The chip giant has managed to create prototype interconnects--microscopic metallic wires inside of chips that link transistors--out of carbon nanotubes and measure how well the interconnects perform. In essence, the experiments are a way to test whether the theories about the properties of carbon nanotubes are accurate.
Mike Mayberry, director of components research at Intel's labs in Oregon, will discuss the research at the International Symposium for the American Vacuum Society next week in San Francisco. Intel worked with California Institute of Technology, Columbia University, University of Illinois at Urbana-Champaign, and Portland State University on the project.
Chip interconnects have become a looming headache for chipmakers. Under Moore's Law, chipmakers shrink the components inside semiconductors every two years. Shrinking interconnects, however, increases electrical resistance, which in turn reduces performance. Chipmakers switched from aluminum to copper interconnects in theto get around the problem. Unfortunately for Intel and other companies, the resistance will start to become a significant problem in smaller copper interconnects in the coming years.
"With metals, as you reduce the diameter of the interconnect, the resistance can go way up," said Dave Lammers, a director at VLSI Research, a semiconductor analysis firm. "The electrons carom off the metal atoms. That is going to slow things down."
Lammers first wrote about the experimental interconnects in The Chip Insider, VLSI's newsletter.
Carbon nanotubes, the, conduct electricity far better than metals. In fact, nanotubes exhibit what's called , which means that electrons are not scattered or impeded by obstacles.
Nanotubes, which measure only a few billionths of a meter thick, are also far thinner than metal interconnects can be made. Potentially, this eliminates the problem with shrinking interconnects. IBM and others have made transistors out of carbon nanotubes.
In its experiment, Intel aligned bundles of nanotubes by means of an electric field and then measured their frequency with fairly standard equipment.
Devil in the details
There is, of course, a catch. Although they exhibit unusual and beneficial properties, carbon nanotubes are difficult to mass manufacture. Some nanotubes are semiconductors, meaning the transmission of electrons can be controlled, while others are pure conductors, depending on the arrangement of the atoms. Some are long; others are short. Nanotubes produced in the same batch will contain a dizzying array of characteristics.
Since each chip would require thousands of nanotubes for interconnects, researchers are going to have to figure out a way to produce uniform ones, or quickly separate the good ones from the chaff.
"With (contemporary) interconnects, you dig a trench and fill it up with metal," Lammers said.
As a result, carbon nanotube interconnects won't likely appear in a commercial chip for several years at best.
Whether carbon nanotubes make it into chips or not, the basic structures and materials inside semiconductors will change radically in the next. Around 2010 or 2012, researchers will begin to narrow down what changes will have to occur and then chips that combine silicon elements with newer nano elements will likely begin to creep in toward the middle of that decade. In the 2020s, the ability to shrink silicon chips and necessitate a shift to very different materials.