Author Topic: Quantitative Spectral Interference Corrections  (Read 9769 times)

Probeman

  • Emeritus
  • *****
  • Posts: 2596
  • Never sleeps...
    • John Donovan
Re: Quantitative Spectral Interference Corrections
« Reply #15 on: January 10, 2023, 01:31:47 PM »
I wanted to return to my presentation of the correction of Va Ka by Ti Kb in Ti bearing materials (e.g., ilmenites) as begun here:

https://probesoftware.com/smf/index.php?topic=69.msg11503#msg11503

I left with the promise of trying to improve the slightly negative concentration of V Ka after the interference correction in SrTiO3 (extrapolated from TiO2) even though it was less than 2 standard deviations from zero and well below the 140 PPM detection limit determined for those conditions. I reproduce the previous result here:

Now our apparent V concentration is of course zero, simply because this is the standard utilized for the interference calibration.  Now let's apply this same interference calibration to a sample with a very different matrix, SrTiO3, first without an interference correction:

ELEM:        V      Ti      Sr       O   SUM 
   830    .199  26.105  47.632  26.154 100.090
   831    .205  26.169  47.918  26.154 100.446
   832    .207  26.253  47.718  26.154 100.332
   833    .207  26.262  47.751  26.154 100.375
   834    .206  26.371  47.789  26.154 100.520

AVER:     .205  26.232  47.762  26.154 100.352
SDEV:     .003    .101    .105    .000    .163
SERR:     .001    .045    .047    .000
%RSD:     1.54     .38     .22     .00

So an apparent vanadium concentration of ~2000 PPM! And now *with* the interference correction:

ELEM:        V      Ti      Sr       O   SUM 
   830   -.010  26.111  47.618  26.154  99.873
   831   -.005  26.175  47.904  26.154 100.228
   832   -.004  26.259  47.704  26.154 100.113
   833   -.003  26.268  47.737  26.154 100.156
   834   -.006  26.377  47.775  26.154 100.300

AVER:    -.005  26.238  47.747  26.154 100.134
SDEV:     .003    .101    .105    .000    .162
SERR:     .001    .045    .047    .000
%RSD:   -50.13     .38     .22     .00

PUBL:     n.a.  26.103  47.742  26.154  99.999
%VAR:      ---     .52   (.01)     .00
DIFF:      ---    .135   (.01)    .000
STDS:      523      22     251     ---

Now our apparent vanadium content is around -50 PPM which is a slight over correction, but still within 2 sigma (60 PPM) of zero, so we cannot really say that this is statistically significant with 99% confidence. In fact the 99% confidence detection limit for this vanadium in this matrix is around 140 PPM!

So first I decided to tune up on the TiO2 standard (as opposed to the Ti metal I used previously) standard but unfortunately that didn't help much, as I got -100 PPM +/- 40 PPM the first time and -30 PPM +/- 30 PPM the second acquisition. So yeah, still below the detection limit of 140 PPM but I thought maybe I should try re-checking the background positions because here is what I saw:



Not so bad you might say (vertical magenta lines are the original off-peak positions used above, and the vertical green lines are the new off-peak positions). So why did I move them?  Let's zoom in a bit:



As you can see, the original off-peak positions (magenta vertical lines) are slightly on the tails of the emission peaks, so the green lines are the improved positions. So how did that do?  Well we now get -60 PPM +/- 50 PPM and -60 PPM +/- 50 PPM extrapolating from TiO2 to SrTo3 and correcting for the interferance.  Essentially what we had before.

Then I thought maybe the problem is using the vanadium metal as the primary standard for V Ka?  I choose vanadium metal because that is what most people are going to have in their standard collections.  But it just so happens that when I was at Berkeley I obtained a very nice V2O3 from the Purdue crystal lab, so I said, let tune up on that and use that as our primary standard...  so how did that do?

Well on the first set of points I got -80 PPM +/- 60 PPM and on the second try I got -60 PPM +/- 30 PPM, so pretty much the same slight over correction, though still within 2 sigma statistics and considerably lower than our 140 PPM 99% confidence detection limit. Again extrapolating from TiO2 to SrTiO3.
The only stupid question is the one not asked!

