Aggregate Intensity interference correction

[David Steele]

I'm currently setting up the QUT EPMA to get into the monazite chemical dating game (a past-time I've "enjoyed" previously on the UTasmania and Edinburgh Uni SX100s....).

At this stage I'm simply translating my successful CAMECA recipe to the JEOL with the necessary adaptations of Xe detectors and L mm not sin thetas.  I am not yet dabbling with MAN backgrounds but that will come with time.....

I'm measuring Pb (Mb) on two PETs.  OK, there is a UMg ovl on the Pb peak, and there is a ThMz2 ovl on the UMb.  I decided against using the UMa line (due to Th lines on both sides of the U Ma peak), and will stick with the Pb Mb line as I also wish to avoid the Y on Pb Ma interference which is relatively small in monazites but becomes a severe headache in xenotimes.....

My questions:
1) when, in the 'Analyse' (i.e. k ratio/ZAF calcn) steps, are the overlap corrections done?  Put another way, does PfE use "raw" overlap intensities and correct BEFORE the k-ratio step or does it work with concentrations?

2) Do I need to have Th on U and U on Pb interference corrections in place (in std assignments) for BOTH channels that Pb is acquired on.  I suspect yes but would like to have it confirmed that this is the process.

3) do I need to measure Th and U as well as Pb on both channels to have the interference corrections done "correctly", or is having the appropriate interference corrections assigned to the respective channels sufficient?

Having used CAMECA's PeakSight I know that on the SX100s I only had to have ONE Th on U and one U on Pb interference correction in place even though Pb was accumulated on two channels.  The correction was done after the intensities were combined using a weighted k-ratio average (the weighting was done on the basis of the respective calibration factors for the Pb lines measured on the two spectros.

I understand PfE uses a simple arithmetic average to aggregate intensities.  I surmise this could lead to inappropriate averages especially if the sensitivities of the spectrometers are very different for the aggregated line of interest (e.g. Pb measured on an SX100 using a low P det PET and a high P det LPET).  CAMECA's calulcation weights the aggregate intensity towards the more sensitive spectrometer that is accumulating the most counts relative to the other(s).  Any thoughts on refining the aggregate weighting John?

4) how are 'cascade' interference corrections handled, i.e. ensuring that the Th on U correction is implemented before the U on Pb correction in the above?  I will also have to deal with the usual interference nasties associated with the REEs.... but a "nice" analytical challenge

Cheers and happy New Year to everyone,

David
QUT

[John Donovan]

Having used CAMECA's PeakSight I know that on the SX100s I only had to have ONE Th on U and one U on Pb interference correction in place even though Pb was accumulated on two channels.  The correction was done after the intensities were combined using a weighted k-ratio average (the weighting was done on the basis of the respective calibration factors for the Pb lines measured on the two spectros.

Yes, the Cameca interference correction method was determined to be inaccurate by Jercinovic and Allaz a few years ago, but Cameca claims it is fixed in the latest version of PeakSight. 

Our method published here is more accurate because it ratios the concentrations instead of the uncorrected intensities:

http://probesoftware.com/Improved%20Interference%20%28Micro.%20Anal,%201993%29.pdf

Remember your interference standard and unknown specimen have to have a different matrix by definition, therefore one *must* use the matrix corrected concentrations for the interference correction as described in the above linked to paper!

I understand PfE uses a simple arithmetic average to aggregate intensities.  I surmise this could lead to inappropriate averages especially if the sensitivities of the spectrometers are very different for the aggregated line of interest (e.g. Pb measured on an SX100 using a low P det PET and a high P det LPET).  CAMECA's calulcation [sic] weights the aggregate intensity towards the more sensitive spectrometer that is accumulating the most counts relative to the other(s).  Any thoughts on refining the aggregate weighting John?

Think harder!  The "simple arithmetic average" as you call it, is actually rigorously accurate- it's exactly how your spectrometers record the total counts normally!  By adding together the photons from the duplicate element spectrometers, the statistics fall through naturally.  Think of it as simply increasing the geometric efficiency of a single spectrometer, while using the same count time and beam current!

