors, according to: where n is the solvent refractive index at room temperature, α and β are Abraham's hydrogen bond acidity and basicity parameters [74] of the solvent, respectively, γ is the macroscopic solvent surface tension, ψ is the fraction of non‐hydrogen atoms in the solvent that are aromatic carbons, and ϕ is the fraction of non‐hydrogen atoms in the solvent that are F, Cl, or Br.…”