Probeman

  • Emeritus
  • *****
  • Posts: 2596
  • Never sleeps...
    • John Donovan
Re: Quantitative Spectral Interference Corrections
« Reply #16 on: January 10, 2023, 02:41:51 PM »
Then I thought: what if this slight over correction is due to an extrapolation of the V Ka backgrounds across the Ti K absorption edge?

So I changed the background correction from linear interpolation to "high only" off-peaks for all the samples as shown here:



and now we get this result:

St  251 Set   1 Strontium titanate (SrTiO3), Results in Elemental Weight Percents
 
ELEM:        V      Ti      Sr       O
TYPE:     ANAL    ANAL    ANAL    SPEC
BGDS:     HIGH     LIN     LIN
TIME:    60.00   60.00   60.00     ---
BEAM:    30.02   30.02   30.02     ---

ELEM:        V      Ti      Sr       O   SUM 
    11    .005  26.207  47.773  26.154 100.139
    12   -.006  26.168  47.735  26.154 100.050
    13    .001  26.248  47.924  26.154 100.327
    14    .011  26.186  47.610  26.154  99.961
    15   -.011  26.161  47.681  26.154  99.984

AVER:     .000  26.194  47.744  26.154 100.092
SDEV:     .009    .035    .117    .000    .148
SERR:     .004    .016    .053    .000
%RSD: -6796.45     .14     .25     .00

PUBL:     n.a.  26.103  47.742  26.154  99.999
%VAR:      ---     .35   (.01)     .00
DIFF:      ---    .091   (.00)    .000
STDS:       23      22     251     ---

Mystery solved!   ;D

I should have utilized the MAN background correction and the problem would have been solved the first time!
The only stupid question is the one not asked!

sem-geologist

  • Professor
  • ****
  • Posts: 247
Re: Quantitative Spectral Interference Corrections
« Reply #17 on: January 11, 2023, 02:42:06 AM »
Then I thought: what if this slight over correction is due to an extrapolation of the V Ka backgrounds across the Ti K absorption edge?

So I changed the background correction from linear interpolation to "high only" off-peaks for all the samples as shown here:



and now we get this result:

St  251 Set   1 Strontium titanate (SrTiO3), Results in Elemental Weight Percents
 
ELEM:        V      Ti      Sr       O
TYPE:     ANAL    ANAL    ANAL    SPEC
BGDS:     HIGH     LIN     LIN
TIME:    60.00   60.00   60.00     ---
BEAM:    30.02   30.02   30.02     ---

ELEM:        V      Ti      Sr       O   SUM 
    11    .005  26.207  47.773  26.154 100.139
    12   -.006  26.168  47.735  26.154 100.050
    13    .001  26.248  47.924  26.154 100.327
    14    .011  26.186  47.610  26.154  99.961
    15   -.011  26.161  47.681  26.154  99.984

AVER:     .000  26.194  47.744  26.154 100.092
SDEV:     .009    .035    .117    .000    .148
SERR:     .004    .016    .053    .000
%RSD: -6796.45     .14     .25     .00

PUBL:     n.a.  26.103  47.742  26.154  99.999
%VAR:      ---     .35   (.01)     .00
DIFF:      ---    .091   (.00)    .000
STDS:       23      22     251     ---

Mystery solved!   ;D

I should have utilized the MAN background correction and the problem would have been solved the first time!

While MAN probably could give You correct answer...
I am glad that You had found at last on your own why a single background position can be better than HIGH+LOW.
And this is exactly Why I am advocating of using a single background measurements instead of two background ): high and low. (and that is why I absolutely despise multi-background position method as that bring even worse biases and guaranties multi-cross of many absorption edges). Exactly in many cases the single background position is possible and easier to set that there would be no absorption edges in between peak and background measurement. I use HIGH+LOW only for tricky situations (in absolutely non canonical way).

