Questions about MAN background use

[John Donovan]

I cannot over emphasize the importance of properly fitting one's standards for the MAN background fits.

The basic "rule of thumb" (and it is merely a "rule of thumb" so there are exceptions!), is that since we are trying to characterize the continuum as a function of average atomic number, and background is, by definition, generally the lowest intensity we can measure, then any standard which plots "above" the fit as seen here, is a good candidate to <ctrl> click to unselect and update the MAN fit:



In the above case, two standards that both do not contain fluorine, are seen to give an elevated background, why? Because P and Cr both interfere quite well with F ka...

In other situations one might actually find that they have a trace element contamination in their standard that they were not aware of (Probe for EPMA automatically deselects all standards that contain any of the element of interest in the MAN fitting) and therefore that standard will also plot "above" the general trend. Therefore "contaminated" standards should also be deselected and the fit updated similarly as we did with the above apatite and Cr2O3 standards which interfere with F ka.

In practice you do not have to know exactly why a particular standard intensity plots above the general trend of continuum intensities, just unselect and re-fit. Though as is often the case, the physics details might be interesting!



For example, why does the copper standard intensity cause the intensity to "curve over" and decrease?  Because in higher average atomic number materials, the self absorption in the material, for low energy x-rays such as fluorine, is increasing faster than the increase in continuum production due to the slowing down of the primary electrons.

[John Donovan]

John - is it possible to use a combination of MAN and off-peak backgrounds for samples where all elements were originally acquired with off-peak backgrounds? To elaborate, following your and Karsten's example, I acquired all the samples and standards with off-peak backgrounds, assigned the MAN fits after checking "Use Off Peak Elements for MAN Fit", and then applied them to my unknowns by checking "Use MAN correction for Off-Peak Elements". I was running a routine for REE-phosphates similar to Karsten's, and was looking to do a similar comparison (which turned out to be quite informative).

Sounds like you did everything correct. This is exactly why I implemented the ability to change all off-peak acquired elements to be treated as MAN background acquired.  This makes sense since the on-peak data is already there, one simply needs to ignore the off-peak data!  That way one can compare the same data using both off-peak and MAN (one can even treat a multi-point background as a normal off-peak sample!).   

This was the reason for this post:

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

I noticed that checking the "Use MAN Correction for Off-Peak Elements" option seems to apply the MAN background to all elements. Is it possible to post-process the data using a combination of MAN backgrounds for a few elements, and off-peak backgrounds for the rest? For example, could I use the off-peak backgrounds for the middle-heavy REE's, and use MAN backgrounds for La, Ce and P? Or would I need to have specified MAN or off-peak for each element prior to acquisition? Apologies if I'm missing something obvious here.

Yes, you can switch all off-peak elements to MAN to compare, but not selected elements. The only thing obvious you are missing is that if this is what you want to do, then do it!   

That is acquire "off-peak backgrounds for the middle-heavy REE's, and use MAN backgrounds for La, Ce and P".  Remember, you can acquire a different background for every element *and* also acquire different backgrounds for the standards compared to the unknowns- if that became necessary...

You might also take a look at this post by Karsten Goemann on looking at REE elements:

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

[Probeman]

There is a simple method to compare the same intensity data using off-peak, MAN and Nth point background corrections...

The above post:

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

mentions calculating the same data using both off-peak and MAN methods to compare the accuracy and precision.

Please note that one might consider making a copy of the original probe data MDB file to another folder, so that one copy can be treated using off-peak background corrections and the other copy can be treated using MAN background correction methods.

This allows the user to keep the results for each method without having to re-assign and re-calculate.

[JohnF]

I am going to post something about MAN, but first I'll make a suggestion. John, you might consider making a separate MAN discussion page. Cause when you search for MAN or "MAN" you also get hits for many, mania, manuscript, etc.
As I hope MAN use takes off, it might be useful to make it easier for people for find MAN posts.

Just thinking....

[JohnF]

John, can you clarify exactly under what conditions one should click the "Clear all MAN assignments" button?

Case in point. I have set up a 10 element glass MAN setup. I acquire all the MAN backgrounds, and life is good.

