CalcImage Quantitative X-ray Mapping Examples

[John Donovan]

If you want to "play with" a number of interesting geological and material science examples of quantitative  x-ray mapping in CalcImage, you can simply download the CalcImage.ZIP file here:

http://www.probesoftware.com/download/CalcImage.ZIP

Warning: the ZIP file is just over 1 GB in size, so use a fast connection to download!

These examples include the use of off-peak and MAN backgrounds, matrix and interference corrections, and numerous jpg and pdf presentation quality outputs from our automatic scripting feature.

These example files can be unzipped to your SurferData folder which should already exist in your C:\Userdata folder.

To use these examples, you'll need to run CalcImage.exe and first click the File | Select Standard Database menu and browse to the supplied Standard.mdb file for the proper standard compositions.  After that you can open any of the already calculated CalcImage project files (*.CIP) and the images will load automatically.

See attached screenshots from the first example in the Samples2 folder (first after the project is opened and then after clicking the File | Close Off-Peak Images menu).

Have fun and please let us know if you have any questions at all!

[John Donovan]

I just updated, re-calculated and uploaded an all new CalcImage.zip file with many cool new examples showing the power and flexibility of Probe For EPMA, Probe Image and CalcImage for high accuracy quantitative x-ray mapping.

I've also added a new CalcImage2.ZIP file which contains all the raw x-ray data maps in case you want to walk through the simple steps to create a CalcImage project from scratch.  Once Karsten Goemann finishes the Probe Image and CalcImage quantitative mapping tutorials, I'll add a link for that here also. Karsten's quantitative x-ray mapping tutorial for Probe for EPMA (creating a sample setup and acquiring standrad intensities for qaunt mapping) is here:

https://probesoftware.com/smf/index.php?topic=106.0

In the meantime you can download the updated CalcImage.ZIP file which contains all the standard intensity files and quant output files (dat, jpg, bmp, etc) here:

https://www.probesoftware.com/download/CalcImage.ZIP

This CalcImage.zip contains the following folders, so simply extract them to your already existing SurferData folder in your User Data Directory (usually C:\UserData\SurferData). Be careful not to create an "extra"  CalcImage folder when you extract the files!



Optionally you can also download the new CalcImage2.zip which contains the raw x-ray maps for the above examples. The raw x-ray maps for all the CalcImage examples above can be downloaded here:

https://www.probesoftware.com/download/CalcImage2.zip

This CalcImage2.zip file should be extracted to your already existing SurferData folder in your User Images Directory (usually C:\UserImages\SurferData). Again, be careful not to create an "extra" CalcImage2 folder when you extract the files.



Once you have these example CalcImage files, you can use them to walk through the CalcImage tutorial which Karsten Goemann will also be posting here soon. As mentioned in the previous post, after running the CalcImage app, you'll need to first load the "demo" standard composition database (which you should already have extracted to the SurferData folder) by clicking this menu and selecting the standard.mdb file in the User Data Directory SurferData folder as seen here:



Note the CalcImage.ZIP file is now around 1.8 GB (the new CalcImage2.zip file of the raw data is only 62 MB!), so be patient downloading (maybe an overnight download?).

[John Donovan]

Here's a cool new capability. Quantitative slicing of a detection limit map. See attached.

Also see the attached example of "slicing" a feldspar (orthoclase molecule) mineral end-member map.

[Probeman]

I decided to add some x-ray mapping files for some natural zircons I recently acquired using both off-peak and MAN methods. I've attached the files as a zip file below if you want to play around with them yourself (remember you must be logged in as a member to see attachments).

Here are quantitative (background, matrix and blank corrected) x-ray maps of a natural zircon core, first using off-peak bgds:



Unfortunately, there were some small apatite inclusions, so the trace phosphorus map didn't turn out well.  Now here is the same map, but background corrected using MAN backgrounds, which because only the on-peak maps are utilized, the mapping acquisition would take approximately 1/2 the time of on and off-peak maps (that is, ~10 hours instead of 20 hours!):



Not the significantly improved sensitivity, again in about 1/2 the acquisition time of the off-peak map. Next here is another zircon core, again first with off-peak bgds, and this time there were no apatite inclusions so the phosphorus map is more interesting:



and again, the same map, but processed using MAN backgrounds:



Again the MAN x-ray maps show significantly improved precision for trace elements.  My recently published paper on this method is found here:

http://epmalab.uoregon.edu/publ/A%20new%20EPMA%20method%20for%20fast%20trace%20element%20analysis%20in%20simple%20matrices.pdf

[Probeman]

I realize that I should also show the CL images for these two x-ray maps. Unfortunately I forgot to acquire the CL images prior to the x-ray maps, so the CL is quite suppressed from the beam damage.

