Length scale dependence and reliability of x-ray strain measurements in metallic glasses.

摘要

There is a long history of using x-ray scattering techniques for measuring elastic strain in crystalline materials. Only recently have these techniques been applied to metallic glasses. Results emerging from this work indicate that the elastic strain in metallic glasses varies with the distance from an average atom r in an asymptotic fashion. The reason for such anomalous length-scale behavior of elastic strain in these materials remains unclear.;To investigate the origins of this behavior, high energy x-ray scattering experiments were carried out in this work on zirconium and palladium based metallic glass specimens subject to mechanical loading. In both these alloys, the elastic strain obtained from pair correlation function position shifts was found to approach the imposed strain asymptotically. This behavior was found to be independent of the uniaxial compression, uniaxial tension, or pure shear loading geometries.;To further investigate the length-scale behavior, molecular dynamics simulations were carried out on a binary Lennard-Jones glass. Elastic strain in this model binary glass, when subject to uniaxial tensile deformation, qualitatively reproduced the length-scale behavior observed in the experiments when the strain was calculated from pair correlation function positions shifts. Other techniques for calculating strain, however, gave different results. Also, independent of the calculation methodology, imposing hydrostatic deformation resulted in nearly length-scale independent strain. These results seem to indicate that the presence of resolved shear stresses and non-affine displacements are central to the length-scale effect.;Besides the anomalous length-scale behavior, recent data also suggest that the x-ray technique usually underestimates the elastic strain (with respect to macroscopic value), or overestimates the elastic modulus, in metallic glasses. To investigate this modulus discrepancy, a two-phase model, containing low density liquid-like regions and high density relaxed regions, was investigated. This model was found to be unsuitable because the density difference required to explain the observed discrepancy was too large to be physically meaningful. However, assuming presence of voids, as an extreme of the two phase model, was found to explain the observed discrepancy.