Hi John,
I have a couple of questions regarding implementation of your trace element blank correction, as presented in American Mineralogist, v. 96, p. 274-282:
If I write for the general case for element A,
C(A)_unk = C(A)_std * (I(A)_unk/I(A)_std) * ([ZAF](A)_unk/[ZAF](A)_std),
then do I substitute I(A)_corr from equation 5 (below) for I(A)_unk in the above expression when applying the blank correction? Or am I not looking at it correctly?
I(A)_corr = I(A)_unk - I(A)_std * ((C(A)_meas - C(A)_level) / C(A)_std ) * ([ZAF](A)_std/[ZAF](A)_unk),
where each I is a net intensity, C(A)_level is the known concentration of element A in the blank, and C(A)_meas is the measured concentration of element A in the blank.
Second, how exactly is C(A)_meas determined? Is it done in the simplest possible way, with background interpolated linearly between offsets on either side of the peak? Does it really matter how it's done as long as it's done exactly the same way on the blank as on the unknown?
I'd like to apply the blank correction to analysis of tens to hundreds ppm Au in pyrite, which is a fairly intractable case if using a conventional approach, even if non-linear background models are considered. It appears that the tail of S Ka and/or its satellites cause undulations in the spectrum in the vicinity of both Au Ma and Au Mb. The undulations are difficult to model as a function of spectrometer position, and so the blank correction appears to be the best approach.
I've attached a set of wavelength scans on PET showing Au-bearing pyrite in red and Au-free pyrite in green. I should have set the PHA window a little narrower to better suppress Fe Ka(3), but the scans still clearly show the undulations around both Au Ma and Au Mb.
Thanks,
Brian
Hi Brian,
Yes, this Au in FeS2 example is nasty for background fitting, but pyrite is also (usually) a simple matrix where a pure "blank" standard may be obtained, so is therefore a perfect candidate for using the blank correction!
I'll start by describing it in my software, since this is a very simple process and good for the overview of the details explained later:
1. Acquire your unknown sample as usual (using off-peak or even MAN backgrounds as discussed here):
http://smf.probesoftware.com/index.php?topic=29.msg237#msg237
2. Then yes, acquire another unknown using the *exact same* conditions (PHA settings, count time, beam current etc.) on a standard suitable for the blank correction. Note that this "blank" standard should be a material that is ideally exactly the same composition as your unknown, but with a known (zero) concentration of the element of interest. This known concentration is usually easiest if it is say a pure synthetic and therefore zero due to purity, but the blank "level" as it is called can also be non-zero, so long as it is accurately known! Our Ti in SiO2 synthetic has 1.42 PPM Ti (which could usually be rounded to zero!). The Spectrosil silica described here:
http://smf.probesoftware.com/index.php?topic=130.msg606#msg606
is even purer, but is not crystalline, so I'm not sure how that might effect its use as a blank std for quartz...
3. Finally perform a normal quant analysis on the blank standard and see how far off you are from the zero or non-zero concentration that you are expecting. This is your accuracy check.
So let's say, your FeS2 blank standard contains zero PPM of Au (could be checked by ICP-MS for example), but when you measure it on your EPMA (using off-peak or MAN modeling), you obtain -40 PPM (yes, negative 40 PPM, though the artifact could be positive as well, though it is usually negative due to secondary Bragg diffraction (Ti Ka in SiO2) or self absorption in the sample, which I believe is the case for Au Ma in FeS2).
For example, see the Ti in quartz example here:
http://smf.probesoftware.com/index.php?topic=29.msg387#msg387
So, let's see what we have for our blank correction equation now remembering that we are only making an adjustment to the unknown intensity, I(A)_unk in your first equation:
What we basically want to do is to calculate the intensity represented by the blank accuracy characterization, so that it can be added to the matrix iteration to improve accuracy (and be reflected in the matrix correction if it is not a small adjustment!). That is, because the blank correction is being performed with a full matrix correction (just like the interference correction), the "blank" value could be a minor or even major element! It all comes out in the iteration!
So, you'll want to ratio the blank concentration error, I(A)_blank_meas - I(A)_blank_level, to obtain the intensity error (to put it another way), as seen here:
I(A)_unk_corr = I(A)_unk - I(A)_std * (I(A)_blank_meas - I(A)_blank_level)/ C(A)_std * (ZAF_std/ZAF_unk)
where I(A)_blank_meas is the concentration we actually measure in the blank standard (negative 40 ppm in the pure Fes2 in the above example) and I(A)_blank_level is the concentration that we know is there from purity or ICP-MS or other considerations (ideally zero for a pure FeS2).
Now that we have the corrected I(A)_unk_corr value, that is simply plugged into the matrix correction using the general case (your first equation)...
However, since you aren't using Probe for EPMA (yet!), you won't be easily able to do that, so you could just do a one step subtraction since the matrix correction will not change significantly for a 40 PPM correction in the matrix.
Bottom line, you could skip the quant aspect completely and just subtract the measured from the expected from your unknown measurements, that is, add 40 PPM to all analyses using the same example above!
I think this is a perfect example for the blank correction, and particularly if you decide to use MAN backgrounds, to avoid the off-peak interferences, as described in this abstract:
http://smf.probesoftware.com/index.php?topic=29.msg706#msg706
Note that in the above equations all concentrations should be in weight fraction.
john
I'm thinking of trying the blank correction with a new glass trace element routine, and I have a couple of questions:
(1) Is there a reason why a sample that has been run as a standard can't be selected in PFE for use as the blank correction sample?
(2) How closely does the blank correction sample have to match the unknowns in practice?
