FIB is modern, but not always uncritically? Why You place the film not on a pod and thinnest them with argon ions in a GATAN PIPS (or DuoMill) with nitrogen cooling, before transporting the thinned sample in the TEM you stick them on a 3-mm ring.

look here: http://www.gatan.com/specimenprep/pips_cols_stage.php

.....

But still you need to definitely avoid steep angles of incidence also briefly!!!

I will check if we have that system in our campus. Btw, I would like to know whether TEM can use to measure the fraction of phase transformation in material

Of course, this can be measured. But I think that the crystallization nuclei will be very small. Therefore you need a HRTEM with very good resolution (0.19 nm or so, for instance a JEOL 200KV with UHR pole piece) . Otherwise you can not reliably distinguish between amorphous and crystalline areas.

If the volume fractions of the crystallites are large, I would work with lower magnification in the diffraction contrast mode (with objective aperture).

However, even then you need a program for image processing. ImageJ is always good!

see: http://rsb.info.nih.gov/ij/index.html

.... for the volume fraction you have no great stereological problem. It is the only parameter for which you can close out of the area without model approach to the volume (if the film is very thin).