The time and power of the plasma is definitely critical (although I
don't know the exact chemistry either). Try exposing for just a couple
seconds (I do 30 W for 10-15 seconds and get inseparable bonds) or even
flashing it with plasma for a second or two. Another thing I noticed is
that the glass is very sensitive to the plasma as well and over-cleaned
glass won't bond. Typically, when I do lithography on glass, I clean the
glass in strong oxygen plasma (350 W for 3 minutes) (it tends to help
with SU-8 adhesion). But when I cleaned my glass slides in this way, the
PDMS didn't bond to it. So to ensure that I have a "fresh" surface of
glass to bond to, I usually give the glass a quick HF dip to expose
fresh material for bonding, and then the glass gets the same 10-15
second plasma exposure as the PDMS.

I hope that helps,
Joe

Gareth Jenkins wrote:
> Hi
>
> I already tried the HCl cleaning suggested by that paper but thanks
> anyway!
> Maybe I'll also try leaving it for a while on a hotplate as I always
> tend to test the bond a minute or so later.
> However, I'm fairly sure my problem is my unreproducible plasma
> conditions. In Beebe's paper they found optimum values for RF power
> and pressure but I have no control over these in my plasma and am not
> even sure what they are (probably 50-100W RF power but there is no
> pressure gauge).
> I have a feeling I "over expose" the surfaces as sometimes the bonding
> is much worse than if no treatment was done - normally you would
> expect some adhesion between clean glass and fresh PDMS but there are
> instances when it doesn't even bond well enough to prevent it falling
> off if I turn it upside down. I have tried shortening the time but
> never seem to find a reliable process.
> Does anyone know the chemistry of what happens if it gets over-cooked?
> I know there is the generation of silanol groups which allows the
> bonding to occur but there must be something else which inhibits the
> bonding if it is over exposed.
> Maybe if I know what it is I can do something to prevent it.
> Best regards
>
> Gareth
>