Quartz and pyrite are both much simpler matrices than glasses. In my case, I am looking for something appropriate to rhyolitic to dacitic glasses. A blank with a reasonable mean rhyolite-dacite composition might look something like 74% SiO2, 13.5% Al2O3, 2.5% FeO, 3% CaO, 4% Na2O, and 4% K2O with everything else at least under 50 ppm. I am not aware of a synthetic glass that looks like this, so I am wondering if something else might work. Besides pure SiO2, the closest that I currently have is a synthetic K-feldspar glass synthesized by Corning that I obtained from Harvey Belkin at the USGS Reston probe lab (USGS code GFOR). This is 64.7% SiO2, 18% Al2O3, and 16.5% K2O with a little Na2O, FeO, and H2O. The trace element content is unknown, though. Alternatively, NIST 612 might work for a couple of the elements that I am interested in, but it does have a rather high Na content.
Quote from: sckuehn on May 21, 2014, 06:23:26 AM
(1) Is there a reason why a sample that has been run as a standard can't be selected in PFE for use as the blank correction sample?
It is just to ensure that the blank sample is run under the exact same conditions as your unknowns. It matters!
Quote from: sckuehn on May 21, 2014, 06:23:26 AM
(2) How closely does the blank correction sample have to match the unknowns in practice?
The blank correction is designed for simple matrices where a sample with the same matrix but with a zero concentration (or where a non-zero concentration of the trace element is known) is utilized to provide an accuracy reference.
Having said that, I have experimented with using the blank correction on a complex glass as a sort of "accuracy" adjustment. This is possible because the code re-calculates the concentration offset into a quantitative intensity and applies that correction during the matrix iteration.
So for example, I used the blank correction to get better accuracy for measuring oxygen in glass for determination of water (Nash, et al.). Here is an example:
Un 17 Withers-N5, Results in Elemental Weight Percents
ELEM: Na K Cl Ba F Ti Fe Mn Ca Si Al Mg O H
TYPE: ANAL ANAL ANAL ANAL ANAL ANAL ANAL ANAL ANAL ANAL ANAL ANAL ANAL SPEC
BGDS: MAN LIN LIN LIN LIN LOW MAN LIN MAN MAN MAN MAN EXP
TIME: 60.00 20.00 10.00 20.00 40.00 10.00 40.00 10.00 20.00 20.00 20.00 60.00 120.00
BEAM: 9.98 9.98 9.98 9.98 9.98 9.98 9.98 9.98 9.98 9.98 9.98 9.98 9.98
ELEM: Na K Cl Ba F Ti Fe Mn Ca Si Al Mg O H SUM
574 2.780 3.536 .205 .030 .086 .108 3.136 .060 .162 32.680 5.404 .012 50.078 .665 98.943
575 3.072 3.557 .207 .017 .084 .124 3.120 .079 .148 32.990 5.365 .007 49.939 .594 99.303
576 2.846 3.314 .180 .032 .100 .137 3.131 .083 .129 32.914 5.316 .000 49.919 .624 98.727
577 2.925 3.550 .258 .041 .111 .098 3.165 .058 .118 32.858 5.437 -.001 49.985 .621 99.224
578 3.011 3.502 .258 -.014 .083 .072 3.173 .081 .116 32.715 5.364 .009 49.812 .626 98.810
579 3.043 3.561 .239 -.079 .070 .124 3.191 .083 .130 32.812 5.428 .006 50.012 .623 99.244
580 2.842 3.531 .223 .016 .089 .101 3.125 .055 .132 32.799 5.335 -.001 49.969 .644 98.859
581 3.179 3.459 .248 -.019 .121 .118 3.100 .050 .152 33.024 5.362 .006 50.053 .605 99.455
582 2.924 3.505 .164 -.025 .080 .134 3.197 .051 .120 33.124 5.375 .009 49.790 .562 99.009
583 2.890 3.436 .218 -.062 .033 .147 3.165 .031 .143 32.919 5.388 .003 49.908 .609 98.830
584 2.952 3.381 .213 .004 .033 .124 3.124 .064 .133 33.068 5.325 .004 49.852 .588 98.865
585 2.699 3.608 .191 .029 .071 .121 3.150 .061 .161 33.000 5.380 .006 50.223 .641 99.341
AVER: 2.930 3.495 .217 -.003 .080 .117 3.148 .063 .137 32.909 5.373 .005 49.962 .617 99.051
SDEV: .133 .084 .030 .038 .026 .020 .030 .016 .016 .139 .038 .004 .122 .028 .248
SERR: .038 .024 .009 .011 .008 .006 .009 .005 .005 .040 .011 .001 .035 .008
%RSD: 4.54 2.40 13.87-1523.47 33.10 17.21 .97 25.48 11.67 .42 .70 80.01 .24 4.47
STDS: 336 374 285 835 835 22 395 25 358 162 336 12 12 0
STKF: .0735 .1132 .0601 .7430 .1715 .5546 .6779 .7341 .1693 .2018 .1331 .4737 .2328 .0000
STCT: 2447.9 2423.7 839.3 8520.6 2398.7 6097.6 14136.8 13590.2 2247.6 34290.6 23223.7 24410.3 8335.9 .0
UNKF: .0154 .0303 .0017 .0000 .0002 .0010 .0261 .0005 .0012 .2688 .0412 .0000 .2315 .0000
UNCT: 513.0 648.1 24.2 -.2 2.7 10.8 545.2 9.5 16.2 45673.8 7190.2 1.7 8290.2 .0
UNBG: 9.9 12.6 5.0 29.0 3.8 5.8 23.4 16.1 4.4 134.1 103.2 15.2 53.7 .0
ZCOR: 1.9017 1.1548 1.2500 1.3725 4.0763 1.1972 1.2041 1.2234 1.1196 1.2241 1.3036 1.4987 2.1582 .0000
KRAW: .2096 .2674 .0289 .0000 .0011 .0018 .0386 .0007 .0072 1.3320 .3096 .0001 .9945 .0000
PKBG: 52.69 52.56 6.39 1.00 1.77 3.03 24.32 1.61 4.71 341.58 70.64 1.11 156.16 .00
INT%: ---- ---- ---- -3.81 ---- -.02 ---- ---- ---- ---- ---- ---- ---- ----
APF: ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- 1.031 ----
TDI%: 79.930 .454 ---- ---- ---- ---- ---- ---- ---- -.870 ---- ---- -3.030 ----
DEV%: 2.7 2.4 ---- ---- ---- ---- ---- ---- ---- .3 ---- ---- .3 ----
TDIF: QUADRA LINEAR ---- ---- ---- ---- ---- ---- ---- LINEAR ---- ---- LINEAR ----
TDIT: 72.58 30.67 ---- ---- ---- ---- ---- ---- ---- 31.00 ---- ---- 129.42 ----
