Ed,

This is a very good list!

Some points to add:

HMDS helps when the surface is hydrophilic. If it is hydrophobic it might
hurt. One needs to test water drop contact angle to know what to do.

Metals or organics need other adhesion promoters.

Down to the 0.5u litho, people used 10% hmds in the resist solvent isn a
spinner, the spin dry and spin resist with no bake in between.

There is a condition know as over-priming. The contact angle is very high
and the resist peels off.

Dehydration bake (Vacuum bake for 30 min at 150 deg or 30 min atmospheric at
250 deg) is always a place to start.

Some processes may cause pure adhesion after them. For example, after dopant
diffusion drive-in, the surface need to be treated with boiling water to
remove the excess leftover on the top layer.

Sometime, the resist peels off if it has too much internal stress, even if
the adhesion is good. (If the adhesion is real good the resist will crack)

Shay

-----Original Message-----
From: mems-talk-bounces@memsnet.org [mailto:mems-talk-bounces@memsnet.org]
On Behalf Of Edward Sebesta
Sent: Sunday, July 19, 2009 2:27 PM
To: mazzolari@fe.infn.it; 'General MEMS discussion'
Subject: Re: [mems-talk] improve photoresist adhesion

Resist adhesion failure during develop is a symptom of severe resist
adhesion failure. Something is really wrong.

There are multiple factors to consider in this problem.

0. Is your process working as you intend it? That is, is the hotplate in
the vapor prime step at the correct temperature, are the heating
elements defunct? Is the HMDS really getting to the vapor prime chamber?
In short, look at every element in your HMDS application to see if the
equipment is broken, heater off, a line crimped, HMDS bottle empty (it
happens!), heaters broken, vacuum not pumping down, something where your
process isn't happening.

1. What was your prior chemical step? If it was an HF or Buffered Oxide
Etch or some flouride chemical step, you need to immerse your wafers in
sulfuric peroxide with heat. I would say 50 deg. C or more will do. HMDS
won't overcome the issues with a surface that has been exposed to a
solution with flouride. A 400 deg. C furnace step will work also, but
this is probably not compatible with your process. Regular oven bakes
won't work.

The symptom with flouride adhesion problems can also be that it works
sometimes and doesn't others. And phosphoric acid won't substitute for
sulfuric peroxide.

In the world of MEMS, there might be other solutions that defeat
adhesion.

2. Also, is your surface clean prior to application of HMDS? There are
three parts to this.

   A. Your wafers are dirty from prior solution used, or storage, etc.
Make sure they are clean. Perhaps a peroxide clean if they are delicate,
or a sulfuric peroxide clean if they can take it. O2 plasma should help
if it is organic, do nothing if it is salts or some other inorganic
crud.

   B. Vapor contamination. I had a case where wafers waiting for
positive resist spin sat on a table next to a negative resist developer
that was poorly exhausted and not isolated from the positive resist
wafers that were waiting for resist spin. The Xylene from the negative
resist developer track provided one or two molecular layers and defeated
the resist adhesion of HMDS. We fixed the exhaust and isolated the
negative resist track and resist adhesion was fine. Spilled machine oils
do have a vapor pressure. Other operations can supply contamination.
Poorly exhausted plasma etchers, are another. O2 plasma should take care
of aromatic and aliphatic vapors though. On the other hand, you might be
contaminating them after the O2 Plasma step. Vapor contamination is
quick!

   C. If your prior step was a solvent immersion or some other chemical,
you might have a few monomolecular layers of contamination. A solvent
might be very pure, but you only need a few molecular layers of crud to
not have good adhesion. However, O2 plasma should take care of aliphatic
and aromatic vapors. HOWEVER, it won't take care of salts, inorganic
crude, or some inorganic-organic junk.

3. Was the HMDS applied through a vapor prime process with the wafers
heated, vapor applied? If the HMDS is applied as a liquid dispensed and
spun off, you need to make sure the wafers are well baked afterwards to
make sure the HMDS on the surfact reacts and the surplus is evaporated
off the wafers. This method isn't used much anymore but I personally
solved a resist adhesion problem this way. You might think you are
baking the wafers, but check that your IR ovens are working if that is
your bake method. A hot plate could work also.

4. You don't say what your substrate is, but it could be that the
developer is attacking the substrate or the surface of the substrate.
Developer is alkali.

5. Of course make sure that you have an adequate softbake. To fail
adhesion in this case you would have to be fairly under softbaked.
However, with some MEMS topologies the resist next to the substrate
might not be very solid. Or you need a 110 deg. C softbake to relax the
resist. I don't think this is your problem, but this is what some people
believe, I have always doubted the idea of relaxing the resist with
softbake. But try softbakes changes, the worse that can happen is a
little wasted time.

6. I don't think they supply HMDS diluted anymore. Make sure the HMDS is
100% in the bottle.

7. For non-silicon substrates there might be some other phenomenon going
on. The surface might oxidize and the oxide is readily soluable in
developer. An plastic might have plasticizers bleeding out.

If you do try some other adhesion agent, apply it by vapor and make sure
it doesn't make your substrates look awful.


Ed Sebesta
Semiconductor & MEMS Process Engineer