Mixed oxidation states in CalcZAF

[Brian Joy]

Hi John,

Here is an example of an analysis (PAP model, accelerating potential = 15 kV, takeoff angle = 40 deg.) of pure magnetite using a pure magnetite standard.  I’ve assumed that all Fe in the unknown is present as FeO.  Under “Calculation Options,” I’ve selected “Calculate with Stoichiometric Oxygen.”



If I then recalculate the result assuming 3 cations and 4 oxygens per formula unit, I obtain 67.996 wt% Fe2O3, 30.591 wt% FeO, and oxide total = 98.586 wt%.  (I’ve used molar mass Fe = 55.845 g/mol and molar mass O = 15.9994 g/mol.)

The standard and unknown both contain 72.359 wt% Fe and 27.641 wt% O, and Fe k-raw is precisely one.  If it is assumed that Fe is present in the unknown as FeO, then the calculated weight per-cent of oxygen is 20.731, producing an oxide total of 93.090 wt%.  However, during the matrix correction iterations, compositions are normalized, and, during the first iteration, the Fe and O contents of the unknown become, respectively, 77.730 and 22.270 wt%.  If matrix corrections are determined based on this composition, then, relative to the standard, the Fe atomic number correction (PAP model) is 0.9851 (=1.0659/1.0820) when in truth it should be precisely one.  This mostly accounts for the anomalously low wt% Fe in the result since the absorption correction is close to unity.

Why does CalcZAF not include a provision to account for mixed oxidation state of a chosen element during the matrix correction iterations?  Or is this somehow possible in CalcZAF, and I just don’t see it?

[Probeman]

Why does CalcZAF not include a provision to account for mixed oxidation state of a chosen element during the matrix correction iterations?  Or is this somehow possible in CalcZAF, and I just don’t see it?

Hi Brian,
CalcZAF mostly certainly does account for different valences in the matrix correction for oxygen. In fact I've spent considerable effort to make sure that excess oxygen is handled properly even in cases where one has halogen replacement of stoichiometric oxygen (e.g., biotites/phlogopites). 

There is an example of magnetite in the CalcZAF.dat sample data file (from the File | Open CalcZAF Input data File menu), but it's assuming 6.9 % or so of excess oxygen and FeO.  Here is the CalcZAF.dat example magnetite with excess oxygen specified:

SAMPLE: 6, TOA: 40, ITERATIONS: 3, Z-BAR: 20.96875

 ELEMENT  ABSCOR  FLUCOR  ZEDCOR  ZAFCOR STP-POW BKS-COR   F(x)u      Ec   Eo/Ec    MACs
   Si ka  1.5172  1.0000   .9336  1.4164   .8910  1.0478   .5969  1.8390  8.1566 1955.60
   Fe ka   .9969  1.0000  1.0666  1.0633  1.0944   .9746   .9877  7.1120  2.1091 55.6969
   Mg ka  2.4698   .9999   .9272  2.2899   .8716  1.0638   .3508  1.3050 11.4943 4477.22

 ELEMENT   K-RAW K-VALUE ELEMWT% OXIDWT% ATOMIC% FORMULA KILOVOL                                       
   Si ka -.00008 -.00003   -.005   -.010   -.005   -.001   15.00                                       
   Fe ka 1.35236  .67675  71.956  92.571  42.452   7.929   15.00                                       
   Mg ka  .00069  .00033    .075    .124    .101    .019   15.00                                       
   Mn                       .054    .070    .032    .006
   Ca                       .000    .000    .000    .000
   Ni                       .000    .000    .000    .000
   Al                       .201    .380    .245    .046
   O                       6.899   6.899  14.207   2.653
   Ti                       .012    .020    .008    .002
   O                      20.861   -----  42.959   8.023
   TOTAL:                100.053 100.053 100.000  18.677

You might find it helpful to run through all the examples in this data file, as it demonstrates many of the CalcZAF modes.