But they are attached below: if you login to see attachments.
john

[Probeman]

As shown in a previous post here:

http://probesoftware.com/smf/index.php?topic=73.msg5130#msg5130

there is a significant improvement in precision for trace elements when performing MAN background processing, as opposed to traditional off-peak methods. We can quantify this improvement using the methods described in my Amer. Min. paper published in August earlier this year (see above post for a link to the full paper).

The first attachment below (remember to login to see attachments!) are the calculated off-peak detection limits for each pixel in the above zircon trace elements maps in 99% (3 sigma) detection limit units. The second attachment are the calculated detection limits for the same map but calculated using the MAN background method.

For off-peak detection limits, the variances of the on-peak and background are roughly similar and add in quadrature. For the MAN method the on-peak variance is also added to the background variance of the MAN calculation in quadrature. However, in this example, because we only measured the trace elements and therefore the matrix was specified as ZrSiO4 by difference, the calculated MAN background variance is essentially zero. Please see the paper for a comparison of off-peak, MAN (with a fixed matrix), and MAN (with a measured matrix) where you will see that both MAN methods yield a background variance close to zero.

In the MAN fixed matrix calculation this is because the matrix (and hence average atomic number) is essentially determined by the average Z of the specified ZrSiO4 composition. In the MAN measured matrix (Zr La and Si ka), the background variance is slightly larger than the MAN fixed matrix, but because the average atomic number is dominated by the measured major element statistics, the variance is still very close to zero when compared to the off-peak variance.

The first thing to note visually is that because the MAN method is more sensitive we can actually see the variation in the calculated sensitivity from the on-peak intensities in the MAN detection limits.  Second, we can note that the detection limits are about two times better (lower) in the MAN method than the off-peak method.

One helpful way to think of this is to consider the Nth point off-peak method, which because there is only one off-peak measurement per N analyses, the background variance is *exactly* zero for that data set.  The downside to the Nth point off-peak method is of course that if the composition changes, the background will change as well, inducing an accuracy issue in the trace element measurement (not to mention possible changes in off-peak interferences!).  The MAN method automatically handles any changes in composition because the average atomic number is calculated based on the actual composition of the material as the major (and minor) element concentrations change!  And of course because the MAN method does not utilize off-peak measurements, there is no possibility for off-peak interferences.   

I know this all seems very unintuitive (it took me two years to understand it), so please let me know if you have any questions at all.
john

[John Donovan]

I added code in the latest CalcImage version to display the area outside the polygon or "freeshape" area of quantitative x-ray maps as "blank" or specifically a gray color. Here are some test x-ray maps from Owen Neill:



Here's the same thing but with oxide output:



What I want to know is, in the SE image, what are little "dew berries" doing on the surface?   
john

[Mike Matthews]

Looks like your coating was bubbling up a little under the beam. You may have had some residues of something on the sample prior to coating and the beam has caused it to expand.

[Probeman]

Looks like your coating was bubbling up a little under the beam. You may have had some residues of something on the sample prior to coating and the beam has caused it to expand.

Or it could be beam damage or more likely carbon contamination spots from previous point analyses.  SE images show all the dirty stuff!     We'll have to wait for Owen to chime in.

[Paul Carpenter]

Here are quantitative compositional maps acquired using the JEOL freeshape routine.
A summary of what is done, here for setup using JEOL 8200 (applies to 8200,8500, freeshape not available in 8900 software). This is not a manual, just a guide.

1. Probe for EPMA: Calibrate quant run setup using MAN background calibration; use analytical setup to allow multiple pass WDS mapping. Make a new sample of type unknown with parameters for mapping, count times and probe current, to be used by CalcImage.

