IN THE world of micromachines, just being small used to be good enough.
In the early 1990s, the novelty and elegance of tiny steam engines and
electrostatic motors fashioned in silicon generated great interest and funding.
Grand visions emerged of robots cruising through the blood vessels of the human
body. Up to now, however, these extravagant machines have found few, if any,
practical uses.
Instead, futuristic thrills are giving way to more pragmatic preoccupations.
Industry has focused on a few, basic micromachines that are starting to have a
profound impact on the lives of ordinary people. They are appearing in cars, in
homes, in operating theatres and in entirely new and unexpected areas. “Up to
now, the research people published visions,” Reiner Wechsung, president of
microParts, a company based in Dortmund, Germany. “What we need are real
products.” Wechsung’s company is striving to produce those “real”
micromachines.
In May, the company’s latest device, a CFC-free inhaler that turns an
anti-asthma drug into droplets smaller than 5 micrometres, was launched by the
German pharmaceuticals company Boehringer Ingelheim. Micromachining produced the
tiny apertures and conduits that create the droplets and propel them towards the
lungs. The firm is also using micromachining techniques to make the nozzles and
inner workings of the next generation of inkjet printers.
Advertisement
And microParts is not the only company exploiting micromachining techniques.
Last year, Analog Devices of Massachusetts, which makes accelerometers that
trigger car airbags to inflate, launched a new micromodel. This fits onto a chip
measuring just 3 millimetres square and includes not only the silicon sensor,
but also the electronics for calibrating and controlling the device. It is ten
times more sensitive to motion than previous versions, and greater sensitivity
means more applications. It is already being touted as a diagnostic tool for
electric motors, which will quickly detect damaging vibrations, and as an
automatic trigger for switching off gas supplies during earthquakes.
In May, Lucas NovaSensor, a Californian subsidiary of Lucas Industries,
unveiled an air-pressure sensor that is small and light enough to fit inside a
car tyre. The device goes hand-in-glove with a new type of tyre that allows a
car to carry on driving with a blowout. These “run-flat” tyres are safe for
about 300 kilometres, but work so well that a driver may not notice the
flat—hence the need for the pressure sensor.
Then there is Texas Instruments which over the past six months has introduced
a series of micromachines based on chips packed with 16 micrometre-square
mirrors that can be moved by electronic signals. These systems are being built
into overhead projectors, home movie equipment and wide-screen TVs.
One of the critical elements of today’s generation of microelectromechanical
systems (MEMS) is that they are simple enough to mass-produce. This makes them
cheap. A single silicon wafer, for example, can hold tens of thousands of simple
pressure sensors and a smaller but still impressive number of accelerometers.
The cost of making one micrometre-scale device is nearly the same as making an
entire wafer-full. A mass-produced, micromachined accelerometer sells for about
$10, compared with up to $1000 for its larger, conventional
cousins.
Mass-production has also helped to decide which applications have been
developed first. Companies have targeted vast potential markets, such as the
automobile industry. System Planning Corporation, a market research firm based
in Arlington, Virginia, estimates that by 1998, American law will make air bags
compulsory for as many as 15 million vehicles.
And once one big market exists for a device, others are free to exploit it.
Lucas NovaSensor, for example, is seeing its pressure sensors included in diving
gauges for measuring depth and air-tank pressure, altimeters for hang-gliders,
and even in computers attached to mountain bikes for those who want to know how
high they’ve climbed.
MEMS may have only started to appear in products over the past few years, but
the idea of using silicon for building machines has been around for much longer.
In 1982, Kurt Peterson of IBM published a paper called “Silicon as a mechanical
material”, in which he noted that as well as being a good semiconductor, silicon
has some remarkable mechanical properties. Pure silicon, for example, has a
hardness, Young’s modulus—a measure of elasticity—and tensile yield
strength similar to stainless steel.
