You might want to consider using an oxygen plasma to improve
adhesion.  I have not done this with SU-8 but I have had adhesion
problems with other photoresists were solved by doing this.  A simple
ashing procedure for less than 5 minutes should be enough.

As for why this happens, I have heard many explanations.  The two
most likely ones are that 1) Ashing produces a minute surface
roughness that improves adhesion and 2) Oxygen ions are imbedded into
the the first few nanometers of the substrate that react with the
photoresist when it coats the wafer, improving adhesion.

While reason 1 is more likely, I have never noticed any surface
roughness produced by ashing (but then again I haven't done any
before/after SEMs of wafers I have ashed).  Either way, it works with
shipley photoresists and should help you out.

Nicolas Duarte
Penn State University
PhD Student in Electrical Engineering

At 3:05 PM -0500 11/18/06, Bin Wang wrote:
>Hi all,
>
>Recently we met a problem on SU-8 adhesion. After a typical SU-8
>process, the ENTIRE SU-8 film peeled off ALL AT ONCE from the Si
>substrate when we developed it. We have dehydrated at 200C for 20
>min before coating. For prebake and PEB, we tried both standard
>length of time and the time 50% longer than the standard, but
>neither worked. Could anyone help me crack this issue? Thanks a
>lot.
>
>Additional info: Wafer cleaned in Nanostrip for 3 hrs before
>dehydration. SU-8 2050 spinned at 2500rpm. We used standard
>exposure time, standard developer and IPA rinsing.