Hi,
Recently I run into some trouble analyzing pyrochlore (ideal formula (Ca,Na)2(Nb,Ta)2O6(OH,F)). I have been using Nb L-alpha, Ca K-alpha, Na K-alpha, Ta L-alpha at 25 kV, 20 nA. The setup includes many other elements (Y, REE, K, Si, Al, Fe, Mg, Mn, Ti, Zr, Ba, Sr, Pb, U, Th) but Ca, Na, Nb, Ti and F makes around 90-95% of the total oxides. My problem concern, I think, the matrix correction. Using Armstrong ZAF correction, all my totals are around 96-98% (including F and recalculated OH based on 1 hydroxyl group site) and the Nb wt-% is apparently too low (around 65-66% element, I would expect more like 68-69%). To check the quality of the results I did perform a mineral formula recalculation assuming a total of 6 oxygens + 1 OH; these yield too low value for the Nb-site (that include all 5+ and 4+ cations plus probably some 3+ cations), but close to 2 for the totals of 1+ and 2+ cations (mainly Ca and Na). However, recalculating the data with either the Heinrich or the P&P matrix correction yield better results, both in term of totals and mineral formula recalculation.
Problems we can rule out...
a) My standards have been recently strongly repolished to remove any oxide layer (~6 months ago), so I doubt this is the issue. Moreover, if this was a problem with the standard (e.g. getting oxidized), I would expect my totals to get too high, not too low.
b) The analyzed samples are well polished, and the standardization is only 2 days old.
c) I've checked the peak position, and there is no significant change between the standard (Nb metal) and the unknown (Nb oxide).
d) An erroneous background position is unlikely. All backgrounds yield coherent values and have been double-checked by numerous scans within the standards. Moreover it would take a huge error on a background value to yield a difference of 3% or so on Nb...
d) I have performed several analyses on the same (unzoned) grain at different beam size (10, 5 and focused) to check for possible beam damage, and all three results are within the error identical.
Unfortunately I have no way to test my analysis using a secondary Nb standard (preferably an oxide), and the only standard I do have for Nb is simply a pure metal.
Does anyone has experience with Nb analysis? Which of the numerous matrix correction available in Probe for EPMA are you using? My tests suggest that Heinrich or Pichou & Pouchoir would be the most appropriate, but...
You will find attached to this email a summary of the various oxide wt-% results (average of 4 points) using the different matrix correction, and (on the second page) a summary of the CalcZAF calculation for one of these points. The most obvious difference is a lower backscatter coefficient (and Z) in Armstrong for Nb, along with a slightly lower MACs calculation and slight difference in absorption coefficient. Any thoughts on this problem would be greatly appreciated!
Thanks in advance for your help!
Julien