The lesson 2, which is also important: as You found out at first attempt - it does not matter to much if background is set on tail or not. Most of my background measurements are on tails! And as far I investigated my experiments brought me to conclusion that if both standard and unknown are measured consistently at tail (natural tail from XTAL-defocusing), not the satellite lines) at exactly same positions it will give same result as if measuring further away (the only downside is slight P/B decrease). That is important and when choosing background position the primary priority should be no any absorption edges in between bkgd and peak positions (at least no edges for major and minor elements expected in mineral group), and "free-from-tails" should come as secondary and optional. The advantage of measuring background not on tail is that background position on unknown can be moved and same standard measurement can be reused. With bkgd on tail the background position should be fixed. Naturally this works reasonably well for not shifting high energy X-rays. The low energy shifting lines have its caveats then measuring bkgd on the tail. BTW some line shifts are not shifts but pseudo-shifts with close absorption edge.

I also think You missed one important part: the slope! From your screenshot I see it is set to 1.0. Due to only Ti and V in the system the interference correction also "corrects" under-interpolated background counts from continuum under peak. Using (warning! a self-advertisement) HussariX I see that at least on my LLIF the slope should be 1.14 for HIGH background position offset of +2084. I think Your presented setup would falsely give some trace amounts of V on substances with low Ti.

P.S. Situation is very similar to that of "on-tail-bkgd" for peak measurements: if standard and unknown is measured with same slight offset from peak top - it will give the same result as if measured both exactly on peak. That gives an opportunity to deal with pathological cases where absorption edge is close to the measured peak.
« Last Edit: January 11, 2023, 03:04:56 AM by sem-geologist »

JonF

  • Professor
  • ****
  • Posts: 108
Re: Quantitative Spectral Interference Corrections
« Reply #18 on: January 11, 2023, 07:28:47 AM »
You might need to look at the suitability of TiO2 as an interference standard for SrTiO3 as discussed here: https://probesoftware.com/smf/index.php?topic=1397.msg10228#msg10228

From your wds plots on page 1, and as you say they're both free of V, it doesn't look like the emission structure is the same in the V Ka energy region

Probeman

  • Emeritus
  • *****
  • Posts: 2596
  • Never sleeps...
    • John Donovan
Re: Quantitative Spectral Interference Corrections
« Reply #19 on: January 11, 2023, 09:14:12 AM »
I am glad that You had found at last on your own why a single background position can be better than HIGH+LOW.

Sorry SG, I have known that a single background can provide a benefit when extrapolating across an absorption edge since I was still "wet behind the ears"!    ;)   Why do you think Probe for EPMA offers the slope-high and slope-low options? 

And this is exactly Why I am advocating of using a single background measurements instead of two background ): high and low. (and that is why I absolutely despise multi-background position method as that bring even worse biases and guaranties multi-cross of many absorption edges). Exactly in many cases the single background position is possible and easier to set that there would be no absorption edges in between peak and background measurement. I use HIGH+LOW only for tricky situations (in absolutely non canonical way).

I think you are too much a "one true thing" guy.   :D  But as you say yourself, different situations require different models. That is why PFE has 8 different off-peak background models, plus MAN and MPB.

Now you may "despise" MPB but I'll bet you've never tried the implementation in PFE! For example the way MPB is implemented in PFE is that it is calculated *iteratively* so that the number of, and the specific off-peak bgd positions for each data point are re-calculated "on the fly" to obtain the fit with the lowest positive deviation. This allows for those "tricky" situations where various trace/minor elements are variably present or not present in complex materials. In addition it allows one to automatically and exactly fit curved backgrounds, which as you know are especially a problem at low sin thetas: 

https://iopscience.iop.org/article/10.1088/1757-899X/32/1/012012/meta

I'm glad you find single backgrounds plus a slope useful for many situations, but I hope you agree that "one size does not fit all"!

I also think You missed one important part: the slope! From your screenshot I see it is set to 1.0. Due to only Ti and V in the system the interference correction also "corrects" under-interpolated background counts from continuum under peak. Using (warning! a self-advertisement) HussariX I see that at least on my LLIF the slope should be 1.14 for HIGH background position offset of +2084. I think Your presented setup would falsely give some trace amounts of V on substances with low Ti.

