Mapping the genome of meta-generalized gradient approximation density functionals: the search for B97M-V

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Materials Science

Background:

• Density functional theory (DFT) is a powerful tool for electronic structure calculations.
• Developing accurate and efficient density functionals is crucial for advancing DFT applications.
• Existing functionals often struggle with specific types of interactions, such as non-bonded forces.

Purpose of the Study:

• To develop a new meta-generalized gradient approximation (meta-GGA) density functional.
• To incorporate the VV10 nonlocal correlation functional for improved accuracy.
• To rigorously screen a vast functional space and identify an optimal form with minimal empiricism.

Main Methods:

• Training and testing over 10^10 candidate functional forms on extensive datasets (1095 training, 1153 primary test points).
• Filtering based on physical constraints, numerical precision, and performance on training and test sets.
• Developing the B97M-V functional with 12 fitted parameters (4 exchange, 4 same-spin correlation, 4 opposite-spin correlation).

Main Results:

• B97M-V demonstrates remarkable accuracy for non-bonded interactions.
• The functional provides very satisfactory results for thermochemical quantities like atomization energies.
• Performance was assessed on a secondary test set (212 points) and large systems, showing good geometry prediction and basis set convergence.

Conclusions:

• B97M-V represents a significant advancement in density functional development, particularly for non-bonded interactions.
• It offers a favorable balance of accuracy and efficiency for many chemical applications.
• Like other local density functionals, it retains limitations in accurately predicting reaction barrier heights.