I would press some pellets using the binder only and take the average carbon content of three pellets as the carbon content, otherwise I would do a carbon sulphur determination using a Leco or something similar on the refectory material so any contamination from the binder could be negated. If this work is for long term analysis I would do both and compare the results to check accuracy.

For very light elements like carbon you are only getting a surface analysis. You need to be sure the same is really homogeneous and flat. The wax will undoubtedly add a blank to the analysis, but if the calibration standards are prepared the same way and the carbon concentration in the unknowns is high enough if might work OK. You should be able to calculate the blank form the intercept of the calibration curve generated from standards prepared with the wax. I am assuming you are doing this by WDXRF so you will also need a good multilayer crystal to get the best sensitivity which is essential in overcoming the blank.

I would advise to use boric acid as a binder (no C content), take a coarse collimator and a decent Carbon analyzer crystal and there you should be able to detect carbon with 200 - 300 ppm precisely.

Thank you all, I have tried the boric acid but its binding ability is not so good. We have a WD XRF of Axios max. We have PX4A crystal of d=12nm and coarse collimator of 4000. No question is that how to calculate the blank form the intercept of the calibration curve generated from standards prepared with the wax?

Did you try this with finely ground boric acid? If this analysis is a real must with XRF (which I do not really advise) then next time you should maybe consider buying a tube above configuration with thin (30µm Be) X-ray tube window; like Rigaku's ZSX Primus II. It would give you about 40% more intensity for the C.

Just to say that you can buy tools for your pellet preparation that will allow you to press 100% of the sample onto a backing of say fine boric acid. This means you get the benefits of the binder without added contamination. When you say you want to calculate the blank from the intercept do you mean standard additions?

Some good advice has already been given. Adding a compound, containing an analyte element, to the sample preparation, will increase the 'equivalent background '-D' of the calibration curve. This increased value for -D ('intercept') will automatically correct for the fixed compound added, but:

1. The increased value for -D will lead to a larger counting statistical error (CSE) en must be established.
2. Any variation in the weight and viewed surface of the added compound (e.g wax in the case of C) produces an equivalent (systematic) error in the analysis result.

Not surprisingly, people try to avoid these problems, by using alternative backings/binders like Boric Acid. It will take some experimentation to establish the optimal sample prep.

I can't accept your idea so easily as the added carbon signal, due to the addition of wax, will be that large, compared to the actual carbon-in-the-sample-signal, that the signal will probably be difficult to subtract and still have good analysis. I do support the idea of pressing a smaller amount of sample directly in a layer of boric acid, at least for the analysis of carbon. That should work.

Putting the powder on a boric acid pellet is a good method that we've used. Another variant suggested is to make a solution of boric acid and use it to wet the loose powder, dry it down and then press. It may be that just mechanically mixing it in isn't giving you good adhesion where coating the particles might work better. I agree that using XRF to measure carbon is pretty sketchy, but sometimes all you have is a hammer and you really do need to make your sample look like a nail. Good luck!

The devil is again in the CSE, which is e.(ROOT)r/t for a calibration curve C=e.r+D

Adding wax will leave 'e' more ore less intact but might change '-D' dramatically e.g from 0.1 wt% to 10 wt% C. Obviously now also the gross countrate for C increases by ~ 100, making the CSE ~10X larger. This is what was meant in my first point.

Another way of saying the same thing: The very high countrate from the added wax, will swamp the originally small signal.

Some people may think, that they can solve the high background problem by including a separate background measurement, However this makes the already high CSE 2X larger. The alternative then is to use a TOTAL counting time for C that is 4X longer compared to a single peak measurement.

A separate background measurement is only wise if the increase in CSE is compensated for by a smaller systematic error.

Thanks a lot, all of you. It was very interesting and knowledgeable for me. I have tried Boric acid fine powder it gives better results.. Thank you so much again..