What are the possible reasons of unvisibility of EBSD pattern? I have measured EBSD on substrate and subsequently on deposited layer (by PVD). There vere only some areas where the Kikuchi lines were visible (and indexed). The measurements were performed on as-deposited layer (with no treatment). It is not clear for me the reason why somewhere the kikuchi lines are quite clearly visible and on the other place they cannot be detected at all. I suppose that the layer should not be amorphous. Moreover the indexed areas have preffered orientation. I know that there are some factors influencing the pattern quality (e.g. dislocation density). Is it possible that the bcc metal has so much better conditions for diffraction in certain orientation? Or is there another possible explanation of such phenomenon?
There are numerous reasons that you only get partially EBSD patterns. The major problem is the very low information depth. Here I have to disagree with Zuzana. I cannot say that I measured it by myself, but it is general accepted and experimentally proved by different experienced labs that the information depth of EBSD is in the scale of a few tens nanometers. We do pattern simulations using dynamical theory of electron diffraction and there the interaction depth is implemented as well. When you don't adjust the depth dependency correctly you do not get a matching simulation. The information depth (not interaction depth which is certainly deeper) is in many fundamental papers discussed so that I would believe them. This is also the reason, why EBSD is extremely sensitive to surface defects (preparation, amorphization in FIB lamellae (100nm thick but you an do EBSD without any problem...except there is an amorphisation in the scale of a few nm only)), oxidation, pollution (contamination) etc. All these effects cause a decrease of usable diffraction signal. Let us keep this as first parameter: the depth commonly forming the Kikuchi signal is not crystalline. I have no idea how the microstructure in your case looks like, but for PVD it is unlikely. Only the size of grains might be too small. Related to this is the size of coherent domains. They can be either described by small angle boundaries (dislocation walls) or large angle boundaries. If these domains are smaller than the beam size you will collect simultaneously patterns from differently oriented domains which results in disappearing Kikuchi bands. This is well know for martensite measured at lower magnification. There the positioning of the (perhaps much smaller) beam is not that stable as assumed and averages over differently oriented grains. This is something I would assume in your case. Go to higher magnifications (not identical to higher resolutions by changing the step width and keeping the magnification!) in order to check this. Of course, the focusing (beam diameter) is important as well, but I guess this is a trivial and obvious factor. If you cannot see the details during imaging the surface you perhaps really need to optimize your astigmatism, focusing etc. If nothing helps, I would highly recommend to check these areas by TEM or at another EBSD facility.
I also assumed that the area of EBSD signal is lets say 10 nm (I have o exact idea for my case). Anyway it is less than the layer thickness (1 to 2 microns), as I have obtained signal from bcc layer on fcc substrate. Moreover the grain size on one of the samples was ~500 microns and the situation was the same. Looks like I need to the some more measurements.I just didn't want to miss any possible simple explanation. Thanks for help.
Which material? The information depth depends on Z (periodic number). For Fe it should be in the scale of 20nm, for Si around 60-80, for Pt below 10.
If your grains are given by 500µm and you observed the same, then I would assume a high defect density or a nano-crystalline layer on the substrate. Did you try to polish the surface gently in order to exclude this? I know, both explanations actually do not make any sense at PVD but it also makes no sense that it doesn't give any pattern. Therefore we have to thing in any direction.
I guess you have XRD. Did you made a test there as well? t would be helpful to see the peak widths. If you already now that you have a strong texture don't be surprised that you only find reflexes from planes perpendicular to the fiber (plane parallel to the surface).
Could you post a pattern quality image and an SE or BSE image in order to get a rough imagination about the problem? You could also collect a few results in a PPT and send it either to me or post it here.
Good luck
Gert
This is titanium alloy. The band contrast map is below. I have XRD results from this sample (not made by myself). It is supposed that they exhibit crystallic structure. The layer is really more than 1 micrometer thich (as EDS signal exhibit the layer composition - not strongly affected by substrate). I wanted to etch the sample a little - I am affraid that polishing could remove the whole layer.
