Indetermination Fe, Cu,... in simple stoichiometric oxide

[aducharme]

Do you have good wavescans of the Fe Ka peak in the different standards? They don't have to be good enough to discern the small shifts-I want to see the shapes. I'm wondering if there's some chemical something that changes between the oxides and sulfides and breaks the assumption that the peak intensity is proportional to the integrated intensity.

[Probeman]

Do you have good wavescans of the Fe Ka peak in the different standards? They don't have to be good enough to discern the small shifts-I want to see the shapes. I'm wondering if there's some chemical something that changes between the oxides and sulfides and breaks the assumption that the peak intensity is proportional to the integrated intensity.

There are indeed possibilities to have Ka/Kb ratios very. In fact one can almost see something in these Mn Ka1/Ka2 scans:

https://smf.probesoftware.com/index.php?topic=1423.msg14418#msg14418

I put the MDB file in the MAS 2006 Google drive that you should have access to, but we can certainly do some wider scans that include the Kb line.

[aducharme]

Oh I think it's because you didn't repeak Fe Ka between samples. These would imply FeS2 and Fe metal, and FeSiO4 and Fe3O4, have very very similar Fe Ka peaks, but FeS2-Fe304 and Fe2SiO4-Fe only have very similar Fe Ka energies. You could even predict that Fe2SiO4 will not work with Fe metal given the first three results.

[Probeman]

Oh I think it's because you didn't repeak Fe Ka between samples. These would imply FeS2 and Fe metal, and FeSiO4 and Fe3O4, have very very similar Fe Ka peaks, but FeS2-Fe304 and Fe2SiO4-Fe only have very similar Fe Ka energies. You could even predict that Fe2SiO4 will not work with Fe metal given the first three results.

Normally I would say yes. In cases of pure peak shift, using different peak positions for the oxide and the metal (e.g., Si, Al, Mg, S, Na, etc.) will allow high accuracy quantification:

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

And maybe that is the case here. But see these wavescans from Jon Fellowes' SX100:



The Ka1 peaks are normalized in intensity though not shifted. But I'm not seeing a peak shift between the Ka1 peaks, but I am seeing a slight difference in the shape of the Fe Ka2 line between the oxide/silicate and the metal. Could that explain an ~2% difference in the quantification when switching the primary standard from oxide to metal as shown here:

https://smf.probesoftware.com/index.php?topic=1423.msg14418#msg14418

[aducharme]

Maybe we aren't seeing due to the normalization that the Ka1 peak is taller in Fe metal than in fayalite and  hematite b/c the Ka1 peak is slightly preferred in that sample than the other two? I don't think this would be fixed by background correction because a different distribution within the lumped peaks could only slightly affect the intensity of two points far away from the peaks.

I wonder if you still got the two percent difference if you deconvolved the Ka1 and Ka2 peaks before matrix correction.

I think it's hard to see on the peak, but based on how the fayalite is further left than hematite and Fe metal predominantly from about 47900 to roughly 48000 s.u., I would venture the fayalite peak is further to the left than Fe and hematite. I'm having a hard time believing it myself though.

[Probeman]

I think it's hard to see on the peak, but based on how the fayalite is further left than hematite and Fe metal predominantly from about 47900 to roughly 48000 s.u., I would venture the fayalite peak is further to the left than Fe and hematite. I'm having a hard time believing it myself though.

I'm not claiming a peak shift per se. But I am wondering if the areal ratios of the Ka1/Ka2 are slightly different between the metal and the oxide. That could affect the quantification of metal vs. oxide if we only utilize the Ka1 peak intensity...

I think this makes sense when we consider this:

Now let's change the primary standard from magnetite to Fe metal:



Ooops. That doesn't look very good.

The Fe metal Ka2 curve is lower than the oxides as seen here:

And maybe that is the case here. But see these wavescans from Jon Fellowes' SX100:

To me that means slightly more photons are emitted in the Ka1 band (relative to the Ka2 band) in the Fe metal.

If we then use Fe metal as the primary standard that would decrease the Fe quant in the oxide.  Right?

[aducharme]

Yes, that's largely what I said. I think it's a peak height issue instead of a peak area issue though.

Maybe we aren't seeing due to the normalization that the Ka1 peak is taller in Fe metal than in fayalite and  hematite b/c the Ka1 peak is slightly preferred in that sample than the other two? I don't think this would be fixed by background correction because a different distribution within the lumped peaks could only slightly affect the intensity of two points far away from the peaks.

I wonder if you still got the two percent difference if you deconvolved the Ka1 and Ka2 peaks before matrix correction.

[Probeman]

Yes, that's largely what I said. I think it's a peak height issue instead of a peak area issue though.

Maybe we aren't seeing due to the normalization that the Ka1 peak is taller in Fe metal than in fayalite and  hematite b/c the Ka1 peak is slightly preferred in that sample than the other two? I don't think this would be fixed by background correction because a different distribution within the lumped peaks could only slightly affect the intensity of two points far away from the peaks.

I wonder if you still got the two percent difference if you deconvolved the Ka1 and Ka2 peaks before matrix correction.

OK, cool. 

We'll need to do more tests but to my eye the difference in the Ka2 peak heights for the oxide/silicate vs. metal appears to be (very) roughly around 2%.

