Bragg order k-ratio testing

[sem-geologist]

We have a mystery that I need your help with...

My preliminary calculation from wavescan taken at 25kV some time ago on our SXFiveFE for Mg 2nd order K-ratio in MgO/MgAl2O4 is 0.267 which seems to be close  in agreement what you got with 1st order and second order for 3rd and 4th spectrometers.

Again, You are right – the background positions should not influence k-ratios up to 6%. I still think there could be some spectrometer resolution differences between those which could cause your observations. At 1st order all the measurements from all spectrometers will be similar as KA1 and KA2 will be convolved into visible and mathematically indistinguishable single peak (also amplitude includes the tail  from KA3). On 2nd order the Mg KA peak asymmetry is starting to show up due to KA1 and KA2 small energy difference (there is still not clearly observable 2 peaks, but shape of the peak (a buldge) starts to show up, depending from resolution and lack of segmentation of TAP).

Proposition for experiment: Instead of calculating k-ratios on measurements made with Pk-bkgd method I would suggest try doing k-ratios from measured integral/area of the peak.

My other experiences (which somehow is connected with issue):
* Some time ago due to extreme overlap-hell (42+ measured elements) (before Peaksight introduced possibility of negative interference correction) I was for some time measuring Si Ka on the second order line to get away with it. It however appeared to be much much much less stable than the first order, and required constant re-calibration, thus I had dropped that completely after upgrading to Peaksight 6.5. As far I could remember, both amplitude and peak shift was very unstable.
* fluorine measurement on TAP vs PC0, or "when worse spectral resolution "is better"". The major problem with fluorine on TAP is that its shape will change from measured phase to phase, and pk-bkgd measurement method is not reliable (CaF2, Topaz-F, Biotite-F, Apatite-F – every phase would have different shape and standardization based on pk-bkgd would not work). On PC0 standardization works as expected and i.e. flourine can be mesured in CaF2 and Biotite-F and Apatite-F using the same Topaz-F standard. That is thanks to worse spectral resolution, where all those minute details of F KA shape observed on TAP is merged/convolved into single Voigt function shape on PC0 - and thus pk-bkgd due to worse resolution then works correctly on PC0, where due to too good spectral resolution on TAP does not work. TAP could work in limited cases for Florine if there is possibility to use peak/area method.

[Probeman]

Proposition for experiment: Instead of calculating k-ratios on measurements made with Pk-bkgd method I would suggest try doing k-ratios from measured integral/area of the peak.

My other experiences (which somehow is connected with issue):
* Some time ago due to extreme overlap-hell (42+ measured elements) (before Peaksight introduced possibility of negative interference correction) I was for some time measuring Si Ka on the second order line to get away with it. It however appeared to be much much much less stable than the first order, and required constant re-calibration, thus I had dropped that completely after upgrading to Peaksight 6.5. As far I could remember, both amplitude and peak shift was very unstable.
* fluorine measurement on TAP vs PC0, or "when worse spectral resolution "is better"". The major problem with fluorine on TAP is that its shape will change from measured phase to phase, and pk-bkgd measurement method is not reliable (CaF2, Topaz-F, Biotite-F, Apatite-F – every phase would have different shape and standardization based on pk-bkgd would not work). On PC0 standardization works as expected and i.e. flourine can be mesured in CaF2 and Biotite-F and Apatite-F using the same Topaz-F standard. That is thanks to worse spectral resolution, where all those minute details of F KA shape observed on TAP is merged/convolved into single Voigt function shape on PC0 - and thus pk-bkgd due to worse resolution then works correctly on PC0, where due to too good spectral resolution on TAP does not work. TAP could work in limited cases for Florine if there is possibility to use peak/area method.

I will do the integrated area measurement, but I very much doubt there's a peak shape change because the bonding physics is the same in both measurements.

More importantly, if the change between 1st and 2nd order Bragg k-ratios is due to a change in spectrometer resolution, why does this effect only appear in one spectrometer?

[Probeman]

Continuing from the previous post...

