APF corrections

[Joe Boesenberg]

John

I was doing some reading last week and came across APF corrections. I looked on the Forum and at your UofO website for additional information. I don't want to offend anyone who uses this technique since it does work, but given that the goal of most of the microprobe experts is to get away from matrix matching and universally standardize the analysis of materials and minerals from a true physical point of view (based primarily on atom and electron interactions), aren't APF corrections just fudge factors? They have to be calibrated to a specific standard on a specific spectrometer on a specific crystal. There is no inter-lab uniformity. Isn't this a kind of matrix matching? Actual true matrix matching may be easier in this case especially since you need to know the mineral (peak shape and position) in order to even calibrate and correct for it. Given that most of these problems occur with boron and carbon in material science labs, most of these endmember standard materials can be purchased relatively easily. Are APFs really the best way of doing this? Seems like we might need some form of artificial intelligence to recognize the peak shape and peak shift and provide corrections for them, rather than asking the inexperienced user to do some rather machine intensive measurements.

[John Donovan]

Hi Joe,
That's a fair point.

But there are really two types of APF corrections, one is what I call a "specified" APF and the other is what I call "compound" APFs.  They are described in detail here:

https://probesoftware.com/smf/index.php?topic=536.0

The "specified" APF is exactly as you describe, a perfectly matrix matched peak shift/shape factor that is applied when using a "specific" primary standard anjd a "specific"unknown.  In this case, one is making a peak shift/shape correction factor for a specific unknown material, for a specific element and they need to be acquired and applied on a case by case basis. Usually as you say for quantifying a light element such as boron, nitrogen, oxygen or carbon.  See Bastin's publications from the 90s on these:

https://probesoftware.com/smf/index.php?topic=1422.msg10470#msg10470

See this white paper for an example of using "specified" APFs:

https://epmalab.uoregon.edu/reports/Preliminary%20work%20on%20MgB2%20and%20MgB4.pdf

But compound APFs are not matrix matched. They are created by combining single element APFs, which are grouped into families, for example oxides or carbides. One example for measuring oxygen one might use MgO as the primary standard for oxygen k-ratio and then combine fractions of the other oxides APFs (MgO to Al2O3, MgO to SiO2, MgO to Fe2O3, etc., etc.) that are present in the unknown to obtain a "compound" APF for your particular unknown composition.  This calculation is performed in the matrix iteration in PFE.

These compound APFs are my own invention but they seem to work great and don't require performing a bunch of peak shape measurements, usually.  See the Empirical APFs dialog in Probe for EPMA. Because the first order effect for chemical effects is the specific element chemical bonding, the secondary effect is valence (small effect) and the third order effect is coordination, which is usually an even weaker effect. 

Yes, as you say, although one would think that these binary to compound APFs would vary from spectrometer to spectrometer, it's not as much of a problem as one might have thought.  Partly I think because if one is (for example) measuring oxygen, you're probably using a W/Si LDE with a 2d around 60 angstroms which is what Bastin used (ignoring the stearate diffractors which due to their higher spectral resolution yield more larger APF effects than LDEs, but they are rarely if ever used these days). And again, for example nitrogen, we're all probably using an LDE specific for nitrogen.  But of course this assumption should be tested.

Here is a test I did myself on MgO to Al2O3 and SiO2 and compared to Bastin's APF measurements:

https://epmalab.uoregon.edu/reports/APF%20measurements.doc

And yes, these peak shift/shape effects tend to become smaller with increasing emission line energy (it's basically a problem for emission lines that involve shells that are also involved in chemical bonding). And since we're talking at most a few percent APF corrections this only applies to major elements, not trace elements. 

So for example we can in general ignore APF issues for Si ka by simply using a silicate standard as a primary standard, as opposed to using Si metal as a primary standard. If the Si atoms in the primary standard and unknown are both bonded to oxygen (e.g., SiO2 and Mg2SiO4), the chemical effects are very small and the APF issues can be ignored. The matrix correction can handle the physics.

The question is at what point can we use a metal as a primary standard for oxide materials.  This is something discussed here:

https://probesoftware.com/smf/index.php?topic=1423.0

[John Donovan]

It should also be pointed out that one is not required to utilize area peak factors (APFs) when there are chemical peak shift/shape effects between ones primary standard and unknowns.

For example, in the case of sulfur the chemical effects seem to produce primarily a peak shift rather than peak shape effect as described here:

https://probesoftware.com/smf/index.php?topic=127.0

So that means that one can simply utilize different peak positions for ones standard vs ones unknowns.

Another alternative is to utilize integrated peak areas for quantification. That is because even if the chemical effects modify the peak shape, if one uses the entire peak area, the quantification will be correct.  This method is described here:

https://probesoftware.com/smf/index.php?topic=536.msg2992#msg2992

This integrated intensity scanning method of course is much more time consuming than using APFs, but it always works for any chemical peak shift/shape situation.

