Most of us understand the effects of "unanalyzed" elements in the matrix correction and how it is important to include these elements either by specification, difference or by stoichiometry to other analyzed elements, in order to obtain an accurate matrix correction.
For example the issue of unanalyzed water in hydrous glasses, where one gets a results with a total of say 95% and one might be forgiven to think that, OK, I've got 5 wt% H2O there because 100 - 95 = 5 in Excel. But that would be wrong because Excel doesn't know anything about matrix correction physics. In fact what we see when we specify water by difference is described in this and other posts in this topic:
https://probesoftware.com/smf/index.php?topic=92.msg7701#msg7701Basically adding oxygen (and hydrogen) to our glass matrix correction results in a significant change in the concentration of other elements. In a generic example of a silicate glass, the Si (and other element) concentrations increases significantly. Why is this? Because it turns out that Si Ka is not well absorbed by Si atoms, but they are well absorbed by oxygen atoms! Once the matrix correction "knows" about the extra oxygen (water), the absorption correction increases correspondingly.
So in our generic example of a 95% total in a hydrous glass, once we add H2O by difference and include it in the matrix correction, because the Si (and other elements) increase by about 1%, we instead obtain about 4% H2O by difference. This effect was described in the Roman et al. paper referenced here:
https://probesoftware.com/smf/index.php?topic=61.msg4303#msg4303Now this makes some intuitive sense to me because Si Ka is a fairly low energy emission line and therefore fairly well absorbed by other elements, but I never expected this to be the case for Fe Ka at 6.4 keV!
Here's an example for a magnetite sample where the Fe is expressed, as we usually do in EPMA, as FeO:
Un 96 7138_PPO-2_Mgt4_HO-1, Results in Elemental Weight Percents
ELEM: Fe Mg Si Ti V Mn Cr Al O
TYPE: ANAL ANAL ANAL ANAL ANAL ANAL ANAL ANAL CALC
BGDS: MAN MAN MAN MAN EXP LIN MAN MAN
TIME: 40.00 120.00 40.00 80.00 30.00 30.00 80.00 120.00 ---
BEAM: 45.93 45.93 45.93 45.93 45.93 45.93 45.93 45.93 ---
ELEM: Fe Mg Si Ti V Mn Cr Al O SUM
146 60.944 1.463 .058 5.329 .265 .422 .375 1.542 23.842 94.240
AVER: 60.944 1.463 .058 5.329 .265 .422 .375 1.542 23.842 94.240
SDEV: .000 .000 .000 .000 .000 .000 .000 .000 .000 .000
SERR: .000 .000 .000 .000 .000 .000 .000 .000 .000
%RSD: .00 .00 .00 .00 .00 .00 .00 .00 .00
STDS: 396 396 14 22 23 25 396 396 ---
STKF: .1836 .0330 .4101 .5547 .6328 .7341 .3060 .0469 ---
STCT: 1119.3 3499.0 4307.2 9290.4 8344.6 14435.6 1338.8 1992.0 ---
UNKF: .5710 .0066 .0004 .0548 .0028 .0040 .0043 .0088 ---
UNCT: 3481.1 701.1 4.3 917.6 37.1 77.9 18.9 373.9 ---
UNBG: 12.2 47.8 2.8 32.9 15.3 29.2 6.1 29.4 ---
ZCOR: 1.0674 2.2108 1.4127 .9728 .9416 1.0658 .8682 1.7520 ---
KRAW: 3.1102 .2004 .0010 .0988 .0044 .0054 .0141 .1877 ---
PKBG: 286.31 15.65 2.55 28.88 3.43 3.67 4.10 13.70 ---
INT%: .00 ---- ---- ---- -15.76 ---- -1.38 ---- ---
As one can see our Fe concentration is 60.944 and our total is 94.240. So we're missing about 5% of something and that something of course is mostly the excess oxygen in the Fe2O3 molecule in magnetite. Now how much of an effect can this ~5% missing oxygen have on our Fe concentration? What would you guess? I wouldn't have guessed this much:
Un 96 7138_PPO-2_Mgt4_HO-1, Results in Elemental Weight Percents
ELEM: Fe Mg Si Ti V Mn Cr Al O
TYPE: ANAL ANAL ANAL ANAL ANAL ANAL ANAL ANAL DIFF
BGDS: MAN MAN MAN MAN EXP LIN MAN MAN
TIME: 40.00 120.00 40.00 80.00 30.00 30.00 80.00 120.00 ---
BEAM: 45.93 45.93 45.93 45.93 45.93 45.93 45.93 45.93 ---
ELEM: Fe Mg Si Ti V Mn Cr Al O SUM
146 61.421 1.451 .057 5.379 .267 .425 .383 1.532 29.085 100.000
AVER: 61.421 1.451 .057 5.379 .267 .425 .383 1.532 29.085 100.000
SDEV: .000 .000 .000 .000 .000 .000 .000 .000 .000 .000
SERR: .000 .000 .000 .000 .000 .000 .000 .000 .000
%RSD: .00 .00 .00 .00 .00 .00 .00 .00 .00
STDS: 396 396 14 22 23 25 396 396 ---
STKF: .1836 .0330 .4101 .5547 .6328 .7341 .3060 .0469 ---
STCT: 1119.3 3499.0 4307.2 9290.4 8344.6 14435.6 1338.8 1992.0 ---
UNKF: .5710 .0066 .0004 .0548 .0028 .0040 .0044 .0088 ---
UNCT: 3481.6 701.6 4.3 918.5 37.1 77.9 19.1 374.1 ---
UNBG: 11.8 47.3 2.8 32.0 15.3 29.2 5.9 29.2 ---
ZCOR: 1.0756 2.1904 1.4061 .9808 .9506 1.0740 .8773 1.7396 ---
KRAW: 3.1106 .2005 .0010 .0989 .0044 .0054 .0143 .1878 ---
PKBG: 296.98 15.83 2.54 29.68 3.43 3.67 4.25 13.80 ---
INT%: .00 ---- ---- ---- -15.75 ---- -1.36 ---- ---
Holy cow! Our Fe concentration went from 60.944 to 61.421 which is almost 0.5 wt% absolute. And note how the ZCOR (matrix correction) for Fe Ka went from 1.0674 to 1.0756. But even more mind blowing is to compare the *other* elements we measured: Mg, Ti, Al, etc. What happened to them? Well they went slightly *down* in concentration! Why? Because they are more absorbed by Fe than by O. It just goes to show you that one cannot intuit physics, you have to just run the darn calculation.
Of course we really should calculate the excess oxygen in our magnetites/ilmenites using a ferric/ferrous calculation (and more on that later), but the oxygen by difference calculation demonstrates that once we obtain our ferric/ferrous ratio, we really need to calculate the excess oxygen from that and specify it in Probe for EPMA, to obtain a more accurate Fe concentration for our mineral thermodynamic calculations.
Now, I'm no geologist but perhaps one should then recalculate our ferric/ferrous ratios, this time using our improved Fe concentration. Again, more on this later...