Andrea,

At Redwood Microsystems, we used a combination of timed KOH anisotropic
etching, and timed HF:nitric:acetic isotropic etching, to achieve 4.5 mm
square membranes for microvalve production, with thickness between 3 and
60 um thick (wafer-to-wafer:  each individual wafer had the same
membrane thickness target), processed on 100 mm Si lightly boron-doped
wafers.  We were able to achieve +/- 1.0 um thickness variation across a
wafer (0.5 um on occasion, but that was difficult).

Initial wafer thickness uniformity was very important.

Like the other responder from IMTEK, we used roughly the same KOH
temperature and concentration.  The bath was either stirred, or the
wafers were moved mechanically on a paddle assembly.  We did look at
ultrasonic agitation, but you must use the correct acoustic frequency in
order to avoid membrane fracture:  resonances can be set up in the
membranes otherwise.  FEA can be used to find the resonances for your
particular design.

We used the HF:nitric:acetic etch, in order to remove sharp corners
after KOH etch.  Unless one performs this step, the stress in the sharp
corners easily exceeds the fracture strength of the crystalline Si.  So,
this became the critical etch, because it is a triple-constituent bath,
and control of the stoichiometry is very important.  To achieve this, we
used a commercial, pre-mixed source for the chemical, and a mechanical
paddle to move the wafer through the bath.

Metrology was important.  We used a laser interferometer to measure the
membrane thicknesses across the wafer, in order to set etch times and/or
membrane thickness targets.

In sum:  temperature control is extremely important, probably most
important; followed by control of reactant concentration/stoichiometry.

Al Henning



Andrea Mazzolari wrote:
Hi all,

i need to realize silicon membranes having thickness 14um and lateral
size
5x5 mm in 4'' silicon wafers, starting thickness is 300um.
I can not use any stop layer or SOI wafers.

I already realized some membranes using KOH etch, but thickness is not
so uniform (thickness variation is about 4-5um).
I tried to use magnetic agitation, but it did not helped much.

Any suggestion on how to improve thickness uniformity ? I tried also
ultrasonic agitation, but membranes went broken along the (100) planes.

Best regards,
Andrea