By Peter Weiss
Physicists have long known that quantum computers have the potential to race through calculations trillions of times as fast as ordinary computers do. Now, it seems that those machines may not have to calculate at all to deliver answers.
That seemingly absurd possibility, which was advanced as a theory several years ago, has now received experimental verification. What’s more, although previous calculations indicated that such an approach would work only half the time at best, the new study suggests that it could become completely reliable.
Onur Hosten and his colleagues at the University of Illinois at Urbana–Champaign present their findings in the Feb. 23 Nature.
“This is a beautiful experiment. It verifies … one of the strangest aspects of the nature of physical reality that is presented to us by quantum theory,” comments theorist Richard Jozsa of the University of Bristol in England, who dreamed up the scenario in 1998.
Built so far only in laboratories and on a limited scale, quantum computers exploit the quantum-mechanical properties of tiny objects, such as photons and ions, to perform calculations (SN: 1/7/06, p. 5: Available to subscribers at Quantum Chip: Device handles ions as if they were data). Such properties include being in a so-called superposition, where an entity simultaneously exists in two or more states that seem mutually exclusive.
For the new experiment, the Illinois team, led by Paul G. Kwiat, built a rudimentary quantum computer from optical components such as mirrors and beam splitters. The researchers first mark one of four locations in a miniature database. When triggered by an incoming red photon with certain traits, the computer searches for the marked location and checks for a match between the location indicated by the photon and that target (SN: 6/3/00, p. 356: Available to subscribers at Quantum quirks quicken thorny searches).
When there’s a match, the computer emits a red photon with specific traits. If there’s no match, the outgoing photon has different characteristics.
The team incorporated the computer into a larger setup that included a beam splitter upstream to provide a path around the computer. Given its quantum nature, a trigger photon simultaneously enters and doesn’t enter the computer. “This puts the quantum computer in a superposition of running and not running,” Hosten explains.
Downstream photodetectors then record light signals in the various paths, which indicate whether the photon went into the computer and what its target location was. When such measurements are taken, however, the computer can no longer maintain its multiple states and the superposition collapses, leaving evidence that the computer ran or didn’t run.
Indeed, the detectors indicated about a third of the time that, with no photon going into the computer, and thus no search, the computer had yielded the correct answer to the question: Was there a mismatch between the incoming photon and the chosen database location?
Kwiat’s team also presents new theoretical calculations showing a way to boost the computer’s accuracy to nearly 100 percent and to specifically identify the selected location rather than determining whether there was a mismatch.
Charles H. Bennett of the IBM Thomas J. Watson Research Center in Yorktown Heights, N.Y., praises the new work for “exploring the places where quantum prediction seems most at odds with common sense.”
