THE pool is carved from pure silicon and the liquid is sealed in by a window
of quartz. The fluid is so thick that any swimmer would barely be able to move.
Suddenly, heating elements embedded in the quartz window begin to glow. A great
bubble of vapour forms and grows, forcing the liquid down a narrow silicon
passageway leading away from the pool.
The bubble blocks the way back, so the only way is forward, along silicon
corridors and past silicon pillars. In some places, waves ripple along the
walls, pushing the fluid along. In others, the flow is governed by the shape of
the corridors and how they twist and turn. Strangest of all, no matter how
convoluted the flow, it is entirely free of turbulence.
In a swimming pool or a teacup none of this could happen, but this is the
brave new world of microfluidics, the science of handling liquid on a tiny
scale. This emerging discipline is set to revolutionise the way scientists
analyse the world around them. Today, most chemical and biological analyses are
traditionally done on the macroscopic scale with test tubes, pipettes and Bunsen
burners, slowly and clumsily handled by humans. In future, analyses will be
controlled by computer and carried out with less than a teardrop of liquid
racing through the corridors of silicon laboratories the size of pennies.
Advertisement
But working on this scale is not easy. The sorts of machines that manage
fluids in the macroscopic world鈥攚ater taps, bicycle pumps and cocktail
shakers鈥攋ust don鈥檛 work with millionths of a millilitre. Even if these
devices could be made on a microscopic scale, they would not work because fluids
behave differently in the miniature world. Water flows like syrup, for example.
It does not splash or churn but moves like a smooth-running river, free of
eddies.
And without turbulence, fluids do not mix. Add milk to a cup of tea and it
mixes with a few turns of the spoon because of the turbulence that stirring sets
up. But with a few millionths of a microlitre of tea and milk, stirring does not
help. The two liquids flow independently in two smooth-running layers. Just
making fluids mix on this scale is one of the great challenges for microfluidics
engineers.
To handle quantities of fluid smaller than the amount of ink in the full stop
at the end of this sentence, engineers have had to come up with some clever
machines. Their designs tend to have few if any moving parts. And they often
rely on forces such as surface tension that seem insignificant in the
macroscopic world.
Bubbles turn out to be one of the most valuable tools for microfluidics
engineers. Dorian Liepmann, a mechanical engineer at the University of
California, Berkeley, uses them as pistons鈥攁s the bubble expands, it
pushes liquid out of its way. 鈥淲e use bubbles because they don鈥檛 have any moving
parts,鈥� he says.
Under the clear quartz top of one of Liepmann鈥檚 chips, tiny circles and
channels are etched in silicon, like the blueprint of a distillery. Liepmann
points to one of the circles, a tiny chamber, with the tip of his pen. 鈥淲e
generate a bubble in here by heating up the fluid,鈥� he says.
Tiny elements in the quartz boil the liquid in a circular chamber below,
creating a bubble of water vapour that pushes the fluid along a channel. Surface
tension keeps the bubble intact. When the heat is turned off, the vapour
condenses and the bubble contracts.
Bubbles can also serve as one-way valves. Liepmann鈥檚 idea is to place three
silicon pillars in a channel with a heater downstream. The heater generates
bubbles that are washed away when the liquid flows in the required direction.
But if the flow reverses, the bubbles are forced against the pillars, forming a
seal. The overall effect is a one-way valve.
Bubble trouble
At least, that鈥檚 the theory. In practice, Liepmann has been unable to make
the valves work effectively. The bubbles, which are shaped like inverted domes,
do not grow large enough to fill the channels and create a seal. For the moment,
his valves leak badly.
What鈥檚 more, Liepmann has discovered that liquid can flow through the bubble
itself. At the surface of a bubble, water is constantly evaporating and
condensing again. If the bubble is perfectly symmetrical, the rate at which this
happens is the same all over its surface. But if the bubble becomes slightly
misshapen鈥攂y pressing against the channel walls for example鈥攚ater
will evaporate more quickly on one side and condense faster on the other. This
creates a net flow of water across the bubble. Liepmann鈥檚 group is now making
shallower channels and more symmetrical spaces for the bubbles. He is confident
that the bubble valves will soon be working properly.