2a. Set up JEOL mapping (Analysis - Map Analysis):
(Measurement - Element Condition): Setup list of WDS elements (Element), use PFE peaking options to drive spectrometers to peak positions for each pass, then read positions using the JEOL mapping software (Condition) and set up for (in this case) two passes.
Make sure in IMS Signal Compo to read BSE brightness and contrast applicable to sample at mapping probe current.
2b. (EOS Condition) Set EOS using probe current, spot mode, and probe diameter for mapping.
2c. (EDS Condition) EDS parameters as needed for mapping run.
2d. (Stage Condition, position input, opens Area Input window) Drive to center of stage map area, click Read Store to Center, and enter map parameters (number of pixels, pixel size, dwell time), then Confirm and set z-axis focus at UR, LR, LL, and UL corners. Now *before* clicking Apply, select Special map - Free shape map. click New in Free Shape Map window; drive to start point and click JEOL joystick STOR button at each point to define polygon line boundary.  When done click End (*not* Close!), should have free shape polygon as shown here. Click Ok. Now click Apply in Area Input window, Close.

2e. Use Measurement - Preset Measurement to run the maps.
2f. When done, as necessary, copy group-sample from JEOL Sun workstation to your PC. It is the .map directory that contains all the binary Jeol map data and the 0.cnd file used for parsing out the data. On your PC you may need to select folder options show hidden files in order to see directories starting with a dot (.).

3. Probe Image: File - Convert JEOL or Cameca etc. to Prbimg (temporary fix), navigate to location of .map and fine 0.cnd, follow directions from there to generate PrbImg files.
4. Use CalcImage to correct the maps!

Here is the result for an irregular polygon map collected by the JEOL software, processed by Probe Image and CalcImage (please login to see attachments).

This, coupled with the ability to generate all Surfer plots for all requested data in CalcImage, is a major step forward.

Cheers,

Paul

[Probeman]

Just a pretty picture from a Al, Mg, Gd, Sn alloy I did several years ago, displayed in the RGB processing window in CalcImage:

[Probeman]

People have been discussing color scales and our visual sensitivity to color here:

http://probesoftware.com/smf/index.php?topic=1003.msg6583#msg6583

As discussed by Anette, there can also be issues with color transitions with complex color scales such as the rainbow color scale, in that we see contrast differently depending on the levels in the data.  For example here is a quant map of Fe Ka in a Ti implant that we recently ran:



After I saw this output I decided to re-calculate the data without the off-peak maps, and simply use the on-peak map data and the other standards as an MAN bgd calibration, to see if I could bring out these features more clearly, as seen here:



And I went, whoa, that is just too good to be true.  I mean we know the MAN bgd method is less noisy than the off-peak background method, but I became concerned about the data levels interacting with the color levels. So I plotted both up again, but this time using the thermal palette which is a much more gradual color change without the abrupt color transitions in the rainbow scale:



So yes, there is a significant improvement in the sensitivity with the MAN method, but not that much!

[John Donovan]

Karsten Goemann sent me these nice quantitative polygon x-ray maps he did on his JEOL 8530.





He acquired these maps using the "free shape" mapping feature in the JEOL software and acquired standard intensities in Probe for EPMA and then used Probe Image to convert the JEOL maps and finally calculate the quantitative results in CalcImage.

[Karsten Goemann]

Here is another example for those olivine maps, same method (map acquisition with JEOL software, conversion with Probe Image, quantification in CalcImage using MAN backgrounds.

We've been doing quite a few of these maps now and the procedure is complicated (including 2 things in the JEOL software I need to work around). Being able to do these free shape maps in ProbeImage would be a big improvement.

[John Donovan]

Once in a while someone measures oxygen as an analyzed element, and then for whatever reason wants to then try and instead calculate oxygen by stoichiometry.  So they disable oxygen for acquisition, and then acquire samples with oxygen to be calculated by stoichiometry during the quantification.



Obviously one doesn't want to include both measured oxygen and oxygen by stoichiometry in the matrix correction, so if one has measured oxygen as an analyzed element, and then from the Calculation Options dialog, one also specifies Calculate Oxygen by Stoichiometry, the program will warn you that you cannot have oxygen both analyzed and calculated by stoichiometry. So one simply specifies that the measured (analyzed) oxygen channel is disabled for quantification as seen here:



This method of being able to switch samples back and forth between measured oxygen and oxygen calculated by stoichiometry has long been a feature in Probe for EPMA, but until recently we guess no one had tried this in CalcImage when quantifying x-ray maps.  Until now.



See the lower left map in the above image which is the oxygen calculated by stoichiometry, even though oxygen was measured on one of the WDS spectrometers. This is now possible in CalcImage after one has specified that the measured oxygen is disabled for quant as seen in the Elements/Cations dialog.

Thanks to Brad Johnson at PNNL for pointing this out.  This update is ready to download now.