Ores are the hardest on the user stand point in the fact that as the veins differ so can the sample matrix making you have to adjust your method/calibration from time to time.

In terms of materials, Phosphines, extremely difficult and dangerous to handle. In terms of sample preparation, borate fusion Molybdenite, pre-oxidation process and precision because that is a challenge.

Sounds very interesting thank you for sharing. Would you be interested in sharing some more detail about the application ? Elements of interest, reason for the application, specific issues and solutions/ analytical approach.

The solution(s) too would depend on what type of XRF you are using.

Yes, there are some ores (like igneous phosphate ores) that are hard to calibrate. In these samples there can be a lot of Fe, REE and high content of phosphates, which also give you trouble when you have to choose which K or L-lines will be measured. Beside all that, Reference Materials are not available since these ores are highly variable in composition and type.

You raise an important point, especially for the analysis of geological materials. There are many many calibration standards of different matrices, but these are all secondary standards and the "official " concentrations are based on lots of analyses by different labs rather than being primary standards. While the data has gotten better over the years , the data quality varies by matrix, age of the standard and by element. There are still lots of reference materials with great major minor and some trace elements, but often stands have poor quality data for more exotic ( in a geological sense) trace elements and REE's. We should all remember that XRF is a comparative technique and the after sample prep, accuracy of the standards used for calibration is a the biggest source of error. Thus a great standards is not always equally great for all elements.

As someone who explored the non traditional uses of XRF I would say that the most challenging application was Fluoride in Toothpaste. I was surprised to learn the complexity of toothpastes, varying polymers and oils and of course TiO for abrasion. In the end it was the sample preparation which was the most challenging. Obtaining a flat and uniform sample meant some mixing and stabilization. This could be achieved in a sample cup. But then the X-ray film would absorb all the F counts. New SDD technology and advanced optical paths demonstrate F capabilities for samples with elevated F concentrations (E.g. Teflon). But I am afraid that the concentration levels, which range around 0.15% still present a challenge for EDXRF.

I am the analyst on a project trying to get accurate measurements on trace elements in ancient (2500 year old) silver and gold artefacts in museum collections around the world. The main difficulties are;

1. Not enough standards -> there is a lack of suitable metal standards, particularly for the concentration ranges of the elements I am looking at (e.g. Bi is particularly informative).
2. No preparation -> with artefacts being unique and therefore irreplaceable, there can be no subsampling or surface preparation. Not even at the micro-scale.
3. On site -> being so valuable, the artefacts are not permitted to leave the museums (this really limits equipment that can be used).
4. Transportability -> the instrument *has* to be transportable for 1-2 people. In one case the lift was broken in a particular ancient museum in Europe and we had to lift the spectrometer up multiple flights of stairs and through tight corridors to the work area.
5. Safety -> in the small enclosed spaces of museum collection rooms, helium isn't permitted (as their safety officers explain, helium isn't toxic but it displaces air, which is required for breathing!). Since constraints 4 & 5 preclude instruments with helium or heavy vacuum pumps, all measurements are in air. Also, museum radiation safety officers strongly prefer fully enclosed instruments and hand held XRFs would require a lot of extra hurdles.
6. Expense -> due to freight expense, I have borrowed spectrometers from local vendors in the host countries and demonstrating data comparability between instruments is a QA/QC challenge.
7. Power off -> in 4 out of the 5 museums so far, the spectrometer has to be turned off each night (fire risk, they argue) - which impacts on analytical quality.

This combination of constraints is particularly challenging and we spend a lot of time on QA/QC. We use 50 kV bench top EDXRF spectrometers, with long count times and the sample spinner to get the best quality measurements possible. We bought ten test artefacts (degraded and of little archaeological value) and performed a comparative study using WDXRF, 3DEDXRF, bench top EDXRF and grazing-incidence XRD before and after abrasion to help understand what elements are in the patinae of the museum collections but not the underlying metal. We then post-process those elements out (mainly Row 3 but a little of Row 4) and normalise back to 100%.

Previously, metals in plastics and polymers was the most challenging I have worked on. The light matrices of pure hydrocarbons change dramatically with the addition of colours, bulkers, stabilisers and mould releasers, so that real samples bear little resemblance to the calibration standards, particularly in terms of analytical depth. And I've never understood why some standards are supplied wafer-thin for such materials.

Yes, ores analysis is pretty difficult and good sample prep is must.  I have worked on REE (high and light) with HHXRF XRAY tube ,was pretty difficult to calibrate for few REE's. The anode slime powders (Pb) and to analyze PGM metals and precious metals.i worked on Rigaku edXRF. sample prep was very difficult for the same and mainly the standards for calibration. analyzing this samples has further matrix effect on SDD detector on PGM Re, Os, Ir, Pt, Ru, Rh, Pd/Au and Ag and heavily overlapped by the K-lines of Ni, Cu, Zn, Ga, Ge, As, Se and Br. done further development and analysis and sample prep.finally it was a good application and results were fine with customer as compared to other bench-top in the market.

