@Evgeny: you are certainly right, but if you do not have a TEM...or the vertical resolution of TEM is perhaps worse, or radiation damages...? The list could be long to find applications. I am not very sure how many people really need TKD, but this is another question. Here I only want hear what people think about a screen below the sample and which advantages or disadvantages it possibly has?

Dear Dr. Nolze,

I have read a paper published in 2013 in Microscopy Today about this topic (doi:10.1017/S1551929513000503). The paper is authored by researchers from NIST (USA).

"Transmission EBSD in the Scanning Electron Microscope"

Roy H. Geiss, Katherine P. Rice, and Robert R. Keller

Regards,

Hugo

Dear Dr. Nolze,

I have been regularly using TEM samples inside SEM and doing TKD or T-EBSD. You need sample holder to hold thin TEM foil. The EBSD detector is placed close to the samples. You can go Oxford Instruments site to get details about the sample holder. TKD is works well for the samples having nanocrystalline features. If  you need any further information you can write to me.

Gert,

Bruker just launched their new type of TKD camera at M&M this summer. With this system, you can change the standard EBSD camera set-up for a transmission set-up with the camera below the specimen and normal to the beam. Spot diffraction is now available in the SEM! I didn't use it already but I guess from what I've seen that indexing Kikuchi bands with the spots around will necessitate to modify the indexing procedures and algorithms.

Nicolas

@ A. G. Rao: Why do you use TEM samples in SEM and not in TEM? Is there any advantage from your point of view, or don't you have any TEM?

@Nicolas: I've seen a presentation during the RMS meeting in Glasgow this spring. Looks interesting, I only did not know that it hasn't been launched before M&M.

The problems I see from the presented images at their webpage are TEM like spot patterns in the background (which one can perhaps filter out), but also high-order the asymmetric higher order interferences and from my point of view also high-indexed reflections so that the indexing can become quite difficult.Therefore I was asking who already has a system running and can give some more details about the indexing reliability. I wonder, which are the specific advantages of such an arrangement. For spot patterns you need to have quite thin samples. Any unclear overlay of a diffraction signal containing spot patterns, CBED and Kikuchi patterns is a really formidable challenge. What do you think? Were do you see the advantages of this arrangement (compared to the standard configuration?  

Thanks!

I don't know how Bruker does it but filtering any spot like (or in general non-linear) patterns shouldn't be a problem. You might have to play with the hough settings a bit or use a pre-filtering procedure if the intensity of the brightest points is in the same range as the integral intensity of an entire band which you want to detect but even that should be relatively straightforward. One could for instance invert the intensity of all pixels that exceed the grayscale on the upper end or set their level to some lower number before going into the hough transformation. Using the patterns directly for cross-correlation EBSD might be more difficult though.

I certainly see the advantage over the standard config in terms of speed and beam damage. Our scan speeds with TKD are painfully slow as high currents are not an option if you want high spatial resolution and work with systems that have a low thermal stability.

@ E. Bruder: What is painful slow? I've only seen maps where they claimed it is comparatively fast (50...100 patterns). I would only expect for very special cases faster maps using TKD. Theoretically many possibilities exist, but practically often people are disappointed when they start to use it. The data presented somewhere are certainly not wrong but clever mixed so that one has the impression everything works so as described for many phases. Fast measurements you can only expect if the SEM has enough current, your sample survives it, the signal is so strong that you can use a fast (and often less sensitive camera) which is able to save patterns in high speed. With a weak signal you can even measure faster with a high-resolution detector than with a "fast" and low-resolution detector. I don't know many people who really measure fast. I assume that for many people fast already means 100 patterns per second since the materials do not deliver such nice images as used for fast maps (Ni, recrystallized).    

For ultrafine grained or nanocrystalline copper based alloys we can't really go beyond 30fps (typically 10-20) without using very high gain settings / extreme binnings or loosing spatial resolution and risking to fry the sample.

I strongly agree that based on literature and company adverts TKD appears much easier to use than it actually is.

