MEMS (micro-electromechanical-systems) technology is set to
explode into a number of new and different areas, including RF,
optoelectronics, and biomedical applications. MEMS, however,
are not new. MEMS research and development has been going for
decades throughout the world. During the early 1970s,
bulk-etched silicon wafers were being used to commercially
produce pressure sensors and in the early 1980's experiments
began using surface-micromachined polysilicon actuators for
disc drive heads. Analog Devices began shipping the first fully
integrated, single-chip MEMS accelerometers for automotive
airbags in 1991.
While MEMS manufacturing leverages IC manufacturing, the
principal users of MEMS are expected not to be limited to the
traditional electronics and computer companies. Instead system
integrators and commercial or defense product manufacturers who
make automobiles, scientific analytical instruments, consumer
goods, medical devices, aerospace navigational systems, and
similar products will benefit from MEMS technology.
Boom Markets Predicted
According to Cahners In-Stat Group MEMS devices accounted for
$4 billion in revenue in 2000 and by 2005 the MEMS market will
quadruple to $11 billion. In addition, MEMS-based photonic
switches are predicted be the first MEMS device to surpass $1
billion by 2004. In the past year, 30 new optical-switch
companies have emerged, garnering $1.2 billion in funding,
according to consultant Jeff D. Montgomery of ElectroniCast.
Consultant Roger Grace estimates that over 600 university,
government, and industry labs are currently working on
One of the first MEMS applications was a tire pressure sensor.
MEMS are now used in many more automotive applications ranging
from airbag accelerometer sensors to fuel sensors, engine and
brake controls, and noise cancellation. Another innovation is
"intelligent tires" that alert the driver to a flat.
Multi-axis accelerometers, inkjet printheads, scanners,
gyros, force and displacement transducers, and a wide range of
electromagnetic devices such as solenoids can be made using
MEMS techniques. Tiny inertial sensors fabricated out of
silicon, which can withstand more than 15,000 g, show promise
of converting "dumb" artillery shells into "smart" guided
munitions. A wide range of MEMS devices for RF applications
include switches and relays, inductors, transmission lines,
filters, antennas, and transformers. MEMS devices for
optoelectronics include 2-D and 3-D micromirrors, optical
attenuators, and fiber alignment and packaging aids. MEMS will
be used in microscopic assembly lines to build the next
generation of optical communications equipment, biotechnology,
and micro-electronic and micro-mechanical devices. Microfluidic
devices with miniature pumps and valves could serve as small
refrigerators that could remove heat from electronics.
The pharmaceutical industry is adopting MEMS devices rapidly
for testing new drugs. Another potential medical application is
blood-screening sensors that can perform complete lab tests at
bedside. Biomedical applications include networks of channels,
pumps, valves, and mixers for analytical devices. MEMS can also
be used as molds for plastic microfluidic parts. MEMS devices
are envisioned for miniature surgical tools, fluid dispensing
heads, and drug delivery, and implantable sensors.
Government-Sponsored Research and Development
During the 1990s the U.S. government began sponsoring MEMS
projects. The Air Force Office of Scientific Research (AFOSR)
was supporting basic research in materials and the Defense
Advanced Research Projects Agency (DARPA) started its foundry
service in 1993. Additionally, the National Institute of
Science and Technology (NIST) began supporting commercial
foundries for CMOS and MEMS devices. Government interest in
MEMS continues, with significant ongoing funding through
agencies such as DARPA. The Japanese and European Governments
have increased their support of MEMS R&D to a combined
level of over $70 to $100 million a year. The combined European
and Japanese industrial annual investments in MEMS R&D are
estimated to be over $200 million and growing.
MEMS Produced by a Variety of Means
MEMS manufacturing takes advantage of high volume batch
processing based on semiconductor techniques. There are many
different processes currently being used to manufacture MEMS.
They include surface micromachining, bulk micromachining,
electro-discharge micromachining (EDM), and high-aspect-ratio
micromachining (HARM) technologies such as LIGA (a German
acronym for Lithographie, Galvanoformung, Abformung or
lithography, electroplating, and molding).
