Doubling down on density-functional theory.

Area of Science:

• Quantum Chemistry
• Computational Chemistry

Background:

• A novel double hybrid density functional, DH23, was introduced by Becke et al.
• DH23's coefficients were trained on the GMTKN55 database, achieving a WTMAD2 error of 1.76 kcal/mol.

Purpose of the Study:

• To revise the DH23 density functional.
• To evaluate the revised DH23's performance on data outside the GMTKN55 training set, specifically organometallic reaction energies and barrier heights.

Main Methods:

• Revisions were made to the DH23 functional.
• The revised DH23 was tested against reference data for organometallic reactions.
• Performance was quantified using the WTMAD2 error metric.

Main Results:

• The revised DH23 demonstrated robustness and accuracy on organometallic reaction energies and barrier heights.
• A slightly improved GMTKN55 WTMAD2 error of 1.73 kcal/mol was achieved.
• The functional performed well on data beyond its original training set.

Conclusions:

• The revised DH23 functional is a reliable tool for computational chemistry.
• DH23 shows broad applicability in predicting chemical reaction properties, including those involving organometallic compounds.