Coulomb interactions and conventional superconductivity: when going beyond the random phase approximation is essential

Abstract: In ab initio calculations of superconducting properties, the Coulomb repulsion is accounted for at the GW level and is usually computed in RPA, which amounts to neglecting vertex corrections both at the polarizability level and in the self-energy. Although this approach is unjustified, the brute force inclusion of higher order corrections to the self-energy is computationally prohibitive. We propose to use a generalized GW self-energy, where vertex corrections are incorporated into W by employing the Kukkonen and Overhauser (KO) ansatz for the effective interaction between two electrons in the electron gas. By computing the KO interaction in the adiabatic local density approximation for a diverse set of conventional superconductors, and using it in the Eliashberg equations, we find that vertex corrections lead to a systematic decrease of the critical temperature (Tc), ranging from a few percent in bulk lead to more than 40% in compressed lithium. We propose a set of simple rules to identify those systems where large Tc corrections are to be expected and hence the use of the KO interaction is recommended. Our approach offers a rigorous extension of the RPA and GW methods for the prediction of superconducting properties at a negligible extra computational cost.