But if you set the excess oxygen to zero in the magnetite example, and then specify Fe as Fe3O4, then you will get this:

SAMPLE: 6, TOA: 40, ITERATIONS: 2, Z-BAR: 20.97229

 ELEMENT  ABSCOR  FLUCOR  ZEDCOR  ZAFCOR STP-POW BKS-COR   F(x)u      Ec   Eo/Ec    MACs
   Si ka  1.5173  1.0000   .9335  1.4164   .8909  1.0478   .5969  1.8390  8.1566 1955.23
   Fe ka   .9969  1.0000  1.0665  1.0632  1.0943   .9746   .9877  7.1120  2.1091 55.6884
   Mg ka  2.4701   .9999   .9272  2.2900   .8716  1.0638   .3508  1.3050 11.4943 4476.34

 ELEMENT   K-RAW K-VALUE ELEMWT% OXIDWT% ATOMIC% FORMULA KILOVOL                                       
   Si ka -.00008 -.00003   -.005   -.010   -.005   -.001   15.00                                       
   Fe ka 1.35236  .67675  71.953  99.438  42.476   7.929   15.00                                       
   Mg ka  .00069  .00033    .075    .124    .101    .019   15.00                                       
   Mn                       .054    .070    .032    .006
   Ca                       .000    .000    .000    .000
   Ni                       .000    .000    .000    .000
   Al                       .201    .380    .246    .046
   O                        .000    .000    .000    .000
   Ti                       .012    .020    .008    .002
   O                      27.732   -----  57.142  10.666
   TOTAL:                100.022 100.022 100.000  18.666


The ZAF window should look like this:

[Probeman]

One can also simply enter Fe3O4 using the "Enter Composition As Formula String" button and you will get this after clicking the calculate button:

ELEMENT  ABSFAC  ZEDFAC  FINFAC STP-POW BKS-COR   F(x)e
   Fe ka  1.0157  4.3900  4.4588   .2087   .9161   .9846
   O  ka  1.4270  3.9154  5.5873   .2438   .9546   .7008

SAMPLE: 32767, TOA: 40, ITERATIONS: 0, Z-BAR: 21.02464

 ELEMENT  ABSCOR  FLUCOR  ZEDCOR  ZAFCOR STP-POW BKS-COR   F(x)u      Ec   Eo/Ec    MACs
   Fe ka   .9969  1.0000  1.0660  1.0626  1.0935   .9748   .9877  7.1120  2.1091 55.6319
   O  ka  1.6478   .9937   .8648  1.4160   .7989  1.0826   .4253   .5317 28.2114 3227.13

 ELEMENT   K-RAW K-VALUE ELEMWT% OXIDWT% ATOMIC% FORMULA KILOVOL                                       
   Fe ka  .00000  .68095  72.359   -----  42.857   1.000   15.00                                       
   O  ka  .00000  .19520  27.641   -----  57.143   1.333   15.00                                       
   TOTAL:                100.000   ----- 100.000   2.333


Or just enter pure Fe with a concentration of 72.359 wt%, and 3 cations and 4 oxygens and and then click the Calculate with Stoichiometric Oxygen checkbox and you will get this:

ELEMENT  ABSCOR  FLUCOR  ZEDCOR  ZAFCOR STP-POW BKS-COR   F(x)u      Ec   Eo/Ec    MACs
   Fe ka   .9969  1.0000  1.0660  1.0626  1.0935   .9748   .9877  7.1120  2.1091 55.6318

 ELEMENT   K-RAW K-VALUE ELEMWT% OXIDWT% ATOMIC% FORMULA KILOVOL                                       
   Fe ka  .00000  .68095  72.359 100.000  42.857   1.000   15.00                                       
   O                        .000    .000    .000    .000
   O                      27.641   -----  57.143   1.333
   TOTAL:                100.000 100.000 100.000   2.333

[Probeman]

Then I thought to myself: Brian Joy is a very smart guy, so what am I missing?   Maybe there's a problem with the PAP corrections, so I used the magnetite example in CalcZAF.dat with PAP and I get this;

