I am making SU-8 2025 masters on Si for PDMS replica molding. I use a
mylar mask and the structures are ~25um tall. I have a problem that I
do not get good vertical walls of SU-8. A 15um feature comes out as a
flat length of ~30um and then a linearly decaying thickness that makes
a total feature size of ~60um.

The processing is as follows:

1) spin at 3000rpm
2) soft bake 65C-1min, then 95C-4min
3)Exposure-700mJ/cm2 using a mylar mask
4)post bake 70C oven - 90min
5)develop - 3min

Please let me know if anyone has any ideas of why feature sizes are
enlarged and walls not vertical at all..

-Abhishek Jain