matrix correction comparison

[Probeman]

It can probably be determined from the compositions, but someone(!) should make a list correlating the Oregon Dahlheim glass letter designations with the Cal Tech P standard numbers.

Since we actually have mounts and material available for these glasses, further investigations would be quite useful I think.

By the way, the glass measurements I did in the MDB file above are just a first attempt, but measurements at multiple keVs would be even more useful from a matrix correction comparison perspective, so if someone would like to borrow my Dalheim glass mount I would be pleased to share it.  The only anomaly I saw in my initial efforts some 10 years ago was that glass "J" seemed to be a duplicate of glass "C", but this could a mistake by the student mounting the glasses. Specifically it appears that what was labeled as glass "C" is actually glass "J", but this can be double checked.

Paul has all these materials (as do I), so this can be double checked.  But they do seem to be very accurate and homogeneous glasses. It would be nice to get some "round robin" k-ratio measurements of these glass at multiple keVs from multiple instruments...  if anyone is interested!

john

PS If you are looking for the original history of old EPMA instruments discussion that was here, I moved those comments here to a new thread for History of EPMA:

http://probesoftware.com/smf/index.php?topic=924.0

[wrigke]

Hello everyone!
  I have been analyzing some U-Pu-Zr alloys using PFE.  I am using Heinrich MACs, with extrapolation to actinide M lines.  When I run "all matrix corrections", the differences between corrections are surprising to me. 

U ranges from 54.3-55.9 wt.%, Pu ranges from 15.9-17.5 wt%, and Zr ranges from 22.0-25.6 wt%.  The correction that I think is closest to "truth" is the Heinrich/Duncomb Reed correction, which reports (note totals are not close to 100% because fission products are not included here):

U 55.0 wt%
Pu 16.4 wt%
Zr 22.0 wt%

In contrast, the full PAP correction, which I think is fairly widely used and accepted reports:

U 55.5 wt%
Pu 16.8 wt%
Zr 23.9 wt%

When I look at all the correction factors what sticks out to me is the absorbance correction for Zr.  It is much higher in most of the matrix correction algorithms than in the Heinrich Duncumb Reed (HDR).

What I would like to do is try to understand why there is such a large difference (especially for Zr) between the PAP and HDR corrections.  It seems to me that the most obvious way to do this is to find out what equations are used for each correction and compare them to one another to see how the k-ratios and MACs are treated differently. 

So my questions are as follows:
1.  Does this seem like a reasonable approach?
2.  Does anyone know if the HDR correction was intended to be used with Heinrich MACs?
3.  What literature specifically discusses the HDR correction?  What I see from PFE is that HDR is composed of parameters from many different researchers.  For example, its absorption correction is from Philibert [FRAME] and its mean ionization parameter is from Berger and Selzer.  Do we know that these parameters developed by different researchers are internally consistent and are meant to be used together?  I can't find literature that specifically discusses HDR in depth.
4.  On what types of samples was HDR designed and tested?

Any insight will be greatly appreciated!

Cheers,

Karen

[John Donovan]

Hi Karen,
I moved your topic to this existing topic (with the exact same name!)...  so you might want to check the previous posts in this topic.  I should also suggest that you check out the CalcZAF feature shown here:



This requires you to export a composition from Probe for EPMA (which was analyzed as a standard, but you said you thought you knew what the composition should be...) using the PFE Output | Save CalcZAF Standard Format menu, and then run CalcZAF to apply all 10 matrix corrections and all 7 MAC tables as shown here:



The output file (or at least a piece of) is shown here loaded into Excel, so you can plot histograms of the different matrix corrections and MAC combinations to see what the distributions are:



Maybe this will be useful.

[wrigke]

Hi John,

I had done what you suggested before I posted here.  That is what showed me that there is a significant absorbance correction difference between HDR and PAP for Zr.  Why that is, is what I would like to understand.

Karen

[John Donovan]

It's a good question. One would have to examine the details of the equations utilized, which can be found on Github. The ZAF.BAS routine contains the source code for all the corrections:

https://probesoftware.com/smf/index.php?topic=570.0

In general one can only say that the Heinrich-Duncumb-Reed method is an older ZAF type correction that is quite different from the more modern PAP phi-rho-Z method which combines the absorption and atomic number corrections. Realistically one should combine the absorption and atomic number corrections in HDR and compare those combined absorption and atomic number correction values in the PAP method.  Of course just because a method is newer doesn't mean that it's better...  it depends on the compositions involved.

That's because, the largest differences in these two methods are probably the standard materials utilized to "tune" these the equations at the time they were created. PAP used the POUCHOU.DAT and POUCHOU2.DAT k-ratio databases (provided with CalcZAF and found in the CalcZAFDATData folder), while the HDR method probably utilized a different k-ratio database (I'm not sure which, but it could have been the NISTBIN.DAT k-ratio database which is also found in the CalcZAFDATData folder).

You might want to contact Brian Joy at Queens University as he has looked into these equations in some detail.   

My final comment is that with high atomic number materials like this, the backscatter correction can be quite large and because current backscatter corrections are improperly based on mass concentrations, there are significant errors associated with them in some compounds. This is the motivation behind our recent paper:

https://probesoftware.com/smf/index.php?topic=1111.msg11954#msg11954

And we are working on implementing a new backscatter correction in CalcZAF/PFE as I write this. You might want to look at the A/Z ratios of the elements in this material and see how different they are.  The more difference in the elemental A/Z ratios in a compound, the larger the mass bias effects in the backscatter correction.

[Anette von der Handt]

Hi,

I am trying to compare some old data sets, done on our previous probe (Cameca SX-50) with the Geller software, with some new data.

I am not familiar with the Geller software and the processing computer is around but in storage. Does anyone know what the Geller software likely used as the MAC set and is the PAP different somehow or the same as the one in CalcZAF/PFE?

[Probeman]

Hi,

I am trying to compare some old data sets, done on our previous probe (Cameca SX-50) with the Geller software, with some new data.

I am not familiar with the Geller software and the processing computer is around but in storage. Does anyone know what the Geller software likely used as the MAC set and is the PAP different somehow or the same as the one in CalcZAF/PFE?

As far as I know, the Geller dQuant used one of the matrix corrections from Armstrong's CITZAF code. If they used the PAP code from Armstrong' CITZAF there are two things I can mention. 

First, there were some typos in the CITZAF PAP code which Paul Carpenter and Brian Joy and I fixed in CalcZAF.  This is documented in the code on the CalcZAF GitHub.

Second there was another bug in the fluorescence correction in CITZAF that Donovan fixed in CalcZAF very early on, but I think Geller fixed that bug around the same time as he reported it to them.