This K-Ratio Consensus project as proposed by Nicholas Ritchie was followed up by Will Nachlas et al., and published here:

https://academic.oup.com/mam/article/29/Supplement_1/225/7228643

The pdf is attached below.

A further followup of the quantification accuracy for the MgO-Al2O3-MgAl2O4 system is found in this topic:

https://smf.probesoftware.com/index.php?topic=1823.0

It sure would be nice to see someone besides Probeman post some quant results..

Note the accuracy of this quant very much depends on properly aligned WDS spectrometers as described here:

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

Of course since there is a significant difference in count rates between the primary (MgO and Al2O3) primary standards and the (MgAl2O4) secondary standard, the accuracy also depends on tuning the PHA parameters properly as described here:

https://smf.probesoftware.com/index.php?topic=1854.msg14601#msg14601

Next most important would be the MACs, especially for O ka.  The NIST FFAST MACs are recommended:

https://smf.probesoftware.com/index.php?topic=1831.0

https://smf.probesoftware.com/index.php?topic=1831.msg14017#msg14017

I'm serious: the most important post I've made on this forum is the previous post in this PHA tuning topic:

https://smf.probesoftware.com/index.php?topic=1854.msg14601#msg14601

The reason is that we know we have a problem with natural mineral (e.g., Smithsonian) standards that are heterogeneous and are only in very limited quantities. However, if we try to use globally distributed high purity synthetic oxide/silicate end-member minerals as primary standards (as proposed by many) where there are significant differences in count rates between these primary standards and our unknown samples, we will not obtain accurate results using traditional PHA tuning methods (as described in the previous post).

The traditional PHA tuning method only works when the standard and unknown are approximately "count rate matched", though it is often misleadingly referred to as "matrix matched".

Now of course other instrumental parameters also matter for accuracy, primarily the spectrometer alignments, especially for lower energy x-rays, Mg Ka, Si, Ka, and Al Ka. And at high count rates the dead time correction matters:

https://smf.probesoftware.com/index.php?topic=1466.msg11173#msg11173

and for extrapolations in average atomic number the Z-based backscatter correction matters:

https://smf.probesoftware.com/index.php?topic=1566.0

Also modern MAC tables like the FFAST tables help accuracy. But it should be pointed out that good standards and this new integral-baseline PHA tuning methods are the most important factors in EPMA accuracy.  To demonstrate take a look at the analysis of the NIST K-411 and K-412 mineral glasses at the end of the previous post in this topic. All within 1% accuracy.

Now let's turn *off* the logarithmic dead time correction, the FFAST MACs and the Z-based (DAM) backscatter correction and see what happens to our accuracy. Here's K-411 again still using SiO2, MgO and Fe3O4 as primary standards:

St  162 Set   2 NBS K-411 mineral glass, Results in Elemental Weight Percents
 
ELEM:       Si      Mg      Fe      Ca      Na       K      Mn      Al      Ti       O
TYPE:     ANAL    ANAL    ANAL    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC
BGDS:      LIN     LIN     LIN
TIME:    60.00   60.00   60.00     ---     ---     ---     ---     ---     ---     ---
BEAM:    19.95   19.95   19.95     ---     ---     ---     ---     ---     ---     ---

ELEM:       Si      Mg      Fe      Ca      Na       K      Mn      Al      Ti       O   SUM  
   167  25.492   8.846  11.230  11.057    .000    .000    .077    .053    .000  43.558 100.314
   168  25.513   8.831  11.095  11.057    .000    .000    .077    .053    .000  43.558 100.183
   169  25.472   8.873  11.115  11.057    .000    .000    .077    .053    .000  43.558 100.205
   170  25.496   8.859  11.217  11.057    .000    .000    .077    .053    .000  43.558 100.317
   171  25.404   8.842  11.112  11.057    .000    .000    .077    .053    .000  43.558 100.104
   172  25.450   8.820  11.153  11.057    .000    .000    .077    .053    .000  43.558 100.168

AVER:   25.471   8.845  11.154  11.057    .000    .000    .077    .053    .000  43.558 100.215
SDEV:     .039    .019    .058    .000    .000    .000    .000    .000    .000    .000    .085
SERR:     .016    .008    .023    .000    .000    .000    .000    .000    .000    .000
%RSD:      .15     .22     .52     .00     .00     .00     .00     .00     .00     .00

