This topic is for discussing trace element data interpretation which is often problematic.
To start I will discuss the trace elements in the RbTiOPO4 standard materials that Owen Neill, Brian Joy and I have been discussing recently in this topic:
http://smf.probesoftware.com/index.php?topic=301.msg4267#msg4267
Here for reference are the results I obtained from a quant measurement at 400 seconds on-peak and 400 seconds on the off-peaks:
Un 7 RbTiOPO4, Results in Elemental Weight Percents
ELEM: K Cs Na Ca Mg Rb Ti P O
TYPE: ANAL ANAL ANAL ANAL ANAL SPEC SPEC SPEC SPEC
BGDS: LIN EXP EXP LIN LIN
TIME: 400.00 400.00 400.00 400.00 400.00 --- --- --- ---
BEAM: 100.94 100.94 100.94 100.94 100.94 --- --- --- ---
ELEM: K Cs Na Ca Mg Rb Ti P O SUM
360 .015 .013 -.014 .000 -.001 34.979 19.604 12.676 32.741 100.014
361 .016 .012 -.016 .000 -.001 34.979 19.604 12.676 32.741 100.012
362 .016 .015 -.012 .000 .000 34.979 19.604 12.676 32.741 100.019
363 .015 .013 -.017 .000 -.002 34.979 19.604 12.676 32.741 100.010
364 .016 .015 -.010 .000 .000 34.979 19.604 12.676 32.741 100.021
365 .015 .013 -.012 .000 .000 34.979 19.604 12.676 32.741 100.017
366 .016 .012 -.013 .000 -.001 34.979 19.604 12.676 32.741 100.014
AVER: .016 .014 -.013 .000 -.001 34.979 19.604 12.676 32.741 100.015
SDEV: .000 .001 .003 .000 .001 .000 .000 .000 .000 .004
SERR: .000 .000 .001 .000 .000 .000 .000 .000 .000
%RSD: 2.86 9.44 -18.91 273.41 -74.31 .00 .00 .00 .00
I'll start with the Na because a negative result is almost always an off-peak interference. But how can we know that? If we go into the PFE Elements/Cations dialog we can use the Hi and Lo off-peak interference buttons to calculate the nominal interferences for this element as seen here:
(https://smf.probesoftware.com/gallery/1_19_03_16_8_59_49.png)
The text result are here:
For Na ka (hi-off), TAP at 12.5222 angstroms( 48685.8), at an assumed concentration of 1 wt.%
Interference by P KA1 II at 12.3180 ( 47890.8) = 5.4%
Interference by P KA2 II at 12.3240 ( 47914.2) = 3.8%
Interference by Ca SKB`` IV at 12.3280 ( 47929.8) = .2%
Interference by Ca SKB^5 IV at 12.3460 ( 47999.8) = .5%
Interference by Ca KB3 IV at 12.3610 ( 48058.2) = 6.6%
Interference by Ca KB1 IV at 12.3610 ( 48058.2) = 11.6%
Interference by Ca SKB` IV at 12.4050 ( 48229.5) = 5.2%
Interference by Ca SKBN IV at 12.5300 ( 48716.1) = 31.7%
Interference by Cs LB2 V at 12.5610 ( 48836.8) = 267.7%
Interference by Ti KB3 V at 12.5720 ( 48879.6) = 79.1%
Interference by Ti KB1 V at 12.5720 ( 48879.6) = 140.3%
Note the significant interference from the 2nd order P Ka tails. For the low off-peak side we obtain this:
For Na ka (lo-off), TAP at 11.3676 angstroms( 44191.1), at an assumed concentration of 1 wt.%
Interference by Cs LG3 V at 11.1660 ( 43406.4) = .3%
Interference by K SKA3`` III at 11.1710 ( 43425.9) = .1%
Interference by K SKA` III at 11.1850 ( 43480.4) = .2%
Interference by K SKA`` III at 11.2150 ( 43597.1) = .8%
Interference by K KA1 III at 11.2270 ( 43643.9) = 123.0%
Interference by K KA2 III at 11.2370 ( 43682.8) = 89.0%
Interference by Na SKB^4 at 11.3140 ( 43982.5) = .5%
Interference by Cs SLA9 IV at 11.3830 ( 44251.1) = 31.0%
Interference by P SKB^4 II at 11.3900 ( 44278.4) = 12.5%
Interference by P SKB`` II at 11.4230 ( 44406.8) = 8.9%
Interference by Cs SLA8 IV at 11.4310 ( 44438.0) = 18.8%
Interference by Cs SLA7 IV at 11.4660 ( 44574.2) = 8.9%
Interference by Cs SLA6 IV at 11.5040 ( 44722.1) = 2.8%
Interference by Cs SLA5 IV at 11.5160 ( 44768.8) = 1.8%
Interference by Cs SLA3 IV at 11.5520 ( 44909.0) = .4%
Interference by Cs LA1 IV at 11.5710 ( 44983.0) = 13.7%
Interference by Cs LA2 IV at 11.6090 ( 45130.9) = .2%
On the low off-peak side we see significant interference from the 2nd order P Ka satellite lines. This situation definitely calls for the multi-point off-peak background method!
Ok, so the negative Na result is an obvious problem. But how do we know that the positive results from K and Cs are real?
On the K front, Brian presented the K wavescan, and since the K Ka and K Kb peaks are both present that is highly suggestive that trace K is present in the RbTiOPO4 material. Here is my wavescan (6 seconds per point), and even with that short a scan time, the K Ka peak is clearly visible (I didn't scan wide enough to catch the K Kb peak):
(https://smf.probesoftware.com/gallery/1_19_03_16_9_27_39.png)
But could this be a secondary (or higher order) peak from one of the major elements? Let's go back to PFE and this time use the Standard Assignments dialog to check for nominal on-peak interferences.
