I have a question on low carbon (e.g., 0.1, 0.2, 0.4 wt% C) steel standards as utilized by some people as a multi-standard calibration curve.
I know that carbon in steel is meta stable, but at room temperature does the carbide segregation continue to occur on the micro or nano scales?
I once heard from someone that for homogeneity one needs to re-manufacture ones low carbon steel standards every 10 years or so even
when stored at room temperature. Does this make any sense at all to anyone?
john
Sorry John, I am not able to address this, but I have a related question.. I have a facility user who is now working on nitriding and carburizing of steel. There ends up being iron nitride phases, and they want to know something of the possible carbon content of these phases. As we are not really setup (at least yet) to do these ultra-light elements well, she would like to know if anyone out there might take on this sort of tricky analysis? So, desperately seeking some expertise in C and N in steel! I will be happy to pass along contact information.
Hi Mike,
I got a response to my question from Philippe Pinard on the room temperature stability of these carbides in steel, and he says yes, it is an issue. Hopefully he will post his response here for all to see.
On your question, Philippe is definitely the expert, but he is at Oxford Instruments now. I have a little experience with this sort of thing and as long as she doesn't require ultra high spatial resolution (we just have a W gun), I would be pleased to give it a try in our lab.
Some years ago I did some trace carbon and nitrogen analyses for a customer that I thought turned out very nice. They are here:
http://smf.probesoftware.com/index.php?topic=48.msg270#msg270
Now maybe I've learned a thing or two since then, or maybe not. But today I would probably apply the TDI scanning method to get best results form trace carbon as seen here:
http://smf.probesoftware.com/index.php?topic=933.msg5990#msg5990
Trace nitrogen is actually much easier than one would think.
Thanks John,
I will let her know you are willing to give it a try. That is interesting to know that nitrogen is easier. In this case, nitrogen will be major, with minor carbon. I don't know anything about the dimensions of these phases, but hopefully this is something not too unreasonable.
Quote from: Mike Jercinovic on December 27, 2017, 02:27:06 PM
I will let her know you are willing to give it a try. That is interesting to know that nitrogen is easier. In this case, nitrogen will be major, with minor carbon. I don't know anything about the dimensions of these phases, but hopefully this is something not too unreasonable.
I should add that while trace nitrogen is relatively easy (because it's merely the background determination that is critical for traces), major nitrogen is considerably harder to do accurately, since it depends more on the matrix and peak shape corrections. Haven't we all at one time or another attempted to analyze TiN compositions and found the Ti L absorption edge to be in the worst place possible...? >:(
At least with trace nitrogen analysis we don't deposit nitrogen during the acquisition as we see with carbon. I still want an "in situ" ultra-violet laser in my instrument for surface cleaning during data acquisition:
http://smf.probesoftware.com/index.php?topic=140.msg643#msg643
Even just a small UV LED with focusing optics might work.
The Nitriding of steel is done at a temperature where the steel is Ferritic or in the BCC crystal form. In this phase it is relatively insoluble and eventually reaches the activity where it forms iron nitride.
The iron nitride is often present as a very hard phase at the surface which can be revealed by etching. It is also often evident within the diffusion gradient as needles of iron nitride. These needles should not be confused with Iron carbides which may also be present. The following online ref is OK.
https://vacaero.com/information-resources/metallography-with-george-vander-voort/1138-microstructure-of-nitrided-steels.html
If the temperature is too high, the steel becomes austenitic or FCC and the nitrogen rapidly dissolves into the structure. In terms of a wear scenario this is a problem as wear may generate locally high temperatures which can locally austenitize the surface (form the FCC structure) and the coating diffuses away.
Note: if you examine the Fe-C phase diagram, you will see that C solubility in ferrite is very very low at nitriding temps. What this means is that the C will be tied up in carbides or in pearlite. Unless there is a free energy driving force for the carbides to dissolve and enter the iron nitride, I would imagine that iron nitride would be relatively devoid of carbon. The increase in nitrogen content in the steel below any nitride layer is also likely to displace the carbon and it would be 'pushed into the coating (see formation of carbides on grain boundaries in above ref). This may be a load of rubbish if the nitriding process is also carburising at the same time.
Note2: Do not do the analyses on an Nital etched sample as the nitric acid etch products swamp any Nitrogen in solution. Been there done that.... The question was "is it nitride?". Analysis on an etched sampled said "yes it was", analysis on a polished sample said "no it wasn't".
Note3. The microstructures in the ref are mostly martensitic in which the C is frozen in a BCT distorted transformation from the high temp FCC form. This adds another layer of complexity as the C is relatively homogeneous instead of essentially zero in ferrite and 0.8 in pearlite.
Lotsa luck and be careful.
Regarding the wandering of standards for microanalysis over time....
Hi John,
Carbon "should" be quite stable at room temperature.
Do you know if the standards are martensitic or ferrite/pearlite structures?
Theoretically if they are as-quenched or fully hard martensitic structures, the carbon could diffuse and form Epsilon carbide which are just localised clusters that require atom probes to see.
This typically forms at temperatures of 100-200 C over an hour or so but I imagine could perhaps be detectable with an extended time at room temp.
Wikipedia suggests that Epsilon carbides will form 100 C to 200 C and decompose above that (this meets my understanding)
https://en.wikipedia.org/wiki/Cementite
This might give some useful background:
https://en.wikipedia.org/wiki/Martensite
Good Old Harry Bhadeshia has something to say here:
https://www.phase-trans.msm.cam.ac.uk/2007/Epsilon/Epsilon.html
The following ref suggests that these Epsilon carbides can form at room temp:
https://www.researchgate.net/publication/248138123_An_interpretation_of_the_carbon_redistribution
I haven't seen inside the article.
What is probably perceived to be going on is a peak shift from the location for C caught in the metastable BCT structures in the interstitial positions to the location (or shape) in Epsilon carbides.
Whether it is really going on, I cannot say.
I'll have a bit of a hunt.
If the structures are ferrite and pearlite, they are probably not good for microanalysis anyway as they are locally variable (0.008 in the ferrite and 0.8 in the pearlite) but should be stable for millions of years until the Fe3C decomposes to graphite.
An interesting page with a heap of steel metallurgy in an interesting context.
http://products.asminternational.org/fach/data/fullDisplay.do%3Fdatabase%3Dfaco%26record%3D1910%26search%3D
This also may give some insight into my comments on the Carbon and Oxygen analysis requests in your other problem.
Cheers
Les
PS. I work in the steel industry and never do (kicking and screaming) Quant C or N in steel.
These elements are best analysed by a bulk technique such as LECO gasses in metal analyser.
Locally, the C distribution is driven by variations in the alloying elements distribution.
I am more concerned with interaction volume, contamination, localised sub micron carbide formation within the steel; all these render the relatively bulk analysis techniques nonsensical.
PPS if you really want a challenge, try finding out if submicron boron entities in steel are boron carbides, boron nitrides or a mixture. Never got that one sorted.
Quote from: Les Moore on December 29, 2017, 06:10:17 AM
PS. I work in the steel industry and never do (kicking and screaming) Quant C or N in steel.
These elements are best analysed by a bulk technique such as LECO gasses in metal analyser.
Locally, the C distribution is driven by variations in the alloying elements distribution.
I am more concerned with interaction volume, contamination, localised sub micron carbide formation within the steel; all these render the relatively bulk analysis techniques nonsensical.
Hi Les,
I find this all very interesting. But I have to proceed with EPMA as some clients seem to require micron spatial resolution of these elements.
Trace nitrogen by EPMA is actually pretty easy. Trace carbon of course challenging due to the carbon contamination issue, but I think our TDI correction takes are of this quite well. The carbon background is the hardest part, but MAN backgrounds plus a blank correction as shown here, seems to work well I think:
http://smf.probesoftware.com/index.php?topic=48.msg270#msg270
For the MAN background on carbon one has to choose standards that don't interfere with the carbon emission line. Interestingly I just had an academic, studying heat treatment of surfaces (in cross section), contact my lab yesterday, asking for trace nitrogen and carbon analysis on steels again on micron scales. So I guess I will give it a go again...
John, Les, Mike,
As far as I know, the calibration steel reference materials are martensitic and martensite is metastable at room temperature. I attached a figure from my thesis, where I performed a line scan on 4 steel reference materials. Besides the overall shape due to contamination (large increase in the first 4um), I think the small spikes in the line scans confirm the carbon segregation and the formation of small carbides. All the quant work in my thesis was done using Fe3C standard, where the Fe3C "islands" are artificially enlarged to several micrometers.
Philippe
Phil - at room temp (and from recollection) its a slow reaction. The temp rise in a conductor must be very low and I suspect that the the cracking of carbon in this case is not due to temperature rise on the surface but some reaction in the vacuum and the presence of a cool surface.
I suspect that a lot of the bean damage on an insulator is not related primarily to temp rise, but related to the dielectric strength of the material and the imposed electric field. This would explain, for instance, the very different behaviour of quartz and amorphous silica under the beam.
The figure that Philippe attached in the previous post, is very similar to what I saw in my own efforts:
http://smf.probesoftware.com/index.php?topic=48.msg339#msg339
The idea he and Sylvia came up with some time ago was to acquire the sample and also the standard for the (MAN) background calibration, using the exact same acquisition parameters. So if you scan 10 seconds per pixel on your sample, also acquire 10 seconds per pixel for the background standard. This is the basis of the correlated pixel quantification method (CPQ) in CalcImage. The idea being that carbon contamination might normalize out between the sample and the background standard.
The main issue I saw with this method was that the carbon contamination reproducibility in the beginning of the CPQ scan isn't as consistent from sample to background standard) as we hoped for. However, Phillipe soon realized that if they started the scan acquisition well before the area of interest, it seems the carbon contamination eventually "settles down" and is much more reproducible after the initial pixels.
Basically there are several possible methods for quant of carbon. I am not sure which is the best. Philippe's method is a single line scan that relies on starting the scan far enough before the area of interest so that the carbon contamination from the previous pixels is allowed to "equilibrate" to a stable level on the subsequent pixels. For example, a single scan that utilizes 5 seconds per pixel will deposit carbon on each pixel, and as the scan proceeds, the subsequent pixels will "encroach" on the carbon contamination from the previous pixels. The difficulty then is quantifying the carbon when there is carbon being added by beam contamination from the previous pixels.
The other method, is to use what I call TDI scanning, where multiple scans over the same area are performed, but using a much shorter pixel dwell time as seen here:
http://smf.probesoftware.com/index.php?topic=933.msg5990#msg5990
For example, 10 separate 500 msec per pixel scans on the same area (for a total of 5 per pixel dwell time). The idea being that the carbon contamination is deposited in a more even and controlled manner by using multiple fast scans, and because we acquired multiple scans, we can extrapolate the carbon signal back to zero time on a pixel by pixel basis. Aside from the tests linked to above, I haven't performed enough analyses to fully test this scanning TDI method for trace carbon but it looks very promising.
I will be using this method on the heat treated samples I get next week and will post results here. On the issue of carbide stability, I should also quote Philippe who responded with this addition info:
QuoteYes the carbon segregates at room temperature. Very slowly but over several years it is measurable.
So the average carbon concentration doesn't change, but it does continue to segregate over time apparently. Does any one have measurements showing how this carbon segregation progresses and at what size/time scales? Also what effect does this carbon segregation over time have on the mechanical properties of the carbon steel? I'm sure someone has looked into this... on a TEM perhaps?
Les
Thanks for your post - we've done a lot of ATP analyses on 'odd' high pressure irons and have been trying to correlate these to chemical maps derived from the FEG EPMA. Its interesting - these things have small oxide blebs in them, on the order of around 100nm's which are mappable. Whats interesting is that they appear to be correlated to C content on the probe. Except the ATP says they aren't. There is a small, atoms thick, 'rim' of carbon around the oxide blobs, but the signal from the probe is significantly bigger.
It transpires that C deposition from the FEG is significant (even with LN2 / Cryocooler) and this together with the change in substrate (from iron to oxide) causes C-estimates derived from the probe to be excessive.
Les - do you know of any good ATP papers that have looked at this Martensite phase transformation as this seems to be, analytically at least, analogous?
Im also stumped as to why our probe is so filthy with Carbon. Its dry scroll pumped, grubby finger free, but the C build up rate makes even quick maps impossible without LN2 and even this ain't great. Its almost like its bleeding P10 into the column.....any clues folks?
Quote from: jon_wade on January 03, 2018, 01:16:53 PM
Im also stumped as to why our probe is so filthy with Carbon. Its dry scroll pumped, grubby finger free, but the C build up rate makes even quick maps impossible without LN2 and even this ain't great. Its almost like its bleeding P10 into the column.....any clues folks?
I looked at carbon contamination a bit since we couldn't afford a turbo pumped system when we bought our SX100 in 2006, but we did specify a 100K cryo pumped baffle which has worked great since then.
The tests I did some time ago, showed that this system wasn't much worse than a normal turbo, dry scroll pump system:
http://smf.probesoftware.com/index.php?topic=140.0
I don't know where the carbon comes from either but it could be from P-10 (another reason to replace flow detectors), native hydrocarbons on the sample and/or vacuum system.
I still like the idea of an "in situ" UV LED laser to clean continuously during acquisition.
Quote from: jon_wade on January 03, 2018, 10:22:47 AM
I suspect that a lot of the bean damage on an insulator is not related primarily to temp rise, but related to the dielectric strength of the material and the imposed electric field. This would explain, for instance, the very different behaviour of quartz and amorphous silica under the beam.
I don't want to get off-topic but I've always thought that SiO2 glass is more robust under the beam than SiO2 crystal simply because the glass is already at a lower state of entropy. Here is a discussion on silica glass and crystal:
http://smf.probesoftware.com/index.php?topic=130.msg610#msg610
My thinking is that the crystal can only get more disorganized due to phonon excitation by the electron beam, but the glass not so much.
john
Quote from: Probeman on January 03, 2018, 02:52:34 PM
Quote from: jon_wade on January 03, 2018, 10:22:47 AM
I suspect that a lot of the bean damage on an insulator is not related primarily to temp rise, but related to the dielectric strength of the material and the imposed electric field. This would explain, for instance, the very different behaviour of quartz and amorphous silica under the beam.
I don't want to get off-topic but I've always thought that SiO2 glass is more robust under the beam than SiO2 crystal simply because the glass is already at a lower state of entropy. Here is a discussion on silica glass and crystal:
http://smf.probesoftware.com/index.php?topic=130.msg610#msg610
My thinking is that the crystal can only get more disorganized due to phonon excitation by the electron beam, but the glass not so much.
john
I haven't got any pics but when we use high purity Silica capsules in experiments, they are a pig to analyse - you can see a little crater develop as the analysis progresses. Sorry - off topic. :)
An issue that I have also run into when measuring carbon (and nitrogen) in cross section near a surface edge is due to rounding of the sample edge where the steel meets the mounting medium. This can be mitigated by utilizing a "surround" of a similar material hardness to the actual sample that surrounds the sample, and prevent rounding of the edge during polishing.
Here is a nice explanation of the issue:
https://books.google.com/books?id=eP6p3zQNjDYC&pg=PA268&
The problem for EPMA (and EDS) is that because our spectrometers measure the intensity at an angle (take off angle) to the surface of the sample, if the sample surface is tilted towards or away from the spectrometer, the emitted intensity will be strongly affected, especially for low energy emissions (such as carbon, oxygen, nitrogen, etc).
I'm sure others know a lot more about these sorts of effects than I do, so please feel free to chime in.
I've been continuing to work on developing a method for trace carbon in steel and I am discovering that there's more that I don't know, than I do know! :)
It's quite interesting how variable these carbon contamination effects are. For example last June I was performing some carbon measurements using the TDI scanning method (replicate x-ray map frames and extrapolation to zero time on a pixel by pixel basis), and found that the carbon behaves much as we would expect as described here:
http://smf.probesoftware.com/index.php?topic=933.msg5990#msg5990
However, on some carbon point analyses in the past (2015) shown here:
http://smf.probesoftware.com/index.php?topic=48.0
and recent point analyses performed this month, I instead see a downward trend over time as seen here:
(https://smf.probesoftware.com/gallery/395_22_01_18_3_18_36.png)
These are from a traverse of a heat treated sample displaying every 5th point. Each point is several microns apart so they should not be intersecting any carbon from a previous analysis position. Why would the carbon levels decrease over time? The system is quite clean (we use a 100K cryo trapped system) and the sample was carefully cleaned in ethanol and dried in a 60C oven just prior to being placed in the instrument.
Anyway, if this is of any interest to anyone, I thought I would step through the process of analyzing these samples for trace carbon and share what I have learned about characterizing trace carbon in steel. First off, here are the results for the first 12 data points, without any specialized corrections. That is, merely off-peak and matrix corrected results:
Un 13 1 trav (diag)
TakeOff = 40.0 KiloVolt = 12.0 Beam Current = 50.0 Beam Size = 0
Un 13 1 trav (diag), Results in Elemental Weight Percents
ELEM: C N Al Mo Cr Fe Mn Ni Si V
BGDS: EXP EXP EXP LIN LIN LIN LIN LIN EXP LIN
TIME: 40.00 40.00 80.00 80.00 80.00 20.00 40.00 40.00 80.00 60.00
BEAM: 50.42 50.42 50.42 50.42 50.42 50.42 50.42 50.42 50.42 50.42
ELEM: C N Al Mo Cr Fe Mn Ni Si V SUM
769 1.561 12.073 .012 1.117 4.902 85.393 .309 .126 .858 1.342 107.693
770 1.666 11.584 .016 .989 4.854 86.282 .348 .204 .894 .592 107.430
771 1.018 5.889 .011 1.104 5.087 89.852 .324 .149 .950 1.121 105.506
772 .974 5.972 .020 1.048 5.116 89.821 .358 .175 .965 1.063 105.512
773 1.088 5.141 .016 1.015 5.085 91.173 .360 .095 .993 .661 105.625
774 1.230 6.735 .015 1.032 5.100 88.904 .371 .108 .977 .866 105.339
775 1.316 5.323 .013 1.070 5.165 90.751 .372 .154 .994 1.019 106.177
776 1.054 5.001 .018 1.096 5.233 89.668 .346 .118 .984 1.401 104.920
777 .966 5.065 .014 1.125 5.247 89.899 .330 .075 .976 1.524 105.220
778 1.012 4.642 .012 1.010 5.167 91.519 .371 .146 1.007 .509 105.395
779 .925 4.452 .011 1.057 5.196 91.330 .365 .163 1.023 .472 104.995
780 .942 4.759 .008 1.158 5.237 91.516 .321 .189 1.014 .599 105.743
Pretty awful as one can see. Oh, I guess I should show the background fits for carbon (and nitrogen), so here they are, first for carbon:
(https://smf.probesoftware.com/gallery/395_22_01_18_3_45_54.png)
Now we can assume that are not seeing the actual carbon background here. The tails in this region extend a long way, especially when using LDE/PC multi-layer Bragg crystals for these emission lines. But measuring too high a background, would give us *lower* carbon results than we expect, not higher. Either way, the blank correction should take care of it. And here are the results for the nitrogen scan:
(https://smf.probesoftware.com/gallery/395_22_01_18_3_46_15.png)
So, now let's add in some other corrections that might be useful. Let's start with the TDI correction, which (unfortunately) we would expect to further increase our carbon concentration since we are seeing (for whatever reason!), negative slopes in our carbon TDI plots:
Un 13 1 trav (diag)
TakeOff = 40.0 KiloVolt = 12.0 Beam Current = 50.0 Beam Size = 0
Un 13 1 trav (diag), Results in Elemental Weight Percents
ELEM: C N Al Mo Cr Fe Mn Ni Si V SUM
769 1.687 12.426 .012 1.124 4.905 85.939 .310 .126 .858 1.343 108.731
770 1.825 12.013 .016 1.006 4.864 85.644 .348 .204 .894 .593 107.408
771 1.182 5.907 .011 1.115 5.094 88.742 .324 .149 .950 1.122 104.597
772 1.084 5.959 .020 1.049 5.117 89.920 .358 .175 .965 1.063 105.710
773 1.364 5.181 .016 1.030 5.092 90.421 .360 .095 .993 .662 105.214
774 1.393 6.908 .015 1.048 5.108 88.082 .372 .109 .976 .867 104.879
775 1.570 5.366 .013 1.054 5.170 90.220 .372 .154 .994 1.020 105.933
776 1.311 5.339 .018 1.146 5.242 89.817 .347 .118 .984 1.403 105.725
777 1.086 5.105 .014 1.139 5.248 90.361 .330 .075 .976 1.524 105.858
778 1.272 4.520 .012 1.017 5.165 92.675 .371 .146 1.007 .509 106.695
779 1.070 4.407 .011 1.047 5.200 90.444 .366 .163 1.023 .472 104.204
780 1.110 4.931 .008 1.182 5.242 91.641 .321 .189 1.014 .599 106.237
And sure enough we're seeing a small increase of carbon, for example the first data point goes from 1.56 wt.% to 1.69 wt.%. And the average TDI parameters are here:
TDI%: 15.038 1.388 -.169 .025 ---- -.044 ---- ---- ---- ----
DEV%: 2.7 8.2 3.0 4.5 ---- .3 ---- ---- ---- ----
TDIF: LOG-LIN LOG-LIN LOG-LIN LOG-LIN ---- LOG-LIN ---- ---- ---- ----
TDIT: 102.90 104.42 137.20 137.16 ---- 71.79 ---- ---- ---- ----
TDII: 6.13 2.30 3.40 2.59 ---- 87.2 ---- ---- ---- ----
TDIL: 1.81 .832 1.22 .952 ---- 4.47 ---- ---- ---- ----
So a 15% increase in carbon at the ~1.5% wt.% level. But why are our totals so darn high? Well first of all, we appear to have some "native" carbon on our sample, as a "blank" measurement on pure Fe gives us about 1 wt.% carbon. Interestingly, if we look at the carbon TDI on our blank we see almost no TDI effect:
(https://smf.probesoftware.com/gallery/395_22_01_18_3_57_38.png)
How weird is that? Same instrument, different sample...
Now after the blank correction is applied we now obtain these quant results:
Un 13 1 trav (diag)
TakeOff = 40.0 KiloVolt = 12.0 Beam Current = 50.0 Beam Size = 0
Un 13 1 trav (diag), Results in Elemental Weight Percents
ELEM: C N Al Mo Cr Fe Mn Ni Si V SUM
769 .557 12.112 .012 1.124 4.889 85.662 .309 .126 .859 1.339 106.987
770 .696 11.708 .016 1.006 4.848 85.365 .347 .204 .895 .591 105.675
771 .048 5.751 .011 1.115 5.077 88.448 .323 .149 .951 1.119 102.991
772 -.051 5.802 .020 1.048 5.099 89.625 .357 .175 .966 1.060 104.101
773 .231 5.045 .016 1.030 5.075 90.126 .359 .095 .994 .660 103.631
774 .260 6.727 .015 1.048 5.091 87.793 .370 .108 .977 .865 103.254
775 .439 5.227 .013 1.054 5.153 89.931 .371 .153 .995 1.016 104.352
776 .178 5.199 .018 1.145 5.224 89.525 .346 .118 .985 1.399 104.137
777 -.049 4.971 .014 1.138 5.230 90.067 .329 .074 .977 1.520 104.272
778 .137 4.403 .012 1.016 5.148 92.378 .370 .146 1.008 .508 105.127
779 -.065 4.290 .011 1.046 5.183 90.147 .364 .163 1.024 .471 102.634
780 -.026 4.803 .008 1.181 5.225 91.344 .320 .188 1.015 .597 104.656
Now we are seeing carbon pretty much hovering around zero, but our totals are still pretty bad. What could be going on?
Well I didn't mention this yet, but most of you probably guessed that the standards are carbon coated, but the unknown sample is *not* carbon coated- because, you know, we're trying to analyze for carbon here! :)
And at 12 keV, the low overvoltage affects the Fe Ka emissions pretty strongly. And I should also mention that this is why we should always utilize standards- because if we normalize to 100% every analysis looks perfectly fine... So let's specify a carbon coat for the standards by checking these boxes in the Analysis Options dialog:
(https://smf.probesoftware.com/gallery/395_22_01_18_4_05_27.png)
The standards are by default specified as 20 nm of carbon, now how do we specify no coating for our unknown sample? We use the calculation Options dialog as seen here and simply uncheck the Use Unknown Conductive Coating correction for the sample as seen here:
(https://smf.probesoftware.com/gallery/395_22_01_18_4_08_09.png)
And now we obtain these results, with the TDI, blank and coating corrections:
Un 13 1 trav (diag)
TakeOff = 40.0 KiloVolt = 12.0 Beam Current = 50.0 Beam Size = 0
ELEM: C N Al Mo Cr Fe Mn Ni Si V SUM
769 .548 10.071 .012 1.093 4.714 82.164 .297 .119 .836 1.294 101.147
770 .682 9.734 .015 .978 4.674 81.887 .334 .193 .871 .571 99.940
771 .050 4.766 .011 1.085 4.904 84.977 .311 .141 .925 1.083 98.253
772 -.045 4.809 .019 1.020 4.925 86.107 .344 .166 .940 1.026 99.311
773 .226 4.180 .015 1.002 4.903 86.609 .346 .090 .967 .639 98.978
774 .256 5.578 .015 1.019 4.915 84.323 .357 .103 .951 .837 98.354
775 .427 4.331 .013 1.025 4.979 86.419 .358 .146 .968 .984 99.648
776 .175 4.308 .018 1.114 5.047 86.028 .334 .112 .958 1.354 99.448
777 -.044 4.118 .014 1.108 5.053 86.555 .317 .071 .950 1.471 99.613
778 .136 3.647 .011 .989 4.975 88.794 .357 .139 .980 .491 100.520
779 -.059 3.553 .011 1.018 5.008 86.649 .351 .155 .996 .456 98.137
780 -.021 3.979 .008 1.150 5.048 87.788 .308 .179 .988 .578 100.004
Holy Toledo, that's not looking too bad! Maybe we can actually perform trace carbon analyses after all... ;D
Edit by John: Moved this discussion into the carbon in steel topic
My half penneth....
I read raw X-Ray count mapping data direct into Excel.
I start fom the csv file output from the JEOL 8530F+
It works a treat.
You can setup linear calibration on the counts to create a concentration matrix.
You then use conditional formatting to set a min concentration = black and a max concentration/number to white.
Resize the cell size size to 1X1 pixels (or 2).
Copy and paste the grey scaled data into Corel Paintshop Pro (I can't speak for any other program)
You then have the grey scaled image in PSP with, courtesy of Excel knowing best, an extra pixel on the top and the LHS.
This allows you infinite flexibility; you could do
1. ratios of quant concs (dating)
2. Linear and non-linear functions of concentrations (CEQ, Liquidus, phase boundaries etc)
3. Thickness maps of nanofilms - who needs GDS, Auger or HRTEM.
Excel copes readily with 500x500 pixels but grinds very slowly at 1000x1000 pixels.
I suppose it's not optimized for such things but its flexibility is legend.
I enclose one done just this week.
35mm square (1024x1024pixels) quant Ni in steel ALL done in excel.
I intend to combine the Quant Ni, Cr, Mn, Mo and Si into an IDiam map (hardenability map)
Wish me luck :-)
Hi Probeman,
Glad to see you could demagnetise your samples. :)
I note with some interest/concern/amazement the Nitrogen content of these steels. ???
Is that percent?
If so, you should have only Nitrides present at these levels and not a Martensitic steel at all.
In fact, Nitrogen is an Austenite (high temperature phase) stabilizer and you should not be able to quench these steel to Martensite.
For some of the problems you have identified, I refuse to do quant C (or N) in any steel sample. :'(
Other things to consider....
1. The C contamination may not be laid down where the spot is; if the spot heats up the surface, it self cleans and sheds the C to a ring around the spot.
2. If the steel truly is Martensitic then its carbon content may be inhomogeneous on a sub micron scale due to carbide formation.
3. If the steel was originally banded i.e. ferrite and pearlite bands and the Austenising treatment of insufficient time, (homogenising at high temp) then the C distribution may be a remnant of this banding.
NB ferrite has ~0.008 wt% C at RT and Peralite ~0.8 wt% C depending on the CEQ (Metallurgical black art measurement).
Lastly a minor quibble... a carbon content of 0.4 and above would put the steel into a "high carbon content" range and calling this a trace level analysis may raise a few Metallurgists' eyebrows.
For steels, low carbon steels are desired to have ~20ppm C and even these few and far carbon atoms are tied up in Ti(CN).
Quote from: Les Moore on February 20, 2018, 09:57:04 PM
I note with some interest/concern/amazement the Nitrogen content of these steels. ??? Is that percent?
If so, you should have only Nitrides present at these levels and not a Martensitic steel at all.
In fact, Nitrogen is an Austenite (high temperature phase) stabilizer and you should not be able to quench these steel to Martensite.
Hi Les,
Thanks for the info. Yes, there's a lot of nitrogen. I probably shouldn't call it a steel. What would you steel dudes call an Fe alloy composition with 10% N, 1% Mo and 4.5% Cr?
ELEM: C N Al Mo Cr Fe Mn Ni Si V SUM
769 .548 10.071 .012 1.093 4.714 82.164 .297 .119 .836 1.294 101.147
770 .682 9.734 .015 .978 4.674 81.887 .334 .193 .871 .571 99.940As I think I mentioned, the sample is a heat treated nitrided alloy from a academic research lab, so they are expecting significant N near the edge. But I have no idea what is normal for these materials.
The t-test values are unreliable because the composition varies from the edge, but the single point statistics are quite good:
Detection limit at 99 % Confidence in Elemental Weight Percent (Single Line):
ELEM: C N Al Mo Cr Fe Mn Ni Si V
769 .043 .228 .005 .020 .011 .104 .054 .091 .005 .035
770 .034 .222 .005 .020 .011 .123 .052 .088 .005 .038
771 .039 .244 .005 .021 .011 .111 .055 .093 .005 .037
772 .034 .216 .005 .020 .011 .107 .052 .090 .005 .038
773 .034 .227 .005 .020 .011 .101 .054 .095 .005 .036
AVER: .037 .227 .005 .020 .011 .109 .053 .091 .005 .037
SDEV: .004 .011 .000 .000 .000 .009 .001 .003 .000 .001
SERR: .002 .005 .000 .000 .000 .004 .001 .001 .000 .001
Percent Analytical Relative Error (One Sigma, Single Line):
ELEM: C N Al Mo Cr Fe Mn Ni Si V
769 3.0 1.2 17.5 1.3 .2 .4 8.4 30.6 .4 2.1
770 2.2 1.2 13.7 1.4 .2 .4 7.4 18.7 .4 3.9
771 15.6 2.1 20.1 1.3 .2 .3 8.2 26.4 .4 2.4
772 -496.4 2.0 11.3 1.3 .2 .3 7.2 22.0 .4 2.5
773 5.0 2.3 14.3 1.4 .2 .3 7.3 41.7 .4 3.4
AVER: -94.1 1.7 15.4 1.3 .2 .3 7.7 27.9 .4 2.9
SDEV: 224.9 .5 3.5 .0 .0 .0 .5 8.9 .0 .7
SERR: 100.6 .2 1.5 .0 .0 .0 .2 4.0 .0 .3
For me this primarily an exercise in trying to understand how carbon and nitrogen might be quantified in these matrices. Got to love a challenge! :)
Quote from: Les Moore on February 20, 2018, 09:57:04 PM
Lastly a minor quibble... a carbon content of 0.4 and above would put the steel into a "high carbon content" range and calling this a trace level analysis may raise a few Metallurgists' eyebrows.
For steels, low carbon steels are desired to have ~20ppm C and even these few and far carbon atoms are tied up in Ti(CN).