Cameca's aggregate method of weighting the concentrations, like their interference correction method, was also determined by Jercinovic and Allaz to be inaccurate. You next need to read this paper:

http://probesoftware.com/Ti%20in%20Quartz,%20Am.%20Min.%20Donovan,%202011.pdf

4) how are 'cascade' interference corrections handled, i.e. ensuring that the Th on U correction is implemented before the U on Pb correction in the above?  I will also have to deal with the usual interference nasties associated with the REEs.... but a "nice" analytical challenge

Also answered in the 1st paper! Read the two papers above and I'll answer your remaining questions.

[Julien]

At this stage I'm simply translating my successful CAMECA recipe to the JEOL with the necessary adaptations of Xe detectors and L mm not sin thetas.  I am not yet dabbling with MAN backgrounds but that will come with time.....

That is fine, just make sure to carefully scan your sample. With a smaller Rowland circle, the peak width will be slightly bigger, so your background positions may not be adequate anymore. For Pb analysis (and U, Th), I would recommend to use the multipoint background acquisition. Everything is clearly (?) explained in a PDF document that I uploaded in this post - along with other comments.

http://probesoftware.com/smf/index.php?topic=186.0;topicseen

My first post contains some comment, the second one has the link to the aforementioned PDF. Note that using the TDI corrections can help, as using a high current (we used 200 nA) with a focused beam do damage the monazite. A little less if you have an aluminum coating, but still, you can see some changes. I was thinking for a long time about doing a first "major element analysis" at lower current (20 nA) and then performing another analysis at higher current just for the minor and trace elements. You can certainly get only one (or two) major element analysis, and then apply this analysis to the associated 5-8 trace element analyses (just make sure you are in a homogeneous domain - I assume you do have mapping for some elements: Y, Th, Ca, Si and U? if not, please do get element maps first). This is relatively easy on PfE...

I'm measuring Pb (Mb) on two PETs.  OK, there is a UMg ovl on the Pb peak, and there is a ThMz2 ovl on the UMb.  I decided against using the UMa line (due to Th lines on both sides of the U Ma peak), and will stick with the Pb Mb line as I also wish to avoid the Y on Pb Ma interference which is relatively small in monazites but becomes a severe headache in xenotimes.....

Your "fear" of Y interference on Pb Ma is legit for xenotime, however, I would still recommend Pb Ma for monazite: the yttrium content remains usually low (1-2%) and can be easily corrected using Probe for EPMA. John's comment about PeakSight is totally right: PeakSight is [was?] not doing a great job in term of interference correction, we tested this by getting an interference factor for U analysis in ThSiO4 standard, then we analyze for U in brabantite/cheralite, and we constantly got negative values around minus 1% U if I recall correctly. Maybe they fixed this...

If you consider Pb Ma, just be aware of a small but significant shift of Y Lg2,3 on monazite compare to YAG or YPO4. See my post linked above.

1) when, in the 'Analyse' (i.e. k ratio/ZAF calcn) steps, are the overlap corrections done?  Put another way, does PfE use "raw" overlap intensities and correct BEFORE the k-ratio step or does it work with concentrations?

John's paper is pretty clear about this: the interference correction is actually implemented WITHIN the matrix correction, and is therefore more accurate. So, the correction is first made on the k-ratio, and in the matrix correction iterations. This is the most logical approach: as the matrix correction goes on, the concentration of the interfering and the interfered element varies, so you cannot just do a single peak interference correction before (or after) the k-ratio and matrix correction calculations.

2) Do I need to have Th on U and U on Pb interference corrections in place (in std assignments) for BOTH channels that Pb is acquired on.  I suspect yes but would like to have it confirmed that this is the process.

3) do I need to measure Th and U as well as Pb on both channels to have the interference corrections done "correctly", or is having the appropriate interference corrections assigned to the respective channels sufficient?

