In general, spatial resolution of TEM is better, than SEM. The most important and obvious reason is in geometry: in SEM we have a bulk specimen and the depth of penetration of beam is about 1000 nm, in other words, the source of X-rays is a sphere with diameter about 1000 nm. In TEM we have just a thin section (50-100 nm) and the source of X-rays is a disk with beam's diameter and section's thickness.

I think, it is a good idea to use TEM, but it is important to remember: EDX isn't a tool for precision chemical analysis, it's just an instrument for estimation of elements distribution in a specimen.

Good luck!       

Dear Denis Korneev,

Thanks for your explanation. 

SEM/EDX  can accomplish accuracy and precision equivalent to WDS and at considerably lower electron dose, even when severe X-ray peak overlaps occur, provided adequate counts are documented. Achieving this rank of performance is now much more realistic depends on: element identification that must precede quantitative analysis, at least some current and many legacy software systems are helpless to sporadic misidentification of main constituent peaks, with the frequency of misidentifications rising significantly for minor and trace constituents. Then, the standard analysis for example our specimens to much broader systematic errors, leads our quantitative results have sufficient accuracy to solve critical problems. EDX also has such a large volume of acceptance that it seems that credible spectra can be obtained from specimens with complex topography that modify X-ray propagation, resulting systematic errors

Hi Ahmad,

I don´t think that EDS in TEM can help you to distinguish between Sb and Te. The emission energies of both Sb and Te are very close to each other. Difference between them is only 164.2eV on L alpha 1 lines which almost reach the energy resolution of your detector. For example Oxford Si SDD detector has energy resolution 124eV measured on FWHM of Mn K alpha 1. If we consider the peak broadening the 164.2eV is over interpretation limit of EDS analysis (in my opinion).

On M-lines family there are differences less than 50eV so you can´t rely on it. Your only hope may be find out the K alpha 1 peaks. The related energies for Te and Sb on this level are 27.473keV and 26.359keV respectively, which gives contrast of 1.114keV. That sounds much more better but the probability of K alpha 1 transition is more less than for L alpha 1. So I would expect much smaller peaks from K alpha 1 in the spectrum. And if you can´t see the K alpha 1 peaks in SEM EDS spectrum, where the signal from the sample is far rich for a photons in compare to TEM EDS, I don´t share the view that TEM EDS may be useful for you. The beneficial way should be to use the TEM with EELS spectrometry.

Dear Abdillah Imron Nasution ,

Thanks for your clear explanation. 

I think you should try TEM. For SEM accelerating voltage (usually up to 30kV) is too low to excite K lines of Sb and Te, and you do need K lines because other series (as shown by Vladimir Girman) are to close, at the the resolution limit (or below). Accelerating voltage of TEM is high enough to excite K-lines, where you have a good chance to identify Sb and Te successfully. 

to Denis:

Your comparison between the difference excitation volumes for X-rays in SEM and TEM is correct to first order. However, there are detailed models that allow you to calculate it more precisely (for SEM: phi-rho-z; for TEM: Goldstein model etc).

The major difference is that all electrons get stuck in a bulk SEM specimen, hence the excitation volume can be calculated analytically by integration, and so the self-absorption effect can be included EXACTLY; as a result, the chemical analysis can be a HIGHLY REPRODUCIBLE mean value averaged over a larger volume,

say some cubic microns, but the mean result is meaningless if the specimen is inhomogeneous.

In TEM most electrons go through the specimen and one needs Monte Carlo simulations to model the interaction volume numerically, which is more complicated. Self-absorption depends heavily on take-off angle, specimen tilt and surface roughness and can only be calculated APPROXIMATELY. As a result, the value is an approximation and OFTEN NOT ACCURATELY REPRODUCIBLE but

the spatial resolution is better (several 10 nm^3 or bigger) and so inhomogeneities can be mapped reliably.

to Abdillah:

I don't think EDX(S) in an SEM will be anywhere near as sensitive as WDX.EDX will be OK to detect ~at%, perhaps ~0.1at%,  but WDX can be 1-2 orders of magnitude more sensitive (100ppm)!

to Vladimir:

If your energy calibration of the X-ray Spectrometer is well calibrated, the software should be able to distinguish lines that differ by much less than the energy resolution of the system, because it is the peak position of the line

that counts and can be determined to better than a single channel precision by weighted interpolation. So check the system calibration with a known line!

If you want to distinguish K lines from heavy elements like Sb and Te, then your acceleration voltage must be higher than the corresponding X-ray line energies to be able to excite then. So try SEM-EDX with >40kV (or TEM with >80kV) and, importantly,  a Ge detector that has better detection sensitivity than Si at higher energies! Especially modern Si-SSD detectors detectors are so thin

that most 30kV X-rays will penetrate them without being detected!

Dear Thomas, thank you for the interesting comment. Of course, the interaction between a beam and a specimen is very complex and that geometrical think was just a simplistic illustration of fundamental differences in the spatial resolution.   

@Thomas Walther

It looks like we understand question in different ways.

Your reasoning is for single element identification (whether it is Sb or Te), mine - for resolution of lines when both Sb and Te are present. By the way, it's not an easy task to find an SEM with>40kV accelerating voltage.

So, back to original question: yes, TEM looks better whether you have one or both elements in your specimen. Or at least it is easier for an operator (you need rather good operator for this task with SEM).

Many thanks to all for great discussions

.  I should mention that our studied material is a composite; one of the phases contains Sb element and another one contains Te element.