The FIGMAS MgO, Al2O3, MgAl2O4 challenge!

[Probeman]

Andrew Ducharme and I re-ran the MgO, Al2O3 and MgAl2O4 FIGMAS mount from Will Nachlas again this last weekend (after a blown filament!), and we got very similar accuracy results to our previous run earlier this month:

https://smf.probesoftware.com/index.php?topic=1442.msg13896#msg13896

I am starting a new topic specifically on the FIGMAS mount and calling it the "FIGMAS Accuracy Challenge" , because a number of labs have the same mount and we would like to compare results with them. Especially comparing the quant data at different accelerating voltages (15, 20 and 25 keV). Because we are seeing some trends and we have to wonder if these are due to the matrix corrections or perhaps a high voltage calibration issue?  We suspect the latter...

Once again, the new results are all within ~1% relative accuracy or better, but I am interested if other labs are seeing these same trends with respect to beam energy.  We used 20 nA for all measurements, but it might worth using higher beam currents if your dead times are properly calibrated and your picoammter response is linear (our SX100 has a non-linear response when going from below 40 nA to above 40 nA so we can't test this at the moment).

So if you have the FIGMAS mount from Will Nachlas containing the MgO, Al2O3 and MgAl2O4 materials you can participate. If you don't have this mount Will might have some extra mounts, though these high purity materials can be found from commercial crystal suppliers such as Princeton Scientific Corporation, which is where they were originally obtained. 

First of all, we are interested to see if other labs can attain similar levels of accuracy, when measuring MgAl2O4 for Mg, Al and O using MgO, Al2O3 and MgO as primary standards respectively, by assuming formula stoichiometry in the MgAl2O4.

And second, do you also see this accuracy trend with respect to beam energy that we are seeing?

Here are the most recent data from Andrew and I for all matrix corrections at 15 keV:

Summary of All Calculated (averaged) Matrix Corrections:
St 3100 Set   7 MgAl2O4 FIGMAS
LINEMU   Henke (LBL, 1985) < 10KeV / CITZMU > 10KeV

Elemental Weight Percents:
ELEM:       Mg      Al       O   TOTAL
     1  17.094  38.080  45.078 100.252   Armstrong/Brown/Scott-Love (prZ)
     2  16.981  38.384  44.644 100.009   Philibert/Duncumb-Reed
     3  17.051  38.343  44.785 100.179   Heinrich/Duncumb-Reed
     4  17.099  38.254  45.108 100.461   Love-Scott I
     5  17.094  38.073  45.161 100.329   Love-Scott II
     6  17.026  37.867  44.839  99.732   Packwood Phi(prZ) (EPQ-91)
     7  17.266  38.220  45.453 100.939   Bastin (original) (prZ)
     8  17.143  38.594  45.358 101.095   Bastin PROZA Phi (prZ) (EPQ-91)
     9  17.125  38.471  45.322 100.919   Pouchou and Pichoir-Full (PAP)
    10  17.098  38.286  45.138 100.522   Pouchou and Pichoir-Simplified (XPP)
    11  17.090  38.093  45.086 100.269   PAP/Donovan and Moy BSC/BKS (prZ)

AVER:   17.097  38.242  45.088 100.428
SDEV:     .072    .207    .249    .418
SERR:     .022    .062    .075

MIN:    16.981  37.867  44.644  99.732
MAX:    17.266  38.594  45.453 101.095

Percent Variances:
ELEM:       Mg      Al       O
PUBL:   17.084  37.931  44.985
STDS:     3012    3013    3012

ELEM:       Mg      Al       O
     1     .06     .39     .21           Armstrong/Brown/Scott-Love (prZ)
     2    -.60    1.19    -.76           Philibert/Duncumb-Reed
     3    -.19    1.09    -.45           Heinrich/Duncumb-Reed
     4     .09     .85     .27           Love-Scott I
     5     .06     .38     .39           Love-Scott II
     6    -.34    -.17    -.33           Packwood Phi(prZ) (EPQ-91)
     7    1.06     .76    1.04           Bastin (original) (prZ)
     8     .35    1.75     .83           Bastin PROZA Phi (prZ) (EPQ-91)
     9     .24    1.42     .75           Pouchou and Pichoir-Full (PAP)
    10     .08     .93     .34           Pouchou and Pichoir-Simplified (XPP)
    11     .04     .43     .22           PAP/Donovan and Moy BSC/BKS (prZ)

AVER:      .08     .82     .23        
SDEV:      .42     .54     .55        
SERR:      .13     .16     .17        

MIN:      -.60    -.17    -.76        
MAX:      1.06    1.75    1.04        

And here for 20 keV:

Summary of All Calculated (averaged) Matrix Corrections:
St 3100 Set   8 MgAl2O4 FIGMAS
LINEMU   Henke (LBL, 1985) < 10KeV / CITZMU > 10KeV

Elemental Weight Percents:
ELEM:       Mg      Al       O   TOTAL
     1  17.217  37.877  45.259 100.353   Armstrong/Brown/Scott-Love (prZ)
     2  17.053  37.882  44.765  99.701   Philibert/Duncumb-Reed
     3  17.129  37.864  44.881  99.874   Heinrich/Duncumb-Reed
     4  17.232  38.381  45.245 100.858   Love-Scott I
     5  17.252  38.125  45.532 100.909   Love-Scott II
     6  17.140  37.550  45.026  99.716   Packwood Phi(prZ) (EPQ-91)
     7  17.382  38.167  45.647 101.195   Bastin (original) (prZ)
     8  17.276  38.520  45.606 101.402   Bastin PROZA Phi (prZ) (EPQ-91)
     9  17.267  38.523  45.619 101.408   Pouchou and Pichoir-Full (PAP)
    10  17.224  38.134  45.401 100.758   Pouchou and Pichoir-Simplified (XPP)
    11  17.216  37.896  45.273 100.386   PAP/Donovan and Moy BSC/BKS (prZ)

