Evaluate Application

[John Donovan]

I should also point out that the Evaluate program which is described here:

http://probesoftware.com/smf/index.php?topic=71.msg400#msg400

and which is ostensibly for evaluating standards can also be used to quickly plot a matrix of variation plots to look for intensity or compositional trends as seen here:



simply by clicking the Window | Variation Plot menu in the Evaluate.exe main window. To view any of the variation trends in more detail, simply click the graph in question to see it enlarged:



One can plot standard or unknown intensities or compositions and multiple selection is allowed as seen here:



and again, zoom on any one of the plots as seen here:

[John Donovan]

I've recently improved the variation (correlation) plotting feature in the Evaluate application a bit more as seen here in a REE run:



And here is one of the correlation plots (when clicked on):

[Probeman]

If there's one thing we can be sure of in this world, it's that pure synthetic quartz is one atom of Si to two atoms of O (SiO2). That yields a theoretical composition of Si 46.74 wt%  O 53.26 wt%.

Now in the case of our synthetic quartz, I have the following composition noted in our standard database:

St   14 SiO2 synthetic
TakeOff = 40.0  KiloVolt = 15.0  Density =  2.650  Type = oxide  Mount = alkali, hydrous, carbonate, sil-sul, misc

Specimen from ESPI, 99.99%, EPMA (UCB): Al2O3 ~ 0.01%
Catalog #K4699M
ICP-MS (Alan Konig, PPM) Ti=1.42, Li=3.18, Cu=1.26, Mn=2.93, Pb=0.11, Zn=2.09, K=0.03
Flame AA (Chris Lewis, PPM):
Al=15 +/- 5
Fe=6 +/- 3
Mn=1.5 +/- 0.3
Na=5 +/- 3
Li= 2.3 +/-
Oxide and Elemental Composition

Average Total Oxygen:       53.255     Average Total Weight%:  100.000
Average Calculated Oxygen:  53.258     Average Atomic Number:   10.805
Average Excess Oxygen:       -.003     Average Atomic Weight:   20.029

ELEM:     SiO2   Al2O3       O
XRAY:      ka      ka      ka
OXWT:   99.994    .009   -.003
ELWT:   46.740    .005  53.255
KFAC:    .4101   .0000   .2664
ZCOR:   1.1397  1.2239  1.9993
AT% :   33.332    .004  66.665
24 O:   12.000    .001  24.000

Because we have around a few tens of PPM of Al in the matrix (from hydrothermal growth) depending on the technique, it's not exactly stoichiometric.  But as to the Si content we can have pretty darn good accuracy just by assuming single crystal stoichiometry.

Now I'm not claiming these are the best analyses that could be performed, as I was in a rush trying to get some results before our instrument undergoes a week of maintenance and the beam was not as stable as it should be. 

And also I'm not claiming these synthetic silicates are perfectly accurate compositions, as I've simply assumed formula stoichiometry on them, though the NIST glasses should be reasonably accurate. The natural end member diopside is also probably reasonable accurate from assuming stoichiometry, but does have some small tremolite inclusions... hey, it's a natural material!

Now because these analyses were performed on an LPET crystal at 30 nA, we had lots of counts coming into the detector, so when extrapolating from these high count rates in SiO2 to lower Si concentrations in silicates and glasses, the dead time calibrations really do matter. But in any event, assuming our dead time calibrations are reasonably accurate (that is, up to date!), we might ask: do our modern matrix corrections really allow us to extrapolate from pure SiO2 to various silicates and glasses?   

The matrix corrections in these various secondary standards ranges from ~20 to 30%. Here is a screen shot of the Evaluate application showing a number of silicates and glass standards using SiO2 as the primary standard:



What do you think?  Here's the key:

263 = synthetic Fe2SiO4 (Oak Ridge, Boatner)
275 = synthetic Mn2SiO4 (Oak Ridge, Boatner)
273 = synthetic Mg2SiO4 (Inst. Solid State Physics, H. Takei)
160 = NIST K-412 mineral glass
162 = NIST K-411 mineral glass
215 = BIR-1G basaltic glass (USGS)
358 = natural end member diopside, 600 PPM Fe (Chesterman)

In fact all these secondary standards show 2% or better relative accuracy when using SiO2 as a primary standard.  I think this data bodes well for our high purity synthetic standards global efforts as described here:

https://probesoftware.com/smf/index.php?topic=1415.0

Do you have the Probe for EPMA software?  It might be instructive to run a bunch of your silicate and oxide standards and calculate the results in the Evaluate application. It won't tell you whether it's your dead time calibrations, or your standard compositions, that are inaccurate, but the first one is easily remedied:

https://probesoftware.com/smf/index.php?topic=1160.0

and the latter we're working on:

https://probesoftware.com/smf/index.php?topic=1442.0

Please join us in our efforts to place microanalysis on a more sound footing by joining the FIGMAS group and/or contacting Will Nachlas at the University of Wisconsin to see how you can participate in these efforts. 

Oh, and please run some dead time calibrations on your instrument!  Your analyses will be better for it no matter what standards you utilize.

[Probeman]

I'm posting these images to make a general point about how much more sensitive it is to plot deviations are when they are plotted on a horizontal slope, rather than a diagonal slope (yes I will be making a link to this post from the constant k-ratio topic next!     ).
 
So here's a plot from the Evaluate application plotting up a number of Fe standard using pure Fe metal as a primary standard:



This is the data from the discussion in this thread:

https://probesoftware.com/smf/index.php?topic=1423.0

Although one can discern that the magnetite standard (#395) is plotting slightly below the general slope of the line, the deviations in the other standards are not at all very clear.  So now let's plot the same points, but using the Plot Relative Difference checkbox option:



Now the magnetite deviation is very easily observed, and even the other even smaller deviations of the other Fe oxide standards are visible:

#160    NIST K-412 mineral glass
#162    NIST K-411 mineral glass
#396    Chromite (U.C. 523-9)

While we can see that the pyrite standard (#730) is actually quite close to it's expected value.  This is the reason that the default mode of the Evaluate application is to plot these relative differences. 

It is also worth noting that these errors for the oxide standards (plotted above in absolute weight percents) fit fairly close on a line as shown here:



Which indicates to me that these errors are related to the choice of the pure Fe metal standard.

[Joe Boesenberg]

John

In addition to the negative trend, it also appears there is a positive trend running from 162 through 730 to 526.  Also due to the choice of Fe metal as the standard??     Joe

[Probeman]

I don't think so. Note, the y axis is plotted in weight percent absolute error, so standards with smaller concentrations of Fe will trend towards zero error. 

And as far as the value for 396 (chromite) being slightly below the trend I drew, I suspect that is just a little error in the Fe "published" value for the chromite. To be perfectly honest, the Fe value in that chromite is based on a bunch of probe analyses I did 30 years ago.  So it's not really a very significant measurement. I just threw it into the mix because it was an Fe standard in the same mount.

Note that the pyrite analysis errors come out almost exactly as expected (close to zero). See the analysis output of the pyrite analyses in the post above.

What is significant is that the relative error for the Fe oxide and NIST glasses are all around 2 to 3% relative. That is what is interesting and makes me suspect this is some sort of subtle peak shift/shape issue for extrapolating metal to oxides.