[Rom]

Here is a test I did myself on MgO to Al2O3 and SiO2 and compared to Bastin's APF measurements:

Could you suggest the methodology for APF measuring with extra peaks inside BG range.
The example is on the picture. We need estimate APF for Oxygen Ka on Al2O3 with MgO like a standard.
But WS for oxygen on Al2O3 has been littered by AlKaIII line and integrated intensity for O Ka will be wrong.

[JonF]

Here is a test I did myself on MgO to Al2O3 and SiO2 and compared to Bastin's APF measurements:

Could you suggest the methodology for APF measuring with extra peaks inside BG range.
The example is on the picture. We need estimate APF for Oxygen Ka on Al2O3 with MgO like a standard.
But WS for oxygen on Al2O3 has been littered by AlKaIII line and integrated intensity for O Ka will be wrong.

You'll want to export the wavelength scans and fit multiple Voigt peaks. Once fitted, you then just use the area of the oxygen peak. I use Origin Pro to do this, e.g. below:



Green lines = individual peaks
Red line = sum of fitted peaks
black line = measured spectrum

[John Donovan]

You probably already noticed that there are two values for the APF of oxygen (k-alpha) in Al2O3 in the EMPAPF.DAT file, that is visible from the Analytical | Empirical APFs menu in Probe for EPMA.

One is relative to MgO (mine) and the other relative to Fe2O3 (Bastin). They are slightly different as seen here:

  "o"     "ka"    "mg"       1.0000       "MgO/Fe2O3/WSi/59.8"
  "o"     "ka"    "al"       1.0213       "Al2O3/Fe2O3/WSi/59.8, Bastin"
  "o"     "ka"    "al"       1.0285       "Al2O3/MgO/WSi/59.8, Donovan"

Note that this difference in the two values (1.0213 vs. 1.0285) is probably *not* because the primary standards are different, because the APF of O Ka in MgO relative to Fe2O3 is 1.000 (see first APF above).  Which is very convenient because Bastin used Fe2O3 as a primary standard for measuring O Ka APFs, while I used MgO as the primary standard for O Ka.

The small difference I suspect is because these two LDE diffractors are probably somewhat different in their actual 2d spacings.

[Rom]

You'll want to export the wavelength scans and fit multiple Voigt peaks. Once fitted, you then just use the area of the oxygen peak. I use Origin Pro to do this, e.g. below:

It is great, thank you very much! I'll search this direction. How can I find this App with explanations of calculation and operation logics?
But this is a bit different topic, may be we can use PFE for APF calculation in cases I exampled.

[Rom]

You probably already noticed that there are two values for the APF of oxygen (k-alpha) in Al2O3 in the EMPAPF.DAT file, that is visible from the Analytical | Empirical APFs menu in Probe for EPMA.
.....
Note that this difference in the two values (1.0213 vs. 1.0285) is probably *not* because the primary standards are different, because the APF of O Ka in MgO relative to Fe2O3 is 1.000 (see first APF above).  Which is very convenient because Bastin used Fe2O3 as a primary standard for measuring O Ka APFs, while I used MgO as the primary standard for O Ka.

The small difference I suspect is because these two LDE diffractors are probably somewhat different in their actual 2d spacings.

Thank you, of course I researched of EMPAPF.dat file with your and Bastin's data .
I gave the example of Al2O3 WS only like very common illustration for task which I want to able to solve.
My question is: could you suggest any approaches to APF measuring in cases like this.   

[John Donovan]

For the situation where you have interfering peaks, yes, you should utilize the deconvolution method that Jon suggests.  There are a number of software packages that can perform such operations.

PFE can calculate integrated to peak intensity ratios from the Model Background dialog from the Plot! window, but it doesn't de-convolve overlapping peaks.
john

[Rom]

Ok, thank you.
May be you can suggest some simpler program than Origin Pro - it looks quite expensive and difficult.

And one more question: can I download WS data back to PFE after their correction in other programs?

[xllovet]

To fit WDS spectra I use a program called "fityk", it is free:

https://fityk.nieto.pl/

Xavier

[John Donovan]

And one more question: can I download WS data back to PFE after their correction in other programs?

You can only import wavescans from other probe database files:

https://probesoftware.com/smf/index.php?topic=74.msg274#msg274

If you think about how much meta data is associated with each element wavescan and how dependent that data is on the version number of the probe database, you will understand that importing external data into Probe for EPMA is a rather  complicated process.

The PFE database is designed to handle raw data acquired from the instrument.  I'm afraid to even ask why you want to do this: what exactly would be the point of importing deconvolved wavescans back into PFE? 

[Rom]

what exactly would be the point of importing deconvolved wavescans back into PFE?

Just for calculate integral intensity of peaks on corrected (deconvoluted) WS.

[John Donovan]

what exactly would be the point of importing deconvolved wavescans back into PFE?

Just for calculate integral intensity of peaks on corrected (deconvoluted) WS.

Oh, I would just use the fityk app that Xavier suggested above.

But remember, both the integrated and peak intensities should be corrected for background before calculating the APF...

[Rom]

Yes, this is why I wanted to download corrected data back in PFE. BG can be non linear and I am not sure, is it possible to use exponential BG in the Fityk App.