Having used CAMECA's PeakSight I know that on the SX100s I only had to have ONE Th on U and one U on Pb interference correction in place even though Pb was accumulated on two channels.  The correction was done after the intensities were combined using a weighted k-ratio average (the weighting was done on the basis of the respective calibration factors for the Pb lines measured on the two spectros.

a) An interfering factor for EACH elements on EACH channels MUST be acquired; for instance, for Pb Ma interfered by Th Mg, Y Lg2,3 and La La(2) [minor interference - you can minimize it with appropriate PHA setting], you will need to acquire a peak interference factor on the two channels you are using for Pb Mb analysis: e.g., analyze Th Mg on your Th standard on BOTH channels used for Pb Ma.

b) You will need to define in Probe for EPMA (in std assignments) the list of interfering element(s) on EACH channels.

Again, this is all explained in the PDF linked in the post about Microprobe Age Dating (Monazite).

I believe John has (or will) answer the remaining questions... Enjoy, "good luck", and Happy New Year!

Julien

[David Steele]

Thank you Gents for your replies.

I stand "corrected" on the PeakSight fluff, but would like to read the Jercinovic and Allaz paper on why the CAMECA corrections were, erm, incorrect....

Re my question 4:
John, I've read both papers as requested and can find NO mention of EXACTLY how to implement cascade interference corrections in the PfE software (hence my last question #4 in my original post).  Perhaps question 4 was not worded specifically enough.
The first paper ("An Improved Interference...") goes through the matrix correction and overall calculation logic, but the first four examples cited in that paper involve ONE interfering element on ONE interfered element (P-Zr, Ba-Ti, Fe-F, Fe-Co).  Only the last, fifth, example mentions the cascade Ti on V then V on Cr inteferences, but the paper does not inform me how to implement the cascade interference correction procedure in PfE.  Reading Julien's "Quick Instructions on Mnz and Xno Analyses" pdf does that to some extent....

Section 12 of Julien's Quick Instructions shows his standard assignment for UMb including Th and K interferences.  In MY case, assuming I stick with using the PbMb line, I will also have U interferents on PbMb on two channels (plus Th on UMb on both channels - from Julien's earlier reply). 
My followup question:  does PfE look through the list of elements and respective interferents then decide which to correct in the appropriate, correct cascade sequence, i.e. do I just put in the interferences in the std assignments window and let PfE do the rest, OR is there another box/window somewhere that I have to specify the cascade correction order? My experiences with the SAMx software were that the cascade interference correction sequence could be explicitly specified by the user, e.g. Th on U then U on Pb (although it could not do aggregate intensities at that time, circa 1998-2003). 

Re my earlier question 3:
"do I need to measure Th and U as well as Pb on both channels to have the interference corrections done "correctly", or is having the appropriate interference corrections assigned to the respective channels sufficient?".... has not been answered.
Yes I know I need to pre-acquire the appropriate interference standards on both channels and have them assigned on both channels but my question should possibly have been "do I need to acquire Th and U on both channels during an 'unknown' acquisition together with Pb on both to have the appropriate interference corrections applied?", OR is John implying by refering me to read the "Improved Interference.." paper that once the concentrations of Th and U are estimated in the matrix (possibly acquired using a third or fourth spectrometer, i.e. non-Pb channel acquisition) that it is not necessary to acquire Th and U intensities on the Pb-acquisition channels to have the appropriate cascade interference corrections done?

I am simply trying to work out exactly how to implement all the stuff in PfE to get it to do what I require!  I have tripped over and been tripped over enough times in the past 8 months with making assumptions about the PfE logic to not want to waste further time and effort trying to implement a method an inappropriate way!

(to paraphrase John, "The only silly question is one that isn't asked"....)
   