TDII: 522. 660. ---- ---- ---- ---- ---- ---- ---- 45860. ---- ---- 8327. ----
BLNK#: ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- 19 ----
BLNKL: ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- 43.5580 ----
BLNKV: ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- 44.9622 ----
Here is a white paper on TDI effects when measuring Na, Si and O in hydrous glasses (method of B. Nash, et al.):
http://epmalab.uoregon.edu/reports/Withers%20hydrous%20glass.pdf
You can try with a complex glass for traces but I can't predict how helpful it will be for this situation. But please see this post for the technical details:
http://smf.probesoftware.com/index.php?topic=204.msg919#msg919
Greetings all,
This is not an entirely new topic, but perhaps a new take on an old one.
I have been experimenting with using the blank correction to remove fictive boron from analyses of B-absent glass that results from internal fluorescence of the B4C interlayer of the 200-ansgtrom LSM device, and it has produced a question about how the correction actually works or could work. The main problem with this kind of analysis is that at HV greater than 3-5 kV (which I need for other elements, and am already using two beam conditions) the B Ka intensity is extremely weak. Using a B-free granitic glass for the blank correction, which yields about 0.2 cps/nA at B K-alpha, I get great results for glasses I know to be B-absent (~0.02-0.07 wt% B2O3, relative to a calculated MDL of ~0.3 wt%), a spot-on match for an E-glass I know to contain 5.5 wt% B2O3, and very reasonable concentrations and totals for anhydrous glasses with 1-2 wt% B2O3.
The issue I have is that if the blank correction is applied to my Pyrex standard it yields a value significantly lower than its known value. This tells me that the blank correction works simply by subtracting the concentration of the blank sample from every analysis it is applied to, which isn't a problem for its intended application in trace element analysis using hard x-ray emissions; after all, it isn't intended to be applied to the standard, and subtracting a few ppm from a standard in which the element is of interest is a major element makes no difference to the result. But in the case of boron, the 0.2 cps/nA of the blank can translate to 1-2 wt% fictive B2O3 (a Pyrex with near 13 wt% yields only about 1.4 cps/nA at 15 kV - I told you, the intensity is weak).
So although the blank correction apparently yields good results for low-boron glasses, its application to the standard yields a poor result(1-2 wt% low in B2O3). Could, or does, the blank correction simply work such that the intensity corresponding to the bank represents zero concentration and then the intensity corresponding to the standard (or standard intensity minus the intensity of the blank) is then used to calculate the cps/na/wt% used for calculating concentration?
In other words, rather than subtracting a concentration, subtract the intensity and then calculate composition? If the intensity of the blank were defined as zero, the process would be much more like scaling image intensities via histogram stretching (changing the "display range"). I wonder if such a method wouldn't better linearize the intensity-concentration relationship, and could be applied to all analyses (including standards and standards analyzed as unknowns).
Quote from: gmorgan@ou.edu on September 25, 2015, 07:11:25 AMGreetings all,
This is not an entirely new topic, but perhaps a new take on an old one.
I have been experimenting with using the blank correction to remove fictive boron from analyses of B-absent glass that results from internal fluorescence of the B4C interlayer of the 200-ansgtrom LSM device, and it has produced a question about how the correction actually works or could work. The main problem with this kind of analysis is that at HV greater than 3-5 kV (which I need for other elements, and am already using two beam conditions) the B Ka intensity is extremely weak. Using a B-free granitic glass for the blank correction, which yields about 0.2 cps/nA at B K-alpha, I get great results for glasses I know to be B-absent (~0.02-0.07 wt% B2O3, relative to a calculated MDL of ~0.3 wt%), a spot-on match for an E-glass I know to contain 5.5 wt% B2O3, and very reasonable concentrations and totals for anhydrous glasses with 1-2 wt% B2O3.
The issue I have is that if the blank correction is applied to my Pyrex standard it yields a value significantly lower than its known value. This tells me that the blank correction works simply by subtracting the concentration of the blank sample from every analysis it is applied to, which isn't a problem for its intended application in trace element analysis using hard x-ray emissions; after all, it isn't intended to be applied to the standard, and subtracting a few ppm from a standard in which the element is of interest is a major element makes no difference to the result. But in the case of boron, the 0.2 cps/nA of the blank can translate to 1-2 wt% fictive B2O3 (a Pyrex with near 13 wt% yields only about 1.4 cps/nA at 15 kV - I told you, the intensity is weak).