My selection of a single high side background was just a quick test. But let's try your 1.14 slope because well, PFE has a "slope high" background calculation also. So here is the same sample quoted above with a slope of 1.14:

St  251 Set   1 Strontium titanate (SrTiO3), Results in Elemental Weight Percents
 
ELEM:        V      Ti      Sr       O
TYPE:     ANAL    ANAL    ANAL    SPEC
BGDS:     S-Hi     LIN     LIN
TIME:    60.00   60.00   60.00     ---
BEAM:    30.02   30.02   30.02     ---

ELEM:        V      Ti      Sr       O   SUM 
    11   -.005  26.208  47.772  26.154 100.129
    12   -.017  26.168  47.734  26.154 100.039
    13   -.009  26.249  47.923  26.154 100.316
    14    .002  26.186  47.609  26.154  99.951
    15   -.022  26.161  47.680  26.154  99.973

AVER:    -.010  26.194  47.744  26.154 100.082
SDEV:     .009    .035    .117    .000    .148
SERR:     .004    .016    .053    .000
%RSD:   -89.86     .14     .25     .00

PUBL:     n.a.  26.103  47.742  26.154  99.999
%VAR:      ---     .35   (.00)     .00
DIFF:      ---    .091   (.00)    .000
STDS:       23      22     251     ---

So we get -100 PPM +/- 90 PPM, which is still within 2 sigma of zero, but a bit more negative than I'd like to see. I think Jon Fellowes comments above might also be relevant in this situation.
« Last Edit: January 11, 2023, 09:37:35 PM by Probeman »
The only stupid question is the one not asked!

Probeman

  • Emeritus
  • *****
  • Posts: 2596
  • Never sleeps...
    • John Donovan
Re: Quantitative Spectral Interference Corrections
« Reply #20 on: January 11, 2023, 09:22:22 AM »
You might need to look at the suitability of TiO2 as an interference standard for SrTiO3 as discussed here: https://probesoftware.com/smf/index.php?topic=1397.msg10228#msg10228

From your wds plots on page 1, and as you say they're both free of V, it doesn't look like the emission structure is the same in the V Ka energy region

I had forgotten about this discussion! Yeah, I think you are correct. In some situations at least, for ultimate accuracy the interference standard may need to be somewhat matrix matched if the interfering emission line is affected by chemical bonding.

Of course the interference correction in PFE does already perform a matrix correction when extrapolating from the interference standard to the unknown composition, so the matrix of the interference standard will not need to be exactly the same as the unknown, but it does probably need to be roughly the same bonding environment if one is seeking sub 100 PPM accuracy with large spectral overlaps.

Since I don't have a synthetic ilmenite standard with a known zero vanadium content, I had to try something and SrTiO3 was all I could come up with.

This again only points out the need for a full suite of high purity synthetic minerals globally distributed as discussed in the open letter here:

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

I hereby declare the need for a high purity synthetic FeTiO3 standard. Not only are these high purity synthetic minerals necessary for interference standards, but also for use as blank standards and also MAN standards.
« Last Edit: January 11, 2023, 02:08:47 PM by Probeman »
The only stupid question is the one not asked!

sem-geologist

  • Professor
  • ****
  • Posts: 247
Re: Quantitative Spectral Interference Corrections
« Reply #21 on: January 12, 2023, 03:10:20 AM »