My chemical dating method implemented via PeakSight on the Edinburgh SX100 multiple beam conditions, e.g. La, Ce, P, Al and Nd at 20kV, 20nA then U, Th, Pb(*2), and the other (high) REEs plus Si and Y at 20kV, 100nA. In total a circa. 8 minute acquisition.
The caveat with PfE and multiple conditions is that the first column conditions used are by default the conditions specified for the first element on channel 1 (as you'd know).  So you have to fiddle with the element sequence to get PfE to acquire using the appropriate conditions sequence.  In my first attempt using PfE and multiple conditions I got caught trying to measure C in steel: C at 7kV, the other elements (Fe, Mn, Cr, Mo etc at 15kV).  Unfortunately my LDE2 xtl is on channel 3, so PfE runs the 15kV elements first then does C.  So much for avoiding C contamination.  I had the anticontamination trap working that day!  For my next iteration I'll have to try C at the top of the LDE1, which is on channel 1, unless JD will change the PfE code.
John, could you change the code to allow the user to specify the beam condition order irrespective of the element/channel combination, rather than the current enforced "default" as indicated above?  If the user can specify the element acquisition order, and we already link the elements to specific conditions in a multiple condition setup, why can't we go a step further and be able to specify the multiple condition order (irrespective of the first element on channel 1...)?     

PS: I like Julien's "Quick Instructions for Monazite and Xenotime Analyses using PfE".... all 50 pages of it! Very thorough, but very necessary....

Possibly a question more appropriate in the monazite chemical dating topic, but....  Julien are you measuring UMb on a highP detector to avoid the detector Ar K abs edge (my numerous CAMECA scans over that edge show that the edge is very obvious on the low P detectors but is alomst non-existent on the high P detectors on the CAMECA EPMAs), or are you positioning your multi-point bgds for UMb above that edge, and also avoiding sample major element abs edges....??

Thanks again and cheers,
David

[John Donovan]

I stand "corrected" on the PeakSight fluff, but would like to read the Jercinovic and Allaz paper on why the CAMECA corrections were, erm, incorrect....

Hi David,
The Cameca interference corrections are not incorrect per se, they are just inaccurate in cases where the trace element overlap is large and matrix effects are significant. But as to precisely "why" you'll probably have to ask Julien and Mike Jercinovic as they dealt with Cameca on that issue.  My guess is that they weren't doing the proper matrix correction to deal with the difference in physics between the unknown and the standard for the interference correction as described in my paper.

John, I've read both papers as requested and can find NO mention of EXACTLY how to implement cascade interference corrections in the PfE software (hence my last question #4 in my original post).  Perhaps question 4 was not worded specifically enough.

The first paper ("An Improved Interference...") goes through the matrix correction and overall calculation logic, but the first four examples cited in that paper involve ONE interfering element on ONE interfered element (P-Zr, Ba-Ti, Fe-F, Fe-Co).  Only the last, fifth, example mentions the cascade Ti on V then V on Cr inteferences, but the paper does not inform me how to implement the cascade interference correction procedure in PfE.  Reading Julien's "Quick Instructions on Mnz and Xno Analyses" pdf does that to some extent....

The Ti, V, Cr interference in the paper is a nice example of a cascade interference and shows why the matrix correction is necessary is these cases. As to "EXACTLY" how the correction is made, the flow chart in the paper also shows that the interference correction calculation is performed in an *additional* outer iteration loop so the order of the elements doesn't matter at all!  Elegant, no?   

If you need more detail than that I'll have to send you the source code.

Section 12 of Julien's Quick Instructions shows his standard assignment for UMb including Th and K interferences.  In MY case, assuming I stick with using the PbMb line, I will also have U interferents on PbMb on two channels (plus Th on UMb on both channels - from Julien's earlier reply). 

My followup question:  does PfE look through the list of elements and respective interferents then decide which to correct in the appropriate, correct cascade sequence, i.e. do I just put in the interferences in the std assignments window and let PfE do the rest, OR...

As I said above, the element order doesn't matter because the concentrations in the interference correction are iterated additionally.

Re my earlier question 3:
"do I need to measure Th and U as well as Pb on both channels to have the interference corrections done "correctly", or is having the appropriate interference corrections assigned to the respective channels sufficient?".... has not been answered.

