Blank correction for analysis of vanadium in rutile

[Brian Joy]

I’ve attached a couple of plots that show graphically how useful the blank correction can be.  I have a rutile standard (NMNH 120812) that I know contains vanadium, but I’ve had a great deal of difficulty verifying the amount precisely due to interference from Ti Kβ_1,3, Ti Kβ₅, associated satellites, and also the Ti K absorption edge.  Of course, this is a worst-case scenario in which the Ti concentration is very high and the V concentration is very low.  Recently, I acquired a high-purity synthetic TiO₂ standard, and I’ve overlaid its spectrum (using LiFL) on that of the natural rutile in the first plot below.  In the second plot, I’ve essentially done a manual blank correction by subtracting the high purity TiO₂ spectrum from that of the natural rutile (difference illustrated in green).  There is absolutely no way to perform an overlap correction with confidence in this case, and the blank correction provides the only useful result (variable between ~0.2 and ~0.3 wt% V₂O₃ in this case), especially since natural rutile is generally close to end-member composition.  In this particular case, the rutile contains ~98.9 wt% TiO₂.  I hope that someone will find this useful for both the general case and for this specific case.  I often find that people minimize the importance of interference from Ti at the V Kα peak position when using LiF.

[Probeman]

I often find that people minimize the importance of interference from Ti at the V Kα peak position when using LiF.

Absolutely. The interference on V Ka from Ti Kb is observable even on LiF.

But I am surprised that you are doing this manually for two reasons. One, you posted this in the Probe for EPMA board, and two, don't you own the Probe for EPMA software? Why don't you utilize the quant iterated interference correction in PFE?  I think you will find it works excellently.

Interestingly it was seeing reports of vanadium in rutile, from when I started out in EPMA in the early 1990s, that was the impetus for developing the quantitative (iterated) interference correction:

http://www.geology.wisc.edu/~johnf/g777/777MicrobeamAnalysis/1993-JJD-InterferenceTrace.pdf

The problem back then was the use of a simple intensity ratio (e.g, Gilfrich) for the interference (overlap) correction procedure (that is, if an interference correction was utilized at all!). The point being that only by including the matrix effects between the unknown, standard *and* the standard used for the interference correction, is the interference correction truly quantitative.

Without an interference correction, the apparent concentration of V in TiO2 is around 0.1 to 0.2 wt% using LiF crystals, and of course larger for PET crystals.  So I agree, it certainly cannot be ignored.

But I am also curious why you found the need for a blank correction as I have found the PFE interference interference correction method to be more than sufficient for this rather small interference correction. Even using a PET crystal, in my experience, the iterated quant interference correction seems to work well for the interference of Ti Kb on V Ka.

Of course if the issue is interpolating the background intensity across an absorption edge, then the solution is the MAN background correction.

[Brian Joy]

I often find that people minimize the importance of interference from Ti at the V Kα peak position when using LiF.

Absolutely. The interference on V Ka from Ti Kb is observable even on LiF.

But I am surprised that you are doing this manually for two reasons. One, you posted this in the Probe for EPMA board, and two, don't you own the Probe for EPMA software? Why don't you utilize the quant iterated interference correction in PFE?  I think you will find it works excellently.

Interestingly it was seeing reports of vanadium in rutile, from when I started out in EPMA in the early 1990s, that was the impetus for developing the quantitative (iterated) interference correction:

http://www.geology.wisc.edu/~johnf/g777/777MicrobeamAnalysis/1993-JJD-InterferenceTrace.pdf

The problem back then was the use of a simple intensity ratio (e.g, Gilfrich) for the interference (overlap) correction procedure (that is, if an interference correction was utilized at all!). The point being that only by including the matrix effects between the unknown, standard *and* the standard used for the interference correction, is the interference correction truly quantitative.

Without an interference correction, the apparent concentration of V in TiO2 is around 0.1 to 0.2 wt% using LiF crystals, and of course larger for PET crystals.  So I agree, it certainly cannot be ignored.

But I am also curious why you found the need for a blank correction as I have found the PFE interference interference correction method to be more than sufficient for this rather small interference correction. Even using a PET crystal, in my experience, the iterated quant interference correction seems to work well for the interference of Ti Kb on V Ka.

