Hello Aris,

I am not very knowledgeable but happy to get the comments going.

• Would lowering the accelerating voltage (e.g. 10–15 kV) improve accuracy by reducing the interaction volume?

Yes - try it and see. The beam is sampling non-inclusion material at 20kV for your smallest inclusions. Even at 5kV you should see Niobium lines.

• What are the tradeoffs of lowering voltage (e.g. signal strength, resolution)?

CPS would decrease but more of the counts would be from the inclusion. You might need to "tell" the software that you have changed kV.

• Would reducing the working distance help? Are there tradeoffs there too?

No. Please use the standard WD to maximise counts. My Oxford Aztec system uses 10mm. Is yours optimised for 7mm WD?

• Is beam drift during EDS point analysis common, and how do people usually minimize it?

I compare the image before collection with one after collection to look for drift. Aztec has a drift correction function but I have not tried it. You could also (not point/spot analysis) look at very high magnification with a reduced area box and move the sample very slightly to get the highest V and Ni peaks

• Would mapping or linescans give more reliable results than point analysis?

My guess is no. They would both usually spend a shorter time per each pixel. I suppose this could be overcome by acquiring for longer times.

Good luck

Dave

Hi Aris

Can you please provide the technical details of your SEM EDS tool so that I may be able to help you more?

But in general, a few tips which can be followed are:

• Before reducing the accelerating voltage, you can start increasing the acquisition time so that we allow the detector more time to gather the signal.
• Please see if the WD 7mm is right for your detector. I have observed that for an SE2 detector, the working distance falls between 8 and 11mm (assuming that you are using an SE2 image).
• If you feel your sample is drifting, there is an option in the EDS software to lock the site of interest where you are performing the scan.

I have already recommended David Patton's answer.

Please simulate the interaction volume at various voltages and see what it takes to keep the beam within the desired depth. I like CASINO from the University of Sherbrooke. (I limit myself to version 2.48; version 3 was too hard to figure out.) Remember, you are working with a polished section, and you probably did not section the inclusion exactly at its equator. It could be thinner.

For EDS, you would like to use twice the energy of the x-ray line for Vo. You could do Nb analyses at 5 kV. However, the V-L line will be hard to use. It overlaps with the O-K line. I would probably try 10 kV.

For working distance, use what your EDS system recommends. Ours is aimed at 10 mm working distance. If we raise the sample to 7 mm, we cut off the EDS signal due to misalignment.

It is good to probe larger features as well as smaller ones. It is also good to probe the matrix so you know when it is contributing. Then when you probe various points, you can tell which ones are more influenced by the neighborhood and they can be rejected as outliers.

Maps could be good to see if all your inclusions are the same composition. Many times they are not. However, I would not even think of using them for a quantitative analysis.

Linescans can be good for seeing if you ever got up to a stable composition in the middle of an inclusion. You expect the composition to change gradually at the edges, but you should see the composition plateau when you are away from edge effects. I expect, you will never see a plateau at 20 kV.

I do wonder about the results in your second attachment.

Figures 2, 6, and 9 do not report Mn in the EDS table, but it includes C. Why? Mn is definitely present. You may have to force its inclusion. I would not trust the C analysis.

Figure 8 shows Mo in the spectrum, but S in the table. Which is it? I'm inclined to think it is really S, perhaps in the form of a MnS inclusion.