Actually this is a good question which arises very often. What is the advantage of TKD by SEM in comparison to TKD by TEM? The most common argument of higher spatial resolution is true if one makes a comparison to backscattered Kikuchi diffraction (BKD) by SEM (bulk samples). I think you are right; we have to admit no one will prepare a TEM-sample just for TKD in the SEM. At least you will not have the technique of sample preparation if you don’t have a TEM available. Then the argument of higher spatial resolution is nullified if a modern scanning transmission electron microscope (STEM) is available. But some labs just have a conventional TEM without scanning unit and then the spatial resolution is limited by the aperture size when using selected area diffraction (SAD) or by the spot diameter of the converged beam (CBED) which can be quite large (up to 20 nm). So indeed in this special case, TKD in the SEM will have an advantage of spatial resolution in comparison to (conventional) TEM. Besides some rather trivial advantages of TKD in SEM (e.g. easy to use SEM, comfortable EBSD-Software, no need to buy expensive equipment or software for analyzing randomly oriented pattern in the TEM) a big advantage might come up when it turns out that the sample is not stable under the electron beam in TEM at high voltage (e.g. 200 - 400 kV). Then the SEM with 30 kV is a nice alternative to observe (optimally SEM equipped with STEM-detector) and analyze sensitive samples (e.g. glass-ceramics).

I will just add to the previous answer that TKD might have a better angular resolution compared to ASTAR (diffraction spots) technique. I don't know if the reason is physical because of the interaction or if it's due to the processing (more precise to fit lines compare to spots ?) .. but it's seem hard to get less than 1 degree of precision in the grain orientation using ASTAR technique in a classic way, whereas in EBSD i have something around 0.1 degree in mind for the sensitivity (What i said needs to be confirmed :P). The main reason to use a TEM is the spatial resolution (some nm) ... i would imagine it's difficult to characterize grains smaller than 10nm2 in any SEM based technique.

However, i agree on the fact that perform TKD on a thin lamella with a SEM @ 30keV could be very interesting. 

Good question Gert.

My opinion is that the TKD has a high potential for automatic analysis since the degree of automation is higher in a SEM than a TEM. A higher automation would open for better quantitative analysis (better statistics), which is one of the drawbacks with TEM as a technique. Good statistics can then also compensate for less precision. This would also motivate a development in sample preparation towards larger suitable areas.

For grain size analysis (single phase) I guess it depends on wether the comparision is done with phase-contrast imaging in TEM or STEM. However, in both cases I see an advantage with TKD in how to interpret the data, even if the advantages may be different depending on what you compare with.

For qualitative phase analysis I don´t know.

Well, as usual there is no black and white. TKD (or however you want to call it) has certainly a theoretical potential and there are a few applications which are hard or impossible to do with EBSD or TEM. However this does not mean that everybody can do it if an SEM and an EBSD is available. There are certain limitations regarding the sample preparation. TEM usually shows the interaction with the "bulk" of lamella whereas TKD only shows the signal of the last few nanometers. This means that the requirement regarding sample prep are even higher than for TEM.  It is definitely not a  simple technique...and because of the sample size the definition of the sample coordinate system as reference is quantitatively very poor. If one does not need absolute orientations or simply nice and colorful images to impress people it is OK. :-)

Hi, I came across this discussion a bit late but I wanted to chime in.  I think there are a growing number of researchers who have had success with t-EBSD.  Obviously, just like regular EBSD, sample preparation is an issue that needs to be carefully thought out before good data can be obtained.  For transmission EBSD in particular, trying to map through multiple grains or a thick sample may not be successful.   With atom probe specimens, many researchers are currently using t-EBSD to identify positions of grain boundaries or specific orientations.  In terms of speed, I would say that maps of an atom probe tip (maybe 1 micron square area) take around 3 minutes, which is acceptable and fast.  So to say that it only has theoretical potential is doing this technique a disservice.  We are using it routinely and it's very helpful. 

 Katherine: Every technique has certain benefits, and everything is relative. I, for instance, only accidentally know that here in Berlin is probably only one institute which has an atom probe. I assume these instruments are even more rare then TEMs, right?

Especially for you TKD is quite useful, or let's say: convenient since otherwise you have to determine your interface with other tools as the years before. Your big advantage is that you know exactly the orientation of your sample which is not the case if you are working with TEM samples. They are often somewhere fixed or resided on a grid (FIB) or somehow bend so that in maximum only relative information can be extracted, and even this only roughly because of sample relaxation, preparation artifacts etc. Mostly they give nice and colorful images :-).  Better than nothing, definitely, but the limitations regarding information content should be clear to everyone. It is not only the technique which is limited, it is the postprocessing tool box which cannot be used with the same reliability as for EBSD where the alignment of the sample surface (flat and  tilted by a very specific angle around a definite axis) is a very important requirement. If this is not correct, half of the tools will produce incorrect or even wrong results. So the orientation determination which is derived with respect to your SEM coordinate system and not to your sample coordinate system. Again: for an application in atom probe the conditions are quite perfect.

Oops, I posted in the wrong discussion, I meant to post in one about speed and mapping.  There are some ways to mitigate the angle problem to know more exactly the angle of electron incidence.  For example, the specimen can be aligned to be normal to the ion beam in the FIB, providing a known tilt angle of (90-52) deg or 38 degree tilt from horizontal for EBSD.  Taking into account your elevation angle, the orientations can be correctly calculated with the normal direction at the bottom surface.  This doesn't work for every sample, but it can help. 

Dear Katherine, from my point of view you are right as long as you are using a micromanipulator, or a prepared sample for 3D EBSD. The you still have you sample position under control. However, immediately if you put your sample on a carbon layer on a TEM grid you don't know anymore the tilt of the FIB lamellae. All described orientations are then wrong by at least this unknown angle. Or am I wrong? 

One way that you can determine the correct angle is by imaging the lamella (breifly) with the ion beam.  For example, if you have a TEM lamella mounted perpendicular to the half grid post as is most often the case, you can tilt the grid and image the lamella such that you determine the position where the lamella is exactly normal with the ion beam.  This often differs from what you have nominally set the stage tilt to be because the grid can have some bend it.  When you have set the tilt to make the lamella normal to the ion beam, the real EBSD tilt from horizontal is 38 degrees for an ion beam mounted at 52 degrees.  We've found a good correspondence with expected results using this method.

Kathrine, in fact you only need a few images and you can calculate the sample surface (e.g. by Alicona or any other software). This is clear, but no standard in EBSD software. It needs some manual operation and a lot of knowledge, how to correct the Euler angles or how samples are tilted. But I appreciate, that you are seriously thinking about it. Many thanks!