LPCVD Nitride Update
mcvittie at stanford.edu
Mon Dec 17 14:46:33 PST 2001
Tylan Nitride Users,
I want to update you on what is being done to address some of the
reported problems with our low stress LPCVD nitride films. There are
three problems that we are looking at: 1. Bubbles (1 um range) in
nitride films, 2. Increased nitride etch in hot KOH, and 3. Higher
stress (> 150 MPa in our low stress process).
Our low stress nitride process was developed in the mid 80s by Sung-Tae
Ahn, who was a grad student in material science. He found that the
nitride stress decreases as the SiH2Cl2 / NH3 ratio (R )increases. His
results were summarized in a short paper published by the Pattie Beck,
who was an Applied Physics grad student, in a MRS conference proceedings
in 1990. Sung-Tae Ahn and I were co-authors of this paper. The original
low stress process was run at 785C with an R of 6:1 at a pressure of 300
mT. At this flow ratio, the film had a reflactive index of 2.55 and a
stress of around 50 MPa on oxide and around 150MPa on bare silicon. No
bubbles were seen in the film and the etch rate in 33% KOH was 8A/hr at
R= 6.0 and 16 A/hr at R=6.5. Increasing R above 6 increases the KOH
etch rate significantly. By 1996 our low stress process (SIN5.5 with NH3
= 30 and DCS = 165) was running at a DCS/NH3 ratio of 5.5 with the
stress still around 150 MPa on bare silicon.
In 1998 the flowmeters for the nitride tube were changed from the
Mono-block construction as used in all the other Tylan tubes (except in
the Tystar) to individual flowmeters. I have no record of stress
problems being reported at that time. During the summer of 1999, several
users reported stress levels between 600 and 700 MPa for films deposited
with R at 5.5. At that time a series of nitride runs were done by the
staff and several users with R increased up to 15. In October of 1999,
we settled on LONH78 (NH3 = 13.5 and DCS = 189) which had R at 14.0.
The rational for allowing the process to be change was that we thought
that the flowmeters had most likely gone out of calibration. At the time
we did not a means of doing in-situ flow calibrations like what is done
on most of the plasma etch equipment.
---Bubbles in films: About a month ago a remote user sent us an SEM
showing 1um sized bubbles in a low stress film deposited for him by us.
Upon checking, we found out that bubbles have been seen in our films for
sometime. Checking low stress films from 1996 with R = 5.5 showed no
sign of bubbles. In addition, recent tests using R= 11, 8 and 5.5 showed
that the bubbles disappear as R is reduced to 5.5. The bubble problem
indicates that the 1999 increase in R was likely a mistake.
---Flowmeter Calibration: During the summer we together a measurements
system for in-situ calibrating flowmeters in LPCVD systems. The method
uses rate of pressure increase with the gate valve closed and N2 feed
into the flowmeter under test. This system was recently used to
calibrate all the flowmeters on the nitride tube. The flowmeter error
was found to be < 10% for all the flowmeters. For LONH378 the actual R
value was found to 11.7 instead of the set value of 14. While there is
an error in R, the error is not large enough to explain the need for the
large change made is 1999. Our actual R value of 11.7 is consistent with
our current bubble problem. For SIN5.5, which was the pre-1999 process,
the actual R value is 4.6 instead of the set value of 5.5. This
explains, why an increase in R of around 20% was needed.
---High etch rate in KOH: At least one user has found that the current
(deposited during summer) low stress nitride does not hold up in KOH as
well as films deposited in 2000. XPS measurements showed that the 2000
film had slightly less oxygen, perhaps due to a lower leak rate during
deposition. From the data in the 1990 paper, it is likely the etch rate
would be greatly reduced by reducing R back to 6.
--- High process pressure: Currently, the process pressure is running
over 500 mT. We suspect that the pump is the problem since it may have
been damaged last month when a user let the system pump on air for
sometime before the burn box failed. The pump will be replaced during
the shut down.
--- High Stress: We still have not answered why the stress went up so
high in 1999 and why we are still seeing higher stress than expected
when we lower R back to around 6. Note that both Berkeley and Cornell
run their low stress processes with R near 6. Some the issues that we
are looking into are:
1. Calibration of our stress measurement tool: Our FSM system is old and
a number of users have noted that it gives inconsistent measurements. We
are going to do a low stress run and measure the stress on at least 2
other systems to verify that our FSM system is not a source of our
2. Deposition temperature: While we have always run our nitride
depositions around 780C, Both Berkeley and Cornell around 880C. We will
do a 880C run to see its affect.
If you have any comments and ideas on the nitride problems, please send
them to me.
Thanks, Jim McVittie
More information about the tylannitride