Bulk analysis from stage scanning

[Frank Spear]

Has anyone developed a method to obtain an estimate of a bulk rock analysis by scanning the stage during an analysis? I've tried a modified version of this using a set of (many) spot analyses, but it is not time efficient and it seems that stage scanning would be far more efficient and allow covering more of a thin section. In case you might think me crazy to try and get a bulk analysis this way, the application is on layered heterogeneous metamorphic rocks where a bulk separate would not be representative. Thanks. Frank.

[John Donovan]

Hi Frank,
This is exactly the right way to do this. Sometimes people try to save time by defocusing the beam while scanning the stage but that only creates systematic errors for light elements as described in the M&M poster attached to the post linked here:

http://probesoftware.com/smf/index.php?topic=44.0

Yes, you should be able acquire arbitrary points or grids of points in PFE and acquire them as described here:

http://probesoftware.com/smf/index.php?topic=60.0

and perform modal analysis on the quantitative compositional results as described here:

http://probesoftware.com/smf/index.php?topic=41.msg254#msg254

You can also of course acquire beam or (preferred) stage scans over whatever areas and quant, filter and export them in CalcImage and  finally average the results in Excel.  This last method I will describe in more detail here...

We start with a compositional map of one of our CalcImage Mt St Helens samples as seen here:

http://probesoftware.com/smf/index.php?topic=41.msg195#msg195

Quantification using PFE, PI and CalcImage is described here:

http://probesoftware.com/smf/index.php?topic=146.0

Now the easy thing is that once CalcImage has finished the quantification, it automatically outputs a tab delimited .DAT that can be easily imported into Excel as seen here with some explanation of the column labels (see the CalcImage.pdf for complete details):



Note that all quant map data types (elemental, atomic, formula, oxide, end-members, etc) are output if selected as described here:

http://probesoftware.com/smf/index.php?topic=41.msg195#msg195

and other posts in that topic.

And once you are in Excel you can average the data any way you like!

And finally one can also perform horizontal and/or vertical "strip averaging" on any x-ray maps as shown here:

http://probesoftware.com/smf/index.php?topic=41.msg897#msg897

Ok, one more thing... you can also average arbitrary polygon areas to obtain average and variance of the quantitative pixels as described here:

http://probesoftware.com/smf/index.php?topic=41.msg543#msg543

[John Donovan]

I wanted to pop this topic forward for those who may have missed some work Julie Chouinard ne Barkman did back in 2013 where she showed that using a defocussed beam to determine a "bulk" analysis of heterogeneous material, results in significant systematic errors in the average estimated composition.

The reason for these errors are due to the fact that averaging x-ray intensities is simply not equivalent to averaging composition. This is because the matrix correction physics assumes a homogeneous volume and the emitted x-rays from a defocused beam volume are *not* representative of actual bulk composition. In other words, the absorption correction is extremely non-linear.

For the poster attached below (please login to see attachments), Julie "synthesized" a defocused beam by averaging x-ray intensities for "strips" of pixels. She then applied matrix corrections to these averaged intensities to obtain a pseudo defocused beam "bulk" composition. These results were then compared to a simple averaging of the average compositions of each fully quantified pixels in these same "strips".

Edit by John:

Also it should be noted that once you have fully quantified maps, one can also simply draw arbitrary "profiles" and get accurate quant information as shown here using CalcImage:

https://probesoftware.com/smf/index.php?topic=1144.msg8000#msg8000