As promised I ran some Penepma/Penfluor Monte-Carlo simulations overnight, using the PenepmaMinimumElectronEnergy keyword in the probewin.ini file set to 0.5 keV as described above in order to model oxygen emissions, and here are the results for Al2O3 matrix corrections for Al Ka and O Ka:
First Al Ka at 15 keV using the CalcZAF JTA-Reed analytical matrix correction (Meas. Wt%):
Now the same but using the CalcZAF XPP-Reed analytical matrix correction:
XPP definitely agrees better with the Penepma MC!
Now for O Ka, again first with JTA-Reed CalcZAF for the "Meas. Wt%":
Reasonable agreement for a 500 eV emission line (though remember we are *not* considering any chemical bonding effects with either model) and you will have difficulty seeing the difference for CalcZAF XPP seen here:
What we are probably seeing here are the limits to our knowledge of the photo-elastic absorption physics (scattering factors for Penepma and MACs for CalcZAF).
However, just for fun I loaded in an empirical Mass Absorption Coefficient (MAC) value from Bastin (1992) for O Ka in Al (6720) into CalcZAF (see Analytical | Empirical MACs menu) and the ZAF correction factor changed from 1.876 to 1.875, so maybe not!
Should we try O Ka in TiO2?