In the years after Peterson’s paper, MEMS research took off, helped in large
measure by the army of experts who had learnt to manipulate silicon for the
electronics industry. Most MEMS are built the same way as
microprocessors—using photolithography. In this process, ultrathin layers
of silicon, silicon oxide or metal are painted with a substance called
photoresist, the desired pattern is masked off and the surface exposed to light,
which chemically alters the photoresist. The unchanged areas of photoresist are
then etched away to leave, say, an electronic circuit. By etching away
“sacrificial” layers of silicon oxide, engineers can create structures that have
gaps beneath them, such as a cantilevered “beam” that overhangs a pit in the
silicon wafer, or a cog that turns.
Up to now, the biggest impact of MEMS has been in devices that sense forces
and pressures around them, particularly where size and weight need to be kept to
a minimum. Lucas NovaSensor’s blood pressure sensor, for example, is only 700
micrometres across. It attaches to a saline-filled catheter that is inserted
into a blood vessel. The sensor itself is a diaphragm of silicon, typically less
than 10 micrometres thick, which covers a sealed pit etched out of the wafer.
Embedded in the diaphragm are piezoresistors which change their resistance when
subjected to stress. When the external pressure alters, the diaphragm deforms
and the variation in resistance gives a measure of the pressure change.
The other main commercial sensor, the accelerometer, is a more complex
machine. Analog Devices’ design consists of a silicon beam mounted a micrometre
or so above a wafer. It is connected to the wafer by “tethers”, each about 2
micrometres thick and 200 micrometres long, which act as springs (see
Diagram).
The beam has dozens of “fingers” sprouting from its sides, each of which acts as
a capacitor plate, and sits between two other plates that are fixed to the
wafer. Normally, each pair of capacitors is balanced, but when the beam
moves—in response to a sudden acceleration—the capacitance of one
changes relative to the other, giving a measure of the change in motion.FIG-20364401.jpg
While sensors are now being widely exploited, actuators are still being
developed. Motors and pumps, and in general anything that must work hard against
the larger world, present troublesome technical challenges. Try moving a mound
of dirt with a micromachine, for example, and you won’t get very far. Outside of
a few specialist niches, actuators are still rare animals.
One of those specialist niches is manipulating light, a task that needs only
tiny forces. Take Texas Instruments’ digital light processor (DLP), a component
of projectors and wide-screen TVs. A DLP chip, which measures 1.5 by 1
centimetres, is covered by some 500 000 aluminium mirrors each of which can be
moved individually. Each mirror is attached to a thin metal strip which acts as
a hinge, with its position controlled by a switch and two electrodes etched into
the silicon below it (see
Diagram over page). The whole structure is built using
standard photolithographic techniques.FIG-20364402.jpg
The mirror’s position is controlled by electrostatic attraction to the
electrodes. To project a large TV picture, a bright light is shone onto the chip
and a TV signal fed to the switches. The signal controls the pattern of
reflected light that passes through a projection lens and onto a screen.
At standard resolutions, the DLP gives a better image than a cathode ray tube
because it has no scanning lines and no visible pixels, says Larry Hornbeck, the
Texas Instruments Fellow who invented the concept. “At high resolutions, where
CRTs drop out, we get brighter and liquid crystal displays get dimmer,” he
says.
Another intriguing application for actuators is in controlling large-scale
effects that begin as tiny perturbations. An aircraft’s wing, for example,
creates and sheds vortices as it travels through the air. When those vortices
start they are miniature progenitors of what they will become. MEMS could detect
and then modify these vortices using tiny flaps that would flick up or lay flat.
A bank of such devices could change the air flow over an entire wing.
Microflaps
“It’s been demonstrated in the last two years in a wind tunnel test at the
University of California in Los Angeles with a scale model that yes, distributed
MEMS devices can cause enough of a change in fluid vortices that you can
generate a tilt in a model,” says Ken Gabriel, deputy director of the
Electronics Technology Office at the US Department of Defense’s Advanced
Research Projects Agency, which is funding the research.
With MEMS to control tilt, roll and pitch, an aircraft could consist of a
single wing with only tiny flaps. But such an aircraft, although possible, is
not going to roll onto the apron in the near future. “We’re talking ten years in
terms of real implementation,” says Gabriel.