A month later, an occasion comes to use the MAN backgrounds but I need to add another element, P. Interesting, I think, as I want to add it on TAP, where it will be at a low angle and thus the background is curved, and since P in a minor element in the basaltic glasses, I have to worry about getting the proper background curvature. But wait.... if I use MAN, I don't have to worry about the modeling of the curvature, since the MAN takes it all into account. Another reason to use MAN, even for minor amounts (I'm expected 0.5-0.9 wt%).

But my question. Since I've added an element, do I need to "start all over" in the MAN assignments (click the clear button), or, preferably, just select "Assign MAN fits" and go thru the procedure. 

Note also that in this procedure, I  am re-acquired (after 2-3 months) the backgrounds on many of the MAN stds I acquired earlier this summer, so I can check how the bkg many have changed, as I assume it probably hasn't but would be nice to be sure.

So the question is, do you recommend clearing the prior MAN background assignment and "starting" fresh?

[John Donovan]

I am going to post something about MAN, but first I'll make a suggestion. John, you might consider making a separate MAN discussion page. Cause when you search for MAN or "MAN" you also get hits for many, mania, manuscript, etc.  As I hope MAN use takes off, it might be useful to make it easier for people for find MAN posts.

I appreciate the thought.  Right now I've tried to contain the MAN discussions into a main area. The main topic is for questions about the current MAN features and how to use them. The correct topic for these types of questions is the current topic:

http://probesoftware.com/smf/index.php?topic=4.0

which was started by Mike Spilde last summer.  Your questions below in the next post are perfectly appropriate for this topic.

In addition there are some "specialty" topics on using MAN bgd corrections, such as...

1. Quantifying carbon using MAN bgd calibration curves:

http://probesoftware.com/smf/index.php?topic=48.0

Ironically, the carbon in steel quantification method used by many European EPMA users is a very simple form of our MAN method, but they only utilize a single standard (pure Fe) to obtain a background for carbon and do not perform a continuum absorption correction.

The main downside with using a single standard for the MAN regression is that if the average Z in the sample changes (e.g., unknown alloy steels and/or the carbon standard which is normally Fe3C), the background will be incorrectly calculated because the composition is different and therefore the average Z is different from our pure Fe standard used for the background. Our multi-standard MAN calibration curve allows the user to accurately apply the regression over a range of compositions and hence over a range of average Z.

2. Another MAN topic is using the MAN correction along with the blank correction to improve precision and reduce integration time for trace elements:

http://probesoftware.com/smf/index.php?topic=29.0

See especially reply#5 in the above Ti in Quartz topic.

3. There is also a topic here discussing the relationship between accuracy and precision in the MAN background correction:

http://probesoftware.com/smf/index.php?topic=307.0

I look forward to more discussion here as it is not a simple thing to think through.

[John Donovan]

John, can you clarify exactly under what conditions one should click the "Clear all MAN assignments" button? Case in point. I have set up a 10 element glass MAN setup. I acquire all the MAN backgrounds, and life is good.

A month later, an occasion comes to use the MAN backgrounds but I need to add another element, P. Interesting, I think, as I want to add it on TAP, where it will be at a low angle and thus the background is curved, and since P in a minor element in the basaltic glasses, I have to worry about getting the proper background curvature. But wait.... if I use MAN, I don't have to worry about the modeling of the curvature, since the MAN takes it all into account. Another reason to use MAN, even for minor amounts (I'm expected 0.5-0.9 wt%).

But my question. Since I've added an element, do I need to "start all over" in the MAN assignments (click the clear button), or, preferably, just select "Assign MAN fits" and go thru the procedure. 

It's really a six or one half dozen of the other sort of thing. Either way is fine, it's just that if one made many changes to the setup such as adding several elements and changing the background types, etc., etc., it is entirely possible to correct everything in the Assign MAN Fits menu dialog, but it just might be easier to clear them and let the application re-assign everything automatically.

Then simply review all the MAN fits again to make sure they are good.
john

[UoL EPMA LAB]

Hi All

I have a slightly more parochial question on MAN usage - or rather NOT usage.