(https://smf.probesoftware.com/gallery/1_19_03_16_9_30_51.png)
So there is a significant interference from Rb but it is 4th order so it should be very weak, and the interfering Rb peak positions are hundreds of spectrometer units away from the K on-peak position... and the presence of both the Ka and Kb peaks (in Brian's scan) is very consistent with potassium actually being present. And his K Kb peak is smaller than his K Ka peak, again consistent with potassium being present.
But ideally we'd probably want to perform an interference correction for Rb on K using any Rb standard that we know doesn't contain potassium (!), to be sure. Or perform a blank measurement on a pure RbTiOPO4 standard. Which is what we are trying to create here so round and round it goes!
On the Cs front, I'm going to agree with Brian again.
A close look at the Cs La on LPET wavescan shows a nasty tail from the Ti Ka on the Cs La peak position.
(https://smf.probesoftware.com/gallery/1_19_03_16_9_53_50.png)
A measurement using a large LIF crystal (instead of the PET crystal I used) would be a much better measurement as the peak position goes from a very low spectrometer angle to a very high spectrometer angle.
This is a great example of why trace element interpretation can be problematic! :-[
I'm going to try again with multi-point bgds on Na and Cs on a LLIF crystal over the weekend.
Quote from: John Donovan on March 19, 2016, 09:38:24 AM
(https://smf.probesoftware.com/gallery/1_19_03_16_9_30_51.png)
So there is a significant interference from Rb but it is 4th order so it should be very weak, and the interfering Rb peak positions are hundreds of spectrometer units away from the K on-peak position... and the presence of both the Ka and Kb peaks (in Brian's scan) is very consistent with potassium actually being present. And his K Kb peak is smaller than his K Ka peak, again consistent with potassium being present.
But ideally we'd probably want to perform an interference correction for Rb on K using any Rb standard that we know doesn't contain potassium (!), to be sure. Or perform a blank measurement on a pure RbTiOPO4 standard. Which is what we are trying to create here so round and round it goes!
Note that Rb K_edge = 15.199 keV. I used beam energy = 15 keV.
Quote from: Brian Joy on March 19, 2016, 10:34:51 AM
Quote from: John Donovan on March 19, 2016, 09:38:24 AM
(https://smf.probesoftware.com/gallery/1_19_03_16_9_30_51.png)
So there is a significant interference from Rb but it is 4th order so it should be very weak, and the interfering Rb peak positions are hundreds of spectrometer units away from the K on-peak position... and the presence of both the Ka and Kb peaks (in Brian's scan) is very consistent with potassium actually being present. And his K Kb peak is smaller than his K Ka peak, again consistent with potassium being present.
But ideally we'd probably want to perform an interference correction for Rb on K using any Rb standard that we know doesn't contain potassium (!), to be sure. Or perform a blank measurement on a pure RbTiOPO4 standard. Which is what we are trying to create here so round and round it goes!
Note that Rb K_edge = 15.199 keV. I used beam energy = 15 keV.
Excellent point! This certainly explains why you didn't see that interference! I should have the nominal interference calculation in PFE check for that!
Edit by John: I actually ran my trace elements at 20 keV (100 nA), so that is why PFE printed the Rb Ka interference out!
But still it would be a low overvoltage (~1.3), so unlikely to show up well as a 4th order interference.
I also will run my UC Berkeley RbTiOPO4 against the CalChemist material and see if there's any difference in the apparent K signal
I ran a more careful characterization on both the CalChemist and the original UC Berkeley RbTiOPO4 material and thought I would share how I set up the analysis for the same trace elements as before, that is K, Cs, Na, Ca and Mg.
For starters I decided to measure using the same crystals, specifically the large PET (LPET) crystal for Cs la, which puts the spectrometer at a fairly low sin theta and due to the lower spectral resolution of the PET crystal compared with LIF or LLIF, to just deal with the over lap from the tail of the Ti ka line. So I also measured Ti for this interference correction.
Also I utilized multi-point bgds (MPB) on all the trace elements so we have more leeway in fitting the backgrounds in post processing. Measuring traces in RbTiOPO4 isn't as bad a monazite, which is a REE "zoo", but it does have its interesting background issues.
So first let's start with K Ka. Here is a wavescan plot of the RbTiOPO4 (CalChemist) material using 12 secs per point with the original (traditional) off-peaks shown in magenta:
(https://smf.probesoftware.com/gallery/395_21_03_16_1_50_52.png)
Now the same, but with the multi-point bgds I selected (four on each side):
(https://smf.probesoftware.com/gallery/395_21_03_16_1_52_34.png)
Here is an example of one point analysis showing the MPBs selected automatically by the software, though they can be manually selected as well (I overlaid the wavescan from above to show the K Ka peak):
(https://smf.probesoftware.com/gallery/395_21_03_16_1_54_15.png)
Note the exponential fit based on the 4 selected multi-point background measurements (320 sec on-peak, and 160 seconds on each of the 4 off-peaks).
Now let's look at Cs La.
Here is the Cs wavescan (20 keV, 100 nA, 10 um 200 points):
(https://smf.probesoftware.com/gallery/395_21_03_16_2_02_37.png)
Now the same plot zoomed in Y and showing the MPBs for Cs La. Note the exponential fit and the Ti ka tails:
(https://smf.probesoftware.com/gallery/395_21_03_16_2_03_40.png)
Now if we plot up the MPBs with the wavescan data we can immediately see the background fit problem. And that is that there is *no background* between the Ti ka and Ti Kb peaks!
(https://smf.probesoftware.com/gallery/395_21_03_16_2_06_22.png)
We can have the program only use 1 off-peak on the low side of the cs La peak to avoid the background between the Ti ka and Ti Kb peaks, but the single low side background is still probably on the tail of the Ti KLb peak... yes, yes, we should be measuring Cs La using an LIF crystal, but this is still instructive to see what we can do in adverse circumstances...