I see. From the pattern quality I would assume that you have some oxidation on the top. Oxidation - as many other properties - is anisotrop, i.e. I would assume a slightly more oxidation for the other grains. My recommendation is a very short and gentle polishing with silica. Since silica polishing is a combination of mechanical and chemical surface attack, I will not exclude that it removes the layer, but if you use a microindenter (measurement of hardness) and only polish for a few seconds, you can compare, how much you removed. So you can control the polishing.
Wat I would exclude is a mechanical reason like a high dislocation density.
Thaks a lot for advice. I will try it. This could be rational explanation
Good luck, and please post how it went. It is always good to see if something works or was misleading or non-successful.
You can lean to optimise the surface quality by finding a proper time, keV, and angle for your sample. I did it with PECS polishing and found a fantastic solution for my work, even though a bit of time consuming but worth it. You do not need to loose much of the material but you may get flat surface that is what important for EBSD. Good luck
Also PECS polishing removes material (not that much as grinding of course), works for different orientations differently, and causes remarkable radiation damages even for 1.5kV tested on Fe. In order to show this we manually prepared a sample and when satisfied with the EBSD quality we added a PECS preparation. We always found a measureable decrease of pattern quality after PECS prep. Therefore we don't use it anymore because or manual preparation is less time consuming and less damaged. Exclusions are multiphase materials where it is hard to find a polishing procedure which works for all in a comparable way, e.g. solders. Only in these cases we are using PECS.
Refer this link:www.ebsd.com/ebsd-analysis/types-of-ebsd.../pattern-quality-maps
Refer this link:www.ebsd.com/ebsd-explained/...ebsd/interpreting-the-diffraction-patter...
Refer this link:ournals.cambridge.org/article_S143192761005467X
Refer this link:journals.cambridge.org/article_S1431927605507347
Refer this link:www.springer.com/cda/content/.../cda.../9780387333250-c2.pdf?...0...
DEAR BABU,
PLEASE GET SOME INFORMATION FROM THE WEBSITE GIVEN BELOW
BEST WISHES
http://www.ebsd.com/ebsd-analysis/types-of-ebsd-experiment/pattern-quality-maps
Dear Yang. i didn't want to influence the layer for the first measurement and this is the result. Yesterday I tried to polish the sample with silica. I think that about 200 nm (estimated approximately from the change of HV indentation). It means about 1 minute polishing. The result was even worse. Maybe etching could help, but anyway certain grains are still "unvisible" for EBSD. I have no PECS available, which could also help.
This is interesting. Could you post the SE images aquired before both EBSD measurements? Moreover, what does this strange feature at the buttom of your BC map reflect? A scratch? Why the surface looks that "dirty" (in your BC map)?
There are all three SE images (substrate, layer and layer after polishing). The band contrast is OK for substrate (about 90% indexed). The scratch at the bottom is from me (to have an orientation point on the sample). It is really strange. for me. I do not know why the surface looks dirty (on band contrast). anzwaz the lazer has worse contrast than polished substrate.
What is the probe current you are using?
Try to increase it to above 4-5 nA and try again
Hi Jaroslav,
Very interesting discussion. EBSD does frequently cause confusion on why the software is not indexing the bands.
First of all if you see a defined XRD spectrum on the same sample does NOT imply that all your sample is crystalline. There might still be amorphous areas. These will, obviously not diffract, so they simply will not show in your spectrum.
Second consideration, especially valid for depositions. If there is an intense residual strain in your material the software might have a hard time indexing the bands simply because the parameters of the elemental cell will be different in the strained areas.
Third consideration: surface preparation for EBSD is the more difficult the smaller the grains. So if you expect a very small grain be careful, go beyond silica and try ion milling or electrochemical etching. Here you really need to try different solutions to see which one fits best for your sample-
Fourth and last consideration: if the grain size of the sample or of second phases inside the sample is below 300-400 nm the percentage of indexed areas will be small. This is a limit of current EBSD technology due to the size of the beam and the resolution of the sensor.
