Title: Non-perturbative theory of the electron-phonon coupling and its first-principles implementation

Abstract: The harmonic approximation of ionic fluctuations and the linear coupling between phonons and electrons provide the basic approach to compute, from first principles, the contribution that the nuclei dynamics and its interaction with electrons have on materials' properties. These approaches are questionable, at least, whenever quantum and anharmonic effects on the nuclei vibrational properties are large, such as in hydrogenous systems, high-$T_c$ superconductors, and when systems are close to displacive phase transitions. Here we propose a novel non-perturbative approach to compute the electron-phonon interaction from first principles that includes non-linear effects and takes into account the quantum nature of nuclei. The method is based on the $GW^{(en)}$ approximation for the electron self-energy, given by the effective nuclei-mediated electron-electron interaction $W^{(en)}$ and the electron Green's function $G$. The electrons are treated at a mean-field level and the nuclei dynamics is described with a Gaussian distribution function, which can effectively take into account anharmonic effects at a mean-field level, e.g. within the self-consistent harmonic approximation. Pivotal quantities of the Gaussian $GW^{(en)}$ self-energy are the renormalized average vertices, which are computed in supercells with a stochastic approach, using the the self-consistent electronic potential computed for different atomic configurations. In order to validate the method, $GW^{(en)}$ calculations are performed on aluminum, a highly harmonic system with weak electron-phonon coupling. The obtained results coincide with the ones obtained with standard linear electron-phonon calculations. However, calculations performed on palladium hydride, a very anharmonic system, show a highly non-linear electron-phonon interaction, with $GW^{(en)}$ bringing corrections as high as the linear-order result.