Re: Microprobe Age Dating (Monazite)

[Julien]

In addition to the mapping of monazite, you might be interested in getting more accurate QUANTITATIVE analyses to obtain U-Th-Pb ages. There are several points to be aware of for such analyses - always keeping in mind that we are here dealing with TRACE element analysis. Among numerous potential issues, there is the need for...

- Careful background characterization. Whereas a two-point background acquisition works for a major (or minor) element analysis, the need for more accurate background is an absolute necessity. This can be achieved for instance by acquiring a WDS scan and fitting a background curve under the peak of interest. This method is tedious and slightly subjective. A more objective method is to acquire a MULTIPOINT BACKGROUND and fit the background to a set of at least 3 (or preferably 5-6 points). Thanks to Probe for EPMA and John D here, since this is the only software offering this method.

- Pay attention to the peak interferences. There are notably interferences on Pb Ma from Y Lg2,3 (with a shift of -50*10^5 sin-theta on monazite compare to xenotime / YPO4 standard), from Th Mz, and La La (II). U Mb is also interfered by Th Mg.

- Using U Ma is problematic when considering an Ar-counter, since the Ar absorption edge is on the tail of U Ma. There are also other absorption edges from Th around U Mb which can be problematic especially with Th-rich samples.

- Pb Mb could be used and has less interferences (pay attention to Ce La (II), though). However, doing so one lose ~40% of counts compare to Pb Ma. However, Pb Mb measurement is recommended for xenotime, as in this case the interference of Y on Pb Ma is HUGE. However, Pb Ma is still recommended for monazite, even with significant amount of Y.

- There can be issues regarding beam damage / diffusion of P and "abnormal" increase of other cations, especially U, Th and Pb. However, thanks again to Probe for EPMA, you can consider using the TDI (Time Dependent Intensity correction) to correct for such effect (however, you might still have problem for correcting all elements, as the TDI can only be applied to the first analyzed elements). Let me quote here a recent conversation with Michael J. Jercinovic from UMass Amherst:

The magnitude depends on the composition, so very little to substantial.  About 1my in a recent 508Ma Moacyr monazite run , but for and older one I ran yesterday (Grand Canyon), 6my TDI vs. no TDI.  One thing here is that it would be nice to have the button (in std assignments) that now says "remove TDI correction" to default to "add TDI correction", so that it would be done for all TDI-specified elements, sor even better, to invoke the TDI correction by default if it has been checked to use it in Special Options (then you can default the button in std assignments to "remove TDI", but if you do click it, then the button should change to "use TDI".  Right now we still have to click on each element and specify the use of TDI even though we have said to acquire using TDI in Special Options in Acquire.  Every time we run an analysis, we have to do this, even though the stored setup had TDI invoked.  If we do point by point acquisition and do the TDI specification on the first point of the file, it will turn it off again on the next point, so if we want to look at the numbers as we go, we have to redo the TDI on this file (in std. assignments) after each point.  As the TDI effect in monazite probably reflects phosphorous loss, it is seen over lengthy acquisitions for U, Th, and Pb (fairly monotonic increase).

And from my own experience:

Well, monazite at 200 nA can see some beam damage. It seems to be more damaged with high Th content, and I think MJJ got also serious beam damage (diffusion) issue with As-rich monazite. Basically, there is a progressive loss of P over time, and a "gain" of heavier element. This is one reason in a near future, I would try to develop a two-way analysis setup, with major element analyzed at low current (with TDI for critical element, such as P and some light REE), and a second run at high current with only U, Th, Pb (and maybe Y and La which are used for overlap correction and making sure we analyzed the same domain during the second run).

U, Th and Pb all show a slight increase, but not always. The example I show is kind of extreme. The effect is also seen with U, but as most of the time U is very low (< 1000 ppm), it is often less obvious than with Th (or Pb when Pb-rich). Because U, Th and Pb increase, this is essentially resulting in "minor" age change - most of the time but not always. The "extreme" example I show is resulting in ages 15 Ma younger compare to age without the TDI (for a 2200 Ma old monazite with 10% Th and 1% Pb). However, we should be able to resolve age difference of 10 Ma, especially with such Th and Pb rich monazite. It then become crucial... Moacyr (one of the age reference material) is less affected with only changes of a couple Ma for an expected age at 507 Ma, but this monazite has much less Th and Pb (6% and 0.5% respectively).

