Hi Julien,
I merged your topic with an existing topic on the alternating on/off background method. The answer to your regression question is in the posts above.

That said, I agree that it is a little misleading to show the regression of the alternating on and off-peak (and P-B) measurements, since the regression is not actually utilized in the background correction. It's just displayed as a aid for visualization.

But, I think your alternating on/off measurements look to be doing exactly what they should be doing.  That is tracking the relative changes in the on and off-peak intensities over time. And as shown in your on and high and low intensity plots, we see significant drift in the x-ray counts as the glass is damaged.   This is a very appropriate application for this acquisition method.

The beauty of the alternating on/off method is that it tracks these intensity drifts through the entire acquisition, therefore characterizing the thing we actually care about, which is the delta between the on and the off-peak intensities (i.e., the net intensity). Look at the P-B plot. The fact that the P-B plot shows more scatter is exactly what one would expect when subtracting two noisy signals from each other.  That is, the noise increases in quadrature.  In fact what is particularly wonderful about your particular P-B plot example is that it shows that in spite of the large changes over time in the raw data, the P-B values are relatively constant, albeit a little noisy. So I think this is working exactly as intended.

Now, as a reminder, these P-B values are recorded (for documentation purposes) at acquisition time, but the sum of all the sub intervals for the on and high and low off-peak intensities are also summed and stored at acquisition time for subsequent quantification. The basic idea is that one is spreading the beam damage effects on intensity over both the on and off-peak measurements as the sample is damaged, thus producing a more accurate trace element measurement.

You are a coder, so if you are interested in the gory details, I can send you the acquisition code module (it is over 30K characters so I can't post it here).

When I activated the blank correction, it warned me that maybe I should remove the peak interference correction to avoid an over correction. I can understand this is correct (that it will overcorrect) IF the blank standard AND the unknown have exactly the same Ti content, but… The blank correction reference material and the unknown samples have different Ti-content (up to 0.5-1.0 wt% difference)! Can the blank correction handle this? Would you say this is not possible at all to correct for this?

Hi Julien,
You always have such great questions!   

I have to admit that this is a topic that I haven't thought about as much as it probably should be.   It's complicated, at least to me.  But in fact in thinking about this a bit more today, I've decided that maybe the warning you are getting about applying both an interference correction and a blank correction may not even be necessary!  I'm not exactly sure, but keep reading. In any case it's always best to check using actual data to check one's assumptions!

So I recently was testing some integrated EDS and WDS analyses where I used WDS for the trace elements and EDS for the major elements, the idea being to reduce the number of elements by WDS, thus speeding up the analysis and damaging the sample less, as described here a few months ago for those that are interested:

https://probesoftware.com/smf/index.php?topic=79.msg7818#msg7818

As you know for some situations, e.g., Rb or Sr La are interfered by Si, so it's important that we can correct for spectral interferences between WDS *and* EDS elements, particularly if one is doing traces by WDS and major elements by EDS!  Anyway, the point being that we can do a quick test using this data set for the interference correction and the blank correction separately, and also together, to try and  answer your questions. It's not a perfect test because there were a limited number of samples, but let's see what we find anyway... please note I'm going to pitch this post to a general audience so I'm going to go over some issues that I know you are very much aware of, probably much more than myself!

So (starting at the beginning) if we do not apply the interference correction for Si interfering with Sr La, nor apply a blank correction, we get the following results on a SiO2 standard (which has zero Sr from ICP-MS by Alan Koenig):

ELEM:       Na      Ba      Rb      Sr      Fe      Si      Al       K       O   SUM 
    60    .001    .002   -.054    .063    .014  46.717    .048    .000  53.282 100.074
    61    .002    .000   -.036    .076    .011  46.971    .045    .000  53.571 100.640
    62    .000   -.008   -.053    .068   -.009  46.890    .042    .000  53.466 100.396
    63    .002    .006   -.057    .064    .000  46.767    .035    .000  53.323 100.139

