In what may prove a major step toward nanoscale electronics, researchers at IBM have built the first array of transistors that use carbon nanotubes.

The discovery "may be tremendously important in nanoelectronic devices-if we ever get to that point," says Richard Smalley, winner of the Nobel Prize in 1996 for his discovery of the related carbon molecules dubbed buckyballs.

Wonder Wire

For nearly a decade, the carbon nanotube-also known as a buckytube-has been heralded as the wonder wire of the future. Excellent conductors with amazing strength, the tiny tubes come in all manner of shapes, sizes and electrical properties. That has led scientists to envision a wide range of applications-from nanoscale electronics to super materials, to tiny machines.

A carbon nanotube is a sheet of carbon atoms joined in a pattern of hexagons and rolled into a cylinder (think chicken wire). Where the two ends wrap around and meet determines the conductive properties of the nanotube. Line the ends up one way, and the nanotube conducts electricity like a metal. But line them up another, and the nanotube behaves like a semiconductor. Roll one nanotube around another, and you get a multi-walled nanotube-a metal-type inside a semiconductor inside a metal-type, for example.

But the great variety of nanotubes is also a source of great frustration.

When scientists grow nanotubes in the lab, they get every possible variety. To find the desired type, they must examine the tubes one at a time-an excruciating process. To compound the problem, nanotubes tend to bunch together into "ropes," and a metal-type nanotube will short-circuit its semiconducting neighbors.

"The stuff is like sticky spaghetti," says Philip Collins, a member of the IBM research team who now works for Covalent Materials, a nanotechnology startup. "Within each bundle there might be 10 or 100 nanotubes, but some fraction are going to be metal, and some fraction are going to be semiconducting."

Constructive Destruction

In a paper published in this week's Science, researchers at IBM's T. J. Watson Research Center in Yorktown Heights, NY, described a new method to separate the wheat from the chaff-in this case, the semiconducting nanotubes from their metal-type neighbors.

The scientists-project leader Phaedon Avouris, Michael Arnold and Collins-call their separation method "constructive destruction." Constructive because they scatter a layer of nanotubes over a silicon wafer and then create a pattern of electrodes, with some electrodes intersecting the nanotubes and some electrodes nearby. The nearby electrodes act as gates, switching off the semiconducting nanotubes. Destructive because the intersecting electrodes zap the metal-type tubes with a current large enough to burn them up. What's left are only the semiconducting tubes, which can be turned on and off by the gate electrodes.

"It's the difference between being able to make a thousand of these devices in a day and making one such device over a couple of days," says Tom Theis, director of physical sciences at the Watson Center.

Theis predicts that "constructive destruction" will let researchers design and study thousands of nanoscale devices on a single chip. But, he notes, the method is still too laborious to make its way to assembly lines.

In the long run, engineers need to be able to grow the kind of nanotube they want where they want it, says Mark Ratner, professor of chemistry at Northwestern University. "Instead of making the red and green M&Ms and separating them," he says, "you want to make just the red or green M&Ms."