Now you can understand why it becomes critical for us to get a *clean* background measurement: with prolonged beam exposure, there is a risk of changing the background value because the light element (P) diffuses out and there is a "gain" of heavier one. As usual, not a big deal for major to minor element, but when it comes to traces, this is critical. I think MJJ still has some issue getting good ages with the on peak time factor (too old ages on the standard I think - suggesting the background is *too* low), but hopefully this will be resolved soon. Meanwhile, it turns out that the TDI is enough to get good ages.

Edit by John: For those wishing more information on EPMA age mapping of monazite, please also see this topic:

http://probesoftware.com/smf/index.php?topic=100.0

[Julien]

And for those who are interested in the detail, here is a DRAFT (!!) of the method we employed at UMass. This draft do NOT contain the TDI option, but we realize it is better to also activate the TDI for U, Th and Pb analysis, and always keep the TDI activated.

(sorry for the bad quality of this PDF, cannot upload file larger than 2 Mb... contact me at julien.allaz ---at--- colorado.edu for a better quality one [original is 10 Mo, or 5 Mo for slightly lower quality but still much better than this attached document]...).

Julien

Edit by John: I uploaded a 5 MB version which looks better.

[BenjaminWade]

Hi Julien (and all)
That pdf is an awesome resource for people interested in setting up chemical dating. I am about to try and resurrect it in my lab, as it was used last about 6 years ago on our older SX51. We have since obtained an SXFive with large crystals and I am interested to see how it performs with monazite analysis.

I do have one question though. Have yourself, or anyone, experimented with using MAN correction to get around all the crazy background position problems for the REE? I know that theoretically this is the perfect example of when you could/should apply the MAN correction, but has anyone actually done it for monazite and shown that it works fine? And uses it routinely for this purpose?

Or is the monazite community still sticking with background modelling?

cheers

[John Donovan]

Hi Benjamin,
Using MAN would be a good idea to avoid off-peak background issues, but...  there are a few caveats.

First you should refer to this post here where Karsten Goemann uses the MAN background on some REE minerals:

http://probesoftware.com/smf/index.php?topic=4.msg17#msg17

The main issue is that as the average atomic number rises, the background intensity increases also, so that accuracy becomes more important then precision. The MAN correction has an over abundance of precision, but might suffer from accuracy issues when analyzing trace concentrations in silicates and oxide below 100 to 200 PPM (a little higher for higher Z minerals).

This post here shows how one can perform trace element analyses with the MAN correction down to the level of precision (under 10 PPM in SiO2), but only when also applying the blank correction for improving accuracy:

http://probesoftware.com/smf/index.php?topic=29.msg237#msg237

And obtaining a suitable blank material for SiO2 or ZrSiO4 is relatively easy since these can be synthesized, but a suitable blank would be quite  difficult for a complex mineral such as monazite.  So I would stick with the multi-point background for traces in monazite analyses but definitely try the MAN correction for more simple high Z minerals where a suitable blank is available.

Edit by John: Of course, you don't need to use a blank standard if the MAN accuracy is good enough. How does one know if the accuracy is good enough? Well one trick is to run the same acquisition with both off-peak and MAN background corrections as discussed here:

http://probesoftware.com/smf/index.php?topic=4.msg189#msg189

That is a direct comparison of the two background methods utilizing the exact same intensities for both methods. Of course one can also just acquire MAN samples and then check how well one is measuring "zero" by analyzing a standard where we know the element is not present  After all, that is almost how we do a blank correction (but the standard is acquired as an unknown exactly similar to our unknown samples).

Does that answer your questions?

[BenjaminWade]

Hi John
Many thanks for the detail answer. I have had the pleasure of seeing Karsten present that UREE talk at one of the workshops here, and it is great stuff. I agree with your points on potential poor accuracy for low concentration elements, and every time I use the MAN correction I have run it with off-peak first to check consistency. Not using the MAN correction for high level elements in the monazite method seems like an ultimate waste to me. Surely it is a great situation to use it?

My follow up question to you and others would be more of a philosophical one, and probably difficult to answer...

If you obtained a paper for review which was focussed on monazite dating, in which someone had used a combination of the MAN correction for higher intensity elements (LREE, P), and off-peak for others (U,Th,Pb possibly HREE), what would you want to see in this paper to demonstrate that using the MAN correction was suitable? Comparison of off-peak and MAN results? Obviously refs to Donovan MAN papers?