AVER:     .001    .000   -.050    .067    .004  46.836    .043    .000  53.410 100.313
SDEV:     .001    .006    .010    .006    .010    .115    .006    .000    .133    .259

and here for an orthoclase standard (which has 12 PPM (0.0012 wt.%) Sr from isotope dilution by John Christensen):

ELEM:       Na      Ba      Rb      Sr      Fe      Si      Al       K       O   SUM 
    56    .718    .049    .039    .052   1.350  30.095   8.552  12.896  45.190  98.942
    57    .747    .071    .027    .043   1.320  30.075   8.544  12.888  45.160  98.875
    58    .753    .060    .015    .056   1.344  30.123   8.575  12.930  45.259  99.115
    59    .731    .064    .017    .039   1.331  29.870   8.467  12.808  44.837  98.165

AVER:     .737    .061    .025    .048   1.336  30.041   8.535  12.881  45.111  98.774
SDEV:     .016    .009    .011    .008    .013    .115    .047    .052    .188    .419

Please ignore the Rb results.  It would be nice to also check the Rb values but there clearly is a problem with the background measurement, so at least for now, let's just focus on these two Sr measurements.

So it would appear that there is an interference from Si on the Sr La emission line. First because the SiO2 shows a greater amount of Sr than the orthoclase, but also because our nominal overlap calculation model from the Elements/Cations dialog in Probe for EPMA shows this:

For Sr la   LPET at  6.86280 angstroms, at an assumed concentration of 1 wt.%
  Interference by Si SKB`           at  6.81610 ( 77894.2) ( -533.78) =      2.0%
  Interference by Rb LB4            at  6.82360 ( 77979.9) ( -448.06) =     16.5%
  On Peak Position   -------------  at  6.86280 ( 78428.0)
  Interference by K  SKB``    II    at  6.88250 ( 78653.2) ( 225.180) =     10.0%
  Interference by K  SKB^5    II    at  6.89860 ( 78837.2) ( 409.211) =      2.0%
  Interference by K  KB1      II    at  6.90910 ( 78957.2) ( 529.227) =      4.5%
  Interference by K  KB3      II    at  6.90910 ( 78957.2) ( 529.227) =      2.5%

When one has an interfering emission line from a major element, those tails can extend a long way...

Now, if we apply the interference correction in Probe for EPMA for Sr interfered by Si using the SiO2 standard as the standard for the interference correction, we get the following results for SiO2 measured as an unknown:

ELEM:       Na      Ba      Rb      Sr      Fe      Si      Al       K       O   SUM 
    60    .001    .002   -.054   -.004    .014  46.721    .048    .000  53.274 100.003
    61    .002    .000   -.036    .008    .011  46.975    .045    .000  53.563 100.569
    62    .000   -.008   -.053    .000   -.009  46.894    .042    .000  53.458 100.325
    63    .002    .006   -.057   -.004    .000  46.771    .035    .000  53.316 100.069

AVER:     .001    .000   -.050    .000    .004  46.840    .043    .000  53.403 100.242
SDEV:     .001    .006    .010    .006    .010    .115    .006    .000    .133    .259

Which is pretty darn good, though expected since the SiO2 standard and the SiO2 unknown were the same material, though measured in different spots on the standard.  And here is the orthoclase standard measured as an unknown:

ELEM:       Na      Ba      Rb      Sr      Fe      Si      Al       K       O   SUM 
    56    .718    .049    .039    .009   1.350  30.098   8.553  12.896  45.186  98.898
    57    .747    .071    .027    .000   1.320  30.077   8.545  12.888  45.155  98.830
    58    .753    .060    .015    .013   1.344  30.126   8.575  12.930  45.255  99.071
    59    .731    .064    .017   -.004   1.331  29.873   8.467  12.808  44.832  98.121

AVER:     .737    .061    .025    .004   1.336  30.043   8.535  12.881  45.107  98.730
SDEV:     .016    .009    .011    .008    .013    .115    .047    .052    .188    .418

The isotope dilution gave us 12 PPM Sr, but our variance is 80 PPM, so statistically a zero concentration.