@ E. Bruder: Well, unfortunately also science is finally only a business :-). It is certainly more serious than selling carpets but the principle is the same: Expect that missing data are intentionally not mentioned. If they would be amazing, they would be specifically pointed out. Take this into account and usually you will get a more realistic view :-). The same is also valid for many papers. Not everything what is written there is valid for any other sample. Mostly this measurement has been done only once, or the approach demonstrated only expresses that it works in principle, but not in general. Otherwise, everybody would do it. Therefore, I am not surprised that also TKD will finally only have a small user group. Perhaps many people will try it but finally give up. One major problem is already that many samples are bent or somehow aligned so that the used reference coordinate system does have a defined relationship to the determined orientation but not to the sample. Misorientation is fine, but orientations are hard to interpret. Not talking about the statistical significance. Nevertheless, for nanocrystalline materials it might be a useful alternative to TEM, especially if you do not have access to TEM. But it is certainly not an easy to go technique for main street applications as claimed by some manufacturers.

Dear Dr. Nolze,

we did launch the new TKD detector head (OPTIMUS) in July and we showed it live in Portland during M&M. It was so well received by people that we practically run just one EBSD demo and for the rest of the show we had one SEM fully booked with TKD demos.

Joe Michael from Sandia National Laboratories in Albuquerque NM, USA is using OPTIMUS on his HR detector and he will be presenting results next week at the TKD seminar organized by Bob Keller at NIST in Boulder CO. During our  webinar on August 26th I showed a few maps that Joe acquired with his system. One nice test Joe did was to acquire two maps from the same area of a Au sample first with the "vertical" standard EBSD screen and then with the on-axis screen. The big improvement is spatial resolution and data quality is beyon any doubt. You can see it for yoursef by watching the recording of that webinar or by downloading the slides. Here is the link:

https://www.bruker.com/de/service/education-training/webinars/eds-ebsd-and-micro-ct-for-sem/quantax-ebsd-with-optimustm-tkd.html

If you have time you are of course welcome to pass by our offfice and we'll demonstrate for you the advantages of using the on-axis detector geometry.

Your comments on the difference between data presented by EBSD users and data showed by EBSD manufactures is totaly true and I believe that this is also very natural with a few exceptions of course,i.e. top experts that have the time to run measurements. EBSD is what we do all day so we know how to get the best out of an EBSD system while most EBSD results we see in conferences or papers are acquired by master and PhD students that have very little experience and not always just EBSD to stress about every day. I have noticed this at the time I was working in Hobro and it is still the case now. I guess that until EBSD/TKD will become more or less automatic we will continue to see this significant discrepancy in data quality.

Best regards,

Daniel

It's a clever idea and Daniels' webinar was excellent.

Re. The bright central spot.

I showed the attached slide at St Etienne in May 2010 - I think we were all there.  It's basically a scintillator with a hole, and a diode below for the bright-field imaging; plus a mirror & camera for TKD/tEBSD and other diodes for bright field imaging.

http://www.emse.fr/ebsd2010/

http://www.emse.fr/ebsd2010/pdf/day.pdf

Interesting comment, Austin. I did not realize that the idea of the screen in transmission position came actually from you :-).

Nevertheless, although TKD will contribute to the solution of some very specific problems, the general impression stays that it is not a technique for everyone. Either the instrumentation is not sufficient, or the users experience (as Daniel pointed out). Moreover, for the real user who often does not have the high-current instrument as EBSD companies will commonly be disappointed about the the difference between promised and real power. It is not that much different to 3D-EBSD, another overestimated technique. As long as we are not able to measure in 2D a stable map (please give a message who can do it and how), how reliable a 3D map in high resolution can be? In principle you have to correct every voxel position. No doubt, there are some contributions as well which would be very hard to produce by conventional preparation techniques, but it is also not a technique for everyone. There are many FIBs with EBSD which never generated a useful dataset.  And as I pointed out: we have to make clear what one can measure with TKD and what does it tell you. The remark to use simple imaging techniques for crystal size determination can't be denied. We already have too many colorful maps in papers which nobody really discusses, see Daniels remark about master and PhD works. Here I see an enormous potential for all these EBSD system attached at SEMs or FIBS which are not working because the operating people have no or only a very limited knowledge how to use it and - which is even more critical - how to interpret the resulting data or to have any impression how accurate and precise they are..