Nanowires Will Make Computers Smarter

How much more powerful can computers get? Much more than we thought, argue American and British engineers. They cite the recent development of nanowire transistors and transistor-simulation tools.

“As a society, we have come to expect the continuation of Moore’s law,” says Eric Stach, associate professor of materials engineering at Purdue University. Moore’s law is the 1965 assertion by Intel co-founder Gordon Moore that the number of transistors that could fit on one integrated circuit would double every two years, and so computers would continue to expand their processing capacity while the price of computer processing would decrease.

Most engineers today doubt that the law can hold true forever, since the silicon-based semiconductors that constitute transistors can only get so small. Stach expects that transistors will reach their limits in the next five to 10 years.

But something else will also take place in five to 10 years, he adds: commercial use of nanowire tranistors.

Stach and other Purdue researchers are working with engineers from IBM and UCLA to develop these new transistors, which are made with silicon and germanium nanowires. The silicon nanowires stand vertically, not horizontally like traditional transistor components, so many more nanowire transistors can fit on one circuit.

The new transistors are also more powerful. The elements that compose them sharply delineate at the atomic level — one solid layer of germanium atop one solid layer of silicon. This is more conducive to effective transistor performance than conventional transistors, whose layers gradually transition from one element to the next. These special features enable them to bypass traditional transistors’ size limitations, Stach explains.

“These structures are being investigated to continue Moore’s law,” he says. Where it might lead is anybody’s guess.

“What will people do with enhanced computation power? Lots!” says Stach.

Existing transistors are already pretty small. Some are no more than a few molecules in size, according to Asen Asenov, University of Glasgow professor of device modeling.

Even smaller ones are possible, he explains, but certain technical problems need to be overcome. One major obstacle to making smaller transistors is that, the smaller they get, the more affected they are by atomic-scale imperfections and variations within the transistor, and thus the more likely the entire microchip will fail to perform as well as it could. To keep shrinking the transistors and improving performance, it is necessary to account for the variability.

Asenov and his Glasgow colleagues have found a way to do so. Working with scientists at Edinburgh, Manchester, Southampton, and York universities, they have devised “simulation tools” that predict how billions of microscopically different transistors will perform if placed together on a computer chip.

These simulation tools will help researchers place nanotransistors in the most optimal arrangements possible. Smaller transistors and more powerful microchips can result.

Moore’s law will reach its end point somewhere, according to Asenov, because no transistor — not even a nanowire one — can be smaller than an atom.

“Because we are hitting the atomic limits, Moore’s law cannot continue forever,” says Asenov. Nevertheless, vast improvements in microchip design and function are still possible. “The end of scaling is not the end of the microchip improvements,” he says.

— Rick Docksai

Sources: Eric Stach, Purdue University, www.purdue.edu.

Asen Asenov, University of Glasgow, www.gla.ac.uk.