PUBL:   25.382   8.847  11.209  11.057    n.a.    n.a.    .077    .053    n.a.  43.558 100.183
%VAR:      .35    -.02    -.49     ---     ---     ---     ---     ---     ---     ---
DIFF:     .089   -.002   -.055     ---     ---     ---     ---     ---     ---     ---
STDS:       14      12     395     ---     ---     ---     ---     ---     ---     ---

Our accuracy for all three elements is still within 1%! OK, what about K-412:

St  160 Set   2 NBS K-412 mineral glass, Results in Elemental Weight Percents
 
ELEM:       Si      Mg      Fe      Ca      Na       K      Mn      Al      Ti       O
TYPE:     ANAL    ANAL    ANAL    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC
BGDS:      LIN     LIN     LIN
TIME:    60.00   60.00   60.00     ---     ---     ---     ---     ---     ---     ---
BEAM:    19.95   19.95   19.95     ---     ---     ---     ---     ---     ---     ---

ELEM:       Si      Mg      Fe      Ca      Na       K      Mn      Al      Ti       O   SUM  
   161  21.532  11.724   7.730  10.899    .043    .000    .077   4.906    .000  43.597 100.509
   162  21.496  11.756   7.782  10.899    .043    .000    .077   4.906    .000  43.597 100.556
   163  21.504  11.724   7.784  10.899    .043    .000    .077   4.906    .000  43.597 100.535
   164  21.434  11.730   7.747  10.899    .043    .000    .077   4.906    .000  43.597 100.433
   165  21.429  11.709   7.736  10.899    .043    .000    .077   4.906    .000  43.597 100.396
   166  21.451  11.696   7.720  10.899    .043    .000    .077   4.906    .000  43.597 100.389

AVER:   21.474  11.723   7.750  10.899    .043    .000    .077   4.906    .000  43.597 100.470
SDEV:     .042    .020    .027    .000    .000    .000    .000    .000    .000    .000    .073
SERR:     .017    .008    .011    .000    .000    .000    .000    .000    .000    .000
%RSD:      .20     .17     .35     .00     .00     .00     .00     .00     .00     .00

PUBL:   21.199  11.657   7.742  10.899    .043    .000    .077   4.906    .000  43.597 100.120
%VAR:     1.30     .57     .10     ---     ---     ---     ---     ---     ---     ---
DIFF:     .275    .066    .008     ---     ---     ---     ---     ---     ---     ---
STDS:       14      12     395     ---     ---     ---     ---     ---     ---     ---

OK, the Si accuracy is now only 1.3% relative, but still, not bad at all.

The point being that for typical geological minerals (and typical count rates) you don't require the logarithmic dead time correction, or the FFAST MACs or the DAM backscatter corrections.

Just use high purity synthetic oxide/silicate minerals and this new integral-baseline PHA tuning method and you can obtain 1% relative accuracy in your analyses.

If you don't get ~1% relative accuracy you might want to check your spectrometer alignments using the Bragg order k-ratio test method:

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

I just wanted to mention again, because someone asked me recently "why do you make an element by difference negative when the total is greater than 100%?"

The reasoning is analogous to measuring a trace element that is at zero concentration:

https://smf.probesoftware.com/index.php?topic=392.msg12619#msg12619

When measuring a zero concentration one must include both the positive values and the negative values in order for the trace element average to approach zero.

In the same way, when calculating an element by difference, when the element by difference is zero, one must include both the positive values for the element by difference (when the total is less than 100%) *and* also the negative values (when the total is greater than 100%).  This way the element by difference averages to zero if it is not present.

For example this glass that contains no water:

