I need to define the amount of ferrite in austenitic stainless steel. Which technique is better: EBSD or XRD?
Dear Oskana,
in principle both techniques are powerful enough:
For a more statistical evaluation I would suggest XRD, but to quantify the phase fractions you should do Rietveld refinement. Furthermore, fractions of less than 5 vol% are hardly to detect by XRD as well as small coherent volumes. Please keep in mind that you should avoid using copper K_alpha radiadion, due to fluorescence of iron.
On the other hand you can perform EBSD to vizualize the phase distribution in the microstructure and to estimate the fraction. Here you have to considder, that you only analyze a small area of your sample, that might not be representative.
I wolud suggest to perform both, if possible and then to correlate your results.
An alternative is to do chemical etching to contrast both phases and to do light microcopy and subsequent image analysis. One of my colleagues did this metallographic approach pretty well on Duplex steels. The specimens were etched electrolytically in a 20% solution of sodium hydroxide for 20 s at 6 V in order to achieve contrast between austenite and ferrite.
Cheers
Martin
http://www.hanser-elibrary.com/doi/pdf/10.3139/147.110216
The XRD technique is generally a qualitatively phase analyses. In some cases which you have so many samples having same phases you could do quantitative phases analyzing by XRD.
I think you should use a simple way to measure the amount of phases. Image analyzing by some software such as ImageJ is suggested by me.
Or also you should use ferritoscope to measure the amount of ferrite in your steel.
Regards
EBSD is more visualized! Moreover, more precise depending on the software what you selected. Many related references you can find from Sciencedirect.com
If you want a reliable quantitative phase analysis you should use XRD. If you have many phases, then the Rietveld-Method should be used for data evaluation.
If you want to analysis ferrite phase in steel then use ebsd technique.
Dear Klok,
It depends on the amounts of the phases and the homogeneity of your sample. Generically speaking, the XRD is more reliable for the phase quantification when the amount is more than approximately 5 wt.%. EBSD is powerful for the quantification of the minor phases in the homogeneous sample. But you should keep in mind that, the EBSD method needs stereological assumptions.
Researchers have given several answers to your question. Just to reiterate, XRD analysis will do the job. It will reveal all the crystalline and amorphous domains in your sample.
Ferrite forms at very less percentage in austenitic stainless steel. In XRD you can scan less area and you cannot be sure is complete area is analysed. In EBSD it scans an area where ferrite formation in between the austenite grains can be analysed easily. Now if you use XRD X Ray should fall on the ferrite grain which is minutely formed between grains, it would hard to detect via XRD.
In addition to XRD, you might consider neutron diffraction to characterize the phases if you want to sample large volumes.
Dear Oskana,
in principle both techniques are powerful enough:
For a more statistical evaluation I would suggest XRD, but to quantify the phase fractions you should do Rietveld refinement. Furthermore, fractions of less than 5 vol% are hardly to detect by XRD as well as small coherent volumes. Please keep in mind that you should avoid using copper K_alpha radiadion, due to fluorescence of iron.
On the other hand you can perform EBSD to vizualize the phase distribution in the microstructure and to estimate the fraction. Here you have to considder, that you only analyze a small area of your sample, that might not be representative.
I wolud suggest to perform both, if possible and then to correlate your results.
An alternative is to do chemical etching to contrast both phases and to do light microcopy and subsequent image analysis. One of my colleagues did this metallographic approach pretty well on Duplex steels. The specimens were etched electrolytically in a 20% solution of sodium hydroxide for 20 s at 6 V in order to achieve contrast between austenite and ferrite.
Cheers
Martin
http://www.hanser-elibrary.com/doi/pdf/10.3139/147.110216
EBSD technich is better ,because XRD technich for your work is not nessasery , that it is hard and not careful.
I agree with Wei-Ying Chen - for XRD you need a powder or you will be performing XRD on a single crystal and will only get one peak for your matrix phase. I think if you crunch your sample into a powder it would be good for identifying which phases are present.
EBSD, on the other hand, is more for visualizing crystallographic orientation of different phases or matrix grains. You certainly can use it for quantifying phase fraction, but its only a 2D image.
Dear Dr.Oskana,
Dr.Seyring is correct. He answered in great detail.
Best regards
Koichi Kato
Dear Colleagues!
