While at least 10 years away from commercialization, the new chip prototype is significant because it may be a way for chipmakers to continue to follow Moore’s Law, the observation by Intel co-founder Gordon Moore that the complexity of chips doubles roughly every 18 months.
Moore’s Law is the pulse of the chip industry and is thought to be sustainable for another several chip generations of silicon chips. For beyond then, Intel, IBM and others are furiously working on new types of materials.
IBM’s announcement is promising for a number of reasons, not least because it was built using standard chip-making processes. Because silicon likely won’t completely be replaced with another type of material, at least not initially, being able to use existing chip-making techniques for some kind of future silicon-hybrid chip would be a boon. After all, chip-manufacturing equipment costs many millions, and a single chip-fabrication facility sets chipmakers back at least $1bn.
Also, IBM’s work was done using a single carbon nanotube molecule, rather than linking together a series of components, which promises relatively simple manufacturing and testing, said one of the IBM researchers, Joerg Appenzeller.
Carbon nanotubes are cylindrical molecules of carbons, which under a high-powered microscope look something like a roll of chicken wire. They are roughly 50,000 times thinner than a single strand of human hair. A human hair is about 80,000 nanometers wide.
As often is the case with nanotechnology, which may be defined as the manipulation of matter at a scale of typically between 1 and 100 nanometers, carbon nanotubes exhibit unique quantum qualities.
Carbon nanotubes are a hollow cylinder made only of carbon atoms. The circumference of a carbon nanotube is just a couple of nanometers. The dot in this letter i is roughly one million nanometers in diameter. It is this tiny size that gives nanotubes extreme quantum conditions for the electrons traveling on them, Appenzeller said.
These quantum conditions mean nanotubes are very efficient conductors of heat. Nanotubes don’t resist a flow of electricity, unlike a regular semiconductor wire that becomes heated as current flows through it. It ensures the nanotube behaves like a ballistic conductor, Appenzeller said.
If carbon nanotubes were used as channel material in chips instead of silicon, chipmakers may not be facing the heat and power issues they do today. It also means carbon nanotubes may carry higher current densities than the silicon channels used in existing chips.
With their tiny size, carbon nanotubes might also allow for further miniaturization and enable chipmakers to keep up with Moore’s Law. The problem with silicon is that to make chips smaller, the channel through which electrons pass needs to be made shorter.
In turn, a chip’s gate, which essentially opens or closes the flow of electrons, also has to shrink. For a gate has to be very close to the channel in order to work. There also is an insulation layer between them, and that oxide layer must become thinner in order to shrink both the channel and the gate. However, when the insulation becomes too thin, electrons tend to leak – and all proverbial hell can break loose.
But, the smallness of the nanotubes allows you to be more relaxed on that oxide thickness, yet still getting the same scaling advantage for channel length, Appenzeller said. In other words, leakage and other problems go away with carbon nanotubes.
By integrating a complete chip around a single nanotube, the IBM team saw circuit speeds nearly one million times faster than previous demo chips with multiple nanotubes. While this is a couple of hundreds times slower than existing silicon chips, Appenzeller pointed out that semiconductor technology takes time, manpower and a lot of effort.
Look at the time it takes to make progress in the silicon world, from the first 30 megahertz chip to gigahertz, he said. Nobody can expect a small research group to come up with something that bypasses almost 50 years of silicon.
He likens the research to building a car. Now we are going to make it faster. And that is less of a problem now because we have learned so much during the process of trying to build this car, which is the circuit, that the next step is pretty clear.
The research work took roughly 18 months to complete, and included input from scientists at the University of Columbia and the University of Florida, who grew the nanotubes.
IBM and other corporate and academic scientists also are looking at various other potential future chip materials, including carbon nanowires.
Appenzeller and his team are evaluating nanowires for the same purposes as nanotubes.
Nanowires are a much closer cousin of silicon than nanotubes, making them a potentially easier material to deal with. Nanotubes are more difficult to grow and to use them with the kind of precision a chip requires. However, nanowires are not as exciting in terms of electrical properties, Appenzeller said.
While their nanotube work is further developed, the jury is still out on which technology will make it to market, he said. We definitely have not made any decision in that respect, that’s why we are actively looking at both.