So although the blank correction apparently yields good results for low-boron glasses, its application to the standard yields a poor result(1-2 wt% low in B2O3). Could, or does, the blank correction simply work such that the intensity corresponding to the bank represents zero concentration and then the intensity corresponding to the standard (or standard intensity minus the intensity of the blank) is then used to calculate the cps/na/wt% used for calculating concentration?
In other words, rather than subtracting a concentration, subtract the intensity and then calculate composition? If the intensity of the blank were defined as zero, the process would be much more like scaling image intensities via histogram stretching (changing the "display range"). I wonder if such a method wouldn't better linearize the intensity-concentration relationship, and could be applied to all analyses (including standards and standards analyzed as unknowns).
Hi George,
You could have saved yourself a little typing if you'd read the blank correction paper first... :(
In fact the blank correction in Probe for EPMA does subtract the corrected intensity from the measurement as described in the paper. Here's how: the concentration difference between the "known" blank standard value, and what you actually measure on that standard, is calculated as a k-ratio intensity and then subtracted during the matrix iteration for unknown samples. See the link here:
http://smf.probesoftware.com/index.php?topic=29.0
Even so, the blank correction should only be applied when the blank standard and the unknown are very similar in matrix. In the case of boron measurement, the situation is even more critical due to the huge absorption corrections present.
Try reading the original paper
http://probesoftware.com/download/Ti%20in%20Quartz,%20Am.%20Min.%20Donovan,%202011.pdf
Then try reading the above posts in this topic and also try reading the boron analysis topic here:
http://smf.probesoftware.com/index.php?topic=248.0
What you are attempting to do is possible I suspect (though I've never tried analyzing silicate glasses for boron!), but I'm pretty sure it's not going to be easy!
But you've come to the right place to ask... John Fournelle I know has looked at boron...
Let me discuss an example that I think will help.
There are so many way one can "die" using the blank correction, but here is one out of many: I'm analyzing a trace element in a mineral or glass, and the standard I decided to use for making the blank correction is almost exactly the same as my unknown, but in fact has a spectral interference from a minor element on the trace element I am attempting to measure. An interference that my actual unknown matrix does not have...
So my blank registers a significantly higher blank "value" (measurement) than the expected blank "level" (known) and therefore when applied to my actual unknown, over corrects the measured intensity resulting in a negative k-ratio.
Remember the standard used for the blank correction should be as exactly the same matrix as the unknown as possible, but with a zero or known non-zero concentration of the element one is trying to measure. It will not always be possible to obtain a proper blank standard for many sorts of compositions.
Quote from: gmorgan@ou.edu on September 25, 2015, 07:11:25 AM
So although the blank correction apparently yields good results for low-boron glasses, its application to the standard yields a poor result(1-2 wt% low in B2O3). Could, or does, the blank correction simply work such that the intensity corresponding to the bank represents zero concentration and then the intensity corresponding to the standard (or standard intensity minus the intensity of the blank) is then used to calculate the cps/na/wt% used for calculating concentration?
Hi George,
If you are asking above if the blank correction can be based on, not only a zero concentration standard, but also on a *non zero* concentration in the blank standard, the answer is : yes!
See the "Blank Level" field circled here:
(https://smf.probesoftware.com/oldpics/i41.tinypic.com/2w6xf8x.jpg)
This is discussed in the post linked to here:
http://smf.probesoftware.com/index.php?topic=29.msg387#msg387
john
Okay, so I see that the blank correction does an intensity subtraction to the samples. But again, it does not remove the blank intensity from the intensity of the primary standard to yield a new, blank subtracted, intensity for it: i.e., such that I(Std)corrected = I(Std)original - I(blank). If this procedure were done - in other words also blank correcting the standard intensity - then Analyzing the standard with the blank correction would still yield the correct concentration for the standard, which the present procedure does not do.
As an example, imagine using a Pyrex standard with 12.8 wt% B2O3 for boron intensity, and that intensity acquisition and analysis of a B-free glass yields fictive 1.5 wt% B2O3. Using the B-free glass as the "blank" during Analysis removes the fictive boron from the samples just fine, but if the Pyrex standard is Analyzed using the blank correction it now yields 11.3 wt% B2O3.
So I get that this is a somewhat different kettle of fish than what the blank correction was intended for, and that such concerns are unimportant for analyzing most heavier trace elements at very low concentrations. I am grateful to see that, as presently applied, the blank correction does seem to yield a good result for glasses of low to intermediate boron content, but just find it a bit disconcerting that the blank correction seems to yield correct results for everything BUT the primary standard. Given such poor counting statistics for boron, I worry about the accuracy of concentration values between the detection limit and that of the primary standard - just how linear is the intensity-concentration relationship? Without being able to reproduce the standard concentration when using the blank correction I am left somewhat unable to evaluate this.
Quote from: gmorgan@ou.edu on September 28, 2015, 09:57:23 AM
Okay, so I see that the blank correction does an intensity subtraction to the samples. But again, it does not remove the blank intensity from the intensity of the primary standard to yield a new, blank subtracted, intensity for it: i.e., such that I(Std)corrected = I(Std)original - I(blank). If this procedure were done - in other words also blank correcting the standard intensity - then Analyzing the standard with the blank correction would still yield the correct concentration for the standard, which the present procedure does not do.
As an example, imagine using a Pyrex standard with 12.8 wt% B2O3 for boron intensity, and that intensity acquisition and analysis of a B-free glass yields fictive 1.5 wt% B2O3. Using the B-free glass as the "blank" during Analysis removes the fictive boron from the samples just fine, but if the Pyrex standard is Analyzed using the blank correction it now yields 11.3 wt% B2O3.