I have built a pretty large setup for a wide variety of elements, many of which were copied in from old projects or elements.mdb
I decided to set everything up using off-peak backgrounds, so set both my unknowns AND my standards to use off-peak backgrounds in both the acquire>elements/cations AND acquire>acquisition options windows (see attached). I have then run a full set of standards and now when I try and analyse my standard data in the analyse window I get a message telling me I have no MAN data to fit for Ni (also attached). I don't just see this for Ni but all standards in this batch. Why is PfE looking for MAN data when I have set it to use off-peak BGs across the board?

Whats going on, what am I missing here?

Confused
Leeds

[Probeman]

I decided to set everything up using off-peak backgrounds, so set both my unknowns AND my standards to use off-peak backgrounds in both the acquire>elements/cations AND acquire>acquisition options windows (see attached). I have then run a full set of standards and now when I try and analyse my standard data in the analyse window I get a message telling me I have no MAN data to fit for Ni (also attached). I don't just see this for Ni but all standards in this batch. Why is PfE looking for MAN data when I have set it to use off-peak BGs across the board?

Whats going on, what am I missing here?

Hi Leeds,
One possibility is that because of the standard intensity drift correction feature, the program will scan for all undeleted standards and try to load them for the drift interpolation.

So double check that all MAN standards are deleted and try again.  You can also clear the MAN assignments but that shouldn't be necessary.

I also assume the Analytical | Use Off-Peaks For MAN Fits menu is unchecked?

[Richard Walshaw]

Thanks for the reply.

Will try that tomorrow.

Leeds

 

[UoL EPMA LAB]

Hi Again

I've done as you suggested and deleted the old standards, some of which were imported from projects using MAN BG.
My problem has gone away. All is now good.

Thanks Probeman!
 

[Ben Buse]

Hi John,

I have a question regarding the collection of backgrounds for quant maps of a single phase  e.g. Mg Calcite or Plagioclase- which we do often. In a single mineral backgrounds are constant & for majors peak to background ratios are high so background represents small percentage - making it tempting to ignore backgrounds. What we should do is MAN backgrounds the only reason we dont is to save time by not collecting standard data for standards which do not contain the elements. But I suppose the machine's pretty stable and we could just calibrate for MAN backgrounds every few months as larger errors for backgrounds would not matter? Have you thought about having an option were you could specify a single background in counts per second for the whole map - for this case were your mapping elements within a single phase e.g. here Mg-calcite for minor and trace elements - and the background could be specified for the map from a single quant point.

Thanks

Ben

[Probeman]

I have a question regarding the collection of backgrounds for quant maps of a single phase  e.g. Mg Calcite or Plagioclase- which we do often. In a single mineral backgrounds are constant & for majors peak to background ratios are high so background represents small percentage - making it tempting to ignore backgrounds. What we should do is MAN backgrounds the only reason we dont is to save time by not collecting standard data for standards which do not contain the elements. But I suppose the machine's pretty stable and we could just calibrate for MAN backgrounds every few months as larger errors for backgrounds would not matter? Have you thought about having an option were you could specify a single background in counts per second for the whole map - for this case were your mapping elements within a single phase e.g. here Mg-calcite for minor and trace elements - and the background could be specified for the map from a single quant point.

Hi Ben,
Good question.

Basically you want to do an MAN background calibration curve using a single point... which assumes of course that you have a constant composition (or at least a constant average Z which is essentially the same). However, as you point out, for major elements the background (if it is smaller than the measurement precision) can be ignored, or at least roughly approximated.

So here's an example of what to do if you wanted to say x-ray map (or point analyze) Ca and Mg on say dolomites that are (roughly) constant in composition. By the way, to include CO3 in the matrix calculation by various methods, see this topic:

http://probesoftware.com/smf/index.php?topic=92.0

Let's start with adding Ca and Mg to a spectrometer and then go to the Acquisition Options dialog as seen here and specify that the standard acquisition use off-peak backgrounds but the unknowns (your dolomite point analyses or maps) use the MAN background. Here is the Acquisition Options dialog:



This gets more important as one starts to add additional elements to analyze for, but then one will have multiple standards that *do not* contain the element of interest and can therefore be utilized in the MAN calibration for variation in average Z!

