Apparent shift in Mg concentrations in standards on a new microprobe

[Probeman]

By the way, if one is concerned about the possibility of the MgO having a surface layer of hydrated MgO (brucite), and contributing to the low k-ratios in the previous post, that doesn't make sense because if the primary standard had a lower concentration of Mg, that would make the k-ratios in the Mg secondary standards higher, not lower.

Another point: why are we using Mg Ka and Al Ka for these tests?  It's because the absorption corrections for these extrapolations from their primary standards are quite large, which is why they were selected in the first place (since we are interested in testing the geometry/alignment, that is, the effective takeoff angle of our instrument!). Here is a matrix model for Al and Mg Ka in MgAl2O4:

 ELEMENT  ABSCOR  FLUCOR  ZEDCOR  ZAFCOR STP-POW BKS-COR   F(x)u      Ec   Eo/Ec    MACs
   Al ka  1.3734  1.0000  1.0450  1.4351  1.0647   .9815   .6461  1.5600  9.6154 1603.98
   Mg ka  1.3033   .9810  1.0120  1.2938  1.0233   .9889   .6648  1.3050 11.4943 1464.54
   O  ka  1.9089   .9988   .9574  1.8254   .9420  1.0164   .3671   .5317 28.2114 3976.77

Even better is the MgCr2O4 synthetic that Joe had obtained from Oak Ridge lab some time ago:

 ELEMENT  ABSCOR  FLUCOR  ZEDCOR  ZAFCOR STP-POW BKS-COR   F(x)u      Ec   Eo/Ec    MACs
   Cr ka   .9971  1.0000  1.0967  1.0935  1.1389   .9629   .9827  5.9900  2.5042 73.0353
   Mg ka  1.9447   .9996   .9507  1.8481   .9194  1.0340   .4456  1.3050 11.4943 3200.17
   O  ka  1.4877   .9937   .8922  1.3189   .8445  1.0565   .4711   .5317 28.2114 2754.86

An ~90% absorption correction is pretty impressive!  Does anyone know where we can obtain more of this MgCr2O4 synthetic material?

Anyway to continue... so during a discussion with Pete McSwiggen and Joe Boesenberg about what might be going on with these low Mg Ka k-ratios, Pete suggested we try and measure Mg Ka k-ratios using both 1st order and 2nd order Bragg reflections.  Because the k-ratios should be the same for both 1st and 2nd orders, if the spectrometer is properly aligned. Here are these nominal spectrometer positions as calculated by CalcZAF:

Spectro position for mg ka on TAP (140 mm), is 107.798 (with refractive index correction, k= 0.00218)
Spectro position for mg ka (II) on TAP (140 mm), is 215.2433 (with refractive index correction, k= 0.00218)

What a brilliant idea by Pete! I suggest we call this the Bragg order k-ratio test and include these in our instrument acceptance testing from now on:

https://epmalab.uoregon.edu/pdfs/Donovan_Community%20Specifications%20for%20EPMA-SEM.pdf

also see the topic here:

https://smf.probesoftware.com/index.php?topic=369.0

So then Joe tuned up Al and Mg Ka (I) and Al and Mg Ka (II) and here is what he obtained for Al Ka extrapolating from Al2O3 to MgAl2O4:



So pretty reproducible!  Next for Mg Ka:



Ok, I'd say there's a spectrometer/crystal alignment problem that gets worse at higher sin thetas!

The point being that whatever the two material compositions actually are, the k-ratios for these two different Bragg orders should be the same within statistics.

JEOL is now checking Joe's spectrometers, but in the meantime if anyone would like to perform some of these Bragg order k-ratio tests themselves, we would be very interested in seeing what you obtain!

And do check this new topic devoted to this spectrometer Bragg order-ratio alignment testing:

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

[Joe Boesenberg]

Hi All

So I thought I would update you on the microprobe adventure/disaster I had. As many of you know, a new JEOL ISP-100 was installed at Brown in Sept-Oct. 2024. After several delays, the microprobe and Bruker EDS were installed by mid-December. Scott Boroughs visited in early January 2025 to install PfE, which was essentially the last part of the installation. While teaching me PfE, Scott corrected the dead times on the spectrometers for me, and then we started doing quantitative test analyses on olivines and pyroxenes. Problems showed up immediately. We were getting low totals (low 98s), with Mg consistently low (4-5% relative), Si sometimes coming out high and sometimes low, and Al seemed okay nearly every time. We couldn't explain what was going on. Scott left after the PFE installation and I talked to JEOL about my poor analyses, who basically suggested it was a problem with my standards, not my spectrometers. I did some additional analyses and soon realized the errors with Si and Mg on TAPL were duplicated on the TAPJ. Both crystals were giving the same wrong answers and to the same magnitude. Sometime around this point I chimed in on the forum and started asking questions and many of you suggested various tests. I re-carbon coated my standards in the meantime and initially thought I had solved my problems. I soon realized however that I had simply had a fortuitous accident in analysis that I misinterpreted. So I continued to do various test analyses to no avail. In February, JEOL sent a service engineer to tweak the two TAP spectrometers and to see if they could find anything wrong. The engineer found nothing wrong, so nothing changed analytically. I was still always getting low totals and specifically low Mg. My suspicion was the TAP crystals were either somehow not properly aligned or the program that the JEOL engineers used to align the spectrometers was somehow giving consistent, but wrong results for the TAP crystals.