So what is causing the Mg Ka k-ratios on spectrometer 1 to be so different for the 1st and 2nd order measurements?  Using the CalcZAF effective angle k-ratio calculator I get these results for Mg Ka at 20 keV for MgAl2O4 as the secondary standard and MgO as the primary standard:

Effective K-Ratios for Primary Standard: 3012 MgO FIGMAS
Secondary Standard: 3100 MgAl2O4 FIGMAS
Emission line: Mg ka at 20 keV
Absorption Correction Method: Phi(pz) Absorption of Armstrong/Packwood-Brown 1981 MAS
MAC File: LINEMU  Henke (LBL, 1985) < 10KeV / CITZMU > 10KeV

Absolute k-ratio change per degree at 40 degrees:  .000294
Percent (relative) k-ratio change per degree at 40 degrees:  .107318

Takeoff Angle:  35.0000, K-Ratio:  .271588
Takeoff Angle:  35.5000, K-Ratio:  .271767
Takeoff Angle:  36.0000, K-Ratio:  .271943
Takeoff Angle:  36.5000, K-Ratio:  .272115
Takeoff Angle:  37.0000, K-Ratio:  .272283
Takeoff Angle:  37.5000, K-Ratio:  .272448
Takeoff Angle:  38.0000, K-Ratio:  .272609
Takeoff Angle:  38.5000, K-Ratio:  .272767
Takeoff Angle:  39.0000, K-Ratio:  .272922
Takeoff Angle:  39.5000, K-Ratio:  .273073
Takeoff Angle:  40.0000, K-Ratio:  .273221
Takeoff Angle:  40.5000, K-Ratio:  .273367
Takeoff Angle:  41.0000, K-Ratio:  .273509
Takeoff Angle:  41.5000, K-Ratio:  .273648
Takeoff Angle:  42.0000, K-Ratio:  .273784
Takeoff Angle:  42.5000, K-Ratio:  .273918
Takeoff Angle:  43.0000, K-Ratio:  .274049
Takeoff Angle:  43.5000, K-Ratio:  .274177
Takeoff Angle:  44.0000, K-Ratio:  .274302
Takeoff Angle:  44.5000, K-Ratio:  .274425
Takeoff Angle:  45.0000, K-Ratio:  .274546

The thing is, if I plot this range of Bragg (1st and 2nd) order k-ratios from the new data as seen here:



there's no reasonable change in effective takeoff angle that could explain this much of a variation in the k-ratios for the 1st and 2nd order reflections on spectrometer 1 for Mg Ka.

Note: the absolute values of the calculated k-ratios (e.g., the horizontal green lines), will depend on the specific matrix corrections and MAC tables selected. 

And from a quant perspective, my spectrometer 1 shows a ~3.5% difference for Mg Ka, which is similar to what Joe Boesenberg at Brown is seeing on both his TAP spectrometers, but I'm seeing only on one of my three TAP spectrometers...

So what could be causing this?  Ideas?

[Probeman]

Here's the complete data from both sets of Bragg order k-ratio testing I did a few weeks ago...

To review, we've all known about the simultaneous k-ratio tests, which attempts to determine if the k-ratios produced from our spectrometers agree with each other within statistics. And the physics indicates that these 1st and 2nd order k-ratios should be statistically similar.

And of course that extends to comparisons with spectrometers on other instruments in other labs!  That is the essence of the "consensus k-ratio" tests that have been proposed by several of us:

https://smf.probesoftware.com/index.php?topic=1239.0

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

For such testing of the consistency of spectrometer k-ratios on your own instrument, I suggest running the "constant k-ratio" test which tests three separate aspects: dead time, picoammeter linearity and (if you measure your k-ratios on more than one spectrometer) also simultaneous k-ratios, depending on how you assign the primary standards to your data:

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

But how do we know if a single spectrometer is producing consistently accurate k-ratios?  Well, Peter McSwiggen suggested we measure k-ratios for both 1st order reflections and also the 2nd order reflections as described here at the beginning of this topic:

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

Because regardless of the Bragg reflection, we should obtain the same k-ratios within statistics.  And there's two things to consider: first the consistency between the 1st and 2nd order k-ratios, but also, how do the k-ratios we measure compare to our physics models in an absolute accuracy sense?  Yes, it's a bit circular...