Now you may "despise" MPB but I'll bet you've never tried the implementation in PFE! For example the way MPB is implemented in PFE is that it is calculated *iteratively* so that the number of, and the specific off-peak bgd positions for each data point are re-calculated "on the fly" to obtain the fit with the lowest positive deviation. This allows for those "tricky" situations where various trace/minor elements are variably present or not present in complex materials. In addition it allows one to automatically and exactly fit curved backgrounds, which as you know are especially a problem at low sin thetas: 

https://iopscience.iop.org/article/10.1088/1757-899X/32/1/012012/meta

I'm glad you find single backgrounds plus a slope useful for many situations, but I hope you agree that "one size does not fit all"!
Yes I had never tried MPB, as You probably are aware we have no PFE license. However, I am looking forward and evaluating features at fundamental level which are in PFE, especially that Cameca is not going to advance its software any more further.
With absolutely whole seriousness I can't find a single usage case where MPB would solve any problem better, than a single background (MPB is disadvantageous in every possible perspective: precision, accuracy and time). Contrary to MPB, MAN looks like very useful option and that is one of PFE selling point (Which sways me toward considering to buy the license). As I got aware that for trace analysis it is more important to measure correctly the background, the investigation of background measurement types was a first thing I implemented in my HussariX software. As I first got to know existence of MPB and read about it - it looked very promising tool. However after collecting extensive WDS Wavescan database and looking to the reality of clutter of REE (especially on large XTALS) I came to conclusion - the MPB tool is very hideous and only pretends to overcome the problems especially for REE cases - its solution hides the problem (of clutter) and does not deal with it. Thus I see MPB too be dangerous (as any statistic tool used not properly) as it checks for best fit, while in reality (i.e. REE minerals) there is very limited uncluttered background positions for every element, and often those are on tail of very measured peak (often a single spot if any). MBP will check for best fit and produce nice statistics embracing majority of overestimated (by densily packed other element lines) biasing background, while it will reject the only closest to correct background measurement as an outlier (accuracy disadvantage); MPB requires to subdivide the background measurement time and thus it takes more time to get something statistically sensible (time disadvantage); There rather is no possibility that all MPB measurements would be used to fit the background in multi-element minerals (in particularly those REE bearing) and thus set background time is wasted for something what is discarded - using the same (total) time for a single background measurement would give much more precise background measurement (MPB precision disadvantage). But maybe I am wrong. Can anyone point me to/ share the clear situation where MPB solves anything better than a single background?

And then lets move to the largest false selling point of MPB - curved backgrounds. YES, for a moment I agree - the traditional two point background (HIGH + LOW) measurement has its problem with that, as linear inter/extra-polation would under/over-shot, and exponential modeling can over or under estimate background too and gets nasty with peak interference. And I guess this is probably why MPB idea was born and got its reason to exist. BUT, it missed much more easier solution: the single background measurement, with well defined slope (the exactly same consistent slope for light, medium and heavy mineral matrixes), has absolutely no problem with curved backgrounds! (if there is no absorption edges in between background and peak - slope is the same - the most important findings thanks to HussariX). Additionally MPB clutters GUI, complicates the code providing absolutely no advantage compared with single background measurement.

So getting back on the topic of spectral interferences, I use two background measurements very rarely only to escape the limitation of Cameca's Peaksight - it is to escape circular correction situations. I.e. in for REE to break out such interference circular loop the Nd La is corrected from Ce Lb interference by setting the background positions which would give peak-bkgd near 0 independently from Ce intensity - it is self-correcting:

(These are not ultimate positions, they do not correct Cs (cesium) and Lu (lutetium) interferences (albeit they are very minor); If any if those elements are presented then this approach fails.)

 If I would be using PFE, where circular interference correction is possible, I would use single background measurements  explicitly (and probably MAN too). In that case I would be advocating that "One size fits all".

So we get -100 PPM +/- 90 PPM, which is still within 2 sigma of zero, but a bit more negative than I'd like to see. I think Jon Fellowes comments above might also be relevant in this situation.
interesting... was the interference correction measurement acquired also with the same slope?

Now as I looked into my Wavescans I got an a bit counter-intuitive idea: Would not V Ka measurement in this case be better to do on LPET instead of LLIF? Ti Kb1 will be much broader and it will dominate the interference with V Ka, and Ti Kb'' and Ti Kb5 will be more diffused with less influence. I call it counter-intuitive as that would use worse spectral resolution as an advantage.
« Last Edit: January 12, 2023, 06:36:32 AM by sem-geologist »

Probeman

  • Emeritus
  • *****
  • Posts: 2596
  • Never sleeps...
    • John Donovan
Re: Quantitative Spectral Interference Corrections
« Reply #22 on: January 12, 2023, 09:56:40 AM »
Yes I had never tried MPB, as You probably are aware we have no PFE license. However, I am looking forward and evaluating features at fundamental level which are in PFE, especially that Cameca is not going to advance its software any more further.

