THIS PAST WEEKEND, The New York Times took the temperature of Moore’s Law, the fifty-year-old notion that the number of transistors packed into a computer chip will double every 18 months.
The “law” has proven true, time and again, since it was laid down in the mid-1960s by Gordon Moore, co-founder of Intel, now the world’s largest chip maker, and this played an essential role in the rapid evolution of computing devices over the past five decades. More transistors mean you can juggle more data at greater speed, which, in turn, drives the progress of everything from smartphones, smart watches, and smart thermostats to the vast networks of computers that underpin Google, Facebook, and Twitter. But now, Moore’s Law is slowing down. Transistor counts double only every two years or so, and some fear that progress will soon hit a wall. In The Times, the headline read: “Smaller, Faster, Cheaper, Over.”
'It's a major step towards having reassurance that we'll have semiconductors working beyond the limits of silicon. Moore's Law can continue on.'RICHARD DOHERTY, ENVISIONEERING
The way forward is unclear. But today, IBM unveiled research that could provide some guidance. In a paper published in the journalScience, company researchers described a new means of building transistors with carbon nanotubes—microscopic sheets of carbon rolled into cylinders—and this could eventually yield viable transistors that are significantly smaller than what we have today. To wit, it provides new hope for Moore’s law.
“I don’t think that chip-scaling will be ended by physics,” says Wilfried Haensch, a senior manager in IBM’s research arm who helps oversee the company’s chip work. In other words, Haensch believes that materials like carbon nanotubes can continue to yield smaller transistors for years to come. he only thing that will stop Moore’s Law, he says, is the economics of building these transistors for a mass audience.
The Incredible Shrinking Transistor
Scientists have already shown that carbon nanotubes can operate as transistors—electrical switches—when shrunk down to fewer than 10 nanometers, about 1,000 times smaller than a human hair. And the thinking is that we can shrink these tiny tubes much further. But below the 10nm range, carbon nanotubes haven’t maintained the performance of larger transistors—until now. IBM has shown that it can build 10-nanometer carbon nanotubes without sacrificing speed.
This is very much a research project. “It’s still in the concept phase,” Haensch says. IBM has yet to prove it can use its methods to produce commercial chips en masse. But its research shows that carbon nanotubes could provide an alternative when other materials hit the proverbial wall.
Today, companies like Intel build chips from silicon, offering commercial products whose smallest features are about 14 nanometers. But Intel plans to release 10nm silicon chips over the next two years. And IBM has previously said that can build a chip from silicon and germanium whose smallest features are in the 7nm range, planning to license its designs to various chip makers. With its carbon nanotube work, the company hopes to take this transistor trend even further. IBM believes it can eventually build carbon nanotube transistors as small as 5 nanometers—and maybe even smaller.
Out the Door
A transistor is essentially a switch made from three basic components: a source, a gate, and a drain. When a certain voltage is applied to the gate, current flows through a channel from the source to the drain, and the transistor is “on.” Apply another voltage, the current stops, and the transistor is “off.”
In shrinking carbon nanotubes, which serve as the channel of the transistor, scientists must also shrink the source and the drain, known as “contacts.” In the past, Haensch says, IBM has successfully shrunk these contacts to less 10 nanometers, but this came at a cost. As the contacts got smaller, electrical resistance increased, and that hindered the transistor’s performance. “The resistance shoots up so high, the device that you want to control doesn’t really matter anymore,” Haensch says. “Everything is dominated by the contacts.”
'You always have enough space to get out of the car. There is no limiting factor.'WILFRIED HAENSCH, IBM
To explain this, Haensch compares a chip to a parking garage. Just as he and his team are working to squeeze as many transistors as possible into a chip, a dude running a parking garage wants to squeeze in as many parking spots as he can. But he can’t shrink the spots without thinking about people getting in and out of their cars. “If park your car, you get really upset if you can’t open the door,” he says. “The space that you have to open your door is kind of like the contact. You have to make sure the current can enter and leave.”
So, Haensch and his team have changed the way they build contacts, welding a metal called molybdenum to the ends of those microscopic carbon nanotubes, and this keeps resistance low. “You always have enough space to get out of the car,” he says. “There is no limiting factor.”
This also means that the team can adjust the size of the channel and the contacts to suit their needs. “We can scale channel and contact in the same way. Or we can make the contact longer and the channel shorter,” Haensch says. “This gives you more flexibility as you try to reach your scaling goal.”
Richard Doherty, the director of technology consulting firm Envisioneering and an electro-physicist by training, sees IBM’s work as a vitally important breakthrough. It gives carbon nanotubes significantly more potential. “It’s a major step towards having reassurance that we’ll have semiconductors working beyond the limits of silicon,” he says. “Moore’s Law can continue on.”
0 Comments