AVER:   17.217  38.083  45.296 100.596
SDEV:     .087    .306    .304    .637
SERR:     .026    .092    .092

MIN:    17.053  37.550  44.765  99.701
MAX:    17.382  38.523  45.647 101.408

Percent Variances:
ELEM:       Mg      Al       O
PUBL:   17.084  37.931  44.985
STDS:     3012    3013    3012

ELEM:       Mg      Al       O
     1     .78    -.14     .61           Armstrong/Brown/Scott-Love (prZ)
     2    -.18    -.13    -.49           Philibert/Duncumb-Reed
     3     .27    -.18    -.23           Heinrich/Duncumb-Reed
     4     .87    1.19     .58           Love-Scott I
     5     .98     .51    1.22           Love-Scott II
     6     .33   -1.01     .09           Packwood Phi(prZ) (EPQ-91)
     7    1.74     .62    1.47           Bastin (original) (prZ)
     8    1.12    1.55    1.38           Bastin PROZA Phi (prZ) (EPQ-91)
     9    1.07    1.56    1.41           Pouchou and Pichoir-Full (PAP)
    10     .82     .53     .92           Pouchou and Pichoir-Simplified (XPP)
    11     .77    -.09     .64           PAP/Donovan and Moy BSC/BKS (prZ)

AVER:      .78     .40     .69        
SDEV:      .51     .81     .67        
SERR:      .15     .24     .20        

MIN:      -.18   -1.01    -.49        
MAX:      1.74    1.56    1.47        

and here for 25 keV:

Summary of All Calculated (averaged) Matrix Corrections:
St 3100 Set   9 MgAl2O4 FIGMAS
LINEMU   Henke (LBL, 1985) < 10KeV / CITZMU > 10KeV

Elemental Weight Percents:
ELEM:       Mg      Al       O   TOTAL
     1  17.274  37.569  45.370 100.213   Armstrong/Brown/Scott-Love (prZ)
     2  17.086  37.271  44.950  99.307   Philibert/Duncumb-Reed
     3  17.158  37.259  45.085  99.502   Heinrich/Duncumb-Reed
     4  17.277  38.318  45.216 100.810   Love-Scott I
     5  17.346  38.192  45.798 101.336   Love-Scott II
     6  17.204  37.208  45.243  99.655   Packwood Phi(prZ) (EPQ-91)
     7  17.434  38.005  45.804 101.243   Bastin (original) (prZ)
     8  17.344  38.316  45.776 101.436   Bastin PROZA Phi (prZ) (EPQ-91)
     9  17.340  38.493  45.809 101.643   Pouchou and Pichoir-Full (PAP)
    10  17.294  37.921  45.633 100.848   Pouchou and Pichoir-Simplified (XPP)
    11  17.275  37.598  45.392 100.265   PAP/Donovan and Moy BSC/BKS (prZ)

AVER:   17.276  37.832  45.461 100.569
SDEV:     .097    .472    .317    .829
SERR:     .029    .142    .096

MIN:    17.086  37.208  44.950  99.307
MAX:    17.434  38.493  45.809 101.643

Percent Variances:
ELEM:       Mg      Al       O
PUBL:   17.084  37.931  44.985
STDS:     3012    3013    3012

ELEM:       Mg      Al       O
     1    1.11    -.95     .85           Armstrong/Brown/Scott-Love (prZ)
     2     .01   -1.74    -.08           Philibert/Duncumb-Reed
     3     .43   -1.77     .22           Heinrich/Duncumb-Reed
     4    1.13    1.02     .51           Love-Scott I
     5    1.54     .69    1.81           Love-Scott II
     6     .70   -1.91     .57           Packwood Phi(prZ) (EPQ-91)
     7    2.05     .19    1.82           Bastin (original) (prZ)
     8    1.52    1.01    1.76           Bastin PROZA Phi (prZ) (EPQ-91)
     9    1.50    1.48    1.83           Pouchou and Pichoir-Full (PAP)
    10    1.23    -.03    1.44           Pouchou and Pichoir-Simplified (XPP)
    11    1.12    -.88     .90           PAP/Donovan and Moy BSC/BKS (prZ)

AVER:     1.12    -.26    1.06        
SDEV:      .57    1.25     .70        
SERR:      .17     .38     .21        

MIN:       .01   -1.91    -.08        
MAX:      2.05    1.48    1.83       

Yes, things get a bit worse at 25 keV but that isn't too surprising given that the matrix corrections get much larger at these higher beam energies. But more interesting I think is that there seems to be a trend in the data as a function of energy. I will reveal these plots in the next post...

oh, and for those that are members of the forum I placed our PFE MDB data file from our first run below as an attachment, if anyone wants to look at the config and data.

[Probeman]

Continued from above...

Here are the quant results as function of keV, first Mg:



Al:



and O:



These deviations look pretty significant, but these relative errors are around 1% or less.  Here are the FIGMAS MgO, Al2O3, MgAl2O4 challenge steps:

1. Perform the constant k-ratio calibration method here on your instrument to confirm your dead time constants, picoammeter linearity and k-ratio agreement between your spectrometers:

https://smf.probesoftware.com/index.php?topic=1466.msg11102#msg11102

If your dead time constants are not providing accurate corrections for dead time then you will need to edit them. Of course above 30 to 40 kcps the counting electronics responds non-linearly, so those without Probe for EPMA and the logarithmic dead time expression will need to limit their beam current.  But at least check your dead time calibrations please.

Donovan, John J., et al. "a new method for dead time calibration and a new expression for correction of WDS Intensities for microanalysis." Microscopy and Microanalysis 29.3 (2023): 1096-1110.

2. Perform Bragg order k-ratio tests to check your spectrometer alignments:

https://smf.probesoftware.com/index.php?topic=1739.0

Your WDS spectrometers should be able to produce similar k-ratios within statistics for all Bragg orders!  If not then your spectrometers need an alignment calibration.  Call an engineer if you do not know how to calibrate your spectrometer mechanicals!

3. Then perform careful PHA adjustments where the gain/bias are set high enough so that at the highest count rates you will be measuring, the PHA peaks are well above the baseline level. At lower count rates your PHA peaks will shift to the right, so as long as you are running in Integral mode, you will be OK.

