Computers of the future could be even smaller, faster, and more efficient — all by doing away with the very material that currently comprises their core.
In a finding that builds on earlier work from several institutions, including IBM, a team of electrical engineers at Stanford University today announced the creation of the first-ever computer based on carbon nanotubes. The nanotubes are used to create a novel kind of transistor, one that doesn't rely on conventional silicon. "This is the most complex electronic ever built with carbon nanotubes," said Max Shulaker, a co-author on the paper announcing the progress, which is published in Nature. "There's been a lot of hype around this field, but people weren't actually sure if you could use them in a practical way like this."
Enhanced speed and energy efficiency
Carbon nanotubes, or CNTs, are infinitesimally small cylinders made from sheets of carbon atoms. When arrayed into transistors, the nanotubes are small enough that engineers can fit many more of them onto a single chip compared to silicon transistors. Their size, combined with other properties of the nanotubes — including high conductivity and rapid on-off ability — would mean enhanced speed and energy efficiency. That's particularly important given the inherent limits of silicon-based transistors: researchers have been doubling the number of transistors on a chip approximately every two years (a process known as Moore's Law) but that progress will within decades reach an end.
"When one chip has billions of transistors, that 2 percent becomes a big problem."
The computer that Shulaker and his colleagues build is extremely basic: it contains only 178 transistors, and runs a rudimentary operating system that allows for the completion of counting and number sorting tasks (as well as the ability to toggle between the two). Still, getting to this point has been years in the making — and required that researchers overcome two daunting challenges. Carbon nanotubes tend to self-assemble unpredictably, and that assembly process also yields some nanotubes with metallic properties, which will cause a transistor to short-circuit. "We had to do a lot of work to get these nanotubes to align how we wanted, and to be perfectly reliable," Shulaker said. "Even 98 percent isn't good enough; when one chip has billions of transistors, that 2 percent becomes a big problem."
To overcome those challenges, the Stanford team used what they've dubbed as "an imperfection-immune design." The approach essentially uses electricity to vaporize metallic nanotubes during the fabrication process, and relies on an algorithm to create circuit designs that function despite any misalignment of the nanotubes. According to Shulaker, the methods can be scaled to support industrial manufacturing of carbon nanotube transistors in the future. "You can't expect to make these transistors one-by-one," he said. "I think what this shows is that we can actually start to compete with silicon."
"Maybe one day, Silicon Valley will become Carbon Valley."
Despite the exciting progress, Shulaker is also realistic about the prospects of electronics running on carbon nanotubes anytime soon. Additional research will be necessary to yield more sophisticated CNT-based computers, and the nanotubes are only one of several emerging technologies with the potential to replace silicon. Not to mention that any major disruption to the silicon-based electronics industry would be a formidable challenge. "I'd be naive to think that one fine morning, the industry will just wake up and toss silicon out the window," he said. "But I like to think that, maybe one day, Silicon Valley will become Carbon Valley."