Roger,
      Composite answer. The first paper and technical work on image
reversal was by Scott McDonald of IBM.  He created a reversal resist
using a 1% additive imadizol to a standard 1350J Shipley resist.  It
worked well, but it deteriated over time and Scott would keep it in a
fridge for 2 weeks then scrap it because of critical dimension changes.
His paper and work with us led to the ammonia process.  This led to no
dimensional changes and the use of one resist.  The second customer we
worked with had just had an operator put the wrong resist in the wrong
track, and had an expensive rework and clean up problem.  As their
reversal resist was a lot more expensive then using the standard resist
they saw advantages to better control of the production area, better
C.D. control and consistency.  The third customer was using a reversal
resist and was trying to achieve 0.1 micron metal gates. Using reversal
resist he could achieve 0.25 micron gates +/- 0.1, and this was making
engineers run the process to ensure there was a fast time to resist bake
to fix the dimensions.  When he switched to ammonia image reversal he
achieved 0.1 micron  gates in 1.2 micron thick resist with better than
0.08 micron control.  The fourth customer had been running with reversal
resist for over 4 years when his resist company discontinued the
reversal resist.  It became commonplace for resist companies to do small
runs of reversal resist then find it was uneconomical to run the
reversal resist and discontinue it. He was delighted to find he could
run with his standard resist with very little change in processing with
better control.  For all of these reasons ammonia reversal seems to be
the accepted lift off method.
Bill Moffat, CEO
Yield Engineering Systems, Inc.
2185 Oakland Rd., San Jose, CA  95131
(408) 954-8353
cell 408 590 4577
bmoffat@yieldengineering.com
www.yieldengineering.com