Dear Jay,

   I had a similar problem for Cr/Au on polySi. I was able to improve
the adhesion of evaporated Cr/Au on c-Si and polySi in 49% HF by one
of more of the following methods (based on internal discussions and
memstalk replies, esp. the one from Stafford Johnson of MEMScAP a
month or so ago...):

1. depending on your lithography process, you want to make sure your
silicon surface is scum and oxide free. You can do this by making sure
the surface is clean to start with (RCA1 or Piranha does the trick
very nicely); then an oxygen ash that removes ~5-10% of your resist
thickness, then a quick BOE dip before evaporation.

2. Going to a lower base pressure before evaporation seemed to improve
the adhesion of the metal on Si. I noticed a substantial improvement
going from ~4 uTorr to 0.1 uTorr.

As far as I can tell based on numerous conversations with other
experienced processing guys, Cr is the only commonly used adhesion
metal that is not attacked by HF. I suspect from your observation of
the metal lifting after only 3 minutes, that it is a scum or oxide
layer that is being etched (causing the metal to lift), since the
thermal oxide would presumably take longer than 3 minutes to release
(unless you happen to have a very thin oxide in your SOI wafer)?

Good luck.

Felix
Applied Quantum Technologies
Durham, NC

On Apr 30, 2010, at 1:33 AM, Jie Zou wrote:

> Hi folks,
>
> I defined a gold pattern by e-beam lithography and lift-off. The
> gold was
> evaporated on the SOI with ~8nm Cr as the adhesion. After several
> other
> fabrications (mainly DRIE), I had some free-standing gold structures
> sitting
> on a silicon structure made by the device layer of a SOI wafer. My
> last step
> was to release the whole structure by HF etching the buried oxide
> layer. I
> used 49% HF for 3min to achieve the necessary undercutting to make
> sure my
> structure was free. However part of the gold structure was peeled off.
>
> How can I improve it? Is there another adhesion layer that can
> resist HF
> better? I tried BOE instead of 49% HF. It didn't help much.
>
> Thanks a lot.
>
> Jay

Felix Lu
felix_lu@yahoo.com