Chris-

I strongly suspect what is happening is that the nitride mask
on the "bad" channels has broken off during the etch.  As the
channel etches it severely undercuts the very thin mask, which
can break at any time.  Once the overhanging mask breaks off a
channel- particularly if it happens late in the process- the
etch
rate increases significantly, resulting in a wider and somewhat
deeper channel.  And because it could happen at any time, you
end up with channels of various widths.

A standard nitride mask is under tremendous tensile stress;
when
the overhanging portion breaks it snaps off the entire channel
and curls up at the ends, just like a string breaking on a
guitar.
The fractured edge of the mask follows the edge of the channel
almost perfectly, which is why the channel ends up uniformly
wider.

I have seen this happen on my wafers. To verify it on your
wafers,
try agitating more vigorously to "break" even more channels.

About the only way I have found to control this problem is to
lower the stress of the mask film.  It may also help some to
round
the ends of your mask features so as not to  concentrate the
stress
in sharp corners.

I hope this helps you, and good luck.  If you want to discuss
it further, drop me a line.

--
Marc Straub
Visteon Automotive Systems, Ford Motor Company
Dearborn, MI
mstraub1@ford.com

-----------------------------------------

On Mar 11,  5:10pm, Chris Turner wrote:
> Subject: Isotropic etch problem
>      Hello,
>
> We are producing a micro-chemical reactor module that
> requires isotropic etching of 100 micron wide channels in
> silicon.
>
> We do this using a standard HF/Nitric/Acetic acid mixture
> with a silicon nitride masking layer. This results in several
> of the channels etching differently to the rest. The
different
> channels are slightly wider by about 5-10 microns and have
> a much rougher, almost crystalline, surface finish. There are
> 120 channels on a wafer and between 1 and 20 per cent can
> be different. This effect runs the whole length of the
channel,
> but neighbouring channels can be unaffected.
>
> Has anyone seen this sort of thing before and if so is there
a
> way of preventing it?
>
> Any help gratefully received.
>
> Thanks
>
> Chris Turner
> Senior Research Engineer
>
>      =========================
>      Central Research Labs
>      Dawley Road
>      Hayes
>      Middlesex
>      United Kingdom
>
>      Tel. +44 (0)181 848 6465
>      Fax. +44 (0)181 848 6442
>      e-mail cturner@crl.co.uk
>      Web. www.crl.co.uk
>
>
>-- End of excerpt from Chris Turner