Of course if the issue is interpolating the background intensity across an absorption edge, then the solution is the MAN background correction.

Hi John,

I never measure background on opposing sides of a K absorption edge when the interfering element is present in high concentration; it's bad practice to do so.  Also, I've tried modeling background radiation (for a purpose different than this), and I've not yet come up with a satisfactory approach, especially when absorption edges are included; this makes me reluctant to use the MAN background correction, especially when I'm interested in quantification in the 2000-3000 ppm range.

I have in fact used PFE with the blank correction to address the problem of analysis of V in rutile, and I find that it works very nicely.  Despite this, I like to have a visual aid to show myself (and others) the exact nature and magnitude of the problem.  Considering that, in this case, Ti contributes more net radiation at the V Ka peak position than V itself indicates the interference is not "small."  This term is too subjective.  The reason I posted these plots is that I found myself referring to them today while analyzing chromite, as I needed to remind myself of the exact nature of the interference.  I can delete the plots if you'd like.

I'm still not able to use PFE routinely, as a communication error occurs between the JEOL computer and the operator electronics.  (I thought we were upgrading the JEOL software to Windows 10 two years ago, but it turns out I was mistaken.)

Brian

[John Donovan]

Yes, I agree that extrapolating (or interpolating) across absorption edges is a bad idea.

That's the cool thing about the MAN background correction as it doesn't use off-peak measurements at all.  Also it does include a correction for continuum absorption, so in silicates and oxides the accuracy is good to around 200 to 300 PPM. Of course if one utilizes the blank correction in addition to the MAN correction, the accuracy will be equal to the precision of the measurement. So easily down to 50 to 100 PPM with a reasonable integration times.  I personally feel this is the best combination: MAN and blank correction (if a suitable blank is available) or MAN and interference correction (if a matrix matched blank is not available).

Nothing wrong with "visual aids" as I use them myself often. I was just trying to understand why you posted this in the PFE board...

As for communication issues between PFE and JEOL, we have dozens of labs running PFE on JEOL 8230 and 8530 instruments with no issues whatsoever on Windows 7 and Windows 10. I would be happy to work with you off-line to get this working on your instrument/computer systems.

[Brian Joy]

We still have the Windows XP version of the JEOL software.  Owen has been trying to help me troubleshoot, but I hardly have any time to do anything but analyze stuff these days.

Yes, I realize that MAN doesn't require any off-peak measurements.  It was my experience trying to model background radiation at the As Lα peak position that turned me off to the idea of accurate modeling of background radiation.

EDIT:  I did the above-mentioned work in 2014 while collecting k-ratio data on GaAs and InAs (relative to elemental As).  I used an approach based on Kramer's Law as modified by Small et al. (1987), Smith (1975), and Smith and Reed (1981), which I've attached.  I simply couldn't get a calculated background that matched wavelength scans to my satisfaction.  In using the approach of Smith and Reed, I normalized to the Si spectrum on TAP, as it contains no characteristic peaks near As Lα and Lβ₁.  Maybe I need to revisit the problem.

[John Donovan]

EDIT:  I did the above-mentioned work in 2014 while collecting k-ratio data on GaAs and InAs (relative to elemental As).  I used an approach based on Kramer's Law as modified by Small et al. (1987), Smith (1975), and Smith and Reed (1981), which I've attached.  I simply couldn't get a calculated background that matched wavelength scans to my satisfaction.  In using the approach of Smith and Reed, I normalized to the Si spectrum on TAP, as it contains no characteristic peaks near As Lα and Lβ₁.  Maybe I need to revisit the problem.

I completely agree that modeling the continuum over a range of photon energies is a "tough nut to crack".  Especially across absorption edges. Best of luck in your efforts on this.

The cool thing about the MAN background correction method is that we only need to apply the continuum absorption correction at a *single* photon energy. In fact, the continuum photon energy one is working with, is *exactly* the same photon energy as the characteristic photon energy one is trying to correct for background!  How lucky is that!