Oksana asks about definition of quantity of a phase. Therefore more exact answer is given by Martin Seiring.
based on my experience for phase quantitation in AISI301, each technique has it advantages and drawbacks- XRD can investigate a significant amount of the volume but is very sensitive to texture and specific orientation; to use it, you need to include this parameter in your Rietweld refinement and/or deploy texture "removal process" (GADDS system from Brüker for example)
EBSD is a powerfull technique but a local one - sample preparation can be tedious and local orientatton and misorientation estimation might lead to errors in phase fraction estimation - it is also time consumming and require a very stable electron beam since you will have to cover a large enough part of your cross section to get statistical sounded results
I think that you should consider first optimised metallography on different samples or , if you can take advantage of the fact that ferrite is ferromagnetic and the rest of your structure might mainly contain austenite and carbides, that are amagnetic. Thus, if you use a ferritescope or any other magnetic measurement system (magnetic permeability ....), you should be able to quantify the amount ferrite you have in your sample
I guess it totally depends on what you need.
In EBSD, you can determine the phases in the localized scanned area. The distribution of phases can be identified in the localized scan area, but this depends on a lot of factors. Firstly, it depends on your step site and the size of the phases. For example you have one phase with one dimension like 2um, and you use step size of 3um; the results are not trustworthy at all. Meanwhile for a large area with relatively small step size, the time you spent simply on the scanning takes several hours, even though the EBSD technique nowadays can make it relatively fast. Secondary, the sample preparation is a bit tricky. Depends on what kind of material you have, generally it takes efforts to prepare the samples good enough for EBSD. However, to use EBSD can provide you the local phase distribution, which cannot be provided from XRD.
For the XRD technique, it could generally detect a large volume and provide you with the diffraction. The results from XRD are in 'macro-level', while the results from EBSD are more 'micro-level'.
Hope this could help.
Actually they can be used in a complementary manner: determine the existing phases by XRD first and introduce these phases to EBSD software to see their distribution and morphology quite accurately. This is the way I usually use for those materials that I don't know initially. But for a well known material like steel which you know the constituents there is no need for XRD before EBSD.
The other point is your purpose of test: do you just wanna know the quantity of the phases or you need also to know their location, sizes and morphology? For the first aim XRD is enough and more accurate thanks to its volume of test which is statistically higher; provided that the quantity of phase to be determined be more than 5% (maybe in new machines, this limit is lower).
Dear Oksana Klok
I think the two techniques are suitable for define the amount of ferrite in austenitic stainless steel but the XRD technique is more suitable.
Practically, all aspects are already discussed from many different point of views. From my opinion, there is no easy answer. Both techniques display different things. Actually, XRD is made for quantitative analysis, and not EBSD. However, also XRD has very specific requirements which have been pointed out and which are typically not always fulfilled by steel, i.e. grain size, grain size distribution, texture, minimal phase fraction, fluorescence, sample alignment of a bulk material within a powder-dffractometer. Some effects of them can be reduced but not really totally excluded. This means, XRD has to be very carefully applied as well in order to get correct results. And this already points out another question which is not in focus here: How accurate is "accurate" (not mention the precision)?
EBSD is a "digital" technique and not an "analogue" like XRD, i.e. because of the step size you perhaps exclude already grains which are lower in dimension, or forming a non-interpretable pattern. The statistic is quite problematic since often you cannot extrapolate a small field to a large area requited for a reliable phase description. The preparation can remove selectively phases or produces topography so that along grain boundaries precipitates a no more visible under 70 degree. The preparation commonly does not work for both phases in the same quality. And I certainly forgot a few other details which speak again EBSD as technique for quantitative phase fraction.
Summarizing, I would always prefer XRD (full-pattern analysis, often mistakenly called Rietveld analysis which is actually a technique for structure analysis) as more reliable technique if the major requirement of the techniques matches to the microstructure of your steel, i.e. grain sizes smaller than 1µm, a low texture (e.g. no welded material), phase fractions bigger than 10% (i.e. clearly bigger than the detection limit of XRD). Bigger grain sizes are not better processable with EBSD but only with neutrons or synchrotron as mentioned above as well). Scanning small areas as done with EBSD is statistically even less reliable, although it tells us, WHERE which phase and in which environment it appears. But also there the grain size is limited because of pattern overlapping. Precipitates along grain boundaries are less reliable or not at all detected, and the practically always applied "data cleaning" artificially removes them or replace them by the matrix phase. My very personal point of view is, that the precision of phase determination by EBSD is usually not better than +-10%, and depends on many, very hardly handable errors, systematic and non-systematic.
