Putting error bars on density functional theory.

Area of Science:

• Computational Materials Science
• Quantum Chemistry
• Materials Informatics

Background:

• Accurate prediction of material properties using Density Functional Theory (DFT) relies heavily on the choice of exchange-correlation (XC) functional.
• Estimating the a priori error associated with different XC functionals is challenging, hindering efficient high-throughput screening.
• Understanding functional-specific errors is crucial for developing more accurate predictive models.

Purpose of the Study:

• To develop a materials informatics approach for predicting errors in DFT calculations arising from different XC functionals.
• To analyze the systemic errors of common XC functionals (LDA, PBE-GGA, PBEsol, vdW-DF) for binary and ternary oxides.
• To provide quantitative error estimates ('error bars') to guide functional selection for high-throughput materials discovery.

Main Methods:

• Computation of structural and elastic properties for binary and ternary oxides using four distinct XC functionals.
• Application of materials informatics techniques to analyze and predict systematic errors associated with each XC functional.
• Correlation of predicted DFT errors with intrinsic functional properties like electron density and hybridization.

Main Results:

• Systematic errors were predicted for LDA, PBE-GGA, PBEsol, and vdW-DF (with C09 exchange) functionals.
• The predicted errors were successfully used to refine DFT-calculated lattice parameters.
• A clear link was established between computed errors and the functional's performance in describing electron density and hybridization.

Conclusions:

• Materials informatics provides a robust framework for estimating DFT XC functional errors.
• The findings offer practical 'error bars' for selecting appropriate functionals in high-throughput materials screening.
• This work paves the way for developing improved XC functionals and accelerating the discovery of novel materials.