EDS analysis and simulations newbie

[Walter]

Hello,

I'm quite new to EDS analysis and I have some basic questions for you experts.

Now I'm trying to get the same quantitative results between real specimens and simulations. My samples are homogeneous material (metal or glass) with a homogeneous metal coating.

For the simulation I've tried pyPENELOPE, CASINO, DTSA-II and CalcZAF.

CASINO doesn't give me the spectrum but maybe I could use the emitted intensities and put them in CalcZAF, is this a correct procedure?

DTSA-II for now is my favourite, even if I've found that if I put its k-ratio in CalcZAF the results are better, anyone have noted that?

I've excluded pyPENELOPE because I'm not able to extract the k-ratios and the spectrum doesn't look like real because the peaks look like lines.

I'd like to understood better the workflow behind the EDS quantification, that in general should be: spectrum -> k-ratio -> ZAF correction -> quantification. The physics and algorithms behind the spectrum simulation can wait for now, I just need to know the essentials. I've not understood how to manually calculate the k-ratios: I should use peak intensity (height), peak area (I think so), or something else (I've read something about Top-Hat filter)?

I've read something about the ZAF correction and it doesn't seem to be something easy, well defined and unequivocal. That's right?

Question summary:
CASINO + CalcZAF should give me the same results as DTSA-II and DTSA-II + CalcZAF?
How could I extract quantitative information from pyPENELOPE?
How could I manually calculate k-ratios?

Thank you so much

Walter

[Probeman]

For the simulation I've tried pyPENELOPE, CASINO, DTSA-II and CalcZAF. CASINO doesn't give me the spectrum but maybe I could use the emitted intensities and put them in CalcZAF, is this a correct procedure?

Hi Walter,
If they output the emitted intensities, then yes, you can calculate the unknown composition by using the unknown emitted intensity (and also the emitted intensity from a standard) and form a k-ratio and input that into CalcZAF.  Remember, the k-ratio must be constructed using the *net* emitted intensities, that is, the unknown and standard emitted intensities must be background corrected.

DTSA-II for now is my favourite, even if I've found that if I put its k-ratio in CalcZAF the results are better, anyone have noted that?

I'm not an expert with DTSA-II (Nicholas Ritchie) or PyPenelope (Philippe Pinard), so I can't really say.  However I should point out that the CalcZAF installation also comes with the Standard.exe app which contains a full implementation of Penelope/Penepma (it's essentially the same FORTRAN) as shown here:

http://probesoftware.com/smf/index.php?topic=202.msg5378#msg5378

I've excluded pyPENELOPE because I'm not able to extract the k-ratios and the spectrum doesn't look like real because the peaks look like lines.

pyPenelope is a GUI for Penelope/Penepma and the calculation shows the natural widths of the lines, which are much higher resolution than what you see with EDS detectors.  The Penepma GUI in Standard.exe (part of CalcZAF) will show the same natural widths of the emitted lines.  But the peak widths don't matter. In fact the emitted (net) intensities from pyPenelope/Penepma are output to a file called pe-intens-01.dat. So just run a Penelope/Penepma simulation for the unknown and standard compositions, ratio the emitted net intensities from the pe-intens-01.dat files and you have your k-ratio for quantification.

I'd like to understood better the workflow behind the EDS quantification, that in general should be: spectrum -> k-ratio -> ZAF correction -> quantification. The physics and algorithms behind the spectrum simulation can wait for now, I just need to know the essentials. I've not understood how to manually calculate the k-ratios: I should use peak intensity (height), peak area (I think so), or something else (I've read something about Top-Hat filter)?  I've read something about the ZAF correction and it doesn't seem to be something easy, well defined and unequivocal. That's right?

That's basically it, but I would say the steps are: spectrum simulation, obtain net emitted intensities (unk and std), ratio them, perform quantification. Now the quantification could be ZAF, could be phi-rho-z, could even be the fast Monte Carlo method in CalcZAF!   You will note that there are 10 ZAF and phi-rho-z quantification methods in CalcZAF, that give 10 different results (not to mention which mass absorption coefficients utilized), so yes, they are certainly not "easy, well defined and unequivocal"!

One last point. For simulating standard intensities I always just use pure elements, that way one doesn't have to calculate a standard k-factor. Using a compound standard would force one to calculate the matrix effects for that standard composition, and that requires additional assumptions about the physics. That is, for pure element standards, the standard k-factor is always 1.0, so just use pure element standards for these simulations.
john