That means we might be able to utilize an absorption correction intended for characteristic photon energies to correct for continuum absorption (because they are measured at that same energy, and a photon is a photon!). Of course continuum photon emissions are spatially slightly non-isotropic compared to characteristic photon emissions, but I thought it's worth a try. And what we have found is that the characteristic photon absorption corrections give better results for the MAN curve fits, than the continuum (specific) absorption corrections of Small, Myklebust and Heinrich.

One can experiment with these different continuum absorption corrections options as shown in the MAN fit dialog in Probe for EPMA:

https://probesoftware.com/smf/index.php?topic=1221.msg8459#msg8459

In fact, PFE utilizes the same characteristic absorption correction as the currently selected matrix correction (ZAF, phi-rho-z, etc.), so one has many absorption models to play with.

But, interestingly, the largest effect we found is in utilizing a Z-fraction (rather than mass fraction) based calculation for average atomic number (Kramer's Law):

https://probesoftware.com/smf/index.php?topic=1221.msg8475#msg8475

https://probesoftware.com/smf/index.php?topic=4.msg10036#msg10036

Because of course neutrons (atomic mass) don't factor into continuum production as it's all electro-dynamics.   

[Brian Joy]

EDIT:  I did the above-mentioned work in 2014 while collecting k-ratio data on GaAs and InAs (relative to elemental As).  I used an approach based on Kramer's Law as modified by Small et al. (1987), Smith (1975), and Smith and Reed (1981), which I've attached.  I simply couldn't get a calculated background that matched wavelength scans to my satisfaction.  In using the approach of Smith and Reed, I normalized to the Si spectrum on TAP, as it contains no characteristic peaks near As Lα and Lβ₁.  Maybe I need to revisit the problem.

I completely agree that modeling the continuum over a range of photon energies is a "tough nut to crack".  Especially across absorption edges. Best of luck in your efforts on this.

The cool thing about the MAN background correction method is that we only need to apply the continuum absorption correction at a *single* photon energy. In fact, the continuum photon energy one is working with, is *exactly* the same photon energy as the characteristic photon energy one is trying to correct for background!  How lucky is that!

That means we might be able to utilize an absorption correction intended for characteristic photon energies to correct for continuum absorption (because they are measured at that same energy, and a photon is a photon!). Of course continuum photon emissions are spatially slightly non-isotropic compared to characteristic photon emissions, but I thought it's worth a try. And what we have found is that the characteristic photon absorption corrections give better results for the MAN curve fits, than the continuum (specific) absorption corrections of Small, Myklebust and Heinrich.

One can experiment with these different continuum absorption corrections options as shown in the MAN fit dialog in Probe for EPMA:

https://probesoftware.com/smf/index.php?topic=1221.msg8459#msg8459

In fact, PFE utilizes the same characteristic absorption correction as the currently selected matrix correction (ZAF, phi-rho-z, etc.), so one has many absorption models to play with.

But, interestingly, the largest effect we found is in utilizing a Z-fraction (rather than mass fraction) based calculation for average atomic number (Kramer's Law):

https://probesoftware.com/smf/index.php?topic=1221.msg8475#msg8475

https://probesoftware.com/smf/index.php?topic=4.msg10036#msg10036

Because of course neutrons (atomic mass) don't factor into continuum production as it's all electro-dynamics.   

OK, I need to play around with MAN once PFE and JEOL are cooperating better, and I need to read through some of your papers.  The modeling that I did should have worked at a given arbitrary energy, but yes, I was indeed more concerned (and frustrated) with the "bigger picture."  Also, I was using simplified atomic number and absorption corrections and found that the chosen reference spectrum (I used at least Si, Fe, and Nb, maybe others) made a significant difference in the results.  I couldn't characterize the As L1, L2, and L3 absorption edges even remotely accurately.

Edit:  Or maybe it's better to say that I was trying to evaluate how good the result was at the As La peak position by evaluating how well I was able to model adjacent parts of the spectrum.  It's really hard to judge where the background "should" be at As La due to the fact that it and As Lb1 are incompletely resolved and also due to the presence of absorption edges on the short-wavelength side of As Lb1.

