Download mems

MEMS —
A Small World with
Big Opportunities
SwRI engineers take on the big challenge
of working on the very small scale
By Heather Hanson
I
n modern technology applications, sometimes smaller is better — and microscopic is
better yet. Microelectromechanical systems
(MEMS) have moved out of the basic research
arena to become a useable technology.
What are MEMS? The short answer is microscopic machines; however, they are much more
diverse than that. The current driver in MEMS
applications is the ability to make an existing
device microscopically small, or to create a new
device that would not work if it were inches in
size but that works well at the micron scale. MEMS
devices also potentially can be manufactured for
low cost at high volumes, mirroring the semiconductor industry upon which they are based.
MEMS are an enabling technology, a building
block for solving problems in nearly every technical field. They often are used to make sensors,
including the air bag sensor in most modern automobiles. In other applications they interact with
their environment to change it in some way, such
as an ion propulsion device that can move tiny
satellites in space or an optical system that diverts
light beams. Sometimes, they interact with themselves, such as the timed-locking mechanism on a
nuclear warhead. Such a mechanism comprises
6
Technology Today • Winter 2004
D1M014463_1533
An SwRI-developed MEMS device undergoes testing
in a probe station. The one-centimeter-square chip
contains more than 60 different mechanical devices.
The microprobe tips provide drive signals and sensing
connections to the chip.
MEMS methodologies
D014463_1578
MEMS devices are fabricated using
methods originally created to process the
small features necessary for integrated
circuits. These processes require a layerbased design — meaning that threedimensional structures are created by
Heather Hanson, shown in front of a scanning
electron microscope, is a senior research engineer in the Medical Systems Department in the
Automation and Data Systems Research
Division. Hanson has designed mechanical
components, assemblies, packaging and fixtures for single unit, low volume and high volume
production. Her work in microelectromechanical
systems includes actuator and sensor design
and development.
RELIEVING STRESS
MEMS technology detects crack growth
Many engineering structures, such as pressure vessels, pipelines, bridges,
boats and aircraft, are subjected to stress in a corrosive environment. Because
these conditions can cause structural failures, experimental models have been
developed to determine a structure’s susceptibility to stress corrosion cracking
(SCC) and to measure the rate of crack propagation. One limitation of these
models is that crack propagation may occur at a rate slower than the current
limit for measurement. Rates below that level are too small to accurately predict whether SCC may lead to component failure during the structure’s
expected life cycle, given current test methods. However, small, lightweight
sensors installed in critical areas
of an aircraft could alert maintenance crews to replace a part
before failure occurs.
Using MEMS technology, a
miniature SCC test beam can be
fabricated out of the structural
material of interest and mounted
on the MEMS chip. Such a test
beam can detect crack growth
based on either a resistive or
capacitative measurement as a
static stress load is applied to the
test beam prior to installation,
and either a static or a cyclic
load is imparted to the beam
Numerous cases of structural failures from
during testing.
stress corrosion cracking (SCC) have been
reported in boilers, pressure vessels, oil and
Engineers at SwRI have built
gas production and transmission piping,
a system that can detect the
nuclear power generation components,
small change in resistance that
bridges, sea craft, aircraft and more. SwRI
results from a MEMS-based SCC
developed a MEMS device using structural
sensor made of brass, operating
engineering materials for sensitive crack growth
in a corrosive environment. The
rate measurements that facilitate SCC sensing
system’s usefulness depends on
and monitoring. A U.S. patent is pending.
the structural materials available
for the MEMS fabrication process.
As a crack grows, the cross-sectional area of the test beam is decreased.
This change in cross section affects the resistivity of the test beam, which can
be detected by the sensor. Because the test beam is so small, a small crack has
a relatively large effect. This method is useful provided that the structural material is electrically conductive.
To make a capacitive measurement, a conductive plate is attached to one
end of the test beam. This plate overlaps a series of stationary conductive
plates on the chip. After a constant load is applied, the conductance can be
measured between the test beam and the stationary plates to determine the
position of the end of the test beam. As a crack begins to form, the test beam
weakens and the applied load causes it to move more than it did initially. This
additional motion is correlated with the crack growth to obtain a growth rate.
Again, because the test beam is so small, a small crack has a relatively large
effect on the beam’s motion under a given load.
The MEMS fabrication process used by SwRI engineers is derived from the
semiconductor industry and is material-dependent. The established processes
are designed for conductive materials that have a minimized corrosion risk.
The fabrication processes would have to be improved for materials such as
aluminum and stainless steel before the sensor would be commercially viable.
SwRI is pursuing funding to achieve this goal.
