Advancement and applications of the template matching approach to indexing electron backscatter patterns

Abstract

Electron backscatter diffraction is a well-established characterisation technique used to determine the orientation and crystal phase of a crystalline material. A pattern is formed by dynamical interaction of elections with the crystal lattice, which can be understood and simulated by using Bloch wave theory. The conventional method of indexing a diffraction pattern is to use a Hough transform to convert the lines of the pattern to points that are easily accessible to a computer. As the bands of the pattern are direct projections of the crystal planes, the interplanar angles can then be computed and compared to a look up table to determine phase and orientation. This method works well for most examples, however, is not well suited to more complex unit cells, due to the fact it ignores more subtle features of the patterns. This thesis proposes a refined template matching approach which uses efficient pattern matching algorithms, such as those used in the field of computer vision, for phase determination and orientation analysis. This thesis introduces the method and demonstrates its efficacy, as well as introducing advanced methods for pseudosymmetry analysis and phase mapping. A new metric for phase confidence is also proposed and the refined method is shown to be able to correctly determine phases and pseudosymmetric orientations. Finally, preliminary work on a direct electron detector stage is presented. Work on the development, testing the pattern centre reliability, modulation transfer and an example map is shown.