Of course we should still check the Fe Ka/Kb ratios on oxides and metals, but here's my question:

If the ratio of the Fe Ka1/Ka2 is higher in the metal than in the oxide/silicate, why is that?  What atomic bonding mechanism would affect the metal/sulfide Ka1/Ka2 ratio compared to the oxide/silicate?

[Probeman]

As seen here:



One can obtain excellent accuracy for Fe Ka analyzing oxide and silicates using an oxide primary standard and excellent accuracy analyzing sulfides (and presumably alloys) using an Fe metal primary standard. Even when utilizing the same spectrometer positions for all materials!

But exactly why do we need to use different Fe primary standards?  Is there a very subtle peak shift involved between the oxide/silicate and the metal/sulfide? And what exactly is the physics mechanism that causes this? 

We certainly see such peak shifts in lower Z elements, e.g., Si Ka, Al Ka, S Ka, etc.  But it has been generally believed that these subtle peak shifts would "fade away" by the time we reach the first row transitional metals.  But data presented in this topic would seem to indicate that there is a peak shift occurring although only on the order of 1-2% relative for Fe Ka.  For example here is magnetite analyzed using Fe metal as a primary standard:

St  895 Set   4 Magnetite (std #395), Results in Elemental Weight Percents
 
ELEM:       Fe      Fe      Al       O
TYPE:     ANAL    ANAL    SPEC    SPEC
BGDS:      LIN     LIN
TIME:    60.00     ---     ---     ---
BEAM:    30.08     ---     ---     ---
AGGR:              ---     ---     ---

ELEM:       Fe    Fe-D      Al       O   SUM  
XRAY:     (ka)    (ka)      ()      ()
   494  71.149     ---    .200  27.640  98.989
   495  70.839     ---    .200  27.640  98.679
   496  70.908     ---    .200  27.640  98.748
   497  71.042     ---    .200  27.640  98.882
   498  70.986     ---    .200  27.640  98.826

AVER:   70.985     ---    .200  27.640  98.825
SDEV:     .120     ---    .000    .000    .120
SERR:     .054     ---    .000    .000
%RSD:      .17     ---     .00     .00

PUBL:   72.080    n.a.    .200  27.640  99.920
%VAR:    -1.52     ---     ---     ---
DIFF:   -1.095     ---     ---     ---
STDS:      526     ---     ---     ---

Here the peak position was the Ka1 peak on the Fe metal standard as seen here:



And we are off by ~1.5% relative in the quant.

But what if we moved the Fe Ka peak position to the flat region between the Fe Ka1 and Ka2 peaks (which are partially resolved)?  Like this:



And now let's analyze both oxides and sulfides using Fe metal as a primary standard:



Yes, the NIST glasses are off by ~1% relative, which is only around 800 PPM absolute, but the oxides and sulfide are now in agreement! 

Andrew Ducharme thinks this indicates that a very subtle peak shift is occurring between the oxide/silicate and metal/sulfide even for Fe Ka and I think I agree with him.

This metal to oxide peak shift should be more apparent for Mn Ka and Cr Ka as indicated in some of the previous posts above.

[aducharme]

What do you mean be peak shift effect? I think this demonstrates the higher energy resolution unlocks a much more consistent measurement method for Fe Ka through Fe Ka2.

Here are the wavescans of the Fe Ka peak in Fe metal measured with an LiF (Sp5) and LLiF (Sp3)



The total Ka1,2 peak observed with LiF crystal very quickly drops in intensity, especially on the right side, so very small differences in where the spectrometer is placed can have highly non-negligible effects on the quantification. I suspect this sharp drop is not as apparent during typical peaking alignments as the continuous ROM scanning will further decrease wavescan resolution. The LLIF crystal, however, reveals a fairly flat shelf made by the Ka2 peak that makes it hard to misalign the spectrometer. Using the Ka2 peak across all measurements allows you to approximate standard WDS-EPMA operation and get high-accuracy results.

[Probeman]

What do you mean be peak shift effect? I think this demonstrates the higher energy resolution unlocks a much more consistent measurement method for Fe Ka through Fe Ka2.
...
The total Ka1,2 peak observed with LiF crystal very quickly drops in intensity, especially on the right side, so very small differences in where the spectrometer is placed can have highly non-negligible effects on the quantification. I suspect this sharp drop is not as apparent during typical peaking alignments as the continuous ROM scanning will further decrease wavescan resolution. The LLIF crystal, however, reveals a fairly flat shelf made by the Ka2 peak that makes it hard to misalign the spectrometer. Using the Ka2 peak across all measurements allows you to approximate standard WDS-EPMA operation and get high-accuracy results.

Yes, I agree.

The broader shelf present on the LLIF near the Ka2 peak I think allows for some degree of peak shifting (metal to oxide) without affecting the quantitative accuracy.  But the fact that using the narrower Ka1 peak affects accuracy indicates to me that there must be some minimal amount of peak shifting occurring with Fe Ka1 between Fe metal and Fe oxides.

But that it is very difficult to see this in wavescans in this case of Fe Ka or other higher Z transition metals.

That is why I suggest we perform wavescans on Mn, Cr, Ti etc. Ka. As a first effort we could compare the metal to the oxide for these other elements.