Hope you make it, keep us informed!
Alessandro
If you compare all three images...didn't you see always the grain boundaries of the substrate? And don't they match the band contrast? in other words....Do you really have a layer ? :-)
Hello Mr. Malek,
I aggree with Mr. Fais concerning the fact that maybe not all of your material might be crystalline, although I think I can see a bit of your grain boundaries in your SEM-Images. Concerning the possibility that your material might have high residual stresses which disturb your EBSD-Measurement, you should perform a residual stress analysis via XRD. This works pretty fine on PVD coatings in the "as deposited" condition. Then you will see, if the residual stress might cause your problem.
I am also curious what kind of material you are using, because mechanical polishing, even if performed very carefully, will cause very small plastic deformations on your surface, if the material is rather soft. Of course, if you do not polish in any way, you cannot get any EBSD information at all. We had the problem with some Aluminium material. We solved the problem using Ion-Polishing. We've got an Ion-Beam polisher from Jeol, which can polish small areas of 1.5 x 1 mm without applying any kind of mechanical force to the material. The method is quite similar to Focused Ion Beam Milling, but less expensive and complicated. The machines are called Cross-Section-Polishers and are available from Jeol, Gatan and some other companies. I am sure you can find an institute near your university, which has got one. If you polish your sample this way and still get no clear EBSD patterns, you can be sure the material is amorphous or nanocrystaline in these areas or has extremely high residual stresses (which you should test first by XRD).
I hope this will help you solve the problem. If you cannot find an institute where you can use an ion-polisher, please contact me again. Sometimes we are also doing wagework polishing if someone has a difficult problem.
Best regards
I suppose there is definitely a layer. We have more substrates (made simultaneously) and thanks to EDS analysis it is sure that there is a layer (it is even sure that more than 1 micron). On other substrates the situation is similar, however the grain size is much different and therefore the result on this specime is the most representative. I suppose this is more likely problem of layer growth than EBSD. Ijust wonder what could be the reason. I will ask my colleague about the XRD results more precisely. Unfortunately I¨m leaving today for 2 weeks so there will be probably no further results from me during that time:-(.
XRD makes no sense since your grains are much too big. I guess that the pattern quality map displays the big grains from the substrate but also from the layer, right? If the grains of the layer (which should be actually displayed by BC) are clearly smaller, then I would agree with some of the comments made above.
Again, if you are sure that there is a layer on the top (by EDS) then it does not mean that EBSD reflects the same information (since only from 20nm) whereas your EDS gives you the information from a few microns, i.e. 100 times bigger depth. Oxidation might be a reason since O attacks crystals orientations differently- This would explain, why do you see several grains in an acceptable quality, and starting from the boundary you do not see other grains. It also seems to be very unlikely (not impossible) that exactly from there a nanocrystalline microstructure appears.
Also the message that crystal of 300-400nm size are the lower limit of EBSD measurements are far away from beeing true. Of course it depends on the material, but Ti is not Si, and even there the lower limit is smaller. Otherwise you couldn't resolve martensitic transformation. For illustration I simply add a snapshot of a Ti-map (Ti-AL) with "fcc" /red), hcp (blue) and bcc (yellow) and TiB (green) and the respective IPF-X map scanned with 80nm at 20kV (sorry for the inaccurate snapshots, but both maps shows the same microstructure). You can clearly see grains smaller that 300-400 nm. We did not used ion polishing, simple standard metallographic preparation with silica as surface finishing. And...these are the original data, i.e. no data cleaning has been applied.
Also please consider, that the question is not, why the EBSD system does not index. This is another question! The shown BC map in one of Jaroslavs contribution shows that there are no bands visible, i.e. the diffraction signal is missing so that no software can index.
