The Best DFT Functional Is the Ensemble of Functionals.

Area of Science:

• Computational chemistry
• Materials science
• Quantum mechanics

Background:

• Density Functional Theory (DFT) is crucial for molecular and materials simulations.
• Current research often focuses on improving single exchange-correlation functionals.
• There is a need for more robust and accurate DFT methods.

Purpose of the Study:

• To propose a method for creating transferable ensembles of density functionals.
• To demonstrate that these ensembles can surpass the performance of individual functionals.
• To introduce practical, accurate, and cost-effective DFT methods.

Main Methods:

• Constructing transferable ensembles of density functionals.
• Utilizing density functionals predating the GMTKN55 benchmark (2017).
• Integrating ensembles into the SCF procedure to create mixed functionals.

Main Results:

• Ensembles achieved a record-low weighted error of 1.62 kcal mol⁻¹ on the GMTKN55 benchmark.
• This represents a significant improvement over the best constituent functional (3.08 kcal mol⁻¹).
• DENS24 ensembles demonstrate consistently accurate performance across various simulations.
• Mixed DENS24 functionals offer comparable accuracy with increased speed.

Conclusions:

• Ensembles of density functionals provide a superior approach to improving DFT accuracy.
• DENS24 ensembles offer a practical, open-source, and accessible solution for computational simulations.
• Mixed DENS24 functionals present an efficient alternative for accelerating simulations without sacrificing accuracy.