SAMPLE: 6, TOA: 40, ITERATIONS: 3, Z-BAR: 20.96556

 ELEMENT  ABSCOR  FLUCOR  ZEDCOR  ZAFCOR STP-POW BKS-COR   F(x)u      Ec   Eo/Ec    MACs
   Si ka  1.5129  1.0000   .9264  1.4016   .8896  1.0415   .5956  1.8390  8.1566 1948.07
   Fe ka   .9966  1.0000  1.0828  1.0791  1.1103   .9752   .9871  7.1120  2.1091 55.4812
   Mg ka  2.4533   .9999   .9148  2.2442   .8644  1.0583   .3504  1.3050 11.4943 4460.17

 ELEMENT   K-RAW K-VALUE ELEMWT% OXIDWT% ATOMIC% FORMULA KILOVOL                                       
   Si ka -.00008 -.00003   -.005   -.010   -.005   -.001   15.00                                       
   Fe ka 1.35236  .66415  71.671  92.204  42.430   7.929   15.00                                       
   Mg ka  .00069  .00032    .071    .118    .097    .018   15.00                                       
   Mn                       .054    .070    .032    .006
   Ca                       .000    .000    .000    .000
   Ni                       .000    .000    .000    .000
   Al                       .201    .380    .246    .046
   O                       6.899   6.899  14.256   2.664
   Ti                       .012    .020    .008    .002
   O                      20.778   -----  42.935   8.024
   TOTAL:                 99.682  99.682 100.000  18.688


Which is a slightly lower total (using Fe2SiO4 as an Fe standard), but it's still within normal accuracy.

And if I set the excess oxygen to zero and specify Fe3O4 for the cations/oxygens I get this with PAP:

ELEMENT  ABSCOR  FLUCOR  ZEDCOR  ZAFCOR STP-POW BKS-COR   F(x)u      Ec   Eo/Ec    MACs
   Si ka  1.5130  1.0000   .9264  1.4016   .8894  1.0415   .5955  1.8390  8.1566 1947.30
   Fe ka   .9966  1.0000  1.0827  1.0790  1.1102   .9753   .9871  7.1120  2.1091 55.4632
   Mg ka  2.4537   .9999   .9147  2.2443   .8643  1.0584   .3503  1.3050 11.4943 4458.32

 ELEMENT   K-RAW K-VALUE ELEMWT% OXIDWT% ATOMIC% FORMULA KILOVOL                                       
   Si ka -.00008 -.00003   -.005   -.010   -.005   -.001   15.00                                       
   Fe ka 1.35236  .66415  71.663  99.039  42.478   7.929   15.00                                       
   Mg ka  .00069  .00032    .071    .118    .097    .018   15.00                                       
   Mn                       .054    .070    .033    .006
   Ca                       .000    .000    .000    .000
   Ni                       .000    .000    .000    .000
   Al                       .201    .380    .247    .046
   O                        .000    .000    .000    .000
   Ti                       .012    .020    .008    .002
   O                      27.619   -----  57.143  10.667
   TOTAL:                 99.617  99.617 100.000  18.667

All seems to be good.
john

[Brian Joy]

Hi John,

I overlooked some of the options in CalcZAF.  What I was looking for was a way to specify the total number of cations (3) along with the total anion charge (-8) during the matrix correction iterations (which is how I've approached it).  Note that specifying molar Fe:O ratio = 3:4 in magnetite does not necessarily work because other cations can substitute for either Fe2+ or Fe3+ so that the molar ratio of FeO to Fe2O3 is not always 1:1.  Let me dig up an example.

Brian

[Probeman]

Hi John,

I overlooked some of the options in CalcZAF.  What I was looking for was a way to specify the total number of cations (3) along with the total anion charge (-8) during the matrix correction iterations (which is how I've approached it).  Note that specifying molar Fe:O ratio = 3:4 in magnetite does not necessarily work because other cations can substitute for either Fe2+ or Fe3+ so that the molar ratio of FeO to Fe2O3 is not always 1:1.  Let me dig up an example.