Next we select our standard positions and run them:



Once the standards are acquired we can analyze them as unknowns because we utilized off-peak backgrounds as seen here:

St  135 Set   1 Calcite (analyzed), Results in Elemental Weight Percents
 
ELEM:       Mg      Ca       C       O
TYPE:     ANAL    ANAL    SPEC    SPEC
BGDS:      LIN     LIN
TIME:    20.00   20.00
BEAM:    30.01   30.01

ELEM:       Mg      Ca       C       O   SUM 
     1   -.017  40.141  12.000  47.952 100.076
     2    .028  39.956  12.000  47.952  99.936
     3   -.010  39.994  12.000  47.952  99.935

AVER:     .000  40.030  12.000  47.952  99.982
SDEV:     .024    .098    .000    .000    .081
SERR:     .014    .057    .000    .000
%RSD:     ----     .24     .00     .00

PUBL:     n.a.  40.031  12.000  47.952  99.983
%VAR:      ---   (.00)     .00     .00
DIFF:      ---   (.00)    .000    .000
STDS:      222     135       0       0


Now we go to assign the MAN background for our unknowns using the Analytical menu as seen here:



But whoa, we now get this error message:



Why? Because our standards were acquired using off-peak backgrounds and so the program ignores the off-peak standards by default for the MAN calibration. So we can change that default by clicking this menu in the same Analytical menu as seen here:



Now we click the Analytical | Assign MAN Fits menu again and now we see this for the first element Mg:



and this for Ca:



Yes, we're only using a single point for each MAN calibration curve so it is a constant background intensity for that element, but depending on what standard you pick you'll get a reasonable background estimate for major and maybe even minor elements.  Or you could just acquire one or two more darn standards and get a proper slope for your MAN calibration curves! 

Helpful hint: if you aren't sure which of your standards might have similar average Z values compared to your unknowns, you can use this handy menu the the Standard.exe app as seen here:

[Probeman]

In practice you do not have to know exactly why a particular standard intensity plots above the general trend of continuum intensities, just unselect and re-fit. Though as is often the case, the physics details might be interesting!



For example, why does the copper standard intensity [cuprite #114] cause the intensity to "curve over" and decrease?  Because in higher average atomic number materials, the self absorption in the material, for low energy x-rays such as fluorine, is increasing faster than the increase in continuum production due to the slowing down of the primary electrons.

After seeing the plot above some might ask, why stop there?  Maybe the std 140 (Rutile) is also either contaminated with fluorine (unlikely I'm guessing) or perhaps the Ti La line slightly interferes with the F Ka line? After consideration of the fact that Ti L emission overlaps *strongly* with N ka, we might guess that perhaps the tail of a Ti L emission line could cause a slight interference for F Ka?

The easiest way to try and answer this question is by examining an x-ray database such as the NIST 2005 database in PFE available from the Xray | Xray Database menu as seen here:



Note that the wavelength range was specified by entering F Ka in the Specify Range controls as seen above. Now to see if any Ti emission lines could in principle, interfere with the F Ka line, we can select an element to highlight, in the case Ti, by using the Highlight Element control as seen here:



Or one can select multiple emitters by clicking the Periodic Table button and using the periodic table dialog as seen here:



In either case no Ti lines are found near the analytical emission line for fluorine.  Another method which can provide a nominal estimate of the possible overlaps is by using the Options | Interferences menu in the Standard.exe app. Begin by selecting a the potentially problematic Ti standard or another similar standard such as synthetic TiO2 as seen here:



This will cause the following window to open with the TiO2 composition loaded automatically as seen here:



Now, using the Unknown button, add a small amount of fluorine to the material because the software only check for elements in the material as seen here:



Finally, click the Calculate button to checking for overlaps where as we can see here, nothing is reported for F Ka:

[Ben Buse]

Hi John,

Thanks. I get you where the average atomic number is constant (as in a single phase) so we don't have to correct for changing atomic number within the map we can use a single MAN standard. With best MAN data given where atomic number of MAN standard is same as unknown.

By using even a single MAN standard with similar atomic number (as is the case for most silicates/carbonate standards) we improve the quantification compared to no background. The error due to difference in atomic number of MAN standard and unknown is much smaller than the error with no background.

A good compromise for major elements where p/b ratio is large and background is small % (<1%) of total counts?