Hopefully you already know how to do this because you had to do it for the constant k-ratio tests above:

https://smf.probesoftware.com/index.php?topic=1466.msg11636#msg11636

4. Now acquire wavescan samples on all three materials (MgO, Al2O3 and MgAl2O4) and make sure that your background positions are measuring the actual backgrounds for all three samples (extend the range of your wavescans as much as necessary to obtain the actual backgrounds):

https://smf.probesoftware.com/index.php?topic=68.msg13913#msg13913

5. And only then, perform quantitative measurements of your MgAl2O4 using MgO and Al2O3 (MgO as the oxygen standard) as standards and see if you can beat our accuracy as seen here:

https://smf.probesoftware.com/index.php?topic=1442.msg13896#msg13896

Please acquire the Mg, Al and O data at 15, 20 and 25 keV.  Yes, there may be a small APF correction for the oxygen peak shape change between MgO and MgAl2O4 but we can deal with that later.  Andrew Ducharme and I are working on the O Ka peak shape APFs...

Then plot your quant results as a function of keV, with the published (assumed stoichiometric values) on the plots. The FIGMAS challenge has been issued!

[sem-geologist]

These deviations look pretty significant, but these relative errors are around 1% or less

Would it be not nice to reduce the errors down to 0.5%? It would be absolutely nice, especially for applications like thermobarometry (i.e. based alone on amphibole chemistry, without boundaries and thermo-chemical ion exchange mineral pairs (the 2ndary fluorescence questionable stuff)).

Anyway, I think I have a plausible hypothesis why you get this trend vs HV. You are measuring background positions across absorption edge, and especially using exponential background fit will "cut under peak into background". The absorption edge background differences in particularly will get enlarged when differing the HV.

I would go with your challenge right away (especially that I also participated in that FIGMAS study, and have a mount at hand, and would try to show that dependency. However at the moment I am quite handicapped to do that, as our SX100 needs some custom electronics to unlock 20-25kV range of our new HV generator. But I am kept busy with our SXFive FE trying to iron out its major design flaws (practically it is not operational at the moment).

picoammeter linearity

Oh this bit me today. We have a client doing some Venetian like glassware. The analyses went pretty well in all previous sessions. Today it just started randomly do poor (low totals ~95-96%). There were many simultaneous trouble sources, i.e. due to cold winter room humidity fallen down to 10%RH, and all kind of other de-calibrations started to show off. Where recalibration kindly fixed them, but this random inaccuracy... We measure these glassware with multiconditions, 5nA for Si, Ca and Na with 0-time intercept function (peaksight implementation for volatile drift), and 15nA for rest. So randomly Si (in glasses 70-60%) and Ca was undermeasured, and randomly was within accuracy of secondary reference material. 

So I had remind myself that Cameca SX* EPMA has 5 ranges: <0.5nA, 0.5-5nA, 5-50nA, 50-500nA and >500nA. These measuring ranges correspond to different negative feedback resistors to OPAMP which sense the beam current. But however, as I found out today, the ranges are not strict at these boundaries. And even measuring 5.1 nA the feedback of 0.5-5nA can be used, or 4.9 nA value could be from measurement using the 5-50nA configuration of picoamperometer! It depends from which side C1,C2 values iterate to the final value. The bad part is that while 50nA and 500nA boundaries can be theoretically calibrated, as there are physical potentiometers, the 5nA boundary can't be calibrated as there is no potentiometers for that. One more caveat is freshly cleaned regulator aperture - at low beam currents it will go in much larger iterative steps (of C1,C2), and much easily gets into 0.5nA-5nA measuring range, when asking for 5nA. Even requesting 5.5nA in about 25% got measured with 0.5nA-5nA. In contrast to that, when our regulation aperture was dirty (a month ago) - it was setting 5nA within +/- 0.05nA and using only 5-50nA measurement range.

[Probeman]

These deviations look pretty significant, but these relative errors are around 1% or less

Would it be not nice to reduce the errors down to 0.5%? It would be absolutely nice, especially for applications like thermobarometry (i.e. based alone on amphibole chemistry, without boundaries and thermo-chemical ion exchange mineral pairs (the 2ndary fluorescence questionable stuff)).

Indeed, that is exactly the goal!

Anyway, I think I have a plausible hypothesis why you get this trend vs HV. You are measuring background positions across absorption edge, and especially using exponential background fit will "cut under peak into background". The absorption edge background differences in particularly will get enlarged when differing the HV.

It's possible I guess, but these are very large P/B ratios:

PKBG:  283.52  461.05  186.36    <-- 15 keV
PKBG:  305.99  504.89  192.46    <-- 20 keV
PKBG:  323.13  524.71  183.03    <-- 25 keV

So the entire background correction here for Mg and Al are around 0.2 to 0.3 % relative to the peaks, and the absorption edge effect is surely only a fraction of that.

I guess I need to run MAN background corrections on (though MAN is a pain for O Ka!) and see what we get...

I would go with your challenge right away (especially that I also participated in that FIGMAS study, and have a mount at hand, and would try to show that dependency.

We look forward to your results!

[Probeman]

In addition to the synthetic MgO, Al2O3 and MgAl2O4 acquisitions we did last weekend at 20 nA, we also did acquisitions at 60 nA just to see how well the dead time correction handled things. Note that the Al Ka on LTAP at 25 keV on Al2O3 was over 150 kcps at 60 nA!