There are, however, some problems that must be overcome before actuators and
more complex sensors become commonplace. One big advantage of MEMS is that they
are made with the same materials and methods as microelectronic devices, and so
can be integrated with the control circuits that are needed to make them work.
For some devices, such as microaccelerometers, this closeness is vital.
“The signals that are generated by these little structures are tiny,” says Jim
Doscher, of Analog Devices. “We measure attofarads of capacitance change. To do
that you can’t take the signals off the chip.”
And integrating electronic and mechanical components helps to keep down
costs. By exploiting advances in the electronics industry, MEMS can reap the
benefits of those advances without footing the whole research bill. That’s
important, at least in the US where last year the Department of Defense found
that 60 of the 80 companies engaged in MEMS activities were small concerns, with
typical revenues of less than $5 million a year.
But in some ways, the marriage of microelectronics and micromachines is not a
happy one. For example, certain components of MEMS, such as tiny cogs, need to
be treated at high temperatures. Without this treatment they “curl up like a
potato chip”, says Paul McWhorter, manager of the intelligent microsystems
department at Sandia National Laboratory in New Mexico, one of the pioneering
centres for MEMS research. But exposing nearly completed microelectronic
circuits to high temperatures ruins them because it melts the aluminium wiring
and drives off the chemical dopants that are added to semiconductors to create
transistors and other components.
Pioneers in a microworld
At first sight, the solution appears to be to build the MEMS first, putting
all the destructive heat at the start of the manufacturing cycle. But this works
only in some cases. The problem is that micromachines are often not level with
the plane of the wafer, which disrupts the focusing of the patterns of light and
shade that are cast onto the wafer as part of the photolithographic process.
To avoid this, engineers have tried a number of approaches, such as lowering
the temperature of the heat treatment. But this limits the designs that can be
used. Now, however, a new technique developed at Sandia may provide a better
solution.
McWhorter and his colleagues etch a trench in a silicon wafer, build their
micromachine inside it, and then heat treat it. The trench is then filled in
with silicon dioxide to leave a flat surface. The microelectronics are then
built nearby and, as a last step, the silicon oxide is etched away to free the
machinery. This innovation promises to improve the sensitivity of MEMs by
allowing the machinery and the electronics to be optimised independently.
“Each time you get more sensitivity you go from being able to detect a car
crash to having something that can detect head position for virtual reality
goggles or hand positions for power gloves in virtual reality systems or input
devices for computers,” says McWhorter.
Another problem facing engineers using photolithography is that they are
largely restricted to two dimensions. The height of microelectronic circuits and
MEMS is typically between 2 and 10 micrometres. Also, silicon is not the best
substance for every task. It would be useful to be able to use a polymer or a
ceramic for certain tasks.
There are two promising solutions here. The first is a technique called
reactive ion etching. This process, which uses chlorine or fluorine-based
plasmas rather than a liquid to etch, is driven by radio frequency power and can
cut out deep structures with nearly vertical sidewalls. “We’re going about 200
micrometres deep now,” said Dale Gee, director of new products at Lucas
NovaSensor, which is working on the technique with Greg Kovacs at Stanford
University. The third dimension opens up a range of new potential applications,
such as building relatively large capacitor plates, which boosts capacitance.
This in turn could help to make accelerometers more sensitive.
The second solution does away with silicon altogether. Developed by
researchers at the Karlsruhe Research Centre in Germany, the technique is known
as LIGA. The process uses X-rays to carve out deep patterns in a polymer. These
patterns can then be electroplated to produce tiny, 3D metallic structures, or
metallic moulds. The moulds can be injected with a variety of materials—
including a number of polymers—to produce micromechanical devices. Among
the devices made with LIGA are gear trains and magnetic motors.
Although the technique is still experimental, microParts has licensed the
process and is using it to make inkjet nozzles for the next generation of
printers, which will produce 2400 dots per inch. For such high resolution, these
printers will need four times the nozzle density of today’s printers, moving
them into the realm where high precision, low-cost micromachining comes into its
own. Whether LIGA will be widely adopted isn’t yet clear. Silicon micromachining
is more mature and offers easier integration with microelectronics.