Also consider, that even highly deformed materials like martensite can be still indexed as long as bands are visible. Hough transform is perhaps not the best way for high accurate orientation determination (required for GND analysis) but for orientation determination it is sufficient since a quite big range of acceptance (deviations) are implemented in the software which still enables to index. There it might happen that a software gets problems if the patterns are too good since the clear seperation of bands becomes more and more difficult (too many peaks in Hough transform).
Dear Yang, have you seen the size of grains (if the BC map expresses a unique orienitation within the "grain")? Using XRD you will perhaps only see 10 grains (during rotation perhaps 100). It would be very surprising that you would see then any peak. What do you would conclude from such a measurement? From my point of view I would not waste time with XRD, except of Laue technique :-).
But the comment regaring optical microscopy is very useful. This we do always! I can only recommend this for any application, even if the grains are perhaps nanocrystalline. Devices are commonly macroscopic so that any failure must not automatically be related to properties which are detectable only in nano scale.
Hi Gert,
Nice EBSD images. Well resolved. Where's the scale?
I've talked about my direct experience with an iron BCC alloy and a pure FCC copper, both ball milled and then sintered. The grain size was checked with XRD from lab scale, synchrotron radiation from Grenoble's beam line and with HR-TEM and it was 25 nm for the copper and 43 nm for the iron alloy. The indexing was below 50%, in some cases below 30%. The Kikuchi lines where just not visible enough for the instrument's software to consider them. The samples for EBSD were polished and prepared in Hobro, at the company HKL that produced and produces the sensors. The company had been just acquired by Oxford Instruments.
Also, there is some quite cited and well known referenced literature on the resolution problems of EBSD for small grains (see "Characterisation of fine-scale microstructures by electron backscatter diffraction (EBSD)" by F.J. Humphreys, Scripta Materialia, 2004, 8 pp 771-776, http://dx.doi.org/10.1016/j.scriptamat.2004.05.016 ). I do have to say that this, the publication and the analysis, is "dated", so I hope there has been progress. Do you have any publications on 20-50-100-200 nm grains resolved with EBSD? I am interested and haven't found neither them nor someone that could resolve the microstructures I had produced.
Hi Allesandro,
Oxford bought HKL 8 years ago :-).
The paper from Humphreys is 13 years old. You certainly heard something about TKD. There they often compare EBSD (a bit conservative since the main interest is to show the big advantage of TKD) and the resolution is not that bad. You can also read old papers from D. Dingley (from 2000 or slightly later) where the mapped Pt with step widths of 2-5nm using EBSD. But also HKL showed many (about 10) years ago maps of Cu interconnects where they applied 20nm step size with EBSD. The easiest way is certainly to look at the websites of Oxford and EDAX to find some official examples. The useful step width is certainly dependent on the material. Therefore I selected the mentioned Ti. For steel we can go also to 50nm step size (at 18mm WD which is definitely a bad setup for high res measurements.) The software packages even compensate the the worse physical resolution since they interprete the averaged pattern, so that the effective resolution seems to be better than the physical. But this is always the same, EDS signal from deeper regions, XRD (overlapping of peaks) etc.
Anyway, what I only wanted to say that EBSD has resolution problems is you have tiny grains, and yes...you need to adapt the measurement routines (high magnification in order to give the SEM a chance for a correct positioning (small disadvantage of the digitized world, and the used amplifiers for the scan generator), a perfect focusing and a stable SEM etc. But it is not that bad as you mentioned... I simply wanted prevent that people read this and conclude that they cannot measure smaller grains than 300nm.