So I get that this is a somewhat different kettle of fish than what the blank correction was intended for, and that such concerns are unimportant for analyzing most heavier trace elements at very low concentrations. I am grateful to see that, as presently applied, the blank correction does seem to yield a good result for glasses of low to intermediate boron content, but just find it a bit disconcerting that the blank correction seems to yield correct results for everything BUT the primary standard. Given such poor counting statistics for boron, I worry about the accuracy of concentration values between the detection limit and that of the primary standard - just how linear is the intensity-concentration relationship? Without being able to reproduce the standard concentration when using the blank correction I am left somewhat unable to evaluate this.
Hi George,
Yes, the blank correction is intended only for correction of unknown samples.
This is because the specimen (standard) used for the blank correction must be run at exactly the same conditions, etc as the unknown to be blank corrected. Allowing only unknown samples for the blank correction helps to ensure this.
What you are attempting to do, is not only a "different kettle of fish", but I think you are "barking up the wrong tree"! I think you need to first look at all the other light element issues for example, background fitting, area peak factors and MACs. Quant boron is not for the "faint of heart". Here are some links to some magnesium boride analyses I performed, maybe this will help:
http://smf.probesoftware.com/index.php?topic=248.0
http://smf.probesoftware.com/index.php?topic=536.0
As you say, the blank correction is designed for situations where one wants to first, check if they can measure zero (or a known non-zero) in a standard material that is exactly like their unknown, but contains a zero (*or known non-zero) concentration. Then subsequently (and optionally) apply that observed offset to correct for actual unknowns.
So I think you've got "lots of other fish to fry" with the light element backgrounds, APFs, and MACs, not to mention beam sensitivity and sample damage...
In any case, just run your standard as an unknown and then you can see how good a job the blank correction does on it, right?
john
Hi George,
It just occurred to me what might be very helpful to you for your boron analyses. Assuming that you can obtain a set of boron silicate glass standards with a similar matrix to your unknown boro-silicate glasses and covering the range of boron concentrations you have in your unknowns...
That is try the "multi-standard" calibration curve option as described here:
http://smf.probesoftware.com/index.php?topic=461.msg2528#msg2528
By the way, this calibration curve method is what many Japanese investigators utilize for measuring trace carbon in steel. Some European steel companies utilize a different method for background determination of carbon, that is they measure a "background" on pure Fe metal and then simply subtract that intensity from both the carbon standard and the Fe unknown. It's essentially a very crude version of my MAN background method, albeit only using a single standard and no continuum absorption correction!
Now, if you can't obtain a suitable set of standard boro-silicate glass standards, you'll have to go "full monty" in PFE with fancy background corrections, APFs, MACs, etc.
Note however, that you can acquire either MAN or off-peak measurements for this multi-standard calibration curve method. If you utilize the off-peak measurements for boron Ka, you can also take advantage of the zero point fit option as described here:
http://smf.probesoftware.com/index.php?topic=461.msg2531#msg2531
john
Hi John,
If your calibration curve reference materials are similar to your Unknown are background measurements needed at all? The carbon in steel is a good example - the bulk is essentially the same for the CC RM's and the sample with just the trace C level changing. What's going to make the background change?
Quote from: Mike Matthews on September 29, 2015, 01:20:48 PM
Hi John,
If your calibration curve reference materials are similar to your Unknown are background measurements needed at all? The carbon in steel is a good example - the bulk is essentially the same for the CC RM's and the sample with just the trace C level changing. What's going to make the background change?
Hi Mike,
That is exactly right. It shouldn't matter.
But I allow the user to utilize background corrected intensities for the multi-standard calibration curve so one can optionally include a 0,0 point for fitting trace elements. In other words, if one is utilizing background corrected intensities for the calibration curve method, a zero concentration *should* yield a zero intensity!
But as you say, on-peak only (MAN) acquisitions should also be fine when using the multi-standard calibration curve method.
john
Hey John,
With the calibration curve method for boron, the only thing that can be measured at one time would be boron, correct? That is, unless all elements are analyzed by calibration curve? I mean, can boron be quantified with a calibration curve while other components are quantified using off-peak intensities from single standards? I am measuring compositional gradients in crystal-glass systems, where the glass represents silicate liquid quenched at particular stages of crystal growth, to look at chemical systematics (long-range diffusion, etc.). So measuring only boron is not really an option (yes, the matrices are similar, but composition does vary with position in each experiment - and that is what we are evaluating).
Much of the problem with measuring boron is due to the internal fluorescence from the Mo-BC4 LSM device. In the old days I could minimize this by acquiring and overlaying broad range WDS scans from different standards to select the peak position and background offsets that minimize or eliminate that fluorescence in the glasses, but I haven't found a convenient way to perform near full range scans and overlay them with PFE as was easily available with my previous automation system. In the PFE documentation I only see how to perform wavescans for individual elements based on their setups in the Elements menu, and to plot them one at a time. I presume it should be possible, and that's one thing I want to work on with Gareth during his next visit.
Quote from: gmorgan@ou.edu on September 30, 2015, 10:41:28 AM
With the calibration curve method for boron, the only thing that can be measured at one time would be boron, correct? That is, unless all elements are analyzed by calibration curve? I mean, can boron be quantified with a calibration curve while other components are quantified using off-peak intensities from single standards? I am measuring compositional gradients in crystal-glass systems, where the glass represents silicate liquid quenched at particular stages of crystal growth, to look at chemical systematics (long-range diffusion, etc.). So measuring only boron is not really an option (yes, the matrices are similar, but composition does vary with position in each experiment - and that is what we are evaluating).
