Low totals in oxides

[Probeman]

I was analyzing Fe-Ti-oxides a few months ago with a very similar routine as you -- see attached spreadsheet with analysis conditions. I was also getting low totals when calculating O by stoichiometry, then I decided to measure O directly and it seemed to fix the problem. O Ka was analyzed on PC0.

Hi Will,
It's not a bad idea to measure oxygen, especially for unknown chemistries (valences). Certainly measuring oxygen might be essential if you are dealing with glasses, for example the method of Nash described here:

https://probesoftware.com/smf/index.php?topic=922.msg5937#msg5937

Though measuring oxygen can be problematic, given that unless one has a properly matrix matched oxygen standard, there may be peak shape/shift effects requiring the use of Area Peak Factors (APFs) per Bastin:

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

and also possible need for empirically measured Mass Absorption Coefficients (MACs):

https://probesoftware.com/smf/index.php?topic=8.msg5258#msg5258

But rather than measuring oxygen and all of the issues with that, I have found that using the method of Droop (1987), which is now built into Probe for EPMA, we now get amazingly accurate totals for oxides and other minerals without actually measuring oxygen (at least when the mineral's cation to oxygen ratios are known!):

https://probesoftware.com/smf/index.php?topic=92.msg8593#msg8593

I suspect the secret to this surprising accuracy is that when the excess oxygen from iron is calculated using the method of Droop, the excess oxygen is included into the normal matrix correction (along with the stoichiometric oxygen from the other cations). This means that the matrix effects of this excess oxygen, (which are surprisingly large even for Fe Ka!), get handled automatically.  The above link demonstrates these matrix effects nicely.

I would welcome the posting of other examples of iron bearing minerals such as pyroxene, garnet, etc., especially mixed ferrous/ferric minerals in this topic...

what about Bruker  eds-wds  quant mapping?

Bruker EDS mapping is well-integrated with PeakSight, but from what I can tell EDS quant accuracy is not particularly good because it uses on-peak calibration only.

Are you referring to major element accuracy or trace element accuracy?  I don't know what PeakSight does with the EDS spectra from Bruker, but in the EDS integration with Probe for EPMA, the EDS spectrum peak stripping to obtain net intensities is handled by the Bruker software, while the actual quantification is performed using standards acquired by PFE (also stripped for net intensities by Esprit), using whatever matrix corrections specified in PFE. basically the same matrix corrections used for your WDS elements.

Here's an old post that describes in this in slightly more detail:

https://probesoftware.com/smf/index.php?topic=226.msg1052#msg1052

We have found that the EDS accuracy (co0mpared to WDS) to be quite good, that is, at least until the EDS deadtime gets too large, and then well, you know, the wheels come off.   

Attached to the post linked below is a PPTX I presented at M&M a few years ago that shows what sort of accuracy one can obtain using EDS on an EPMA (remember to login to see attachments):

https://probesoftware.com/smf/index.php?topic=79.msg6195#msg6195

The cool thing is that because we using EPMA instruments have a much better control over the instrument geometry (no tilting stage for example), we can actually get much better accuracy than an SEM. Not to mention that in EPMA we tend to use standards!  Nicholas Ritchie makes these points often in his presentations also.