566 doi:10.1017/S143192761800332X Microsc. Microanal. 24 (Suppl 1), 2018 © Microscopy Society of America 2018 Automated Prediction of Pseudo-Symmetry Issues in EBSD P.G. Callahan 1 , Saransh Singh 2 , M. Echlin 1 , J.C. Stinville 1 , T.M. Pollock 1 , and Marc De Graef 2 1 Dept. of Materials, University of California at Santa Barbara, Santa Barbara, CA, USA 2 Dept. of Materials Science and Engineering, Carnegie Mellon Univ., Pittsburgh, PA, USA Pseudo-symmetry (PS) issues arise in the analysis of electron back-scatter diffraction (EBSD) patterns when similarities between patterns for different crystal orientations cause the indexing algorithm to select the incorrect orientation. In inverse pole figure (IPF) maps, the result often appears as grains with a random distribution of two colors instead of a single color. While PS issues are readily recog-nizable in IPF maps, and can be corrected manually by post-processing of the map, the prediction of the occurrence of pseudo-symmetry is not as straightforward. In this contribution, we present a new algorithm capable of predicting for which crystal orientations PS issues may occur. Using the physics-based forward model described in [1], along with an orientation space sampling approach, and a set of detector parameters, we generate a dictionary D of simulated EBSD patterns. The dictionary indexing approach [2] is then employed to match the dictionary against itself, i.e., each pattern in D is compared to all patterns in D using the dot product as a similarity metric, and the top N dot product values along with the indices of the corresponding patterns are kept. For the set of N best matches, we expect the top match to occur when a pattern is compared to itself; the other N - 1 matches are generally expected to correspond to orientations that are close to the correct orientation. Hence we compute the disorientation between the top match and the other matches; if all disorientation angles fall below a threshold value, then there is no risk of pseudo-symmetry issues. If, on the other hand, there are matches among the top N for which the disorientation angle is substantially larger than the threshold value, then there is the potential for an indexing algorithm to incorrectly assign this lower ranked orientation to the original pattern. The algorithm was tested using the SrTiO 3 cubic perovskite structure. A total of 333, 227 patterns was simulated for a microscope accelerating voltage of 20 kV, pattern size 480 × 480, detector pixel size 50 μm, detector tilt 10 ◦ , sample tilt 70 ◦ , and pattern center coordinates (x ∗ , y ∗ , z ∗ ) = (0.5, 0.5, 0.625). After binning by a factor of 8×, the 60 × 60 pixel patterns were stored in a dictionary D and indexed against themselves, keeping the top N = 20 matches. In all cases, the top dictionary match was the pattern itself. For all other 19 patterns, the disorientations with respect to the best matching pattern were computed; disorientation angles larger than 4 ◦ were considered to be due to pseudo-symmetry. Of the 6, 664, 540 misorientation angles, 99.35% were smaller than the threshold of 4 ◦ ; 40, 649 angles were centered around a value of 45 ◦ , 2, 716 around 60 ◦ , and 171 near the maximum possible misorien-tation angle for cubic symmetry, 62.7994 ◦ . Fig. 1(a) shows the histogram of misorientations using bins of 0.1 ◦ width; the dashed lines indicate the 4 ◦ threshold (left) and the maximum cubic misorientation angle (right). The orientations for which pseudo-symmetry is possible (for the given detector param-eters) are shown in four views of the cubic Rodrigues fundamental zone in Fig. 1(b); each blue-gray sphere represents a single orientation. The corresponding misorientations cluster in three regions on the outer surfaces of the cubic misorientation MacKenzie fundamental zone shown in Fig. 1(c). The bottom portion of the figure shows six pairs of pseudo-symmetry related patterns, two pairs for each of the peaks in Fig. 1(a). In each case, a portion of the two patterns is virtually identical and the dif-ferences are typically located near the outer edge of the patterns. For real experimental patterns, in the presence of noise, it is entirely possible https://doi.org/10.1017/S143192761800332X Downloaded from https://www.cambridge.org/core. IP address: 192.3.180.147, on 22 Apr 2020 at 02:40:03, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms.