Correcting for carbon coating thickness by C-Kα counts

[Jens Andersen]

Dear Colleagues. We had to purchase a new carbon coater around a year ago, as our trusty old Emitech K950x gave up. The new coater is from Agar (Agar Auto Carbon Coater), and we’re really struggling to control the thickness of the carbon coating and get substantial variations between runs. We thought it might be a clever idea to use the C-Ka signal measured on our LDE2H crystal to create a general calibration curve for individual sections, but what appeared to be a simple idea, doesn’t seem to quite work. There are obviously some matrix effects, and after testing the method on glass, I resorted to making calibration curves on plagioclase, augite and ilmenite (fig. 1). The curves look okay, but when I read off the signals and add a manual thickness estimate in the ‘Analyze’ ‘calculation options’, then the software seems to vastly over-correct. Has anybody had success with a similar idea, or might in any other way share insight into what I may be doing wrong?

[Mike Matthews]

Hi Jens,

How are you determining the coating thickness? Film thickness monitor? If this is located to the side rather than at the ‘same’ position as the samples this can produce a large but consistent error.

[Jens Andersen]

We’re using the purple/blue transition on brass as a thickness monitor. We did an emission pattern on a white sheet of paper, which shows us clearly that we shouldn’t coat more than one thin section at a time (although the sample table will take four). On my graphs it would appear that the thinnest coat is a little less than 20nm. The 40 and 60 nm were done by two and three repeat coatings. Regrettably we don’t have any more precise methods of measurement.

[John Donovan]

...We thought it might be a clever idea to use the C-Ka signal measured on our LDE2H crystal to create a general calibration curve for individual sections, but what appeared to be a simple idea, doesn’t seem to quite work. There are obviously some matrix effects, and after testing the method on glass, I resorted to making calibration curves on plagioclase, augite and ilmenite (fig. 1). The curves look okay, but when I read off the signals and add a manual thickness estimate in the ‘Analyze’ ‘calculation options’, then the software seems to vastly over-correct. Has anybody had success with a similar idea, or might in any other way share insight into what I may be doing wrong?

Yes, the thin film calculation in Probe for EPMA (and CalcZAF) is really a modified particle geometry correction and is not designed for samples on a substrate:



Or at least a very low Z substrate, such as a carbon plancette which wouldn't work for your effort!  See here for more info:

https://probesoftware.com/smf/index.php?topic=1241.msg8679#msg8679

However, I've done what you are attempting and it works great but I used STRATAGem (or BadgerFilm) and got excellent results that compared very well with critical angle diffraction XRD.  See here for STRATAGem:

https://probesoftware.com/smf/index.php?board=33.0

and here for BadgerFilm (which is free!):

https://probesoftware.com/smf/index.php?board=37.0

Note that most modern carbon coaters are not evaporators and instead are sputter coaters, and these devices will *not* provide a uniform coating over any reasonably large area.

[Jens Andersen]

Thanks for the replies John and Mike, much appreciated. Just to clarify, it is not my aim to measure the carbon coating thickness, but to measure the minerals in the substrate. My estimate of the coating thickness is merely a means to allow me to correct my quantitative results for minerals in the substrate for variations in the coating thickness, which are substantial. The quantification of minerals in the substrate is the important aspect, not the coating itself. I’ve used the ‘Analyze’ ‘Analysis options’ and changed the thickness to be different to the suggested 200Å based on my measurements, and when I thin hit ‘analyze’ I don’t get the expected result. Suddenly my ilmenite analysis adds up at 105 wt%. I suspect I’m misinterpreting how this function works.
I read from this, that I should perhaps not use the coating thickness to try and make corrections, but instead get the users to repolish, recoat and remeasure their samples. Would that be correct?

[John Donovan]

Ah, OK. Based on your "measuring carbon coat thickness" topic and the fact that you posted this in the sample prep board you got me off-track.  So when you mentioned the Calculation Options dialog I thought you were going into the Particle and Thin Film calculations dialog from there trying to determine coating thickness!

OK, if you are only trying to correct for different carbon thicknesses between your standards and your unknowns you will need to specify the carbon thickness for the unknowns (from the Calculation Options dialog), and also for the standards in the Standard | Standard Coating menu dialog, and then you will need to turn on the carbon coat correction options in the Analytical |Analysis Option menu dialog.  Both of them!

This coating correction works well, so if you don't know your carbon coat thickness you'll have to adjust it until you get totals you are expecting.  Not an ideal situation as I'm sure you know.

See here for some details:

https://probesoftware.com/smf/index.php?topic=23.0

This summer/fall we fixed a few bugs in this code (thanks to some excellent sleuthing by Rom), so you'll want to be sure you are updated to the latest PFE v. 13.7.x by updating using the Help menu as usual.

[Jens Andersen]

Thanks John. You describe the steps that I used, and which seems to lead to the over correction. I will check for updates, we haven’t updated anything since Gareth installed it last year.

[John Donovan]

Yeah, check with the updated app, I'm pretty sure that will fix your issues if you have specified all the parameters correctly:

https://probesoftware.com/smf/index.php?topic=40.0

If not, then your coating thicknesses on your standards and/or your unknowns are not what you think they are.

[Anette von der Handt]

Hi Jens,

The C counts on your three phases are likely affected by a) the carbon deposition/erosion behavior during the analysis as I usually find this is partially affected by the thermal conductivity of the material and b) there is a second order interference by O Ka on C Ka on the LDE2 so you are seeing the effect of different O contents in your phases.

Why don't you calibrate your C counts on the thin section glass? That should have a consistent compositions and C contamination behavior and I would hope that the coating thickness is not that variable across the sample. I know that some people used that method and would get consistent results.

[Jens Andersen]

Hi Anette, thanks for your answer. Yes of course, the oxygen interference will be significant. I did try at first on a standard thin section glass, but our polished thin sections have had their glass ‘buffed’ to effectively bond with the rock slices. So they are significantly pitted as a consequence. That’s why I thought that calibration curves on specific minerals would be better - ideally quartz, but I don’t have much quartz in my samples. I’ll try to upgrade the software next week, as John suggested, when my visitor has finished his work and see how that will work. It does seem like a promising way forward, so I’m not going to despair just yet, it hopefully just needs a few more adjustments. Getting a better coater and resolve the problem alltogether would obviously be the ideal solution.

[Anette von der Handt]

Hi Anette, thanks for your answer. Yes of course, the oxygen interference will be significant. I did try at first on a standard thin section glass, but our polished thin sections have had their glass ‘buffed’ to effectively bond with the rock slices. So they are significantly pitted as a consequence. That’s why I thought that calibration curves on specific minerals would be better - ideally quartz, but I don’t have much quartz in my samples. I’ll try to upgrade the software next week, as John suggested, when my visitor has finished his work and see how that will work. It does seem like a promising way forward, so I’m not going to despair just yet, it hopefully just needs a few more adjustments. Getting a better coater and resolve the problem alltogether would obviously be the ideal solution.

Ah, yes. Rough thin section glass will get you nowhere. Do you have control over the thin section process (in-house versus external)? For example, Spectrum Petrographics puts a rim of quartz sand around each thin section billet that then helps monitor the thickness of the sample. A comparable procedure could provide you with quartz grains irrespective of the rock itself. Just a thought.

Otherwise, we use brass as a thickness monitor as well and I go for a nice orange myself. However, you should always aim to have a comparable thickness to the coating on the standards.

When you use the coat thickness correction in PFE it is important that the coating thickness of the standards is defined correctly in the standard database as well. For most Ka X-ray lines (except the ultralight elements), small differences in coating thickness should be relatively forgiving. If you get 105% totals with the correction, I would make sure that both standards and unknown are correctly defined.

[John Donovan]

Otherwise, we use brass as a thickness monitor as well and I go for a nice orange myself. However, you should always aim to have a comparable thickness to the coating on the standards.

Yes, I agree. Color on brass is quite accurate when using a carbon evaporator.

When you use the coat thickness correction in PFE it is important that the coating thickness of the standards is defined correctly in the standard database as well...

Note that the coating thickness values for standards are not defined in the standard database itself, but instead for each probe run database from the Standard | Edit Standard Coating Parameters menu dialog.

[JonF]

The new coater is from Agar (Agar Auto Carbon Coater), and we’re really struggling to control the thickness of the carbon coating and get substantial variations between runs.

We've got a Cressington 208 carbon https://www.cressington.com/product_208c.html coater, which seems to be the same as the turbo carbon coater from Agar (https://www.agarscientific.com/instrumentation/vacuum-coaters/turbo-carbon-coater) and we've been happily using it for the best part of a decade now. We also use the colour change on brass to produce consistent coating thicknesses - I've even had TEM people come to me to get those very thin carbon films they use that they couldn't otherwise produce with the variety of other carbon coaters we have. 

One thing to note with these systems is that they are very sensitive to when the spring pushing the rods together loses its... springiness. We usually end up changing ours out after a couple of months - the old spring will be notably shorter than the new one.

If you've got the version with a big enough chamber, I recommend the rotary-planetary-tilting stage (https://www.agarscientific.com/rotary-planetary-tilting-stage-for-auto-sputter-carbon-coaters) as this helps give a decent and even coat.

[Probeman]

One thing to note with these systems is that they are very sensitive to when the spring pushing the rods together loses its... springiness. We usually end up changing ours out after a couple of months - the old spring will be notably shorter than the new one.

Yes. After swapping these damn springs out every year or so on our old Edwards carbon coater I finally located some Inconel springs of similar size and stiffness and we've been using those for years now. 

They can get almost red hot and still stay springy.  I think I found them on McMaster-Carr's website.

[JonF]

The curves look okay, but when I read off the signals and add a manual thickness estimate in the ‘Analyze’ ‘calculation options’, then the software seems to vastly over-correct. Has anybody had success with a similar idea, or might in any other way share insight into what I may be doing wrong?

Another thing you might want to look at is the density of the carbon coat. There was a thread a while back (here: https://probesoftware.com/smf/index.php?topic=995.msg7125#msg7124) where it was discussed that a density of ~1.35 g/cc was more appropriate for evaporated carbon films.