I've got two questions regarding isotropic etching of <100> wafers:

The situation:

We're etching v-grooves on 6" <100> wafers.  The mask lines are 148um
center-to-center, and are 10 microns wide.  We need to etch all the
way to the bottom of the v-groove, and we need the <111> surfaces to
be very flat and very regularly spaced.  (This is a grating for an
optical application, so grating efficiency is important.)

Right now we're using KOH etchant on 6000A SiO2 coated wafers.  This
limits our etching time since KOH degrades SiO2, but for our test
wafers (using a 25um mask) this has worked reasonably well.

The problem:

We've run into problems with the 148um line spacing.  Etch times for
30% KOH at 60 C are around 4-5 hours, and by the time we reach the
bottom of the grooves the <111> planes look pretty ragged.

Ideas:

One problem I can identify right now is that we aren't as closely
aligned as we'd like.  We used an alignment technique with the 25um
mask that worked quite well, but with the coarser mask it doesn't work
as well (more on that in a sec.)  We need to align our mask to the
lattice better than we are.

Another problem is that we got the 30% concentration and 60 C
temperature from another paper on KOH etching, but really don't know
if those are ideal for our purposes.  I've spent the past few days
combing a number of journals, hoping to find an article with a
parametric survey for KOH with respect to the <111> surface quality,
but haven't had any luck.

A final idea is that KOH may not be the etchant of choice.  We've
looked into TMAH to some extent, but haven't tried it yet since
switching etchants would set us back by quite a bit.

Questions (now for the fun part):

Is there a better way to align our mask to the lattice?  Here's how
we've aligned in the past:  We used a circular mask to etch a small
pyramidal pit in the silicon, then went back and re-coated the wafer.
By shining a laser on the pit, we can get a reflective pattern back
from one of the pit facets (a <111> plane).  The mask returns a
linear diffraction pattern.  Lining up the two patterns lets us get
very nice alignment.  The problem is that with the coarser mask (148um
as opposed to 25um), by the time we illuminate enough of the mask to
get a strong diffraction pattern, it's so spread out that the pit
doesn't return anything.  Catch 22.

The only constraint we have is that we've only got about 1" x 1" on
the wafer surface that we can sacrifice to alignment marks.  I'd be
interested in splay patterns provided we can get one that small that
would still be useful.

Next question:  Is KOH really the etchant we want to use for this?
Primarily we need very very clean <111> surfaces, and we need the
spacing of the facets to be very very regular.  Our mask is good down
to better than 0.01um (don't ask), so if we could hold that tolerance
we'd be set.

Next question:  Is there another concentration and temperature we
could use that might yield better results?

Final question:  Is there anything else we could do that we might not
know about to make this work a little better?

Any advice or references would be a great help.

Thanks,

Tom Benedict                          McDonald Observatory at Mount Locke
[email protected]               Department of Astronomy
Tel: 512-471-3337                     University of Texas at Austin
Fax: 512-471-6016                     Austin, Texas  78712