durusmail: mems-talk: silicon membranes
silicon membranes
2009-12-21
2009-12-21
2009-12-21
2009-12-21
silicon membranes
Chilcott, Dan - NV
2009-12-23
I agree with the posts below. This is how I would proceed.

If you can tolerate a variation in thickness in the 5 um range then time etch
starting with thin wafers, good temperature and condensation control, and expect
a variation across the wafer. You should get some parts. I would pick KOH and
water at as high of a concentration (20-40 percent) that the mask can stand.
Better thickness control with higher concentration but slower etch rates of
silicon and worse selectivity to oxide.

If you want tight thickness control then use SOI wafers and balance out the
stress. You can use thinner oxides at the SOI interface or match the stress
using deposited films or implants. This technique has been used in high volume.

If you want tighter thickness control and don't want to use SOI wafers then you
have several choices but they will all take some effort.

1) Electrochemical etching of P-type wafers with N-type epi or N diffusion
regions can repeatedly produce diaphragm thickness variations of less than 1 um.
Of course this depends on the etchant temperature and concentration control.
Lots of literature available. This technique has been used in high volume.

2) You can produce 29 um thick diaphragms with a boron doped etch stop. You
first must use a boron doped layer or an epitaxially grown layer and then grow a
lighter doped epi layer to the desired thickness. You will get stress
dislocations in the epi until you reach a certain thickness but 29 um should be
fine. You can also use stress compensated boron doped regions as an etch stop
which will allow you to grow thinner epi regions without stress dislocations but
you would need to order special wafers from epi vendors. You then would need to
pick the proper etchant and conditions to stop on the boron layer. There a few
choices and they can all work. Lots of literature available. This technique has
been used in high volume. EDP has the best boron selectivity.

3) You can create a cavity in one substrate and then bond and thin another wafer
over the cavity. You still need to have an etch stop and you have to do bonding.
This has also been done in high volume.

These are probably the most common ways but there are several more techniques
that have been used.

I would use SOI wafers and balance the stress. I would stay away from EDP unless
you need a high volume process using a boron etch stop. TMAH works well in most
instances but I would only use it if my mask can't take the KOH.

Dan Chilcott
ITT

-----Original Message-----
From: mems-talk-bounces@memsnet.org [mailto:mems-talk-bounces@memsnet.org] On
Behalf Of Albert Henning
Sent: Tuesday, December 22, 2009 4:19 PM
To: General MEMS discussion
Subject: Re: [mems-talk] silicon membranes

I have worked on membranes 10-100 um thick, KOH-etched mostly, 4.5 mm or so
square.

29 um is too thick for boron etch stop, via either diffusion or ion
implantation.

SOI is probably the best bet, even with the anticipated stress issue.  You could
actually compensate for the oxidation-related stresses, using appropriate
implantation, per Brian's description.  (B is tensile stress, As is compressive
stress, if the concentration is high enough.)  If, however, the stress is truly
a problem, then you need to use other means.

Timed etch gives good results, but be aware that if 29 um is your target
thickness, then you will end up with variation of thickness across the membrane
of 29 +/- 3 um or +/- 5 um, depending upon the quality of your starting
material, and the control of KOH stoichiometry and temperature during the etch.

I have heard that timed etch is much easier to control with EDP.  However, EDP
is very tough stuff to work with, from an MSDS perspective, and not everyone is
willing to take the risk.

I have also heard electrochemical etch of n-type doped silicon can be tightly
controlled.

---
Albert K. Henning, PhD
Director of MEMS Technology
NanoInk, Inc.
215 E. Hacienda Avenue
Campbell, CA  95008
408-379-9069  ext 101
ahenning@nanoink.net


-----Original Message-----
From: Brian Stahl [mailto:bstahl@mrl.ucsb.edu]
Sent: Monday, December 21, 2009 12:25 PM
To: General MEMS discussion
Subject: Re: [mems-talk] silicon membranes

Morten is right - heavily boron-doped silicon etches much slower than undoped or
lightly-doped silicon in anisotropic wet etchants like KOH and TMAH, and makes a
nice etch-stop.  There is a significant amount of literature out there on this
subject.  Keep in mind that because boron is a substitutional impurity and has a
smaller covalent radius than silicon, boron doping creates a residual tensile
stress in the silicon.

The more significant issue is creating an etch-stop at your target thickness of
29µm.  I would say that thermally diffusing boron from a solid or gaseous source
would be prohibitively time consuming (email me if you'd like a detailed
explanation).  I can't speak to the feasibility of ion implantation because I
don't know what kind of penetration depths can be achieved with ion beams.

I fabricated ~5um thick membranes as part of my MS thesis research.  I doped one
side of the wafer with boron to create an etch stop, and I measured the etch
progress by periodically removing the wafers from the etchant (refluxing 25wt%
TMAH @ 90°C) and measuring the etch depth with a stylus profilometer.  This
method, or something similar (without the etch-stop), could work for your
application provided that you know the initial thickness of your wafer with an
accuracy of +/- 1µm or so.  Please contact me if you'd like to discuss this
further.

Good luck,

--
Brian C. Stahl
reply