Furnace Protocol

Eric Perozziello eap at gloworm.Stanford.EDU
Wed Apr 1 03:03:16 PDT 2009

This is obviously a bigger issue than just furnace non-uniformity-

FIRST:  SNF began reducing the Tylan oxidation tubes' idle temperature to
500C from 800C during periods when tubes are not reserved.  This change was
only recently discovered after tubes were noticed idling under the recipe
"500AN."  The reason given was "to save electricity."   Users are now told
to turn up the heat "right before you clean your wafers," but the temperature is far
from uniform in the tube in one hour.
There were no prior announcements or discussions with the general lab user
base, and no calculations or tests were done before making this drastic change.  
As we might expect, this change has resulted in big variations in both in-wafer
and wafer-to-wafer uniformity due to insufficient heat soak of the equipment.
(My cat never studied physics, but even he is well aware, it takes a lot longer
to warm some places in the house even though the thermostat reaches its setpoint.)

Here are some reports from lab users:
"It seems that the oxide thickness uniformity was still pretty good until
at least Feb 2009.  Now the oxide thickness increases gradually from the top
to the bottom of wafer (flat at bottom), with variations of as much as 15nm
within wafer (used to be < 5 nm)."
And another:
"very big variation in quals: According to the quals on 3/18, both in-wafer
and wafer-to-wafer variation are very large. I also observed non uniform oxidation
recently. We may need inspection."
What is the purpose of running quals when nothing is done or said when they fail?
Apathetically, the tubes are still being ramped down (31-March) almost two weeks
after a bad qual, reiterated by users.
SECOND:   TLC CLEANS used to be run twice a week on each tube because contamination
diffuses in from the the tube, and not just because of tool usage.  This seems
to be misunderstood by staff, given recent comments posted. Why bi-weekly cleans
ceased is not clear, other than a lack of necessity on the part of the lab to
make real, working devices along with the lab users.  We've gone from technician-run
TLCCLEANS twice a week, to not running them often, to now technicians complaining
when users run them.
Can we please have a bona-fide engineer or scientist review any changes like this,
AND POST THEM FOR COMMENT before wrecking people's experiments because of hasty,
ill-conceived decisions?  (Especially when a modification is made to a manufacturer's
design spec!)   A few simple calculations/tests could be made by qualified people
before costing users lots of time and effort. To paraphrase "the letter" drafted by
students in Feb 2008, Is anyone accountable/responsible for these decisions?  More
than a year later, we really need to close the loop on these kinds of things. 
Obviously "quality circles," "name tags," and "shadowing" didn't resolve the basic issues.

More information about the tylan1 mailing list