Riassunto analitico
Vanadium dioxide is a prototypical material undergoing a coupled structural and metal-insulator transition at Tc = 340 K. This transition has been extensively studied, for both its fundamental interest and its possible technological uses, ranging from optoelectronics to neuromorphic computing. Due to the complex interplay between electronic and structural effects involving dimerization and correlation, it has been one of the paradigmatic challenges for first-principles calculations. Standard approaches like (approximate) density-functional theory (DFT) fail to capture the insulating nature of the monoclinic phase of VO2, thus pointing to the need of beyond-DFT methods. In this work, we study VO2 using a novel dynamical Hubbard functional, designed to simultaneously address ground state and spectral properties for correlated materials. This functional can be seen as the dynamical generalization of the static Hubbard extension of DFT, hosting a frequency-dependent screened potential U(ω), where the fundamental variable is the Green's function---instead of the density of DFT. Furthermore, we exploit the recently introduced algorithmic-inversion method on sum-over-poles (AIM-SOP) to deal with dynamical functionals involving the Green's function of the material. Within this framework, the Dyson equation is solved exactly by mapping it into a nonlinear eigenvalue problem and diagonalizing an Hamiltonian with augmented degrees of freedom. For the application to VO2, we choose a set of bond-centered Wannier functions as the active localized manifold, in agreement with recent literature on DFT augmented by dynamical mean-field theory. This seems to be a most natural way to capture both intra- and inter-atomic effects driving the transition, since it allows for the formation of a molecular bonding state between dimerizing V atoms. To work with Wannier functions, as a first step in our work, we generalize the in-house implementation of the dynamical Hubbard extension. Our study allows us to move towards an improved description of the monoclinic phase of VO2, by breaking some band degeneracies and inducing a small scissor effect. Remarkably, we are also able to draw in the case of VO2 some interesting similarities between different static DFT corrections, namely DFT+V and DFT+U using bond-centered orbitals.
|
