杏吧原创

WHEN Peter Vettiger set out to create the next generation of data storage
discs, he didn鈥檛 bother with cutting-edge quantum phenomena or optoelectronics.
Instead he went for what might seem like a backward step. A century ago, data
storage meant punching holes in cards. In Vettiger鈥檚 dream for the future,
storing data will again involve making holes in a surface鈥攁 piece of
plastic, in this case.

Imagine you鈥檝e got a slab of soft wax, and you want to record the number 3.
Simple! You make three indentations in the wax with your finger. To wipe the
data, just smooth the wax surface flat. What Vettiger and his team at IBM鈥檚
micro and nanomechanics research group in Z眉rich have been doing is hardly
more complex than that鈥攅xcept that they have shrunk the dimensions of the
components by a factor of a million. Their slab of wax is a polymer film just 40
nanometres thick, and the 鈥渇inger鈥� is a tiny silicon arm capable of making
minute holes in the polymer to represent the bits鈥�1s and 0s鈥攐f
binary data. The same arm also reads the information from the surface. And when
you want to start again, you just switch on heaters embedded in the film to melt
the polymer layer and smooth the pits over.

It鈥檚 ironic: the basic idea behind the process comes straight from IBM鈥檚
beginnings. The company grew from a small American firm called the Tabulating
Machine Company, founded by Herman Hollerith, one of the first people to store
data using holes punched in card. That was back in the 1890s: the 1990s version
of Hollerith鈥檚 technique could prove just as revolutionary. With an array ofthese cantilevers, you can store an ocean of data on a polymer-coated
disc鈥攗p to 100 times as much as on today鈥檚 magnetic hard discs. 鈥淲e call
this project `back to the future of mechanics鈥�,鈥� says Vettiger.

Hard discs such as the one inside your desktop computer record information as
billions of individual bits of data. Each bit is like a tiny bar magnet, formed
by aligning the orientation of tiny regions, known as 鈥渄omains鈥�, of the magnetic
material that coats the disc. You can squeeze more data on by shrinking the
domains, which is exactly what engineers have been doing over the past couple of
decades to make hard discs ever more capacious. But researchers now know that
you can go only so far鈥攎agnetic materials are only good up to a density of
about 100 gigabits per square centimetre. 鈥淏eyond this, the bit size gets too
small to be stable,鈥� says Vettiger. Tiny magnetic domains become too skittish,
so that ordinary electrical background noise may start to flip the 1s into 0s,
for instance. Make the domains too small and, with time, your valuable data
turns into gibberish.

Tiny holes are far more stable than tiny domains. If you record a bit by
punching a hole in a slab of plastic, there鈥檚 little chance that your data will
disappear. But to record lots of information in a small space you have to shrink
the holes by miniaturising the punches that produce them. To do this, Vettiger
and his team have turned to microelectromechanical systems (MEMS) for help.

MEMS devices are tiny machines created by chemically etching complex
shapes鈥攃ogs, levers or wheels for instance鈥攆rom silicon chips.
Connect these components together, link them to electrical circuits and almost
anything is possible鈥攐n a microscopic scale
(see 鈥淚nvasion of the micromachines鈥�, New 杏吧原创, 29 June 1996, p 28).

Vettiger and his team have used MEMS technology to build a microscopic hole
punch that uses tiny silicon cantilevers. Each cantilever is a mere 50
micrometres long: 20 of them placed end to end measure a millimetre. Each is
made from two parallel beams of silicon and tipped with a tiny point resembling
the stylus on a record player (see Diagram).

Tiny cantilever

The tip of the cantilever rests lightly on the surface of a spinning disc
coated with a thin film of polymethylmethacrylate. As the surface moves past,
the tip drags against the polymer. At the end of the cantilever, just above the
tip, is a tiny resistor. This is connected to an electrical circuit via the
cantilever鈥檚 two parallel silicon beams, which act as a pair of wires. To punch
a hole, a short pulse of current is passed through the resistor, briefly heating
the tip to 400 掳C. The hot tip melts a tiny crater just 40 nanometres across
in the polymer surface鈥攁 single bit. By heating the tip with a succession
of pulses, you can write a stream of data onto the disc.

