Proper PHA tuning for high accuracy quantitative analysis

Started by Probeman, June 14, 2026, 10:46:17 AM

Previous topic - Next topic

Probeman

Starting a new topic to focus on the correct method to tune one's PHA settings because proper tuning of PHA is a "black box" to most analysts!  Including myself until quite recently!  >:(

This link is to the "Limits of EPMA Accuracy" which explores this topic in broader detail ("peak shift" matching, dead time calibration, and better standards):

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

Let's start by analyzing Ti Ka in synthetic SrTiO3 using TiO2 as a primary standard (without "count rate" matching our primary standard to our secondary standard or unknown!).

Remember, on the Cameca the PHA range is 0 to 5 volts, while on the JEOL the PHA range is 0 to 10 volts!

Running a PHA scan using the default PHA settings we obtain this scan on the TiO2 standard (remember, we always want to adjust our PHA settings on a material with the highest concentration (generally the primary standard) at the highest beam current we expect to utilize (that is, at the highest count rate we expect to see, that is, tuning our PHA to avoid pulse height depression effects which can cause non-linear responses)



Note first that we are in "integral" PHA mode. In integral PHA all photons to the right are still counted, even those that plot (invisibly!) to the right of the maximum x axis displayed!

Next we clearly see that the baseline level is too high and is intersecting the PHA peak. Why is that a problem?  Because if the peak shifts to the right or left, the baseline level cuts off more or less of the PHA peak and we obtain a non-linear response in our counting electronics!

However, we can make two adjustments to avoid this: first by lowering our baseline level below the PHA peak, and second, increasing our gain (for *both* Cameca and JEOL instruments), we can shift the PHA peak to the right *completely above* the baseline level, thus ensuring a linear response in our counting electrics. 

Yes, one might need to increase their gain by a factor of 2x or 3x or more in oprder to see the PHA peak fully above the baseline level, but again, all photons are counted to the right in integral mode, so even if you can't see them in the PHA display, they are still counted!

One caveat: the Cameca SXFive instrument new PHA electronics will force a minimum PHA baseline based on the gain level, thus making this procedure more difficult. See here for more details:

https://smf.probesoftware.com/index.php?topic=1466.msg14450#msg14450

Here is the PHA scan after adjusting our baseline level and our gain:



Better, but the baseline level still intersects the PHA peak tail a bit on the left.  After raising our PHA gain still further, we obtain this PHA scan:



This is what we will utilize for our quantitative analyses in the next post.  The same PHA tuning procedure was utilized for Fe Ka as seen here:


The only stupid question is the one not asked!

Probeman

OK, so how does this rather unintuitive PHA tuning method, shown in the previous post actually perform?

Let's start with Ti Ka in SrTiO3 using TiO2 as the primary standard (showing three different matrix correction models) The raw k-ratio for this pairing is around 0.43 so over a 100% difference in count rates:



All three models perform quite well and are all within ~1% relative accuracy, though the DAM BSE correction does somewhat better.

Now let's look at Fe Ka in SRM K-411 glass using Fe3O4 as a primary standard:



Again, all three models do very well using the integral baseline PHA tuning method, though again, the DAM BSE models is somewhat more accurate. This pair has a raw k-ratio of 0.1405, so an even more extreme extrapolation than the TiO2 example above.

Do you care about EPMA accuracy? Why don't you try this PHA tuning method and do some intensity extrapolations from your primary standard to a secondary standard and see what you learn... send me a private message (or email) if you are too shy to share your results.  I'm here to help!    :)
The only stupid question is the one not asked!

sem-geologist

Quote from: Probeman on June 14, 2026, 10:46:17 AMStarting a new topic to focus on the correct method to tune one's PHA settings because proper tuning of PHA is a "black box" to most analysts!  Including myself until quite recently!  >:(

I think one of primary things which should be very much bolded, underlined, SHOUTED, SCREAMED, printed in large font and send all over the earth with paper mails... and I don't know what else... is this:

DO NOT RELY ON OEM AUTO PHA FUNCTION - THOSE ARE MADE BASED ON HISTORICALLY WRONG ASSUMPTIONS AND IT LEADS TO VERY HUGE ERROR!

We could ask EMAS, AMAS, MAS and other EPMA-related societies to update their mouse pads, wall-cheat-sheets/posters with this disclamer – maybe it would start changing something...

