Statistical Analysis of Grain Boundaries in the Space of Macroscopic Boundary Parameters
Current issues concerning the characterization of grain boundary networks via five-dimensional (5D) grain boundary distributions are considered. A quantitative measure of reliability of such distributions is adapted from conventional texture analysis. Application of the measure shows that with the currently available size of experimental data sets of boundary parameters, only strong components of the boundary distributions can be reliably evaluated. Improvements of the computational part of the analysis are possible if the the binning based on Euler and polar/azimuth angles is replaced by searching the data sets based on a suitably defined distance. Moreover, it is indicated that for textured materials the stereological approach has limited reliability. Finally, it is suggested that coherent twins can be used for estimating experimental errors, and that the distributions cannot be a basis for conclusions about tilt boundaries unless additional restrictions are applied. The approach used in the paper is theoretical with support by computer simulations.