Author Topic: REE M lines  (Read 122 times)

sem-geologist

  • Professor
  • ****
  • Posts: 67
REE M lines
« on: April 07, 2021, 05:04:23 AM »
I am really bombarding this channel with topics... This topic actually is much broader, and tackles the problem for other xray software too. But as this problem affects me directly in my attempt of REE fluoride measurements on DTSA-II, I am writing this here.
What is striking me is that simulated REE M lines with DTSA-II is so different from what is measured. The lighter REE the bigger discrepancy, albeit even Lu is not perfect. There is few different stuff bothering me which is interconnected.
1. Where are Ba and La M alpha lines? They are not listed in DTSA-II line selector. But that is not the only place where its existence is ignored. I see that Cameca Peaksight is ignoring them too. Bruker Esprit is not taking them from existence, but still is weakly neglecting it (see further points)
2. The relative line intensities for REE M lines are bizarre, where dominating line is Mζ, and Mα for LREE is given ridiculously low weight. The Esprit does exactly the same, and existing there Ma of Lanthanium and Barium have smallest weight.
3. Simulated spectra on DTSA-II follows those intensity weights and additionally to that introduce some peak shifts at Ma position, which is not observed in real spectras.

I thought that it is related with self absorption known for REE M lines (see the attached publication which demonstrates that), and those software automatically disables those lines for high voltage. But if I simulate spectra for low voltage (3kV) it still keeps the bizarre intensity ratios, and Ma lines are non-existing or neglected. This is troubling me, as ignoring the existence of those lines or neglecting them probably leads to neglecting them in fluorescence simulation. That means that near surface generated REE Ma by fluorescence (thus would have not much absorption) is not accounted. It probably is the cause why Auto referencing of M lines works so poorly.

I  have one more interesting observation, I tried to modify EDS spectra by stripping Oxygen (removing equivalent amplitude of Oxygen spectra component generated by autofit for SiO2)... however I forgot that there is perfect standard for M lines of REE, at least I have such for Lanthanum - that is LaB6. Interestingly LaB6 gives different proportions between zeta and alpha(or collection of bizarre IUPAC line names if we ignore existence of alpha). I made WDS scan on LPC0, and compared that to LaPO4 scan, and there are some amplitude differences, albeit at LaB6 M alpha peak gets some flanks - so probably this is some crystallization thing. And that is probably source where those theoretical mark intensities with higher M zeta peak comes from. This can have terrible consequences for fluorine measurement in REE bearing minerals, where interference correction would depend from state of Ce in the sample. This however does not explain why M Alpha peak in simulation is so shifted, while theoretical marker positions fit well with measured M alpha positions, it looks those position converges at LuPO4 (see attachements).
Maybe La M zeta in LaPO4 seems to be relatively smaller to Malpha as it is partly absorbed by Oxygen?

If I strip oxygen peaks from my REEPO4 spectras and use those as reference for M lines, will DTSA-II respect relative line weights or will it try to fit theoretical intensities? If it ignores/neglects Ma lines for LREE, how is fitting going to work?

NicholasRitchie

  • Moderator
  • Professor
  • *****
  • Posts: 68
    • NIST DTSA-II
Re: REE M lines
« Reply #1 on: April 07, 2021, 05:34:15 AM »
The physics of M-lines is poorly known and *may* even change from compound to compound.

You can select which lines are visible on a given detector using the "Detector Calibration" alien.  This creates a new "Detector calibration" which you will need to apply to your unknown spectra using the "Spectrum Properties" dialog.
"Do what you can, with what you have, where you are"
  - Teddy Roosevelt

sem-geologist

  • Professor
  • ****
  • Posts: 67
Re: REE M lines
« Reply #2 on: April 07, 2021, 05:48:34 AM »
I mean not all M lines are missing, only the M alpha lines are missing (M zeta is present, and M UIPAC notation lines are there present, calibration wont change that). M lines are hard due to self absorption. See the attached paper about that - it is quite enlightening.

sem-geologist

  • Professor
  • ****
  • Posts: 67
Re: REE M lines
« Reply #3 on: April 08, 2021, 06:54:40 AM »
What is that "BestFit" spectra when I create a bundle?
I am still fighting with REE and F;
It appears only after I create a bundle, but if I try to load a bundle again, it is not there. Is this trying to do quanti alien on primary standard spectra using everything in the bundle?

For CePO4 I stripped the Oxygen, by subtracting Oxygen peak generated by SiO2 standard bundle, and trimmed the remaining part left to where O peak was. I saved that trimmed spectra, and when making standard bundle for CePO4 I had used that trimmed spectra for Ce M lines (after doing auto for O and C coating).
I tried to do quanti alien on the same standard (CePO4), albeit with a few aberations - O from SiO2, and for seeing that I would get 0% of Fluorine (like a blind standard for F to check overlap correction), I had added to quanti list of elements F where CaF2 as standard.
The result is excellent, with O just 1% off, but still in the error margin, Most importantly Fluorine is 0.0004 +/- 0.0005. So basically it is zero as it should be! Yahoo! The residual also looks exceptionally good. But this simulation looks off, is this analytical simulation as from simulation alien?