Probeman

Quote from: sem-geologist on June 15, 2026, 07:27:28 AM
Quote from: Probeman on June 14, 2026, 10:46:17 AMStarting a new topic to focus on the correct method to tune one's PHA settings because proper tuning of PHA is a "black box" to most analysts!  Including myself until quite recently!  >:(

I think one of primary things which should be very much bolded, underlined, SHOUTED, SCREAMED, printed in large font and send all over the earth with paper mails... and I don't know what else... is this:

DO NOT RELY ON OEM AUTO PHA FUNCTION - THOSE ARE MADE BASED ON HISTORICALLY WRONG ASSUMPTIONS AND IT LEADS TO VERY HUGE ERROR!

We could ask EMAS, AMAS, MAS and other EPMA-related societies to update their mouse pads, wall-cheat-sheets/posters with this disclamer – maybe it would start changing something...

I could not agree more, though I think only the Cameca has such an "auto PHA" button. The JEOL might have it but apparently it's only accessible by the service engineer for running tests.

But regardless, the auto OEM or traditional manual PHA tuning methods basically *force* the user into utilizing "matrix matched" or as I call more properly: "count rate matched" natural standards (with all their documented heterogeneity and inclusions). Because if there is a difference in the count rate between the standard and the unknown, the traditional PHA tuning will cause a non-linear response, due to the baseline/window filtering.

Yup, at this point I think we mostly have a sociological, or at least an educational endeavor (or for some stubborn souls, perhaps a psychological hurdle) to overcome traditional PHA tuning methods.  Once people overcome their initial reluctance they will find that they can use high purity synthetic end member mineral standards and still achieve ~1% relative accuracy on their unknown samples.

By the way, if anyone with a JEOL instrument is interested in running PHA tests similar to those in first post above:

https://smf.probesoftware.com/index.php?topic=1854.0

I would be very interested in seeing a series screen captures of PHA adjustments where they attempt the integral-baseline PHA tuning method described above.

Then see what quantitative results they get when extrapolating Fe Ka from say Fe3O4 to the NIST glasses or Ti Ka extrapolating from TiO2 to SrTiO3...
The only stupid question is the one not asked!

Probeman

Quote from: Probeman on June 15, 2026, 08:21:31 AMBy the way, if anyone with a JEOL instrument is interested in running PHA tests similar to those in first post above:

https://smf.probesoftware.com/index.php?topic=1854.0

I would be very interested in seeing a series screen captures of PHA adjustments where they attempt the integral-baseline PHA tuning method described above.

Then see what quantitative results they get when extrapolating Fe Ka from say Fe3O4 to the NIST glasses or Ti Ka extrapolating from TiO2 to SrTiO3...

Not to put too fine a point on this integral-baseline PHA tuning method, but this next example is pretty graphic. Here is Fe Ka measured at multiple keVs in NIST SRM K-412 mineral glass using Fe3O4 as a primary standard:



This measurement yields a raw k-ratio of ~0.97 at 20 at keV, which means that this is a count rate extrapolation of 10x from the primary standard to the secondary standard!

St  160 Set  12 NBS K-412 mineral glass, Results in Elemental Weight Percents
 
ELEM:       Ti      Fe      Sr      Rb      Mn      Cr      Si      Mg      Ca      Al       O
TYPE:     ANAL    ANAL    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC
BGDS:      LIN     LIN
TIME:    60.00   60.00     ---     ---     ---     ---     ---     ---     ---     ---     ---
BEAM:    19.94   19.94     ---     ---     ---     ---     ---     ---     ---     ---     ---

ELEM:       Ti      Fe      Sr      Rb      Mn      Cr      Si      Mg      Ca      Al       O   SUM 
  1169    .003   7.756    .000    .000    .077    .000  21.199  11.657  10.899   4.906  43.597 100.094
  1170    .000   7.732    .000    .000    .077    .000  21.199  11.657  10.899   4.906  43.597 100.067
  1171    .002   7.713    .000    .000    .077    .000  21.199  11.657  10.899   4.906  43.597 100.050
  1172    .002   7.727    .000    .000    .077    .000  21.199  11.657  10.899   4.906  43.597 100.064
  1173   -.002   7.761    .000    .000    .077    .000  21.199  11.657  10.899   4.906  43.597 100.094
  1174   -.001   7.750    .000    .000    .077    .000  21.199  11.657  10.899   4.906  43.597 100.085