If we select only one background on the low side we get this, which is an improvement, but still not good enough:
(https://smf.probesoftware.com/gallery/395_21_03_16_2_23_29.png)
I'm going to perform a wider scan this time going further on the low side to see if we ever get to actual background past the Ti Kb peak...
Just for completeness, here are the plots for Na, Ca and Mg. First here is Na Ka. There could be an interference on Na Ka from the tail of the P Ka 3rd order peak as seen here:
(https://smf.probesoftware.com/gallery/395_21_03_16_2_47_10.png)
And indeed if I analyze my YPO4 standard get about 600-700 PPM of Na apparently from an interference by P Ka 3rd order:
ELEM: K Cs Na Ca Mg Ti P Rb O Y SUM
275 .000 -.007 .063 .000 -.003 .006 16.836 .000 34.805 48.350 100.050
276 .001 -.015 .070 -.002 -.001 .004 16.868 .000 34.805 48.350 100.081
277 .005 -.010 .068 -.002 .002 .004 16.842 .000 34.805 48.350 100.064
AVER: .002 -.011 .067 -.001 .000 .005 16.849 .000 34.805 48.350 100.065
SDEV: .002 .004 .004 .001 .002 .001 .017 .000 .000 .000 .016
SERR: .001 .002 .002 .001 .001 .001 .010 .000 .000 .000
%RSD: 119.51 -39.50 5.75 -88.93 -523.22 22.07 .10 .00 .00 .00
There does indeed appear to be a small Ca peak..
(https://smf.probesoftware.com/gallery/395_21_03_16_2_47_53.png)
(https://smf.probesoftware.com/gallery/395_21_03_16_2_48_46.png)
But not much Mg. The quant will tell us.
Quote from: Probeman on March 21, 2016, 02:21:49 PM
We can have the program only use 1 off-peak on the low side of the cs La peak to avoid the background between the Ti ka and Ti Kb peaks, but the single low side background is still probably on the tail of the Ti KLb peak... yes, yes, we should be measuring Cs La using an LIF crystal, but this is still instructive to see what we can do in adverse circumstances...
If we select only one background on the low side we get this, which is an improvement, but still not good enough:
(https://smf.probesoftware.com/gallery/395_21_03_16_2_23_29.png)
I'm going to perform a wider scan this time going further on the low side to see if we ever get to actual background past the Ti Kb peak...
But keep in mind that the Ti K absorption edge produces a discontinuity in the continuum on the low sine(theta) side of the Ti Kb peak. Measurement of background on opposing sides of an absorption edge could cause the concentration of a trace element to be overestimated significantly (aside from problems with the high sine(theta) tail on Ti Ka). Why not just choose "background" offsets (using a non-linear model) on either side of the Cs La peak position and all on the high-sine(theta) tail of Ti Ka, especially since the matrix is essentially constant in composition?
Quote from: Brian Joy on March 21, 2016, 05:56:00 PM
But keep in mind that the Ti K absorption edge produces a discontinuity in the continuum on the low sine(theta) side of the Ti Kb peak. Measurement of background on opposing sides of an absorption edge could cause the concentration of a trace element to be overestimated significantly (aside from problems with the high sine(theta) tail on Ti Ka). Why not just choose "background" offsets (using a non-linear model) on either side of the Cs La peak position and all on the high-sine(theta) tail of Ti Ka, especially since the matrix is essentially constant in composition?
Absolutely. That might work too. Here's an example from my Amer. Min. paper (2011) of Al Ka in quartz where the tail of the Si Ka peak sits on the Al peak:
(https://smf.probesoftware.com/gallery/395_21_03_16_8_38_11.jpeg)
It will be interesting to see the shape of the continuum on the low side. Unfortunately it will have to wait a few days as our x-axis has just failed on our SX100 stage. :'(
I'm excited to attempt an interference correction from the Ti overlap on Cs La on this PET crystal. You know- "failure mode" testing... :D
Good news, my Sx100 stage is operational again and I acquired a nice multi-point background run this weekend on the RbTiOPO4 (both UC Berkeley and CalChemist material).
I'll post the quant results tomorrow... in the meantime here is the MPB background fit for the Cs La line in this matrix:
(https://smf.probesoftware.com/gallery/395_03_04_16_3_33_46.png)
Brian Joy is absolutely correct: a sane person would run this element on an LiF crystal due to the nasty Ti ka interference, not to mention the absorption edge issue on the other side of the Ti Kb peak, but when the quantitative iterated interference correction is applied from the TiO2 standard, a very interesting thing happens (but only if the same MPB positions were utilized for both the standard and the unknown). Tomorrow...
Strange when you think about it, that we often run our EPMA standards and unknowns differently, because every other analytical field I know of generally treats unknowns and controls as much the same as possible to rule out problems from things we don't know about. Of course I understand the reason (to save time), but still.
So we'll get back to the Cs La trace characterization in a moment because it is quite interesting, but in the meantime I ran a wider wavescan on the RbTiOPO4 to see the Cs La background on the low side of the Ti Kb peak and it appears to be measuring the actual background (whether the Ti K edge is complicating things or not remains to be seen, as I have an "ace up my sleeve" with the quantitative interference correction!).
