Benchmarking density functional methods against the S66 and S66x8 datasets for non-covalent interactions

Area of Science:

• Computational chemistry
• Quantum chemistry
• Materials science

Background:

• Non-covalent interactions are crucial in chemistry and biology.
• Accurate calculation of these interactions is challenging for standard density functional theory (DFT).
• Dispersion corrections are known to improve DFT performance.

Purpose of the Study:

• To assess the performance of dispersion-corrected DFT on the S66 and S66x8 benchmark sets.
• To evaluate 17 different density functionals with two flavors of the DFT-D3 dispersion correction.
• To compare the accuracy of DFT-D3 and DFT-D3(BJ) methods.

Main Methods:

• Dispersion-corrected density functional theory (DFT-D3 and DFT-D3(BJ)).
• Evaluation of 17 density functionals on the S66 and S66x8 benchmark sets.
• Statistical analysis of interaction energies and equilibrium distances.

Main Results:

• Dispersion corrections are crucial for reliable non-covalent interaction energies and distances.
• DFT-D3 and DFT-D3(BJ) methods show similar performance, with DFT-D3(BJ) slightly better in most cases.
• Double-hybrid functionals like PWPB95-D3 and B2-PLYP-D3(BJ) are the most accurate.
• Minnesota functionals show improvement with the D3 correction.

Conclusions:

• Most dispersion-corrected DFT methods are recommended for calculating non-covalent interactions.
• DFT-D3 and DFT-D3(BJ) approaches are competitive with perturbation methods and outperform MP2.
• The S66 benchmark set is statistically robust and useful for method development and validation.