Hi George,
That is correct that it's either one quantification method or the other. But you can post process the same data using both methods and then combine them later on.
Quote from: gmorgan@ou.edu on September 30, 2015, 10:41:28 AM
Much of the problem with measuring boron is due to the internal fluorescence from the Mo-BC4 LSM device. In the old days I could minimize this by acquiring and overlaying broad range WDS scans from different standards to select the peak position and background offsets that minimize or eliminate that fluorescence in the glasses, but I haven't found a convenient way to perform near full range scans and overlay them with PFE as was easily available with my previous automation system. In the PFE documentation I only see how to perform wavescans for individual elements based on their setups in the Elements menu, and to plot them one at a time. I presume it should be possible, and that's one thing I want to work on with Gareth during his next visit.
If you select multiple wavescan samples you can plot them all at once. See here:
http://smf.probesoftware.com/index.php?topic=42.msg2833#msg2833
This is a feature available since v. 10.9.x or so, but you should update to v11 just to be sure.
john
John,
You were definitely right about a better background model helping. Although it cuts off a little of the peak, widening the background on the high energy (low sine-theta) side of the peak seems to have removed most of the intensity from the internal fluorescence (due to the high intensity of background on that side of the peak in aluminosilicates). What previously gave me ~0.24 cps/nA at zero wt% B2O3 - that made me try the blank correction - now is averaging very close 0.00 cps/nA. So the blank correction now seems unnecessary (okay, inappropriate)...
Thanks.
Greetings Probe (for EPMA) Guru :)
Reviving this topic with an additional complication: peak interference (briefly mentioned in a sub-message of this post) and slightly different matrices compositions.
I tried for the first time tonight a blank correction on some glass analyses with the goal to measure traces (Rb, Ba, and Zr, all in the range of ca. 100 to 2000 ppm), not really great so far, but I have other analyses running overnight that I can "play" with tomorrow morning. All analyses have the same conditions, and I'm using glass materials that are of similar chemistry, although not perfectly the same... Most of these glasses are Ti-bearing, with variable content from around 0.5% to 1.5%, and thus I should run a peak interference correction on Ba when using the H-type spectrometer, PET monochromator. I'm using the MAN and to perform "good" analysis, I'm also analysing a synthetic glass that has either nothing or very low content of these elements (one is supposedly pure, the other sample has LA-ICP-MS data so I now the content at the sub-ppm level).
When I activated the blank correction, it warned me that maybe I should remove the peak interference correction to avoid an over correction. I can understand this is correct (that it will overcorrect) IF the blank standard AND the unknown have exactly the same Ti content, but... The blank correction reference material and the unknown samples have different Ti-content (up to 0.5-1.0 wt% difference)! Can the blank correction handle this? Would you say this is not possible at all to correct for this?
Other question: how "close" should the blank standard be from the analyzed material? I know you developed the blank correction for simple matrices (e.g. quartz), but it might be very useful for beam sensitive materials with a little bit more complex matrix, too, usually with similar SiO2 content, but with a couple to maybe max 3-4 wt% element change on the other elements (especially Al, Ca, Na, K, Fe and Mg)... Have you tested this? I really would like to use the MAN on these glass materials, as (a) they are beam sensitive, (b) I am forced to use a small beam size (5 um), and (c) I need to reach sub-100 ppm detection limit in the least amount of time...
If not possible, then I will either use a simple two-point background acquisition in the time / current / beam size that will prevent any beam damage, or I will consider the Alternating On-and-Off background correction... Not that it is not possible in such samples to have for instance a background acquired on the first point and then only the peak on the n-th point, as the inclusion are often too small to set two points without having them overlapping.
Julien
P.S. Yes, I know, I'm probably asking for the impossible, but you like a good challenge from time to time, no? :D
Quote from: Julien on February 18, 2019, 09:58:12 AM
When I activated the blank correction, it warned me that maybe I should remove the peak interference correction to avoid an over correction. I can understand this is correct (that it will overcorrect) IF the blank standard AND the unknown have exactly the same Ti content, but... The blank correction reference material and the unknown samples have different Ti-content (up to 0.5-1.0 wt% difference)! Can the blank correction handle this? Would you say this is not possible at all to correct for this?
Hi Julien,
You always have such great questions! :-*
I have to admit that this is a topic that I haven't thought about as much as it probably should be. It's complicated, at least to me. But in fact in thinking about this a bit more today, I've decided that maybe the warning you are getting about applying both an interference correction and a blank correction may not even be necessary! I'm not exactly sure, but keep reading. In any case it's always best to check using actual data to check one's assumptions!
So I recently was testing some integrated EDS and WDS analyses where I used WDS for the trace elements and EDS for the major elements, the idea being to reduce the number of elements by WDS, thus speeding up the analysis and damaging the sample less, as described here a few months ago for those that are interested:
https://smf.probesoftware.com/index.php?topic=79.msg7818#msg7818
As you know for some situations, e.g., Rb or Sr La are interfered by Si, so it's important that we can correct for spectral interferences between WDS *and* EDS elements, particularly if one is doing traces by WDS and major elements by EDS! Anyway, the point being that we can do a quick test using this data set for the interference correction and the blank correction separately, and also together, to try and answer your questions. It's not a perfect test because there were a limited number of samples, but let's see what we find anyway... please note I'm going to pitch this post to a general audience so I'm going to go over some issues that I know you are very much aware of, probably much more than myself!
