Author Topic: detection limit CAMECA  (Read 322 times)

stanroozen

  • Student
  • *
  • Posts: 1
detection limit CAMECA
« on: May 01, 2019, 09:05:11 am »
Hello all,
I am Stan Roozen, Phd student in geochemistry at the McGill university. I am using a Cameca SXFive electron microprobe to measure major elements of the mineral tourmaline. I am looking at the forum as i am trying to find out how the detection limits are calculated in the CAMECA software. Two procedures are named: Ancey and Goldstein,  which give very different answers for detection limit and i am currently unsure which one is the correct one to use. This is especially important for F as this one is close to the detection limit.

Greetings, Stan Roozen

Mike Jercinovic

  • Professor
  • ****
  • Posts: 67
    • UMass Geosciences Microprobe-SEM Facility
Re: detection limit CAMECA
« Reply #1 on: May 01, 2019, 03:06:17 pm »
Hi Stan, which version of Peaksight is running there?  I don't see the equations in the 6.x users guide, but I think you should still have the initial version of the SXfive users guide somewhere there.  The equations relevant to the Ancey method are given in the guide to Results.  I have a pdf of the original paper from the famous St-Martin-d'Heres Summer School on Microanalysis and Scanning Electron Microscopy in 1978 if you need it.  We also outlined the basic things in our IOP paper ( Jercinovic, M.J., Williams, M.L., Allaz, M.J., and Donovan, J.J. (2012) Trace analysis in EPMA. IOP Conference Series: Materials Science and Engineering 32 (J.L. Labar, C.T. Walker, F. Brisset, O. Dugne, and F. Robaut eds.). Institute of Physics Publishing, Bristol, UK.  Doi:10.1088/1757-899X/32/1/012012. P.1-22).

I am not sure of the exact Goldstein equation Cameca might be using now, but I would presume it is the one used in Goldstein (1967) which is also outlined in the Goldstein et al textbook (editions 1-3, not v4!).  You can also find the Ziebold approximation in that book also, which gives a result pretty similar to the Goldstein calculation.  Might check on that implementation with Carl Henderson at Ametek. 

Anyway, a lot of this depends importantly on the confidence level you need (to get the student's T value) and the number of measurements involved in the test.  We like the statistics of the Ancey approach overall.  Remember you are just manipulating X-ray counting statistics based on Poisson stats and relating that to concentration.  The question of detectability is somewhat subjective (i.e. 3x the standard deviation of the background).  This means the concept of "correct" is not so straightforward.  Detection limit sounds like it should be a hard number, but it is really just a statistical estimate from a confidence envelope that makes no presumption of accuracy.  When does a signal emerge from the noise?  You can say probably this value or that value depending on your assumptions and how you define that magical signal level.  I am just making this worse now, sorry.  One thing for sure, these sorts of measurements imply that, as your peak intensity approaches the background intensity, the precision on the background measurement needs to be excellent.  As you will see in the equations, the calculation of detection limit depends mostly on the precision of the background.
« Last Edit: May 01, 2019, 07:23:00 pm by John Donovan »

Probeman

  • Emeritus
  • *****
  • Posts: 1945
  • Never sleeps...
    • John Donovan
Re: detection limit CAMECA
« Reply #2 on: May 02, 2019, 08:41:07 am »
Hello all,
I am Stan Roozen, Phd student in geochemistry at the McGill university. I am using a Cameca SXFive electron microprobe to measure major elements of the mineral tourmaline. I am looking at the forum as i am trying to find out how the detection limits are calculated in the CAMECA software. Two procedures are named: Ancey and Goldstein,  which give very different answers for detection limit and i am currently unsure which one is the correct one to use. This is especially important for F as this one is close to the detection limit.

Greetings, Stan Roozen

Hi Stan,
If you want to compare the Cameca detection limits to the Goldstein calculations you can also try the detection limit calculations in the free CalcZAF utility as described here:

https://probesoftware.com/smf/index.php?topic=121.0

It can be downloaded here:

https://probesoftware.com/Technical.html
« Last Edit: May 02, 2019, 08:44:46 am by Probeman »
The only stupid question is the one not asked!