As you certainly know the chalcopyrite structure is a klassengleiche subgroup of sphalerite. In orther words, caused by ordering the cations the lattice needs to be thinned out in order to describe the translation symmetry correctly. Since practically two cube-shaped unit cells are stacked over each other, the resulting c/a ratio is often very close to 2. Therefore, all bands visible in sphalerite also appear in chalcopyrite, only the indexing is different (cubic hkl become tetragonal hk2*l ) but the band widths are practically identical. Since also the structure is nearly the same all band intensities doe not change considerably so that a correct indexing is challenging. If the cations are not that much different only the deviation of c/a from 2 can help. If the deviation is below 2% you cannot get reliable results. Standard EBSD indexing might have a chance if the deviation of c/a is bigger than 4%. However then the projection center should be very well-known. Good pattern quality is essential. Use the full dynamic of the camera. Only if the full dynamic still results in noisy patterns use frame averaging. I do not know why did you start with low-energetic electrons. If your crystal are not in the scale of nanometers use 20kV or higher. If your crystals are in the scale of µm, the higher the better! The patterns become more brilliant, the measurement time is shorter, the band are more narrow and therefore better detectable.

Don't hope higher resolved patterns are better than lower resolved. Internally, indexing software bins down all patterns in order to get acceptable processing time for the Hough transform. The often observed increase of orientation accuracy has nothing to do with the pattern resolution but with the reduces noise. The higher resolved the patterns the longer the acquisition. A twice lower binned pattern increases the acquisition time by a factor of 4. The down-binned image has then a for-times lower noise which improves the band detection and so the orientation precision. You can test this using a 4x4 binned images and then 8x8 binned with 4 frames. The acquisition time is the same ...and you will see....also the orientation noise.    

Thanks Gert

This was what I expected. We ion polished the sample today and got great indexing of sphalerite and isocubanite which were both indexed as sphalerite. It seems also that cubanite lammellae could be distinguished from the other phases, whereas we had to leave out chalcopyrite structure, because it was preferred by the OIM indexing software because of more bands and hence also more votes. My impression from the overlayed indexing suggestions of sphalerite is that main bands are common, however there are some additional bands in the OIM chalcopyrite structure.

I have the same experience as you with the binning of patterns for most minerals and I agree that it is caused by the noise reduction. For some silicate minerals however there are some narrow bands that requires high pattern resolution. It is always a challenge setting the Hough such that we get both the narrow and broad bands without detecting the noise as peaks in the Hough. 

For silicates with high lattice parameters you should test a smaller butterfly filter. Then it should also work with higher binning. Never spend more time than necessary! It is wasted time. However, if you have to deal with phases of small and big lattice parameters it might be problematic (although the filter anyway does not fit even for the bands of a single phase). It would not be any problem, if TSL would develop a tool which automatically selects the best settings. When I worked for Bruker and developed the system i forced them to do this from the beginning. You only select the phase, nothing else. No bands, no reflectors, no filter size... And you can see from their mappings: it works without serious problems. The postprocessing is still a mess...however my opinion is anyway that post-processing is not the topic of EBSD companies. They should deliver most reliable data...and some fundamental post-processing tools. Everything else is always "problematic" since nobody really knows what is hidden in those black-box software packages. Too slow, too complicated, too mysterious... Software like MTEX are much better since open software and script-controlled. You always know what you did and you can use the same scripts again if you want compare data.  

I have another question to you: Do you really need a separation of all these phases? If yes, could you send me the sample? I could apply a pattern matching. If nothing works, this will do it (in case it s solvable). Since we compare dynamically simulated with experimental patterns and not only a few bands which are possibly also unsuitably detected, the chance to get the best matching phase is quite high. Solid solutions are of course also not that easy since uncertain....

Did you ever tried to combine EBSD and EDS? For step sizes in the scale of 100 or more nm it delivers suitable results.  

I'm sorry for offtop, but could you please write how did you perform an ion milling for EBSD sample preparation? Which gas did you use and angles of incident plasma beam? Would be highly pleased for answer.

hi Vladimir

We Ar ions 2 kev for 20 mins, but I dont think you need that much time. I dont remember gas flow rates but I think the standard recommended by the supplier. for EBSD you need a high tilt angle because low angles gives to much etching effect.

Thank you wery much, I tried zero tilt, obtained good patterns from a few points and totally ruined surface. Now will try tilted.