Here is an example at 20 keV and 60 nA:

St 3100 Set   5 MgAl2O4 FIGMAS, Results in Elemental Weight Percents
 
ELEM:       Mg      Al       O
BGDS:      EXP     EXP     EXP
TIME:    60.00   60.00   60.00
BEAM:    60.04   60.04   60.04

ELEM:       Mg      Al       O   SUM  
    83  17.178  37.522  45.078  99.778
    84  17.178  37.583  45.079  99.840
    85  17.168  37.583  45.158  99.909
    86  17.129  37.549  45.074  99.751
    87  17.134  37.546  45.040  99.721

AVER:   17.158  37.557  45.086  99.800
SDEV:     .024    .026    .043    .075
SERR:     .011    .012    .019
%RSD:      .14     .07     .10

PUBL:   17.084  37.931  44.985 100.000
%VAR:      .43    -.99     .22
DIFF:     .074   -.374    .100
STDS:     3012    3013    3012

As you can see our variances from the "published" values are still around 1% or less relative accuracy. Here are the full data sets, first Mg Ka:



Now Al Ka:



And O Ka:



We had to separate our 20 nA measurements from our 60 nA measurements because our picoammeter has a serious "glitch" going from below 50 nA to above 50 nA (around 3 to 4%) as can be seen here:

Drift array standard intensities (cps/1nA) (background corrected):
ELMXRY:    mg ka   al ka    o ka
MOTCRY:  1   TAP 2  LTAP 4   PC1
STDASS:     3012    3013    3012
STDVIR:        0       0       0
          612.37 1836.39  276.30        <-- 20 nA
          592.72 1779.64  270.39        <-- 60 nA
          615.27 1827.84  277.01        <-- 20 nA
          595.01 1780.23  271.32        <-- 60 nA

[Probeman]

These deviations look pretty significant, but these relative errors are around 1% or less

Would it be not nice to reduce the errors down to 0.5%? It would be absolutely nice, especially for applications like thermobarometry (i.e. based alone on amphibole chemistry, without boundaries and thermo-chemical ion exchange mineral pairs (the 2ndary fluorescence questionable stuff)).

Well, it's a good question... 

Here's the answer. From the output below, we can see that we are now better than 0.5 % relative accuracy error!  How is this possible when previously we were only able to attain ~1% relative error?

For historical reasons the default matrix correction in Probe for EPMA is the Armstrong/Brown/Packwood pr(z), but one can specify any of the 11 matrix corrections as the default if desired, in the Probewin.ini file.  Likewise, the default mass absorption coefficient table is the Henke (LBL, 1985) + CITZMU (CalTech) for x-rays > 10 keV. Though this default can also be changed in the Probewin.ini file.

Personally I have found that the Armstrong/Brown/Packwood pr(z) matrix correction is probably the best option for oxide and silicate quantification, only surpassed by the new PAP/Donovan and Moy pr(z) matrix correction which offers a Z based backscatter correction for materials with large differences in average Z:
Donovan, John, et al. "An improved average atomic number calculation for estimating backscatter and continuum production in compounds." Microscopy and Microanalysis 29.4 (2023): 1436-1449.

Since in these MgO, Al2O3 and MgAl2O4 materials the Z corrections are pretty insignificant, and since the main accuracy issue we're dealing with (once the dead time and spectrometer alignment issues are calibrated) is the matrix absorption correction, I thought: why not use the best mass absorption coefficients we have, and those would be the Chantler FFAST MACs from NIST (2005). Check these new plots here starting with the Mg data:







Now the FIGMAS challenge has been upped! Can anyone beat this 0.5% accuracy with these materials on your own instruments?  🙂  If you can't, what is the problem?  🙁

Once again, please check the end of this post (linked below) for links on how to calibrate your detector dead times and spectrometer alignments. Then be sure to properly set your PHA peaks completely above the baseline at the highest count rate and use differential mode. The background correction isn't so important for major elements, but be sure anyway to check your off-peak background positions using a wide enough scan as described in the post linked here!

https://smf.probesoftware.com/index.php?topic=1823.msg13918#msg13918

Here are the actual accuracy variances for the first half of the MgAl2O4 data shown above:

15 keV:

Summary of All Calculated (averaged) Matrix Corrections:
St 3100 Set   1 MgAl2O4 FIGMAS
FFAST    Chantler (NIST v 2.1, 2005)

Elemental Weight Percents:
ELEM:       Mg      Al       O   TOTAL
     1  17.098  38.052  44.764  99.914   Armstrong/Brown/Scott-Love (prZ)
     2  16.988  38.376  44.366  99.731   Philibert/Duncumb-Reed
     3  17.057  38.334  44.505  99.897   Heinrich/Duncumb-Reed
     4  17.102  38.216  44.811 100.129   Love-Scott I
     5  17.097  38.037  44.820  99.954   Love-Scott II
     6  17.029  37.827  44.525  99.381   Packwood Phi(prZ) (EPQ-91)
     7  17.269  38.186  45.158 100.613   Bastin (original) (prZ)
     8  17.141  38.565  45.037 100.743   Bastin PROZA Phi (prZ) (EPQ-91)
     9  17.124  38.435  44.992 100.551   Pouchou and Pichoir-Full (PAP)
    10  17.099  38.256  44.821 100.176   Pouchou and Pichoir-Simplified (XPP)
    11  17.095  38.064  44.771  99.930   PAP/Donovan and Moy BSC/BKS (prZ)

AVER:   17.100  38.214  44.779 100.093
SDEV:     .071    .211    .239    .408
SERR:     .021    .064    .072

MIN:    16.988  37.827  44.366  99.381
MAX:    17.269  38.565  45.158 100.743

Percent Variances:
ELEM:       Mg      Al       O
PUBL:   17.084  37.931  44.985
STDS:     3012    3013    3012

ELEM:       Mg      Al       O
     1     .08     .32    -.49           Armstrong/Brown/Scott-Love (prZ)
     2    -.56    1.17   -1.38           Philibert/Duncumb-Reed
     3    -.16    1.06   -1.07           Heinrich/Duncumb-Reed
     4     .10     .75    -.39           Love-Scott I
     5     .08     .28    -.37           Love-Scott II
     6    -.32    -.27   -1.02           Packwood Phi(prZ) (EPQ-91)
     7    1.08     .67     .38           Bastin (original) (prZ)
     8     .33    1.67     .12           Bastin PROZA Phi (prZ) (EPQ-91)
     9     .23    1.33     .02           Pouchou and Pichoir-Full (PAP)
    10     .09     .86    -.37           Pouchou and Pichoir-Simplified (XPP)
    11     .06     .35    -.48           PAP/Donovan and Moy BSC/BKS (prZ)

AVER:      .09     .75    -.46        
SDEV:      .41     .56     .53        
SERR:      .12     .17     .16        

MIN:      -.56    -.27   -1.38        
MAX:      1.08    1.67     .38        

20 keV:

Summary of All Calculated (averaged) Matrix Corrections:
St 3100 Set   2 MgAl2O4 FIGMAS
FFAST    Chantler (NIST v 2.1, 2005)

Elemental Weight Percents:
ELEM:       Mg      Al       O   TOTAL
     1  17.101  37.885  45.021 100.007   Armstrong/Brown/Scott-Love (prZ)
     2  16.943  37.911  44.512  99.366   Philibert/Duncumb-Reed
     3  17.012  37.884  44.648  99.544   Heinrich/Duncumb-Reed
     4  17.114  38.395  45.092 100.601   Love-Scott I
     5  17.125  38.122  45.286 100.534   Love-Scott II
     6  17.020  37.529  44.784  99.333   Packwood Phi(prZ) (EPQ-91)
     7  17.268  38.182  45.421 100.871   Bastin (original) (prZ)
     8  17.151  38.539  45.376 101.066   Bastin PROZA Phi (prZ) (EPQ-91)
     9  17.141  38.533  45.388 101.062   Pouchou and Pichoir-Full (PAP)
    10  17.100  38.145  45.158 100.403   Pouchou and Pichoir-Simplified (XPP)
    11  17.100  37.905  45.034 100.039   PAP/Donovan and Moy BSC/BKS (prZ)

AVER:   17.098  38.094  45.065 100.257
SDEV:     .085    .311    .308    .645
SERR:     .026    .094    .093

MIN:    16.943  37.529  44.512  99.333
MAX:    17.268  38.539  45.421 101.066

Percent Variances:
ELEM:       Mg      Al       O
PUBL:   17.084  37.931  44.985
STDS:     3012    3013    3012

ELEM:       Mg      Al       O
     1     .10    -.12     .08           Armstrong/Brown/Scott-Love (prZ)
     2    -.83    -.05   -1.05           Philibert/Duncumb-Reed
     3    -.42    -.13    -.75           Heinrich/Duncumb-Reed
     4     .18    1.22     .24           Love-Scott I
     5     .24     .50     .67           Love-Scott II
     6    -.38   -1.06    -.45           Packwood Phi(prZ) (EPQ-91)
     7    1.08     .66     .97           Bastin (original) (prZ)
     8     .39    1.60     .87           Bastin PROZA Phi (prZ) (EPQ-91)
     9     .34    1.59     .89           Pouchou and Pichoir-Full (PAP)
    10     .10     .56     .38           Pouchou and Pichoir-Simplified (XPP)
    11     .10    -.07     .11           PAP/Donovan and Moy BSC/BKS (prZ)

AVER:      .08     .43     .18        
SDEV:      .50     .82     .68        
SERR:      .15     .25     .21        

MIN:      -.83   -1.06   -1.05        
MAX:      1.08    1.60     .97        

25 keV:

Summary of All Calculated (averaged) Matrix Corrections:
St 3100 Set   3 MgAl2O4 FIGMAS
FFAST    Chantler (NIST v 2.1, 2005)

Elemental Weight Percents:
ELEM:       Mg      Al       O   TOTAL
     1  17.195  37.794  45.268 100.258   Armstrong/Brown/Scott-Love (prZ)
     2  17.003  37.507  44.780  99.290   Philibert/Duncumb-Reed
     3  17.072  37.494  44.913  99.479   Heinrich/Duncumb-Reed
     4  17.204  38.586  45.210 101.000   Love-Scott I
     5  17.260  38.421  45.694 101.376   Love-Scott II
     6  17.117  37.408  45.109  99.634   Packwood Phi(prZ) (EPQ-91)
     7  17.356  38.241  45.710 101.306   Bastin (original) (prZ)
     8  17.260  38.558  45.667 101.485   Bastin PROZA Phi (prZ) (EPQ-91)
     9  17.256  38.740  45.715 101.711   Pouchou and Pichoir-Full (PAP)
    10  17.211  38.148  45.496 100.854   Pouchou and Pichoir-Simplified (XPP)
    11  17.196  37.824  45.289 100.308   PAP/Donovan and Moy BSC/BKS (prZ)

AVER:   17.194  38.065  45.350 100.609
SDEV:     .098    .483    .333    .863
SERR:     .030    .146    .100

MIN:    17.003  37.408  44.780  99.290
MAX:    17.356  38.740  45.715 101.711

Percent Variances:
ELEM:       Mg      Al       O
PUBL:   17.084  37.931  44.985
STDS:     3012    3013    3012

ELEM:       Mg      Al       O
     1     .65    -.36     .63           Armstrong/Brown/Scott-Love (prZ)
     2    -.48   -1.12    -.46           Philibert/Duncumb-Reed
     3    -.07   -1.15    -.16           Heinrich/Duncumb-Reed
     4     .70    1.73     .50           Love-Scott I
     5    1.03    1.29    1.58           Love-Scott II
     6     .19   -1.38     .28           Packwood Phi(prZ) (EPQ-91)
     7    1.59     .82    1.61           Bastin (original) (prZ)
     8    1.03    1.65    1.51           Bastin PROZA Phi (prZ) (EPQ-91)
     9    1.01    2.13    1.62           Pouchou and Pichoir-Full (PAP)
    10     .74     .57    1.14           Pouchou and Pichoir-Simplified (XPP)
    11     .66    -.28     .67           PAP/Donovan and Moy BSC/BKS (prZ)

AVER:      .64     .35     .81        
SDEV:      .58    1.27     .74        
SERR:      .17     .38     .22        

MIN:      -.48   -1.38    -.46        
MAX:      1.59    2.13    1.62        

Remember, this is extrapolating from synthetic MgO and Al2O3 to MgAl2O4 at 15, 20 and 25 keV where the matrix corrections are 30 to 140%!

How well does your instrument perform on this FIGMAS challenge?

[Probeman]

I should probably mention that all the data from the FIGMAS MgO, Al2O3 and MgAl2O4 samples shown in this topic were acquired using an 8 um beam diameter.  Even though I would consider all three materials quite refractory, I wanted to be sure that there was no damage to the sample coat to eliminate that variable.