AVER:     .000   7.740    .000    .000    .077    .000  21.199  11.657  10.899   4.906  43.597 100.076
SDEV:     .002    .019    .000    .000    .000    .000    .000    .000    .000    .000    .000    .018
SERR:     .001    .008    .000    .000    .000    .000    .000    .000    .000    .000    .000
%RSD:   402.07     .24     .00     .00     .00     .00     .00     .00     .00     .00     .00

PUBL:     n.a.   7.742    n.a.    n.a.    .077    n.a.  21.199  11.657  10.899   4.906  43.597 100.077
%VAR:      ---    -.03     ---     ---     ---     ---     ---     ---     ---     ---     ---
DIFF:      ---   -.002     ---     ---     ---     ---     ---     ---     ---     ---     ---
STDS:       22     395     ---     ---     ---     ---     ---     ---     ---     ---     ---

STKF:    .5611   .6860     ---     ---     ---     ---     ---     ---     ---     ---     ---
STCT:  1833.94 1035.67     ---     ---     ---     ---     ---     ---     ---     ---     ---

UNKF:    .0000   .0666     ---     ---     ---     ---     ---     ---     ---     ---     ---
UNCT:      .01  100.60     ---     ---     ---     ---     ---     ---     ---     ---     ---
UNBG:     2.64    1.38     ---     ---     ---     ---     ---     ---     ---     ---     ---

ZCOR:   1.1885  1.1615     ---     ---     ---     ---     ---     ---     ---     ---     ---
KRAW:    .0000   .0971     ---     ---     ---     ---     ---     ---     ---     ---     ---
PKBG:     1.00   73.96     ---     ---     ---     ---     ---     ---     ---     ---     ---

And here even at 10 keV we are well within 1% relative accuracy:

St  160 Set   1 NBS K-412 mineral glass, Results in Elemental Weight Percents
 
ELEM:       Ti      Fe      Sr      Rb      Mn      Cr      Si      Mg      Ca      Al       O
TYPE:     ANAL    ANAL    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC    SPEC
BGDS:      LIN     LIN
TIME:    60.00   60.00     ---     ---     ---     ---     ---     ---     ---     ---     ---
BEAM:    19.99   19.99     ---     ---     ---     ---     ---     ---     ---     ---     ---

ELEM:       Ti      Fe      Sr      Rb      Mn      Cr      Si      Mg      Ca      Al       O   SUM 
   707    .008   7.781    .000    .000    .077    .000  21.199  11.657  10.899   4.906  43.597 100.124
   708   -.015   7.945    .000    .000    .077    .000  21.199  11.657  10.899   4.906  43.597 100.265
   709    .015   7.659    .000    .000    .077    .000  21.199  11.657  10.899   4.906  43.597 100.009
   710    .009   7.778    .000    .000    .077    .000  21.199  11.657  10.899   4.906  43.597 100.122
   711   -.011   7.827    .000    .000    .077    .000  21.199  11.657  10.899   4.906  43.597 100.152
   712    .011   7.800    .000    .000    .077    .000  21.199  11.657  10.899   4.906  43.597 100.147

AVER:     .003   7.798    .000    .000    .077    .000  21.199  11.657  10.899   4.906  43.597 100.136
SDEV:     .013    .092    .000    .000    .000    .000    .000    .000    .000    .000    .000    .082
SERR:     .005    .038    .000    .000    .000    .000    .000    .000    .000    .000    .000
%RSD:   438.46    1.18     .00     .00     .00     .00     .00     .00     .00     .00     .00

PUBL:     n.a.   7.742    n.a.    n.a.    .077    n.a.  21.199  11.657  10.899   4.906  43.597 100.077
%VAR:      ---     .73     ---     ---     ---     ---     ---     ---     ---     ---     ---
DIFF:      ---    .056     ---     ---     ---     ---     ---     ---     ---     ---     ---
STDS:       22     395     ---     ---     ---     ---     ---     ---     ---     ---     ---

STKF:    .5446   .6648     ---     ---     ---     ---     ---     ---     ---     ---     ---
STCT:   317.84   88.25     ---     ---     ---     ---     ---     ---     ---     ---     ---