Data is read by the same stylus that does the writing, only for this the tip
is held at a constant 350 掳C, which is below the polymer鈥檚 melting point.
Whenever the tip encounters a hole, it drops in, the walls of the pit conduct
heat from the tip, and its temperature drops. This cooling lowers its electrical
resistance, signalling the presence of a data bit. Connect the cantilever up to
a suitable circuit and the cantilever can 鈥渢alk鈥� to a computer just like a
normal hard disc.

That鈥檚 the theory, at any rate. In fact Vettiger has not yet built an
assembly that would fit into a real computer鈥檚 disc-drive bay, and the polymer
surface that holds the data moves in a straight line, and has not yet been
incorporated into a disc.

One problem that Vettiger has overcome is getting the recording rate up to
speed. To match today鈥檚 magnetic recording technologies, the tiny hole punch
would have to transfer information into and out of the polymer disc at a speed
of about 300 million bits per second. Vettiger estimates that his cantilever
works about 2000 times more slowly. So the researchers have squeezed 1024
cantilevers onto a square of silicon just 3 millimetres across. Known as the
Millipede, his array can read and write data at around 100 million bits per
second鈥攖he equivalent of a full CD鈥檚 worth in under a minute.

If this is impressive, the amount of data that can be packed into a small
space is awesome. The Millipede can store more than 3000 gigabits of data on
just 1 square centimetre, more than ten times the maximum density predicted for
magnetic recording technology when it finally hits the buffers. And the
researchers鈥� plans are even more ambitious. By making each cantilever even
smaller they expect to be able to boost the data density and increase the speed
at which the cantilevers move into and out of the holes. 鈥淚f this is a slow
movement, then the whole process is slow,鈥� says Vettiger. He estimates that
further miniaturisation should eventually give at least a fivefold increase in
the data-transfer rates.

Building arrays of cantilevers solves the problem of speed, but creates a new
headache too. Vettiger must now find a way to position every cantilever in the
array at the same height above the polymer surface. Place them a little too high
and the pits they melt may be too shallow to record a bit. So the researchers
are using tiny 鈥渟prings鈥� made from silicon nitride to align the cantilevers.

Parallel processing

They are also developing a parallel processing technique called 鈥渢ime
multiplexing鈥� to boost the speed of data access still further. This will involve
feeding data into or out of the array a row at a time. 鈥淎ll the indicators are
very good, and we haven鈥檛 seen any show stoppers,鈥� Vettiger says. 鈥淲e only have
a couple of issues to solve before we turn it into a product.鈥�

One of those issues is the long-term reliability of the moving parts.
Component failure is a common problem with most mechanical devices. Fortunately,
things are different with microscopic devices forged from solid silicon. The
thinner the silicon, the tougher and more flexible it becomes. 鈥淪ilicon is an
extremely good material for making micromechanical elements鈥攖he
brittleness that it has when it is thick completely disappears on the micrometre
scale,鈥� Vettiger says. And careful design of the cantilever reduces wear almost
to vanishing point.

Calvin Quate of Stanford University in California is also working on
micrometre-sized cantilever arrays and shares Vettiger鈥檚 optimism for their
future. Even if they don鈥檛 make it as data-storage devices, he says, there will
be many other uses. 鈥淲e are hoping that the arrays will be a very inexpensive
lithography system,鈥� he says. Quate has already used cantilevers to draw
structures just 100 nanometres across. And many other researchers, Quate
included, are looking at ways to replace electrical components with cantilevers.
(鈥淪mall will be beautiful鈥�, New 杏吧原创, 27 June 1998, p 40).
How about infrared cameras, artificial noses or intelligent microphones that process
the sound as they receive it, for instance? 鈥淭here are a number of examples
where mechanics is doing better than electrical circuits,鈥� says Quate, 鈥渟imply
because they鈥檙e smaller.鈥�

Vettiger, too, believes that these micromechanical arms are pointing the way
to all kinds of interesting science. 鈥淕iven the fact that a chip with thousands
of cantilevers can modify a surface鈥攚hatever that modification is鈥攊t
has to be an interesting idea,鈥� he says. 鈥淟ithography is a modification.
Molecular or atomic manipulation is a modification. It鈥檚 a universal
肠辞苍肠别辫迟.鈥