快猫短视频

A DUST mite trying to build a miniature car would have a hard time copying
one of ours. The reason is that large machines are designed to rely on the
large-scale properties of matter. A flywheel that keeps an engine turning relies
on its momentum to keep going. And motor oil relies on its low viscosity to
lubricate an engine.

But change the scale and the properties of matter change too. In the dust
mite鈥檚 world, momentum is of little consequence compared with forces like
friction, so a microscopic flywheel would soon grind to a halt. And the same
motor oil that keeps a large engine turning behaves like caramel on the
microscopic scale. A single drop would prevent a tiny engine working at all.
Even if a mechanically minded dust mite could get around these problems, it
would come up against others thrown up by the strange laws of physics that
govern the behaviour of machines on the microscopic scale.

While dust mites will not be losing sleep over the difficulties in building
these machines, the same cannot be said of human engineers. These engineers have
been telling anyone who will listen that microelectromechanical machines (MEMs)
are about to change the world. For just over a decade now, they have made
amazing claims for their tiny offspring鈥攖hat they will monitor pollution
in places too hazardous for human life, perform impossible operations inside the
human body and even control the airflow over aircraft wings. And all more
efficiently and at a fraction of the cost of the bulky equipment that carries
out only some of these tasks today.

So where are these devices? A few have hit the market鈥攖he
accelerometers that control airbags and the nozzles that spray ink in some types
of printer are good examples. But the truth is that the promised MEMs Revolution
is running way behind schedule. The problem is that engineers are struggling to
work with the strange laws of physics that govern the world on the smallest
scale. Practically anything that works on a large scale has to be rethought for
micromachines. 鈥淎lmost everything you know about designing machines you had
better forget,鈥� says Terry Michalske, a chemical engineer at the Sandia National
Laboratories in New Mexico, one of the big pioneering centres of MEMs
development (鈥淚nvasion of the micromachines鈥�, 快猫短视频, 29 June
1996, p 28).

Without solutions to the dust mite鈥檚 dilemmas, researchers find it difficult
to design MEMs that work well and reliably. Engineers need new tools to help
with their designs. They need ways to test their machines once they have been
built. And if they do work, they need to find out how reliable the machines will
be. How long will a silicon seal survive? Will a tiny gear still be spinning
after 20 years inside a human body?

MEMs devices are tiny machines carved out of silicon. They are made in much
the same way as integrated circuits (ICs)鈥攊n several layers. First, a
sliver of silicon is coated with a layer of plastic that protects it. Some of
this layer is then cut away, exposing the chip with the desired pattern. The
chip maker can then etch away the chip or deposit a layer in the shape of this
template. The protective layer is then washed away and the process begins
again.

The difference for a micromachine is that in the first layer of silicon, the
chip maker carves out the desired parts鈥攕prings, cogs, connecting rods and
suchlike鈥攖hen embeds them in a layer of silicon dioxide. The next layer of
parts is created on top of this and then embedded in silicon dioxide, and the
next layer on top of this and so on. When all the parts of the machine have been
created, a final chemical bath strips away the supporting layers of silicon
dioxide and the silicon parts are left to move freely, pushing, pulling and
interlocking as the design requires. That鈥檚 when the trouble starts.

One of the biggest problems that MEMs engineers face is 鈥渟tiction鈥�. For a
start, tiny electric charges can force components together, like nylons in a
clothes dryer. Weak chemical bonds can also form between the surfaces. And
liquid residues are a problem. In the macroscopic world, surface tension is a
negligible force. But in the microscopic world, it can glue components together
irreversibly. 鈥淪uppose we lived in this microscopic world,鈥� says Sam Miller, a
MEMs engineer at Sandia, 鈥渢o hang a picture on the wall, you would just put it
up and it would stick.鈥�

Stiction can be particularly nasty during the very last stage of MEMS
fabrication. After the parts are free to move, the 鈥渞elease鈥� chemicals that
strip away the silicon dioxide must be removed. But researchers cannot let the
liquid evaporate or drain away because surface tension tends to pull the
components together into a useless jumble of cogs and springs.