What of the future? In 1994, System Planning Corporation estimated that the
total MEMS market would be worth $13.9 billion by the year 2000.
McWhorter reckons that much of this expansion will come from an explosion of new
applications for today’s sensors. And these systems will have a much bigger
impact than expected, argues Wechsung. A micromachine may be only a small
fraction of the cost of a new device—such as an altimeter for
mountainbikers—but its action, size and weight will be the key to its
competitiveness, he says.
Increasingly sophisticated sensors will also appear, says McWhorter. Much
attention is now being given, for example, to developing inertial guidance
systems for cars. Along with accelerometers, these systems will need gyroscopes,
which require constantly rotating or vibrating parts. Such active devices will
help to bridge the gap between sensors and actuators.
More complex actuators, born of the tiny steam engine and motors designed at
Sandia, will also arrive eventually, reckons McWhorter. Their first use is
likely to be in the defence industry, for locking and arming weapons, because
they would be difficult to defeat and should improve reliability, he says.
Meanwhile one of the more popular images of MEMS is still the tiny robot free
to roam human blood vessels. How close is such an device? McWhorter is in no
doubt: “I think that would be further down the road than ten years.”
* * *
Half-time in Tokyo
AT the First International Micromachine Symposium held at the Science Museum
in Tokyo last November, the organisers were turning them away at the door. More
than 500 people had turned up for the symposium, far more than anticipated, and
a further 3500 flocked to the accompanying micromachine exhibition.
The events were staged to present results from the first half of Micromachine
Technology, a $250 million, 10-year project set up by Japan’s Ministry of
International Trade and Industry (MITI). The project, which began in 1991,
supports research at 24 companies. They include leading representatives from
Japan’s car parts, electronic, medical equipment, and robotics industries. Two
small American firms also took part, as well as one Australian
university—the Royal Melbourne Institute of Technology.
“The first five years of the Micromachine Project has been exploratory,
proof-of-principle research,” says Dinesh Sood, a materials scientist who is
jointly in charge of the work at RMIT. Research thus far has been conducted at
individual sites.
Allowing groups to work in isolation is a new departure for large-scale
projects run by MITI. But the companies’ reluctance to collaborate with each
other combined with concerns about ownership of intellectual property rights
made the change inevitable. A second departure was the use of committees of
independent academics to assess the research before giving the go-ahead to the
second half of the project—which began in April.
This phase will bring researchers together to work on two projects. One is to
make a catheter intended to carry a camera and other sensors into the brain. The
second is a microrobot intended to travel through the pipes of power stations,
diagnosing and fixing problems.
A fringe benefit of participation in the MITI project for Sood and his
colleague Ron Zmood is that their counterparts at Japanese companies have opened
the doors of their laboratories for them. Sood came away impressed by the level
of activity he observed. “The Japanese system is very different [from that of
the US and Europe],” he says. “The companies are very solid in their commitment,
and they mean business.” In Tsukuba, at the lab of the control components maker
Omron, for example, Sood found an entire research team assigned to an optical
scanner to be used as a guide for microrobots.
Japanese universities are also pulling their weight. Sood reckons that Tokyo
University alone musters six or seven well-equipped micromachine labs. But even
Tokyo pales in comparison to Tohoku University, selected as a centre of
excellence, which is “four or five times stronger in micromachine research than
any university in the US”.
Another difference Sood noted is that “the Japanese believe the future is not
in silicon. Silicon is just one of the materials that you can work with.”
Indeed, Sood believes that MITI chose to fund the RMIT project precisely because
the Australian team didn’t want to work in silicon.
Sood and Zmood are tackling one of the knottiest problems of
micromachines—friction. At the micrometre scale, friction can destroy the
machines. At conferences, “no one talks about how long their motors last”, Sood
says, “and it’s considered unethical to ask.” The ugly truth is that most
micromotors fall apart first time out. The RMIT researchers have adopted an
approach to the problem which avoids friction altogether. Rather than have two
surfaces come into contact with each other, they use tiny electromagnets to
levitate objects and keep them apart.
Bob Johnstone, Melbourne