Silicon Valley might be the current hotbed for tech, but silicon will likely one day be a thing of the past. To increase speed and power of computers, researchers make switches on microchips called transistors smaller and smaller—but silicon is reaching its limit. Now, many scientists are hyper-focused on building hardware using carbon and other materials that can be manipulated at an even smaller scale. The promise of carbon lies with a material called graphene, which is only one carbon atom thick.
This week MIT researchers made a huge step in that direction: they created a new 16-bit microprocessor made of carbon nanotubes, which are made of graphene cylinders, they report in the journal Nature.
The microchip has 14,000 transistors, making it the largest of its kind to date. It’s certainly a big move from the first carbon chip, which only had several hundred, but it doesn’t even come close to the current power of silicon-based devices, have hundreds of millions of transistors.
As silicon transistors got smaller and more powerful over the past 50 years, things like home computers, smartphones and self-driving cars have became possible. But recently, that process has slowed as engineers approach the limits of how small they can make silicon transistors.
One solution to the end of transistor shrinkage is to make the transistors out of another material. Enter graphene, which is the strongest known material in the world and ten times more energy efficient than silicon. It’s estimated that a carbon nanotube processor could run three times faster than current chips using about one-third the energy.
The problem is, producing the nanotubes, called carbon nanotube field-effect transistors (CNFET) is messy and current processes create tubes with too many defects to scale up the process.
According to an MIT press release, the team tackled several major hurdles in the new project. They developed a circuit design that compensates for some of the natural defects found in the nanotubes. Prior to this work, nanotubes had to be 99.999999 percent pure to work in a processor, a standard that is currently unreachable. The new design means the tubes can be just 99.99 percent pure, a standard that is doable with current technology.
The team also developed a manufacturing process that eliminates more defective tubes. When the tubes are deposited on a chip, they often clump up.
“It’s like trying to build a brick patio, with a giant boulder in the middle of it,” co-author Max Shulaker, an electrical engineer at MIT, tells Maria Temming at Science News.
The team coated the chips with a special polymer and then washed it off, carrying away the bigger chunks and leaving the single nanotubes. They also developed a process for creating the two different types of transistors, n-type and p-type, needed in a microprocessor.
When they put it all together into a chip, it worked, and the processor was able to execute a set of instructions, printing out “Hello, World! I am RV16XNano, made from CNTs.”
“This work takes a big step forward and gets much closer to a commercial chip,” physicist Yanan Sun of the Shanghai Jiao Tong University in China, not involved in the study, tells Elizabeth Gibney at Nature.
The technology still has a long way to go, and, in the end, it may not prove feasible. Katherine Bourzac at Chemical & Engineering News reports that the first carbon nanotube transistor was created at IBM in 1998. But the difficulties in producing the nanotubes at scale dampened enthusiasm for the technology. Over the past decade, teams of scientists at Stanford and MIT have continued to plug away at the problems of carbon nanotubes. In July 2018, the team received $61 million from the Defense Advanced Research Projects Agency to refine the tech, resulting in the new techniques.
One big hurdle, however, is moving the technology to a factory. Most manufacturers want to be able to use the same machines and equipment they currently use to produce silicon chips. In reality, however, the material needs to be adapted to those industrial processes. “They have outstanding results in the research and lab space,” Greg Schmergel, CEO of Nantero, a company that makes carbon nanotube memory modules, tells Bourzac. “But it can be quite a shock as you’re moving from the lab to production stages.”
Shulaker, however, is optimistic that his team will be able to make that leap, estimating in the press release that commercial nanotube microprocessors could be available in less than five years. “We think it’s no longer a question of if, but when,” he says.