Once he has working pumps and valves, Liepmann plans to build a drug delivery
system on a microchip that could give patients smaller doses more often. Such a
system would be a boon to diabetics, for example, because in concentrated doses
insulin damages the tissues at the site of the injection. The lack of moving
parts makes the device inherently safer than other designs. 鈥淚f you are
delivering insulin in a controlled manner and something breaks, you鈥檙e in
trouble,鈥� says Liepmann. The wrong dose can be lethal. And if a conventional
pump goes wrong, there is no way of telling what dose has been given. But a
bubble pump can break down only if the whole unit burns out, says Liepmann. In
that case, it would stop working altogether and deliver nothing.
While bubble pumps have tremendous potential, they do have limits: big
particles tend to clog such a system. So bubble pumps cannot be used to move red
blood cells or viruses, just the kind of things used in DNA analysis. One
solution to this problem has been devised by a colleague of Liepmann鈥檚 at
Berkeley, Richard White. He has managed to get a technique called acoustic
streaming, in which a fluid is driven by waves that deform the wall of a
channel, to work on the microscopic scale.
White鈥檚 pump consists of a silicon canal sealed by a membrane covered with
zinc oxide. Zinc oxide is a piezoelectric material: apply a voltage across it
and it will change shape. So by applying a rapidly alternating voltage to the
zinc oxide membrane, White can make it flex first one way then the other,
setting up a wave that travels along it. This wave motion sets up pressure waves
in the liquid, forcing liquid near the wall to move in the direction of the
waves at up to several centimetres per second.
Since the main flow is relatively close to the wall, only particles that are
small enough to fit in it can be moved. White has generated flows roughly 10
micrometres wide, about the same size as red blood cells and many bacteria.
Acoustic streaming could be particularly useful for counting or analysing
cells one by one, says White. When the pressure waves in the liquid hit a wall
at an angle of 45 degrees to the flow, they bounce off the wall and move away at
right angles to the original flow
(see Diagram). Where the crest of a reflected
wave encounters the crest of an incoming wave, they combine to give a doubly
large peak. Those points of constructive interference move along the channel,
coaxing particles into rows parallel to the channel wall.
White has built a tiny octagonal test chamber in which every side is at 45
degrees to its neighbours. Inside, it is possible to see tiny beads streaming
around in invisible lanes like cars on a motorway. In theory, a laser shining
through the chamber could count each one as it flowed past, or even analyse its
chromosome content.
Microscopic valves
Shape also plays a crucial role in other devices. The miniature Tesla valve
created by Fred Forster, a mechanical engineer at the University of Washington
in Seattle, is a good example. In 1920, the Serbo-American scientist Nikola
Tesla, inventor of fluorescent lights and power grids based on alternating
currents, patented a valve with no moving parts. Essentially, it is the fluidics
version of a ratchet: the fluid can move easily in one direction but not the
other. The valve is a pipe with a series of side channels that reconnect to the
pipe further along. The channels connect at a sharp angle in one direction but a
shallow one in the other.
Any fluid flowing in the preferred direction is unlikely to be diverted into
the side channels, but fluid moving in the opposite direction is likely to be
diverted into the channels, creating turbulence that disrupts the flow. Tesla
claimed in his patent that the flow in the forward direction was 90 times that
in the reverse direction鈥攁 remarkably effective valve, considering that it
has no physical seals.
Forster had been trying for years to make a microvalve with no mechanical
parts when he thought of adapting the Tesla valve. At first, he didn鈥檛 really
expect it to work because of the way fluids behave on the small scale. Two
properties govern the motion of liquids: inertia, a liquid鈥檚 tendency to keep
moving, and viscosity, the tendency of the molecules of a liquid to interact
with their neighbours and slow down. In the macroworld inertia is the dominant
effect, but in the microworld tiny volumes have very little mass and thus low
inertia. Here, viscosity is the important factor.
Tesla鈥檚 valve depends on inertial forces. When the divided flows meet up
again they collide, losing their momentum like a waterfall hitting a river. This
ought to mean that a microflow would not be disrupted to anywhere near the same
degree by this mechanism because of its low inertia. But Forster went ahead and
built a scaled-down Tesla valve anyway.
He made a valve in which the main channels were around 100 micrometres
across, the width of a human hair, and included several side channels
(see Diagram).
