A chip made with carbon nanotubes, not silicon, marks a computing milestone

“Silicon Valley” may soon be a misnomer.

Inside a new microprocessor, the transistors — tiny electronic switches that collectively perform computations — are made with carbon nanotubes, rather than silicon. By devising techniques to overcome the nanoscale defects that often undermine individual nanotube transistors (SN: 7/19/17), researchers have created the first computer chip that uses thousands of these switches to run programs.

The prototype, described in the Aug. 29 Nature, is not yet as speedy or as small as commercial silicon devices. But carbon nanotube computer chips may ultimately give rise to a new generation of faster, more energy-efficient electronics.

This is “a very important milestone in the development of this technology,” says Qing Cao, a materials scientist at the University of Illinois at Urbana-Champaign not involved in the work.

The heart of every transistor is a semiconductor component, traditionally made of silicon, which can act either like an electrical conductor or an insulator. A transistor’s “on” and “off” states, where current is flowing through the semiconductor or not, encode the 1s and 0s of computer data (SN: 4/2/13). By building leaner, meaner silicon transistors, “we used to get exponential gains in computing every single year,” says Max Shulaker, an electrical engineer at MIT. But “now performance gains have started to level off,” he says. Silicon transistors can’t get much smaller and more efficient than they already are.

Because carbon nanotubes are almost atomically thin and ferry electricity so well, they make better semiconductors than silicon. In principle, carbon nanotube processors could run three times faster while consuming about one-third of the energy of their silicon predecessors, Shulaker says. But until now, carbon nanotubes have proved too finicky to construct complex computing systems.

One issue is that, when a network of carbon nanotubes is deposited onto a computer chip wafer, the tubes tend to bunch together in lumps that prevent the transistor from working. It’s “like trying to build a brick patio, with a giant boulder in the middle of it,” Shulaker says. His team solved that problem by spreading nanotubes on a chip, then using vibrations to gently shake unwanted bundles off the layer of nanotubes.

Another problem the team faced is that each batch of semiconducting carbon nanotubes contains about 0.01 percent metallic nanotubes. Since metallic nanotubes can’t properly flip between conductive and insulating, these tubes can muddle a transistor’s readout.

In search of a work-around, Shulaker and colleagues analyzed how badly metallic nanotubes affected different transistor configurations, which perform different kinds of operations on bits of data (SN: 10/9/15). The researchers found that defective nanotubes affected the function of some transistor configurations more than others — similar to the way a missing letter can make some words illegible, but leave others mostly readable. So Shulaker and colleagues carefully designed the circuitry of their microprocessor to avoid transistor configurations that were most confused by metallic nanotube glitches.

“One of the biggest things that impressed me about this paper was the cleverness of that circuit design,” says Michael Arnold, a materials scientist at the University of Wisconsin–Madison not involved in the work.

With over 14,000 carbon nanotube transistors, the resulting microprocessor executed a simple program to write the message, “Hello, world!” — the first program that many newbie computer programmers learn to write.

The newly minted carbon nanotube microprocessor isn’t yet ready to unseat silicon chips as the mainstay of modern electronics. Each one is about a micrometer across, compared with current silicon transistors that are tens of nanometers across. And each carbon nanotube transistor in this prototype can flip on and off about a million times each second, whereas silicon transistors can flicker billions of times per second. That puts these nanotube transistors on par with silicon components produced in the 1980s.

Shrinking the nanotube transistors would help electricity zip through them with less resistance, allowing the devices to switch on and off more quickly, Arnold says. And aligning the nanotubes in parallel, rather than using a randomly oriented mesh, could also increase the electric current through the transistors to boost processing speed.