I can see how nicely the MAN background would pair with the blank correction for the case of Ti interference at V Ka (since I do have single-crystal, high purity TiO2), especially since background counting error would be eliminated.

[JonF]

Thought I might chip in with some user experience of a similar situation (page 7 here: https://probesoftware.com/smf/index.php?topic=4.90).

I was measuring Co in Fe-Ni meteorites, for which there is a very well known Ni:Co solar ratio.

I acquired my data with off peak backgrounds (green lines below), but had "messed up" the background measurement by putting the low side off peak measurement on the Fe Kb and trying to subtract the background using an exponential fit:



But there was method in my madness as, like you, I was trying to avoid a large absorption edge (Fe K [yellow line] in my case):



Turns out, there is no good position to put the low side background for Co Ka in a major Fe phase.

The problem came to light several months after the user's session, when they were sifting through their data and plotting everything up.

Because I'd acquired everything with off peaks, I could construct a MAN curve and use those backgrounds instead (rather than repeat the session). After a lightning speed bug fix (thanks again, John!), I could assign an Fe on Co interference with the MAN backgrounds and the results went from plotting ~250ppm below the solar Ni:Co line to sitting exactly on it (with better accuracy and precision than some of the other published measurements they had plotted, too!).   

I should mention that this was using the interference correction rather than a blank correction, as I hadn't measured a pure Fe metal as an unknown, and I don't have any Fe-Ni reference materials (anyone have any recommendations for any?). The interference correction could be applied having measured Co in Fe metal for the standards (i.e. I didn't use quick standards). 

[Brian Joy]

Thought I might chip in with some user experience of a similar situation (page 7 here: https://probesoftware.com/smf/index.php?topic=4.90).

I was measuring Co in Fe-Ni meteorites, for which there is a very well known Ni:Co solar ratio.

I acquired my data with off peak backgrounds (green lines below), but had "messed up" the background measurement by putting the low side off peak measurement on the Fe Kb and trying to subtract the background using an exponential fit:

Hi Jon,

I have to deal with that interference quite a lot, as I frequently analyze for small amounts of Co in pyrrhotite.  Thankfully, on the LiFL crystal here, the "background"/tail on the high-Bragg-angle-side of Fe Kβ is almost perfectly linear, and so I'm able to place background offsets at -0.6 mm and +1.0 mm and interpolate linearly.

Brian

[Probeman]

I’ve attached a couple of plots that show graphically how useful the blank correction can be.  I have a rutile standard (NMNH 120812) that I know contains vanadium, but I’ve had a great deal of difficulty verifying the amount precisely due to interference from Ti Kβ_1,3, Ti Kβ₅, associated satellites, and also the Ti K absorption edge.  Of course, this is a worst-case scenario in which the Ti concentration is very high and the V concentration is very low.  Recently, I acquired a high-purity synthetic TiO₂ standard, and I’ve overlaid its spectrum (using LiFL) on that of the natural rutile in the first plot below.  In the second plot, I’ve essentially done a manual blank correction by subtracting the high purity TiO₂ spectrum from that of the natural rutile (difference illustrated in green).  There is absolutely no way to perform an overlap correction with confidence in this case, and the blank correction provides the only useful result (variable between ~0.2 and ~0.3 wt% V₂O₃ in this case), especially since natural rutile is generally close to end-member composition.  In this particular case, the rutile contains ~98.9 wt% TiO₂.  I hope that someone will find this useful for both the general case and for this specific case.  I often find that people minimize the importance of interference from Ti at the V Kα peak position when using LiF.

Hi Brian,
Looking through my old data I realized I've never really looked closely at this specific interference.  So yesterday I acquired a test run analyzing a bunch of Ti standards for vanadium and today I took a look at the data. It's more interesting than I thought it might be... there is something going on that I don't understand. Interestingly I get very similar results using both off-peak and MAN background corrections, so I don't think the issue is extrapolation across an absorption edge.

Conditions were 15 keV, 30 nA and 60 seconds on-peak and 60 seconds off-peak. The background positions for V Ka on LiF are here:



And for V ka on PET here:



Where the red line is the on-peak position for V Ka, the purple lines are the old off-peak positions and the green lines are the off-peak position utilized for the acquisitions.

