Sayonara, silicon? Engineers build first carbon nanotube computer

Tech soothsayers have long predicted the demise of computers as we know them today, as their shrinking sizes approach the limits of silicon’s ability to take the heat. Now, researchers at Stanford University — in the heart of Silicon Valley — have tossed the essential element aside and built a basic computer out of carbon nanotubes.

The engineering feat, described this week in the journal Nature, could herald the birth of a whole new generation of carbon-based computing devices, experts said.

For decades, the transistors that serve as the building blocks of computer chips have been based on silicon. And they’ve been getting smaller and cheaper at a pretty consistent rate — that’s the premise behind Moore’s Law.

But among the many advantages of silicon lies one significant drawback, said Max Shulaker, a graduate student in electrical engineering at Stanford who led the development of the newfangled computer.

“When we keep going smaller in silicon, it becomes less efficient,” he said. “You waste a lot of power.”


The tinier the silicon transistors get, the more heat they leak. They potentially could get so small that the chip would actually melt.

Engineers have long recognized the need to find a replacement for silicon. They’ve been working since the early 1980s on quantum computing, which takes advantage of the fact that individual particles of matter and light can be in more than one state at a time. They’ve also been experimenting since the 1990s with DNA computing, which stores data in double helixes.

But some researchers in the burgeoning field of nanotechnology had a simpler idea: Why not use carbon nanotubes in place of silicon? That would allow engineers to take advantage of silicon computing’s existing architecture.

“It’s good because it makes it much more easily commercialized,” Shulaker said. “There’s so much money and time invested in silicon. We’re not going to throw it away all of a sudden.”

Carbon nanotubes are rolled-up sheets of graphene, a form of crystalline carbon that’s only 1 atom thick. The tubes are extremely small — about a nanometer across — and have a host of fascinating properties that scientists have been exploring for a range of purposes, from super strength to electrical conductivity.

Their thinness — thousands of nanotubes lined up can fit across a human hair — is what makes them potentially useful to build transistors, Shulaker said. By cramming more transistors into a small space, carbon nanotubes could improve on silicon’s performance by an order of magnitude.

Other teams have managed to create carbon nanotube transistors, but not an entire functioning computer, Shulaker said. That’s because carbon nanotubes come with a few drawbacks. They have to be grown on a quartz substrate, and it’s difficult to make sure they’re all perfectly aligned. Even a few stray nanotubes could create a host of bad connections.

On top of that, some of the nanotubes act like they’re metallic and conduct electricity. To make a computer, the researchers needed to find a way to get rid of those and leave behind only the ones that were semiconducting, and thus be able to turn electrical current on and off.

The Stanford team, led by professors Philip Wong and Subhasish Mitra, came up with a two-pronged solution. First, they turned all of the properly functioning, semiconducting nanotubes “off” and then ran an electrical current through them. The current heated the remaining nanotubes until they fried in a puff of carbon dioxide, leaving only the properly functioning ones intact.

Even with the metallic ones gone, any remaining disorderly nanotubes could still cause a short-circuit. To work around the issue, the scientists came up with an algorithm for a circuit design that is guaranteed to work, even with a few mislaid strays.

The result is a computer with 178 transistors and only a one-bit processor. (For the sake of comparison, the computer on your lap or desktop contains a 32-bit or 64-bit processor that use millions or billions of transistors.) Building anything more complex will require an industrial facility, not a university lab, researchers said.

But it’s an impressive start.

“This computer is far from being competitive with current standards, but had this machine been made in 1955, it would have been,” Franz Kreupl of the Technical University of Munich wrote in a commentary that accompanied the Nature report. The team’s smallest carbon transistor is about 8 micrometers, he said, and scientists still need to slim that model down to a mere 20 nanometers.

However, Kreupl said, if researchers can build more powerful processors and scale down the size, “we might be able to type on one soon.”

For his part, Shulaker estimated that it would probably take 10 to 20 years to bring a functioning carbon nanotube computer to market.

“I don’t think Silicon Valley will become Carbon Valley overnight,” he said.