Technology Today • Winter 2004
D013892
gears and linkages that can open a switch
with the correct electrical input. Because
they are micro-scale, they can be installed
in small spaces. They also can interact
with molecules, opening up a new realm
of possible chemical, biological and medical applications.
MEMS researchers at Southwest
Research Institute (SwRI) outsource the production of MEMS, enabling them to select
the process steps best suited to the design
of a particular device. The selection of a
fabrication facility is an early part of the
design process. This process determines
such requirements as layer thicknesses,
number of layers, materials, minimum feature sizes, residual stress and final package
size. As a multidisciplinary research and
development organization, SwRI provides
the wide range of technical expertise
needed to design, develop and package
MEMS devices for a variety of applications.
7
IN MOTION
MEMS actuators fabricated for optical
and mechanical applications
D014584_9398
One of the key abilities of MEMS devices is that they
can produce mechanical motion on a very small scale.
This provides the means to create compact actuators with
motions in the range of nanometers to millimeters and with
the ability to generate forces up to the milli-newton range.
MEMS actuators are generally based on either electrostatic
or thermal operation, although piezoelectric, magnetic and
even hydraulic types also have been demonstrated. An electrostatic actuator relies on the potential difference between
a movable and a fixed surface and exerts a force on the
movable surface. These actuators are very low power
(microwatts) but require high voltage, sometimes in excess
of 100 V, to generate a sufficient electric potential. Their
response time is fast — tens to hundreds of microseconds
— but they can be limited in their maximum displacement. Such actuators have a “pull-in” voltage limit
whereby they fully deflect after moving a certain portion of
their range, making this type of actuator most suitable for
binary positioning. The well-known comb drive actuator is
of this type. Thermal actuators, on the other hand, use Joule
heating to produce thermal expansion. They have only a
slightly slower response time than electrostatic actuators
and can be operated over their entire range of motion.
One of the first MEMS projects at SwRI was an internally funded development study of several of these
micro-actuators and their use in switching applications.
Actuators were designed based on both electrostatic and
thermal methods, with a variety of force-displacement
characteristics and the ability to produce both horizontal
and vertical motion. Also developed were two novel actuators that have since been patented by SwRI — a singlematerial, in-plane, bi-directional thermal actuator and a
vertical thermal actuator based on a bi-metal design.
One switching device produced with the actuators
from this project was an optical switch. It was designed to
switch light between any of several input collimated or
parallel optical fibers to any output fiber. The switch was
based on a two-dimensional array of micromirrors
mounted on the end of electrostatic out-of-plane actuators.
The mirrors, which are smaller than the head of a pin,
were fabricated flat, but were flipped and latched into
position by a hinge mechanism on the end of the actuator
arm. The switch was scaleable to allow additional input or
output fiber channels. Such a switch can be used as a
cross-connect in fiber optic communications systems or in
optical scientific instrumentation.
Several sponsored programs for development of
unique microstructures and sensors have benefited from
this internal research. Some of the actuator designs have
been used directly in the development of the SwRI “materials lab on a chip.” The handling and testing techniques,
design procedures and measurement tools developed
under the internal research project laid the groundwork for
many subsequent MEMS projects at SwRI.
Although this MEMS device has more than 100 functioning mechanical devices, the entire chip
can rest on a fingertip. MEMS devices are incredibly small yet highly functional.
8
Technology Today • Winter 2004
stacking layers of material(s). Initially the
materials were polysilicon and metals, but
the technology has progressed to include
numerous other materials, including polymers. Typically, the layers range in thickness
from 1 micron to 1,000 microns, depending
on the fabrication process. The diameter of
a human hair is around 90 microns, fitting
easily in the realm of MEMS.
The most common processes in use
today involve photolithography, or etching a
design into photosensitive materials. In this
process, a structural material of constant
thickness is deposited onto a substrate
chip. A photoresistant material is applied
onto the structural material in a particular
pattern of interest. The structural layer is
then etched according to the photoresist
pattern, and the photoresist is removed.
This process is repeated until the desired
layers have been placed. Some of the
materials deposited throughout this process
are “sacrificial.” When the deposition and
photolithography processes are complete,
the entire chip is exposed to an etchant,
which removes the sacrificial layer(s), thus
making room for the other layers to
move. There are many variations of this
basic process, which can be categorized
into bulk micromachining, LIGA (a process
that makes parts with significant depths
from metals, metal alloys, plastics or
ceramics) and surface micromachining, to
name a few.
SwRI has extensive experience in
designing MEMS devices. Outsourcing
allows Institute staff members to custom
design the devices and select the
MEMS methodology best-suited for a
particular project.
must have time to cool between cycles. It
can be used for discrete motions, similar
to a mechanical relay switch.