If I would be using PFE, where circular interference correction is possible, I would use single background measurements  explicitly (and probably MAN too). In that case I would be advocating that "One size fits all".

I'm all for "rules of thumb", but there are always exceptional circumstances and various approaches to them as I hope you will agree. Remember, we don't know, what we don't know.   :D 

The approach taken by Jercinovic et al., is to utilize MPB background measurements separately from the on-peak measurements and apply those background measurements to the on-peak measurements using the "Nth Point" methods in Probe for EPMA. This has the effect of vastly improving sample statistics much like the MAN method, without incurring additional sample damage.  The paper I cited above doesn't go into this variation of method in super detail but I think it gets a mention.

OK, I hope you have the opportunity to try Probe for EPMA some day. The good news is that it has something for everybody, but a person can certainly utilize only those options and methods that they prefer.  It's definitely a "team effort"!

So we get -100 PPM +/- 90 PPM, which is still within 2 sigma of zero, but a bit more negative than I'd like to see. I think Jon Fellowes comments above might also be relevant in this situation.
interesting... was the interference correction measurement acquired also with the same slope?

Yes, I selected all the samples and switched them from 2 point linear to high side only for all V Ka measurements.  It's just two clicks of the mouse!     :)    OK, OK, maybe 4 clicks!   :D

Now as I looked into my Wavescans I got an a bit counter-intuitive idea: Would not V Ka measurement in this case be better to do on LPET instead of LLIF? Ti Kb1 will be much broader and it will dominate the interference with V Ka, and Ti Kb'' and Ti Kb5 will be more diffused with less influence. I call it counter-intuitive as that would use worse spectral resolution as an advantage.

Measuring V Ka in SrTiO3 using a PET crystal without an interference correction yields an apparent vanadium concentration of ~1.39 wt% (13,900 PPM) as seen here (TiO2 as the primary standard):

St  851 Set   2 SrTiO3 (strontium titanate), Results in Elemental Weight Percents
 
ELEM:       Ti      Ti       V       V      Sr      Fe      Cr      Mn       O
TYPE:     ANAL    ANAL    ANAL    ANAL    SPEC    SPEC    SPEC    SPEC    SPEC
BGDS:      LIN     EXP     LIN    HIGH
TIME:      ---   60.00     ---   60.00     ---     ---     ---     ---     ---
BEAM:      ---   30.03     ---   30.03     ---     ---     ---     ---     ---

ELEM:     Ti-D      Ti     V-D       V      Sr      Fe      Cr      Mn       O   SUM 
XRAY:     (ka)    (ka)    (ka)    (ka)      ()      ()      ()      ()      ()
   179     ---  26.153     ---   1.398  47.740    .000    .000    .000  26.150 101.441
   180     ---  26.222     ---   1.388  47.740    .000    .000    .000  26.150 101.501
   181     ---  26.215     ---   1.372  47.740    .000    .000    .000  26.150 101.477
   182     ---  26.144     ---   1.397  47.740    .000    .000    .000  26.150 101.431
   183     ---  26.173     ---   1.395  47.740    .000    .000    .000  26.150 101.459

AVER:      ---  26.182     ---   1.390  47.740    .000    .000    .000  26.150 101.462
SDEV:      ---    .035     ---    .011    .000    .000    .000    .000    .000    .028
SERR:      ---    .016     ---    .005    .000    .000    .000    .000    .000
%RSD:      ---     .14     ---     .77     .00     .00     .00     .00     .00