(https://smf.probesoftware.com/gallery/395_04_04_16_10_25_29.png)
In the meantime here are the results starting with potassium. First a plot of the multi-point-backgrounds (MPB) with a fairly high precision wavescan using 12 seconds per point:
(https://smf.probesoftware.com/gallery/395_04_04_16_10_59_40.png)
The Ar absorption edge is obvious but there appears to be no other continuum artifacts to be concerned with as seen in this closer look:
(https://smf.probesoftware.com/gallery/395_04_04_16_11_01_09.png)
Of most concern would be the possibility that the Rb 4th order reflections could be interfering with the K emission line and creating an apparent K concentration. But I am not concerned for three reasons:
1. The Rb K edge is 15.2 keV and although we ran the acquisition at 20 keV, that is still a very low overvoltage (~1.31) so our sensitivity should be very poor.
2. Any 4th order reflections that are present should be dramatically suppressed by the use of differential mode in PHA.
3. The position of the 4th order reflections are off enough to suspect that they are not at the K Ka position, though sometimes, these calculations can be off a little. But the best evidence against the Rb Ka 4th order reflections being present is that we should be able to resolve the Ka1/Ka2 split and instead we see a single clean peak which makes sense for K ka.
Finally, we should expect some K to be present as this is a RbTiOPO4 material and potassium would be expected as a contaminant.
Now for Na it is a little more messy. We have some P K 2nd order reflections that could easily interfere with the Na backgrounds. The P Ka1/Ka2 2nd order reflection is obvious, while the small peak to the left could be the satellite P SKA6 2nd order and that seems to be confirmed by the larger P SKBX 2nd order reflection further to the left.
(https://smf.probesoftware.com/gallery/395_04_04_16_11_22_13.png)
But all in all, the backgrounds selected by the MPB method (points circled in red), seem to be otherwise pretty good, though it is a nasty part of the spectrum.
Now let's turn to Ca and Mg. Both appear to be pretty clean with respect to backgrounds. First Ca Ka:
(https://smf.probesoftware.com/gallery/395_04_04_16_11_40_11.png)
This scan was 20 seconds per point, so again fairly high precision.
For Mg Ka, we can see a small Ti Ka satellite (4th order!) reflection, but otherwise the background appears free from artifacts (again 20 seconds per point):
(https://smf.probesoftware.com/gallery/395_04_04_16_12_10_43.png)
Now for the trace characterization of Cs La using a PET crystal in our RbTiOPO4 matrix.
As Brian Joy has pointed out, we really should be using an LIF crystal for this analysis, due to the higher resolution and higher sin theta of the LIF for Cs La, but let's run the experiment. Recall the significant interference from Ti ka due to the tail of the Ti Ka peak:
(https://smf.probesoftware.com/gallery/395_04_04_16_10_25_29.png)
If we calculate the nominal interferences using the Standard Assignments dialog, we obtain this prediction assuming Gaussian peak overlaps:
(https://smf.probesoftware.com/gallery/395_04_04_16_1_38_39.png)
So, without an interference specified let's see what we obtain for our trace elements. First a report on the analytical setup:
Un 4 CalChemist RbTiOPO4 #2
TakeOff = 40.0 KiloVolt = 20.0 Beam Current = 100. Beam Size = 10
(Magnification (analytical) = 20000), Beam Mode = Analog Spot
(Magnification (default) = 1000, Magnification (imaging) = 1000)
Image Shift (X,Y): .00, .00
Compositional analyses were acquired on an electron microprobe (Cameca SX100 (TCP/IP Socket)) equipped with 5 tunable wavelength dispersive spectrometers. Operating conditions were 40 degrees takeoff angle, and a beam energy of 20 keV. The beam current was 100 nA, and the beam diameter was 10 microns.
Elements were acquired using analyzing crystals LPET for K ka, Cs la, Ti ka, P ka, PET for Ca ka, LPET for K ka, Cs la, Ti ka, P ka, and TAP for Na ka, Mg ka.
The standards were TiO2 synthetic for Ti ka, Nepheline (partial anal.) for Na ka, Diopside (Chesterman) for Mg ka, Ca ka, Orthoclase MAD-10 for K ka, YPO4 (USNM 168499) for P ka, and Pollucite, PA-031 for Cs la.
The counting time was 80 seconds for Ti ka, P ka, 320 seconds for K ka, Cs la, and 400 seconds for Mg ka, Na ka, Ca ka. The off peak counting time was 80 seconds for Ti ka, P ka, 320 seconds for K ka, Cs la, and 400 seconds for Mg ka, Na ka, Ca ka. Off Peak correction method was Exponential for Ti ka, P ka, and Multi-Point for Na ka, Ca ka, Mg ka, K ka, Cs la.
Unknown and standard intensities were corrected for deadtime. Standard intensities were corrected for standard drift over time.
Results are the average of 12 points and detection limits ranged from .001 weight percent for K ka to .002 weight percent for Na ka to .003 weight percent for P ka.
Analytical sensitivity (at the 99% confidence level) ranged from .044 percent relative for Ti ka to 4.937 percent relative for Cs la to 70.071 percent relative for Na ka.
The exponential or polynomial background fit was utilized.
