Oct 11 Specmat Reply

Jim McVittie mcvittie at cis.Stanford.EDU
Tue Oct 11 10:38:20 PDT 2005

>1) Jun-Fei Zheng:  Poly deposition on a Vapor Jet Deposition Si3N4 film
>from outside CIS.  Jun-Fei is willing to provide TXRF data to show the
>film meets our contamination requirements.  The second part of the
>request is "to avoid H2SO4:H2O2 process as our wafer has never been
>through Lithographic process and this is what the H2SO4:H2O2 mainly for
>but H2SO4:H2O2 process might destroy the gate quality Si3N4. We also do
>not see the need for 50:1 HF dip, which is used for clear oxide formed on
>Silicon during H2SO4:H2O2 process. We believe SC1 and SC2 in combination
>would be able to address potential contamination during wafer handling
>and shipping."

This is OK with me.

>2) Ching-Huang: McIntyre ALD Hf02 processing in Semiclean Tools.  TXRF 
>analysis was provided with the request.  The data shows the sample to be 
>above our SemiClean contamination specification for both Fe and Ca (along 
>with S).  Do to peak overlap the levels of Ni, Co and Cu (possibly Cl) 
> could not be determined.

Sometime back we agreed to set the contamination spec for semi-clean
equipment at the spec for the lowest purity target we have for the SCT
sputtering system. When I was purchasing Hf and W targets, I consulted
Nishi and Saraswat regarding the trade-offs between cost and purity for 
these targets. They said that 99.9% purity or 1000ppm would be sufficient
for these materials. I actually purchased 99.93% or 700 ppm material. For 
our Hf target, the material certification lists the Cu contamination at 
209ppm and Fe at 0. Assuming a penetration depth of 10 monolayers or 28A,
I calculate a TXRF measurement of this material should give Cu measurement
of 2.5E12 cm2. I also checked the purity of Hf targets from Kurt Lesker at
their website. They spec their Hf at 99.9% and give data for a typical Hf 
target, which has a Fe level of 235 ppm and Cu at <25ppm. This should give
a TRXF measurement closer to 3E12cm2.  The bottom line is Hf deposited
from the SCT most likely will not meet a 1E12cm2 spec. I propose we move
our spec up to at least 2E12cm2 and do a TRXF measurement of Hf deposited 
using the SCT. Note, I suspect there could be a Cu/Fe error in the
certification for our target considering the difference with Lesker

With the semi-clean spec level to 2E12cm2, Ching-Huang ALD wafers will 
past and should be allowed. Hafnium is a very important material for
advanced CMOS devices, let its purity is still not very good compared to  
the other materials we deal with. His purity is not out of line for what  
is expect for this material. Note, I did talk this over with Mike and he 

>3) Mike Weimer (for the III-V Users):  Limiting CF4 Etchs in pQuest.
>Etches using CF4 must contain high amounts of O2. As a guideline (this
>number is not perfectly strict, but is a good guideline) any etch with
>CF4 should be ~20% CF4 and 80% O2. Small variations around this % are OK.
>No etches containing large amounts of CF4 and small amounts of O2 are
>allowed.  This basically eliminates etching of Si3N4 and SiO2 with CF4.

As for item 3, I feel that the III-V user's request to limit CF4 in the   
Pquest should be the rule. There are a number of reasons why this should  
be so:
1. F is well known to degrade Cl based processes. I can probably get a set
of sides from a recent AMAT talk on the subject if you are interested.
Well F can be removed with proper chamber cleaning and conditioning. We do 
not have any characterized F cleaning procedures for the tool. Who going  
to say what the worst case F chamber contamination will be? Who is going  
is to do the characterization for the cleaning process? Who is going to
monitor the chamber? The III-V users want a guarantee that their processes 
will not be affected by heavy CF4 use.

2. The tool has had a history of F processes negatively impacting the GaAs
users, whose processes are all based on Cl. While there is no definitive   
data showing F use been the problem, the issue has come up repeatedly over
the last 10 years and has gone away when the F processes have been 

3. Having Specmat set the rules for the use of the Pquest is a major
policy shift and should be discussed with the PIs for the III-V students
using the tool. The agreement, when the tool was placed in the SNF lab,
was that only processes compatible with the III-V etching would be run in
the tool. It has always been left to III-V users to determine what
non-III-V etching could be done in the tool. Initially, it was limited to
only Cl based processes. Three or four years ago, there were no III-V
students limiting it use and it got heavily used by non III-V users to
point that the III-V processes went out of control. At this point, the
III-V users rebelled, the non-III-V use was limited to 3 days per week and
the non-III-V processes were restricted again. The restriction to only
allow CF4 in dilute concentrations is refinement of restrictions already
on the tool.

4. As with any other plasma etching tool, one needs to limit the range of
chemistries if you want control. It is very common to limit etch
chemistries in etcher to improve
reproducibility and to reduce chamber seasoning time. MIT limits F in one  
of their Cl based Si etchers for this reason. In our Si etching area, we  
have Si etchers, oxide etchers and a metal etcher each with limited 
chemistries mainly because of large chemistry changes often affect process 
control in subsequent etching. The III-V users are only asking for what 
the Si users take for granted.

More information about the specmat mailing list