Silicon surface micromachining uses the same equipment and
processes as IC manufacturing, so it was one of the first
processes adapted to MEMS. Applications for surface
micromachining include actuators and electrostatic motors as
well as mirrors and accelerometers. Polycrystalline silicon is
a good material from which to make MEMS. Polysilicon has a
strength of 2 to 3 GPa, depending on surface flaws, while steel
has a strength of 200 MPa to 1 GPa, depending on the process
parameters. Polysilicon is extremely flexible; its maximum
strain before fracture is ~0.5% and it does not readily
Figure 1: This microphotograph shows a MEMS
Precision Instruments HEXSIL polysilicon tweezers gripping a
polysilicon gear. The tweezers can manipulate micron-scale
In the LIGA process, polymethyl methacrylate (PMMA) plastic
is selectively exposed to radiation through a mask. This allows
some of the PMMA to be washed away, leaving structures that are
then electroplated with metal. The structures can be the actual
MEMS device or can be used as molds. The LIGA technique has
been used to produce electrostatic motors and gears.
Figure 2: The Sandia National Laboratories' SUMMIT
four-level poly process allows many gear stages to be chained
into a transmission to give the desired reduction. The output
gear engages the rack on the side of the tensile tester
The TiNi Alloy Company is developing heat actuated
shape-memory alloy (SMA) thin film microdevices. The initial
application of these microactuators is in miniature valves, but
other potential applications include miniature connectors,
switches, and end effectors for microrobotic manipulators. SMA
film actuators can produce large forces and displacements
within small spaces at voltages compatible with
Argonne National Laboratory researchers have developed a
process for growing diamond film that promises to bring the
superior mechanical, tribological, and thermal properties of
diamond to MEMS. The properties of silicon are not suitable for
some potential MEMS applications. This is especially true for
devices that require extensive sliding and rolling contact,
such as micromotors for aerospace applications, because silicon
wears too quickly.
MEMS Miniaturize RF Components
The wireless industry is faced with a number of tough design
challenges. A 3G "smart" phone, PDA, or base station, could
require as many as five radios for TDMA, CDMA, 3G, Bluetooth,
and GSM. Such additional functions entail an increase in
component count. Yet at the same time the industry must meet
consumer demand for smaller form factors, lower costs, and
reduced power consumption.
RF MEMS offer reductions in size, power, and signal loss
thereby extending battery life and reducing weight.
Applications for RF MEMS include RF front ends, microwave
filters, antenna arrays, phase shifters, transmit/receive
switches, inductors, resonators, transmission lines, and
Analog Devices and Cronos now make MEMS relay/switch devices
that combine the low-loss and signal-fidelity benefits of
electromechanical relays with the small size and low power
consumption of solid-state devices. Recent measurements have
demonstrated less than 0.5 dB of loss and greater than 30 dB of
isolation for signal frequencies up to 10 GHz.
MEMSCAP's MEMS-based RF switch uses the company's
proprietary membrane process to create a low-loss, low-power
device. The RF switch is composed of a movable metallic
membrane. When an electrostatic force is applied, the membrane
is pulled down to complete the circuit. The "above-IC"
configuration can offer higher isolation and lower power
consumption than electronic switch technology. MEMSCAP's
MEMS-based tunable capacitor also uses the membrane process to
create a device offering an improved high Q value and tuning
range compared to traditional technology.
Miniature passive RF devices are becoming available too.
MEMSCAP's MEMS-based "above-IC" high-Q inductor designs achieve
cost reduction, high integration density, and performance
improvement using IC-compatible processes. These inductors
eliminate discrete inductors from the circuit board and place
them on top of the chip, helping to shrink form factors.
MEMS in Optical Communications
Significant progress has been made in increasing the amount of
data that an optical fiber can carry with the amount of data
that can be transferred doubling every eight months. This data
increase sprung from advances in DWDM (Dense Wavelength
Division Multiplexing) technology. DWDM enables the
transmission of a large number of wavelengths over a single
fiber"currently up to 160 discrete channels, with 1000 channels
under development"and increased transmission speed from OC48
(2.5 Gbps) through OC192 (10 Gbps) to the emerging OC768 (40
Gbps) and planned 100 Gbps.
