Authors: Ageo Meier de Andrade, Jolla Kullgren and Peter Broqvist
We evaluate the qualitative and quantitative accuracy of various flavors of density functionals with and without accounting for dispersion corrections. Our test system is nickel in the form of bulk, surfaces, and nanoparticles for which we compute structural properties, bulk cohesive energies, surface energies, and work functions and compare to experimental data. We find that the inclusion of any dispersion, either by an a posteriori correction or by a self-consistent treatment by explicitly computing the nonlocal correlation contribution to the total energy, has a significant effect on the calculated properties and improves the quantitative comparison to experiments. Besides the quantitative agreement, we also investigate qualitative features by comparing Wulff shapes of metal nanoparticles as obtained using the different density functionals. We find that all tested functionals predict similar Wulff shapes for nickel nanoparticles but still have some small differences. These results show that the relative energies calculated using the semilocal GGA and meta-GGA functionals, with and without dispersion, are quite similar. Our findings can also be generalized to other systems when rationalized in terms of the computed reduced density gradients. We find that the distribution of reduced density gradients in a material is correlated to the steepness of the exchange enhancement factor and propose that this information can be used as a quantitative guide when it comes to picking the most appropriate density functional for specific target systems as well as when it comes to extrapolating DFT data to predict experiments.
Phys. Rev. B 102, 2020, 075115