I should have put this info in the first post, but here's something I
had mentioned in one of the followups.  Wish this was a newsgroup or had
a web archive, then I could just point to a link...(oh, using RIE):

> We're using a foundry service (through MOSIS) that gives us a standard 5
> metal layer CMOS chip back.  We define the top level interconnect metal
> for use as a post processing etch mask.  So, the aluminum mask is a
> must, and can't modify it to any great extent.  The depth of the etch is
> such that we are beginning to see ARDE effects, and need the lower
> pressure/higher bias to clear out the SiO2 from the bottom of the
> trenches.  So, mainly looking at options for modifying the aluminum
> (i.e., oxidizing or something) to increase its sputtering resistance.
> We would like the aluminum layer to still be mostly aluminum when we're
> done, and it's only 1micron thick to start with.  So, we need something
> that will controllably oxidize 100-300nm.  Most anodization schemes I've
> looked at are likely to turn the whole layer into Al2O3 in a very short
> time.  maybe I'm wrong. I started looking a bit at plasma oxidation, and
> it would be very nice if we could do it in the RIE, as then we wouldn't
> have to break vacuum.  We could even oxidize a bit, etch, oxidize some
> more, etc.  But I have no idea where to start as far as plasma
> parameters, and oxidation times are concerned.  If they are
> self-limiting that would be great, because then we could set up the
> conditions, and the thickness wouldn't be very sensitive to exposure
> time.  Another reason I'm shying away from the wet processes for
> oxidation is that the majority of the milling occurs at the end of the
> etch when we have to drop pressure and turn up the bias to clear out the
> last of the SiO2.  At that point I wouldn't want to put the chip in a
> wet environment if I didn't have to.  We could possibly even wait to do
> the oxidation until this point, but all of that means I need to find
> some rate temp dependence data, and so far I've come up short :)