I would like to do some Quant analysis of Carbon chars. Optical analysis shows they are a mixture of initial chars (fusinite, semi-fusinite) but EPMA has shown some sulphur, chlorine and even oxygen enrichment. I wondered what I might use for a standard for S, Cl & O in a C matrix - using the elemental standards I have would presumably have a huge MAC difference. But I don't know this for sure.
Hi Les,
Here's my take on trace standards in general, and of course that would apply to trace S, CL and O standards as well.
The take away message is simply: it doesn't matter, really.
Here's an example: measuring trace S in something. You're right, there is a difference in the mass absorption coefficients (MACs) between the pure element and any compound of that element. But how large is that difference? With regard to our example of trace sulfur in something, here are some matrix corrections (ZCOR) in a number of compounds:
St 730 Pyrite UC # 21334
TakeOff = 40.0 KiloVolt = 15.0
ELEM: Fe S Ni Co Cu As Zn Pb Cr Ti
XRAY: ka ka ka ka ka la ka ma ka ka
ELWT: 46.550 53.450 .000 .000 .010 .009 .009 .010 .000 .058
KFAC: .4278 .5061 .0000 .0000 .0001 .0001 .0001 .0001 .0000 .0005
ZCOR: 1.0881 1.0560 1.1017 1.1090 1.1513 1.6740 1.1478 1.1278 1.0993 1.0829
St 707 Chalcopyrite U.C. #1232
TakeOff = 40.0 KiloVolt = 15.0
ELEM: Cu Fe S
XRAY: ka ka ka
ELWT: 34.620 30.430 34.950
KFAC: .3152 .3027 .3151
ZCOR: 1.0984 1.0052 1.1093
St 710 Sphalerite, Bingham
TakeOff = 40.0 KiloVolt = 15.0
ELEM: Zn S Fe Pb Cd
XRAY: ka ka ka ma la
ELWT: 67.000 32.920 .110 .170 .500
KFAC: .6285 .2805 .0011 .0014 .0039
ZCOR: 1.0661 1.1737 .9575 1.2455 1.2714
So let's assume that the matrix correction magnitude of S Ka in common materials is around 20% and that the accuracy of this matrix correction is around 10% relative accuracy (at worst). So 10% of 20% is 2%. That means our analytical accuracy for S Ka in any number of materials will be around 2%- for major elements! So yes, "matrix matching" of standards might sometimes be worthwhile for high precision measurements of these elements if they are present in high concentrations.
What about trace elements? Well in our example above, whatever the analytical accuracy of our system (S ka in something), that will remain roughly the same (2%), as the concentration decreases. But the counting statistics will reciprocally worsen at a rate of Sqr(2) per halving of the intensity/concentration, that is until the precision of the signal is no worse than the precision of our background characterization.
At some point, the accuracy of our example system matrix correction (2%) is small compared to the precision of a typical trace element acquisition (10-30%). Instead, the accuracy of the background correction now becomes the dominant source of error (interfering secondary emission lines, absorption edges, and other continuum artifacts which become more significant as the signal decreases in intensity).
Moral of this story? If you are measuring trace elements of moderate energy emission lines, in most common matrices, the accuracy of the background correction becomes the dominant source of error as opposed to the matrix corrections.
Caveat: low element emission lines (e.g., oxygen Ka), can have such extreme matrix correction effects (100% or more), that even trace element intensities can be affected significantly. Of course for these low Z emission lines, the background intensities are also much higher and hence the backgrounds have correspondingly worse precisions!
Bottom line: I would just do a quick check in CalcZAF for the matrix corrections (ZCOR) of the standards you have, with a significant concentration of the element of interest (the higher the intensity per concentration the better for your statistics anyway), relative to the pure element and select something with a reasonable matrix correction such as any of the three standards I discussed above.
The point is: being able to quantify these sources of error allows the analyst to make choices that improve the quality of our analyses. Here is an example of using CalcZAF to make related judgements:
http://probesoftware.com/smf/index.php?topic=121.msg490#msg490And the full Penepma Monte-Carlo modeling GUI is nicely implemented in Standard.exe for answers to all sorts of "what if" questions... but be forewarned, just as it takes time to make sensitive measurements- it takes even longer using Monte-Carlo methods!
I will end with this: for ultimate trace elements accuracy, we can hope for specimens where "blank" standards (with a similar matrix and known zero or non-zero levels) are available:
http://probesoftware.com/smf/index.php?topic=29.0In your case of S, Cl and O in carbon, you are in luck! Get thee a polished diamond "facet" and mount it (and carbon coat it). It will have close to zero oxygen, chlorine and sulfur and can therefore be used as a quantitative "blank" standard as just described.
or we can utilize one of the advanced off-peak characterization methods pioneered by Mike Jercinovic, Julien Allaz and Kartsen Goemann such "multi-point" backgrounds:
http://probesoftware.com/smf/index.php?topic=131.0and, alternatively, "shared" background measurements:
http://probesoftware.com/smf/index.php?topic=9.0