Freeze-dried

To get around this, Miller and his colleagues have tried eliminating surface
tension altogether by literally freezing the release chemicals in situ. They
then lower the pressure, allowing ice to become a gas without becoming a liquid,
a process called sublimation. This tackles the stiction associated with surface
tension but doesn鈥檛 eliminate the other causes.

So Michalske and his colleagues have tried coating the microscopic parts with
perfluorinated hydrocarbons, such as Teflon, which are water-repellent and
chemically inert. These compounds consist of a carbon atom surrounded by atoms
of fluorine. To attach them to the silicon components, the Sandia team created a
molecule shaped like a dumb bell consisting of a perfluorinated hydrocarbon
connected by a rod of carbon atoms to a silane (an atom of silicon surrounded by
hydrogen atoms). The silane end of the dumb bell attaches to the silicon
components, leaving the Teflon-tipped carbon rods sticking out like bristles on
a brush. The result is a fur-like coating of Teflon bristles that repel water
and each other. This non-stick coating works so well that most engineers who
design MEMs use something similar, says Michalske.

But controlling stiction is only one step in making a device that will change
the world. Equally important is a way of testing the devices once they have been
made. Testing ordinary integrated circuits is relatively straightforward since
both the input and the output of these devices is electronic. Putting a signal
in and measuring what comes out is easy. But the input and output of MEMs can be
anything from a change in acceleration or an increase in concentration of
hydrogen gas to the angle of a movable mirror or a dose of insulin pumped out
through a needle. What kind of test can accommodate all of these?

Bill Miller, another Sandia engineer, says one solution is simply to watch
the way the device works using a microscope. 鈥淎n IC chip running looks a lot
like an IC chip not running,鈥� he says, 鈥渂ut with micromachines you can tell
quite a bit by watching.鈥�

The problem with watching is that micromachines move very quickly. Some of
the tiny engines developed at Sandia run at 400 000 revolutions per minute. 鈥淚t
just looks like a blur,鈥� says Miller. So he flashes a strobe light in synchrony
with the engine to 鈥渇reeze鈥� the gears once during each revolution. In this way,
the engine appears stationary and so it can be checked using pattern recognition
software for misalignments, loose fits, or any other visible problems. If Miller
runs the strobe light a little faster, the gears appear to turn very slowly and
a computer can compare the motion with images of what it should look like.

Not every problem is necessarily visible. A microengine developed several
years ago at Sandia contains a 鈥渃omb drive鈥� that looks like two hair combs with
their teeth interlocked. By creating an alternating electric field between the
interlocking teeth, the combs can be forced to pull apart and then slide
together again. It is important that the tines never touch because this would
short-circuit the device, but the gaps are too small to be picked out easily
using a microscope. So Miller checks the electrical resistance across the combs
with tiny electrodes. If all is well, the resistance will be high, millions of
ohms. But if two parts are touching, the resistance is much lower, hundreds of
ohms. 鈥淚t鈥檚 a real go, no-go, kind of test,鈥� he says.

These kinds of tests ensure that a micromachine works when it鈥檚 new, but
cannot tell how long it will last in future. Knowing the reliability of a device
is crucial because many MEMs devices will have to work under challenging
conditions, rotating inside car tyres for thousands of miles, or working for
decades as insulin dispensers implanted in the body of a diabetic. Replacing
these devices if they break could turn out to be far more expensive than fitting
them in the first place, so researchers want to be sure that they last longer
than the lifetime of a car or a human being.

The human body is a particularly harsh environment, says Khalil Najafi, an
electrical engineer at the University of Michigan in Ann Arbor. For him, it鈥檚 a
pool of warm, salty water, full of enzymes and corrosive ions鈥攃onditions
very difficult to protect against. Najafi has developed a microstimulator that
produces electrical signals similar to those produced by the nervous system. The
idea is that such a device would be implanted in the limbs of people who have
been paralysed to restore function. 鈥淚t doesn鈥檛 take much moisture to disrupt
the circuitry,鈥� says Najafi. 鈥淪o our main effort now is testing the long-term
蝉迟补产颈濒颈迟测.鈥�

One way to test reliability is to accelerate the ageing process. Computer
chips tend to fail when electromagnetic fields and heat gradually break down
insulating layers causing the circuits to short out. This process can be speeded
up by heating the chips. The principle is that if the chip lasts, say, 20 hours
at 100 掳C, it should work for 20 years at room temperature.

