Efficient GW calculations for metals from an accurate ab initio polarizability

Abstract: Despite its success in the study of spectroscopic properties, the $GW$ method presents specific methodological challenges when applied to systems with metallic screening. Here, we present an efficient and fully ab-initio implementation for the calculation of the screened potential, specifically designed for 3D and 2D metals. It combines a Monte Carlo integration with an appropriate interpolation of the screened potential between the calculated grid points (W-av), complemented with an extrapolation to the long-wavelength limit, able to seamlessly account for the so-called intraband term. This method greatly accelerates the convergence of GW calculations for metals while improving their accuracy, due to the correct description of the intraband transitions in the long wavelength limit, as shown here for 3D metals and doped monolayers, such as MoS$_2$ and graphene. The use of W-av results in an excellent agreement with ARPES measurements for monolayer doped MoS$_2$. Furthermore, for graphene we show that more robust results are found with the use of higher-order Lorentzians in the description of the self-energy, together with the solution of the QP equation beyond the linearized approximation.