Author Topic: Binary K-ratios  (Read 239 times)

JohnF

  • Professor
  • ****
  • Posts: 94
    • UW Madison EPMA Lab
Binary K-ratios
« on: October 17, 2019, 08:03:30 am »
PfE supplies a POUCHOU.DAT listing representing those developed by Pouchou (and presumably with Pichoir). There are 756 binaries. My question is, the Green Book's article by P and P gives an appendix with 826 binaries. Were the 70 not included, excluded for some cause?

Probeman

  • Emeritus
  • *****
  • Posts: 1955
  • Never sleeps...
    • John Donovan
Re: Binary K-ratios
« Reply #1 on: October 17, 2019, 09:58:44 am »
It's a good question.  To be honest it's been so long ago that I can't really remember.

But in searching the Probe for EPMA Reference manual I see a note on the POUCHOU.DAT file saying "compiled by Heinrich".  So I assume that I received the file from Kurt Heinrich some 25 years ago.  I have no idea why he didn't include all the k-ratios from the "green book".

One additional note. Some of you may have seen the paper recently published with Philippe Pinard and Hendrix Demers using k-ratios generated from Penepma (Penfluor/Fanal) and fit to binary alpha factors for fast Monte Carlo based matrix corrections:

https://probesoftware.com/smf/index.php?topic=152.msg8276#msg8276

We utilized the POUCHOU.DAT k-ratio file for our accuracy testing, but we did remove the Cu La and B Ka k-ratios to eliminate problematic k-ratios affected by chemical bonding. I attached that slightly smaller k-ratio file below if anyone is interested.
The only stupid question is the one not asked!

Probeman

  • Emeritus
  • *****
  • Posts: 1955
  • Never sleeps...
    • John Donovan
Re: Binary K-ratios
« Reply #2 on: October 17, 2019, 11:10:54 am »
I just noticed that there is another (more recent) file, called Pouchou2.dat, that is also distributed with PFE (and CalcZAF), and that file does contain 826 binary k-ratios.  I think maybe I received this file from Philippe Pinard and/or Hendrix Demers.

You'll have to compare them to see what the differences are.

I also noticed that the Pouchou file, without the Cu La and B Ka k-ratios that I attached to the previous post, contains 777 binary k-ratios, so it must be based on the Pouchou 826 binary k-ratio file.

What we really need are some tertiary (or more) empirical k-ratio measurement files. Binary compositions are great but they don't account for some effects. Also the Pouchou k-ratio database was carefully selected to avoid fluorescence effects, since no one at that time felt we were dealing properly with fluorescence in the matrix corrections.

My "XTREME" k-ratio database proposal from several years ago is here in case anyone is interested:

https://probesoftware.com/smf/index.php?topic=115.0
« Last Edit: October 17, 2019, 12:36:10 pm by Probeman »
The only stupid question is the one not asked!

Probeman

  • Emeritus
  • *****
  • Posts: 1955
  • Never sleeps...
    • John Donovan
Re: Binary K-ratios
« Reply #3 on: October 31, 2019, 10:14:23 am »
Quote
What we really need are some tertiary (or more) empirical k-ratio measurement files. Binary compositions are great but they don't account for some effects. Also the Pouchou k-ratio database was carefully selected to avoid fluorescence effects, since no one at that time felt we were dealing properly with fluorescence in the matrix corrections.

My "XTREME" k-ratio database proposal from several years ago is here in case anyone is interested:

https://probesoftware.com/smf/index.php?topic=115.0

I thought I would take a minute to explain how a database of k-ratios with more than two elements can be handled in CalcZAF. The idea being that we acquire intensities in Probe for EPMA (or any application really), on known standards, and then compare our measured k-ratios to the k-ratios calculated from our matrix corrections.

To do this, first acquire intensities on pairs of primary standards and secondary standards just as one would for a normal probe run. Then from the Probe for EPMA | Output menu, export the standard k-ratios and composition information as seen here:



This so called "standard" k-ratio output was developed by Paul Carpenter and I some years ago, and is described in the Probe for EPMA Reference manual as seen here:



Then from the CalcZAF application, open this output file using this menu:



The error distribution results for each element in each standard matrix will be calculated output to a .out file which can opened in Excel as seen here:



The idea being that we can have any number of elements measured in our standards, not just binary compositions.
« Last Edit: October 31, 2019, 02:02:43 pm by Probeman »
The only stupid question is the one not asked!