EDS Detection Limits

[DavidAdams]

On my new 8530F+ I've been playing around with combined WDS/EDS quantitative analyses and have been relatively happy so far with the results using the Thermo EDS. One thing that I would like is to have the limits of detection for the EDS elements available in Probe for EPMA just like it is for WDS analyses.

I've been chatting with John and Steve Seddio (@Thermo) about this and am posting it here for broader discussion to see if others might have an idea of how this might be accomplished.

Edit by John: Dave, I removed your pasting of Steve's response because the images are not visible in line. Let me work on that...

[John Donovan]

One relevant point I can make, because I had discussed EDS detection limits previously with Keith Thompson before he left Thermo, is that if one is using the Bruker EDS interface, with Probe for EPMA for integrated WDS and EDS point quantification, Bruker does return the background intensities (needed for calculating detection limits), along with the net intensities from the EDS spectrum.  However Thermo does not.

My recollection, from what Keith told me, is that Bruker can easily return the background intensities along with the net intensities because they construct a full model of the continuum. At that point, it's simply a subtraction problem.  Therefore, based on whatever the current ROI is for the element in question, Bruker can return the net intensities and also the background intensity. However, because Thermo is using a "filter fit" method for obtaining the net intensities, calculating the background intensities is more difficult.

That is about as much as I know so I'm going to wait for some experts to weigh in on this question.

[emma_fisi]

Excellent timing! I was just about to post on the forum about exactly this same topic...

[SteveSeddio]

My questions on the topic were: What is the appropriate "background" to use for EDS CDL's?

Is it the intensity of a single channel at the X-ray energy, the summed intensity of all channels in the processed region of interest for an element's peak(s), or the average intensity of all channels in the processed region of interest for an element's peak(s)?

In the attached Fe K example, the top hat filter is processing the spectrum between 5.6 and 7.75 keV. Which channels do we care about?



Also, what is the correct way to calculate a CDL in the presence of interferences?

From a spectrum of xenotime (YPO4+REE buddies), I have removed the Y L counts used for quantification (xenotime1, attached). Is that what should be used for the Y detection limit?



Or would it be more appropriate to use just the continuum with all peaks removed (xenotime2, attached)?



Then I did some light reading. Goldstein et al. 2017 (starting at Page 341) and Newbury and Ritchie (2015) have much to say on the topic. They seem to agree that the answer is using just the continuum. I'm still not entirely sure what energy range should be used, though.

Edit by John: Steve I put your attached images "in-line" for easier reading.