Very good alternatives are also mentioned already: magnetic measurements. They seem to me comparable to neutron measurements.
XRD is suitable for phase detection for relatively large specimen (considerable amount of powder or particles or even bulk material). However the EBSD, Electron Backscatter Diffraction Analysis, is used to perform quantitative microstructure analysis in the Scanning Electron Microscope (SEM), on a millimetre to a nanometre scale. it require a special surface preparation. I believe the XRD is more suitable for such task.
@ Martin Seyring: Could you tell us, why your colleagues applied chemical etching and did not perform EBSD or XRD? Or did they? What is the minimal grain size, how imaging artefacts along phase boundaries affect the result? And, how did they know the correct phase fractions?
Thanks a lot for your valuable contribution!
Both techniques are reliable, and one has to do both techniques for comparative study, although both techniques provide different piece of information. For a material scientist both techniques are important as they provide insight about the material - phases, morphology, grains, orientation. I like to vote both and difficult to point out one is better than other.
@ R. Krishna: How both techniques can be reliable when they provide different piece of information? She is only interested in one piece of information: the phase fraction.
How do you define "reliable" ? A software always delivers a number as long as there is no division by zero. How reliable this number is, strongly depends on the raw data interpretation, e.g. smoothing of data, experimental setup, or whether the applied theory in the software matches "somehow" the presently investigated material. XRD is made for powder in order to make sure that statistically enough crystals of ideal size and shape match the assumptions made in the coded theory. Here we have a steel which certainly does not fulfill this conditions so that there is a big question-mark behind the reliability of the given number. And I am not yet talking about the parameters used during refinement, or which software approach has been used. And the reliability of EBSD is not better but statistically more worse. The precision of the result might (!) be very good if all patterns have been solved without "tricks" (data cleaning). But the result itself can be far away from the correct one...and very different to the one collected by XRD (relatively).
Dear Oksana,
First of all I must notice that I do not have any experience to deal with steel but in our group I learned something useful with could be useful to characterize steel's structure such as different phases and so on. Why don't you use Ferritoscope as a method to identify the magnetic phase (which is ferrite in your case). However I know that XRD is the very definite method to use in your case. You can comprise the XRD data with ferritoscope result. Hope you have successful in your researches.
Amirali
First you must be careful about percent of this phase in the steel as in XRD Phases with wt% lower than 3% can not be detected. On the other hand XRD can detect the phase between than XRD. But if you need phase map EBSD is better
Definitely XRD because it gives the information of the volume of material (the powder).
Dear Marcin, "Definitely" is a very strong statement and from my point of view only then correct if all conditions for an exact XRD-interpretation are fulfilled. This is NOT the case here, so that I would recommend to read first ALL former contributions.
X-ray diffraction is a powerful technique and it can be utilized to find out phases in a sample. Since it's discovery, X-ray diffraction has remained a starting point to find out the composition and crystal structure of metallic and non metallic samples. Atomic arrangement, type of atoms can be found in the unit cells of simple and complicated molecules. Bond lengths can be found. Phases can be detected and their amounts can be calculated. Nearly all aspects of science of materials can be analysed to have an idea about further studies. So in my opinion, this technique is useful and strong.
If you have the EBSD available with you it can confirm the finding of X-ray diffraction with more precision.
@ Tahir: Do you really think, EBSD is more precise? Or more accurate than XRD? And if yes, regarding what? Phase identification, phase fraction, lattice parameters...?
for quantification analisie both xrd and ebsd can be used , but i prefere ebsd , because results of xrd and convert to quantification analisise sometimes is hard and with low careful.
Both XRD and EBSD will work fine together, but you can also try the ASTM A800 method (ferritescope test) if you can not destroy the sample.
I am of the opinion that under conditions two methods give the same result.
Hi
The proposal and the detailed explanations supplied by Kai Zhang are the most suitable : EBSD provide localised informations and is extremely time-consuming process compared to XRD technique that enables a large depiction
The condition "...phase crystallographical orientations are randomly..." is extremely important. However, in ALL the materials in engineering practice this condition is not realized completely. All the engineering materials are anisotropic (sometimes even glasses!); if you will tell me about your "isotropic" samples it means for me that only your measurements have not been done perfectly and thoroughly...