See John J. Donovan, Heather A. Lowers and Brian G. Rusk, Improved electron probe microanalysis of trace elements in quartz, American Mineralogist, 96, 274282, 2011
And here are the results (without the interference correction for Ti on Cs La):
Un 4 CalChemist RbTiOPO4 #2, Results in Elemental Weight Percents
ELEM: K Cs Na Ca Mg Ti P Rb O
TYPE: ANAL ANAL ANAL ANAL ANAL ANAL ANAL SPEC SPEC
BGDS: MULT MULT MULT MULT MULT EXP EXP
TIME: 320.00 320.00 400.00 400.00 400.00 80.00 80.00 --- ---
BEAM: 99.63 99.63 99.63 99.63 99.63 99.63 99.63 --- ---
ELEM: K Cs Na Ca Mg Ti P Rb O SUM
37 .018 .028 .001 .003 .000 19.427 12.506 34.979 32.741 99.702
38 .017 .028 .000 .003 .000 19.454 12.522 34.979 32.741 99.744
39 .017 .028 .000 .001 -.001 19.485 12.505 34.979 32.741 99.756
40 .018 .029 .001 .003 .000 19.516 12.542 34.979 32.741 99.828
41 .019 .027 .002 .004 .001 19.465 12.500 34.979 32.741 99.736
42 .018 .028 -.002 .002 -.001 19.446 12.511 34.979 32.741 99.723
43 .018 .027 .001 .002 -.001 19.471 12.568 34.979 32.741 99.805
44 .018 .026 .002 .002 .000 19.445 12.496 34.979 32.741 99.707
45 .018 .026 -.002 .002 .000 19.488 12.478 34.979 32.741 99.730
46 .018 .024 .001 .003 .000 19.527 12.567 34.979 32.741 99.860
47 .018 .026 -.001 .003 -.001 19.489 12.544 34.979 32.741 99.797
48 .019 .028 .001 .002 -.001 19.507 12.506 34.979 32.741 99.781
AVER: .018 .027 .000 .002 .000 19.477 12.520 34.979 32.741 99.764
SDEV: .000 .002 .001 .001 .001 .031 .029 .000 .000 .050
SERR: .000 .000 .000 .000 .000 .009 .008 .000 .000
%RSD: 2.29 5.61 712.86 31.93 -247.65 .16 .23 .00 .00
STDS: 374 1125 336 358 358 22 1016 --- ---
The Ti interference is around 270 PPM in the RbTiOPO4 matrix. Now let's check our TiO2 standard which will be used for the interference correction:
St 22 Set 2 TiO2 synthetic, Results in Elemental Weight Percents
ELEM: K Cs Na Ca Mg Ti P Rb O
TYPE: ANAL ANAL ANAL ANAL ANAL ANAL ANAL SPEC SPEC
BGDS: MULT MULT MULT MULT MULT EXP EXP
TIME: 40.00 40.00 40.00 40.00 40.00 40.00 40.00 --- ---
BEAM: 99.85 99.85 99.85 99.85 99.85 99.85 99.85 --- ---
ELEM: K Cs Na Ca Mg Ti P Rb O SUM
279 .001 .081 .001 -.003 -.001 59.882 .001 .000 40.050 100.011
280 .003 .076 .002 .002 .003 59.955 .000 .000 40.050 100.090
281 .002 .080 .003 .000 .000 59.991 .001 .000 40.050 100.126
282 .002 .079 -.002 .003 .002 59.906 -.002 .000 40.050 100.039
AVER: .002 .079 .001 .000 .001 59.934 .000 .000 40.050 100.067
SDEV: .001 .002 .002 .003 .002 .049 .001 .000 .000 .052
Note the we see an apparent Cs concentration of 790 PPM in our TiO2 due to the interference from Ti Ka.
Now we apply the interference correction by simply selecting Ti as the interfering element and TiO2 as the interference standard (because it contains no Cs, and contains a known amount of the interfering element Ti, and no other interfering elements).
Remember, because the interference correction in Probe for EPMA is iterated and quantitative, the differences in the matrix corrections between the unknown (RbTiOPO4), and the interference standard (TiO2) are properly and automatically dealt with.
Un 4 CalChemist RbTiOPO4 #2, Results in Elemental Weight Percents
ELEM: K Cs Na Ca Mg Ti P Rb O
TYPE: ANAL ANAL ANAL ANAL ANAL ANAL ANAL SPEC SPEC
BGDS: MULT MULT MULT MULT MULT EXP EXP
TIME: 320.00 320.00 400.00 400.00 400.00 80.00 80.00 --- ---
BEAM: 99.63 99.63 99.63 99.63 99.63 99.63 99.63 --- ---
ELEM: K Cs Na Ca Mg Ti P Rb O SUM
37 .018 .002 .001 .003 .000 19.428 12.506 34.979 32.741 99.677
38 .017 .002 .000 .003 .000 19.454 12.521 34.979 32.741 99.719
39 .017 .003 .000 .001 -.001 19.486 12.505 34.979 32.741 99.731
40 .018 .003 .001 .003 .000 19.517 12.542 34.979 32.741 99.803
41 .019 .001 .002 .004 .001 19.466 12.499 34.979 32.741 99.711
42 .018 .002 -.002 .002 -.001 19.447 12.511 34.979 32.741 99.698
43 .018 .001 .001 .002 -.001 19.472 12.568 34.979 32.741 99.780
44 .018 .000 .002 .002 .000 19.445 12.496 34.979 32.741 99.682
45 .018 .000 -.002 .002 .000 19.489 12.478 34.979 32.741 99.704
46 .018 -.002 .001 .003 .000 19.528 12.567 34.979 32.741 99.835
47 .018 .001 -.001 .003 -.001 19.489 12.543 34.979 32.741 99.772
48 .019 .002 .001 .002 -.001 19.507 12.506 34.979 32.741 99.756
AVER: .018 .001 .000 .002 .000 19.477 12.520 34.979 32.741 99.739
SDEV: .000 .002 .001 .001 .001 .031 .029 .000 .000 .050
SERR: .000 .000 .000 .000 .000 .009 .008 .000 .000
%RSD: 2.29 128.51 712.86 31.93 -247.65 .16 .23 .00 .00
STDS: 374 1125 336 358 358 22 1016 --- ---
STKF: .1102 .2652 .0583 .1676 .0644 .5616 .1496 --- ---
STCT: 9129.3 11088.8 1550.4 7022.5 3286.5 64371.5 4913.6 --- ---
UNKF: .0002 .0000 .0000 .0000 .0000 .1753 .0763 --- ---
UNCT: 12.9 .4 .0 .9 -.1 20095.1 2504.8 --- ---
UNBG: 45.8 156.1 11.1 40.4 22.7 132.4 7.6 --- ---
ZCOR: 1.1552 1.1604 2.6811 1.0545 1.8314 1.1110 1.6413 --- ---
KRAW: .0014 .0000 .0000 .0001 .0000 .3122 .5098 --- ---
PKBG: 1.28 1.00 1.00 1.02 1.00 152.73 330.98 --- ---
INT%: ---- -95.91 ---- ---- ---- ---- ---- --- ---
Note that with the interference correction turned on, the apparent 270 PPM of Cs now disappears, and the interference correction magnitude is -95.91 %, which tells us that all the apparent Cs was in fact spurious.
