The effectiveness of automated electron backscatter diffraction in rock-forming minerals
Magda Lopez-Pedrosa1, Dave J. Prior1, John Wheeler1, Niels-Henrik Schmidt2
1 Department of Earth Sciences,4 Brownlow street, Liverpool,L69 3GP
2
HKL technology ApS, Majsmarken 1, Hobro, Denmark, DK-9500Electron Backscatter Diffraction (EBSD) in the Scanning Electron Microscope (SEM) provides the crystallographic orientation of individual rock-forming minerals from grains and subgrains as small as < 1µm. Errors may occur when the EBSD data is processed using computer based packages.
We present an assessment of the effectiveness of EBSD using the CHANNEL 5 software program (FLAMENCO, TANGO and MAMBO). The FLAMENCO program collected and indexed thousands of EBSD patterns in a rectangular grid mode from several rock-forming minerals. The same program was used to analyse the EBSD patterns off-line (away from the SEM) but with different acquisition parameters in the FLAMENCO software program, such as the hough resolution (resolution of the pattern), the number of bands picked by the hough transform and the number of reflectors in the match units. Approximately 20 maps were generated for each sample using the TANGO program and only one was selected to be the most correct map. The selected map was compared with SEM image, Orientation Contrast Image (OCI) and Secondary Electron Image (SEI) and processed in the TANGO software program to reconstruct the microstructure of the sample. The selected map was called "the reference map". Comparisons between the maps with the reference map allowed us to assess the errors in the EBSD data and the effects of acquisition parameters in the different crystal systems. For example, the optimum acquisition parameters in a sample of Pyrite (Fig.1) differ from those applied to a sample of Garnet to smooth the indexing problems, although both are cubic. The lack of the tetrad rotational axes in Pyrite give rises to severe misindexing problems in the EBSD maps even with the most optimal acquisition parameters in the FLAMENCO program.
a)
b)
c)
Fig.1 50000 EBSD patterns forming orientation maps from a sample of Pyrite, a) the reference map, b) 100 hough resolution, 5 bands and 50 reflectors gave rise to 21% misorientated points of all dataset b) 150 hough resolution, 8 bands and 75 reflectors gave rise to 7% misorientated points of all dataset. Either of them was compared with the reference map
.In contrast, the indexing problems in Garnet can be reduced considerably applying either high hough resolution or up to 6 bands in the FLAMENCO program.
On the other hand, less symmetric systems, such as Quartz, highlights severe indexing problems when the number or bands and the hough resolution in the FLAMENCO program are too low. However, these can be reduced considerably applying higher acquisition parameters. Therefore, the degree of complexity in the crystal systems cannot be easily treated by automated EBSD and thus orientation mapping often requires manual intervention to smooth the indexing problems.