Un   12 Withers-NSL, Results in Oxide Weight Percents

ELEM:     Na2O     K2O      Cl     BaO       F    TiO2     FeO     MnO     CaO    SiO2   Al2O3     MgO     O-D     H2O       O   SUM  
   514   5.756   4.302    .213   -.046    .111    .163   4.109    .033    .150  74.652  10.425    .000     ---    .133    .000 100.000
   515   5.112   4.352    .234   -.079    .108    .239   4.131    .109    .207  74.252  10.396   -.004     ---    .942    .000 100.000
   516   5.500   4.110    .215    .054    .038    .228   4.121    .082    .204  74.470  10.459    .005     ---    .513    .000 100.000
   517   5.278   4.351    .291   -.009    .165    .168   4.107    .122    .160  74.722  10.485    .000     ---    .161    .000 100.000
   518   5.678   4.351    .234    .018    .069    .152   4.143    .114    .157  74.664  10.420    .001     ---    .000    .000 100.000
   519   5.553   4.501    .153    .007    .091    .168   4.074    .082    .180  74.356  10.368    .001     ---    .466    .000 100.000
   520   5.876   4.449    .210    .011    .097    .249   4.038    .088    .174  74.691  10.400   -.001     ---   -.283    .000 100.000
   521   5.784   4.547    .210    .022    .085    .222   4.113   -.004    .167  74.619  10.407    .009     ---   -.180    .000 100.000
   522   5.757   4.464    .221   -.034    .056    .211   4.073    .069    .169  74.899  10.482    .003     ---   -.370    .000 100.000
   523   5.571   4.507    .288   -.046    .079    .217   4.005    .049    .180  74.064  10.410    .001     ---    .674    .000 100.000
   524   5.690   4.486    .231    .031    .084    .260   4.071    .080    .180  74.332  10.399   -.002     ---    .160    .000 100.000

AVER:    5.596   4.402    .227   -.006    .089    .207   4.089    .075    .175  74.520  10.423    .001     ---    .202    .000 100.000
SDEV:     .230    .126    .038    .040    .033    .038    .042    .037    .018    .246    .037    .004     ---    .412    .000    .000
SERR:     .069    .038    .011    .012    .010    .011    .013    .011    .005    .074    .011    .001     ---    .124    .000
%RSD:     4.11    2.86   16.62 -617.22   36.76   18.36    1.02   49.87   10.18     .33     .36  315.23     ---  204.24  466.43
STDS:      336     374     285     835     835      22     395      25     358     162     336      12     ---     ---     ---

Note that the average H2O by difference is within a standard deviation from zero as one would expect.

Of course this statistical consideration applies to all element(s) by difference that approach zero concentrations, not just H2O by difference in glasses.

Here's a tip when adjusting your off-peak backgrounds from the Plot! window after plotting up a wavescan:



If the current sample is a wavescan, or a unknown or standard which contains intensity data, Probe for EPMA cannot assign the new off-peak positions to the current sample. Instead you will get this warning:



When you see this message you should immediately create a new unknown sample from the Acquire! window. Then your changes to the off-peak positions will be saved.

Note that you can also assign the modified off-peak position using these option here from the Model backgrounds button:



Even easier, just create a new unknown sample before opening the Plot! window before making any changes to the off-peak positions.

Quote from: jon_wade on October 27, 2026, 01:53:22 PMjust putting it out there, but theres no V in your unknown perchance?  I mean, you'd have to have a percent or so to explain the totals and thats ...unlikely, and wouldn't explain the lower kV highs. Much more likely its a coating difference and/or a standard issue

That's an interesting point.

Since Dominik is using the Ti Kb line, there is a possible interference from the V Ka line. Here calculated using the Standard app (Misc Options menu):

On Peak Interferences for : St    1 Ti99V1
For Ti kb    LiF at  2.51391 angstroms, at an assumed concentration of 99 wt.% 
  Interference by V  KA1            at  2.50430 ( 174.150) ( -.66806) =      4.1%
  Interference by V  KA2            at  2.50780 ( 174.393) ( -.42468) =      2.8%
  On Peak Position   -------------  at  2.51391 ( 174.818)

https://smf.probesoftware.com/index.php?topic=598.msg13394#msg13394

just putting it out there, but theres no V in your unknown perchance?  I mean, you'd have to have a percent or so to explain the totals and thats ...unlikely, and wouldn't explain the lower kV highs. Much more likely its a coating difference and/or a standard issue

It doesn't make any sense that high totals in the rutile inclusion could be caused by fluoresecence from the CaTiSiO5 matrix, but I ran a quick simulation in Standard:

https://smf.probesoftware.com/index.php?topic=58.msg212#msg212

It doesn't make any sense that this high Ti in the inclusion could be from the matrix since there's less Ti in titanite than TiO2, but sometimes physics unintuitive so I ran it.

