Pepsi Challenge

[Randy True]

Mark-

There are simple ways not to exclude people from the discussion, use the 
log book, the pquest at snf email list, and coral. I'm asking for the 2nd 
time to please document the details of these experiments you refer to. 
As an example, look at the summary from last July, it's in the pquest 
email archive. Here, I will copy it to the end of this email. I do not 
know if you are aware of this, but almost identical problems to those 
you are reporting were encountered last July by Vijit Sabnis. Initially, 
my process was blamed and Jim McVittie forbid me to run it. I think that 
the resolution of the issue was that the use of copper tape to attach 
pieces was the culprit, not my process

Since August of 2002 up until the recent problems, no one has reported 
(in any public format) problems with their etches that they suspected 
were due to interactions with other processes. Basically it has been 
smooth sailing. In the last month, usage of the pquest has quadrupled, 
including several new processes. I'm not blaming these new processes for 
the problems, but given the number of different materials being etched 
and the lack of monitoring and characterization, it's not a suprise that 
standard processes have gone south (my process included!).

The only GaAs user I have talked to, Gigi, said that he has only had one 
abnormal etch, and in this case he changed some of his experimental 
parameters. With regard to ruining one of your valuable samples, you 
should certainly be running test samples beforehand, and if these turned 
out fine but your "real" chip didn't, I don't think it's logical to 
blame chamber conditions. Looking at the log book, I only saw one entry 
under your name (along with sbanks) on 6/12. How many times have you 
used the pquest in the last month? Why aren't you logging your runs? 
When did you first start using the pquest and have you established a 
reproducible baseline etch process?

When you mentioned PMMA, were you refering to my polyimide etch process, 
ot perhaps another process that uses PMMA as an etch mask? About my 
process: I etch polyimide (PI) not PMMA, though I do not think there 
would be a difference since PI and PMMA are very similar polymers. I use 
Al as a hard mask and the gases in the etch are Ar and O2. My wafers are 
"clean" at the point that they go into the pquest, they are coming from 
the P5000 metal chamber. I have been doing this same etch for over a 
year, running consistently (at least several times a month and sometimes 
several times a year) and concurrently with many GaAs users (gigi, 
ethrush, kaimeifu) and some non-GaAs users (cmfalukner).  It is not 
impossible that my process (perhaps in an interaction effect with 
another) is now causing problems, but I think it is quite obvious that 
above information shows that it's unlikely. Is using the machine 
frequently for 6 months with GaAs users enough of a demonstration for 
you that I am not "ruining the machine for everyone else". It's not 
unfair to ask me to demonstrate this (though it does take the guilty 
until proven innocent approach), but no one has asked, they've just 
accused. So I invite you to take the Pepsi Challenge: you run your 
season process, etch a test GaAs chip, etch another test GaAs chip, 
measure the etch rates and confirm that everything is normal. Then I 
will run my process on the Si dummy wafer (my season step), then etch a 
PI wafer, then you repeat your part and we see if the results compare 
with those from the first round. Sound good? I reserved the Pquest next 
Thurs, starting at 7am. Does that day work for you? If not perhaps, we 
could have someone else from your group run the GaAs chips. It may be 
easier to arrange things by phone, my cell phone number is 415-269-2974 
. Let's do this, it will go along way toward settling this issue, at 
least with my PI etch process.

--Randy

 From the Pquest archive July 02:

Here is a summary of the etches done since the etcher came back up last
Thurs:

 > PQuest Etches
 >
 > Date & Time     User    Recipe                  Material        
Brown?  Temp    Bias
   Etch Rate
 > 7/25 4pm        true    Ar=40,O2=20             Dummy Si        
No      0 C
 >                         Ar=40,O2=20             Polyimide       
No      0 C     150
V   Normal, >5000A/min
 > 7/25 9pm        sabnis  O2                      Carrier         
NA      85 C
 >                         Ar=15,BCL3=10,CL2=3     Carrier         
Brown   85 C
 >                         Ar=15,BCL3=10,CL2=3     InP             
Brown   85 C    
      10% of Normal, 250A/min
 > 7/26 3pm        gigi    O2                      Dummy Si        
NA      5 C
 >                         Ar=15,BCL3=10,CL2=2     Dummy Si        
No      5 C
 >                         Ar=15,BCL3=10,CL2=2     GaAs            
No      5 C     
      Normal, 1000A/min
 > 7/27 2pm        sabnis  O2                      Carrier         
No      25 C
 >                         Ar=15,BCL3=10,CL2=3     Carrier         
No      25 C
 >                         Ar=15,BCL3=10,CL2=3     InP             
No      25 C    
      Normal, 330A/min
 >                         Ar=15,BCL3=10,CL2=3     InP             
No      25 C    
      Normal, 260A/min
 >                         Ar=15,BCL3=10,CL2=3     Carrier         
No      85 C
 >                         Ar=15,BCL3=10,CL2=3     InP             
Brown   85 C    50
V    10% of Normal, 260A/min
 >                         Ar=15,BCL3=10,CL2=3     InP             
Brown   85 C    50
V    10% of Normal, 290A/min
 >                         Ar=15,BCL3=20,CL2=6     InP             
Brown   85 C    55
V    10% of Normal, 265A/min
 >
Mark Wistey wrote:

>Randy -
>
>Although it's true that several of us have been discussing our
>respective problems with the PQuest, it's certainly not been our
>intention to keep you--or any other user--out of the discussion.  What
>happened was that several of us GaAs folks were having problems with
>etch rates, junk being deposited, etc., but each just assumed it was
>something we were doing wrong.  That has changed over the past
>2 weeks, as we've started swapping notes with each other.  So the
>circle of discussion has been gradually expanding, until it finally
>reached Jim McVittie and, now, you and all the other PQuest users.
>
>It boils down to this: Most of the GaAs users are having severe and
>unrepeatable problems with the PQuest.  I'm not talking about a few
>percent change in etch rates.  I'm talking about wafers that come out
>so contaminated that they're grey.  Some of these wafers (in my case
>in particular) are irreplaceable on the short term.  With all due
>respect, writing comments in the log book doesn't solve our problem.
>Conventional wisdom has been that a long O2 descum, followed by a long
>seasoning, returns the PQuest to "clean" condition.  If you want to
>take a look at my ash-colored wafer, you will see that this is not
>true.  (90min O2 descum, 1+ hour seasoning.)
>
>Anecdotal evidence suggests that the worst of the problems arise after
>someone etches PMMA in the PQuest.  We are trying to be as accommodating
>as possible to you and other users, and as far as I know, no existing
>processes have been banned yet.  Jim McVittie will be doing tests to
>try to reproduce the problems we're seeing, so we can make informed
>decisions from there.
>
>The PlasmaQuest was purchased by III-V people for clean III-V etching,
>and at the moment, it's not serving that purpose.  I don't believe
>it's unfair to ask you to demonstrate that your own process--however
>long it may have taken you to develop--is not ruining the machine for
>everyone else.  I can assure you, we've invested considerably (!) more
>"time and money" developing growth and fabrication processes which are
>worthless while the PQuest is in such a sorry state.
>
>More in another email.  The most-affected III-V people are madly
>preparing for a conference, so email will have to serve as the
>"meeting" for the time being.
>
> - Mark
>
>  
>