Here we graphically compare the data processed using both Henke and FFAST MACs, and I have to say, it's not as obvious as looking at the averages in the post above, but changing the MAC should only push the concentrations one way or another, and the fact that the FFAST MACs push the MgAl2O4 concentrations closer to their "published" (actually assumed stoichiometry), is worth noting.  But here we go, first Mg Ka at 20 nA:



and Al Ka:



and O Ka:



Again, the FFAST accuracy extrapolating from MgO and Al2O3 to MgAl2O4 is pretty darn good even at 25 keV:

St 3100 Set   3 MgAl2O4 FIGMAS
TakeOff = 40.0  KiloVolt = 25.0  Beam Current = 20.0  Beam Size =    8
(Magnification (analytical) =  20000),        Beam Mode = Analog  Spot
(Magnification (default) =      600, Magnification (imaging) =    200)
Image Shift (X,Y):                                         .00,    .00

Univ Wisconsin
Number of Data Lines:   5             Number of 'Good' Data Lines:   5
First/Last Date-Time: 02/07/2026 02:23:04 PM to 02/07/2026 02:31:53 PM
WARNING- Using Empirical Mass Absorption Coefficients

Average Total Oxygen:         .000     Average Total Weight%:  100.185
Average Calculated Oxygen:    .000     Average Atomic Number:   10.564
Average Excess Oxygen:        .000     Average Atomic Weight:   20.292
Average ZAF Iteration:        8.00     Average Quant Iterate:     2.00

St 3100 Set   3 MgAl2O4 FIGMAS, Results in Elemental Weight Percents
 
ELEM:       Mg      Al       O
BGDS:      LIN     LIN     LIN
TIME:    60.00   60.00   60.00
BEAM:    20.18   20.18   20.18

ELEM:       Mg      Al       O   SUM  
   291  17.037  37.712  45.310 100.059
   292  17.098  37.678  45.380 100.156
   293  17.101  37.703  45.461 100.265
   294  17.083  37.692  45.384 100.160
   295  17.162  37.752  45.370 100.283

AVER:   17.096  37.707  45.381 100.185
SDEV:     .045    .028    .054    .091
SERR:     .020    .012    .024
%RSD:      .26     .07     .12

PUBL:   17.084  37.931  44.985 100.000
%VAR:      .07    -.59     .88
DIFF:     .013   -.224    .396
STDS:     3012    3013    3012

STKF:    .3807   .3710   .1549
STCT:   762.41 2611.36  203.26

UNKF:    .1022   .1897   .1600
UNCT:   204.60 1335.62  209.95
UNBG:      .67    2.57    2.23

ZCOR:   1.6734  1.9874  2.8365
KRAW:    .2684   .5115  1.0329
PKBG:   306.81  521.45   95.27

Note that the ZAF corrections (ZCOR) above (Armstrong/Packwood/Brown) range from ~70% to 180%! That's pretty good for that large an extrapolation.

I don't think we need closely matrix match standards for best accuracy. Rather I suspect that the accuracy crisis we are seeing in geology:

https://smf.probesoftware.com/index.php?topic=1811.msg13868#msg13868

is not due to our matrix corrections, but instead more likely due to heterogeneous, natural standards with grain to grain variation:

https://smf.probesoftware.com/index.php?topic=1415.0
https://smf.probesoftware.com/index.php?topic=1770.0

Poor dead time calibrations:

https://smf.probesoftware.com/index.php?topic=1466.msg11102#msg11102

Misaligned (WDS) spectrometers:

https://smf.probesoftware.com/index.php?topic=1739.0

And inappropriate PHA tunings:

https://smf.probesoftware.com/index.php?topic=1466.msg11636#msg11636

One other thing that is visible from these plots is that our precision is more of the limiting factor here, so maybe next weekend we will acquire the data again, but using a longer counting time (was 60 sec on-peak and 15 sec each off peak in the data shown above).

[Probeman]

One other thing that is visible from these plots is that our precision is more of the limiting factor here, so maybe next weekend we will acquire the data again, but using a longer counting time (was 60 sec on-peak and 15 sec each off peak in the data shown above).

As mentioned in the above post, it might help to perform some higher precision analyses. Then we remembered that the 60 nA acquisitions, might provide a reasonable proxy, since 3 times the beam current (20 nA to 60nA), should yield SQRT(3) times better precision, though these higher count rates will be more affected by the dead time correction, possibly an accuracy hit.

So let's compare Henke and FFAST MACs with one of these data sets. First a 20 keV analysis of MgAl2O4 using the Henke MACs:

St 3100 Set   5 MgAl2O4 FIGMAS, Results in Elemental Weight Percents
 
ELEM:       Mg      Al       O
BGDS:      EXP     EXP     EXP
TIME:    60.00   60.00   60.00
BEAM:    60.04   60.04   60.04

ELEM:       Mg      Al       O   SUM  
    83  17.305  37.643  45.285 100.233
    84  17.307  37.708  45.290 100.305
    85  17.299  37.712  45.373 100.384
    86  17.262  37.680  45.292 100.235
    87  17.269  37.682  45.262 100.213

AVER:   17.289  37.685  45.300 100.274
SDEV:     .021    .028    .042    .071
SERR:     .010    .012    .019
%RSD:      .12     .07     .09

PUBL:   17.084  37.931  44.985 100.000
%VAR:     1.20    -.65     .70
DIFF:     .205   -.246    .315
STDS:     3012    3013    3012

STKF:    .4220   .3997   .1888
STCT:   700.01 2267.21  240.18

UNKF:    .1165   .2219   .1969
UNCT:   193.25 1258.67  250.46
UNBG:      .62    2.53    1.29

ZCOR:   1.4839  1.6981  2.3008
KRAW:    .2761   .5552  1.0428
PKBG:   311.72  497.80  194.91

Not too bad at all. The worst being Mg with a +1.2% relative error. While Al and O are both under 1% relative error.