There was a TiO2 PAR file already in the Compound sub folder, but had to run a titanite material simulation first. Both PAR files are attached below. Here's the PENFLUOR/FANAL calculation:

https://smf.probesoftware.com/index.php?topic=58.msg212#msg212

If we zoom into the Ti Kb signal:



It makes sense that the Ti should decrease due to less internal fluorescence and here we can see the contribution just from the boundary material:



Of course this geometry assumes a vertical plane boundary and the actual inclusion geometry should be modeled in PENEPMA but an inclusion geometry should only increase these effects more.

So the high Ti in the rutile has got to be from a carbon coat difference or maybe a standard that is charging slightly...

So Quorum suggest can use Pt with rotary pump version if use a metal hose https://quorum-configurator.cdn.prismic.io/quorum-configurator/Z5EdK5bqstJ99vno_QSeriesNewBrochure_Printready.pdf so just wondering how use attach a metal hose when it's current fitting is for PVC tube push onto metal tube with wire clip. Has anyone tried this?

I included Ti to demonstrate it is pure rutile (I also included Ca, Si, etc. to monitor contributions from the surrounding). I will try with lower beam current, but have my doubts, also regarding the coating, as the totals scatter around 3-4 wt%. We now started a more systematic study with grain size or other mineral surrounding the rutile.

Thanks for all the comments, these helped a lot already.

I put together some schematics to help explain how this new integral-baseline PHA tuning method allows one to utilize high purity synthetic end member oxide/silicates as primary standards for geological microanalysis.

Here's the problem of the so-called Pulse Height Analysis (PHA) "depression":



As the photon count rate increases, the PHA peak shifts to the left (lower voltages) because at high count rates the gas detector cannot recover its bias (high) voltage completely. At lower count rates the peak shifts to the right. So any differences in the count rates between the primary standard and the unknown need to be properly compensated for.

Here is a schematic of what I would call a "traditional" PHA tuning (what I was taught many moons ago):



As one can see, a significant change in the count rate between the primary standard and the unknown will cause the PHA peak to shift to the left (at higher count rates) or to the right (at lower count rates). This will cause the tails of the PHA peak to be cut causing a non-linear response in the counting system. This is the main reason people have assumed they needed to "matrix match" their primary standard to their unknown, when in fact they were merely "count rate" matching!

Therefore an improvement to this "traditional" PHA tuning is to simple "open up" the baseline and window to allow the PHA peak to shift around as shown here schematically:



But in fact what we really need to do is place our PHA in integral mode and to tune our PHAs at the highest count rate we expect to encounter for that element in the probe session. That is, on a material (e.g., high concentration primary standard) if we tune our our PHA and adjust our bias *or* gain so that our PHA peak is *completely* above the baseline level, then at lower count rates it will only shift to the right. Thus ensuring a more linear response in our counting electronics:



Now some might say "well I need to use the window level in differential mode because I have a high order Bragg interference". And that might be true in some rare instances, but it would be better to correct for the WDS interferences in software and have a linear response in your pulse processing system. Also SEM Geologist correctly points out that the dead time correction is more accurate if applied to the integral mode count rate, as the detector sees all the photons.

So in integral mode all pulses above the baseline will be still be counted. In fact, the surprising thing is that even if the PHA peak on the right is partially "cut off" in the graphical display, all pulses to the right are counted even if they are not visible!  Did you know this?  I didn't until just a few years ago!

In a chat with JEOL engineer John Glass, he confirmed this saying that the JEOL PHA electronics will energy discriminate up to 15 volts (though only display up to 10 volts), but if there are any pulses above 15 voltage they are set to 15 volts and *still counted*! 

The Cameca PHA behaves similarly by only displaying up to 5 volts, but energy discriminating up to 10v, and then setting all pulses above 10v to 10 volts.

Here is an example of a Fe Ka PHA scan that will blow your mind, but it will produce a completely linear response for high accuracy microanalysis, because the Fe "escape" peak is *fully* above the baseline at high count rates (and therefore at lower intensities the PHA peak will shift to the right and still be counted in integral mode):



You might say "prove it!"  OK, so here is an analysis of NIST K-411 mineral glass using SiO2, MgO and Fe3O4 as primary standards:

St  162 Set   2 NBS K-411 mineral glass, Results in Elemental Weight Percents
 
ELEM:       Si      Mg      Fe      Ca      Na       K      Mn      Al      Ti       O
TYPE:     ANAL    ANAL    ANAL    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC
BGDS:      LIN     LIN     LIN
TIME:    60.00   60.00   60.00     ---     ---     ---     ---     ---     ---     ---
BEAM:    19.95   19.95   19.95     ---     ---     ---     ---     ---     ---     ---

