A New System for Storing Data: 
Think Punch Cards, but Tiny

New York Times
June 11, 2002
http://www.nytimes.com/2002/06/11/science/physical/11DATA.html

I.B.M. scientists say they have created a data-storage technology 
that can store the equivalent of 200 CD-ROM's on a surface the 
size of a postage stamp.

Writing in the current issue of the journal IEEE Transactions 
on Nanotechnology, researchers at I.B.M.'s laboratories in 
Zurich report that they have achieved a storage density of 
one trillion bits of data per square inch, about 25 times 
as great as current hard disks.

Dr. James C. Ellenbogen, an expert on molecular electronics 
at the Mitre Corporation in McLean, Va., described the work 
as "incredible engineering."

Still more remarkable, this new technology is a return to 
an obsolete one, at least in concept. Like computer punch 
cards - which were invented more than a century ago and 
went out of vogue in the 1970's, about the same time as 
slide rules - I.B.M.'s system stores data in a pattern of 
little holes.

But I.B.M.'s holes are much, much tinier - half of a 
billionth of an inch across. While mechanical devices have 
steadily given way in recent decades to electronic ones 
that are faster, cheaper and more reliable, that trend
may reverse at the molecular scale, where friction and 
wear and tear act differently.

"Back to the future of mechanics," said Dr. Peter Vettiger, 
leader of the I.B.M. project, known as Millipede. Millipede 
has another advantage over punch cards: the holes can be 
closed up so that data can be rewritten over and over.

Nantero, a start-up company in Woburn, Mass., is also taking 
a mechanical approach. Scientists there are making computer 
memory using nanotubes - rolled-up sheets of carbon graphite - 
that open and close like mechanical switches. 

"It's counterintuitive because people assume in a computer-
type application that if you have moving parts, it will be 
too slow and also it will break," said Greg Schmergel, 
president and chief executive of Nantero.

In electronic devices, data are stored in bundles of 
electrons, and as electronics shrink, the bundles contain 
fewer and fewer electrons. Smaller bundles of charge fall 
apart more easily.

Hard disks run into similar problems as storage densities 
rise. Instead of electrical charge, hard disks store data 
in what is essentially a vast array of tiny magnets. 
But if a magnet is shrunk too small, vibrations of heat 
can make the magnet flip, destroying data.

A hole, even one a few atoms wide, is a more resilient 
structure.

The Millipede project grew out of invention of the scanning 
tunneling microscope in 1981 by Dr. Gerd Binnig and Dr. 
Heinrich Rohrer at the I.B.M. Zurich laboratories. 
By measuring the current passing between a sharp microscopic 
silicon tip and a surface, the new microscope generated images
of individual atoms. Dr. Binnig and Dr. Rohrer won the Nobel 
Prize in Physics in 1986.

Dr. Binnig said he realized early on that the silicon tip 
might be used to poke holes in a surface. "I certainly realized 
the method is usable for storage," he said.

Six years ago, Dr. Binnig and Dr. Vettiger started working 
on the Millipede project in earnest.

The Millipede chip consists of a layer of plexiglass a couple 
of billionths of an inch thick laid on a silicon chip. To write 
a bit of data, a microscope tip, heated to 750 degrees Fahrenheit, 
softens the plexiglass and dents it. 

To read data, the tip is heated to 570 degrees - not hot enough 
to deform the plexiglass - and pulled across the surface. 
When it falls into a dent, the tip cools because more surface 
area is in contact with the cooler plexiglass. That temperature 
drop reduces its electrical resistance, which can be easily 
measured.

To erase data, a hot tip is passed over the dent, causing it 
to pop up.

In the latest work, the I.B.M. researchers show that they 
can now erase individual dents; previously, they had to erase 
large patches at once. They have also shown that they can 
reliably erase and rewrite data. 

Millipede still suffers a big drawback of mechanical systems. 
Reading and writing data with a single silicon tip takes 
about 1,000 times as long as with hard disks.

To compensate, a second prototype chip uses 1,024 silicon tips 
to read and write data in parallel, bobbing up and down like 
a flock of birds pecking at dirt over a square area about 
a tenth of an inch wide.

The work is still years from becoming a commercial product, 
but it has progressed far enough that Dr. Binnig, 55, will 
retire from I.B.M. at year's end to spend more time on 
interests like painting and composing music. "It's in a state 
where all the big problems are solved," he said.

Last year, Dr. Charles M. Lieber, a professor of chemistry 
at Harvard, and Dr. Thomas Rueckes, a graduate student, reported 
that they had made computer memory out of nanotubes. Each bit 
of data is stored in two of nanotubes placed at right angles 
and separated by a small space. Applying a voltage to the 
tubes creates an electric field that pushes them together.
Once stuck, the nanotubes remain held together by molecular 
forces even when the voltages are turned off. That means data 
would remain in memory when the computer was turned off. 

Applying an opposite voltage separates the tubes.

After completing his doctorate, Dr. Rueckes, along with 
Mr. Schmergel, founded Nantero in October. They aim to produce 
a prototype by the end of next year and to go into production 
within a year after that.

They also think they could produce much higher densities 
than could be possible with conventional memory, where 
neighboring bundles ofelectrical charge interfere with one 
another if pushed too close.

With the nanotube memory, they need only to move the nanotube 
pairs closer to each other.

"Neighboring bits don't affect each other," Mr. Schmergel 
said. "It's just mechanical."