Metallic Surface Energies Beyond the Random Phase Approximation

The prediction of surface energies for extended systems is often used as a descriptor for understanding trends in surface properties such as molecular adsorption. Though semilocal functionals can predict either accurate adsorptions or accurate surface energies, non-local functionals, such as the Random Phase Approximation (RPA), are required to deliver accurate predictions simultaneously for both properties. Here we tested the impact of correlation methods beyond RPA to understand the role short-ranged correlation plays in determining the surface energies of several transition metals. We find that RPA overestimates the surface energy due to an overestimation of the bulk correlation energy. Addition of an exchange-correlation kernel corrects this behavior and tends to reduce the predicted surface energies due to an increase in the bulk energy with respect to the surface slab. In comparison to experiment, RPA and beyond-RPA methods both deliver systematically accurate results that are superior to popular semilocal functionals.