Now lets turn off the interference corrections and turn on the blank correction instead. Again, here is SiO2 as an unknown using itself for the blank correction:

ELEM:       Na      Ba      Rb      Sr      Fe      Si      Al       K       O   SUM 
    60    .001    .002   -.054   -.004    .014  46.721    .048    .000  53.274 100.003
    61    .002    .000   -.036    .008    .011  46.975    .045    .000  53.563 100.569
    62    .000   -.008   -.053    .000   -.009  46.894    .042    .000  53.458 100.325
    63    .002    .006   -.057   -.004    .000  46.771    .035    .000  53.316 100.068

AVER:     .001    .000   -.050    .000    .004  46.840    .043    .000  53.403 100.242
SDEV:     .001    .006    .010    .006    .010    .115    .006    .000    .133    .259

As expected, and as near as we can tell we get zero. Which is also the same result we got for the interference correction. But that is merely because we assigned the SiO2 unknown as the blank correction to itself.  Now let's look at the orthoclase standard using the same SiO2 unknown for the blank correction:

ELEM:       Na      Ba      Rb      Sr      Fe      Si      Al       K       O   SUM 
    56    .718    .049    .039   -.015   1.350  30.099   8.553  12.896  45.183  98.873
    57    .747    .071    .027   -.024   1.320  30.079   8.545  12.888  45.153  98.806
    58    .753    .060    .015   -.011   1.344  30.127   8.575  12.930  45.252  99.046
    59    .731    .064    .017   -.029   1.332  29.875   8.468  12.808  44.830  98.096

AVER:     .737    .061    .025   -.020   1.336  30.045   8.535  12.881  45.105  98.705
SDEV:     .016    .009    .011    .008    .013    .115    .047    .052    .188    .419

Whoa!  Now we are getting a *negative* 200 PPM for Sr.  It's a significant *over correction* clearly. Why would that be?

Well, the interference correction is based on the actual concentration of the interfering element, in this case Si. And since there is an actual interference here, the interference correction handles the situation very well.  But the blank correction is *not* based on the concentration of any particular element.  It merely assumes that there is some sort of measurement artifact, what exactly we may not know, but it assumes that the measurement artifact is constant regardless of composition. Perhaps something like a detector absorption edge or a secondary Bragg reflection artifact like this:



So we would expect that since the concentration of Si, which is causing the interference, is different for SiO2 and orthoclase, that the blank correction would be unsuitable for this situation and indeed that appears to be the case.

OK, but what happens if we apply *both* the interference correction and the blank correction to these samples? Here is the SiO2 standard with the interference correction for Si on Sr La using SiO2 as our interference standard and also the blank correction using the SiO2 unknown as the blank correction sample:

ELEM:       Na      Ba      Rb      Sr      Fe      Si      Al       K       O   SUM 
    60    .001    .002   -.054   -.004    .014  46.721    .048    .000  53.274 100.003
    61    .002    .000   -.036    .008    .011  46.975    .045    .000  53.563 100.569
    62    .000   -.008   -.053    .000   -.009  46.894    .042    .000  53.458 100.325
    63    .002    .006   -.057   -.004    .000  46.771    .035    .000  53.316 100.069

AVER:     .001    .000   -.050    .000    .004  46.840    .043    .000  53.403 100.242
SDEV:     .001    .006    .010    .006    .010    .115    .006    .000    .133    .259

And since we're using the SiO2 standard as the standard for the interference correction and we're using the SiO2 unknown as the blank correction sample, we get zero as expected. But now let's try the orthoclase standard with both corrections turned on:

ELEM:       Na      Ba      Rb      Sr      Fe      Si      Al       K       O   SUM 
    56    .718    .049    .039    .009   1.350  30.098   8.553  12.896  45.186  98.898
    57    .747    .071    .027    .000   1.320  30.077   8.545  12.888  45.155  98.830
    58    .753    .060    .015    .013   1.344  30.126   8.575  12.930  45.255  99.071
    59    .731    .064    .017   -.004   1.331  29.873   8.467  12.808  44.832  98.121

AVER:     .737    .061    .025    .004   1.336  30.043   8.535  12.881  45.107  98.730
SDEV:     .016    .009    .011    .008    .013    .115    .047    .052    .188    .418

Weird!  We're getting the same result we got for the interference correction only! How is that possible?  Well a closer look at the blank correction "value" on this orthoclase analysis shows us why:

ZCOR:   1.8699  1.3745  1.2860  1.2359  1.1909  1.2191  1.2592  1.1361     ---
KRAW:    .0536   .0006   .0006   .0001   .0166   .6009   .5086  1.0017     ---
PKBG:    12.74    1.19    1.07    1.06   11.90     .00     .00     .00     ---
INT%:     ----    ----    ----  -92.75    ----    ----    ----    ----     ---
BLNK#:    ----    ----    ----       9    ----    ----    ----    ----     ---
BLNKL:    ----    ----    ---- .000000    ----    ----    ----    ----     ---
BLNKV:    ----    ----    ---- .000070    ----    ----    ----    ----     ---

Two things to look at here. First the interference correction shows a -92% correction leaving us with 40 PPM of Sr. Now our variance is 80 PPM so the 40 PPM result is statistically a zero, but it is a tiny bit suggestive, because according to isotope dilution do we have 12 PPM of Sr in this orthoclase standard. So a higher precision measurement is necessary before we proceed with any further speculation, and I would probably use the MAN method for best trace element precision combined with a blank correction for an accuracy correction, so next time I get a chance I'll try that measurement.

Second, look at the blank correction value (BLNKV). It's 0.00007 wt% or 0.7 PPM! What does this mean?  It means that after the interference correction is applied to the sample, the blank correction calculates the difference between what we measured and what we should have obtained (that's the BLNKL or blank correction level), and that was only 0.7 PPM or essentially zero. So when we apply that blank correction of 0.7 PPM to either the SiO2 as an unknown or the orthoclase as an unknown, there is essentially no effect on the data.

I know, it hurts my brain too but it actually makes sense.  I would be very interested in hearing from you (Julien), and/or any one else on what you find on your samples. Please try some tests on some well characterized standards that have trace elements measured and let's see what we find.

Hi Jakub,
It's a good question. But there are already 10 matrix corrections in CalcZAF/Probe for EPMA, of which several give excellent results. Frankly I think that we have bigger problems in EPMA with beam sensitive samples, peak shape issues, background measurements and interference corrections, particularly the last two as you correctly point out.

I read about Merlet's X-Phi algorithm at the time he presented it, but he did not want to share his source code and in any case I no longer know how to reach him now that he retired.  But to be honest I'm not sure it is any improvement over the Armstrong/Bastin, PAP and XPP models so it would be a very low priority for me. 

Dear John Donovan, thank you very much for your answer.
I did not want to assign you a homework to build a code with X-Phi correction. I was only interested in the reason, because this algorithm is used in the default Cameca software, at least in the version we have. So, to know, if there are any reasons against the algorithm of Merlet. So, it is not possible to run the same correction algorithm in both, CALCZAF and PeakSight; but it is no problem. Now, I understand the reason.
Best regards, Jakub Haifler. 

Hi Nick
Resurrecting an old post...but I now have exactly the same problem with my XFlash detector, in that i have exactly zero count rates and it is exhibiting exactly the same behaviour. Can I ask what the end result was of your problem? Was it electronics in the box?

Cheers

This is all great information and I think will be very helpful as I build a "probe of dreams" instrument. Thanks all for your input!!

I will compile my email responses and share them here as well.

Dawn