Computing's Next Wave: The Light Fantastic?

Researchers are experimenting with superfast data searches that rely on refractions of colored light instead of digital code

Today's computers work somewhat like pinball machines. Fire an electronic ball at a specific location, and it throws a switch that releases bells and whistles of encoded data. The smaller and faster the circuits on a computer chip, the thinner the wires become. Eventually, the quickest chips should be able to reduce the size of the balls to a single electron. But sooner or later, experts believe the ability to build increasingly powerful computers by etching circuits on silicon chips will reach its limits.

So some researchers are trying to change the nature of the game. One new idea, recently demonstrated by a team of researchers from the University of Rochester, would relegate wires and electrons to history. Instead, the data would ride on waves of colored light. Theoretically, this system of computing could conduct huge searches in one fell swoop, rather than making millions of queries.

Although the effort is still highly experimental, if the method lives up to its promise, "We may be able do things that present semiconductor-based computers can't do -- nor will ever be able to," says Ian Walmsley, professor of optics at the University of Rochester, who led the team that invented the device.


  At a recent conference on lasers and optics in Baltimore, Walmsley and his colleagues reported that they had invented a way to use light to do a database search of 50 items in a way that can't be duplicated in any particle-based computer. Rather than relying on a digital system that uses strings of 1s and 0s to encode data, the Rochester machine is analog. It works on a simple principle discovered in the 19th century: When different waves of sound or light combine, they create unique patterns, called interference.

While the properties of waves may be old hat, Walmsley says the technology needed to build a light-wave computer, such as ultrashort-pulse lasers, has only been developed over the past decade. In the Rochester device, information is stored in a transparent crystal made of a material called tellurium dioxide that vibrates in response to sound waves. The sound waves compress some parts of the crystal and expand others, creating a pattern of wave frequencies that encode the data.

To search the database, laser light is split into two identical beams and directed at the crystal. One beam is split into a rainbow of colors by a prism. The crystal bends the various colors of light as they pass through it. Then the two beams are recombined into a single beam. The resulting interference pattern can be read to identify the wavelength that matches the data being sought.

In the Rochester experiment, 50 different frequencies of light shine through the vibrating crystal. If the 20th frequency is altered, then the information sought is located at position 20 in the database. "We are leveraging new physics on the back of optical technology," says Walmsley.


  Most intriguing about the Rochester device is that it can search an entire database at once. A conventional computer would have had to query each site in that database, one at a time. And the larger the database, the greater the advantage. For instance, pinpointing a number in the Manhattan phone directory requires millions of searches, but with a wave computer, just one would be required.

Still, it's a long way from 50 to all the world's telephone numbers. Can wave transactions be scaled up? Walmsley believes they can. Although he says his group "chose a toy problem," he points out that a "single pulse from a laser contains a huge number of wavelengths of light all precisely phased to one another." With more compact methods of encoding data, he believes, "We should be able to identify one element in at least 2 to the power 50 elements in a single run." That's an enormous number -- 1 followed by 15 zeros.

This is a complex system. And it's unlikely that computers based on this technology will be on our desktops anytime soon. But the technology someday could have real-world applications in searching rapidly through huge amounts of data, researchers believe. With the increasing use of optics to connect computers on networks, "we can draw on modern optical communications technology to build such computers," says Walmsley. "This provides a massive and expanding technology base."

Someday, computers may truly be riding the wave.

By Alan Hall in New York

Edited by Douglas Harbrecht