The UC Berkeley Microlab recently had a query from another university research
group for a low-cost method of eliminating particles on wafers prior to anodic
bonding. Perhaps this method is useful to MEMS?

In addition to Microelectronics, the University of California Dept. of
Electrical Engineering and Computer Sciences (EECS) does significant research
in opto-electronics. Historically, large numbers of custom lasers were
fabricated for use by EECS as well as other research groups. Most of these
lasers required brewster windows; windows free of films and particles. A
laminar flow workbench was used for assembly.

The final step in a protocol which yielded laser quality windows consisted of
coating the optical surfaces of brewster windows with collodion and then
peeling the resulting film away. Collodion is a solution of nitrocellulose in
ethyl ether. A cotton swab is saturated and a layer of collodion was worked
evenly over the optical surface. This coating is allowed to dry for 30-60
seconds, then peeled off using office adhesive tape (the old kind, not "magic"
type). Fine particles are bound to the collodion and are removed with it. A
radioactive, Po210 alpha source in the shape of a bar, mounted on a gooseneck
was used to neutralize static charges created by the removal if the collodion.
This alpha source may not be necessary for success.

Laser windows in operation show even small numbers of particles. Although other
techniques were tried, no better method of window preperation was found.

The collodion method of particle removal was used in the late 1970's to bond
clean, thinned Si wafers to Corning 7070 low sodium glass. The completed glass/
Si structures were used for solar cell research. The term for this type of
bonding was field assisted bonding (FAB), coined in a NASA technical paper.
Today the term has been edited to "annodic bonding".

Sincerely,
Bob Hamilton
o