The effect of fluctuation microscopy constraints on reverse Monte Carlo models of metallic glass structure

Although discernible atomic ordering in amorphous materials is not thought to extend much beyond two nanometers, such length scales are beyond the limits of what a two-body correlation from x-ray and neutron scattering experiments can clearly illuminate. In order to incorporate nanometer scale order into current models of amorphous solids, we must turn to experimental characterization techniques that are sensitive to higher order correlation functions.;One such technique is fluctuation transmission electron microscopy (FEM) which is sensitive simultaneously to two-, three-, and four-body correlation functions and has been shown to successfully differentiate between models of amorphous materials that differ on the nanometer scale. We set out to generate a simulated atomic model of a metallic glass that is not only consistent with a number of complementary experimental two-body scattering spectra, but also with measured fluctuation electron microscopy patterns using reverse Monte Carlo computer simulations.;This computational technique refines an atomic structure until it is consistent with experimental scattering spectra within errors without introducing any structure into the model a priori. Resonant x-ray scattering, neutron scattering, and FEM spectra were measured from Pd40Ni 40P20 bulk metallic glass samples and used as constraints on RMC simulations. To determine the influence of the higher order correlation functions from FEM on an atomic simulation, models were run with only two-body constraints and with two-body plus FEM constraints. It was found that a RMC model constrained only by two body-correlations could reproduce qualitative features of the FEM experiment. The two models turned out to be nearly statistically equivalent out to the nanometer scale, suggesting that the short range structural information contained in two-body functions is sufficient to produce nanometer-scale order.;Reverse Monte Carlo was also used to successfully identify subtle structural differences between Pd40Ni40P20 bulk samples quenched at cooling rates. Both samples displayed mostly icosahedral local ordering, with the slowly cooled sample model showing sharpened structural features in the first and second coordination shells. This observation is consistent with the notion that high cooling rates applied to metallic glasses quench in a more disordered atomic structure.