Comparative study of semilocal density functionals on solids and surfaces

TL;DR

This study evaluates the TMTPSS density functional, combining TM exchange with TPSS correlation, demonstrating its competitive accuracy for solids and surfaces, especially in surface exchange-correlation energy calculations.

Contribution

It introduces and assesses the TMTPSS functional, a new combination of exchange and correlation components, for improved accuracy in solid and surface property predictions.

Findings

TMTPSS outperforms LSDA, PBE, and TPSS in most properties

Significant improvement in surface exchange-correlation energy accuracy

TMTPSS is comparable to TM meta-GGA in performance

Abstract

Recently, Tao and Mo (TM) proposed an accurate all-purpose nonempirical meta-generalized gradient approximation (meta-GGA). The exchange part was derived from the density matrix approximation, while the correlation part is based on a modification of TPSS correlation in the low-density or strong-interaction limit. To further understand this density functional, we combine the TM exchange part with the original TPSS correlation and make a comprehensive assessment of this combination, which we call TMTPSS functional, on solids and solid surfaces. Our test includes 22 lattice constants and bulk moduli, 30 band gaps of semiconductors, 7 cohesive energies, and surface exchange-correlation energies for rs ranging from 2 to 3 bohr. Our calculations show that TMTPSS functional is quite competitive to the TM meta-GGA functional, improving upon the nonempirical functionals LSDA, PBE GGA, and TPSS meta-GGA for most properties considered. In particular, it significantly improves the surface exchange-correlation energy calculation, with a mean absolute error of only 1 erg/cm2.