The intent of current electronics research, and the challenge to chip makers, is to find a way to make the transistor small enough to process information at much greater speeds than is possible with today's computers. But researchers have come up against a huge problem: size and shape really matter.
The problems involved in speeding up electronic circuitry were spelled out by Mark Rodwell, professor of computer and electrical engineering at the University of California, Santa Barbara, in a Feb. 24 lecture aptly titled "Making Devices Small to Make Them Fast."
Over the last few years, said Rodwell, chip makers have been producing ever-faster high frequency semiconductor devices simply by making them smaller.
Since one of the most important factors in determining the speed of a transistor is the time it takes for an electron to travel between the transistor's components, miniaturization has enabled a reduction in this travel time. But now a limit to shrinkage seems to have been reached because it has been discovered, said Rodwell, that the characteristics of the transistor change when its components are scaled down beyond a certain point.
As one answer to this problem, Rodwell proposes changing the shape of the transistor from square-shaped to something like a long, narrow line.
To help understand why the new shape works better, think of an elephant, said Rodwell. If an elephant was made 10 times bigger in all dimensions, its legs would simply give way. That's because its total volume would have increased 1,000-fold, but its footprint would only be 100 times larger. And every square inch of the elephant's feet would suddenly be supporting proportionately more weight. An elephant, said Rodwell, is not scaleable.
In the same way, scaleability is important to transistor speed because simply changing the size of the transistor does not guarantee that it will run faster. In fact, some attempts to simply shrink current transistor designs have not yielded any speed increase.
If you change the structure of the transistor, it will get faster as it gets smaller, agreed Joseph Ballantyne, the new Lester B. Knight Director of the Cornell Nanofabrication Facility. This means, he said, that it would be possible for electronics using such transistors to process information at far greater speeds than is currently possible.
Ironically, the possible speed of the transistor proposed by Rodwell would be so high that there might be no market for it. The only practical application would be for research studies of molecules and materials. But as far as an everyday application, "we can't even generate data that quickly," said Ballantyne. For example, the information carried over fiber optic telephone lines is only a ten-thousandth of the bandwidth such transistors could handle.
And these chips would have no application in current personal technology, partly because of their voracious appetite for power.
| Cornell Chronicle Front Page | | Table of Contents | | Cornell News Service Home Page |