Investigation of medium to high strain deformation microstructures using an automated electron back scatter pattern (EBSP) system

摘要

The speed of automated EBSP orientation gathering has now increased to a point where it is possible to examine large areas of deformed material at a resolution close to the length scale of the dislocation substructures involved (typically < 1 (micro)m for strains of greater than (epsilon) = 1). Investigation of such deformed samples invariably gives rise to diffraction images of low quality and to many double patterns. To assist in a critical selection of the data, it is usual to store a measure of the indexing confidence for each point. Here the authors describe how deformed samples, which provide the appropriate range of pattern qualities, can be used to calibrate acceptance criteria for unsupervised EBSP measurements on deformation microstructures. An application in the use of such criteria is also described. Two crystals chosen with orientations that are known to develop a well defined texture spread during deformation were channel die deformed to a strain of (epsilon) = 1.5 (78% reduction). The texture spread was determined for each sample from a large number of transmission electron microscope (TEM) orientation measurements taken using a fast semi-automatic technique. EBSP scans were then made on the samples. Orientations failing within the expected (TEM determined) texture spread are assumed to be correct. While those falling outside the expected spread are assumed to be incorrect. Unsupervised acceptance criteria based on an indexing confidence parameter (in their case the fraction of located Kikuchi bands matching the indexing solution) were then tested by comparing the set of orientations selected using a given approval criterion with the set of the orientations falling within the expected spread. Criteria were appraised by determining both the fraction of all orientations that were misclassified (incorrect orientations selected plus correct orientations not selected) or by the fraction of incorrect orientations in the selected data set.