I am using XRF (Rigaku NEX-CG) to quantify nutritional elements in complex animal feed materials. In theory the Empirical approach to calibration (preparing your own standards by spiking a blank matrix) should be more accurate than the Matching Library approach (using wet chemistry data from real samples to adjust the sensitivity of the on-board XRF calibration). In your experience is this the case in practice? Is an Empirical calibration actually more accurate or not?
Secondly, has anyone had success in creating a single Matching Library for a range of different sample matrices? How have you defined the range of samples matrices which are included (similar ash contents? similar major constituents...)? How similar do these matrices need to be to fit within the one method? Any thoughts gratefully received.
Hi Kevin,
we all know that quantitatative XRF is a challenging task.
To my opinion the empirical approach to calibration by preparing ones own known reference is the more trustfull procedure. In the case you have carefully prepared the reference with respect to geometry and x-ray attenuation & scatter you then know what you have measured. In the case of the 'matching library approach' you do not really know what is going on there and how much the library entries match to your sample system.
With respect to the second part of your question I have got no experience, sorry.
Hi Kevin, a further idea:
please perform your quatitatative XRF measurement via your 'home-made' reference and then compare the result with the matching library approach. So you will see whether you can trust the library approach of your machine for the case of your sample system.
In the case that the results significantly differ I would trust more my results based on the own calibration procedure, as I stated above.
The empirical approach to calibration by preparing ones own known reference is in general the more trustful procedure. But this is very costly in terms of labor.
For the empirical approach to calibration you should reach a high quality for your prepared standards especially with regard to homogeneity. This is not as easy as it sounds. There are other problems to solve like what is an adequate blank matrix for all kinds of complex animal feed materials. Furthermore for the empirical approach you should perform multi-element calibrations at least for all elements that can influence your target elements.This need a large number of standard samples.
Therefore it is often the case that the results of the built-in Library are more accurate then your 'home-made' calibration. The idea of Gerhard Martens to compare your 'home-made' results with the matching library approach is always a good approach.
We have tried in the past to build up a Matching Library for a range of different sample matrices (especially food and plastics). We ended up in using the built-in Library for all kinds of sample matrices and made additionally calibrations for 'single matrices' (like paper, PET, PVC, Polyolefines, food with high starch content, vegetable oil). This may depend also on your kind of software. For our Bruker S8 Tiger and X-Lab 2000 this works very well.
Hi Kevin,
Using known standards (in this case those you have prepared yourself and know the content) is generally the best method. It is how we are taught to do XRF and other related fields So...why would some other method seem more accurate. I see four potential flaws. 1) Mixing dry ingredients, or wet with dry, can sometimes be challenging. The dependent variable is often the mixing. When is the material really evenly mixed, and is that even possible. 2) In foods we find that some ingredients have a natural affinity for each other. Getting those ingredients separated and evenly dispersed can be a real bear. 3) In feed it may be possible for settling to occur. Fine particles tend to settle to the bottom, allowing larger particles to float to the top. This can happen at any point in the process, given a little jostling. 4) Under certain conditions sampling itself can be an issue, if the sample size is big enough (whatever that is for a given material) then it becomes a good test, but if it is too small (and XRF tends to favor smaller sample size), the test may not be representative. We are repeatedly running into these issues in the food industry. Too, foods do not often behave exactly the way me might predict (meaning - foods and more typical materials, though often similar, are not exactly alike). If any of these situations exists within your “materials,” this may explain your results. If such things cannot be corrected you might try this: To get around these faults, I often do multiple samplings from various locations within the total lot, then average them together. The average tends to be more accurate that any given single sampling (your standard, as well as your product).
Thank you gentlemen for all your replies! It is fascinating to hear others' experiences. Making XRF truly quantitative for heterogeneous samples is indeed a real challenge. Consistent over-estimation of certain elements by XRF has been an issue for me, even with my Empirical, "home-made" standards with known concentrations. Nevertheless, XRF has proved a useful rapid screening technique for certain animal feed products. Many thanks.
Hi Kevin,
Some input from Alessandro and Janos in the Nuclear Science and Instrumentation Lab at Seibersdorf...
"Actually there is not straightforward answer. It depends mainly on the instrument and on the quantification software that has been developed by the manufacturer (we do not have this instrument in our lab). The best way to test which method performs better would be to analyse certified reference materials of their application field and compare the results achieved to the ones certified."
"
- Badges
- Science topic
- Similar topics
- Agricultural Science
- Environmental Science