Page 175 - MEMS and Microstructures in Aerospace Applications

Osiander / MEMS and microstructures in Aerospace applications DK3181_c008 Final Proof page 165 1.9.2005 12:05pm

Microelectromechanical Systems for Spacecraft Communications 165

deposition by low-pressure chemical vapor deposition (LPCVD), followed by
patterning and etching, to create the desired structures on the silicon substrate.
Significant progress has been made in manufacturing commercial-quality mirrors
using these methods.
Stress-free optical thin film surfaces are critical for optical networking as well
as free-space beamsteering applications, but film stress is difficult to control in the
fabrication process. It can vary dramatically with a relatively small change in the
number of atoms, and hence, the film’s chemical composition. As a consequence, it
is difficult to make polysilicon mirrors very flat, particularly if they need to be
relatively large (~few millimeters). After a surface is initially deposited and all the
supporting layers are removed, it may not remain flat. Even thin gold over-coatings
can cause substantial deformation of an uncoated plate.
Bulk micromachining is used to form MEMS microstructures by either wet or
dry anisotropic etching. In this case silicon on insulator (SOI) wafers are useful,
especially in separating moving parts from the bulk silicon structure, and this was
determined early. When a plate-type structure is freed in the fabrication process, a
mirror can be produced on either side, with that surface in contact with the oxide
often being superior in terms of scattering properties. The availability of both sides
allows the deposition of perfectly stress-balanced gold reflection layers for en-
hanced reflectivity, which makes manufacturing easier and more predictable.
Leading candidates for optical switches and cross-connects are free-space
micromirror switch arrays, and a scheme to do this using conventional scanning
mirrors was first proposed as early as 1982. 80 Arrays of collimators are positioned
such that light from each collimator is directed toward a dual-gimbaled mirror.
The first mirror reflects the beam toward a corresponding mirror in the opposing
array. The latter mirrors adjust their angles to send their respective beams to
each receiving fiber. Light from each fiber can only be directed toward its corre-
sponding mirror at a given instant. Likewise, the receiving mirror can only send light
to its associated fiber, but both mirror arrays can be virtually infinitesimally adjusted,
so that any mirror that receives a beam can send it to any of the opposing mirrors,
thereby making fully free connections. The supporting parts of each mirror, such
as the hinges and drive structures, are kept small to maximize mirror area fill factors.
For low-loss transmission the mirrors must be very flat, with flatness better than
one fifth the operating wavelength. Mirrors with gold coatings can have reflectiv-
ities over 98%, and mirror arrays can be several square millimeters in size, with
square or rectangular aspect ratios. Fiber-to-fiber losses through the cross-connect
can be as low as 0.7 dB, and mirrors have been exercised over 60 billion cycles
without any failures. Cross-coupling between the various channels also turns out to
be negligible because even a small amount of angular offset between the input
and output mirrors will cause a significant displacement of the inappropriate beam
at a given output fiber entrance. Small-scale cross-connects with fewer optical
switches have switching times as low as 50 ms or less, although larger N N
switches, configured into 2-D crossbar arrays, have switching times on the order of
500 ms.

© 2006 by Taylor & Francis Group, LLC