Correct about HMDS.  It will make things worse.  Even your unexposed
regions will suffer(and stick).

For success with bare silicon processing, I use a 5 minute oxygen plasma
RIE just prior to spin.  The surface interface physics, and chemistry is
different for nitrides, and in this case; the oxide you have.  Hydrogen bonds.

Being SiO2,  piranha, followed by a rigorous bake is best.

Which brings us to AP300 or Omnicoat.  I suggest trying it.  I have used
Omnicoat with success on nitride, and oxide.  The data sheet instructions
are good.  The important thing is to ensure that the correct Omnicoat bake
temperature is indeed reached.  Interesting side about Omicoat is the
tensile stress it exhibited on quartz wafers.

Stress:  Carefully controlled by taking the longer bake ramping/process
temp and cooling times throughout.

Adhesion: Always err toward over-exposure for best results.  Using a filter
as recommended is great,  do increase the dose by at least 40%.  Interval
is best for any films over 20 microns thick for most aligners.

Developing:  I use a two bath method.  I use water to rinse,  less stress
inducing than IPA.

Regards, and hope this helps.

Brian C.

Brian Czimback
Nanobiotechnology Center
Cleanroom Operations
350 Duffield Hall
Cornell University
Ithaca,NY 14853