The effect of five decades of exponential progress with silicon chips doubling in speed every couple years as observed by Intel cofounder Gordon Moore in 1965 cannot be overstated. As silicon-based transistors push against the limits of physics, the death of Moore’s law could pack a devastating blow to the industry and even the global economy. It’s a big problem that has chip makers, like IBM, Intel and others, scrambling for a workaround. One of the most promising strategies for extending Moore’s law involves using carbon nanotube-based transistors.
Currently, Intel makes most of its CPUs on a 22nm manufacturing process, and its smallest silicon transistor measures 14 nanometers. The semiconductor industry group, ITRS, anticipates that the five-nanometer “node” will debut in 2019. It’s a point that may very well spell the death of silicon from a practical standpoint. That’s the opinion of Wilfried Haensch, who heads up IBM’s nanotube project at the T.J. Watson research center in Yorktown Heights, New York.
“That’s where silicon scaling runs out of steam, and there really is nothing else,” says Haensch in an article on MIT’s Technology Review.
When this day comes, IBM wants to have its carbon nanotube-based processors ready to roll out. It’s a plan that’s been many years in the making.
IBM’s history with carbon nanotube transistors dates back to 1998, when company researchers showed that it was a viable approach by building one of the first working prototypes. Now IBM is working to bring the technology to commercialization.
According to simulations carried out at T.J. Watson research center, the design that IBMers are implementing will be five times faster than silicon-based microprocessors using the same amount of power. The technology, while very real, is still in the design stage, however, and there are no guarantees it will pan out.
IBM obviously has a lot of investment sunk into the silicon-based manufacturing process so naturally the company is focusing on building a carbon-based transistor using similar design and manufacturing methods. The research group recently made chips with 10,000 nanotube transistors, using six-packs of nanotubes, each 1.4 nanometers wide and 30 nanometers long. The ends of the tubes make contact with electrodes which supply current, while a third electrode runs underneath and acts as a switch.
At this stage of design, the researchers cannot get the nanotubes close enough because existing chip technology doesn’t operate at that scale. They are working on a solution that would cause the tubes to self-assemble into position. The helper compounds would then be removed, leaving the nanotubes in the proper configuration ready for the electrodes and other circuitry to be added.
A lot is riding on the research. If the nanotube transistors are not ready in time to meet the post-silicon demand, they may miss their market opportunity, according to IBM’s James Hannon, head of the company’s molecular assemblies and devices group. But there’s not a lot of other options out there. Possibilities like spintronics exist, but they’re less mature, and don’t have the advantage of behaving like silicon transistors, so they wouldn’t be compatible with existing semiconductor manufacturing techiques.