Maximizing Spatial Resolution

[Michael Lance]

Hello,

I work a lot with commercial Ni and Fe-base alloys and there is a demand for high spatial resolution in order to quantify sub-micron phases. I have a JEOL 8200 and a W filament. I typically use 10 kV and beam scan to do this.

What are the forum's thoughts regarding the best beam current and pixel dwell time needed to optimize spatial resolution?

Thanks,

Michael

[jon_wade]

Hi Michael

what lines are you analysing?  I ask as there may not be a massive benefit in terms of interaction volume going from 12 to 10kV (can't get casino to run right now) and it may help with beam focus.  running LN2 will definitely help mitigate carbon build up/dwell time issues

[Michael Lance]

I use the L-lines for Fe and Ni at 10 kV. For Mn and below, I use the K-lines.

[JonF]

Running the numbers through Casino, Fe Ka and Ni Ka at 12kV is ~ the same interaction volume as Fe La and Ni La at 10kV and neatly side-steps a lot of the issues with using the L emission lines of the 3d metals whilst also decreasing the beam diameter. This probably will increase the interaction volume for other elements though. 

The JEOL brochure has a nice probe diameter vs probe current table for W, LaB6 and FEG. This suggests (at 10kV) the probe diameter for a W filament is ~100nm at 1nA, 2-300nm at 10nA and 5-600nm at 100nA. 

I like the concept in NIST DTSA-II of working with Probe Dose e.g. nA.seconds. You can get the same counting statistics by doubling the current and halving the time, at the cost of resolution (as the probe diameter increases). In other words, you can get higher resolution if you map at a lower beam current for a longer time (to produce the same probe dose).

I'd suggest working from the other way round in your question - how many analyses do you need and how much time have you got to get it in? 

[Probeman]

There are two basic strategies for improving spatial resolution, low voltage and low overvoltage. Each method has their own tradeoffs.

1. Low voltage: one can reduce the size of the interaction volume by lowering the beam energy. As Jon mentioned, running at 4 or 5 keV (or even lower) significantly reduces the interaction volume. Some quick and dirty calculations from CalcZAF (assuming a 50:50 Ni:Fe composition and density 8.3):

15 keV: 0.93 um
12 keV: 0.64 um
10 keV: 0.47 um

See here for details:

https://probesoftware.com/smf/index.php?topic=86.msg309#msg309

So using the L series lines for Fe and Ni, you will get x-ray emission from almost this entire interaction volume since the ionization energy for Ni La is around 0.7 keV. So usually when using low voltage to improve spatial resolution, we further decrease the beam energy to 4 or 5 keV (or even lower if a FEG gun is available):

 6 keV: 0.20 um
 4 keV: 0.10 um

The tradeoff with using L lines is of course accuracy as Jon alluded to. John Fournelle, Aurelien Moy and Xavier Lovet and others have evalulated accuracy of the matrix corrections for the first row transition metal L lines. I would search the literature for details but basically one either needs a standard of a very similar composition, or one can attempt to utilize the Ll/Ln lines instead.

2. Low overvoltage: the other strategy is to use K lines at a low overvoltage for example Fe/Ni Ka at 12 or 10 keV. Here are some more rough calculations using CalcZAF showing the generation volume (as opposed to the interaction volume) for Ni Ka at various beam energies:

15 keV: 0.58 um
12 keV: 0.29 um
10 keV: 0.12 um

For Fe Ka the spatial resolution will be slightly worse. So by using K lines and a low overvoltage (e.g, beam energy of 10 keV) one can attain almost as good spatial resolution as using a low beam energy (4 keV) with L lines.

The tradeoff with using K lines at low overvoltage is reduction in sensitivity as Jon mentioned, but accuracy can be excellent. But with a large LiF crystal and major concentrations you should be good to go. Modeling in Casino (or running some tests on the instrument), can help you decide which method is best for you.

[Michael Lance]

Thanks! This is very helpful.

-Michael

[jon_wade]

running low overvoltage (my favoured approach) you really do need to run some sort of decontamination. I have some 'comedy'* images of Carbon build up when mapping fancy steels but, more importantly, the landing energy and hence count rate drops off with C deposition. the  LN2 cold finger really helps all round

*yes, it was either laugh or burst out in tears!