Supercomputing plays an essential role, however, in scanning the sky to find and catalog unknown pulsars. A radio telescope pointed at one point of sky receives (and digitally records on magnetic tape) all the noise it "sees" in that direction. "Somewhere buried in that noise," says Taylor, "we're looking for a fixed pattern of signals that involves a periodic repetition of blips with a certain dispersion constant."
It's essentially a multidimensional pattern-recognition problem, says Taylor. "We don't know the direction in the sky to look. We don't know the pulsar period or the dispersion constant. Even once you've chosen to look in a particular direction, you still have to try on the order of a million different periods and a few hundred dispersion measures. You're trying to create in software a set of optimally matched filters that will sort through the noisy data and tell you if there might be something there."
Probably several thousand millisecond pulsars lurk undetected in the galaxy, but these superfast spinning stars accentuate the search problem. The incoming radio signals must be sifted and analyzed in smaller chunks of time, and the size of the dataset grows exponentially as you shorten the pulsar period you want to find. "To find a pulsar with a period of a few milliseconds instead of half a second or a second," says Taylor, "is four to six orders of magnitude more difficult."
Over 25 years of pulsar searching, Taylor and his collaborators have discovered about half of the more than 600 pulsars presently known, and most of the rest have been found by others using methods pioneered by the Princeton group.
go back to the main screen