Accurate Atomic and Molecular Calculations without Gradient Corrections:  Scaled SVWNV Density Functional

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Materials Science

Background:

• The Local Spin Density Approximation (LSDA) was an early Density Functional Theory (DFT) method for calculating atomic and molecular properties.
• LSDA has become less popular for molecular systems due to the development of more advanced methods like BLYP and B3LYP.

Purpose of the Study:

• To revisit the LSDA method and investigate a simple approach to enhance its accuracy for molecular calculations.
• To propose an improved SVWNV density functional suitable for molecular systems, offering a balance of accuracy and computational efficiency.

Main Methods:

• Scaling the local correlation contribution within the SVWNV functional.
• Evaluating the performance of the scaled SVWNV functional for various molecular properties, including heats of formation, ionization potentials, and bond characteristics.
• Comparing the computational speed of the improved SVWNV functional against B3LYP and BLYP.

Main Results:

• Scaling the SVWNV functional significantly reduced average unsigned errors in heats of formation, up to nine times smaller than the standard SVWNV.
• Substantial, though less dramatic, error reductions were observed for other studied properties like ionization potentials and electron affinities.
• The improved SVWNV functional demonstrated notable time savings, being 55% faster than B3LYP and 40% faster than BLYP for a 9-alanine system at the 3-21G* basis set.

Conclusions:

• The scaled SVWNV functional provides improved accuracy for molecular properties compared to the standard LSDA approach.
• The enhanced SVWNV functional is a computationally efficient and reasonably accurate DFT method, suitable for molecular studies, especially at small basis sets.
• This method is particularly well-suited for QM/QM applications requiring a cost-effective treatment of larger system portions.