Precision of orientation and misorientation angle+axis are crucial in terms of analyzing the EBSD data in very smaller ranges. For reference, I am attaching Wilkinson 2001 paper on exactly same subject. But since EBSD systems have improved a lot since 2001, I am interested in knowing the routine/typical values in today's perspective for a well-annealed material under optimum SEM conditions.
Attached: Wilkinson 2001 paper.
I have to disappoint you. EBSD systems are nowadays faster, and new post processing tools have been derived, but in principle from the beginning nothing has been dramatically changed regarding precision and accuracy. The reason is that also nowadays the identical indexing and band detection is used as 20 years ago. Oxford Instr. developed approx. two years ago a better band-edge detection, and TSL recently improved the detectibility by an averaging of adjacent patterns if they are saved, however the last mentioned automatically means a loss in spatial resolution since you assign a solution from a bigger area to a smaller one. This is OK for a simple orientation description since you can improve the signal to a level that it still can be indexed but generate artifacts in misorientation evaluation tools. This means, one should carefully think about the tools you can apply. The band-edge detection developed by Oxford Instr. mainly improves the correct band detection for TKD measurements, i.e. if the gnomonic distortions become remarkable and do not fit to the developed peak search in Hough space. For standard measurements it shall be also better, but a) it needs more time (of course) and b) I havn't seen that much where the effect has been proved. They certainly showed some maps But I havn't heared a remarkably reaction from Aztec users.
Another approach is used in Pittsburgh (Carnegie Mellon University, M. de Graef) where simulations are used to index, similar like ASTAR for SAED patterns in TEM. The limitation is the accuracy of the parsed orientation space. As far as I've heard the accuracy does not become better, only the precision.
A similar approach we are using by pattern matching however we are refining the orientation i.e. we do not use a prepared database containing hundred thousands of pre-calculated patterns. This increases accuracy and precision. However, the major request are sufficient patterns which do not require an averaging, i.e. the better the patterns the better accuracy and precision. In so far, the detector development is important which delivers better diffraction patterns with less noise. Here the new Nordlys detector is obviously the best one: very sensitive and fast enough. The resolution is not that important since diffraction patterns do not show very sharp features. Only for HR-EBSD (strain determination) the image must have as many as possible pixels for mathematical reasons, not for the actual signal resolution.
The common precision of an orientation measurement by EBSD (using "standard" acquisition conditions during mapping and high binning) are still in the scale of 0.5...1°. Changing the conditions (higher resolved and averages frames, higher Hough resolution, refined band detection) Oxford showed that it is possible to become up to 5 times better, but this is not for mappings. The pattern matching approach (which is some kind of brutal-force technique which shall only show that with common EBSD pattern much more can be done) delivers an about 3 times better accuracy and precision.
I have to disappoint you. EBSD systems are nowadays faster, and new post processing tools have been derived, but in principle from the beginning nothing has been dramatically changed regarding precision and accuracy. The reason is that also nowadays the identical indexing and band detection is used as 20 years ago. Oxford Instr. developed approx. two years ago a better band-edge detection, and TSL recently improved the detectibility by an averaging of adjacent patterns if they are saved, however the last mentioned automatically means a loss in spatial resolution since you assign a solution from a bigger area to a smaller one. This is OK for a simple orientation description since you can improve the signal to a level that it still can be indexed but generate artifacts in misorientation evaluation tools. This means, one should carefully think about the tools you can apply. The band-edge detection developed by Oxford Instr. mainly improves the correct band detection for TKD measurements, i.e. if the gnomonic distortions become remarkable and do not fit to the developed peak search in Hough space. For standard measurements it shall be also better, but a) it needs more time (of course) and b) I havn't seen that much where the effect has been proved. They certainly showed some maps But I havn't heared a remarkably reaction from Aztec users.
Another approach is used in Pittsburgh (Carnegie Mellon University, M. de Graef) where simulations are used to index, similar like ASTAR for SAED patterns in TEM. The limitation is the accuracy of the parsed orientation space. As far as I've heard the accuracy does not become better, only the precision.
A similar approach we are using by pattern matching however we are refining the orientation i.e. we do not use a prepared database containing hundred thousands of pre-calculated patterns. This increases accuracy and precision. However, the major request are sufficient patterns which do not require an averaging, i.e. the better the patterns the better accuracy and precision. In so far, the detector development is important which delivers better diffraction patterns with less noise. Here the new Nordlys detector is obviously the best one: very sensitive and fast enough. The resolution is not that important since diffraction patterns do not show very sharp features. Only for HR-EBSD (strain determination) the image must have as many as possible pixels for mathematical reasons, not for the actual signal resolution.