I acquired both Ti Ka and V Ka on both LiF and PET crystals just for fun and acquired 5 points on each standard. That way we can look at LiF data only, PET data only, or aggregate the intensities from both the LiF and PET crystals. For example, looking at the off-peak data on LIF crystals first I get this for TiO2 when no interference correction is applied:

ELEM:       Ti    Ti-D       V     V-D      Sr      Fe      Cr      Mn       O   SUM 
XRAY:     (ka)    (ka)    (ka)    (ka)      ()      ()      ()      ()      ()
AVER:   59.956     ---    .551     ---    .000    .000    .000    .000  40.000 100.506
SDEV:     .073     ---    .010     ---    .000    .000    .000    .000    .000    .083
SERR:     .033     ---    .005     ---    .000    .000    .000    .000    .000
%RSD:      .12     ---    1.84     ---     .00     .00     .00     .00     .00

As you can see I get about 0.5% wt% apparent V in TiO2. Turning on the interference correction and using TiO2 as the interference standard, we of course get a perfect correction because it's analyzing itself:

ELEM:       Ti    Ti-D       V     V-D      Sr      Fe      Cr      Mn       O   SUM 
XRAY:     (ka)    (ka)    (ka)    (ka)      ()      ()      ()      ()      ()
AVER:   59.988     ---    .000     ---    .000    .000    .000    .000  40.000  99.988
SDEV:     .073     ---    .009     ---    .000    .000    .000    .000    .000    .082
SERR:     .033     ---    .004     ---    .000    .000    .000    .000    .000
%RSD:      .12     --- 9536.47     ---     .00     .00     .00     .00     .00

Now if we look at another Ti standard, say SrTiO3 we obtain this with the interference correction on:

ELEM:       Ti    Ti-D       V     V-D      Sr      Fe      Cr      Mn       O   SUM 
XRAY:     (ka)    (ka)    (ka)    (ka)      ()      ()      ()      ()      ()
AVER:   26.329     ---    .010     ---  47.740    .000    .000    .000  26.150 100.228
SDEV:     .223     ---    .005     ---    .000    .000    .000    .000    .000    .222
SERR:     .100     ---    .002     ---    .000    .000    .000    .000    .000
%RSD:      .85     ---   50.80     ---     .00     .00     .00     .00     .00

Not terrible, but still 2 standard deviations from zero. And next we can do an extrapolation to a material containing a larger concentration of Ti such as TiC as seen here:

ELEM:       Ti      Ti       V       V      Sr      Fe      Cr      Mn       C
TYPE:     ANAL    ANAL    ANAL    ANAL    SPEC    SPEC    SPEC    SPEC    SPEC
BGDS:      LIN     EXP     LIN     EXP
TIME:    60.00     ---   60.00     ---     ---     ---     ---     ---     ---
BEAM:    30.03     ---   30.03     ---     ---     ---     ---     ---     ---

ELEM:       Ti    Ti-D       V     V-D      Sr      Fe      Cr      Mn       C   SUM 
XRAY:     (ka)    (ka)    (ka)    (ka)      ()      ()      ()      ()      ()
AVER:   78.622     ---    .012     ---    .000    .000    .000    .000  20.000  98.634
SDEV:     .259     ---    .011     ---    .000    .000    .000    .000    .000    .250
SERR:     .116     ---    .005     ---    .000    .000    .000    .000    .000
%RSD:      .33     ---   92.30     ---     .00     .00     .00     .00     .00

Actually not too bad, just over a standard deviation from zero.  Next we'll look at the results from the PET crystals and that's where the wheels really come off...