Forster pumped liquid through this structure using a piezoelectrode to
move a membrane up and down very rapidly. This alternately pushed and pulled the
fluid through the valve.FIG-mg21425101.JPG
To his surprise the valve worked, although it was nowhere near as effective
as the full-sized version. Instead of a flow ratio of 90 to 1 in the preferred
direction, he got 1.1 to 1, or a tenth more liquid going forward than back. But
with rapidly oscillating fluid, that鈥檚 enough to ratchet the liquid along.
Why the valve works at all remains a mystery. Forster believes that the
liquid has to travel farther on the backward stroke as it passes through the
side channels, but this idea has yet to be proven.
Forster鈥檚 design is just one of many for microvalves, but devices that mix
fluids on this scale are harder to come by. In our macroscopic world, mixing is
easy. Anything that creates the chaotic swirls and eddies known as turbulence
will do鈥攕uch as shaking or stirring.
But neither of these methods works in a channel only 100 micrometres deep,
because the fluid鈥檚 viscosity prevents it becoming turbulent. Instead, the fluid
flows as smoothly and steadily as lava, a process called laminar flow. On the
molecular scale, the liquid molecules are travelling in neat layers that mix
with other layers only by gradual diffusion. Stirring is simply ineffective.
鈥淚t鈥檚 like trying to stir molasses,鈥� says John Evans, a graduate student in
Liepmann鈥檚 lab.
Instead, Evans has come up with a clever way to mix fluids based on the
theory that chaos can be induced in laminar flow even if it never becomes
turbulent in the conventional sense. Chaotic motion can be thought of as any
movement that causes two molecules which are close together to end up far apart.
Turbulence is a form of chaotic motion but there are others too.
Evans鈥檚 microfluidic mixer is an oblong chamber 100 micrometres deep in which
fluid is circulated by briefly sucking it out of the chamber at one point and
pumping it back in at another, then repeating this process at another pair of
points.
This alternated pulsing is the key to chaotic mixing. Without pulsing, a
water molecule would follow a smooth path from a source to a sink, all the while
staying close to the molecules it started out with. But the alternating pulses
give each water molecule the chance to jump from one trajectory, say to the sink
on the left, to another, towards the sink on the right. Crucially, there is no
guarantee that nearby molecules will follow the same course.
Bright future
After passing through the mixer several times, molecules that started out
next to each other end up far apart, says Evans. 鈥淥nce they cross the middle,
they go someplace completely different.鈥�
Evans tested a crude version of the mixer last year. He filled a chamber half
with water and half with Indian ink. Without pumping, it took 30 minutes for the
two liquids to mix by diffusion. But when Evans switched on the pumps, he got
thorough mixing in just 30 seconds. The mixer used Liepmann鈥檚 bubble valves,
which still work poorly, and Evans believes better valves would improve the
speed of mixing by allowing more rapid pulsing. He is currently designing a
flexible flap type of valve.
With the devices needed to control microflow taking shape, many researchers
are already concentrating on the commercial applications of microfluidics.
Becton Dickinson, a New Jersey syringe manufacturer, has helped fund Liepmann鈥檚
research and hopes to use bubble pumping to deliver insulin for diabetics.
Large-scale manufacturing will be the key to the commercial success of
microfluidics. Fabricating such precise machines is expensive, and the required
investments can only be recovered if there is a big market for the products.
However, commercial companies aren鈥檛 the only ones interested in the
technology鈥攁 large amount of funding comes from the Defense Advanced
Research Projects Agency, the research and development arm of the US military.
DARPA envisions a soldier of the future carrying a device the size of a credit
card that鈥檚 loaded with antibiotics or antidotes to poison gas.
Medicines that normally spoil without refrigeration could be stored dry in
single doses, dissolved on demand and injected at the push of a button. And a
microfluidic wristwatch could sample the air continuously and warn of chemical
or biological weapons attacks.
In theory, it could determine the size, shape and genetic content of microbes
in the air to home in on an anthrax germ hidden in a background of harmless
mould spores or skin bacteria. Today, such detectors are so large that they have
to be mounted on Jeeps.
Forster is optimistic about the future. He believes microfluidics will
eventually be commonplace, pointing to the way that microelectromechanical
systems (MEMS) are beginning to invade everyday life. After all, there are
already micromechanical accelerometers in every car air bag system, he says. If
microfluidic machines live up to their potential, they may become an even more
pervasive, if invisible, part of our lives.