Yes I know I need to pre-acquire the appropriate interference standards on both channels and have them assigned on both channels but my question should possibly have been "do I need to acquire Th and U on both channels during an 'unknown' acquisition together with Pb on both to have the appropriate interference corrections applied?", OR is John implying by refering me to read the "Improved Interference.." paper that once the concentrations of Th and U are estimated in the matrix (possibly acquired using a third or fourth spectrometer, i.e. non-Pb channel acquisition) that it is not necessary to acquire Th and U intensities on the Pb-acquisition channels to have the appropriate cascade interference corrections done?

Just run your standards and unknowns with the same element setup and you will be ok. I suspect you are over thinking this issue.

My chemical dating method implemented via PeakSight on the Edinburgh SX100 multiple beam conditions, e.g. La, Ce, P, Al and Nd at 20kV, 20nA then U, Th, Pb(*2), and the other (high) REEs plus Si and Y at 20kV, 100nA. In total a circa. 8 minute acquisition.

The caveat with PfE and multiple conditions is that the first column conditions used are by default the conditions specified for the first element on channel 1 (as you'd know).  So you have to fiddle with the element sequence to get PfE to acquire using the appropriate conditions sequence.  In my first attempt using PfE and multiple conditions I got caught trying to measure C in steel: C at 7kV, the other elements (Fe, Mn, Cr, Mo etc at 15kV).  Unfortunately my LDE2 xtl is on channel 3, so PfE runs the 15kV elements first then does C.  So much for avoiding C contamination.  I had the anticontamination trap working that day!  For my next iteration I'll have to try C at the top of the LDE1, which is on channel 1, unless JD will change the PfE code.

John, could you change the code to allow the user to specify the beam condition order irrespective of the element/channel combination, rather than the current enforced "default" as indicated above?  If the user can specify the element acquisition order, and we already link the elements to specific conditions in a multiple condition setup, why can't we go a step further and be able to specify the multiple condition order (irrespective of the first element on channel 1...)?     

You can specify the spectrometer acquisition order of the elements from the Acquisition Options dialog in Acquire! Just click on the element row in the list to change its spectrometer order. By using a combination of channel order (in the Combined Conditions dialog) and acquisition order (in the Acquisition Options dialog) you should be able to do what you want.

I'll leave your other questions to Julien as he is the expert on monazite.

[David Steele]

Many thanks for the reply and comments John.

By "EXACTLY", I am asking for the specific steps in which windows in PfE to set up the cascade interference corrections.  I know the interference 'stds' are assigned in "standard assignments" - I routinely do the Fe L on F Ka without issue.

My query is really about PfE knowing how to implement the correct order of cascade correction, but you are saying the element order doesn't matter as the PfE iterative correction procedure that includes iterated interferences takes care of it.  I'll have to simply accept that as I don't have time to pursue this further at present.

I AM probably overthinking the issue BUT it is my role here to 'overthink' these issues, because a) I have to understand what PfE is doing and how, AND be able to pass that information on to EPMA facility users (if they ask....), and b) I am not a person who treats an instrument and associated software that I use as a simple black box.

I specify the spectrometer acquisition order quite frequently so am familiar with where and how to do this.  It's the inflexibility of PfE and control of multiple (beam) condition order that is behind my later comments and request.  The first element on channel 1 'sets' what happens beam condition-wise subsequently.... viz my C in steel example. Ideally I'd like to measure C at 7kV first on channel 3 (LDE2/PC2) and minimise potential C contamination from the beam on the steel sample surface.  But PfE currently forces me analytically to EITHER use a less sensitive crystal on channel 1 (LDE1) but get the C acquired first OR use the better crystal (LDE2) on channel 3 but acquire C later in the analysis and run the risk of measuring surface C contamination.  Neither are optimal approaches.....

Cheers,
David
QUT

[John Donovan]

Many thanks for the reply and comments John.

