On the uniqueness of structure extracted from diffraction experiments on liquids and glasses

There is continued interest in the problem of extracting structures from x-ray and neutron diffraction data on liquids and glasses. Traditional Fourier transform techniques, with their inherent weakness of possible systematic and truncation artefacts being introduced into the estimated distribution functions, are increasingly being complemented by computer simulation methods. These allow three-dimensional models of the scattering system to be built, at the correct atomic number density, which are consistent with both the diffraction data themselves and with other known or estimated constraints such minimum particle separations. Here the empirical potential structure refinement (EPSR) method is used to explore structure in supercooled liquid Ni, amorphous Ge and amorphous GeSe₂, and to evaluate alternative versions of the radial distribution functions which are consistent with the diffraction data. In the case of liquid Ni, it is found that there is, based on the diffraction data, some uncertainty on the hardness and shape of the repulsive core of the interatomic pair potential, and this may influence the current debate about the existence of icosahedral order in this liquid. For amorphous Ge two distinct radial distribution functions are generated, both consistent with the diffraction data, one of which has strong tetrahedral local order with the other having a predominantly triangular local coordination. For amorphous GeSe₂ it is found the SeSe and GeSe radial distribution functions can be determined well from the data, but the GeGe distribution is more uncertain, with the best fits implying both GeGe and SeSe homopolar bonds as originally proposed. The results are used to discuss the ambiguities inherent in the structural interpretation of diffraction data, even for one- and two-component systems.