May/June 2000

Quantum Computing

Continued from page 4

By M. Mitchell Waldrop

Making "Magic"

None of this means that we'll be trading in our PCs for quantum laptops anytime soon. Quantum computing is barely into its proof-of-principle stage, with a long way to go before it evolves even to the qubit equivalent of World War II-era vacuum-tube machines like the ENIAC. Much more likely are quantum add-ons to conventional machines-"coprocessors" that will perform specific tasks in the same way that a graphics card takes over the most difficult display chores.

What exactly those tasks will be, however, is an unsettled question. MIT physicist Edward Farhi points out that the study of quantum algorithms is still in its infancy. The two best ones discovered so far-Shor's and Grover's-will doubtless be followed by many more, opening up applications that go well beyond factoring and searching. Nonetheless, says Farhi, "a quantum computer isn't necessarily fast. It's a device that attacks problems in a different way. We're still trying to understand what makes a problem amenable to that kind of attack. You have to choose your problem carefully to take advantage of the quantum magic."

In practice, adds Farhi, how quantum coprocessors are used could depend a lot on how much they cost. And that's a complicated issue. At IBM Almaden, for example, the core of Chuang's quantum computer is small and inexpensive: qubit-containing molecules dissolved in a few drops of colorless solvent, encased in a glass tube smaller than his little finger. But the NMR spectrometer that makes the computer go is a silvery, 10-foot-tall cylinder surrounded by great thickets of wires and plumbing-most of it required to service the liquid helium that chills the spectrometer's superconducting magnets. If future quantum coprocessors follow this pattern, they will be huge, multimillion-dollar behemoths that fill up whole rooms, and that only governments can afford. In that case, quantum computers may well be restricted to hard-core national security tasks such as cryptography and intelligence-gathering.

But such an outsized contraption may not be inevitable. Gershenfeld's group at the MIT Media Lab is working on a compact, room-temperature NMR computer. They hope this device will be a prototype of a quantum coprocessor that will power inexpensive little gadgets-peripherals that will sit on the desktop like a modern-day printer or scanner. If that proves to be the pattern, then we could see a new generation of quantum hackers going to work in much the same way their forebears did at the beginning of the personal computer revolution, creating a profusion of innovative quantum software.

In the meantime, however, Chuang and his fellow experimenters are far less concerned with their machines' physical size than with their qubit count. "The first year we had lots of wonderful one-qubit machines popping up all over the place," he says. "Now at IBM we have a three-qubit computer, and we're planning even larger computers." In March, Los Alamos National Laboratory announced a seven-qubit NMR computer, and Chuang is confident that one lab or another will soon be demonstrating molecules with as many as 10 qubits. He concedes that this will be a tricky task, however. "Say I want a molecule with certain properties. When I draw it and go to the chemists, they just laugh: 'That's not real!' They call it a 'physicist's molecule.'" One complication is that carbon, the key ingredient of all complex molecules, almost invariably occurs as the isotope carbon-12-which is spinless. Chuang's chloroform computer worked only because its molecules were made with the rare and expensive isotope carbon-13, which does spin.