DFT Accuracy on Bulk Transition Metals Properties

Abstract

Density Functional Theory would be exact in estimating a polyelectronic chemical system energy, when the exchange-correlation (xc) functional would be known. Unfortunately, it is not, and has to be approximated, with dozens of xc functionals developed in the last decades, belonging to different rungs of Jacob’s ladder of xc improvement. In the case of Transition Metals (TMs), mostly describing few late TMs and their structural properties. Recent studies expanded the analysis covering different bulk properties and surface features. Here we carry out such performance evaluation on so far ignored xc functionals, either within the most basic local density approximations, including the, Hendin-Lundqvist (HL) and Perdew-Zunger (PZ) xc functionals, or within the Generalized Gradient Approximation (GGA), exploring the revised Perdew-Burke-Ernzerhof (revPBE) and the Armiento-Mattson (AM05) xc functionals. Aside from these, the recent meta-GGA Strongly Constrained Appropriately Normed (SCAN) functional is analysed, characterized by fulfilling all 17 mathematical conditions an xc must comply, plus the Bayesian Error Estimation Functional (BEEF) is explored, a functional where artificial intelligence, in the form of a machine learning algorithm, was used to adjust the mathematical expression to a large and diverse set of experimental results. The present results, acquired for 27 TM bulks in their crystallographic structures —body-centred cubic, face-centred cubic, and hexagonal close-packed—, reveal that none of the explored functionals is best in describing TM bulks, were Viñes-Vega (VV) excels, and highlighting that, while SCAN performance is acceptable, BEEF is not. When accounting for TM surface properties, acquired on 81 low-index Miller surfaces, the same situation applies, not improving the VV xc adapted for solids (VVsol).