I think there is little if any carbon in the matrix of this sample (except maybe for the first few points on the edge and some random others). The variation may represent sub micron inclusions which are beyond my tungsten gun's resolving capabilities.
Someday maybe I will have access to a FEG instrument!
Quote from: Les Moore on February 20, 2018, 09:35:57 PM
I read raw X-Ray count mapping data direct into Excel.
I start from the csv file output from the JEOL 8530F+
It works a treat.
You can set up linear calibration on the counts to create a concentration matrix.
You then use conditional formatting to set a min concentration = black and a max concentration/number to white.
I have to ask: are these raw x-ray maps background corrected?. If not, do all your phases have the same average Z? Different phases will often have different average Zs and hence different zero concentration intensities, and therefore will require an actual background correction.
What about absorption, fluorescence corrections? Interference corrections? We have quantitative matrix corrections for a reason... ;)
Calibration curves have their place for sure (e.g, carbon in steel), but wouldn't a background corrected, interference corrected and matrix corrected x-ray map be more quantitative?
Looking at the phase diagram again, it would appear that you might have the Fe2N on the surface followed by an intergrowth of Fe4N into the matrix.
From the dim dark past, my memory suggests the former of these grows as a nice hard layer, the latter less so and is often seen as needles.
Beneath these 'layers', you probably have a duffusion profile that is highly compromised by enhanced diffusion down prior austenite boundaries.
The link I shared before has some informative Metallography but not technically deep:
https://vacaero.com/information-resources/metallography-with-george-vander-voort/1138-microstructure-of-nitrided-steels.html
You could try etching it and seeing the phases present; just don't map it after etching in Nital :P
Lastly, you could always map first and ask questions later ;D
The steel is ~98%Fe and there are no significant ZAF factors.
Or, more correctly, I should say the ZAF individual factors are way way below the SE of the measurement and any of the raft of errors that can be attributed to the mapping process.
When you start to have two phases such as in stainless steels you have major issues with localised variations in Cr, Mn, Fe & Ni - then you need full correction.
When you have pearlite, a eutectoid mixture of lamellar of Fe3C and iron, the Z & A factors are a function of the scale and the angle of the colony to the surface; nasty indeed.
Quote from: Les Moore on February 21, 2018, 03:09:32 AM
The steel is ~98%Fe and there are no significant ZAF factors.
Or, more correctly, I should say the ZAF individual factors are way way below the SE of the measurement and any of the raft of errors that can be attributed to the mapping process.
When you start to have two phases such as in stainless steels you have major issues with localised variations in Cr, Mn, Fe & Ni - then you need full correction.
When you have pearlite, a eutectoid mixture of lamellar of Fe3C and iron, the Z & A factors are a function of the scale and the angle of the colony to the surface; nasty indeed.
Hi Les,
OK, that makes sense now. I guess I'm asking a general question about accuracy and calibration curves. I appreciate the power of calibration curves applied to specific situations- as I've said in the past, calibration curves account for everything one doesn't know about! :)
So if one has say trace carbon standards going from zero to say, 1.2 wt% carbon in iron, then I can see how one could assign a zero value and a max value, since the standards cover the range of composition.
I can also see that if one has a background corrected map, and one assumes that the lowest pixel value is a zero concentration, then we might fix the zero concentration level, on that zero basis. That is essentially what one is doing with a blank correction (though the blank correction in PFE has some additional flexibility in that the blank level doesn't have to be zero, and the correction is applied as an intensity correction, with a matrix correction relative to the primary standard).
The nice thing about assigning a zero (blank) is that we can have
a priori reasons for assuming zero, for instance the element is below detection from bulk techniques, but without another standard how do we assign a maximum value?
So here's my question: if one doesn't have a non-zero concentration carbon or nickel standard (as in our examples), how does one assign an upper concentration value to the "white" pixels? I'm probably still missing something...
Quote from: Les Moore on February 21, 2018, 02:58:55 AM
Looking at the phase diagram again, it would appear that you might have the Fe2N on the surface followed by an intergrowth of Fe4N into the matrix.
You are most sagacious, sir!
Un 13 1 trav (diag), Results Based on 1 Atoms of n
ELEM: C N Al Mo Cr Fe Mn Ni Si V SUM
769 .068 1.000 .001 .016 .126 2.044 .008 .003 .041 .035 3.342
770 .087 1.000 .001 .015 .129 2.108 .009 .005 .045 .016 3.414
771 .023 1.000 .001 .033 .277 4.467 .017 .007 .097 .062 5.984
772 -.001 1.000 .002 .031 .276 4.487 .018 .008 .097 .059 5.978
Quote from: Les Moore on February 21, 2018, 02:58:55 AM
Lastly, you could always map first and ask questions later ;D
Since I know very little about these materials, that's about all I can do! :)