5 December 2007—Scientists at Sandia National Laboratories, in Albuquerque, have begun a three-year effort that may yield the world’s first silicon quantum bit—the key component in future quantum computers. Making a ”qubit” out of this ubiquitous workhorse semiconductor could pave the road to a practical quantum computer by allowing its construction using existing technology and its integration with ordinary computers.
A lot of scientists are understandably excited about quantum computing. Certain problems are intractable for today’s computers. Factoring a 300-digit number, for example, might take your laptop several decades. But a quantum computer could puzzle out the same problem in hours or days, says Malcolm Carroll, the principal investigator on the silicon qubit project.
That’s because a quantum computer encodes information in a way that is totally different from how your laptop does it. In a conventional computer, the transistor works like a switch—it can be either on or off. The bit encoded by the transistor’s state is mutually exclusive—1 or 0, yes or no, true or false. But a quantum computer works differently. A quantum dot, the analog to a transistor in some quantum computer schemes, traps electrons and measures a quantum property called ”spin.” The direction of that spin can be up, down, or both at once, a state called ”superposition.” Those measurements are analogous to 1, 0, and the superposition of 1 and 0.
Qubits are key to a quantum computer’s theoretical ability to radically outperform its conventional counterpart. Though up, down, and both seem to be three options, they’re not. ”Think of it as percentages,” says Carroll. ”Ten percent up and 90 percent down. Or 20 percent up and 80 percent down, or 50 percent up and 50 percent down. Any combination is possible at any percent. So it’s not three options but nearly infinite possibilities.”
Spin is one of the most promising ways to read and encode information in a quantum computer. But measuring it is a delicate undertaking, because it can be easily disrupted. A qubit’s spin is not eternal; it will inevitably change direction, a process called ”decoherence.” It’s just a matter of how soon that happens. ”Working with qubits is a race against the clock,” says Mark Eriksson, a professor of physics at the University of WisconsinMadison, who is collaborating on the project. ”It’s just a question of how fast things go wrong versus how quickly you can do what you need to do.” Researchers are always looking for ways to stave off decoherence in order to draw out the period when information can be encoded and operations performed.
The fleeting nature of a spin-based qubit is compounded by another problem: how many quantum dots you can make. ”With one or two qubits at a time, you can’t do very much computing-wise,” says Matthew Pelton, a researcher at Argonne National Laboratory, in Illinois. A functioning quantum computer chip needs hundreds or thousands of them. For that kind of integration, he says, ”you’re going to have to build a quantum dot in a semiconductor material.” For years, researchers toiled to do just that. Then, in 2005, the first quantum dotbased qubit was created in gallium arsenide, an exotic semiconductor.
But there’s a fundamental problem with qubits in gallium arsenide: the spin of the material’s atoms’ nuclei affects the state of the qubit’s spin. These nuclear spins ”infect” the spin of the qubit, causing it to change direction and likely spoiling any computation it was involved in.
Carroll thinks the best material for the job is silicon. But though quantum dots have been demonstrated in silicon over the past two years, Eriksson says, ”no one has ever measured spin in a silicon quantum dot.” Lieven Vandersypen is a quantum-computing expert at Delft University of Technology, in the Netherlands, whose latest research, published last week in Science magazine, demonstrates the successful manipulation of electron qubits in gallium arsenide. He says the main obstacle to silicon spin qubits is that the quality of the dots ”isn’t yet good enough.” Carroll is out to change that.
Instead of using the silicon found in a standard microchip, which is made of a mash of different silicon isotopes, Carroll’s group will use isotopically pure silicon-28, which, because of its perfectly balanced 14 neutrons and 14 protons, has a nuclear spin of zero. They are banking on that being an ally in this race against the clock. No interference with the qubit’s spin buys researchers precious time to perform operations before the inevitable decoherence; the information encoded in the qubits remains legible for longer.
Sandia is banking on the fact that creating a silicon qubit will give the nascent technology an enormous boost. ”It makes a huge silicon circuit and fabrication toolbox available that just wasn’t there before,” says Carroll.
To Probe Further
For Lieven Vandersypen’s adventures in building a quantum computer, see ”Dot-to-Dot Design” in the September 2007 issue of IEEE Spectrum.