So after being told repeatedly that it was a problem with my standards through most of March, and now frustrated beyond compare, I asked John Donovan to look over my shoulder while I did some simple olivine analyses. He kindly agreed. My argument was simple. Even if many of my standards are wrong (and I have worked with many of these for near two decades), I should be able to take a synthetic Fo100 and a near endmember fayalite (Rockport), and get nearly 100% totals on any unknown olivine by analyzing Fe, Mg and Si. However, I could not. I kept getting totals of 97-98 and MgO was always low. John was quickly convinced something weird was going on, so he sent me the FIGMAS Al2O3-MgO-spinel standard block via Will Nachlas. When I couldn't get realistic results with those standards, he came to the conclusion that either the two TAP crystals, the column and/or the stage were misaligned. John then graciously talked to Pete McSwiggen at JEOL on my behalf and convinced him that the problem with my instrument was real.

At this point we started doing a test designed by Pete and discussed by John in this forum, the so called Bragg k-ratio test, which involves looking at both the first and second order peaks for both Mg and Al on the TAP crystals. In March, this is what my TAPL looked like using those standards.



Edit by John: The Bragg order k-ratio test that Joe is referring to is discussed here in more detail:

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

The k-ratio value that most labs got for the first order Mg, when they did the round robin with this standard block, was 0.272. The ideal Al value should be 0.605. As you can see the first order Mg is a bit low, but the second order Mg ratio is quite a bit off (and should be theoretically be giving the same ratio value as the first order), while Al appears nearly perfect.

Given the shift on the Bragg test, Pete talked to the JEOL engineers, and they came back in late April to my lab to re-align the two TAP crystals. After they left, I re-ran the Bragg k-ratio tests only to discover that both crystals were still misaligned, and were now slightly worse than before.






Now in May, John and I had a meeting with about 7 people from JEOL via Zoom. The conclusion of the meeting was Pete, as the JEOL microprobe expert, needed to do extensive testing on the instrument to determine the problem. Meanwhile, all of the data from the tests I had run were sent to the engineers main JEOL office in Japan.

So Pete basically started from scratch and documented all of the tests and parameters on the probe. He analyzed a bunch of olivines and re-did the Bragg tests. One of the issues Pete found with the Bragg test was that there was a peak shift in Mg on the MgO versus the spinel. The effect is not drastic, but certainly was making the problem look mildly worse than what it was.

Meanwhile, both John and myself had heard many good things about one of the JEOL engineers, John Glass. I had mentioned John G. to the JEOL service manager, Keith, at some point, and by sheer dumb luck, John G. was sent to my lab in June to attempt to re-align the two TAP crystals for a third time. After re-aligning the TAPL, we initially didn't think much had changed. We ran the Bragg test while John G. was there. The shift difference between the first and second order had certainly gotten better, but the second order was still about 1% low.



Part of the improvement was we were now correcting for the Mg peak shift. John G. left thinking nothing had improved substantially. Before he left, however he mentioned to me that he had done a minor manual re-alignment of the electron column. The next day, almost on a lark, I did a simple three element olivine analysis, wondering how far off MgO still was. To my surprise, the totals were now all clustering around 100%, and the MgO values for the olivines were realistic. John G. came back the following week and re-aligned the TAPJ, and after testing, showed it is nearly perfectly align with respect to both the first and second order peaks of Mg.



Subsequent analysis on olivines showed everything was now analytically reasonable.


Several quandaries remain:

1) What changed during the final re-alignment visits? Was the TAPL crystal that far off from alignment or was there something else weird going on in the column (like a stuck condenser lens), that got freed up when John G. manually re-aligned the column? Up to that time, I had been using the auto alignment function for the tilt and shift of the beam. not manual.

2) Why is there still a 1% shift in Mg for the first order and second order ratios on the TAPL? Is this simply the limit of the alignment accuracy given the rather coarse adjustments possible in aligning the crystal and using the JEOL alignment program?

3) The theoretical value based on CalcZAF, looking at Mg and using MgO and spinel at 15kv should give a k-ratio value of 0.279, not 0.272, which is the avg for all of the FIGMAS labs. A ratio of 0.272 actually correlates with a takeoff angle of around 15 degrees, not 40. Why the difference?

4) Why were the Al ratios always correct, while Mg varied with every new re-alignment attempt?

In the end, the lab is now getting realistic quantitative results, but it took an extra six months, John Donovan, Pete McSwiggen, John Glass, and a whole lot of persistence on my part.