Switching this data is traditionally done by converting the
optical signals to electrical signals, switching them
electronically, and then converting them back to optical
signals (optical-electronic-optical or OEO). Today's OEO
electronic switches throttle the wide bandwidth advantage of
the optical fiber. All optical switches eliminate this
MEMS is an enabling technology for fiber-optic applications
frequently referred to as micro opto electro-mechanical systems
or MOEMS. MEMS-based mirrors, shutters, and fiber aligners are
used in optical switches, tunable filters, tunable lasers and
variable attenuators. Some examples of the network
architectures that can utilize optical add/drop multiplexers
(OADMs) are: linear add-drops for backbone DWDM networks,
hubbed rings in metro access networks, and a logical mesh ring
that allows dynamic path reconfiguration. Large port-count
optical cross-connects (OXCs) will be used in central offices
for dynamic remote network provisioning.
Umachines has introduced a 2-D 1x2 MEMS fiber optic switch.
The MEMS 1x2 switch measures 38 x 16 x 11 mm and provides
channel selection between one input fiber and two output fibers
for a number of network functions, including protection,
Reconfigurable Optical Add/Drop Multiplexing (ROADM),
monitoring, and provisioning. This MEMS switch avoids both
stiction and frictional wear, resulting in a switch that has
successfully cycled over 100 million times without any
detectable change in insertion loss. The device features low
insertion loss (<0.4 db="" typical,="" 0.6="" db="" maximum),="" low="" polarization="" dispersion="" loss="">0.4><0.1 db="" maximum)="" and="" low="" power="" consumption="" (6="">0.1>
Whenever an optical signal is transmitted through free air,
it runs the risk of being distorted when it moves through
turbulence in the atmosphere. If uncorrected, this turbulence
limits the useful range of the signal as the strength of the
signal degrades with distance. A CMDM (Continuous Membrane
Deformable Mirror) from MEMS Optical can correct for this
turbulence in real time greatly extending the useful range of
MEMS Optical also makes moving mirrors for tunable lasers.
When put inside the laser cavity, this mirror can be moved by
up to 20 µm. Actuated by MEMS Optical's patented vertical comb
drive, this allows a simple and effective way of changing the
path length inside the laser cavity.
Micralyne makes Silicon V-Groove Chips that are used to
align fiber optic cable with either an active laser device or
arrayed waveguide (AWG). Micralyne offers standard chip
products for 2, 4, 8, 16, and 32 fiber arrays and manufactures
customized chips based on customer specifications.
Silicon Light Machines, a wholly owned subsidiary of Cypress
Semiconductor, has developed Grating Light Valve (GLV)
technology which is used for creating a high-performance
spatial light modulator on the surface of a silicon chip.
MEMSCAP's variable optical attenuators (VOAs) are based on
the company's switch technology. They use a miniaturized
shutter mechanism located between optical waveguides or fibers
that mechanically blocks the light path, reducing the amplitude
of the signal without distorting the waveform. Further,
MEMSCAP's tunable filters, expected to ship in the second half
of 2002, are based on the principle of having one of two
mirrors fixed to an actuator, enabling variation of the
distance between the two mirrors forming the optical cavity.
These filters are expected to have a steep cut-off to avoid
crosstalk from adjacent channels. In addition, they are
expected to be thermally stable with low insertion loss.
Texas Instruments' DLP (digital light processor) was
originally developed for displays but is being adapted to
optical-communications switches. It is based on a MEMS device
called the Digital Micromirror Device (DMD). TI's DMD is a
fast, reflective digital light switch that is fabricated by
CMOS-like processes over a CMOS memory. Each light switch has
an aluminum mirror that can reflect light in one of two
directions depending on the state of the underlying memory
cell. CiDRA has demonstrated its AgileWave dynamic spectral
equalizer that uses TI's DLP technology.
