As silicon transistors get smaller, with billions now crammed onto a single microprocessor, the chips get less efficient. They run hotter and require more energy to keep cool. For the ever more powerful smartphones and speed-craving servers of tomorrow, silicon is reaching its physical limits. Chip designers may need to find a new substance.
Walt de Heer thinks he has it. In his nanotech lab at the Georgia Institute of Technology, the physicist has been experimenting with graphene, a sheet of carbon one atom thick. It can transmit electrons as much as 100 times faster than silicon. That translates into faster computers while generating less heat and requiring less power. Although graphene was discovered in the 1960s, it wasn't until 2001 that de Heer began tinkering with it as a possible replacement for silicon.
The challenge is mass-producing it. Previous researchers cut graphene sheets into narrow transistors, leaving frayed edges that disrupted the electron flow. In 2006, De Heer had an idea: Rather than cut the stuff, why not produce it in the exact shapes needed by the computer industry? De Heer starts with crystals of silicon carbide, a compound containing equal numbers of silicon and carbon atoms. He arranges the substance into the form of an H-shaped transistor and then heats it to 1,500C. Some silicon evaporates, leaving behind a perfectly shaped graphene transistor atop a thicker layer of silicon carbide. De Heer's findings will be published in the October issue of Nature Nanotechnology.
Chip manufacturers have provided de Heer with about $800,000 in funding since 2003. Mike Mayberry, director of components research at Intel (INTC), says it'll be 5 to 10 years before graphene can be used in large-scale production. "People aren't putting down a billion dollars to build a factory to manufacture it," says Mayberry. "But they might at some point."
De Heer, whose father worked in the oil industry, moved around a lot growing up. He was born in the Netherlands, lived in Saudi Arabia and Aruba, played bass in a rock band that toured the West Indies, and eventually settled at the University of California at Berkeley to pursue a PhD in physics.
He knows new findings often turn out to be letdowns. The 1992 discovery of carbon nanotubes, an unusually strong substance with high conductivity, generated excitement as a possible silicon successor. He saw that their irregular shape would make them hard to manipulate. "I was getting annoyed at the unbridled hype," he says. "It's just snake oil."
Even if graphene proves to be more promising than carbon nanotubes, silicon isn't going away anytime soon, de Heer says. While graphene might make sense for compact devices like cell phones or power hogs like server farms, silicon will be difficult to supplant. "Silicon is dirt cheap," he says. "Graphene can't make things cheaper, but it can go to much higher speeds and use much less energy."
Grew up in the Netherlands, Saudi Arabia, and Aruba
A new process for building transistors out of graphene
Tiny, powerful chips that don't require much energy