Gutzwiller Molecular Dynamics Simulation using Second-Moment Approximation

Molecular dynamics (MD) simulations are crucial to modern computational physics, chemistry, and material science, especially when combined with potentials derived from density-functional theory. However, even in state of the art MD codes, the on-site Coulomb repulsion is only treated at the self-consistent Hartree-Fock level. This standard approximation may miss important effects due to electron correlations. The recently developed Gutzwiller molecular dynamics (GMD) method provides a feasible approach for realistic correlated materials [1]. The Gutzwiller variational method captures the essential physics of correlated electron systems, and is much faster than, for example, the dynamical-mean field theory approach. In its current implementation, however, the GMD method is limited to small system sizes mainly due to the heavy computational cost of solving the renormalized electron Hamiltonian at every time-step. Here we demonstrate significant improvement of the GMD efficiency using the second-moment approximation. Importantly, we show that this approximation still captures the main features of the correlation-induced metal-insulator transition.

[1] G.-W. Chern, K. Barros, C. D. Batista, J. Kress, G. Kotliar, Phys. Rev. Lett. 118, 226401 (2017).