Brian

It sounds like you are wanting a mineral normalization for magnetite.  Probe for EPMA has a number of mineral normalization methods based on site occupancy, but not CalcZAF.  You might want to try Julien Allaz's mineral re-normalization web page which is for many minerals:

http://cub.geoloweb.ch/index.php?page=mineral_formula

I've re-coded his php code in VB, but haven't implemented it into PFE yet.

Andrew Locock also posted an oxide normalization spreadsheet here:

http://probesoftware.com/smf/index.php?topic=92.msg3926#msg3926
john

[Brian Joy]

It sounds like you are wanting a mineral normalization for magnetite.  Probe for EPMA has a number of mineral normalization methods based on site occupancy, but not CalcZAF.  You might want to try Julien Allaz's mineral re-normalization web page which is for many minerals:

http://cub.geoloweb.ch/index.php?page=mineral_formula

I've re-coded his php code in VB, but haven't implemented it into PFE yet.

Andrew Locock also posted an oxide normalization spreadsheet here:

http://probesoftware.com/smf/index.php?topic=92.msg3926#msg3926
john

I understand what you're suggesting, but that's not what I'm looking for.  What I'm looking for is a means of adjusting the oxygen content of the analyzed mineral during the matrix correction iterations.  I wanted to compare my solution to the problem with possible other methods.  Let me post an example later.  Perhaps this is dealt with in Probe for EPMA (which of course I don't have).

[Probeman]

It sounds like you are wanting a mineral normalization for magnetite.  Probe for EPMA has a number of mineral normalization methods based on site occupancy, but not CalcZAF.  You might want to try Julien Allaz's mineral re-normalization web page which is for many minerals:
By
http://cub.geoloweb.ch/index.php?page=mineral_formula

I've re-coded his php code in VB, but haven't implemented it into PFE yet.

Andrew Locock also posted an oxide normalization spreadsheet here:

http://probesoftware.com/smf/index.php?topic=92.msg3926#msg3926
john

I understand what you're suggesting, but that's not what I'm looking for.  What I'm looking for is a means of adjusting the oxygen content of the analyzed mineral during the matrix correction iterations.  I wanted to compare my solution to the problem with possible other methods.  Let me post an example later.  Perhaps this is dealt with in Probe for EPMA (which of course I don't have).

Hi Brian,
Probe for EPMA does adjust the oxygen concentration during the matrix correction iteration, for example when performing the halogen equivalence correction as I mentioned previously. PFE also adjusts the concentrations of interfering elements during the spectral interference correction (which again, is iterated in the matrix correction), and for a number of other corrections which are also compositionally dependent.

So this is something I am very interested in.

[Brian Joy]

Below are a couple of magnetite analyses that I’ve processed (PAP/MAC30) by placing constraints on the total number of cations (3) and total anion charge (-8) per formula unit within the matrix correction iterations; I’ve incorporated these constraints within the subroutine that I use to normalize composition during the iterations.  Of course the JEOL software contains no provision for calculating Fe3+/Fe2+ while processing data, and so otherwise I get stuck recalculating after the fact and getting inaccurate results.  Usually I check my work against CalcZAF, but, like I noted, I couldn’t find a way to apply the constraints in the same manner.  In this case the Fe standard is pure Fe3O4 synthesized by Peter Roeder.  The output is a little rough; for Z=26, wt% FeO is listed first and Fe2O3 second.  In analysis 137, which is close to end-member magnetite, note that Fe Ka Z-std is very close to unity.

