DTSA-II for geological materials - 1 month of results

[orlandin]

Hi all, although there is still a ton left for me to learn about DTSA-II, EDS detectors, and x-ray microanalysis in general, I think that my personal experiment to turn my humble little JEOL 6490LV with an OI X-Max 50mm2 into a probe-lite has been fairly successful. I have attached a spreadsheet of measurements I've made on a variety of geological standards that have served my JEOL 8200 well over the years. Everything is quantified with measured O, even though I'd probably not do that for a real analysis unless there was some pressing reason. Livetimes are marked. All were collected by beam deflection, and most from free-hand rectangular box areas of wildly varying dimensions but usually >10 microns on a side. You'll have to log in to see and download the two Excel files, I think.

A few observations I'd like throw out:

- The standardization I made 32 days ago is still producing excellent accuracy.
- My EDS is FAR more forgiving of beam deflection than the WDS spectrometers. That makes sense but I have not quite tested it as well as I'd like. So far the evidence I have seen suggests that I can deflect the beam >200 microns from stage center without suffering any degradation in quant results.
- Single-line uncertainty has severely diminishing returns at around 40 seconds of collection time at 1nA, which is about a 20-23% dead time for my detector at process time 4 and roughly 18k cps input rate.
- When I propagate single-line uncertainty on a garnet measurement through the garnet-biotite thermometers found in a handy spreadsheet on Dave Waters webpage (https://www.earth.ox.ac.uk/~davewa/pt/th_tools.html), you end up with a 2-sigma error of 2-10 degrees C from >60s EDS spectra, which is WAY smaller than the disagreement between the thermometers themselves. I don't do much petrology so I can't quite judge the utility of that, but it sure seems precise enough to use that particular thermometer.
- The biggest problem I think I have is current drift. The W-filament in this LV instrument gets treated very poorly compared to the pampered existence of the 8200's W-filament, and I don't always succeed in preventing a bit of current drift while collecting spectra. Some of the linear trends with time you see in the data is likely due to current drift.

Please have a look and see what you think. I think that standardized EDS really has the potential to make a lot of the data produced by microprobes available to anyone with SEM+EDS, a way to measure their beam current, and the ability to borrow standards once every few weeks/months.

Best,

Phil

[Probeman]

Excellent work Phil.

I’ve seen that some EDS vendors utilize the zero energy (or strobe) peak to normalize changes in beam current. Is that something DTSA can handle?

Let me just say that I want to live in the world where every SEM EDS user is like Phil Orlandini- that is, they use standards!  😁

[Nicholas Ritchie]

Orlandin: Thanks for sharing your results.


Probeman: WRT zero strobe - I don't know how to use the zero strobe to normalize changes in beam current.  As I understand, the zero strobe is a stream of "0 V" pulses inserted into the pulse processor as though they were x-ray events.  This can be used 1) to stabilize the "offset" (so only the "gain" needs to be otherwise calibrated) and 2) to measure the electronic noise component of the resolution function.  In this model, the number of zero strobe pulses is just a function of the frequency with which they are inserted.  (My mental model could be wrong.)

[sem-geologist]

I just will add one more additional info about zero V peak.
additional to those functions stated before, I am sure that it is used (at least by Bruker) to record information of livetime.
Normally live time is provided in saved single spectra files, but in case of hyper-mapping (spectral image) there is no single dead time which could be applied for all acquired pixels. So live time is encoded by zero V peak at every spectra of every pixel. At Hyperspy we had tried to reverse engineer how that works, but we could not find satisfactory algorithm to get exactly the same values as reported by Esprit software. Maybe it is floating point rounding in Esprit, I don't know. I was inexperienced about pulse processing then I was RE that, now probably I could figure it out if that would be needed.
I should mention that I was looking for possibility to use zero peak width as mean to calculate broadening factor. I had collected spectras at different spectrometer settings on different instruments (with bruker EDS). Zero V peak is following general broadening trends (You get more broadening when setting spectrometer for higher throughput) but it is out of broadening factor curve (Mn Ka,F Ka, C Ka), So at least for Bruker detectors it should not be used as lower energy point for calculating theoretical peak widths.

[Probeman]

Probeman: WRT zero strobe - I don't know how to use the zero strobe to normalize changes in beam current.  As I understand, the zero strobe is a stream of "0 V" pulses inserted into the pulse processor as though they were x-ray events.  This can be used 1) to stabilize the "offset" (so only the "gain" needs to be otherwise calibrated) and 2) to measure the electronic noise component of the resolution function.  In this model, the number of zero strobe pulses is just a function of the frequency with which they are inserted.  (My mental model could be wrong.)

Yeah, that's what I've always assumed too.  It was a Bruker guy we were chatting with recently, because someone was asking if Bruker Esprit can read the beam current, and he was saying no they never utilize beam current for anything, and he added "The zero strobe peak is our proxy for probe current, if I recall."

I asked him to confirm this and will let you know what he says.