Here is a plot for Mg Ka at 20 keV, again showing the range of k-ratios from PAP/FFAST k-ratios from 35 to 45 degrees:



Clearly Spec 1 has a problem with Mg, but the other two are quite reasonable. And here for Al Ka also at 20 keV:



The upper range for Al Ka is off the plot (45 degrees = ~ 0.570, but the measured data is mostly within the 35 degree limit.  As to the accuracy of the analytical calculations, that is the question...

What surprises me though, is how much these k-ratios vary, if the effective takeoff angle is the only variable here!

[Joe Boesenberg]

John

For your post of April 2, for the first and second order Al results, you get a value of about 0.562. I get a value of about 0.605 for all of my Bragg tests with spinel all the way back to March. Are we using such different conditions and correction procedures? Seems like this is really big difference.

Joe

[Probeman]

For your post of April 2, for the first and second order Al results, you get a value of about 0.562. I get a value of about 0.605 for all of my Bragg tests with spinel all the way back to March. Are we using such different conditions and correction procedures? Seems like this is really big difference.

Good eye!  Yes, the expected 15 keV k-ratio for Al Ka (Al2O3/MgAl2O4) is around 0.60:

Effective K-Ratios for Primary Standard: 3013 Al2O3 FIGMAS
Secondary Standard: 3100 MgAl2O4 FIGMAS
Emission line: Al ka at 15 keV
Absorption Correction Method: Phi(pz) Absorption of Armstrong/Packwood-Brown 1981 MAS
MAC File: LINEMU  Henke (LBL, 1985) < 10KeV / CITZMU > 10KeV

Absolute k-ratio change per degree at 40 degrees:  .001716
Percent (relative) k-ratio change per degree at 40 degrees:  .281803

Takeoff Angle:  35.0000, K-Ratio:  .597463
Takeoff Angle:  35.5000, K-Ratio:  .598536
Takeoff Angle:  36.0000, K-Ratio:  .599585
Takeoff Angle:  36.5000, K-Ratio:  .600608
Takeoff Angle:  37.0000, K-Ratio:  .601606
Takeoff Angle:  37.5000, K-Ratio:  .602581
Takeoff Angle:  38.0000, K-Ratio:  .603534
Takeoff Angle:  38.5000, K-Ratio:  .604464
Takeoff Angle:  39.0000, K-Ratio:  .605373
Takeoff Angle:  39.5000, K-Ratio:  .606261
Takeoff Angle:  40.0000, K-Ratio:  .607128
Takeoff Angle:  40.5000, K-Ratio:  .607976
Takeoff Angle:  41.0000, K-Ratio:  .608804
Takeoff Angle:  41.5000, K-Ratio:  .609614
Takeoff Angle:  42.0000, K-Ratio:  .610405
Takeoff Angle:  42.5000, K-Ratio:  .611178
Takeoff Angle:  43.0000, K-Ratio:  .611934
Takeoff Angle:  43.5000, K-Ratio:  .612674
Takeoff Angle:  44.0000, K-Ratio:  .613396
Takeoff Angle:  44.5000, K-Ratio:  .614103
Takeoff Angle:  45.0000, K-Ratio:  .614795

But for 20 keV the expected k-ratio for Al Ka (Al2O3/MgAl2O4) is around 0.56:

Effective K-Ratios for Primary Standard: 3013 Al2O3 FIGMAS
Secondary Standard: 3100 MgAl2O4 FIGMAS
Emission line: Al ka at 20 keV
Absorption Correction Method: Phi(pz) Absorption of Armstrong/Packwood-Brown 1981 MAS
MAC File: LINEMU  Henke (LBL, 1985) < 10KeV / CITZMU > 10KeV

Absolute k-ratio change per degree at 40 degrees:  .002098
Percent (relative) k-ratio change per degree at 40 degrees:  .372878