The condition "...phase crystallographical orientations are randomly..." is extremely important. However, in ALL the materials in engineering practice this condition is not realized completely. All the engineering materials are anisotropic (sometimes even glasses!); if you will tell me about your "isotropic" samples it means for me that only your measurements have not been done perfectly and thoroughly...
The condition "...phase crystallographical orientations are randomly..." is extremely important. However, in ALL the materials in engineering practice this condition is not realized completely. All the engineering materials are anisotropic (sometimes even glasses!); if you will tell me about your "isotropic" samples it means for me that only your measurements have not been done perfectly and thoroughly...
In my opinion...EBSD provides localized information...........analysis volume of XRD in more as compared to EBSD.............both are necessary.....
It is trivial to expect that my answer to this question must have been mentioned by some other users. Anyway, as some colleagues mentioned about a threshold limit of about 5%, which in my opinion is an underestimation to the potential of XRD method. If you have good equipment like: a rotating anode generator, a good detector and can eliminate/minimize fluorescence then you can easily go down to even 0.5% (by volume)!
While EBSD is also a very powerful method with better visualization to the reader but as downside, it is essentially restricted to low deformation state of the specimen. If you are willing to study deformed steel, I am doubtful if it can quantify a high strained steel microstructure. I am not expert in EBSD but the hkl software that I used regularly fails to index the high strain specimens. So, for a low strain specimen EBSD is as good as XRD.
Just a quick warning on using the EBSD technique. The EBSD technique only allows you to determine the surface fraction, not the volume fraction. So if you expect an on average non-symmetrical and/or non-homogeneous morphology distribution of the ferrite particles (e.g. elongated in a certain direction which is quite common in rolling textures), the fraction determined by EBSD could be quite far off. So preferably measure the fraction on different sample planes.
Hi Oksana, please realize that neither techniques is ideal to do what you want, XRD and EBSD will give you a rough estimate of the true volume fraction. It is usually accepted that XRD can only identify phases with more than 5% volume fraction (so you can assume this is the accuracy of your measurement) and EBSD is, essentially, identical to the ones in which the area of the phase is counted in a standard frame (and this assuming your software can unambiguously detecct ferrite and differentiate it from austenite and martensite, which is difficult). Hilliard and Cahn have an old work published in Trans ASM (If I remember right, I cannot find the reference now) in which they proved the most efficient method is using the grid of points, in a magnification such that no two adjacent points fall inside the same particle (this is important to assure statistical independence of the measurements) . This is incorporated in the ASTM standard (E562-11) of volume fraction measumement. You may use EBDS, but be sure to measure many unrelated fields. For ferrite/martensite there is a magnetic technique called "ferritometer" which measures the amount of ferromagnetic phase, but this in imprecise for obvious reasons.
Most mentioned techniques become almost equally poor /for apparent though different reasons/ when applied to low (say
However the point is in what you view. If the phase of interest is rare and its particle is small, you will skip over the latter when scanning with usual scan steps...
@ C. Schön: Only a short comment to the deformation of a phase: EBSD can be do this as well if you consider that you need in this case a higher resolution since otherwise the positioning of the beam because of digitization is very bad, at least worse than the subgrains caused by plastic deformation. This means you get an superimposition of several grains of nearly the same orientation which causes blurring. Check it by martensite, which has small particles with totally different orientations. At lower magnification you get a remarkably worse hit rate but if you scan the same microstructure at higher magnification the hit rate become MUCH better. Moreover, the hkl indexing had many problems with indexing of overlaid patterns. Therefore, along grain boundaries practiaclly 30% of all patterns could not be indexed. This is history, since also hkl is now working with more than 5 or 7 bands. Takes more time but you have a much better hit rate. This has been introducesd by Bruker (detect standardly 12 bands and use as many as possible bands for indexing).
Of course: ferrite and martensite are hard to differentiate, but ferrite and austenite are definitely not, i.e. this is standard nowadays.
A major problem with EBSD are the zero solution, and how to deal with this. I don't want conceal that also XRD has a comparable problem, if the background is badly processed (and this can be especially in case of small fractions a big challenge, despite full pattern refinement (often also mistakenly called Rietveld refinement)). How big or small the detection limit is depends on the phases. For comparably scattering phases like ferrite and austenite and "low defect density I would not acpect a better detection limit than 2% (volume or mass since both are comparable). Defect-rich phases like martensite form very wide peaks so that they are badly detectably so that the detection limit is more in the scale of 5...10%. For EBSD the simple and physically absolutely inacceptable cleaning procedures are especially for this application horrible but common so that beside misindexing also the post-processing affect the final error. This means, if the hit rate is only 90% you have a quite big error also in your phase map. It makes no sense to split the zero solution half/half to both phases or simply ignore the zero solutions, especially when the grain sizes are different or both phases have a different pattern quality (because of deformation, different preparation behaviour etc.)
