Testing interaction models by using x-ray absorption spectroscopy: solid Pb

Structural models obtained using classical molecular dynamics (MD) simulations and realistic interatomic potentials for solid metals are tested using experimental results obtained by x-ray absorption spectroscopy (XAS). Accurate L-edge extended x-ray absorption fine-structure (EXAFS) measurements of Pb grains dispersed in BN and graphite matrices have been collected for temperatures up to the melting point. The thermal expansion of the grains was measured by energy-dispersive x-ray diffraction techniques and found to be coincident with that of pure Pb up to the limit of the present measurements. L 3 -edge EXAFS measurements of solid Pb at various temperatures have been analysed using advanced data-analysis techniques (GNXAS) based on exact spherical-wave multiple-scattering simulation of the absorption cross-section. Realistic structural models for solid Pb were obtained from MD simulations using an empirical pair potential (Dzugutov, Larsson and Ebbsjo (DLE)), a tight-binding (TB) square-root functional, and an embedded-atom (EA) model potential parametrized by us. The short-range pair distribution function g ( r ) reconstructed by means of EXAFS is compared with those obtained by MD simulations. The empirical DLE potential, originally designed for the liquid state, is too soft, showing too-large values for the average distance R , variance σ 2 , and skewness β. The TB and EA potentials are both compatible with XAS data as regards the average distance and skewness of the first neighbours. The distance variance, associated with the thermal vibration amplitudes, is underestimated for the TB potential, while the EA model is found to be in agreement with XAS data. The present results are also compared with those from a previous EXAFS study on solid lead, where the cumulant expansion and a simple one-dimensional anharmonic oscillator model were used. The need for realistic interaction models and appropriate simulation schemes for reliable XAS data analysis is emphasized, while differences from and improvements with respect to previous approaches are only briefly discussed.