Takeoff Angle:  35.0000, K-Ratio:  .548901
Takeoff Angle:  35.5000, K-Ratio:  .550176
Takeoff Angle:  36.0000, K-Ratio:  .551424
Takeoff Angle:  36.5000, K-Ratio:  .552647
Takeoff Angle:  37.0000, K-Ratio:  .553845
Takeoff Angle:  37.5000, K-Ratio:  .555019
Takeoff Angle:  38.0000, K-Ratio:  .556169
Takeoff Angle:  38.5000, K-Ratio:  .557295
Takeoff Angle:  39.0000, K-Ratio:  .558398
Takeoff Angle:  39.5000, K-Ratio:  .559480
Takeoff Angle:  40.0000, K-Ratio:  .560539
Takeoff Angle:  40.5000, K-Ratio:  .561577
Takeoff Angle:  41.0000, K-Ratio:  .562594
Takeoff Angle:  41.5000, K-Ratio:  .563590
Takeoff Angle:  42.0000, K-Ratio:  .564567
Takeoff Angle:  42.5000, K-Ratio:  .565524
Takeoff Angle:  43.0000, K-Ratio:  .566462
Takeoff Angle:  43.5000, K-Ratio:  .567381
Takeoff Angle:  44.0000, K-Ratio:  .568282
Takeoff Angle:  44.5000, K-Ratio:  .569165
Takeoff Angle:  45.0000, K-Ratio:  .570029

Yes, my measurements here:

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

were performed at 20 keV...

Why? Because the 20 keV measurement is more sensitive to the effective takeoff angle, e.g., at 15 keV:

Percent (relative) k-ratio change per degree at 40 degrees:  .281803

and at 20 keV:

Percent (relative) k-ratio change per degree at 40 degrees:  .372878

[Probeman]

As some of you may know, Joe Bosenberg had some trouble getting quantitative results on his TAP spectrometers on his new JEOL iHP200F instrument:

https://smf.probesoftware.com/index.php?topic=1718.0

After six months of continued efforts by him and his local JEOL service engineer without resolution of the problem, Joe was able to get JEOL service engineer John Glass to come out to take a look. John was able to determine the problem and get Joe's spectrometers properly aligned:

https://smf.probesoftware.com/index.php?topic=1718.msg13502#msg13502

I wrote to John to ask him what exactly he did to get Joe's spectrometers aligned and I here post here (with permission) what he said:

Hi John,

The baseplate alignment is done using an alignment script, where it will defocus the stage, peak, and measure. Plots the measured counts out and fits a curve to determine where to adjust the baseplate.

The crystal is usually done with the alignment script in the same way. But I prefer to align the crystal without the script because I find it's faster, and improves reproducibility at the upper middle and upper end of the spectrometer.

I will focus on the standard (Si Metal for PET, Ti Metal for LiF, Mg Metal for TAP). I will then peak the spectrometer. I open up the chart recorder and adjust the count rate to at least 20kcps, usually shooting for the range of 20k-40kcps. I usually use 25kV for this.

Looking at the chart recorder, I will then defocus the stage, going +100um from just focus (I use +50um for LiF), and observe the count rate. Then I go -100um from just focus (-50um for LiF) and observe the count rate. I want to see less than a 1000 cps difference between the defocused count rates. If not, I will adjust the crystal at the -100um (-50um LiF) from focus spot and try to split the difference (usually need to go over a little bit). Then I repeak at just focus, and check again until I'm satisfied. If a large adjustment was done, I will redo the baseplate alignment.

It's also important to check the belt linearity, as this affects the alignment of the spectrometer between the baseplate and the crystal. The spec is a delta Z of 40um for standard spectrometers, and 60um for H type. But the better you get this, the better the performance and reproducibility will be. So even though the belt linearity passed on Joe's CH-3 spectrometer, I tweaked the belt length to get it a little bit better.

I find the problems with the alignment scripts is sometimes there will be an outlier in the measurements (like a noise spike) and it will include that in its calculations. If the engineer is not familiar with interpreting the data, they will end up just following the script's adjustment instructions and end up in a loop where they are chasing their tails. Sometimes the script will fail in peaking, and it's not smart enough yet to know that, and so the count rates will be much different since it will use the previous peak position from the previous step's measurement. The scripts have gotten better, but they are not perfect.

I tend to see a bit of difference aligning the TAP manually compared to the scripts. PET and LiF not so much, where the results between script and manual adjustment is similar.
Hope that makes sense.

Regards,

John