Another comment regarding "surface": In fact also XRD is a surface technique if you consider grain sizes in the scale of some micrometers (not untypically for steel). Then also X-rays do not consider more than the first layer of grains, i.e. exactly the same as EBSD. Neutrons are a better alternative, but....who seriously has access to a neutron reactor? There the chance is bigger to get time at a synchrotron which you can also use btw...
I have access.. . Of course it is not an everyday routine instrument, however it is an indispensable means to calibrate any alternative technique. More so that research reactors work by themselves regardless of our particular issue and, hence, such measurements will not suggest much extra expenses. In a sense synchrotron facilities would be more consumptive though less efficient.
Depends what phases you have. If martensite is also included, XRD is a bit challenging, but EBSD could provide a bit better solution (to differentiate between martensite and ferrite). But on the other hand, EBSD is more "local" compare to XRD (how big are your grains?)
@ Ehsan: I know that some people use EBSD for the separation of ferrite and martensite, but my opinion is: what you cannot distinguish by XRD you cannot by EBSD. What you are doing is, you are making some assumptions which show some correlation to the pattern quality or the band slope. However, from my point of view there are several phenomena which can generate these secondary properties as well (since they are not martensite-specific) so that this assumption is very risky. Especially when you compare this procedure with well established standard techniques like full-pattern refinement used in powder XRD.
@ Gert: Yes it is based on assumptions that are sort of supported by theories. Martensite formation is associated with high density of dislocations, which is not the case for ferrite formation. XRD has a lot of capibilities but very difficult to distinguish these two phases, even if they have different localized stored strain values (if the sample has been prepared properly).
But as you know better than me, EBSD could (qualitatively) compare the amount of stored strain energy in these two phases and show it with relatively significant difference in IQ of the two phases. I have done number of tests years back on different types of steels and the result was 100% true identification of the martensite phase using EBSD technique which was supported by microhardness, and etching/microscopy. XRD, however, did not show such a capability easily (at least in my experience).
I would be glad to see your evidences against this.
Dear Ehsan, I am not very sure whether martensite really contains much more dislocations than ferrite, however, martensite is a defect structure which (from my point of view) crystallographically expresses that the assumption of a periodic lattice does not match to martensite. This means, that carbon-containing martensite should always show broadened peaks in XRD and blurred pattern in EBSD whereas ferrite can generate a perfect lattice, perfect peaks and perfect patterns. Regarding the phase determination: actually martensite is no phase! It is a microstructure description, but I do not know any database which contains the phase martensite. :-)
If we discuss the "phase" fraction using image quality (IQ)...since when IQ is phase specific? Together with Matt Novell (EDAX) and Scott Sitzman (Oxford Instr.) I gave a talk about 5 years ago at M&M demonstrating which factors affecting IQ, or pattern quality or and contrast. We know that people see the remarkably black areas in contrast to the ferrite, but the final histograms do not reflect separate peaks (and how to divide the overlapping?), and bad IQ can be also generated by boundaries, small precipitates, local deformation, different phases. This means, only in case you can absolutely exclude the other effects, and your post-processing does not falsify the boundary effects, your procedure is acceptable. For very specific model steels this is perhaps given, mainly in order to demonstrate the successful application, however, for commercial steel I havn't heard that one of the steel producers is using it. They are definitely interested in and mainly sponsored these PhD works in different countries, but as far as I know (and also believe), it is too risky and the data produced are not trustful. EBSD is really not the technique to quantify martensite reliably, especially when the fraction is below 10%. But this is certainly my personal opinion, but remember the first sentences about "phase" and non-preventable lattice distortions.
Hello, I am facing the same question. I am characterizing a new high strength steel with 200 nm grain size. Which one do you recommend? XRD, EBSD or TEM?
I would always start with lower magnification to get a nice overview. TEM is a nice technique if you know what is inside or you have certain information that some minor phases exist which are below the limit of XRD. If you want count the flowers on a meadow you also don't run around with a magnification glass.
200nm is perfect for XRD and the statistic worth is unbeatable.