[Brian Joy]

I’ve attached a couple of plots that show graphically how useful the blank correction can be.  I have a rutile standard (NMNH 120812) that I know contains vanadium, but I’ve had a great deal of difficulty verifying the amount precisely due to interference from Ti Kβ_1,3, Ti Kβ₅, associated satellites, and also the Ti K absorption edge.  Of course, this is a worst-case scenario in which the Ti concentration is very high and the V concentration is very low.  Recently, I acquired a high-purity synthetic TiO₂ standard, and I’ve overlaid its spectrum (using LiFL) on that of the natural rutile in the first plot below.  In the second plot, I’ve essentially done a manual blank correction by subtracting the high purity TiO₂ spectrum from that of the natural rutile (difference illustrated in green).  There is absolutely no way to perform an overlap correction with confidence in this case, and the blank correction provides the only useful result (variable between ~0.2 and ~0.3 wt% V₂O₃ in this case), especially since natural rutile is generally close to end-member composition.  In this particular case, the rutile contains ~98.9 wt% TiO₂.  I hope that someone will find this useful for both the general case and for this specific case.  I often find that people minimize the importance of interference from Ti at the V Kα peak position when using LiF.

Hi Brian,
Looking through my old data I realized I've never really looked closely at this specific interference.  So yesterday I acquired a test run analyzing a bunch of Ti standards for vanadium and today I took a look at the data. It's more interesting than I thought it might be... there is something going on that I don't understand. Interestingly I get very similar results using both off-peak and MAN background corrections, so I don't think the issue is extrapolation across an absorption edge.

Conditions were 15 keV, 30 nA and 60 seconds on-peak and 60 seconds off-peak. The background positions for V Ka on LiF are here:

The Ti K absorption edge produces a discontinuity in the continuum, especially when the compound is ~99 wt% TiO₂.  It must create problems for off-peak background measurement.

[Probeman]

OK, so here the wheels really come off. I have to say I don't understand what the problem is.  I'm doing something wrong but can't see what it might be...

Here is TiO2 analyzed for V *without* the interference correction on an LPET crystal:

ELEM:     Ti-D      Ti     V-D       V      Sr      Fe      Cr      Mn       O   SUM 
XRAY:     (ka)    (ka)    (ka)    (ka)      ()      ()      ()      ()      ()
AVER:      ---  59.804     ---   3.271    .000    .000    .000    .000  40.000 103.075
SDEV:      ---    .060     ---    .015    .000    .000    .000    .000    .000    .066
SERR:      ---    .027     ---    .007    .000    .000    .000    .000    .000
%RSD:      ---     .10     ---     .46     .00     .00     .00     .00     .00

So that is a big spectral interference for sure.  Now we turn on the interference correction again and get this result which is of course zero because it's analyzing itself (showing all data from the 5 points):

ELEM:     Ti-D      Ti     V-D       V      Sr      Fe      Cr      Mn       O   SUM 
XRAY:     (ka)    (ka)    (ka)    (ka)      ()      ()      ()      ()      ()
   204     ---  60.027     ---   -.016    .000    .000    .000    .000  40.000 100.010
   205     ---  59.999     ---    .012    .000    .000    .000    .000  40.000 100.011
   206     ---  59.886     ---   -.009    .000    .000    .000    .000  40.000  99.878
   207     ---  60.000     ---    .017    .000    .000    .000    .000  40.000 100.018
   208     ---  60.035     ---   -.003    .000    .000    .000    .000  40.000 100.032

AVER:      ---  59.989     ---    .000    .000    .000    .000    .000  40.000  99.990
SDEV:      ---    .060     ---    .014    .000    .000    .000    .000    .000    .063
SERR:      ---    .027     ---    .006    .000    .000    .000    .000    .000
%RSD:      ---     .10     --- 3781.96     .00     .00     .00     .00     .00

OK, now let's look at the SrTiO3 again on the LPET crystals:

ELEM:     Ti-D      Ti     V-D       V      Sr      Fe      Cr      Mn       O   SUM 
XRAY:     (ka)    (ka)    (ka)    (ka)      ()      ()      ()      ()      ()
   179     ---  26.199     ---   -.035  47.740    .000    .000    .000  26.150 100.054
   180     ---  26.268     ---   -.049  47.740    .000    .000    .000  26.150 100.109
   181     ---  26.261     ---   -.058  47.740    .000    .000    .000  26.150 100.093
   182     ---  26.190     ---   -.037  47.740    .000    .000    .000  26.150 100.043
   183     ---  26.219     ---   -.043  47.740    .000    .000    .000  26.150 100.066

