Trapping the light fantastic

QUANTUM mechanics, the theory that explains how the universe's fundamental particles behave, thrives on reconciling opposites. According to its topsy-turvy tenets, it is perfectly all right for a particle to smear away into space like a ray of light, while a beam of light behaves like a collection of tiny, distinct particles. So it is fitting that an idea about quantum mechanics itself that once seemed far-fetched may soon appear just the reverse: it now looks possible that scientists may some day harness quantum principles to build a computer.

A quantum computer could solve problems that a conventional computer could never manage, such as factoring enormous numbers for code breakers or simulating the intricacies of complicated systems. This is because such computers would rely on quantum bits (qubits) to store and process data. Unlike conventional bits (the ones and zeros of the binary arithmetic that existing computers employ), qubits can be in two states at once. Because each qubit can simultaneously process a zero and a one, a quantum computer could zip through lots of calculations in parallel. A regular computer, in contrast, has to carry out each calculation in turn.

Manufacturing devices that can store qubits is therefore the most important part of making a quantum computer. It is also proving the most challenging. The best efforts to date have produced seven- or eight-qubit systemsÑa considerable accomplishment, but far from the number required for solving any interesting problems. Now, a new technique may bring some of those problems within reach. Physicists working at the University of Hanover's Institute for Quantum Optics think they have worked out a way to make qubits that would beat the previous record by a factor of ten.

Wolfgang Ertmer, Gerhard Birkl and their colleagues plan to use rubidium atoms as their storage devices. A lone rubidium atom can exist in one of two different low-energy quantum states, meaning that it can act as a qubit store. But holding enough lone rubidium atoms in place to have a chance of making a working computer is tricky. Going on to read them so that the values of their qubits can be known is doubly so.

As the researchers report in a forthcoming issue of Physical Review Letters , they managed to hold significant numbers of isolated rubidium atoms in place by shining a laser beam on to an array of microscopic lenses. Light passing through each microlens focuses at a point a few tenths of a millimetre beyond it, and thereafter spreads out as a cone. This conical beam is then reflected into a cluster of 1m or so rubidium atoms, which are held trapped in a magnetic field within a vacuum chamber.

One effect of the beam is to polarise the cloud of electrons surrounding each rubidium nucleus, meaning that part of the atom becomes a negatively charged pole, and part a positively charged pole. The result is that the atoms interact with the beam, are drawn into it, and settle at the tip of the coneÑa phenomenon known as dipole trapping.

Using this technique, and a sufficient number of microlenses, the team created an array of 80 dipole traps, each of which captured about 100 rubidium atoms, spaced 125 microns from one another. With sufficient cooling, Dr Birkl reckons, these traps could be made to hold on to just one atom each, thereby producing 80 individual qubits.

One advantage of this method of making a quantum computer is that the atomic qubit stores can be manipulated relatively easily. Using pulses of laser light aimed through the appropriate microlenses, the researchers were able to set each of the qubits into a particular initial stateÑa necessary precursor to any calculation. Another advantage is that when the atoms are left alone, they do not change state too quickly. That means they are likely to prove sufficiently sturdy to perform quantum computations.

Doing actual calculations with these trapped atoms will be hard, because the qubit stores must interact with each other during the process. The researchers plan to solve this problem by using two laser beams focused through the same array of microlenses. By placing the beams at a slight angle to each other, they are able to create an interleaved set of dipole traps. These can be made to overlap with each other by altering the angles of the beams. This means that atoms in different traps could be brought together, made to perform a computation and then separated. Then the result of the computation could be "read" by laser once the qubits were comfortably sequestered again.

At the moment, the researchers are at the beginning of this process. It will be a long time before they are putting any conventional computer makers out of business. But their demonstration of dipole trapping at least shines a captivating light down the path to quantum computing becoming a practical technology.

From The Economist print edition