PUBL:     n.a.  26.100    n.a.    n.a.  47.740    n.a.    n.a.    n.a.  26.150  99.990
%VAR:      ---     .31     ---     ---     .00     ---     ---     ---     .00
DIFF:      ---    .082     ---     ---    .000     ---     ---     ---    .000
STDS:      ---     922     ---     923     ---     ---     ---     ---     ---

This was  a measurement I did a number of years ago, so when I get a chance I might try it again, though I'd be more interested finding a synthetic high purity ilmenite standard to try these tests on! The same data with the interference correction (and a high off-peak only measurement) produces this result:

St  851 Set   2 SrTiO3 (strontium titanate), Results in Elemental Weight Percents
 
ELEM:       Ti      Ti       V       V      Sr      Fe      Cr      Mn       O
TYPE:     ANAL    ANAL    ANAL    ANAL    SPEC    SPEC    SPEC    SPEC    SPEC
BGDS:      LIN     EXP     LIN    HIGH
TIME:      ---   60.00     ---   60.00     ---     ---     ---     ---     ---
BEAM:      ---   30.03     ---   30.03     ---     ---     ---     ---     ---

ELEM:     Ti-D      Ti     V-D       V      Sr      Fe      Cr      Mn       O   SUM 
XRAY:     (ka)    (ka)    (ka)    (ka)      ()      ()      ()      ()      ()
   179     ---  26.181     ---   -.010  47.740    .000    .000    .000  26.150 100.061
   180     ---  26.251     ---   -.023  47.740    .000    .000    .000  26.150 100.117
   181     ---  26.243     ---   -.040  47.740    .000    .000    .000  26.150 100.093
   182     ---  26.173     ---   -.011  47.740    .000    .000    .000  26.150 100.051
   183     ---  26.202     ---   -.015  47.740    .000    .000    .000  26.150 100.077

AVER:      ---  26.210     ---   -.020  47.740    .000    .000    .000  26.150 100.080
SDEV:      ---    .036     ---    .012    .000    .000    .000    .000    .000    .026
SERR:      ---    .016     ---    .005    .000    .000    .000    .000    .000
%RSD:      ---     .14     ---  -61.81     .00     .00     .00     .00     .00

PUBL:     n.a.  26.100    n.a.    n.a.  47.740    n.a.    n.a.    n.a.  26.150  99.990
%VAR:      ---     .42     ---     ---     .00     ---     ---     ---     .00
DIFF:      ---    .110     ---     ---    .000     ---     ---     ---    .000
STDS:      ---     922     ---     923     ---     ---     ---     ---     ---

We now obtain -200 PPM with a variance of 120 PPM, so again within 2 standard deviations of zero!  Again a little more negative than I'd like to see... Frankly I think Jon Fellowes is correct about the bonding environment and that we should seek a high purity synthetic FeTiO3 material for a more accurate test of correcting the Ti Kb interference on V ka in ilmenites (though TiO2 should work absolutely fine as an interference standard for correction of Ti Kb on V Ka in rutiles).  I did find this article on synthesizing FeTiO3 from 1978:

https://www.sciencedirect.com/science/article/pii/0022024878903123

It's sort of funny to find this paper, because back in the 80s (starting out as a fledgling microanalyst), I had written Takei about his synthesis of fayalite (Fe2SiO4) and he very nicely sent me a little crystal which I examined. The only problem was that it had tiny little blebs of Fe metal. But then (I can't remember the circumstances exactly) Lynn Boatner at Oak Ridge sent me a synthetic fayalite which I have in several of my standards mounts at Oregon and I think I even have a tiny scrap left in a bottle in the standard materials collection. The material from Oak Ridge is perfect, and I wrote him a few years ago asking if he still had any additional material to spare or at least the recipe for this synthesis, but he never responded.

Finding significant (commercially produced) quantities of Fe2SiO4 and FeTiO3 would be fantastic. I think Will Nachlas, Aurelein Moy and John Fournelle are working on obtaining these high purity end member materials for use as microanalytical standards.

It sure would be nice.
« Last Edit: January 12, 2023, 11:27:28 AM by Probeman »
The only stupid question is the one not asked!