AVER:      ---  26.227     ---   -.045  47.740    .000    .000    .000  26.150 100.073
SDEV:      ---    .035     ---    .009    .000    .000    .000    .000    .000    .027
SERR:      ---    .016     ---    .004    .000    .000    .000    .000    .000
%RSD:      ---     .14     ---  -21.07     .00     .00     .00     .00     .00

Well that's pretty bad I have to say. But even worse is the TiC as shown here:

ELEM:       Ti      Ti       V       V      Sr      Fe      Cr      Mn       C
TYPE:     ANAL    ANAL    ANAL    ANAL    SPEC    SPEC    SPEC    SPEC    SPEC
BGDS:      LIN     EXP     LIN     EXP
TIME:      ---   60.00     ---   60.00     ---     ---     ---     ---     ---
BEAM:      ---   30.03     ---   30.03     ---     ---     ---     ---     ---

ELEM:     Ti-D      Ti     V-D       V      Sr      Fe      Cr      Mn       C   SUM 
XRAY:     (ka)    (ka)    (ka)    (ka)      ()      ()      ()      ()      ()
   174     ---  80.170     ---   -.354    .000    .000    .000    .000  20.000  99.816
   175     ---  80.007     ---   -.351    .000    .000    .000    .000  20.000  99.656
   176     ---  80.083     ---   -.366    .000    .000    .000    .000  20.000  99.717
   177     ---  80.442     ---   -.372    .000    .000    .000    .000  20.000 100.070
   178     ---  80.261     ---   -.356    .000    .000    .000    .000  20.000  99.905

AVER:      ---  80.193     ---   -.360    .000    .000    .000    .000  20.000  99.833
SDEV:      ---    .169     ---    .009    .000    .000    .000    .000    .000    .163
SERR:      ---    .075     ---    .004    .000    .000    .000    .000    .000
%RSD:      ---     .21     ---   -2.41     .00     .00     .00     .00     .00

Clearly a blank correction is required to fix this, but why is it so bloody horrible?  I even tried the original Gilfrich interference correction method and it's only different by 30 PPM (0.003) which is less than the variance so it's not a matrix effect.

Help!

[Probeman]

The Ti K absorption edge produces a discontinuity in the continuum, especially when the compound is ~99 wt% TiO₂.  It must create problems for off-peak background measurement.

I know, that's what I thought too.  But I get almost exactly the same results (only 10 PPM different!) using the MAN method as shown here looking at V Ka in TiC using LPET again:

St  674 Set   2 TiC (titanium carbide), Results in Elemental Weight Percents
 
ELEM:       Ti      Ti       V       V      Sr      Fe      Cr      Mn       C
TYPE:     ANAL    ANAL    ANAL    ANAL    SPEC    SPEC    SPEC    SPEC    SPEC
BGDS:      MAN     MAN     MAN     MAN
TIME:      ---   60.00     ---   60.00     ---     ---     ---     ---     ---
BEAM:      ---   30.03     ---   30.03     ---     ---     ---     ---     ---

ELEM:     Ti-D      Ti     V-D       V      Sr      Fe      Cr      Mn       C   SUM 
XRAY:     (ka)    (ka)    (ka)    (ka)      ()      ()      ()      ()      ()
   174     ---  80.171     ---   -.350    .000    .000    .000    .000  20.000  99.820
   175     ---  80.024     ---   -.352    .000    .000    .000    .000  20.000  99.672
   176     ---  80.081     ---   -.371    .000    .000    .000    .000  20.000  99.710
   177     ---  80.443     ---   -.372    .000    .000    .000    .000  20.000 100.071
   178     ---  80.268     ---   -.352    .000    .000    .000    .000  20.000  99.916

AVER:      ---  80.197     ---   -.359    .000    .000    .000    .000  20.000  99.838
SDEV:      ---    .165     ---    .011    .000    .000    .000    .000    .000    .161
SERR:      ---    .074     ---    .005    .000    .000    .000    .000    .000
%RSD:      ---     .21     ---   -3.05     .00     .00     .00     .00     .00

I'm really perplexed I have to say...  all I can say for now is don't use the LPET crystals for trace V in Ti compounds without a suitable blank correction!