It is a bit of a strange question, but due to my inexperience I am not aware of how widely accepted to the MAN correction method is in the EPMA community worldwide. If/when I get the monazite method running I don't want to let people collect a whole heap of data that may be correct, but hard for them to publish due to reviewers skepticism.

Cheers

[John Donovan]

It's not an unreasonable concern...

In the 1990s there was a bit of resistance to the idea because it was new and not so intuitive, but by the 2000s pretty much everyone happily cites the original MAN paper and publishes a mixture of off-peak and MAN corrected data without any difficulty at all.

I certainly haven't heard of any cases where reviewers raised a concern... not for over 10 years now.

A Google search shows how widely cited the MAN method is:



So, bottom line, you and your students are safe to publish using this method!

[Julien]

Just to follow up on this, I have personally no experience with MAN on monazite. However, I know that Heather Lowers from USGS does use it for REE analyses. As John said, you might be in trouble for minor/trace elements (HREE in monazite), but it should certainly work for LREE.

On another topic that we already discussed (John), ultimately it would be wonderful if we could apply the multipoint background acquisition to model the entire background curvature under the whole set of REE and then apply this "single" multipoint background acquisition to all REE. I know this might be tricky to code in the present state of Probe for EPMA philosophy. I guess such a method would be incredibly powerful, but it definitely means that you need to create a separate table in the MDB file for "backgrounds", and enable the user to choose the regression he/she wants to a specific set of data (i.e. one specific background regression for the entire set of 13 "common" REE (La to Yb, minus Pm). Also in this case, one needs to worry about possible absorption edges along this large spectrometer range... Maybe one day we can further implement a correction for such absorption edges...

Another advantage of such a philosophy (separating background from peak acquisition) would be to enable the acquisition of a separate background acquisition on a set of multiple "peak only" acquisition; and this would refer to the other discussion we had with MJJ and MLW about running a single background acquisition (WITHOUT acquiring the peak) and then apply this background acquisition to the then n-th peak acquisition. Once again, you realize (I hope) that background can dramatically change over a long analysis time at high current due to beam damage (diffusion of lighter element => increase in background intensity over time). Of course, have the on-time peak factor helps, but I think there is still this other approach mentioned here...

Best,

Julien

Edit by John: It would be a lot of work and as you said, it would not account for modeling the absorption edges. I think this is already posted here as a suggestion by Karsten Goemann and named "background sharing":

http://probesoftware.com/smf/index.php?topic=9.0

[John Donovan]

Julien,
One of the problems with attempting to model a large swath of spectrometer range is that due to the non-linearity of WDS, applying physics to these intensities is not trivial. In fact, this is exactly the reason why the "virtual standard" feature is limited to around 5% relative accuracy (even though I do include a matrix correction in the virtual standard intensity interpolation).

Now a 5% accuracy error is acceptable for a standard intensity when it is being utilized as a primary standard for a trace element where the background dominates the accuracy, especially when the alternative is being unable to perform the analysis at all, but a 5% error in modeling the continuum for a trace U, Th, Pb, etc is quite problematic.

I'm just thinking out loud here, but... perhaps there are other approaches we could discuss. For example, one of the major advantages of the MAN correction is that because your background calibration is based on absorption corrected *on-peak* intensities there are no off-peak modeling issues at all. This is why we can obtain such amazing precision and accuracy for trace analysis in SiO2, TiO2, ZrSiO4, etc as recently described here:

http://probesoftware.com/smf/index.php?topic=29.msg237#msg237

The residual accuracy errors (less than 100 PPM) are easily corrected by including the Probe for EPMA "blank correction". Which is accomplished using a well characterized zero blank or even a well characterized non-zero blank. In the case of SiO2, TiO2, etc this is not so difficult as one may obtain well characterized pure SiO2, TiO2 materials for use as a blank.

In the case of complex materials, such as monazite, this is less easily done. It may even not be possible given the compositional range of monazites.

But, and again I'm just thinking out loud, but could we perhaps find another way to improve the accuracy of the MAN correction for these complex materials?

I'm going to suggest that we might want to acquire some test data on some monazite standards using the MAN correction and see exactly what we can accomplish...