[Probeman]

Below are a couple of magnetite analyses that I’ve processed (PAP/MAC30) by placing constraints on the total number of cations (3) and total anion charge (-8) per formula unit within the matrix correction iterations; I’ve incorporated these constraints within the subroutine that I use to normalize composition during the iterations.  Of course the JEOL software contains no provision for calculating Fe3+/Fe2+ while processing data, and so otherwise I get stuck recalculating after the fact and getting inaccurate results.  Usually I check my work against CalcZAF, but, like I noted, I couldn’t find a way to apply the constraints in the same manner.  In this case the Fe standard is pure Fe3O4 synthesized by Peter Roeder.  The output is a little rough; for Z=26, wt% FeO is listed first and Fe2O3 second.  In analysis 137, which is close to end-member magnetite, note that Fe Ka Z-std is very close to unity.

Hi Brian,
Let me see if I understand what you are doing.

So you are essentially calculating the excess oxygen (over FeO) by adjusting the oxygen concentration so that the total oxygens for all cations is exactly 4 atoms (-8 charge). And that this is because you have some Cr, Ti, Al, Mn and Mg replacing some Fe?

Have I got it right so far?
john

[Brian Joy]

Hi Brian,
Let me see if I understand what you are doing.

So you are essentially calculating the excess oxygen (over FeO) by adjusting the oxygen concentration so that the total oxygens for all cations is exactly 4 atoms (-8 charge). And that this is because you have some Cr, Ti, Al, Mn and Mg replacing some Fe?

Have I got it right so far?
john

Hi John,

Specifically what I’m doing is normalizing first to a given number of cations specified by the user in a settings file.  I then calculate the total charge on the cations assuming, for instance, that all Fe is divalent.  For this case, the number of Fe3+ per formula unit is equal to the anion charge (corresponding to the number of oxygen ions specified by the user in the settings file) minus the total charge on the normalized cations assuming all Fe is divalent.  I then calculate the mass of oxygen required to balance the charge on the cations (whatever they may be) during the iteration while using an “average” Fe charge determined by the above procedure.  It’s essentially the same procedure used in after-the-fact recalculations, but I’ve simply built it into the matrix correction iterations.  I was afraid that it might hinder convergence, but I haven’t run into problems with this.

The scheme that I’ve implemented is the simplest possible one, and one could envision much more elaborate schemes in which cations are assigned to specific sites during the normalization, as is necessary in the amphiboles, for instance.

[Probeman]

Hi Brian,
Ok, so you're essentially performing a mineral formula normalization during the matrix iteration.  As I said before I do this in some cases- e.g., fluor-phlogopites where by not including the change in oxygen from the halogen equivalence, there is about a 15% effect on the fluorine concentration. But that's a 5 or 6% absolute change in oxygen concentration, and fluorine Ka is strongly absorbed by oxygen...

Can you post an example composition where you don't iterate the matrix correction during the mineral normalization and the same composition where you do re-iterate the matrix correction after the oxygen has been modified by the mineral re-normalization? 

I'd be curious to see the magnitude of the effect.
john

[Probeman]

So I tried a test myself. Here is a typical chromite analysis:

ELEMENT  ABSCOR  FLUCOR  ZEDCOR  ZAFCOR STP-POW BKS-COR   F(x)u      Ec   Eo/Ec    MACs
   Fe ka  1.0235  1.0000  1.1012  1.1271  1.1454   .9614   .9620  7.1120  2.1091 175.945
   Cr ka   .9993   .9530  1.0946  1.0424  1.1351   .9643   .9806  5.9900  2.5042 82.3851
   Ti ka  1.0058   .8822  1.0858   .9634  1.1191   .9703   .9665  4.9670  3.0199 135.744
   Al ka  1.6721   .9995   .9812  1.6398   .9537  1.0288   .5307  1.5600  9.6154 2436.91
   Mn ka   .9973   .9958  1.1182  1.1104  1.1617   .9625   .9853  6.5390  2.2939 64.2572