MEMS devices also can be activated
in other ways. A magnetic field can be
applied to attract a magnetic material. A
Sacrificial
Material Layer
Substrate
Methods of actuation
Researchers can make small devices:
but how do they work?
MEMS devices employ several basic
methods of actuation. The most common is
the electrostatic method, which requires
high voltage (more than 100 volts) but little
current. A receptor plate is charged, which
causes it to be attracted to a grounded
plate. One or both of these plates can
move toward the other. Sometimes these
plates are bonded on one end, forming a
cantilever beam or a series of interdigited
comb fingers. In other cases, they can be
spokes on a wheel that the electrostatic
attraction causes to turn. This attraction
produces the motion and is repeated to
obtain a desired effect. This effect may be
raising and lowering mirrors in an optical
device at a particular frequency, pumping
fluid through a microchannel or operating
a micromotor.
Joule resistive heating is another common actuation method that requires a
more moderate voltage and current (about
5 volts, about 5 milliAmperes). This
method can be used to produce devices
containing either a bimetallic construct or
a single material of varying widths. The
bimetallic version requires two materials
to be bonded together that have differing
thermal expansion coefficients. When the
current is applied, the two materials
expand at different rates, causing a straight
beam to bend in the same manner as its
large-scale counterpart, the thermostat
strip. A single-material version must use
varying thicknesses or widths for the two
beams that are attached to each other on
one end. The difference in cross-sectional
area causes the arms to heat at different
rates, producing the same beam-bending
effect. Unlike the electrostatic method,
joule resistive heating cannot be used to
resonate a structure because the structure
voltage can be applied to a device made
from a piezoelectric material to cause
motion. In the area of microfluidics, a stationary actuator may cause the motion of
the fluid. For example, an ultrasonic pump
uses piezoelectric material to create a
Patterned
Photoresist
After
1st Etching
Sacrificial
Layer
Material
Layer
Patterned
Photoresist
Material
Layer
After
2nd Etching
Sacrificial
Layer
Completed
Micromachined
Profile
After
Etching
Sacrificial
Layer
Substrate
The illustration depicts the steps used in the surface micromachining process. A layering process
is used to create MEMS components and obtain air gaps between layers that allow the components
to move.
Technology Today • Winter 2004
9
Small size, big challenges
The small size of MEMS devices creates design challenges. One of these is
“stiction,” the result of two surfaces not
having great enough stiffness to overcome
surface adhesion. During processing, liquids can stay between layers rather than
drying out, pulling the layers together via
D014669
surface acoustic wave. This wave, also
known as acoustic streaming, causes motion
in the contacting fluid. Electrohydrodynamic
pumps use a traveling electric field to pull
a dielectric fluid along a desired channel
path. This pump and similar electrokinetic pumps are used in biomedical
applications for such things as separating
DNA molecules.
The Institute developed a bi-directional horizontal thermal actuator. This patented device can
move in two directions upon the application of the electrical current.
NEW METHODS
SwRI techniques enhance quality,
reliability in MEMS technology
MEMS technology is entering a critical transition period
from research to widespread commercialization. While several of the simpler devices are being mass-produced, such as
inkjet printer heads and airbag sensors, the more complex
designs of new MEMS applications present significant challenges for manufacturing. Although current semiconductor
processes are well suited to reliability at high volumes,
MEMS devices that require movement are affected by additional parameters, such as fatigue strength, failure strength
and the elastic modulus. A material’s elastic modulus is
determined by calculating the ratio of stress to strain or by
determining the relative stiffness of a material within its
elastic range.
SwRI has developed techniques to assist MEMS designers and manufacturers in these areas. Designers need to validate that their designs can meet lifetime requirements,
while manufacturers need to monitor important properties
to ensure that their processes are producing high-quality
components.
The elastic modulus is one of the most important material properties required for MEMS design. It is used to set
design limits on stress, strain and displacement. SwRI developed a technique that employs a resonating cantilever
beam with an end mass on the MEMS chip. The beam is
driven with a sinusoidal motion from a MEMS actuator
comprised of interdigited fingers, or a comb-drive. As the
drive signal frequency approaches the natural frequency of
the beam, the amplitude of the vibration increases dramati-
10
capillary forces. Once the surfaces are held
in contact in this manner, the devices are
useless because the layers are not strong
enough to overcome the capillary forces
holding them together. Even room humidity
can be the source of the liquid agent that
causes this phenomenon. This can happen
in varying degrees, and is most often seen
as a type of friction that requires more
force to overcome than had been planned.
As described earlier, MEMS devices
are designed and fabricated using layers.