The images I simply captured by a clipping program but if you are interested I can send you the original images with the scale marker :-). I only want show that you cannot only scan with these step sizes, you also get reliable information from your EBSD patterns. Not perfect, but in an adequate quality. :-)
Dear Y. Yang, also iron meteorites are often polycrystalline although their grain size can be in the scale of meters. On the other hand, nowadays the structure of phases are performed on single crystals with a dimension down to 100nm (synchrotron even smaller), i.e. much smaller than the crystals here. It is all relative and we have always to keep in mind how the used technique works. Let us assume you don't see any peak. This can be caused by either an x-ray amorphous material, no layer, a single crystal... And especially the last one is the problem here since if you only illuminate 3 or 10 grains it is very unlikely that you will detect any intensity since the diffracting plane for this respective grain must be accidentally exactly (!) parallel to the sample surface (Bragg-Brentano geometry). Do you know, how likely this is? Therefore, powder XRD assumes several 10.000 grains with a preferably homogeneous orientation distribution within the illuminated area in order to make sure to get a statistically relevant signal. Then it is really better to use Laue technique since this works always and gives only for x-ray amorphous (or no layer) no signal.
Well Gert, I've never heard of a positive correlation between age and quality of a good scientific paper (cited 281 times by the way). Humphrey's observations, discussed in more detail in his book, are still valid today. The main problem is the size of the bubble where electrons diffuse. As you probably know when electrons hit the sample they spread by scattering. Back scattered electrons are from that cloud, bubble, whatever you want to call it. This is why with materials that have a higher atomic weight the resolution is higher and you can observe finer microstructures than in materials with a lower atomic weight. So the problem is not step size, it's really the grain size. You can have very small steps but not see Kikuchi bands due to the fact that the signal comes from a bubble large as 2-5-10 grains!
So, again, please point to publications where nanocrystalline bulk materials are resolved with EBSD with an indexing superior to 50% (possibly 70-80%).
And yes, clearly if you have the possibility to do transmission diffraction (the TDK you mentioned) you will increase the resolution and will be able to focus and see a decent signal even with small grains. But it's not the same thing as using a conventional SEM / FEG SEM. You need to prepare a thin sample, similar to what you use for a TEM.
Ah, Jaroslav's original question was on EBSD by the way, not TDK. Different hardware = different results ;-)
You forget a very serious problem. If electrons interact with the material they loose energy and do not contribute anymore to the Kikuchi patterns visible on the screen. They exist (interfering with electrons of a comparable energy) but the major signal comes from electrons of slightly less energy as the initial. This is one reason why (the visible part of) Kikuchi patterns are comparatively weak on a big background. On the other hand it means that only electrons very close to the surface can fulfill this condition (very low energy loss) and this again tells you that the bubble you mentioned is not that important as you think. It is existing, but the effective size forming the extractable Kikuchi signal is clearly smaller.
As I mentioned above, the map of Ti collected 3 years ago was measured at 20keV electrons, with 80nm step size and a hit rate which is higher than 90% (again: these are the original data, i.e. no data cleaning applied). It is the microstructure produced after spark erosion, therefore the different phases and the curved surface. The irregular shape was also the reason why I captured only a part of the map and therefore the scale bar is missing. Moreover, the literature gives enough examples where microstructures have been scanned in comparable and higher resolution. Although from my point of view not very high resolving, I attach a link to a recently published paper where framboidal pyrite has been scanned with 60nm step width. Comparable measurements have been done by the cited Liverpool group ten years ago as well at the same sample...and this sample was even C-coated! FeS2 (Z=58/3) should be comparable to Ti (Z=22) and no exclusive material. As you can see, we even can recognize without any problem a non-existing 4-fold symmetry for the cubic phase on standard patterns (again: no data cleaning applied). If you need more examples...this is no serious problem. Minerals, meteorites, metals and alloy...whatever you want :-). But then we should find another platform :-).
I don't think that I am specialized on high-resolution maps, already because the location of our SEM is not the best to preform them, but my memory tells me that I've read enough papers to make a conclusion that the spatial resolution of EBSD is better than your opinion but probably worse that sometimes claimed, especially if one compares physical and effective resolution. Finally, there are several serious factors which are important like the focusing, astigmatism, any kind of charging, beam stability, the beam diameter (we are often using the maximum current, i.e. a beam diameter which is far away from the high-resolving conditions) etc. However, with a little bit patience in setting up a FEG-SEM (I installed EBSD systems at SEMs for practically every SEM supplier) you can get this resolution practically everywhere.