[Probeman]

Sort of not surprising, but if I aggregate the data from the LiF and PET crystals we get this somewhat better results without a blank correction (but with an interference correction):

ELEM:       Ti      Ti       V       V      Sr      Fe      Cr      Mn       C   SUM 
XRAY:     (ka)    (ka)    (ka)    (ka)      ()      ()      ()      ()      ()
   174  79.950    .000   -.266    .000    .000    .000    .000    .000  20.000  99.684
   175  79.868    .000   -.269    .000    .000    .000    .000    .000  20.000  99.600
   176  79.903    .000   -.276    .000    .000    .000    .000    .000  20.000  99.627
   177  80.240    .000   -.281    .000    .000    .000    .000    .000  20.000  99.959
   178  80.106    .000   -.274    .000    .000    .000    .000    .000  20.000  99.832

AVER:   80.013    .000   -.273    .000    .000    .000    .000    .000  20.000  99.740
SDEV:     .156    .000    .006    .000    .000    .000    .000    .000    .000    .152
SERR:     .070    .000    .003    .000    .000    .000    .000    .000    .000
%RSD:      .19   .0000   -2.16   .0000     .00     .00     .00     .00     .00

Then using the TiC standard acquired as an unknown (matrix matched ) blank but without the interference correction we get this:

ELEM:       Ti      Ti       V       V      Sr      Fe      Cr      Mn       C   SUM 
XRAY:     (ka)    (ka)    (ka)    (ka)      ()      ()      ()      ()      ()
   174  79.904    .000    .618    .000    .000    .000    .000    .000  20.000 100.522
   175  79.823    .000    .612    .000    .000    .000    .000    .000  20.000 100.435
   176  79.857    .000    .607    .000    .000    .000    .000    .000  20.000 100.464
   177  80.193    .000    .618    .000    .000    .000    .000    .000  20.000 100.811
   178  80.059    .000    .619    .000    .000    .000    .000    .000  20.000 100.678

AVER:   79.967    .000    .615    .000    .000    .000    .000    .000  20.000 100.582
SDEV:     .155    .000    .005    .000    .000    .000    .000    .000    .000    .159
SERR:     .069    .000    .002    .000    .000    .000    .000    .000    .000
%RSD:      .19   .0000     .83   .0000     .00     .00     .00     .00     .00

So even worse.

But using the TiC standard acquired as an unknown (matrix matched) blank and also with the interference correction we get an almost reasonable result:

ELEM:       Ti      Ti       V       V      Sr      Fe      Cr      Mn       C   SUM 
XRAY:     (ka)    (ka)    (ka)    (ka)      ()      ()      ()      ()      ()
   174  79.935    .000    .023    .000    .000    .000    .000    .000  20.000  99.958
   175  79.853    .000    .020    .000    .000    .000    .000    .000  20.000  99.873
   176  79.888    .000    .013    .000    .000    .000    .000    .000  20.000  99.901
   177  80.224    .000    .008    .000    .000    .000    .000    .000  20.000 100.233
   178  80.091    .000    .015    .000    .000    .000    .000    .000  20.000 100.105

AVER:   79.998    .000    .016    .000    .000    .000    .000    .000  20.000 100.014
SDEV:     .156    .000    .006    .000    .000    .000    .000    .000    .000    .152
SERR:     .070    .000    .003    .000    .000    .000    .000    .000    .000
%RSD:      .19   .0000   37.27   .0000     .00     .00     .00     .00     .00

I have to say I don't really understand these results...

[Brian Joy]

I have to say I don't really understand these results...

Using LiF, the continuum intensity is very low near Ti Kβ (compared to PET) and varies relatively linearly and with very small slope with respect to Bragg angle, even across the Ti K absorption edge.  Although I haven’t actually checked to verify this, peak:background at V Kα should be greater on LiF than PET.  Further, the Ti K absorption edge will be more difficult to see using PET due to its poorer resolution than LiF.  I never, ever use PET to analyze for vanadium in the presence of easily measurable titanium.