So now we see only around ~180 PPM of potassium and maybe 20-30 PPM of Ca, and that is all.
Let's compare the above CalChemist RbTiOPO4, with some crystals grown many decades ago at UC Berkeley:
Un 5 UC Berkeley RbTiOPO4 #1, Results in Elemental Weight Percents
ELEM: K Cs Na Ca Mg Ti P Rb O
TYPE: ANAL ANAL ANAL ANAL ANAL ANAL ANAL SPEC SPEC
BGDS: MULT MULT MULT MULT MULT EXP EXP
TIME: 320.00 320.00 400.00 400.00 400.00 80.00 80.00 --- ---
BEAM: 99.72 99.72 99.72 99.72 99.72 99.72 99.72 --- ---
ELEM: K Cs Na Ca Mg Ti P Rb O SUM
49 .137 .017 .002 .002 .000 19.558 12.562 34.979 32.741 100.000
50 .137 .019 .002 .002 .000 19.526 12.463 34.979 32.741 99.870
51 .139 .017 .003 .003 .000 19.578 12.475 34.979 32.741 99.935
52 .138 .019 .003 .002 .001 19.567 12.475 34.979 32.741 99.925
53 .138 .015 .002 .003 .000 19.541 12.445 34.979 32.741 99.865
54 .137 .021 .003 .002 .000 19.587 12.481 34.979 32.741 99.951
55 .139 .018 .004 .002 .000 19.553 12.457 34.979 32.741 99.893
56 .136 .021 .003 .001 .000 19.556 12.467 34.979 32.741 99.904
57 .138 .016 .005 .002 .001 19.579 12.479 34.979 32.741 99.940
58 .137 .017 .004 .002 .000 19.591 12.451 34.979 32.741 99.922
59 .137 .018 .006 .002 -.001 19.595 12.456 34.979 32.741 99.932
60 .137 .023 .004 .002 .000 19.610 12.451 34.979 32.741 99.947
61 .137 .021 .006 .001 .000 19.590 12.458 34.979 32.741 99.933
62 .136 .020 .002 .003 -.001 19.637 12.454 34.979 32.741 99.970
63 .137 .018 .003 .002 -.001 19.629 12.486 34.979 32.741 99.993
64 .137 .018 .003 .003 .000 19.649 12.467 34.979 32.741 99.997
65 .136 .018 .002 .001 .002 19.635 12.487 34.979 32.741 100.002
AVER: .137 .019 .003 .002 .000 19.587 12.471 34.979 32.741 99.940
SDEV: .001 .002 .001 .001 .001 .036 .027 .000 .000 .043
SERR: .000 .000 .000 .000 .000 .009 .006 .000 .000
%RSD: .59 10.57 32.94 32.18 1115.88 .18 .21 .00 .00
STDS: 374 1125 336 358 358 22 1016 --- ---
STKF: .1102 .2652 .0583 .1676 .0644 .5616 .1496 --- ---
STCT: 9176.1 11107.4 1548.1 7016.8 3268.1 64424.2 4922.8 --- ---
UNKF: .0012 .0002 .0000 .0000 .0000 .1763 .0760 --- ---
UNCT: 99.0 6.7 .3 .8 .0 20223.5 2500.9 --- ---
UNBG: 46.1 155.7 11.1 40.4 22.3 132.7 7.4 --- ---
ZCOR: 1.1546 1.1605 2.6835 1.0545 1.8328 1.1111 1.6404 --- ---
KRAW: .0108 .0006 .0002 .0001 .0000 .3139 .5080 --- ---
PKBG: 3.15 1.04 1.03 1.02 1.00 153.38 336.85 --- ---
INT%: ---- -58.31 ---- ---- ---- ---- ---- --- ---
Ok, this material is much less pure than the CalChemist material. Specifically around 1400 PPM of potassium, 190 PPM of Cs (note the interference correction magnitude was only some 58%), and possibly 30 PPM of Na and 20 PPM of Ca , but no Mg.
Of course for use as standard for trace Rb, none of these impurities really matter, but it is an interesting exercise.
Now, if we look at the calculated detection limits for these RbTiOPO4 materials we obtain the following for our single point detection limits:
Detection limit at 99 % Confidence in Elemental Weight Percent (Single Line):
ELEM: K Cs Na Ca Mg Ti P
37 .001 .002 .002 .001 .001 .002 .003
38 .001 .002 .002 .001 .001 .002 .003
39 .001 .002 .002 .001 .001 .002 .003
40 .001 .002 .002 .001 .001 .002 .003
41 .001 .002 .002 .001 .001 .002 .002
42 .001 .002 .002 .001 .001 .002 .003
43 .001 .002 .002 .001 .001 .002 .003
44 .001 .002 .002 .001 .001 .002 .003
45 .001 .002 .002 .001 .001 .002 .002
46 .001 .002 .002 .001 .001 .002 .003
47 .001 .002 .002 .001 .001 .002 .003
48 .001 .002 .002 .001 .001 .002 .003
AVER: .001 .002 .002 .001 .001 .002 .003
SDEV: .000 .000 .000 .000 .000 .000 .000
SERR: .000 .000 .000 .000 .000 .000 .000
And for the t-test predictions on the average we obtain:
Detection Limit (t-test) in Elemental Weight Percent (Average of Sample):
ELEM: K Cs Na Ca Mg Ti P
60ci .000 .000 .000 .000 .000 --- ---
80ci .000 .001 .001 .000 .000 --- ---
90ci .000 .001 .001 .000 .000 --- ---
95ci .000 .001 .001 .001 .000 --- ---
99ci .001 .001 .001 .001 .001 --- ---
So basically 10 PPM is our 99% confidence interval t-test for the average. Not too bad.