Thanks
Joe

[Joe Boesenberg]

All

I forgot to mention, following all of the re-alignments, both TAP crystals passed whatever spectrometer qualifications the JEOL engineers use to "certify" the spectrometer. The problem I saw is that they appear to use the intensities of certain elements on specific standards, rather than any sort of alignment test or analytical test to pass the spectrometer/crystal.

Cheers
Joe

[Probeman]

Subsequent analysis on olivines showed everything was now analytically reasonable.

Several quandaries remain:

1) What changed during the final re-alignment visits? Was the TAPL crystal that far off from alignment or was there something else weird going on in the column (like a stuck condenser lens), that got freed up when John G. manually re-aligned the column? Up to that time, I had been using the auto alignment function for the tilt and shift of the beam. not manual.

I find it difficult to believe that the column alignment could be the cause of this problem. It's true that there was an incident back in the 1990s where Cal Tech discovered on their new JEOL 733 that the column had been mounted off-center from the spectrometer circle, so opposing spectrometers produced k-ratios that were either too high or too low. But that was a mechanical offset on the order of many millimeters, so probably a degree of misalignment that could not be duplicated by a mere lens alignment issue:

https://smf.probesoftware.com/index.php?topic=1466.msg11143#msg11143

But it's probably worth recording the current condenser lens values, then run an auto-alignment and re-measure the FIGMAS mount k-ratios.

2) Why is there still a 1% shift in Mg for the first order and second order ratios on the TAPL? Is this simply the limit of the alignment accuracy given the rather coarse adjustments possible in aligning the crystal and using the JEOL alignment program?

Good questions. It would be interesting to hear from John Glass how the spectrometer alignments performed by him differed from the methods used by the previous technicians...

3) The theoretical value based on CalcZAF, looking at Mg and using MgO and spinel at 15kv should give a k-ratio value of 0.279, not 0.272, which is the avg for all of the FIGMAS labs. A ratio of 0.272 actually correlates with a takeoff angle of around 15 degrees, not 40. Why the difference?

Joe is referring to the FIGMAS round robin k-ratio measurements that were performed on these materials a couple of years ago by a number of labs globally:



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

The theoretical value from CalcZAF very much depends on the matrix correction utilized. Badgerfilm (using XPHI?), predicts a value of around 0.272 which is pretty darn close the average of the experimental measurements. The value of 0.279 from CalcZAF is only about 2.5% different from 0.272 but it very much depends on what specific matrix correction (and MACS) are selected in CalcZAF (which ones did you select?).

And looking at all matrix corrections in an elemental calculation of MgAl2O4 (relative to the pure elements) we get these theoretical k-ratios:

Mg2AlO4, sample 1
40 degrees and 15 keV
LINEMU   Henke (LBL, 1985) < 10KeV / CITZMU > 10KeV
            "Al K-RAT"    "Mg K-RAT"    "O K-RAT "
     1     .113792     .265524     .266787    Armstrong/Brown/Scott-Love (prZ)
     2     .112964     .267640     .265805    Philibert/Duncumb-Reed
     3     .110439     .260813     .273097    Heinrich/Duncumb-Reed
     4     .111961     .263550     .264162    Love-Scott I
     5     .113929     .266351     .265634    Love-Scott II
     6     .113103     .264478     .281269    Packwood Phi(prZ) (EPQ-91)
     7     .108068     .251155     .267156    Bastin (original) (prZ)
     8     .106716     .255334     .260955    Bastin PROZA Phi (prZ) (EPQ-91)
     9     .108120     .257614     .260070    Pouchou and Pichoir-Full (PAP)
    10     .109804     .259592     .268052    Pouchou and Pichoir-Simplified (XPP)
    11     .113928     .265921     .265513    PAP/Donovan and Moy BSC/BKS (prZ)

AVER:      .111166     .261634     .267136
SDEV:      .002662     .005235     .005830
SERR:      .000802     .001579     .001758

MIN:       .106716     .251155     .260070
MAX:       .113929     .267640     .281269

Looking at the MIN/MAX for Mg Ka we see about a 7% difference! Even just comparing two modern matrix corrections (Armstrong-Brown and PAP), we see a 3% difference.  So I wouldn't worry too much about this particular theoretical k-ratio from CalcZAF.

4) Why were the Al ratios always correct, while Mg varied with every new re-alignment attempt?

It's another good question. When I performed my own Bragg order k-ratio measurements, I saw both Mg and Al Bragg orders k-ratios looking consistent on two of my TAP spectrometers, but inconsistent only for Mg on one of them:

https://smf.probesoftware.com/index.php?topic=1739.msg13407#msg13407

Why would Mg Ka k-ratio be so much more sensitive to the spectrometer alignment? It's at a higher sin theta than Al Ka and the absorption correction is smaller than for Al Ka...

In the end, the lab is now getting realistic quantitative results, but it took an extra six months, John Donovan, Pete McSwiggen, John Glass, and a whole lot of persistence on my part.

It takes a village!