Assessing the performances of different continuum solvation models for the calculation of hydration energies of molecules, polymers and surfaces: a comparison between the SMD, VASPsol and FDPB models

Abstract

We present a comparison of the performances of three continuum solvation models, namely the Solvation Model Density (SMD), VASPsol and Finite-Difference Poisson Boltzmann (FDPB) approaches, for the calculation of the hydration energies of molecules, polymers and semiconductor surfaces. For finite molecular systems, all three models have been considered, and the computed data have been compared to available experimental solvation energies for a test set of 630 neutral solutes. For infinite periodic systems, due to the lack of periodic implementations of the SMD model and of experimental solvation energies, only a comparison between the data obtained with the VASPsol and FDPB approaches has been performed. As a key criterion of the validation of a periodic implementation, the size-extensivity of the solvation energy of a model of a poly glycine chain has been considered. In addition, the effect of the surface orientation and of the slab thickness on the computed solvation energy has been investigated by considering five low-index surface orientations of galena PbS, due to the importance of this material in environmental-related processes. For finite molecular systems, all models performed satisfactorily on the whole test set with mean unsigned errors (MUE) values close to the target 1 kcal/mol error threshold with respect to the experimental data. However, depending on the chosen solvation model, some solute classes revealed challenging with MUE reaching 4.5 kcal/mol in the worst cases. Overall, the best performing models were found to be: FDPB > SMD ≈ VASPsol. For infinite periodic systems, the size-extensivity of the computed solvation energies has been verified, both with the VASPsol and FDPB models. In addition, calculations performed on PbS surface models further revealed at least a qualitative agreement between the two solvation models. In particular, the trends in surface energy variation between gasphase and solvated cases, computed solvation energies, as well as the surface solvation energy convergence as a function of the number of layers in the slab models have all been found very similar between these two approaches. Overall, therefore, these results are encouraging for the applicability of continuum solvation models to a broad range of solutes, ranging from finite small molecular solutes to extended periodic systems.

Towards a transferable nonelectrostatic model for continuum solvation: The electrostatic and nonelectrostatic energy correction model

“…On the other hand, results with FDPB electrostatics are slightly worse, with a MUE of 1.13 kcal/mol on the training set and 1.07 ± 0.05 kcal/mol on the test set. Interestingly, the MUE on the test set is lower than the one obtained on the training set: this can be traced to the higher number of long chain aliphatic molecules 84 within the Solv@tum set compared to the MNSOL set. Although acceptable results can be obtained by considering a reduced number of parameter, a lower MUE can thus be achieved by introducing solute atomic number dependent parameters, as done in the extended version of the ENE model.…”

Interplay of Solid–Liquid Interactions and Anisotropic Aggregation in Solution: The Case Study of γ-AlOOH Crystallites

“…The validity of the dielectric continuum model implemented in VASP has been benchmarked in ref . Examples of similar calculations to our approach in this work may be found in refs and .…”

Revisiting the Electrified Pt(111)/Water Interfaces through an Affordable Double-Reference Ab Initio Approach

“…The adequacy of this solvation model has been a subject of prolonged debate within the scientific community. Bramley et al and Akinola believed that the DFT-based implicit solvent is problematic. − However, several groups also suggested that the solvation model can effectively predict interfacial properties. − , Figure a shows the calculated electrostatic potential of the Pt(111) in the absence of an explicit water layer but in the presence of the implicit solvation model. The calculated PZC value obtained is 1.09 V vs SHE, which significantly differs from the experimentally reported value of 0.4 V. − Our calculated PZC value from the implicit model is close to a similar work by Mathew et al, which is 0.85 V. The small difference may be ascribed to the use of different XC functionals.…”