I also note that the average Z for monazite (Moacyr) is quite moderate, about 42.  Hmmm.. that number sounds strangely significant!   

http://en.wikipedia.org/wiki/42_%28number%29

 

[BenjaminWade]

Awesome. Thanks for the replies. I will acquire some data with traditional off-peak and compare it to MAN corrected data.
So the answer to the ultimate question is EPMA. Figures~

[John Donovan]

Awesome. Thanks for the replies. I will acquire some data with traditional off-peak and compare it to MAN corrected data.
So the answer to the ultimate question is EPMA. Figures~

Don't forget, you can still utilize the quantitative spectral interference corrections in Probe for EPMA for both off-peak and also MAN background corrected data...

[Dan MacDonald]

Good afternoon, everyone:

Over the last short while, I have ventured into the realm of monazite chemical-age dating using PfE software, and have done a lot (not exhaustive, however) of background reading of topics in the forum, which have been very helpful and informative.  I am not an expert with this method at all, however.

A student using the PfE software noted that he got age results that are about 15% lower than expected for an age-standard we have in the lab (a GSC standard).  In going through his file, there are a few samples where he didn't specify concentrations for the 'major' elements, so I would expect age results from these samples will be off significantly, but I am still searching for answers for his other unknowns, where he assigned the correct values for majors using the "Specified Concentrations" option. (Are some default values used for matrix-corrections where concentrations of majors are not specified?)

First of all, unfortunately, we don't have a well-characterized YPO4 (xenotime) standard in the lab, so we had to resort to a YAG standard for correcting the Pb_Y peak overlap, which is not ideal owing to the difference in chemical states of Y between YPO4 and YAG and associated peak-shifts. If anything, shouldn't one expect apparent calculated ages to be higher than accepted for an age-standard (and for unknowns) using YAG for the Pb_Y versus those obtained using a YPO4 standard?

Secondly, we don't have an anhydrous Pb-phosphate standard for Pb, we have a PbS (galena) and a Pb-chromate (crocoite) standards.  My predecessor had used crocoite in the past, in the inherited methodology.  Is there an inherent shift of the PbMa peak similar to that of the interfering Y-peak that I have missed that would account for a 15% difference in observed PbMa intensity that would carry through into the age calculation?

Thirdly, owing primarily to tight time constraints in trying to adapt someone else's protocol from an old version of JEOL software (i.e., UNIX-based, circa 2004) (and where I suspect the main fault is going to be found in the 15% error in age determination, mentioned above), only 1 upper and 1 lower off-peak intensities were measured, instead of the recommended, more rigorous backgound-modelling methods widely discussed in this forum (and in the scientific literature).  I realize that rigorous background determination for unknowns is more critical in obtaining good analyses for this technique than is standardization/calibration.

My recommendation to the student was to re-acquire minor-element data using a multi-point background model, while I track down a good piece of YPO4 for the Pb_Y overlap correction.  The student then asked if it was possible to use already acquired wavescans and associated DAT files to extract background values and somehow plug those values into the age calculation.  I pointed out that the analytical conditions used during the wavescan acquisition are not the same as those used for the minor element determination, and that, even though the intensities are quoted as being normalized to cps/20 nA for both wavescan and quantitative data, the reliability of the counting statistics for the wavescans is relatively low, owing to the huge difference in dwell times between the two types of measurements.  That being said, I wouldn't know where in the software one would put in 'normalized (to 20 nA)' background values from multiple points taken from wavescans, nor even if it is possible (nevermind advisable!) to do this manually.

Any helpful hints (and strong boots to the rear!) in the right direction are most appreciated!

Best regards

Dan MacDonald

Probe servant
Dalhousie University

[Probeman]

A student using the PfE software noted that he got age results that are about 15% lower than expected for an age-standard we have in the lab (a GSC standard).  In going through his file, there are a few samples where he didn't specify concentrations for the 'major' elements, so I would expect age results from these samples will be off significantly, but I am still searching for answers for his other unknowns, where he assigned the correct values for majors using the "Specified Concentrations" option. (Are some default values used for matrix-corrections where concentrations of majors are not specified?)

Hi Dan,
I'd say that 15% absolute accuracy for a monazite age isn't awful especially for an early effort.  What was the variance?