And now the same sample using the FFAST MACs:

BGDS:      EXP     EXP     EXP
TIME:    60.00   60.00   60.00
BEAM:    60.04   60.04   60.04

ELEM:       Mg      Al       O   SUM  
    83  17.244  37.812  45.060 100.116
    84  17.246  37.878  45.064 100.187
    85  17.238  37.881  45.149 100.268
    86  17.201  37.849  45.068 100.118
    87  17.208  37.851  45.037 100.096

AVER:   17.227  37.854  45.075 100.157
SDEV:     .021    .028    .043    .071
SERR:     .010    .012    .019
%RSD:      .12     .07     .09

PUBL:   17.084  37.931  44.985 100.000
%VAR:      .84    -.20     .20
DIFF:     .144   -.077    .090
STDS:     3012    3013    3012

STKF:    .4274   .4047   .2100
STCT:   700.01 2267.21  240.18

UNKF:    .1180   .2247   .2190
UNCT:   193.25 1258.67  250.46
UNBG:      .62    2.53    1.29

ZCOR:   1.4600  1.6850  2.0585
KRAW:    .2761   .5552  1.0428
PKBG:   311.72  497.80  194.91

Well, that's an improvement, and all in the "right" direction, assuming of course a stoichiometric composition for MgAl2O4.  And this, even though the dead time corrections for Al Ka are ~23% for the MgAl2O4 and almost 50% on the Al2O3 primary standard at ~136 kcps!

But lets see the data graphically, first Mg Ka:



now Al Ka:



and O Ka:



Not too bad, we're really happy that we can make this extrapolation for Mg Ka and Al Ka from pure MgO and Al2O3 so well... but I will confess that there's something Andrew and I don't understand. And that is, for O Ka, without applying an Area Peak Factor correction (APFs):

https://smf.probesoftware.com/index.php?topic=536.msg2993#msg2993

we really should be getting totals around 98% in this material on a 60A W/Si LDE. Because the predicted APF for O Ka in this matrix should be around 1.02 (using MgO as a primary standard). At least based on work done by Bastin and myself 25 years ago (Bastin used an Fe2O3 primary standard for oxygen, but the APF for O Ka in Fe2O3 relative to MgO is 1.00).

We're making more high precision wavescan measurements on MgO, Al2O3, MgAl2O4 (and SiO2 just for fun) this weekend.  But if anyone has any ideas on this apparent O Ka anomaly, please respond.

[Probeman]

Andrew and I ran the MgO, Al2O3 and MgAl2O4 FIGMAS standards again yesterday, but this time using different spectrometers.  That is, in our previous runs we ran Mg Ka (sp1), Al Ka (sp2) and O Ka (sp4), but this weekend we ran O Ka (sp1), Mg Ka (sp2) and Al Ka (sp4).

But from previous Bragg order k-ratio testing we know that sp4 for Al Ka is producing slightly low k-ratios and that is what we saw.

Here is 15 keV:

Summary of All Calculated (averaged) Matrix Corrections:
St 3100 Set   1 MgAl2O4 FIGMAS
FFAST    Chantler (NIST v 2.1, 2005)

Elemental Weight Percents:
ELEM:        O      Mg      Al   TOTAL
     1  44.504  17.031  37.739  99.274   Armstrong/Brown/Scott-Love (prZ)
     2  44.109  16.922  38.062  99.093   Philibert/Duncumb-Reed
     3  44.248  16.991  38.021  99.260   Heinrich/Duncumb-Reed
     4  44.552  17.035  37.902  99.488   Love-Scott I
     5  44.559  17.030  37.725  99.314   Love-Scott II
     6  44.267  16.963  37.516  98.747   Packwood Phi(prZ) (EPQ-91)
     7  44.895  17.201  37.874  99.970   Bastin (original) (prZ)
     8  44.775  17.074  38.251 100.100   Bastin PROZA Phi (prZ) (EPQ-91)
     9  44.730  17.057  38.121  99.908   Pouchou and Pichoir-Full (PAP)
    10  44.560  17.033  37.943  99.536   Pouchou and Pichoir-Simplified (XPP)
    11  44.510  17.028  37.752  99.290   PAP/Donovan and Moy BSC/BKS (prZ)

AVER:   44.519  17.033  37.900  99.453
SDEV:     .237    .071    .210    .406
SERR:     .071    .021    .063

MIN:    44.109  16.922  37.516  98.747
MAX:    44.895  17.201  38.251 100.100

Percent Variances:
ELEM:        O      Mg      Al
PUBL:   44.985  17.084  37.931
STDS:     3012    3012    3013

ELEM:        O      Mg      Al
     1   -1.07    -.31    -.51           Armstrong/Brown/Scott-Love (prZ)
     2   -1.95    -.95     .34           Philibert/Duncumb-Reed
     3   -1.64    -.54     .24           Heinrich/Duncumb-Reed
     4    -.96    -.29    -.08           Love-Scott I
     5    -.95    -.32    -.54           Love-Scott II
     6   -1.60    -.71   -1.09           Packwood Phi(prZ) (EPQ-91)
     7    -.20     .69    -.15           Bastin (original) (prZ)
     8    -.47    -.06     .84           Bastin PROZA Phi (prZ) (EPQ-91)
     9    -.57    -.16     .50           Pouchou and Pichoir-Full (PAP)
    10    -.94    -.30     .03           Pouchou and Pichoir-Simplified (XPP)
    11   -1.06    -.33    -.47           PAP/Donovan and Moy BSC/BKS (prZ)

AVER:    -1.04    -.30    -.08        
SDEV:      .53     .41     .55        
SERR:      .16     .12     .17        

MIN:     -1.95    -.95   -1.09        
MAX:      -.20     .69     .84        

Here at 20 keV:

Summary of All Calculated (averaged) Matrix Corrections:
St 3100 Set   2 MgAl2O4 FIGMAS
FFAST    Chantler (NIST v 2.1, 2005)

