PQuest Reservation Restrictions

Mark Wistey wistey at snowmass.stanford.edu
Fri Dec 19 10:13:04 PST 2003


Dan -

Thanks for your thoughtful message.

I'll explain where we're coming from on the III-V side of things, and
hopefully it will provide some perspective from the other side.  The
pquest seems to go through a two stage life cycle.  In the first stage, the
machine starts in perfect working order but with few users.  Gradually
more users discover it, develop new processes for it, and introduce new
materials, sometimes without clearing them through proper channels (or
even being aware that there are channels to be cleared).  Then in the
second stage, several III-V people share notes and find that they're all
having the same problems. This usually happens when a III-V student is
caught between the PQuest and graduation and truly needs the machine in
peak condition. Accusations fly, suspicious processes are retested or
suspended, and everyone is on their best behavior for a while. Then the
responsible student graduates, and it starts over.

By splitting the week, we were hoping to short-circuit that life cycle.
It appears that the troublesome etches are not being done by III-V people,
but the III-V's are the most affected.  There are also some indications
that AlAs/AlGaAs may be more susceptible than plain GaAs, so certain III-V
folks are affected more than others.  So by cordoning off the troublesome
users from the III-V's, both groups have a wider margin of breathing room.

There's another observation to note, from the users who've been at the
PQuest from the beginning: there seems to be a cumulative or even
threshold effect to certain kinds of contamination.  In other words, if
someone runs a "bad" process once, or for a short time, it might not
affect III-V etching, but repeated "bad" processes eventually cause some
kind of contamination.  So the old process qualification test (an hour
of process X, followed by a short O2 plasma, seasoning, and a test GaAs
etch) might not reveal long term consequences of process X.
Contamination outgassing and particles flaking off the sidewalls are
just two suspected examples: days to build up and days to remove.

So this is a long way of explaining why the week split worked so
effectively.  If indeed there were "dirty" non-III-V processes, by
clustering them together, then after several days of consecutive III-V
growth, the machine will be clean enough for the most critical etches.
And by and large, it seems to have worked.

I hope you can see why we feel reluctant to give up days.  It's not
that we're being pricks, and it's not that the III-V folks are lazy on
Wednesdays.  Some of us don't want to risk being in the first etches
after silicon, due to bad experiences we've had.

I've laid my cards on the table here to help further the discussion.
I'd be interested to hear what suggestions others have to offer,
especially ways to alleviate these collective fears.

On a different note...  A lot of the recent complaints seem to be about
certain people reserving long blocks of prime hours on the machine.  If
someone is violating either the posted policy or the one Jim McVittie sent
out this week, they should be reported to the staff.  After a warning, they
should be suspended for a while.

If you (not Dan personally, I mean y'all) need long blocks of time, they
are available off hours, even Sunday-Tuesday.  This is a shared resource,
so people who want a big share should expect some discomfort (skewed
hours).

					 - Mark


- - - - - - - - - - - - - - - - - - - - - - - - + - - - - - - - - - - - -
Mark Wistey, wistey at snow.stanford.edu (Internet)| the love of the Lord is
575H Stanford Ave, Palo Alto CA 94305 (BikeNet) | the beginning of wisdom





More information about the pquest mailing list