Najafi and his colleagues have developed a technique that recreates this test
with the chips sitting in salty water鈥攁 kind of artificial human body. By
measuring how quickly moisture penetrates the device, he has been able to design
a sealed glass capsule which models a device that lasts on average more than 100
years.

Like all mass-produced materials, some chips will last longer than others, so
Najafi is building a self-tester into the microstimulator. The tester will send
a reading of the humidity level to a detector held to the skin to warn of
potential failure. The sensor should help with future studies in which the
devices will be implanted in the muscles of lab animals. But it could also work
as part of the final package, which may be ready for clinical trials within a
year or two.

While the heat test works for electronic devices like Najafi鈥檚
microstimulator, it does not work for moving parts. But accelerating the ageing
process for gears and cogs is a bigger problem, says Paul McWhorter, who
oversees the micromachine division at Sandia. 鈥淲ith MEMs you don鈥檛 know what the
failure mechanisms are,鈥� he says, 鈥渟o you don鈥檛 know how to test for long-term
蝉迟补产颈濒颈迟测.鈥�

It would be easy to think that springs made of crystalline silicon would snap
if they were bent too often, as they do on the macro scale. But McWhorter says
that repeated bending has almost no effect. Flexible parts seem almost
indestructible. 鈥淵ou can bend them an incredible number of times and not have
failure,鈥� he says.

Friction failure

Nobody is quite sure why MEMs devices fail but friction seems a likely
candidate. In lab tests, McWhorter says, most failures occur where two surfaces
are rubbing on each other. Testing the long-term reliability is difficult,
though, because nobody has worked out how to speed up a machine that is already
designed to turn several hundred thousand times per minute.

Meanwhile, Michalske is trying to make friction less of a problem. And he is
stumbling on some strange effects. Traditional lubricators such as grease and
oil simply don鈥檛 help in the microworld. On that scale, water turns to syrup,
oil becomes caramel and grease behaves more like marzipan. 鈥淭he hundreds of
thousands of revolutions per minute our motors generate just can鈥檛 happen with a
liquid there,鈥� explains Michalske.

But it turns out that the anti-stiction Teflon coating can work like grease.
When two coated parts rub together, the natural repulsion between the
Teflon-tipped bristles reduces friction. But some other effect is also at work.
Last year, Michalske found that bristles that are 18 atoms long turn out to be
better at reducing friction than bristles that are 17 atoms long. Nobody knows
why, although Michalske thinks that the way bristles expend energy as they flex
might be an important factor. The team is now working to optimise the bristle
length.

Whatever the eventual solutions to these problems, everyone agrees that MEMs
devices will only be mass-produced when everybody agrees to make them in the
same way. To understand how important this kind of standardisation is, take a
look at the IC industry. After all, ICs are just MEMs without the mechanical
parts. All microchips made today use the same materials and the same basic
microprocessor design. Although engineers experimented early on with various
semiconducting materials, including germanium and gallium arsenide, as well as
many ways to make transistors, in the end one material, silicon, and one
transistor design, CMOS, became standard.

Today, anyone with an idea for a computer chip can quickly get it designed,
tested and made using CMOS techniques. Investors need never worry about whether
a new chip can be designed and built cheaply, so good computing ideas can always
find financial backing.

But this is not yet the case with MEMs. Since engineers are still working out
how to deal with the problems of building in the dust mite鈥檚 world, there are no
standards. And it鈥檚 not just a transistor that needs to be standardised, but
mirrors, pumps, motors, sensors, gears and valves. 鈥淭oday in MEMs, people try to
design the best technology for their purpose,鈥� says McWhorter, 鈥渨hereas in the
semiconductor industry they start by thinking `how do I use CMOS?'鈥�

But the standards will come, and so will the real MEMS Revolution, says Bill
Miller. 鈥淚t鈥檚 like ICs back in the sixties. They鈥檙e really going to become big,鈥�
he says. 鈥淲hat you鈥檙e looking at here is the tip of a very large iceberg.鈥� The
dust mites of this world would be surprised.