MEMS in Biomedical Applications
Microfluidics, a MEMS technology, enables the fabrication of
networks of channels, chambers, and valves to control the flow
of liquids in amounts as minute as one picoliter. These systems
have few moving parts and require little assembly. They offer
the potential to miniaturize analytical equipment that uses
expensive chemicals and DNA samples. They take advantage of
physical phenomena such as electro-osmosis, dielectro-phoresis,
and surface interaction effects.
Figure 3: Micralyne uses MEMS techniques to make
channels in a substrate for microfluidics applications.
Electrokinetic flow is generated when electrodes attached to
computer-driven power supplies are placed in the reservoirs at
each end of a channel and activated to generate electrical
current through the channel. Under these conditions, fluids of
the appropriate type will move by a process known as
electro-osmosis. Typical flow rates within the channel are
about a millimeter per second and the flow rate can be
controlled with a high degree of precision. Another
electrokinetic phenomenon known as electrophoresis occurs in
the microchannels. This is the movement of charged molecules or
particles in an electric field. Electrophoresis can be used to
move molecules in solution, or to separate molecules with very
subtle differences. Pressure can also be used to move fluid in
the channels. On the microfluidic scale, small amounts of
pressure produce highly predictable and reproducible fluid
Another way to connect to the molecular world is through
control by light. When a molecular design is at the scale of
the wavelength of light, interesting quantum behaviors
emerge"for example, quantum dot lasers that emit light and
bandgap crystals that switch light. Arryx fabricates 10,000
independently controllable "tweezers" that can manipulate
molecular objects in 3D (move, rotate, cut, place), all from
one laser source passing through an adaptive hologram. With
thousands of miniature robot arms, Arryx can sort cells and
proteins as well as manipulate the organelles and DNA inside a
living cell. Arryx's first product is the BioRyx 200 system.
Future products will include dynamically configurable biochips,
cell sorters, purification equipment and optical switch/router
Micralyne makes the Microfluidic Tool Kit, a
user-configurable instrument that is being used in corporate
and academic research laboratories for customized
bio-analytical applications in protein, DNA, and cellular
Verimetra has introduced the "Data Knife," a suite of
surgical tools that incorporate sensing and measuring devices.
The Data Knife combines sensing and data-gathering capabilities
on the edges of various surgical tools. These instruments are
capable of distinguishing tissues, such as cartilage, bone,
muscle, and vascular, and also of measuring tissue properties,
including density, temperature, pressure, and electrical
Figure 4: This Zyvex fully released X/Y stage is
pushed or pulled around by the connectors on its four sides.
After moving it to the desired location, the connectors can
be moved out of contact with the moveable portion of the
Advances in Processing
Eliminating the need for a conventional semiconductor
clean-room fab, MEMGen offers a single tool called an EFAB that
rapidly creates 3-D micromachines from a variety of materials.
EFAB (Electrochemical Fabrication) uses a patterning technology
called Instant Masking to generate microstructures quickly and
without the need for photolithography. Instant Masking makes it
possible to rapidly deposit an unlimited number of
independently patterned layers. Together, these layers form
virtually arbitrary, complex 3-D shapes, overcoming the
geometrical limitations of conventional microfabrication.
MEMGen has used its EFAB technology to fabricate a micron-scale
device having 38 metal layers.
Things to Come
Improved simulation and modeling tools for MEMS design are
urgently needed because currently MEMS design is often a
trial-and-error process, requiring deep knowledge of the
micromachining process employed. Also MEMS packaging presents
unique challenges compared to IC packaging. Where conventional
IC packaging serves to isolate the chip from the environment,
MEMS devices often must be in continuous and intimate contact
with their environment. Consequently a new and specialized
package must be developed for each new device. In fact,
packaging is often the single most expensive and time consuming
task in a MEMS product development program. Packaging
simulation tools need to be developed as well as standardized
About the Author
Charles H. Small is a technical editor based in Waltham, MA.