So (starting at the beginning) if we do not apply the interference correction for Si interfering with Sr La, nor apply a blank correction, we get the following results on a SiO2 standard (which has zero Sr from ICP-MS by Alan Koenig):
ELEM: Na Ba Rb Sr Fe Si Al K O SUM
60 .001 .002 -.054 .063 .014 46.717 .048 .000 53.282 100.074
61 .002 .000 -.036 .076 .011 46.971 .045 .000 53.571 100.640
62 .000 -.008 -.053 .068 -.009 46.890 .042 .000 53.466 100.396
63 .002 .006 -.057 .064 .000 46.767 .035 .000 53.323 100.139
AVER: .001 .000 -.050 .067 .004 46.836 .043 .000 53.410 100.313
SDEV: .001 .006 .010 .006 .010 .115 .006 .000 .133 .259
and here for an orthoclase standard (which has 12 PPM (0.0012 wt.%) Sr from isotope dilution by John Christensen):
ELEM: Na Ba Rb Sr Fe Si Al K O SUM
56 .718 .049 .039 .052 1.350 30.095 8.552 12.896 45.190 98.942
57 .747 .071 .027 .043 1.320 30.075 8.544 12.888 45.160 98.875
58 .753 .060 .015 .056 1.344 30.123 8.575 12.930 45.259 99.115
59 .731 .064 .017 .039 1.331 29.870 8.467 12.808 44.837 98.165
AVER: .737 .061 .025 .048 1.336 30.041 8.535 12.881 45.111 98.774
SDEV: .016 .009 .011 .008 .013 .115 .047 .052 .188 .419
Please ignore the Rb results. It would be nice to also check the Rb values but there clearly is a problem with the background measurement, so at least for now, let's just focus on these two Sr measurements.
So it would appear that there is an interference from Si on the Sr La emission line. First because the SiO2 shows a greater amount of Sr than the orthoclase, but also because our nominal overlap calculation model from the Elements/Cations dialog in Probe for EPMA shows this:
For Sr la LPET at 6.86280 angstroms, at an assumed concentration of 1 wt.%
Interference by Si SKB` at 6.81610 ( 77894.2) ( -533.78) = 2.0%
Interference by Rb LB4 at 6.82360 ( 77979.9) ( -448.06) = 16.5%
On Peak Position ------------- at 6.86280 ( 78428.0)
Interference by K SKB`` II at 6.88250 ( 78653.2) ( 225.180) = 10.0%
Interference by K SKB^5 II at 6.89860 ( 78837.2) ( 409.211) = 2.0%
Interference by K KB1 II at 6.90910 ( 78957.2) ( 529.227) = 4.5%
Interference by K KB3 II at 6.90910 ( 78957.2) ( 529.227) = 2.5%
When one has an interfering emission line from a major element, those tails can extend a long way...
Now, if we apply the interference correction in Probe for EPMA for Sr interfered by Si using the SiO2 standard as the standard for the interference correction, we get the following results for SiO2 measured as an unknown:
ELEM: Na Ba Rb Sr Fe Si Al K O SUM
60 .001 .002 -.054 -.004 .014 46.721 .048 .000 53.274 100.003
61 .002 .000 -.036 .008 .011 46.975 .045 .000 53.563 100.569
62 .000 -.008 -.053 .000 -.009 46.894 .042 .000 53.458 100.325
63 .002 .006 -.057 -.004 .000 46.771 .035 .000 53.316 100.069
AVER: .001 .000 -.050 .000 .004 46.840 .043 .000 53.403 100.242
SDEV: .001 .006 .010 .006 .010 .115 .006 .000 .133 .259
Which is pretty darn good, though expected since the SiO2 standard and the SiO2 unknown were the same material, though measured in different spots on the standard. And here is the orthoclase standard measured as an unknown:
ELEM: Na Ba Rb Sr Fe Si Al K O SUM
56 .718 .049 .039 .009 1.350 30.098 8.553 12.896 45.186 98.898
57 .747 .071 .027 .000 1.320 30.077 8.545 12.888 45.155 98.830
58 .753 .060 .015 .013 1.344 30.126 8.575 12.930 45.255 99.071
59 .731 .064 .017 -.004 1.331 29.873 8.467 12.808 44.832 98.121
AVER: .737 .061 .025 .004 1.336 30.043 8.535 12.881 45.107 98.730
SDEV: .016 .009 .011 .008 .013 .115 .047 .052 .188 .418
The isotope dilution gave us 12 PPM Sr, but our variance is 80 PPM, so statistically a zero concentration.
Now lets turn off the interference corrections and turn on the blank correction instead. Again, here is SiO2 as an unknown using itself for the blank correction:
ELEM: Na Ba Rb Sr Fe Si Al K O SUM
60 .001 .002 -.054 -.004 .014 46.721 .048 .000 53.274 100.003
61 .002 .000 -.036 .008 .011 46.975 .045 .000 53.563 100.569
62 .000 -.008 -.053 .000 -.009 46.894 .042 .000 53.458 100.325
63 .002 .006 -.057 -.004 .000 46.771 .035 .000 53.316 100.068
AVER: .001 .000 -.050 .000 .004 46.840 .043 .000 53.403 100.242
SDEV: .001 .006 .010 .006 .010 .115 .006 .000 .133 .259
As expected, and as near as we can tell we get zero. Which is also the same result we got for the interference correction. But that is merely because we assigned the SiO2 unknown as the blank correction to itself. Now let's look at the orthoclase standard using the same SiO2 unknown for the blank correction:
ELEM: Na Ba Rb Sr Fe Si Al K O SUM
56 .718 .049 .039 -.015 1.350 30.099 8.553 12.896 45.183 98.873
57 .747 .071 .027 -.024 1.320 30.079 8.545 12.888 45.153 98.806
58 .753 .060 .015 -.011 1.344 30.127 8.575 12.930 45.252 99.046
59 .731 .064 .017 -.029 1.332 29.875 8.468 12.808 44.830 98.096
AVER: .737 .061 .025 -.020 1.336 30.045 8.535 12.881 45.105 98.705
SDEV: .016 .009 .011 .008 .013 .115 .047 .052 .188 .419
Whoa! Now we are getting a *negative* 200 PPM for Sr. It's a significant *over correction* clearly. Why would that be?
