The most general way to describe localized atomic-like electronic states in strongly correlated compounds is to utilize Wannier functions. In the present paper we continue the development of widely-spread DFT+U method onto Wannier function basis set and propose the technique to calculate the Hubbard contribution to the forces. The technique was implemented as a part of plane-waves pseudopotential code Quantum-ESPRESSO and successfully tested on a charge transfer insulator NiO.
We implemented the derivative of the free energy functional with respect to the atom displacements, so called force, within the combination of Density Functional Theory and the Embedded Dynamical Mean Field Theory. We show that in combination with the numerically exact quantum Monte Carlo (MC) impurity solver, the MC noise cancels to a great extend, so that the method can be used very efficiently for structural optimization of correlated electron materials. As an application of the method, we show how strengthening of the fluctuating moment in FeSe superconductor leads to a substantial increase of the anion height, and consequently to a very large effective mass, and also strong orbital differentiation.
We report on the implementation of the Wannier Functions (WFs) formalism within the full-potential linearized augmented plane wave method (FLAPW), suitable for bulk, film and one-dimensional geometries. The details of the implementation, as well as results for the metallic SrVO3, ferroelectric BaTiO3 grown on SrTiO3, covalently bonded graphene and a one-dimensional Pt-chain are given. We discuss the effect of spin-orbit coupling on the Wannier Functions for the cases of SrVO3 and platinum. The dependency of the WFs on the choice of the localized trial orbitals as well as the difference between the maximally localized and first-guess WFs are discussed. Our results on SrVO3 and BaTiO3, e.g. the ferroelectric polarization of BaTiO3, are compared to results published elsewhere and found to be in excellent agreement.
We study the electronic properties of GaV4S8 (GVS) and GaTaSe8 (GTS), two distant members within the large family of chalcogenides AM4X8, with A={Ga, Ge}, M={V, Nb, Ta, Mo} and X={S, Se}. While all these compounds are Mott insulators, their ground state show many types of magnetic order, with GVS being ferromagnetic and GTS non-magnetic. Based on their bandstructures, calculated with Density Functional Theory methods, we compute an effective tight binding Hamiltonian in a localised Wannier basis set, for each one of the two compounds. The localised orbitals provide a very accurate representation of the bandstructure, with hopping amplitudes that rapidly decrease with distance. We estimate the super-exchange interactions and show that the Coulomb repulsion with the Hunds coupling may account the for the different ground states observed in GVS and GTS. Our localised Wannier basis provides a starting point for realistic Dynamical Mean Field Theory studies of strong correlation effects in this family compounds.
We present a simple derivation of the Hellmann-Feynman theorem at finite temperature. We illustrate its validity by considering three relevant examples which can be used in quantum mechanics lectures: the one-dimensional harmonic oscillator, the one-dimensional Ising model and the Lipkin model. We show that the Hellmann-Feynman theorem allows one to calculate expectation values of operators that appear in the Hamiltonian. This is particularly useful when the total free-energy is available, but there is not direct access to the thermal average of the operators themselves.
The Feynman-Hellmann (FH) relation offers an alternative way of accessing hadronic matrix elements through artificial modifications to the QCD Lagrangian. In particular, a FH-motivated method provides a new approach to calculations of disconnected contributions to matrix elements and high-momentum nucleon and pion form factors. Here we present results for the total nucleon axial charge, including a statistically significant non-negative total disconnected quark contribution of around $-5%$ at an unphysically heavy pion mass. Extending the FH relation to finite-momentum transfers, we also present calculations of the pion and nucleon electromagnetic form factors up to momentum transfers of around 7-8 GeV$^2$. Results for the nucleon are not able to confirm the existence of a sign change for the ratio $frac{G_E}{G_M}$, but suggest that future calculations at lighter pion masses will provide fascinating insight into this behaviour at large momentum transfers.