For that thickness of etching, I would recommend using a SiO2 hard
mask.  I have done submicron deep trench (AR >40:1 with grating space
<1um) on newer ICP systems, and once I went more than 30um into the
Si, the PR mask began to lose integrity.  You have to transfer the
pattern to SiO2 and then etch into the Si, otherwise the sides of the
PR will be too eroded and you will lose mask integrity.  Particularly
for an ICP environment, you want to use SiO2, not silicon nitride or
anything else, as it holds up the best under ICP etching.  The problem
there is the ability to get a thick layer of PR and get submicron
resolution, and you have to watch resist reflow during hard bake with
very thick resists.  I had resist failing after etching 30-40um into
Si even with PR thickness ~3um, which is extremely hard to do and get
that kind of resolution.

-Eric