A double-hybrid density functional based on good local physics with outstanding performance on the GMTKN55 database

Area of Science:

• Quantum Chemistry
• Computational Chemistry
• Materials Science

Background:

• Previous work compared Kohn-Sham density functionals with power-series fits.
• The best error statistics for a hybrid functional on the GMTKN55 database were previously reported.

Purpose of the Study:

• To develop a new double-hybrid density functional with improved accuracy.
• To evaluate the performance of the new functional on the GMTKN55 chemical database.

Main Methods:

• Incorporation of additional second-order perturbation-theory terms.
• Development of a 12-parameter double-hybrid density functional.
• Testing on the GMTKN55 database for thermochemistry, kinetics, and noncovalent interactions.

Main Results:

• The new functional, DH23, achieved the lowest WTMAD2 error (1.76 kcal/mol) on the GMTKN55 database.
• This represents the best performance to date for any hybrid or double-hybrid density-functional approximation.
• The results surpass previous benchmarks for chemical accuracy.

Conclusions:

• DH23 offers a significant advancement in density-functional theory.
• The functional provides highly accurate predictions for a wide range of chemical properties.
• DH23 is a promising tool for computational chemistry research.