 ELEMENT   K-RAW K-VALUE ELEMWT% OXIDWT% ATOMIC% FORMULA KILOVOL                                       
   Fe ka  .27180  .18431  20.774  26.725  10.317   1.931   15.00                                       
   Cr ka  .47690  .30522  31.817  46.503  16.972   3.177   15.00                                       
   Ti ka  .00550  .00305    .294    .490    .170    .032   15.00                                       
   Al ka  .10830  .04715   7.731  14.608   7.947   1.488   15.00                                       
   Mn ka  .00200  .00147    .163    .211    .082    .015   15.00                                       
   Mg                      6.290  10.431   7.178   1.344
   V                        .122    .179    .066    .012
   O                       1.080   1.080   1.872    .350
   O                      31.956   -----  55.395  10.371
   TOTAL:                100.228 100.228 100.000  18.721


Note the 1.08 wt% excess oxygen. Now I edit the excess oxygen number to 0 wt% and you can see the changes:

ELEMENT  ABSCOR  FLUCOR  ZEDCOR  ZAFCOR STP-POW BKS-COR   F(x)u      Ec   Eo/Ec    MACs
   Fe ka  1.0239  1.0000  1.0997  1.1259  1.1433   .9618   .9616  7.1120  2.1091 175.497
   Cr ka   .9994   .9529  1.0932  1.0410  1.1330   .9648   .9805  5.9900  2.5042 81.9200
   Ti ka  1.0060   .8817  1.0845   .9618  1.1170   .9708   .9663  4.9670  3.0199 134.939
   Al ka  1.6755   .9995   .9801  1.6413   .9519  1.0297   .5296  1.5600  9.6154 2417.43
   Mn ka   .9974   .9957  1.1167  1.1090  1.1596   .9629   .9852  6.5390  2.2939 63.9071

 ELEMENT   K-RAW K-VALUE ELEMWT% OXIDWT% ATOMIC% FORMULA KILOVOL                                       
   Fe ka  .27180  .18431  20.752  26.698  10.510   1.931   15.00                                       
   Cr ka  .47690  .30522  31.774  46.441  17.283   3.175   15.00                                       
   Ti ka  .00550  .00305    .293    .489    .173    .032   15.00                                       
   Al ka  .10830  .04715   7.738  14.621   8.111   1.490   15.00                                       
   Mn ka  .00200  .00147    .163    .210    .084    .015   15.00                                       
   Mg                      6.290  10.431   7.319   1.345
   V                        .122    .179    .068    .012
   O                        .000    .000    .000    .000
   O                      31.936   -----  56.452  10.371
   TOTAL:                 99.069  99.069 100.000  18.371


Obviously the total is now low by some 1% or so, but the absolute changes in the measured elements due to the reduced excess oxygen are pretty small.  On the order of a few hundred PPM for each element.

That is generally smaller than the measurement precision, so is it worth doing? I have to ask myself...

I've attached a CalcZAF input file for this below.  Note that there are two samples in it. The first calculating everything elemental and the second calculating everything as oxides (and excess oxygen).  The 2nd sample in the file is the example I used above.

[Brian Joy]

Hi John,

Yes, I agree that the effect is small unless Fe2O3 is present in large quantity.  This is why I chose magnetite as an example.  As far as silicates, possible problematic cases would be minerals such as aegirine-augite and riebeckite, especially if the latter is fluorine-rich.

Below are the same two magnetite analyses as before with all Fe assumed to be present as FeO during the matrix corrections.  If I recalculate for Fe2O3 after the fact, then for the first analysis I get 30.62 wt% FeO, 67.14 wt% Fe2O3, and total = 98.71 wt%.  For the second analysis, I get 31.68 wt% FeO, 26.16 wt% Fe2O3, and total = 100.04 wt%; wt% Cr2O3 is 31.43 instead of 31.57.

 

[Probeman]

Yes, I agree that the effect is small unless Fe2O3 is present in large quantity.  This is why I chose magnetite as an example.  As far as silicates, possible problematic cases would be minerals such as aegirine-augite and riebeckite, especially if the latter is fluorine-rich.

Hi Brian,
You're showing a magnetite and a chromite in the image of the FORTRAN output.  Did you post the wrong image?