Well, the interference correction is based on the actual concentration of the interfering element, in this case Si. And since there is an actual interference here, the interference correction handles the situation very well. But the blank correction is *not* based on the concentration of any particular element. It merely assumes that there is some sort of measurement artifact, what exactly we may not know, but it assumes that the measurement artifact is constant regardless of composition. Perhaps something like a detector absorption edge or a secondary Bragg reflection artifact like this:
(https://smf.probesoftware.com/gallery/1_19_02_19_12_23_41.png)
So we would expect that since the concentration of Si, which is causing the interference, is different for SiO2 and orthoclase, that the blank correction would be unsuitable for this situation and indeed that appears to be the case.
OK, but what happens if we apply *both* the interference correction and the blank correction to these samples? Here is the SiO2 standard with the interference correction for Si on Sr La using SiO2 as our interference standard and also the blank correction using the SiO2 unknown as the blank correction sample:
ELEM: Na Ba Rb Sr Fe Si Al K O SUM
60 .001 .002 -.054 -.004 .014 46.721 .048 .000 53.274 100.003
61 .002 .000 -.036 .008 .011 46.975 .045 .000 53.563 100.569
62 .000 -.008 -.053 .000 -.009 46.894 .042 .000 53.458 100.325
63 .002 .006 -.057 -.004 .000 46.771 .035 .000 53.316 100.069
AVER: .001 .000 -.050 .000 .004 46.840 .043 .000 53.403 100.242
SDEV: .001 .006 .010 .006 .010 .115 .006 .000 .133 .259
And since we're using the SiO2 standard as the standard for the interference correction and we're using the SiO2 unknown as the blank correction sample, we get zero as expected. But now let's try the orthoclase standard with both corrections turned on:
ELEM: Na Ba Rb Sr Fe Si Al K O SUM
56 .718 .049 .039 .009 1.350 30.098 8.553 12.896 45.186 98.898
57 .747 .071 .027 .000 1.320 30.077 8.545 12.888 45.155 98.830
58 .753 .060 .015 .013 1.344 30.126 8.575 12.930 45.255 99.071
59 .731 .064 .017 -.004 1.331 29.873 8.467 12.808 44.832 98.121
AVER: .737 .061 .025 .004 1.336 30.043 8.535 12.881 45.107 98.730
SDEV: .016 .009 .011 .008 .013 .115 .047 .052 .188 .418
Weird! We're getting the same result we got for the interference correction only! How is that possible? Well a closer look at the blank correction "value" on this orthoclase analysis shows us why:
ZCOR: 1.8699 1.3745 1.2860 1.2359 1.1909 1.2191 1.2592 1.1361 ---
KRAW: .0536 .0006 .0006 .0001 .0166 .6009 .5086 1.0017 ---
PKBG: 12.74 1.19 1.07 1.06 11.90 .00 .00 .00 ---
INT%: ---- ---- ---- -92.75 ---- ---- ---- ---- ---
BLNK#: ---- ---- ---- 9 ---- ---- ---- ---- ---
BLNKL: ---- ---- ---- .000000 ---- ---- ---- ---- ---
BLNKV: ---- ---- ---- .000070 ---- ---- ---- ---- ---
Two things to look at here. First the interference correction shows a -92% correction leaving us with 40 PPM of Sr. Now our variance is 80 PPM so the 40 PPM result is statistically a zero, but it is a tiny bit suggestive, because according to isotope dilution do we have 12 PPM of Sr in this orthoclase standard. So a higher precision measurement is necessary before we proceed with any further speculation, and I would probably use the MAN method for best trace element precision combined with a blank correction for an accuracy correction, so next time I get a chance I'll try that measurement.
Second, look at the blank correction value (BLNKV). It's 0.00007 wt% or 0.7 PPM! What does this mean? It means that after the interference correction is applied to the sample, the blank correction calculates the difference between what we measured and what we should have obtained (that's the BLNKL or blank correction level), and that was only 0.7 PPM or essentially zero. So when we apply that blank correction of 0.7 PPM to either the SiO2 as an unknown or the orthoclase as an unknown, there is essentially no effect on the data.
I know, it hurts my brain too but it actually makes sense. I would be very interested in hearing from you (Julien), and/or any one else on what you find on your samples. Please try some tests on some well characterized standards that have trace elements measured and let's see what we find.
Here's a great example of a sample absorption edge artifact requring special treatment in a recent (Kovalchuk et al., 2024) paper:
(https://smf.probesoftware.com/gallery/395_27_03_25_4_17_42.png)
https://www.mdpi.com/2075-163X/14/2/170
Yes, one could play all sorts of games trying to get one's off-peak positions adjusted exactly right, or... one could obtain a high purity synthetic (or even natural) pyrite with no detectable gold, and utilize the blank correction in Probe for EPMA.
https://smf.probesoftware.com/index.php?topic=454.msg6694#msg6694
https://smf.probesoftware.com/index.php?topic=1397.0
Also see the discussion in the previous post in this topic regarding using the interference and blank corrections together (it seems to be OK in spite of the warning the software pops up with):
https://smf.probesoftware.com/index.php?topic=204.msg8096#msg8096