Scientists at Cambridge University have found a new way of creating carbon nanotubes with properties that promise to revolutionise the design and manufacture of processors and micro-mechanical devices.
The breakthrough has been hailed as a way of extending the life of Moore's Law, which predicts that computing power will double every 18 months. In fact, it makes the law look curiously dated.
Carbon is unique in the variety and uses of its forms, which include graphite, diamonds and 'bucky balls', so called because of their structural similarity to the geodesic domes invented by maverick engineer Buckminster Fuller.
Nanotubes look like a tube of chicken wire patterned in hexagons or pentagons, except that the tubes are as small as 1.6 nanometres (thousand millionths of a metre) in diameter. By contrast, the latest Extreme Ultraviolet Light technique for making silicon chips can draw features 'only' down to 30 nanometers.
The discovery of nanotubes 10 years ago caused a lot of excitement because they can act as semiconductors, conductors and superconductors, or be formed into immensely strong structures for use in nano-robotics. The snag was that they could be made only as a tangled mat of tubes of varying characteristics.
Scientists from IBM in the US published a paper in Science magazine describing a way of separating out semiconductor nanotubes by burning off the conducting variety with a suitable current.
This rather overshadowed news of the breakthrough by a team led by Mark Welland at Cambridge's Nanoscale Science Laboratory, whose paper was posted at the Science website at much the same time. Even more confusing, one of Professor Welland's co-workers, credited on the paper, is Maria Seo of IBM Zurich.
Greater potential
But the project is not connected with the US IBM work and it seems to have much greater potential.
It grew out of a discussion between team leader Welland and Jim Gimsewski of California's UCLA, who also worked on the project. "We were just talking over a beer and began to discuss the possibility of filling a nanotube with a conductor. Then we started looking at ways it could be done," he said.
Using well-established procedures they vaporised bucky-ball carbon and nickel several times in succession through a mask of 300 nanometer holes onto a mollybdenum substrate to form a grid of tiny piles consisting of alternate carbon and nickel layers.
These were then heated within a strong magnetic field with the idea of trying to get the carbon to wrap round the nickel. To their astonishment the nickel disappeared and they got an array of perfectly formed nanotubes. "We spent the next six months making sure we had done what we thought we had done," said Welland.
Their excitement is understandable, as it was a new way of creating processors. The magnetic field can be replaced by a local electric field and the position of the mask altered to nanometer accuracy during vaporisation to create tubes of any required shape or orientation. The electronic characteristics can also be precisely controlled.
The nanotube's dimensions are not the only aid to miniaturisation, and thus to the amount of computing power you can pack on a chip. Nanotube chips, unlike silicon ones, do not require dopants, the correct concentrations of which are hard to sustain when the total number of molecules gets very small.
The repeal of Moore's Law
Moore's Law in this context seems like a measure of an antique technology, reflecting the early priorities of Intel. Processing power is no longer the be all and end all of chip design, as Intel itself has been discovering.
The company has its work cut out to persuade people to pay a premium even for today's fastest processors, and it is hard to see what the average user is going to do with the 10Ghz chips projected for the fairly near future.
Welland sees a need for very low-cost chips using a minimum of power - something, he says, of the capacity of today's PCs that will run for two years on a couple of AA batteries. Screens also drain a lot of power, and even here nanotubes can help: they may be used for electrodes in emerging organic screens.
Nanotubes are causing a rethink in the logical structure of processors, as well as their physical design. The very tasks for which more computing power may well be needed, such as photo-realistic 3D graphics, or handwriting and speech recognition, are those for which classic silicon architecture is least likely to provide the best solution.
And, as Welland points out, if Intel engineers were to start from scratch to design a processor for today's mainstream tasks, they would not come up with the Pentium 4. Computers do not need to be purely binary. "You could use four or five or six levels. You don't need to stick to two. You can even use analog computers," he said.
"The x86 design won't go away. It has too much momentum. There is no doubt that people will use nanotube processors to run x86 code because manufacturers will seize on anything that will make money. But all technology has its optimal uses and it might be that some other architecture will be optimal for nanotubes," he concluded.
See also:
In the wake of Intel's high profile Pentium 4 launch, vnunet.com chips away at the hype and delves deep into the technology which drives the chip giant's next-generation processor architecture. 22 Dec 2000All Computer Components