JEOL SXES Soft X-ray Emission Spectrometer

[JonF]

I had a private message about the quantitative analysis of Li by EPMA-SXES and other SXES queries in general, and I thought I'd copy out the bulk of my reply here. All of the below is my interpretation of things based on my own experience, so read in to this what you will! I'm happy to be corrected by any stray physicists out there.
For the sake of clarity, I'd also say that I'm actually a fan of the SXES, but they aren't drop in replacements for WDS - they're another complementary detector alongside WDS and EDS that can provide a bit more information. To fully appreciate what the JS50XL, JS200N and JS300N gratings can tell you, you'll need to wander down the rabbit hole of "where do X-ray emissions come from and why do the peaks look the way they do?".   

Regarding Li determination:
Quantitatively analysing Li by EPMA is a bit of a non-starter. It breaks one of the golden rules of microanalysis in that the Li Ka intensity varies as a function of not just concentration but also the form that it is in i.e. the Li Ka emission from Li metal and Li2O is not directly proportional to the concentration of Li in each phase (matrix corrections aside). Hovington et al have a paper on this that covers the basics (https://doi.org/10.1002/sca.21302), and in particular figure 2 (note the 1/5th scaling of the Li Ka in Li metal relative to the other Li phases).
To make matters worse, aside from the concentration =/= intensity, the peak shift and the peak shape changes, the Li is also readily beam damaged, forming Li metal (and so the Li Ka becomes more obvious during a course of analysis as the sample becomes damaged!). And then there's interference: for natural materials, the Fe Mz (M2,3-M4,5) is right in the way of Li Ka! A (very) high order reflection of the C Ka is knocking about down there too, which isn't helpful if you need to carbon coat your non-conductive samples (which will also effectively hinder the Li Ka making it out of the sample).

Qualitatively determining presence/absence and phase of Li might be possible depending on the phase that it is in - sometimes the Li just refuses to generate an X-ray. To see Li though, you'll need the JS50XL grating, which is only available on the SXES-LR along with the JS200N grating. The SXES-ER comes with the JS300N and JS2000 gratings. I've been told that the gratings aren't compatible between the -LR and -ER (I'd quite like the JS2000 grating myself!).

Regarding the slow CCD speed, it is definitely a bit of a hindrance, but these light elements (rather low energy transitions) just aren't favourable for X-ray emission, so the slow CCD speed is actually not as bad as it sounds: you need to count for 100s if not 1000s of seconds by WDS to get good counting statistics if you want to quantify. If you're looking for multiple light element/low energy emissions, then measuring for a few seconds to get all of them and the backgrounds isn't too bad. You can also measure using the WDS and EDS during this time, so there's two hyperspectral datacubes (EDS and SXES) plus the WDS maps for a single pass.   
The slow CCD speed would also not be a problem for point analysis, however the software doesn't support automated acquisition of a series of points, so you'll be manually driving around and clicking "go" for each point - but at least the WDS are free to move around and measure as normal.

Regarding regular calibration: the SXES is quite stable relative to WDS as it is insensitive to barometric pressure and relative humidity fluctuations (i.e. why we need to keep re-standardising the GFPC - the sealed Xe detectors are much more stable). Well, at least as far as I've seen, anyway. That doesn't mean it doesn't need calibrating though! The energy axis is quite stable, but the intensity axis is quite sensitive to CCD temperature and the crud that will inevitably stick to the -70C CCD and absorb low energy X-rays (the cryo-pumping problem!).

Regarding cleaning of the SXES CCD: there's no window or anything between the sample and the CCD surface. It is the CCD itself that will become contaminated. The only way that you can clean this up is to turn the cooling off and allow it to warm up and hope everything that was on the CCD boils away in the vacuum. -70C shouldn't be cold enough to significantly capture water vapour under the chamber vacuum, but you shouldn't have the CCD cold during sample exchanges. If the crud on the CCD doesn't boil away, then it'll be an engineer visit to clean it up.
The time before the contamination builds up to the point where the SXES can't "see" will be variable depending on how often its used and how dirty the system is (and what the users try and put in there!). You can monitor the cleanliness of the system using your regular standards: the absorption by the muck layer will become apparent if you look at the intensity ratios between e.g. Al La and O Ka[3] over time (might not bee the best choice as the O Ka will get strongly absorbed as well!).