If I run a magnetite both ways I get this for magnetite with excess oxygen of 6.881 wt%:

ELEMENT  ABSCOR  FLUCOR  ZEDCOR  ZAFCOR STP-POW BKS-COR   F(x)u      Ec   Eo/Ec    MACs
   Fe ka   .9970  1.0000  1.0666  1.0634  1.0944   .9746   .9875  7.1120  2.1091 56.5165
   Cr ka  1.0008   .7464  1.0623   .7935  1.0840   .9800   .9791  5.9900  2.5042 88.8413
   Ti ka  1.0088   .8934  1.0558   .9516  1.0680   .9885   .9635  4.9670  3.0199 148.135
   Al ka  1.8553   .9996   .9587  1.7779   .9077  1.0561   .4783  1.5600  9.6154 2935.69
   Mn ka   .9982   .9707  1.0841  1.0505  1.1097   .9769   .9844  6.5390  2.2939 68.3900

 ELEMENT   K-RAW K-VALUE ELEMWT% OXIDWT% ATOMIC% FORMULA KILOVOL                                       
   Fe ka 1.00000  .67813  72.110  92.770  42.354   7.907   15.00                                       
   Cr ka  .00305  .00195    .155    .226    .098    .018   15.00                                       
   Ti ka  .00013  .00007    .007    .011    .005    .001   15.00                                       
   Al ka  .00279  .00121    .216    .408    .263    .049   15.00                                       
   Mn ka  .00082  .00060    .063    .082    .038    .007   15.00                                       
   Mg                       .072    .119    .097    .018
   O                       6.881   6.881  14.107   2.633
   O                      20.993   -----  43.039   8.035
   TOTAL:                100.498 100.498 100.000  18.668

And if I reduce the excess oxygen to 5.881 wt% (which I think is larger than any change you will see in the oxygen concentration from your charge balancing), the results are:

ELEMENT  ABSCOR  FLUCOR  ZEDCOR  ZAFCOR STP-POW BKS-COR   F(x)u      Ec   Eo/Ec    MACs
   Fe ka   .9971  1.0000  1.0648  1.0617  1.0920   .9752   .9875  7.1120  2.1091 56.2059
   Cr ka  1.0009   .7455  1.0607   .7915  1.0815   .9807   .9790  5.9900  2.5042 88.3326
   Ti ka  1.0090   .8929  1.0543   .9499  1.0656   .9893   .9633  4.9670  3.0199 147.256
   Al ka  1.8604   .9996   .9575  1.7806   .9056  1.0573   .4770  1.5600  9.6154 2915.34
   Mn ka   .9983   .9706  1.0824  1.0488  1.1073   .9775   .9843  6.5390  2.2939 68.0066

 ELEMENT   K-RAW K-VALUE ELEMWT% OXIDWT% ATOMIC% FORMULA KILOVOL                                       
   Fe ka 1.00000  .67813  71.999  92.626  43.232   7.907   15.00                                       
   Cr ka  .00305  .00195    .154    .226    .100    .018   15.00                                       
   Ti ka  .00013  .00007    .007    .011    .005    .001   15.00                                       
   Al ka  .00279  .00121    .216    .409    .269    .049   15.00                                       
   Mn ka  .00082  .00060    .063    .082    .039    .007   15.00                                       
   Mg                       .072    .119    .099    .018
   O                       5.881   5.881  12.326   2.254
   O                      20.961   -----  43.932   8.035
   TOTAL:                 99.354  99.354 100.000  18.289


And again, it's a change on the order of a few hundred PPM for Fe.   If you leave all the excess oxygen out (6.9 wt%), of course you will get a significantly different matrix correction.  But you should be comparing the difference (as I'm trying to show above) between including the mineral normalization in the matrix correction or not.

My point being that whether or not we are performing a mineral normalization for charge balance in the matrix correction, we should *always* be including the oxygen from Fe2O3 in the matrix correction.

I've attached my magnetite example file for CalcZAF below.