Elemental Weight Percents:
ELEM:        O      Mg      Al   TOTAL
     1  44.807  17.108  37.700  99.616   Armstrong/Brown/Scott-Love (prZ)
     2  44.301  16.950  37.726  98.977   Philibert/Duncumb-Reed
     3  44.436  17.019  37.698  99.152   Heinrich/Duncumb-Reed
     4  44.878  17.121  38.209 100.208   Love-Scott I
     5  45.072  17.132  37.937 100.141   Love-Scott II
     6  44.571  17.026  37.344  98.942   Packwood Phi(prZ) (EPQ-91)
     7  45.205  17.274  37.995 100.474   Bastin (original) (prZ)
     8  45.161  17.157  38.352 100.670   Bastin PROZA Phi (prZ) (EPQ-91)
     9  45.173  17.147  38.346 100.667   Pouchou and Pichoir-Full (PAP)
    10  44.943  17.107  37.959 100.009   Pouchou and Pichoir-Simplified (XPP)
    11  44.820  17.107  37.720  99.647   PAP/Donovan and Moy BSC/BKS (prZ)

AVER:   44.852  17.104  37.908  99.864
SDEV:     .306    .085    .311    .643
SERR:     .092    .026    .094

MIN:    44.301  16.950  37.344  98.942
MAX:    45.205  17.274  38.352 100.670

Percent Variances:
ELEM:        O      Mg      Al
PUBL:   44.985  17.084  37.931
STDS:     3012    3012    3013

ELEM:        O      Mg      Al
     1    -.40     .14    -.61           Armstrong/Brown/Scott-Love (prZ)
     2   -1.52    -.78    -.54           Philibert/Duncumb-Reed
     3   -1.22    -.38    -.61           Heinrich/Duncumb-Reed
     4    -.24     .22     .73           Love-Scott I
     5     .19     .28     .02           Love-Scott II
     6    -.92    -.34   -1.55           Packwood Phi(prZ) (EPQ-91)
     7     .49    1.11     .17           Bastin (original) (prZ)
     8     .39     .43    1.11           Bastin PROZA Phi (prZ) (EPQ-91)
     9     .42     .37    1.10           Pouchou and Pichoir-Full (PAP)
    10    -.09     .14     .07           Pouchou and Pichoir-Simplified (XPP)
    11    -.37     .14    -.56           PAP/Donovan and Moy BSC/BKS (prZ)

AVER:     -.30     .12    -.06        
SDEV:      .68     .50     .82        
SERR:      .21     .15     .25        

MIN:     -1.52    -.78   -1.55        
MAX:       .49    1.11    1.11        

And 25 keV:

Summary of All Calculated (averaged) Matrix Corrections:
St 3100 Set   3 MgAl2O4 FIGMAS
FFAST    Chantler (NIST v 2.1, 2005)

Elemental Weight Percents:
ELEM:        O      Mg      Al   TOTAL
     1  45.115  17.181  37.473  99.769   Armstrong/Brown/Scott-Love (prZ)
     2  44.631  16.989  37.188  98.808   Philibert/Duncumb-Reed
     3  44.765  17.059  37.175  99.000   Heinrich/Duncumb-Reed
     4  45.062  17.191  38.263 100.515   Love-Scott I
     5  45.538  17.246  38.098 100.883   Love-Scott II
     6  44.958  17.103  37.091  99.152   Packwood Phi(prZ) (EPQ-91)
     7  45.557  17.342  37.920 100.819   Bastin (original) (prZ)
     8  45.513  17.247  38.236 100.996   Bastin PROZA Phi (prZ) (EPQ-91)
     9  45.561  17.243  38.417 101.221   Pouchou and Pichoir-Full (PAP)
    10  45.342  17.197  37.827 100.366   Pouchou and Pichoir-Simplified (XPP)
    11  45.135  17.182  37.502  99.819   PAP/Donovan and Moy BSC/BKS (prZ)

AVER:   45.198  17.180  37.744 100.123
SDEV:     .330    .098    .482    .860
SERR:     .100    .030    .145

MIN:    44.631  16.989  37.091  98.808
MAX:    45.561  17.342  38.417 101.221

Percent Variances:
ELEM:        O      Mg      Al
PUBL:   44.985  17.084  37.931
STDS:     3012    3012    3013

ELEM:        O      Mg      Al
     1     .29     .57   -1.21           Armstrong/Brown/Scott-Love (prZ)
     2    -.79    -.56   -1.96           Philibert/Duncumb-Reed
     3    -.49    -.14   -1.99           Heinrich/Duncumb-Reed
     4     .17     .63     .87           Love-Scott I
     5    1.23     .95     .44           Love-Scott II
     6    -.06     .11   -2.22           Packwood Phi(prZ) (EPQ-91)
     7    1.27    1.51    -.03           Bastin (original) (prZ)
     8    1.17     .96     .80           Bastin PROZA Phi (prZ) (EPQ-91)
     9    1.28     .93    1.28           Pouchou and Pichoir-Full (PAP)
    10     .79     .66    -.28           Pouchou and Pichoir-Simplified (XPP)
    11     .33     .57   -1.13           PAP/Donovan and Moy BSC/BKS (prZ)

AVER:      .47     .56    -.49        
SDEV:      .73     .58    1.27        
SERR:      .22     .17     .38        

MIN:      -.79    -.56   -2.22        
MAX:      1.28    1.51    1.28        

Not too bad really, again usually within +/1 1%

Tomorrow I'll show some plots of the data...

[Probeman]

Continuing with the FIGMAS MgO, Al2O3 and MgAl2O4 analyses at 15, 20 and 25 keV using different spectrometers from above, we see this for Mg Ka:



These are four separate acquisitions on MgAl2O4 each at 15, 20 and 25 keV.  Not all that bad, most analyses are between +1% and -1% relative error. Now for Al Ka and here we see a problem:



This is in contrast to the Al Ka measured on spectrometer 2 above where we obtained excellent accuracy:

https://smf.probesoftware.com/index.php?topic=1823.msg13944#msg13944

We will re-run these materials again, but I suspect a small spectrometer misalignment on spectrometer 4, since using spectrometer 2 we obtained excellent accuracy several times (at 20 nA and 60 nA).  Now for O ka:



Again, quite good accuracy extrapolating from the pure oxides where at 25 keV we are looking at matrix corrections on the order of 150% for oxygen Ka (ZCOR = 2.43).