ELEM:       Si      Mg      Fe      Ca      Na       K      Mn      Al      Ti       O   SUM  
   167  25.332   8.772  11.190  11.057    .000    .000    .077    .053    .000  43.558 100.039
   168  25.353   8.756  11.055  11.057    .000    .000    .077    .053    .000  43.558  99.909
   169  25.312   8.798  11.075  11.057    .000    .000    .077    .053    .000  43.558  99.930
   170  25.335   8.785  11.177  11.057    .000    .000    .077    .053    .000  43.558 100.042
   171  25.245   8.768  11.072  11.057    .000    .000    .077    .053    .000  43.558  99.830
   172  25.290   8.745  11.113  11.057    .000    .000    .077    .053    .000  43.558  99.894

AVER:   25.311   8.771  11.114  11.057    .000    .000    .077    .053    .000  43.558  99.940
SDEV:     .039    .019    .057    .000    .000    .000    .000    .000    .000    .000    .084
SERR:     .016    .008    .023    .000    .000    .000    .000    .000    .000    .000
%RSD:      .15     .22     .52     .00     .00     .00     .00     .00     .00     .00

PUBL:   25.382   8.847  11.209  11.057    n.a.    n.a.    .077    .053    n.a.  43.558 100.183
%VAR:     -.28    -.86    -.85     ---     ---     ---     ---     ---     ---     ---
DIFF:    -.071   -.076   -.095     ---     ---     ---     ---     ---     ---     ---
STDS:       14      12     395     ---     ---     ---     ---     ---     ---     ---

Note the relative percent errors bolded in red!  All within 1% relative accuracy!  And the Fe results were obtained using the PHA tuning with the PHA peak appearing to be "cut-off" on the right side shown above...

And here is NIST K-412 again using pure oxide primary standards:

St  160 Set   2 NBS K-412 mineral glass, Results in Elemental Weight Percents
 
ELEM:       Si      Mg      Fe      Ca      Na       K      Mn      Al      Ti       O
TYPE:     ANAL    ANAL    ANAL    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC
BGDS:      LIN     LIN     LIN
TIME:    60.00   60.00   60.00     ---     ---     ---     ---     ---     ---     ---
BEAM:    19.95   19.95   19.95     ---     ---     ---     ---     ---     ---     ---

ELEM:       Si      Mg      Fe      Ca      Na       K      Mn      Al      Ti       O   SUM  
   161  21.434  11.633   7.701  10.899    .043    .000    .077   4.906    .000  43.597 100.289
   162  21.398  11.664   7.752  10.899    .043    .000    .077   4.906    .000  43.597 100.336
   163  21.406  11.633   7.754  10.899    .043    .000    .077   4.906    .000  43.597 100.315
   164  21.336  11.639   7.717  10.899    .043    .000    .077   4.906    .000  43.597 100.214
   165  21.331  11.618   7.706  10.899    .043    .000    .077   4.906    .000  43.597 100.176
   166  21.353  11.605   7.690  10.899    .043    .000    .077   4.906    .000  43.597 100.170

AVER:   21.376  11.632   7.720  10.899    .043    .000    .077   4.906    .000  43.597 100.250
SDEV:     .042    .020    .027    .000    .000    .000    .000    .000    .000    .000    .072
SERR:     .017    .008    .011    .000    .000    .000    .000    .000    .000    .000
%RSD:      .20     .17     .35     .00     .00     .00     .00     .00     .00     .00

PUBL:   21.199  11.657   7.742  10.899    .043    .000    .077   4.906    .000  43.597 100.120
%VAR:      .84    -.22    -.29     ---     ---     ---     ---     ---     ---     ---
DIFF:     .177   -.025   -.022     ---     ---     ---     ---     ---     ---     ---
STDS:       14      12     395     ---     ---     ---     ---     ---     ---     ---

Again, Si Ka, Mg Ka and Fe Ka all within 1% relative accuracy! Even with a 10x extrapolation in count rates for Fe ka.

Thus, one can obtain 1% relative accuracy using pure oxide/silicate standards with this improved integral/baseline PHA tuning method.  Try it yourself...