UNKF:    .0000   .0635     ---     ---     ---     ---     ---     ---     ---     ---     ---
UNCT:      .01    8.42     ---     ---     ---     ---     ---     ---     ---     ---     ---
UNBG:      .96     .34     ---     ---     ---     ---     ---     ---     ---     ---     ---

ZCOR:   1.1948  1.2290     ---     ---     ---     ---     ---     ---     ---     ---     ---
KRAW:    .0000   .0954     ---     ---     ---     ---     ---     ---     ---     ---     ---
PKBG:     1.02   25.77     ---     ---     ---     ---     ---     ---     ---     ---     ---

I encourage you to try this new PHA tuning method and convince yourself.  See attached pdf.
The only stupid question is the one not asked!

Les Moore

Regardless of the approach, looking at the PHA at different count rates is an important activity. The peak shift and doubling and even tripling peaks show up at higher count rates. To include or exclude - that is the question. If you run over a pure material you may even saturate the detector and it is blinded and gives no counts for an appreciable time afterwards.   

Probeman

Quote from: Les Moore on June 16, 2026, 05:53:53 PMRegardless of the approach, looking at the PHA at different count rates is an important activity.

I found it to be very educational!

But the most surprising aspect for me, was discovering that in integral mode all photons to the right of the baseline are counted.  Even when they appear to be "cutoff" in the graphical display.  That was the "a ha!" moment for me.

Quote from: Les Moore on June 16, 2026, 05:53:53 PMThe peak shift and doubling and even tripling peaks show up at higher count rates. To include or exclude - that is the question. If you run over a pure material you may even saturate the detector and it is blinded and gives no counts for an appreciable time afterwards.

These PHA distributions were acquired at fairly "normal" beam currents, 20, 30 nA or so which makes sense for major and minor elements. Under these conditions I only see the normal peak and sometimes the escape peak if the gas physics permits it.

For trace elements of course one can acquire the primary standard at a lower beam current than the unknown, but then one must be sure to check their picoammeter linearity using the constant k-ratio method:

https://smf.probesoftware.com/index.php?topic=1466.msg11124#msg11124
The only stupid question is the one not asked!

Ben Buse

Quote from: Probeman on June 14, 2026, 10:46:17 AMStarting a new topic to focus on the correct method to tune one's PHA settings because proper tuning of PHA is a "black box" to most analysts!  Including myself until quite recently!  >:(

This link is to the "Limits of EPMA Accuracy" which explores this topic in broader detail ("peak shift" matching, dead time calibration, and better standards):

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

Let's start by analyzing Ti Ka in synthetic SrTiO3 using TiO2 as a primary standard (without "count rate" matching our primary standard to our secondary standard or unknown!).

Running a PHA scan using the default PHA settings we obtain this scan on the TiO2 standard (remember, we always want to adjust our PHA settings on a material with the highest concentration (generally the primary standard) at the highest beam current we expect to utilize (that is, at the highest count rate we expect to see, that is, tuning our PHA to avoid pulse height depression effects which can cause non-linear responses)



Note first that we are in "integral" PHA mode. In integral PHA all photons to the right are still counted, even those that plot (invisibly!) to the right of the maximum x axis displayed!

Next we clearly see that the baseline level is too high and is intersecting the PHA peak. Why is that a problem?  Because if the peak shifts to the right or left, the baseline level cuts off more or less of the PHA peak and we obtain a non-linear response in our counting electronics!

However, we can make two adjustments to avoid this: first by lowering our baseline level below the PHA peak, and second, increasing our gain (for Cameca instruments) or our bias (for JEOL instruments) we can shift the PHA peak to the right *completely above* the baseline level, thus ensuring a linear response in our counting electrics. 

Here is the PHA scan after adjusting our baseline level and our gain (or bias):



Better, but the baseline level still intersects the PHA peak tail a bit on the left.  After raising our PHA gain still further, we obtain this PHA scan:



This is what we will utilize for our quantitative analyses in the next post.  The same PHA tuning procedure was utilized for Fe Ka as seen here:




Might be missing something but this is how JEOL have always recommended PHA - set to 4 volts. And idealy want to include escape peak, if not exclude escape peak, but ensure baseline is not on peak, it's what Paul Carpenter also has been talking about for decades.

