Revised Damping Parameters for the D3 Dispersion Correction to Density Functional Theory.

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Materials Science

Background:

• The accuracy of density functional theory (DFT) methods, particularly for non-covalent interactions, relies heavily on empirical dispersion corrections like DFT-D3.
• Original DFT-D3 parameters were fitted using limited benchmark sets, potentially lacking representativeness for diverse interaction types and geometries.

Purpose of the Study:

• To re-evaluate and improve the accuracy of Grimme's DFT-D3 dispersion corrections.
• To assess the impact of updated and expanded benchmark datasets on the performance of DFT-D3.
• To provide a more robust description of intermolecular interactions across various chemical systems.

Main Methods:

• Augmentation of conventional benchmark sets with new databases like side chain-side chain (SSI) interactions derived from crystal data.
• Extension of existing databases (e.g., S22×5) to include shorter intermolecular distances.
• Retraining of DFT-D3 damping parameters using an expanded dataset of 1526 interaction energies.
• Validation of the refitted parameters against a new, independent dataset of 6773 interaction energies.

Main Results:

• The refitted DFT-D3 parameters show varying degrees of accuracy improvement across different DFT functionals.
• For PBE-D3, refitting led to an almost two-fold decrease in average error.
• For LC-ωPBE-D3, the accuracy remained largely unchanged, indicating robustness.
• The expanded training set provides a more balanced description of interaction distances and types.

Conclusions:

• Refitting DFT-D3 parameters with larger, more diverse datasets significantly enhances the accuracy of predicting interaction energies for certain functionals.
• The updated approach offers a more reliable description of non-covalent interactions crucial for molecular modeling and materials science.
• The choice of DFT functional critically influences the benefits gained from refitting DFT-D3 parameters.