You can try to "train" (stabilize) sections by irradiating them at lower (than you'll use for you maps) magnification for 3-15 min.

Also you can use thicker sections (gold) for better signal/noise ratio. Of course some loss of resolution is possible, but if you work at low/intermediate magnifications it does not matter. 

You can formvar coat the grids, lower the emission voltage, expand the condenser light, cut your sections thicker (although the resolution will be less) or try embedding in a different epon ratio of A to B component. Formvar is 0.5% polyvinyl formal in ethylene di chloride. Dip a slide in the solution and let dry, then scratch the edge with forceps and let the film come off on  distilled water. Top the film with grids and pick up on another slide by coming from above and submerging the grids in water. They will adhere to the slide. Alternatively just add the formvar solution to water, top with the grids, and pick up on a piece of parafilm.

Dear Jari, IMHO, if you are prdoucing your formvar coating of grids by yourself, you could also try to make the formvar film a bit thicker. This might be perhaps a better solution re: resolution than cutting /mounting thicker ultrathin sections. As I understand you will have your specimens examined by EDX-mapping so it might not be easy to lower the emission voltage [this could be read as 'beam acceleration voltage' (e.g. 50kV, 80 kV, 100 kV etc] or   "deflecting" the beam]. Another possibility would be to use grids higher 400 mesh.

But: My experience (similar as Vladimir proposed to you)  in examining and documenting morphology of ultrathin (70 nm)  sections  of diagnostic samples was to irradiate grids (regular square 75 - 100 mesh, copper) mounted with a bit thicker formvar film (80kV accel.voltage, lowest magnification possible, withdrawn objective aperture and - if existing - also a selected-area aperture) for the first minute with almost totally deflected, thereafter for at least 2-3 minutes with a beam deflection which illuminates a bit more than the diameter of the observation screen. Naturally one also could adjust exposition time for documentation to the shorter side if that is feasible (not knowing whether this is possible with your TEM and for your EDX-task).

Thanks for all comments. I have tested thicker cut lamellas (upto 230 nm), not feasible on my experine. Also, I think the drifting problem is usually far worse with formvar than with the holey carbon mounted specimens. I concluded this might be a problem due to not eletrically conductive system. For the details: I have a 200kV Jeol-JEM2100F and magnifications used are usually 50kx-500kx. I do life science materials and also materials science like nano catalysts etc.

Dear Jari, thank you for listing the parameters for your TEM and the mag-operations. You perhaps are right with "holey-carbon mounted specs" which might be a bit more resilient to charging and/or heat-induced / physical drift due to impaired or - in the worst case - even blocked electrical discharge. So you might achieve better stability in using grids mounted with formvar or pioloform-film + finally additional carbon coating (stability of Epon-sections perhaps depending also a bit on the mixture, infiltration, and polymerisation conditions of the specimen blocks).

NB: When in my active professional work I examined (only formvar-coated) grids I sometimes also noted aggravated risk of "section" drift  

i) when I did not use the LN2 cooling device (cold trap) -  which points to increased specimen heating, and  ii) when the "final" fit between specimen [grid] carrier and the grid carrier's seat / instrument's ground(ing) was not perfect or insulated for a certain  reason (electrical disturbance). But one has to admit that your conditions (measuring / EDX analysis at x 50k to x 500k) are extraordinary. Let me wish you a miracle to happen regarding your nasty problem, regards, WM.

Thanks Wolfgang. I appreciate your concern and wishes. For me, easiest thing to test would be to add some coat on the grid and/or sample. I hope I can report the result near future. If you guys find or know any studies about TEM sample coating in order to eliminate the drifting, I would be more than interested. BR, Jari.

I am not sure coating will eliminate all your problems. It should work in the same way as if you place your section on a grid with carbon film and put it in a TEM upside down.

The highest magnification I ever used while working with EDS with biological sections was about 80 kX. Carbon film was used. Line scans, not maps were acquired (they need much less time and anyway I like line scans better, more information). Some drift did occur, even after pre-irradiation, but results were usable. 

Dear Jari, just to let you know:  I only found the following reference regarding the matter:

Stabilization and clarification of sections mounted on Formvar-coated slot grids by deliberate irradiation in the electron beam, by YANG & MORRISON, J. Microscopy, Volume 103, Issue 2, March 1975 Pages 187–194, 

http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2818.1975.tb03893.x/full

Abstract/Summary (only for convenience): SUMMARY

Epon-Araldite sections supported on Formvar-coated slot grids can be stabilized by exposing the area of interest in the column of the electron microscope to irradiation of 0.5 C/cm2. With this procedure one can obtain electron micrographs of high quality, without drift or distortion, from any part of the open slot area. After treatment the support film is not only stable and transparent but robust enough to permit repeated removal and replacement of the specimen in the microscope column. We illustrate the method with a series of electron micrographs with initial magnifications ranging from × 300 to × 500,000 to show the quality of image that can be achieved. The effect of the irradiation on tissue architecture is checked by resectioning the irradiated sections. We recommend this method of local irradiation of sections for investigations using low or intermediate magnifications for the wide field it permits, for the durability it confers on the specimen and for its convenience.

Should you be in need of that article, please, let me know.

NB: just to compare with SEM-specprep [cf:

(http://www.microscopy-analysis.com/sites/default/files/2013_May_Joubert.pdf)]:

"Finally the specimen is coated with a thin conductive film of metal, typically gold and/or palladium, iridium, or platinum. In cases where heavy metals may interfere with the secondary electron (SE) or backscattered electron (BSE) signals needed for the detection of valuable information, carbon evaporation can be used to provide a conductive coating on the sample. 

So it might be able that you mount ultrathin sections on to formvar-(holey)carbon coated grids and afterwards´coat again with carbon to further stabilize against drift and on the other hand overcome electrical disturbances....

Thank you guys, again, for the knowledge shared. I use carbon coating mentioned above sometimes for SEM specimens. However, I have found low vacuum (~15-25 Pa, N2) work more feasible often. Since the biological samples are prone to rupture in high vac conditions. This article by Joubert is interesting, they also seem to have similar SEM instrument compared to ours :)

To Vladimir: The analogy of the TEM specimen upside down is interesting. I thought this coating could provide a poor man's Faraday's cage to ease the electrical charge accumulation and therefore, to stabilize the sample drifting in the electric field in column. However, it seems that I need to test it.

Dear Jari,

Epon is not the best choice for analytical microscopy. Araldite or at least Epon-Araldite mixture will be better, and sections will be stable. For EDX the thicker sections are to be used, up to 400 nm. Thin sections (40-60nm) are good for EELS. I suspect that neither formvar nor carbon coating would help to stop drifting but maybe if you put additional cuprum grid 50-100 mesh upon your 400 mesh grid with the sections, some charge will flow away.

Alexandra, I'll test the araldite whenever possible. Double grid application might be worth to test. Cheers.