Probeman

Quote from: Ben Buse on June 17, 2026, 06:15:10 AMMight be missing something but this is how JEOL have always recommended PHA - set to 4 volts. And idealy want to include escape peak, if not exclude escape peak, but ensure baseline is not on peak, it's what Paul Carpenter also has been talking about for decades.

Yes, you are missing something.    :)

For one thing these are Cameca plots and the Cameca PHA only goes to 5 volts. So on a JEOL instrument, you would be setting the PHA peak to 8 volts!  Do you get it now?   :D

In addition, with this integral-baseline method you *always* want to include the escape peak no matter how high one has to amplify the PHA.  Not having the escape peak above the baseline means that at some low enough count rate the escape peak could shift to the right, and negating your counting linearity.

The more important point is that we now know that photons that appear "cut off" graphically are still all counted in integral PHA mode. I did not know that until a few years ago, did you?

The final point is that we no longer need to "count rate match" our unknowns to our standards and therefore we can finally begin using globally distributed synthetic end member composition standards.

Combining all these points together will allow us as a global community to finally attain ~1% (or better) relative accuracy throughout the community.

I believe that doing all this (better PHA tuning, better dead time calibrations, better matrix corrections and better globally distributed high purity synthetic standards), as a community, will finally allow us to produce compositions that are in agreement with each other. Because that is certainly not the case today:

Wieser, P. E., Kent, A. J., Till, C. B., Donovan, J., Neave, D. A., Blatter, D. L., & Krawczynski, M. J. (2023). Barometers behaving badly I: assessing the influence of analytical and experimental uncertainty on clinopyroxene thermobarometry calculations at crustal conditions. Journal of Petrology, 64(2), egac126.
The only stupid question is the one not asked!

Ben Buse

Thanks John for explaining. By "count rate match" you mean changing the beam current so count rate same on standards and unknowns? Might be tricky for some multi element phases.

Probeman

Quote from: Ben Buse on June 17, 2026, 08:11:21 AMThanks John for explaining. By "count rate match" you mean changing the beam current so count rate same on standards and unknowns? Might be tricky for some multi element phases.

No, I do not.

I mean the widely utilized "practice" of selecting a natural material (no matter how poorly characterized or heterogeneous), to use as a standard to match the count rate observed in an unknown.

The only benefit being that then the dead time and PHA tuning (and matrix corrections) have small effects, but now the accuracy of the standard becomes dominant. But we also know that these natural "standard" materials are not good enough as has been documented by Vicenzi, Fournelle and Nachlas and others.

Of course one can also utilize different beam currents for the standard and unknown as one commonly does for trace element analysis, though accuracy in trace elements then depends on the picoammeter linearity.
The only stupid question is the one not asked!

Ben Buse


Ben Buse


Probeman

Quote from: Ben Buse on June 17, 2026, 08:18:29 AMGotcha. Thanks. I need to read more carefully!  :)

No worries. If this PHA tuning method was intuitive, I would have figured it out 40 years ago!

 :D
The only stupid question is the one not asked!

Probeman

Quote from: Ben Buse on June 17, 2026, 08:21:55 AMHas anyone tried it on a JEOL?

Anette did some testing a while back and found that the JEOL also counts photons above the JEOL 10v graphical limit like the Cameca above the 5v graphical limit.

But it would be great to see some high accuracy tests extrapolating Fe Ka from say Fe3O4 to the NIST glasses.

If one has calibrated their dead times, and their spectrometers are mechanically aligned to give an effective takeoff angle of 40 degrees, you should obtain ~1% relative accuracy even with a 10x difference in count rates. Test your effective takeoff angles using the Bragg order k-ratio tests described here:

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

Of course using the logarithmic dead time correction will help with large area Bragg crystals and their higher count rates, and when testing with soft x-rays, e.g., Mg, Al and Si Ka, be sure to use the FFAST MACs.  The DAM BSE correction is great for when large atomic number extrapolations are present.  I just always use FFAST with the Armstrong/Donovan and Moy matrix correction...

If using SiO2 as an Si Ka standard be sure to defocus the beam to 10 um or so. Do you have the FIGMAS MgO, Al2O3, MgAl2O4 mount?  Will Nachlas also has some excellent synthetic Mg2SiO4.  It's the Takai Mg2SiO4 recipe, which I have some of the original material from Takai back in the 1980s.

https://www.sciencedirect.com/science/article/abs/pii/0022024874901110
The only stupid question is the one not asked!