Benchmark Databases for Nonbonded Interactions and Their Use To Test Density Functional Theory.

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Molecular Modeling

Background:

• Accurate prediction of binding energies is crucial for understanding nonbonded interactions in molecular systems.
• Existing computational methods vary in their performance for different types of nonbonded interactions.
• Development of reliable and accurate computational tools is essential for advancing chemical research.

Purpose of the Study:

• To establish benchmark databases for evaluating the accuracy of computational methods in predicting binding energies.
• To assess the performance of a wide range of density functional theory (DFT) and wave function theory (WFT) methods.
• To identify the most accurate methods for different types of nonbonded interactions, including hydrogen bonding, charge transfer, dipole interactions, and weak interactions.

Main Methods:

• Creation of four benchmark databases containing binding energies for nonbonded complexes.
• Testing of 44 DFT methods and 1 WFT method against these databases.
• Evaluation of method performance based on relative error for hydrogen bonding, charge transfer, dipole, and weak interactions.

Main Results:

• Specific DFT functionals (PBE, PBE1PBE, B3P86, MPW1K, B97-1, BHandHLYP) excelled in hydrogen bonding predictions.
• MPWB1K, MP2, MPW1B95, MPW1K, and BHandHLYP showed top performance for charge-transfer interactions.
• MPWB1K demonstrated the best overall performance among all tested DFT methods, with an average relative error of 11%.

Conclusions:

• The MPWB1K method is identified as a highly accurate DFT approach for predicting binding energies of nonbonded complexes.
• DFT methods like MPW1K, PBE1PBE, and B98 are recommended for cases where kinetic energy density is not included.
• Advancing along the 'Jacob's ladder' of DFT, such as with PBE and TPSS functionals, shows significant improvements for nonbonded interactions.