These "Simulation", and "BestFit" spectra scared shit out from me every time I was trying to make REE calibrations – I was loosing hope and thought that I am doomed to fail and this method won't work at all. Scrapped all detector configurations twice and began from zero. I am lucky I had tried to do quanti alien to see that actually this works and those mentioned spectras are irrelevant. (Are they irrelevant? please, say yes)

update: "Best Fit" spectra for PrPO4 using same Oxygen stripping stratege for bundle gives very good fit.
« Last Edit: April 08, 2021, 07:31:24 AM by sem-geologist »

sem-geologist

  • Professor
  • ****
  • Posts: 67
Re: REE M lines
« Reply #4 on: April 09, 2021, 04:55:59 AM »
The progress of measuring F is going interestingly. However, I am considering alternative strategy. I would want to calculate F stochiometrically (as unknown substance/mineral has no Oxygen, but is fluoride). That would allow to ignore most of Ba,La,Ce,Pr Mz-lines and Ma-lines completely - the ones which are the most self-absorbed and overlaps F Ka.

I am looking at scripting, and after reading DTSA-II_Scripting_quantification.pdf and trying around in jython console, I feel stuck. is it possible to customize the ox-is-f = epq.Oxidizer() so that it would be fluorine instead of oxygen?

sem-geologist

  • Professor
  • ****
  • Posts: 67
Re: REE M lines
« Reply #5 on: April 12, 2021, 02:45:14 AM »
I hope this is last question as I am getting very good results. I see no perceivable difference for chevkinite or britholite analysis when calculating O stochiometrically, with full PAP, XPP (PAP simplified) or even ZAF, recalculation models. However, I get quite big different results for F when  using different models for Gagarinite. PAP and XPP gives similar result for most of elements, but significantly differs for light elements. What is the reason that XPP is default method (I understand that naive ZAF is outdated)?
As F Ka is highly overlapped with LREE M lines, and there is huge uncertainty how that interference is resolved I think I would better recalculate F by stochiometry (for Gagarinite).

I wish the epq.Oxidizer class would be better documented somewhere, The uncertainties introduced by M-lines to K lines overlaps could be overcome by calculating anions by stochiometry. It would be nice to do stochiometric F, or even bunch of elements, i.e. CO3 (in case of carbonates) and put those through same matrix correction (We can always recalculate composition with stochiometry after analysis, but in case of missing elements, the matrix correction often is biased).

I am also a bit surprised that despite different matrix correction, the residual (for same set of standards) is the same. I know that on EPMA-WDS'es interference intensities needs to go through same matrix correction loops to account for absorbed portions of intensities. in Case of DTSA-II this obviously is not done (as then we would get slightly different residuals)... unless I am missing something obvious again. We don't need to mark where is background, so DTSA-II is determining the bremstrahlung background automatically (auto-magically) - without known composition given, which makes me wondering - how is it done? And then it needs only the reference peaks to fit those in between real emission and calculated bremstrahlung spectra? How can it be done correctly without taking into consideration the absorption - which can be estimated only after getting the composition of material?

Despite some minor shortcomings, I am impressed. This looks so good for my case that I am starting to draft a paper comparing the EPMA-WDS vs SEM-EDS for beam ultra-sensitive complex materials.

PS. The EDS spectra can be exported directly to plain LaTeX script. This is kind of feature which makes this software even more attractive to kind of typography-junkies like me.

sem-geologist

  • Professor
  • ****
  • Posts: 67
Re: REE M lines
« Reply #6 on: April 12, 2021, 04:42:53 AM »
I still can't get why we need reference spectra if it is not used at all. (I am using DTSA-II Microscopium version)
Considering an example with 3 strategies for La M lines:
1) using LaB6 for shapes and intensities, where LaMz is clearly higher than La M alpha (non existent for DTSA-II), no reference fitting is required, as La M lines are far enough from B Ka line. So intensity and reference by LaB6.
2) use LaPO4 intensities, with La M line shapes referenced from LaB6.
3) use LaPO4 intensities, with La M lines referenced with manipulated spectra of LaPO4, where Oxygen is interactively subtracted, and continuum left to it is trimmed. LaPO4 has better pronounced La M alpha line than LaB6.

I would expect those strategies would produce the different residuals for La M region when used, but they do not. They produce different resulting residual for La L -lines region (but reference is asked to be provided for M region). DTSA-II clearly ignores provided reference peaks for residual generation (see the ROI in second picture - those are La to Nd (and Sm) M alpha lines ignored, thus there is a positive plateau-like bump in the residual, as all those lines are neglected with bizarre minor MxNy line substitutes).  This makes F and Na measurements very susceptible to analytical/deconvolutional bias.
« Last Edit: April 12, 2021, 07:56:22 AM by sem-geologist »

jrminter

  • Professor
  • ****
  • Posts: 70
Re: REE M lines
« Reply #7 on: April 12, 2021, 05:05:06 AM »
I'm. confused - you mentiond saving spectra as LaTeX script. I can't find a way to do so. One can save a .csv format. Am I missing something?  I am using Microscopium 2021-03-23 revision.

NicholasRitchie

  • Moderator
  • Professor
  • *****
  • Posts: 68
    • NIST DTSA-II
Re: REE M lines
« Reply #8 on: April 12, 2021, 05:28:03 AM »
To save to LaTeX (or Encapsulated PostScript), you need to have GNUPlot installed.  It was a feature I added many years ago but is rarely used.
"Do what you can, with what you have, where you are"
  - Teddy Roosevelt