Any light-matrix sample would be challenging especially if CRMs are not available for use with the well developed empirical-co-efficient method of quantification.
I think for light matrix samples Compton normalization can be effective. "Turbo-quant" software with Spectro-XEPOS (polarized with HOPG for lighter element analysis) is good. Don't know if the QA/QC requirements are met with ( from a commercial side).
We tried Raleigh/Compton ratio quite successfully with a non-commercial software and set-up...The success I think was heavily instrumental set-up dependent.
The Fundamental Parameters need to be updated covering more elements and with more accurate values in the library for XRF to become a potential primary method.

The most difficult application I have worked on to determine L X-rays of rare earths which coincide with K X-rays of Low z elements. I faced this difficulty while carrying out PIXE studies on tagged ink tagged with rare earths for a forensic study. Please refer to my article in IJPIXE. The most difficult application I worked on is determining the L X-rays of rare earth.which coincide with the K X-rays of low Z elements. I faced this while doing PIXE of rare earths in tagged ink tagged with rare earths for a forensic study. 

I have 2-3 projects where food and drug control unit has asked for XRF.  I have used hhxrf in certain industry to check heavy elements in leaves(dried/ash). They need to check Si and S may for pollution check i guess r may be some elements as the some leaves plants r used in biofuel and cement industry might be this element concentration gives a quality check instantaneously (but my not b good as to conventional methods). I have worked out earlier with one of r & d labs for plantation research(included soil and fertilizer effects on plants)Their purpose for getting hhxrf is for testing Soil/fertilizers in agricultural land as well study of leaves. Their main elements of interest are K, Ca, Fe, Zn, Ni, Cu, Co, Mo, Sn, Sb, Cr, Pb, Se, Au, W, Ti, Ag, Cd, Pd. They also want to check FERTILIZERS mainly Fe, Mn, Zn and Cu .The purpose is not to compete with ICP but to be complimentary.

OUCH! Your approach using XRF for these applications is not correct! Please step away from the idea, you may as well put the samples in your mouth and taste how much of these elements is present. Stick to ICP-MS or high power WD-XRF. The main elements you mention can be analyzed with XRF, but some, like Ag, Cd, Pd will, because of overlaps with the Rh tube lines and therefore the use of a filter, never be analyzable in concentrations lower than 10 - 20 ppm. So, what's the use of XRF in that case....? Sn, Sb will also better use a filter for the K-lines or you have to use the L-lines, meaning LLD go up. You see, HHXRF can never offer the flexibility and the intensities you need to see, as power is way too low and the resolution way too bad. Sorry, not even a complementary technique.

I would have to agree that you might be stretching HH XRF for this application despite the fact that detector resolution and light element performance on these instruments have improved a lot.
I would also be careful about comparing with ICP as there are often biases between the two techniques, often the result of sampling related to sample preparation. Issues such as incomplete dissolution , or inhomogeneity of a pressed pellet can come into play.
In my experience having used a wide range of instruments over the years is that your best performance would be with a polarized EDXRF with a secondary targets.
This type of instrument avoids some of the sample damaged that can be caused by high power WDXRF systems and gives you a lot of analytically flexibility for tuning the excitation and using Compton correction. None of the eleemnts you are looking at are particularly light and this type of XRF system works great on organic and light matrices such as particlaute air filters for obtaining some of the best performance particularly on heavy elements.

And therein lies the value proposition for HHXRF: by supplementing the lab analysis (via ICP, WDXRF, etc) with HHXRF, one may gather considerably more data points on sample sets, rather than relying exclusively on a more limited number of samples at the laboratory. Even if HHXRF lacks the DL's of these lab techniques, it frequently serves as a valuable screening tool. In addition to the soil application Megan references, RoHS and CPSIA compliance are common applications where multiple techniques are employed.

One of the biggest problems that seems to be overlooked here is that foods are extremely heterogeneous. Many elements are simply bound to different components of food (fats and proteins and starches). A great deal of effort must be put into sample prep. This becomes another science in its own. Grinding/blending these materials becomes and art and it is important to fix the composition so that the result is not segregation of materials which of course would then result in an over/under representation in an XRF analysis. As stated earlier the detection limits for most elements require substantial XRF investment. I have yet to see any HHXRF that could rival an ICP/MS for heavy metal analysis. However the 100 kV variety does have some use in the heavy element area as does the higher mA varieties for light element analysis.

Excellent point! If you you are not going to do some sample prep to homogenize the sample then you probably should take an average of a bunch of single spot analyses to come closer to a better overall composition. I am not sure if any of the hand held or small spot systems have this type of averaging capability in their software, but it would be a great thing to have. A running average mode.