Static, dynamic and electronic properties of some trivalent liquid rare earth metals near melting: ab initio and neural network simulations

We report a study on several static and dynamic properties of the early trivalent liquid rare-earth metals at thermodynamic conditions near their respective melting points. It has been performed by resorting to machine learning (ML) techniques, in which the associated neural network-based interatomic potentials were derived from ab initio molecular dynamics simulations within Hubbard-corrected density functional theory. We report the results obtained for the static structural properties, including an analysis of the local short-range order. Single-particle and collective dynamic properties have also been obtained, from which transport coefficients and wavevector-dependent dispersion relations are evaluated. The results show a quite homogeneous behavior of the structural, dynamic, and transport properties throughout the series. The electronic properties have been obtained from the ab initio simulations, and show important discrepancies with respect to the low temperature solids, portraying a more band-like picture of the 4f states in the liquid.