A typical personal computer will size up small bits of data, matching one against another. Usually, it does so in series, one bit at a time. Even parallel computers, which run many processors simultaneously, ultimately fall back on step-by-step analysis.
The ability to compare, say, a million pieces of information in one deft stroke would therefore offer tremendous computational advantages. Optical computers, which rely on transmissions of light rather than electricity to perform calculations, might provide such benefits. To date, though, no good material for making the machinery to handle optical comparisons has been found.
Seeking to supply this missing link, Randall J. Knize, a physicist at the University of Southern California in Los Angeles, and his colleagues describe fabricating an optical correlator, a device that uses a vapor of cesium atoms to compare images. They report their work in the Sept. 22 Nature.
"Optical computing aims to take advantage of the massive parallelism of light," says Knize. "If you expand a laser beam so that it can carry a million pieces of information, then process all of that information at the same time, you could make massively parallel computations in a way that isn't possible with ordinary electronic computers."
"This optical correlator offers a way to process information by comparing two images to see if they are similar," he says. Two laser beams carrying information about two images pass through a glass cell containing a cesium vapor. The cesium cell senses similarities and differences, then emits a third signal that reveals where the two images do and do not overlap.
"We chose cesium vapor because of its enormous sensitivity, which is much greater than that of most semiconductors," says Knize. "It's also very nonlinear optically, meaning that it emits much more light than it receives. And it reacts strongly with the particular wavelengths of light that our diode laser emits."
Knize stresses the importance of nonlinear optical materials in computing. "With linear optics, where you get out the same amount of light that you put in, you can't do very much that's useful. You need a material that will give out more than just the sum of the two laser beams going in. They must interact [to] create new information."
To test their new system, the researchers had the optical correlator compare minuscule images of the first three letters of the alphabet, each letter only 230 micrometers high and printed with lines only 35 micrometers thick. The comparisons took only 30 nanoseconds. "The combination of its low-power operation and its fast response makes cesium vapor a candidate for the most sensitive material demonstrated in a correlator to date," they conclude.
The scientists believe this sort of device could eventually improve computerized vision systems, which rely on quick and subtle pattern recognition. For instance, a security system might need rapid fingerprint comparisons. Or a pharmaceutical company might benefit from a quick inspection of pills on an assembly line.
Next, Knize and his colleagues plan to compare moving images, perhaps at 1,000 frames per second, fed to the correlator by a video system. This will allow them to explore image analysis "in real time," says Knize.
--R. Lipkin