Analyzing Fluor directly in a pressed powder sample will probably not be accurate enough as the mineralogical effects will be very significant. Fusion is possible, but the fusion device must be very reproducible and the recipe must be verified for the whole range of samples that you want to cover before using it as a reference method. I noticed that dependent on the sample composition, the sample preparation may be less reproducible where for the first samples it may have worked well, if you do not check your recipe well. Mind that temperature of the fusion should be moderate as F will otherwise escape. So in the beginning I advice to have both methods run in parallel. Still you will need a high power WD-XRF in order to be able to analyze at those levels. I wish you success with the method. Maybe it is good for the discussion to mention what time it currently takes to get a result, as well as the accuracy that you are looking for.

You will have more chance in analysing Fluorine in a pressed pellet than in a fused bead, like my well-appreciated colleague Armand stated. Especially as you only want to add 0,25% of CaF2 to your cement. I guess that, with a higher power WD-XRF, you may obtain a precision of +/- 0,01% (with a fused bead - if one doesn't loose Fluorine at preparation - it would be a factor 10 - 15 less good). The X-ray spectrometer you currently use will not be that appropriate for the Analysis of F (simultaneous = less sensitivity and also, is the channel in there?)

I have an application note about testing of F in powder form by use of EDXRF. Ability to detect less than 1% concentration. Detection limit of 500ppm. Includes reproducibility.

Fluorine has Kα of 0.677 keV. Does not need high powered XRF. High powered X-Ray will activate lots of heavier elements potentially interfering with low signal of F. I would go for low power and high current to increasing the cps.

I am analyzing Fluorspar by using pressed pellet technique. Its results are very close to the actual one (As i validated by other tehniques). The WD XRF is a good choice to analyse the fluorine even up to ppm level. Mininmum kV and Maximum mA stimulated the Fluorine Kα easily. I am well satisfied by this technique and hope so for you also.

I think you are misinterpreting 'high power' for 'high kV', as power is expressed in Watt (kV*mA). If you have high current (mA), you will automatically arrive at high power. Though indeed you will not need high kV for exciting F, X-ray tubes/generators are limited by the maximum current that can be delivered at a given kV setting. Where EDS systems are not very sensitive for low energies, WDS systems have to be used. Since efficiency of excitation is low for F, you will need high power. Efficient excitation can be reached with 5* the absorption edge energy, which would be less than 5kV here. But the minimum kV a high power system can be run at is usually in the range of 20 - 30 kV, for a 4 kW system this would mean a current 200mA - 133mA, which may exceed the maximum current of the system. In that case one will increase the kV until the maximum power can be used. This is what is done in sequential systems, but for a simultaneous spectrometer this is not practical, a fixed kV must be used that excites all elements, usually 50-60kV.

I mentioned that pressed powder analysis suffers from severe mineralogical effects. In a simultaneous system where one cannot measure the background this will disturb measurement even more, so I think fusion is the only way, unless the measurement range is limited and one can use in-type calibration to circumvent mineralogical effects. But this is hardly ever the case, certainly if materials come from different sources, or a are mixtures of materials from different sources.

F in Cellulose in percentage level is another story than a true cement sample + 0,25%CaF2 (more absorption, etc.). When I look at the scans you provide in the application report and I see a net Peak of 2cps, then I guess accurate analysis of 1000ppm F samples in a heavier matrix doesn't belong to the possibilities of ED-XRF! We all know lab reports very often show "idealised" results. We, as systems providers, should try to keep XRF analysis a bit practical oriented.

Thanks for pointing out. Power measured in W and Voltage measured in V.
Yes Ca definitely has higher matrix effect than Cellulose. Agreed, WDXRF will be definitely be a better instrument for this light elements like F.

As it has already been mentioned, borate fusion is possible for the determination of CaF2 in your samples. The difficulty of this sample preparation resides in the fact that the fluoride is easily lost under heating. Thus, a low fusion temperature and short fusion time are necessary to achieve good measurements. Claisse, has been able to prepare an application and fusion program on our instruments to retain highly volatile elements. 

We routinely analyse F in samples using a borate fusion. It’s the only halide that for practical purposes can be retained in the borate glass quantitatively. From a practical point of view the limit of detection in our laboratory is 200-400ppm in the glass. Taking into account dilution when preparing a bead, this may be a significant problem for your application. Problems do occur with drift and the analysis is highly sensitive to variables such as spectrometer vacuum. It may not be a practical approach for you. For some of our applications for example the analysis of fluoride salts where the concentration is high the glass bead fusion with XRF analysis has been the most reliable and accurate method. For fluoride at low concentration you could consider performing a sodium peroxide fusion, dissolving the fusion mixture in water, diluting with TISAB and then determining the fluoride concentration using an ion selective electrode. I have no specific experience on your sample type for the approach though. You will need to use a nickel crucible. Fluoride reacts quite readily with ziconium.

Pressed pellets is definitely the correct approach for this type of material. grain size will be important for the light elements. I would suggest that you grind to 50um or less. I would go for making pellets 40mm in diameter if your XRF has a sample mask wide enough to take advantage of the larger diameter( eg. 35-37mm). This will provide the best signal for the lightest elements like F.

-of sample and 3g of a wax-cellulose binder pressed at 20-30T will make a nice robust pellet. getting a fine and consistent grind for the powder is likely to be the hardest part. Not all mills can deliver that level of consistency for grinding. You may also want to consider an automated mill and press for the best and most consistent analytical results.

Fluorine analysis is significantly effected by particle size ( shadow effects) effects because the sampling depth is so small. The grain size of the powder you use to make the pellets needs to be less than <50um. Then you can make a robust pellet 40mm diameter pellet using 12g of your sample and 3g of binder. These should be mixed well, preferably in a mill or shaker and then pressed at 25-30T for 2-3 minutes and with a pressure release over abut 1 minute. You can scale down the sample and binder proportionally if you are using 32mm pellets. I would use a cellulose-wax binder, but would measure the binder as a blank as some waxes contain Fluorine. There are many ways to make pressed pellets but this one should work for you.

<40uM particle size, 20+tons of pressure, minimum binder at 5-20%, thickness critical to escape depth ie F is only a few uM depth.

The X-ray sampling depth for Fluorine. While the primary X-ray beam can penetrate deeply in to the sample and generate secondary X-rays of all the elements of the sample, the X-rays can be reabsorbed by the sample before they can travel back out of the sample to the detector. The lower the atomic number of the element, the longer the wavelength of the X-ray and the easier it is re-absorbed by the sample. The sample depth at which an X-ray for a given element will make it out of the sample is called the critical escape depth. It is different for different elements. The escape depth for low atomic number elements such as Fluorine is only a few 10's of microns. This is why it is so critical that the sample be extremely homogeneous when measuring these elements. Having a sample with a consistently small particle size, that is well compacted and has a very flat surface is really important.

The 'nugget effect' can be critical in geological samples. The sample for XRF needs to be both representative and of a particle size smaller than the critical escape depth of the F X-rays. Taking lots of samples won't help much as there will be a self-absorption bias on the results. F is quite a challenge for XRF.