Ben Mercier/D-Wave Systems; (inset) Paul Bunyk/D-Wave Systems.

How’s it work? Debate continues as to whether D-Wave Systems’s computer can do quantum computations or merely classical ones. The inset shows circuit-level details of the device.

Controversial Computer Is at Least a Little Quantum Mechanical

Staff Writer

If the experiment was meant to silence the critics, it didn’t. Four years ago, an upstart tech company created a stir when it claimed to have built a quantum computer—a thing that, in principle, could solve problems ordinary computers can’t. Physicists from D-Wave Systems in Burnaby, Canada, even put on a demonstration. But other researchers questioned whether there was anything quantum mechanical going on inside the device. Now, the D-Wave team has published data that they say prove quantum phenomena are at work within its chip. But even if that’s true, others still doubt that, as D-Wave researchers claim, the chip can do quantum-mechanical computations.

“I think they’re overstating this,” says John Martinis, a physicist at the University of California, Santa Barbara (UCSB). “It’s not obvious that they’ve implemented a quantum algorithm.”

Physicists have been trying to develop quantum computers for more than a decade. An ordinary computer deals with bits that encode a 0 or a 1. As first conceived, a quantum computer would use subatomic particles or other quantum objects as “qubits” that could encoded 0, 1, or, thanks to the weird rules of quantum mechanics, both 0 and 1 at the same time. What's more, a string of qubits in that strange state could encode every possible combination of 1 and 0 values at the same time. As a result, a quantum computer could process myriad inputs at once and crack problems that would overwhelm a conventional computer. However, that approach to quantum computing, called the “gate model,” presents many unresolved practical problems, as scientists must maintain and manipulate the delicate quantum state of many qubits.

D-Wave researchers have taken a different tack, known as “adiabatic quantum computing” or “quantum annealing.” They begin with a set of noninteracting qubits—in their rig, little rings of superconductor that can carry current either one way or the other or both ways at once—and put the rings in their lowest energy “ground state." To perform the computation, the researchers slowly turn on various interactions among the qubits. If they’ve done things right, then the ground state of the noninteracting system should naturally evolve into the ground state of the interacting system and reveal the answer to the problem encoded in the interactions.

In February 2007, D-Wave created a splash when its 16-qubit machine solved several puzzles—although none that a conventional computer couldn’t handle—such as figuring out how to seat guests around a table so that people who dislike each other do not end up side by side. However, “people had serious doubts that this was a true quantum computer,” says Wim van Dam, a theoretical computer scientist at UCSB.

Here’s why: The workings of the machine can be thought of as tracing the trajectory of a marble through a changing energy landscape of peaks and valleys as it finds its way to the lowest point—the solution to the problem. A process called quantum tunneling lets the marble burrow from one valley to another. At the same time, however, plain old “thermal fluctuations” also agitate the hypothetical marble and can push it over the ridges in the landscape so that it reaches the lowest valley. That process is not quantum mechanical, van Dam says, so if that’s how the D-Wave computer works, then it cannot be significantly more efficient than an ordinary computer.

However, new data show that in fact the qubits in the chip can find their lowest energy state quantum mechanically, D-Wave researchers report this week in Nature. Physicist Mark Johnson and colleagues begin experimenting with a single qubit within their latest 128-qubit chip. Current in the ring can circulate either clockwise or counterclockwise, and those two states represent two dips in a very simple energy landscape. By tuning the qubit and applying a magnetic field, the researchers can raise the height of the ridge between those two states and also tilt the entire landscape to make one dip lower than the other. They can also change the temperature—the source of the pesky thermal fluctuations.

The researchers found that the ability of the qubit to get from the higher energy state to the lower one at first decreases as the temperature falls. But below about 45 thousandths of a degree above absolute zero (45 millikelvin), the rate at which the qubit makes the switch levels off. That suggests that even as thermal fluctuations grow too weak to nudge the system over the energy barrier, quantum tunneling remains to allow the qubit through it. The researchers observe a similar phenomenon as a chain of eight qubits with very simple interactions finds its way to its predicted ground state. “The evolution [of the system] is consistent with quantum mechanics and not with classical mechanics,” Johnson says.

Martinis has some quibbles. Still, he says, “I think it’s pretty likely that they’ve got tunneling. I’m not 100% sure, but I’m 90% sure.”

The results won’t end the controversy over D-Wave's technology, however. Quantum tunneling alone is not enough to make the device significantly faster than a classical computer, van Dam says. To whack through really big computations that would take an infinite amount of time on a classical computer, he says, D-Wave’s chip also has to maintain a kind of delicate synchrony between the individual qubits called coherence. But it’s possible that D-Wave’s qubits lose coherence very quickly to act more or less independently but nonetheless tunnel to their collective ground state. And in that case, the computer can’t hope to be any more efficient than a regular one, van Dam says.

Johnson and the D-Wave team are not convinced that coherence is necessary in adiabatic quantum computing. “I think it’s not entirely understood what role coherence plays in quantum annealing,” Johnson says. Martinis says it’s unusual to see a company essentially wager its future on a point of scientific dispute. “In some ways, I kind of respect that it’s a clear corporate strategy,” he says. “On the other hand, I’m not going to invest in their technology because I think they’re wrong.”

Stay tuned. Johnson says the D-Wave team members will have more publications to back up their claim that they really have a quantum computer.

Posted in Math, Physics