Peak Width and Spectral Interferences

[cvidito]

I apologize in advance if this can be found in the manual. How does the software account for the difference in peak widths when correcting for interferences? For example, the width of a Zn Ka peak on an LIFL crystal is wider than on an LIF crystal and even wider on an LIFH. There has to be a difference in how the software makes corrections on these different spectrometers, correct? Is there specific information that needs to be entered into the configuration files for this? Thanks.

Chris

[Probeman]

I apologize in advance if this can be found in the manual. How does the software account for the difference in peak widths when correcting for interferences? For example, the width of a Zn Ka peak on an LIFL crystal is wider than on an LIF crystal and even wider on an LIFH. There has to be a difference in how the software makes corrections on these different spectrometers, correct? Is there specific information that needs to be entered into the configuration files for this? Thanks.

Chris

Hi Chris,
No worries.

The only time my software utilizes the FWHM peak width is when performing a nominal calculation of potential interferences as seen here in Probe for EPMA:

http://probesoftware.com/smf/index.php?topic=71.msg3492#msg3492

There is also a calculate nominal interference calculation window in the Options | Interferences menu in Standard. The peak width utilized in these nominal calculations can be edited in the config files, but it is only a nominal calculation simply for predicting whether an observed interference is actually possible or not.

For the spectral interference correction, the peak widths are not utilized. It may seem weird, but for the actual quantitative interference correction, the peak width doesn't enter into the calculation at all!  How could this be?  Well it's because the quantitative interference correction is based solely on the on-peak intensities. First in a standard used for the interference correction, and second in the actual unknown. And of course the matrix corrections in the unknown and interference standard.

This is explained in some detail in the User Reference manual (as you guessed) and in a published paper described here:

http://probesoftware.com/smf/index.php?topic=69.msg1189#msg1189

The full paper can be seen as an attachment in this post at the very end:

http://probesoftware.com/smf/index.php?topic=69.msg257#msg257

[John Donovan]

This is a small tweak to the code, but when trying to evaluate whether a non-zero value in a standard represents a contamination or an actual interference, one can utilize the Calculate WDS Interferences button in the Standard Assignments dialog as seen here:



The program will then calculate "nominal" interferences for the current (on-peak) emission based on an assumed gaussian peak width. If the suspected interfering element does not show up, then the non-zero value in the standard could actually be a trace contamination.

This is one reason why we should generally acquire all the elements that we are analyzing for in all our standards (at least the first time).  Because if we find a non-zero value in a standard that should be zero, and the Calculate WDS Interferences output shows that it is a potential interference (based on the nominal peak overlaps), then we are already measuring the interfered *and* the interfering element, and we can assign the interference correction to improve our accuracy.

This contamination vs. interference issue also comes up sometimes in the MAN fit curves, where we might see higher intensity outliers in our standard measurements, and the question is: are these high outliers a contamination or a spectral interference? See this old post:

https://probesoftware.com/smf/index.php?topic=307.msg1773#msg1773

It's nice to be able to look at the MAN background curves to evaluate how pure one's standards are...

Anyway, the point is that now the program will output the actual emission line position with the possible interferences. I was doing some software training at PNNL a couple of weeks ago and one of the new users there, Nathan Canfield, suggested it and it's a good idea I think.

Especially for the nominal interferences calculated for the high or low off-peak positions as seen here in the Elementas/Cations dialog:



Also, we added it to the nominal calculation for the Misc Options | Interferences menu dialog in the Standard application as seen here:



Where the program will check if any elements in a standard are likely to interfere with other elements in that standard.

[Probeman]

I don't know if this would be useful to anyone or not, but the intermediate calculations for the quantitative spectral interference correction can be seen by checking the Output | Verbose Mode menu in Probe for EPMA. Here is a tiny portion of what the output looks like when a quant analysis is performed and the Verbose Mode menu is checked:

[Rom]

Could you explain, if program works only with real peak positions of interfered element, what reason to show program the order ("Intf Order I-V") of interference line in the Standard Assignments dialog? 

[John Donovan]

Could you explain, if program works only with real peak positions of interfered element, what reason to show program the order ("Intf Order I-V") of interference line in the Standard Assignments dialog?

That's a very good question. The answer, is described in the original 1993 paper found here:

https://epmalab.uoregon.edu/publ/Improved%20Interference%20(Micro.%20Anal,%201993).pdf

Basically, we apply a matrix correction to account for the difference in composition between the unknown and the standard utilized for the interference correction. Because clearly the standard utilized for the interference correction *cannot* be the same composition as the primary standard. That is, the standard utilized for the interference correction must contain a known concentration of the interfering element, but none of the interfered element, nor any other interfering elements.

So if the interfering line is a first order emission line, the energy of the interfering line will be exactly equal to the interfered line, therefore the absorption and atomic number corrections for the interfering line will be the same as the interfered line in the standard used for the interference correction (but which might not be the same emission line as the emission line causing the observed interference!). 

The fluorescence correction will only be the same for K lines since the Ka and Kb emission line share the same edge, so one may optionally not include the fluorescence correction, though because the fluorescence correction is generally small, we usually leave it in.

On the other hand, spectral interferences from higher order emission lines are higher energies than the normal emission line being observed (a 2nd order interference will have twice the energy of the interfered line!), so the same matrix correction does not apply. Therefore since these higher order emission lines causing an interference will generally be less affected by absorption (because they are much higher energies), the interference correction in Probe for EPMA will not apply the matrix correction to the interfering intensity for these higher order cases. But it's not a bad idea to document the identity of the interfering line!

Jakub Haifler has been working on method to perform a completely rigorous matrix correction and it looks very promising, however, it requires that one must know the exact identity of each interfering line, which can be somewhat ambiguous in many situations.  See here for more details:

https://probesoftware.com/smf/index.php?topic=1175.0