Chem age dating of monazites by EPMA is complicated, the background measurements being the most challenging.   I'm no expert in this area either, but besides the background corrections, the spectral interferences are very important, especially the interference of Y on Pb, but also Th on U.  In addition as Ben Wade pointed out, there is also an interference of La on Pb as seen here:

For Pb ma   LPET at  5.28601 angstroms, at an assumed concentration of 1 wt.%
  Interference by Ce Ln       II    at  5.24140 ( 59771.1) ( -509.87) =      1.8%
  Interference by Th MZ1            at  5.24510 ( 59813.4) ( -467.57) =     30.0%
  Interference by Y  LG3            at  5.28490 ( 60268.4) ( -12.648) =     37.5%
  Interference by La LA1      II    at  5.33280 ( 60815.9) ( 534.859) =     99.1%
  Interference by U  Ll       V     at  5.33680 ( 60861.6) ( 580.582) =      5.7%
  Interference by Th MZ2            at  5.34020 ( 60900.4) ( 619.441) =      5.9%
  Interference by La LA2      II    at  5.35210 ( 61036.5) ( 755.461) =      1.1%

Did you see Ben's recent efforts at Chem Age Dating by EPMA here?

http://probesoftware.com/smf/index.php?topic=1029.msg6757#msg6757

My recommendation to the student was to re-acquire minor-element data using a multi-point background model, while I track down a good piece of YPO4 for the Pb_Y overlap correction.  The student then asked if it was possible to use already acquired wavescans and associated DAT files to extract background values and somehow plug those values into the age calculation.  I pointed out that the analytical conditions used during the wavescan acquisition are not the same as those used for the minor element determination, and that, even though the intensities are quoted as being normalized to cps/20 nA for both wavescan and quantitative data, the reliability of the counting statistics for the wavescans is relatively low, owing to the huge difference in dwell times between the two types of measurements.  That being said, I wouldn't know where in the software one would put in 'normalized (to 20 nA)' background values from multiple points taken from wavescans, nor even if it is possible (nevermind advisable!) to do this manually.

To a first approximation, the standards aren't going to matter near as much as the backgrounds for trace elements, because the background measurement dominates accuracy for trace elements, especially high z materials such as monazite.  Of course the older the monazite the easier, as you'll have more Pb to measure...

Question are you using the U Ma or U Mb line?  I'm sure you realize the presence of the Ar absorption edge makes this measurement difficult.

As for importing wavescans, if the wavescan is from another PFE run, one can import that scan from the New Sample Setup dialog as discussed here:

http://probesoftware.com/smf/index.php?topic=74.msg274#msg274

Once the wavescan is imported into PFE it can be utilized to select background positions for either normal off-peak or multi-point background.  The real experts on this method are Mike Jercinovic and Julien Allaz, so hopefully they will also chime in (it was Mike and Julien that developed the multi-point bgd method that we implemented in PFE for trace element analysis in general, and monazite dating in particular).
john

[neko]

Updating this topic with a paper I've found on the subject, https://www.sciencedirect.com/science/article/pii/S0016703701007530 - which seems to be an update of the CHIME method and seems to only be suitable to magmatic monazite with multiple zones that are the same age (thanks to Julien and Mike for the explanation via email). It does claim a large (~10x) reduction in errors, but according to Mike, you can get lower error using multipoint backgrounds with the tradition method.

It's mostly outside my domain, I was only curious to see if any improvements to the method had been made since the initial 1996 and 1999 papers on the subject.

[Julien]

From the recent discussion on the SX50 email listing (about the method used by Cameca for microprobe age calculation  and about the article of Cocherie & Albarede (https://www.sciencedirect.com/science/article/pii/S0016703701007530) mentioned by Neko in the previous post).

=======

Interesting paper from Cocherie & Albarede. I have only run through the paper quickly, and don’t see an obvious “flaw”. However, it seems to rather be an approach similar to the CHIME method developed by Suzuki & Adachi. With this method, the precision on the age can seriously be improved BUT… it relies on the fact that you do have multiple generations of monazite with different composition (either different absolute U, Th and Pb content, or in the case of Cocherie different U/Pb and Th/Pb ratios). It also assumes that despite the change in composition, the age is the same. This is often valid in the case of magmatic monazite, where the crystal shows some sector zoning with differences in composition but similar ages. I would stress out that this is often NOT the case with metamorphic monazite, or with monazite that have suffered a (partial) resetting or complex post-emplacement magmatic history.

Coming back to Cameca, I believe it does consider the Montel et al. 1996 equation. I have just recently submitted this paper to the European MAS conference last May (in review; accepted paper), and I am attaching it for your convenience. It will soon be published in the IOP conference series journal. I also attach a small Excel macro I wrote to calculate the age (iterative procedure…).

Cheers,

Julien