PXRD vs. XRF + LOI

[puklen]

Dear X-ray colleagues,
I quantify the mineral content of geological samples with Rietveld using the program TOPAS (and rarely BGMN). I cross-check the results with XRF data which works quite well ... except for (soil) samples that have a very high loss on ignition (LOI; determinded at 1050 °C). An LOI higher than approx. 4 wt% results in an overestimation of quartz (?) with Rietveld thus leading to an erroneous SiO2 content.
To put it in other words: The mineral content recaluclated to major element oxides results in too high an SiO2-content. When comparing with XRF analyses, this SiO2 excess exactly corresponds to the LOI. Example: LOI = 5 wt%, SiO2 (XRF) = 50 wt%, SiO2 (Rietveld-recalculated) = 55 wt%
Has anyone ever come across this problem, too? If the LOI "only" represents the (structurally-bound) H2O-content and some organic compounds, why does it affect the SiO2 content?
I'd be grateful for any small hint!
Best wishes, Jasmin

[johnewarren]

This is all new to me so this could be a silly idea. I'm guessing that your refinement methods differ as one can also see the amorphous SiO2 which could also be present (glass) and this would only appear as a background in the PXRD thus the PXRD refinement would not include it?

Perhaps it would be worthwhile trying an experiment where you purposes add some ground up powdered glass to your sample and see if it throws the measurements out?

Depending on the nature of the glass it has greatly different amorphous backgrounds.

[puklen]

Hi johnewarren,

I am glad you answered. Any suggestion or idea is valuable.
I will certainly try the experiment with ground glass, even though I do not really expect amorphous material in the (mostly) soil samples. Neither do I have any organic compound that I can detect in the diffractogram.

We have measured a kaolinite standard with XRF and XRD. The XRF data give me too low an SiO2 content when comparing the element oxides re-calculated from the chemical formula. It cannot be a problem with XRF, because our standard measurements perfectly agree with reference material.

My guess is that it might have something to do with the clay minerals and/or glass bead making. But why does it affect Si and not some other, volatile element (e.g. Na)?