Hi Abhi:
I don't know if your problem is planar in nature or not, but it sounds a little bit like you are trying to simulate a very thin oxide layer on a semiconductor or some type of integrated circuit.
I have not had good experiences with HFSS for things like this because in order to get consistently accurate results, your mesh is dictated by the oxide layer mesh (0.1 micron layer). You end up with many small mesh boundaries that have to match up with descritizations in the thicker layers on both sides of it, and this usually leads to very large problem size and long run times.
This is one of those applications that a planar software tool based on the Method of Moments like Momentum or Sonnet is best suited to do because they are based on planar meshing, not volume meshing. I think you will find that these codes will give you accurate results with much smaller analysis requirements, and do this much, much faster than HFSS.
I don't think there is a free version of Momentum, but I think Sonnet has one on their web site (sonnetsoftware.com).
I have heard that some people get some limited success with mesh seeding in HFSS; you might try looking through the documentation in your software for information on seeding the mesh. Maybe this will help to keep the memory requirements down a little bit.
--Max