Non-Covalent Interactions Atlas Benchmark Data Sets 3: Repulsive Contacts.

Area of Science:

• Computational chemistry
• Molecular interactions
• Quantum mechanics

Background:

• Accurate computational methods are essential for understanding molecular interactions.
• Repulsive contacts in molecular complexes are critical for developing robust computational models.
• The Non-Covalent Interactions Atlas provides valuable data for method development.

Purpose of the Study:

• To evaluate the performance of various density functional theory (DFT) and semiempirical quantum-mechanical methods using the new R739×5 dataset.
• To identify the most accurate computational methods for describing repulsive non-covalent interactions.
• To analyze errors in existing methods and propose corrections.

Main Methods:

• Utilized the R739×5 dataset with highly accurate coupled cluster singles and doubles with perturbative triples (CCSD(T))/complete basis set (CBS) interaction energies.
• Tested selected density functional theory (DFT) methods, including double-hybrid and range-separated functionals.
• Assessed semiempirical quantum-mechanical methods, such as GFN2-xTB and PM6.

Main Results:

• Double-hybrid DFT functionals demonstrated superior performance, with revDSD-PBEP86-D3 being the most accurate.
• Range-separated ωB97X functionals also showed high accuracy.
• GFN2-xTB provided the best results among the tested semiempirical methods.
• An analysis of the PM6 method identified error sources and led to a correction for conformational energies.

Conclusions:

• The R739×5 dataset is valuable for benchmarking computational methods for repulsive interactions.
• Advanced DFT functionals and GFN2-xTB are recommended for accurate calculations of non-covalent repulsive forces.
• Methodological improvements, like the proposed correction for PM6, can enhance the practical applicability of computational chemistry tools.