The 3-D structures are made from these
layers. Therefore, devices that need to
stand up off the chip, such as mirrors, must
be made flat and be able to lift up on their
own or by an actuator on the chip,
because they are too small to be effectively
handled manually. MEMS devices can be
manually activated using a microscope and
special probes, but it is only practical to do
cally. The natural frequency is detected visually, or by using
a laser-driven, optical system. Using the detailed computer
model developed at SwRI, the elastic modulus can be
determined. The test method was validated using atomic
force microscopy.
Using SwRI’s Displacement Mapping Software (DISMAP),
three-point bending notch specimens are used to perform
strain field measurements. A scratch drive actuator (SDA) MEMS
device is used to apply a load to the beam. The beam deflects
under the load and a digital image is captured before and
after the load application. The digital images are imported
into the DISMAP software to measure the motion of the
grains on the surface of the beam and the strain field of the
material. These parameters are important to understanding
when a device will break under a given mechanical load.
Fatigue strength is most important for MEMS devices
that operate in resonance, such as those in telecommunications applications, where large numbers of operational
cycles are quickly accumulated. The Institute designs
notched beams, attached to comb-drives, to determine this
parameter based on the research of Van Arsdell and Brown1.
The devices are operated at their natural frequency until failure occurs. The test devices are combined with advanced
digital signal processing to automatically measure the natural frequency and amplitude, and to count the elapsed
fatigue cycles. Using probabilistic techniques developed by
SwRI, parameters derived from these devices can be used for
reliability assessment and failure prediction.
1
W.W. Van Arsdell and S.B. Brown, “Subcritical Crack Growth in Silicon
MEMS,” IEEE Journal of Microelectromechanical Systems, Vol. 8, No. 3, Sept.
1999, pp. 319–327.
Technology Today • Winter 2004
Technology Today • Winter 2004
11
D006708
this in a research laboratory
setting, such as at SwRI.
The manufacturing
process causes another difficult challenge for MEMS.
When a layer of material is
deposited, all of the atoms
in the layer are not in the
same energy state. This
results in a compressive or
tensile stress inside the
layer, known as residual
stress. After the sacrificial
layers are released from
the chip and the structural
layers can move, structures
made with high residual
stresses will bend upward,
off the chip (tensile) or
downward toward the
chip (compressive).
Devices can be designed
to make use of residual
stress, but researchers usually design around it.
Once devices have
been fabricated and
assembled, they must be
protected from their environment, yet in many
cases they must be able
Institute engineers developed vertical thermal actuators operated by a differential expansion between two layers of
to interact with their envidissimilar materials in each arm. To obtain the initial upward curvature, the device takes advantage of residual
ronment. Thus, the final
stresses in the film layers. An electrostatic version with integrated flip-up mirrors was used to produce an optical
packaging is typically indiswitch. SwRI holds U.S. patent No. 6,608,714.
vidualized for the given
device. For example, a
chemical sensor must be in
sensors and devices. They can interact
workers often replace parts according to a
contact with the chemical it is sensing. If
with the environment at the molecular
given schedule rather than first showing
this chemical is in liquid form, it must be
level to achieve new goals. ❖
that stress corrosion cracking has occurred
injected into the MEMS chip in some manor may occur soon. MEMS fabrication
ner. If the chemical is in gaseous form, the
Comments about this article?
techniques allow for smaller fabrication of
MEMS device must be in contact with air
Contact Hanson at (210) 522-5798
standard test structures that could be
but not clogged by dust particles because,
or [email protected].
employed in small locations.
to a MEMS device, a dust particle is a large
SwRI project teams can consist of a
object. Of course, all devices must be provariety of experts across several technical
tected from vibration and handling.
Acknowledgments
areas. For example, the SwRI team that
The author gratefully acknowledges the
contributions of the following SwRI staff
designed an in vivo drug delivery system
Solving industry problems with MEMS
members to the preparation of this article,
included experts in microencapsulation,
as follow:
chemistry, mechanical engineering, fluid
The SwRI MEMS team seeks ways to
“Relieving Stress,” Manager Dr. Sean Brossia
dynamics, electronics, bioengineering
apply MEMS technology to industry proband Research Engineer Andy Veit.
and MEMS.
lems. An example of this is a stress corro“In Motion,” Research Engineer
sion-cracking sensor (see sidebar, p. 7)
Joseph N. Mitchell.
Conclusion
being developed using internal research
“New Methods,” Institute Scientist
funds. Current methods available for in
Dr. Steve Hudak, Principal Engineer
MEMS have the ability to impact
situ monitoring of stress corrosion crackDr. Dan Nicolella, Research Engineer
almost every technical field. Their small
ing are too cumbersome to install in the
Joseph N. Mitchell and former SwRI staff
member Rick Fess.
size, high volume and low cost enable
minimal spaces on an airplane, for examthe creation of a suite of disposable
ple. Because of this, airplane maintenance