Finally, since we did measure the RbTiOPO4 for Ti using a TiO2 standard (we had to for the interference correction), and also P using a YPO4 standard let's take a look at the major element chemistry:
The ideal formula for this material is:
ELEM: Rb Ti P O
ELWT: 34.979 19.604 12.676 32.741
If we run the default Armstrong/Reed (modified) correction we obtain:
Un 4 CalChemist RbTiOPO4 #2, Results in Elemental Weight Percents
ELEM: K Cs Na Ca Mg Ti P Rb O
TYPE: ANAL ANAL ANAL ANAL ANAL ANAL ANAL SPEC SPEC
BGDS: MULT MULT MULT MULT MULT EXP EXP
TIME: 320.00 320.00 400.00 400.00 400.00 80.00 80.00 --- ---
BEAM: 99.63 99.63 99.63 99.63 99.63 99.63 99.63 --- ---
ELEM: K Cs Na Ca Mg Ti P Rb O SUM
37 .018 .002 .001 .003 .000 19.428 12.506 34.979 32.741 99.677
38 .017 .002 .000 .003 .000 19.454 12.521 34.979 32.741 99.719
39 .017 .003 .000 .001 -.001 19.486 12.505 34.979 32.741 99.731
40 .018 .003 .001 .003 .000 19.517 12.542 34.979 32.741 99.803
41 .019 .001 .002 .004 .001 19.466 12.499 34.979 32.741 99.711
42 .018 .002 -.002 .002 -.001 19.447 12.511 34.979 32.741 99.698
43 .018 .001 .001 .002 -.001 19.472 12.568 34.979 32.741 99.780
44 .018 .000 .002 .002 .000 19.445 12.496 34.979 32.741 99.682
45 .018 .000 -.002 .002 .000 19.489 12.478 34.979 32.741 99.704
46 .018 -.002 .001 .003 .000 19.528 12.567 34.979 32.741 99.835
47 .018 .001 -.001 .003 -.001 19.489 12.543 34.979 32.741 99.772
48 .019 .002 .001 .002 -.001 19.507 12.506 34.979 32.741 99.756
AVER: .018 .001 .000 .002 .000 19.477 12.520 34.979 32.741 99.739
SDEV: .000 .002 .001 .001 .001 .031 .029 .000 .000 .050
If we run all the matrix corrections we obtain:
Summary of All Calculated (averaged) Matrix Corrections:
Un 4 CalChemist RbTiOPO4 #2
LINEMU Henke (LBL, 1985) < 10KeV / CITZMU > 10KeV
Elemental Weight Percents:
ELEM: K Cs Na Ca Mg Ti P Rb O TOTAL
1 .018 .001 .000 .002 .000 19.477 12.520 34.979 32.741 99.739 Armstrong/Love Scott (default)
2 .017 .001 .000 .002 .000 19.529 12.346 34.979 32.741 99.615 Conventional Philibert/Duncumb-Reed
3 .018 .001 .000 .002 .000 19.729 12.257 34.979 32.741 99.728 Heinrich/Duncumb-Reed
4 .018 .001 .000 .002 .000 19.391 12.546 34.979 32.741 99.678 Love-Scott I
5 .018 .001 .000 .002 .000 19.478 12.557 34.979 32.741 99.776 Love-Scott II
6 .017 .001 .000 .002 .000 18.913 12.456 34.979 32.741 99.109 Packwood Phi(pz) (EPQ-91)
7 .018 .001 .000 .002 .000 19.421 12.602 34.979 32.741 99.764 Bastin (original) Phi(pz)
8 .018 .001 .000 .002 .000 19.401 12.718 34.979 32.741 99.860 Bastin PROZA Phi(pz) (EPQ-91)
9 .018 .001 .000 .002 .000 19.340 12.560 34.979 32.741 99.641 Pouchou and Pichoir-Full (Original)
10 .018 .001 .000 .002 .000 19.379 12.505 34.979 32.741 99.625 Pouchou and Pichoir-Simplified (XPP)
AVER: .018 .001 .000 .002 .000 19.406 12.507 34.979 32.741 99.654
SDEV: .000 .000 .000 .000 .000 .205 .130 .000 .000 .206
SERR: .000 .000 .000 .000 .000 .065 .041 .000 .000
MIN: .017 .001 .000 .002 .000 18.913 12.257 34.979 32.741 99.109
MAX: .018 .001 .000 .002 .000 19.729 12.718 34.979 32.741 99.860
Pretty darn good for such an extrapolation.
