Predicting plastic events and quantifying the local yield surface in 3D model glasses

Abstract

By applying the local yield stress (LYS) method to probe local regions of three-dimensional computational glass models, we confirm high correlations between the measured local yield stress (Δτc) and the plastic events when the parameterization of the method is properly optimized. The optimal probing region for this system is found to be ∼5σ in radius, where σ represents the Lennard-Jones length scale, approximately the atomic size. The averaged correlation remains positive through the first 200 identified plastic events or 1/3 of the yielding strain (∼7%). Here we apply only the local probing that aligns perfectly with the loading on the boundary. The LYS measurements converge to a Weibull distribution with a minimum Δτc indistinguishable from zero at larger probing region radii. Analysis of the data in light of an assumption that Δτc is a local quantity that obeys extreme value statistics above a critical length scale bounds the exponent of the underlying partial distribution of Δτc≲0.71. A thorough investigation of the anisotropy of the local yield surface at the location of the first plastic event indicates that the first triggered region does not align perfectly with the loading on the boundary, but is well-predicted by projecting the shear applied at the boundary onto the local yield surface. This implies that the correlation between the local yield stress prediction and the resulting plasticity may be enhanced by performing a more complete assessment of the local yield surface at each sample point.