The common precision of an orientation measurement by EBSD (using "standard" acquisition conditions during mapping and high binning) are still in the scale of 0.5...1°. Changing the conditions (higher resolved and averages frames, higher Hough resolution, refined band detection) Oxford showed that it is possible to become up to 5 times better, but this is not for mappings. The pattern matching approach (which is some kind of brutal-force technique which shall only show that with common EBSD pattern much more can be done) delivers an about 3 times better accuracy and precision.
I basically agree with Gert concerning the accuracy of misorientation angles (of order 0.5°) but wish to make a couple of comments concerning misorientation axes. As Wilkinson observed in 2001 the scatter of misorientation axes increases when the misorientation angle decreases. This was later quantified by Pete Bate et al who gave the axis accuracy as tan-1 (alpha/theta) where alpha is the angular accuracy (about 0.5°) and theta the measured angle. See J Microscopy (2005) vol 220, 36-46.
Secondly when the misorientation is small, misorientation maps become very poor to visualize. However, and perhaps surprisingly, the corresponding maps of misorientation axes are often much better when the axis is given a standard colouring scheme - see Albou et al Acta Mater (2010) vol 58, 3022-34.
@ Gert Nolze: Thanks for enlightening about the present status of EBSD systems.
According to Wilkinson (2001), orientation precision is 0.5 deg and misorientation angle precision is around 0.2 deg for higher misorientation angle, whereas problem stays for misorientation axis (around 1-2 deg precision at misorientation angles > 20 deg), especially at lower misorientation angles.
I also calculated for our EBSD (with Nordlys nano detector, high sensitivity) system, following a similar approach as that of Wilkinson (2001) for conventional EBSD, i.e. no Cross-court and I obtained orientation precision close to 0.15 and misorientation angle precision close to 0.12 deg. So looking at these values, I thought this slight improvement is because of improved technology. But now after reading your answer, I am still wondering whether the values that I obtained are quite reliable/usable or not.
Further, My actual concern is about the precision with which we can measure the misorientation axis, especially at higher misorientation angles, in context of deviation from exact twin configuration in FCC (i.e. misorientation angle as well as axis). What is the minimum angular step for twin-deviation angles we can use for making sense out of it?
Hi Julian, thanks for the additional comments. I did not want to give any advice regarding the axis/angle presentation. From my point of view it should be clear that for very small misorientation angles in the scale of the limit you should not have any misorientation axis preference. If this is not fulfilled it means from my point of view that the orientation determination algorithm has some "problems" and artificially generates a preference. From my opinion this does not have to surprise if a band detection is used selecting only a certain number of bands. On the other hand I would be very careful in interpreting such results.
Misorientation also depends on the pattern quality, and surprisingly it does not always mean that better patterns deliver a higher accuracy and precision. The algorithms are optimized/developed for the band center position, and all patterns deliver fantastic results if they fulfill this assumption of a nice intensity maximum within the center. Therefore, sometimes slightly blurred patterns are "better" than very sharp one, since there often a misinterpretation of a band edge as band center arises (which is one of the major problems during indexing). For blurred patterns this does not happen that often :-). Sounds a bit strange, but I guess, everybody has seen this effect (especially for wide bands since there the internal restrains do not work that nicely anymore). Therefore, Oxford Instr. developed this new algorithm in order to prevent this misdetection of important bands since they mainly decide about precision and accuracy. The problem of post-processing of orientation data is, that you don't see this if you only have the orientation description as Euler angles but no pattern anymore. In so far, any orientation description contains a systematic and orientation dependent error which we do not know, but we perhaps misinterpret them as real existing affect, like for GND evaluation. I can only recommend to have a look at page 18 in the linked presentation. Both KAM images are based on the same patterns and the maps speak for themselves. As long as I do not know, how systematically wrong orientation interpretation are I do not trust misorientations quantitatively. If you see these maps, e.g. in pyrite in Fig. 7 of the attached paper, there are specific grains display a remarkably higher or lower KAM, but a more careful inspection shows that there is nothing and each grain shows (practically the same misorientation.
Conference Paper EBSD: more reliable and precise
Article Pattern matching approach to pseudosymmetry problems in Elec...
- Badges
- Science topic
- Similar topics
- Engineering
- Materials Engineering
- Annealing
- Metallurgy