So, finally another look at the accuracy of the major elements in our RbTiOPO4 material, where we apply the "fast Monte Carlo" quantification method in Probe for EPMA (derived from the Penepma Penfluor/Fanal Monte-Carlo package by Cesc Salvat and Xavier Llovet), and when we do this we get the best accuracy so far:
Un 4 CalChemist RbTiOPO4 #2, Results in Elemental Weight Percents
ELEM: K Cs Na Ca Mg Ti P Rb O SUM
37 .018 .002 .001 .003 .000 19.646 12.685 34.979 32.741 100.074
38 .018 .002 .000 .003 .000 19.672 12.701 34.979 32.741 100.117
39 .018 .002 .000 .001 -.001 19.704 12.685 34.979 32.741 100.128
40 .018 .003 .001 .003 .000 19.735 12.722 34.979 32.741 100.202
41 .019 .001 .002 .004 .001 19.684 12.679 34.979 32.741 100.109
42 .019 .002 -.002 .002 -.001 19.665 12.690 34.979 32.741 100.096
43 .018 .001 .001 .002 -.001 19.690 12.749 34.979 32.741 100.179
44 .018 .000 .001 .002 .000 19.663 12.675 34.979 32.741 100.079
45 .018 .000 -.002 .002 .000 19.707 12.657 34.979 32.741 100.102
46 .019 -.003 .001 .003 .000 19.746 12.748 34.979 32.741 100.234
47 .018 .000 -.001 .003 -.001 19.708 12.724 34.979 32.741 100.171
48 .019 .002 .001 .002 -.001 19.726 12.685 34.979 32.741 100.154
AVER: .018 .001 .000 .002 .000 19.696 12.700 34.979 32.741 100.137
SDEV: .000 .002 .001 .001 .001 .031 .029 .000 .000 .051
:)
Remember, the ideal formula for RbTiOPO4 is 19.604 wt.% for Ti and 12.676 wt.% for P. This is using the polynomial (three coefficient) alpha factor fit method. If we fit non-linear (4 coefficient) alpha factors we get:
Un 4 CalChemist RbTiOPO4 #2, Results in Elemental Weight Percents
ELEM: K Cs Na Ca Mg Ti P Rb O SUM
37 .018 .002 .001 .003 .000 19.667 12.690 34.979 32.741 100.100
38 .018 .002 .000 .003 .000 19.693 12.706 34.979 32.741 100.142
39 .018 .002 .000 .001 -.001 19.725 12.690 34.979 32.741 100.154
40 .018 .003 .001 .003 .000 19.757 12.727 34.979 32.741 100.228
41 .019 .001 .002 .004 .001 19.705 12.684 34.979 32.741 100.135
42 .019 .002 -.002 .002 -.001 19.686 12.695 34.979 32.741 100.122
43 .018 .001 .001 .002 -.001 19.711 12.754 34.979 32.741 100.205
44 .018 .000 .001 .002 .000 19.684 12.680 34.979 32.741 100.105
45 .018 .000 -.002 .002 .000 19.728 12.662 34.979 32.741 100.128
46 .019 -.003 .001 .003 .000 19.767 12.753 34.979 32.741 100.260
47 .018 .000 -.001 .003 -.001 19.729 12.729 34.979 32.741 100.197
48 .019 .002 .001 .002 -.001 19.747 12.690 34.979 32.741 100.180
AVER: .018 .001 .000 .002 .000 19.717 12.705 34.979 32.741 100.163
SDEV: .000 .002 .001 .001 .001 .031 .029 .000 .000 .051
If the crystal is actually exactly stoichiometric, we might conclude our results are very slightly worse with the non-linear method, but still we seem to have excellent accuracy with the fast Monte-Carlo method.
I've decided to start discussing my attempts to analyze trace elements (Na, Ba, Rb, Sr and Fe) in an orthoclase matrix (Si, Al and K). For these attempts I'm using the MAD-10 orthoclase from CM Taylor because I have some trace element chemistry data on it. If anyone else has performed ICPMS or other bulk trace element methods on this material please feel free to post your data.
The composition of the MAD-10 orthoclase in my standard database is:
St 374 Orthoclase MAD-10
TakeOff = 40.0 KiloVolt = 15.0 Density = 2.590 Type = feldspar
Specimen from Chuck Taylor
Fe2O3=2.01% (EPMA by J. Donovan) (as FeO=1.88% + 0.13% O)
K2O=15.49%, Na2O=1.07% (Flame photometry by J. Hampel)
BaO=0.06%, Rb2O=0.03% (EPMA by J. Donovan)
Sr=12 ppm, Rb=600 ppm (Isotope dilution)
Oxide and Elemental Composition
Average Total Oxygen: 45.798 Average Total Weight%: 100.009
Average Calculated Oxygen: 45.671 Average Atomic Number: 11.991
Average Excess Oxygen: .127 Average Atomic Weight: 21.490
ELEM: SiO2 Al2O3 FeO K2O Na2O BaO Rb2O O
XRAY: ka ka ka ka ka la la ka
OXWT: 64.793 16.720 1.880 15.490 .910 .060 .030 .127
ELWT: 30.286 8.849 1.461 12.859 .675 .054 .027 45.798
KFAC: .2480 .0703 .0123 .1132 .0036 .0004 .0002 .1733
ZCOR: 1.2211 1.2594 1.1909 1.1361 1.8720 1.3740 1.2882 2.6422
AT% : 23.171 7.047 .562 7.067 .631 .008 .007 61.507
24 O: 9.041 2.750 .219 2.757 .246 .003 .003 24.000
My trust is in the Rb and Sr by isotope dilution performed many decades ago by John Christensen when he and I were at UC Berkeley. But I would be interested in other bulk efforts in characterizing these trace elements.
I'll post more later this week, but just to start things off here is a plot of the Rb La region using a TAP Bragg crystal.
(https://smf.probesoftware.com/gallery/395_06_03_19_10_45_44.png)
Note that on PET the Rb La line is up against the high spectrometer limit, while on TAP, the emission line is almost to the bottom of the spectrometer range.
The mostly horizontal blue line is the background model using an exponential fit. I'm not sure I'm very happy with it.
Does anyone know where MAD-10 orthoclase originated from? Geological source/region/location?
My memory is saying something to the effect that MAD stands for Madagascar.
I am quite confident that it is Orthoclase from Itrongahy, Madagascar.
https://www.mindat.org/loc-2273.html