By "EXACTLY", I am asking for the specific steps in which windows in PfE to set up the cascade interference corrections.  I know the interference 'stds' are assigned in "standard assignments" - I routinely do the Fe L on F Ka without issue.

My query is really about PfE knowing how to implement the correct order of cascade correction, but you are saying the element order doesn't matter as the PfE iterative correction procedure that includes iterated interferences takes care of it.  I'll have to simply accept that as I don't have time to pursue this further at present.

I AM probably overthinking the issue BUT it is my role here to 'overthink' these issues, because a) I have to understand what PfE is doing and how, AND be able to pass that information on to EPMA facility users (if they ask....), and b) I am not a person who treats an instrument and associated software that I use as a simple black box.

Hi David,
I'm not asking you to accept it as a black box (that would be against everything I stand for)! I'm simply saying that the use of an additional outer iterative loop in the code, as described in a peer reviewed paper I linked to, means that one does not have to worry about the element order when performing the interference correction in PFE! 

I'm sorry but I just think you are used to the clumsy manner in which Cameca and SAMx implemented their interference corrections so this is an issue for users of their software.  It is simply not an issue for PFE that you need to worry about. 

As explanation I referenced a peer reviewed paper and even offered to send you the source code.  I'm not sure what else I can do.  If these offers are not sufficient I suggest you perform an experiment (as I did in the paper), to test whether the element order is a concern or not in PFE for cascade interferences.  I hope you will share your results with us here.

I specify the spectrometer acquisition order quite frequently so am familiar with where and how to do this.  It's the inflexibility of PfE and control of multiple (beam) condition order that is behind my later comments and request.  The first element on channel 1 'sets' what happens beam condition-wise subsequently.... viz my C in steel example. Ideally I'd like to measure C at 7kV first on channel 3 (LDE2/PC2) and minimise potential C contamination from the beam on the steel sample surface.  But PfE currently forces me analytically to EITHER use a less sensitive crystal on channel 1 (LDE1) but get the C acquired first OR use the better crystal (LDE2) on channel 3 but acquire C later in the analysis and run the risk of measuring surface C contamination.  Neither are optimal approaches.....
QUT

I think you are still missing something because between the channel order and the element order you can pretty much do anything you want in the sample setup in PFE.  Maybe you should contact Karsten and see if he can figure out what the problem is.
john

PS You do realize, that the first element in the sample (channel order), can be an element on the last spectrometer?

[John Donovan]

Hi David,
After thinking about your points and discussing it a bit with Karsten, I understand better the situation you are facing. So I've decided to re-work the acquisition code for combined condition samples.

Originally I had intended to make the acquisition of combined condition samples "transparent" to the user. That is, making it so the user didn't need to specify the condition order and simply have the software "work it out", but that does have some drawbacks in that the software then needs to make some assumptions about what is intended. 

Karsten gave you a work around for now, so you can accomplish your analysis in the meantime, but I will I'm going to make the condition acquisition order a user defined parameter, just like the spectrometer acquisition order. That will provide the ultimate flexibility.

It will be a bit of work to do all this so not sure when it will be ready, but I've already got some code almost working so...

I should eventually even be able to add a drag and drop method for condition/spectro acquisition ordering...  that will be cool!

Thanks for your feedback and ideas. This is how the software gets better and better all the time!
john

[David Steele]

Many thanks John, and also thanks to Karsten for his input and suggestions.

I look forward to trying out the improved flexibility, when available 

Cheers,
David

[John Donovan]

I look forward to trying out the improved flexibility, when available 

Hi David,
It's ready to test. See here:

http://probesoftware.com/smf/index.php?topic=42.msg2248#msg2248

john

Edit by John:  I should also mention that if any other labs, who are not official beta test labs, but would also like to beta test this new version, just let me know and I will make it available via Dropbox or something.

[Probeman]

I look forward to trying out the improved flexibility, when available 

Hi David,
It's ready to test. See here:

http://probesoftware.com/smf/index.php?topic=42.msg2248#msg2248

john

I should also mention that absorbed current measurements are now supported for combined condition samples.