Benchmarking the performance of the SCAN meta-GGA functional for solid-state thermodynamics

Constructed to satisfy all known exact constraints and appropriate norms for a semi-local density functional, the SCAN meta-GGA has shown early promise in the accurate description of electronic structure of molecules and solids [1]. One open question is how SCAN performs in describing the thermodynamics of solid-state materials. To answer this question, we perform an extensive benchmark of SCAN by computing the formation energies for a diverse group of nearly one thousand crystalline solids for which experimental values are available. While SCAN substantially decreases the formation energy errors for strongly bound solids (by around 50%) compared to the GGA level of theory, for intermetallic compounds we see no improvement. We find that SCAN leads to significantly more accurate structural parameters, mildly improved band gaps, and enhanced magnetism compared to GGA. Additionally, we present elemental chemical potential corrections for SCAN that are fit to achieve best agreement with experimental formation energies.

[1] J. Sun, A. Ruzsinszky, and J. P. Perdew, Phys. Rev. Lett. 115, 36402 (2015).