Quantum leap: Company claims breakthrough, though skeptics remain
Could a small Canadian start-up usher us into the age of quantum computing?
- By William Jackson
- Feb 17, 2007
ZERO SUM: Refrigeration keeps the D-Wave quantum computer's circutry near absolute zero temperatures.
D-Wave Systems Inc. photo
Quantum computing has long held the promise of harnessing processing power orders of magnitude greater than what's possible with today's most powerful digital computers, but researchers have always considered that goal decades away.
'There is no reason it can't be done, but nobody sees a clear path to it,' said Carl J. Williams, coordinator of the Quantum Information Program at the National Institute of Standards and Technology.
Now, a Vancouver, B.C., start-up says it has found the path and is introducing what it calls the first quantum computer designed to run commercial applications.
D-Wave Systems Inc. was scheduled to unveil the 16-qubit system on Feb. 13 at the Computer History Museum in Mountain View, Calif.
'The initial model is not really fast, but it's a demonstration of the technology,' said CEO Herb Martin. 'We expect to have many more qubits by the end of 2007,' when the company plans to begin offering its platform to outside users as an online service. He said the company expects to begin selling the computers themselves by late 2008.Mechanics Illustrated
A qubit is the basic unit of information in quantum computing, processing and holding data. The more qubits, the more processing power.
Other researchers have been reluctant to comment directly on the D-Wave system before actually seeing it in action, but it is difficult for them to avoid some skepticism.
'If it is true, it is remarkable,' Williams said.
NIST has been a major player in quantum information research since the early 1990s. Commercial R&D centers, including IBM Corp.'s Watson Research Lab, also are pursuing the technology.
'Our research guys consider it a big thing for the future,' said Watson Lab staff member David DiVincenzo. 'It's pretty far from commercial reality at this point.'
How far from commercial reality? No one can say for sure, but researchers commonly throw out numbers such as 20 to 25 years. There is no reason a quantum computer can't be made, they say, but it is a massively complex engineering problem.
That was the attitude that let D-Wave steal a march on commercial development, Martin said.
'The general feeling was [that] it was going to take 20 years to do this,' he said. So no companies were putting big money into the effort. 'Most of the work is being done in universities or in labs that are government funded.'
"You could say we took a hybrid approach rather than a pure quantum approach." Herb Martin, D-Wave
D-Wave, which grew out of work done at the University of British Columbia, took a simplified approach. It uses a traditional digital computer to run applications and submit problems, and a quantum processor for acceleration.
'You could say we took a hybrid approach rather than a pure quantum approach,' Martin said.
The system uses an array of magnetic flux qubits, which are micron-sized loop circuits etched into superconducting niobium. Current can flow clockwise or counterclockwise on the loop, the two states representing a 1 or a 0, much like a bit in a traditional computer. The difference with the qubit is that because of the peculiarities of quantum mechanics, each qubit can represent both states at the same time under the right circumstances.
This increases the processing power exponentially. Even the most massively parallel digital computer has to perform each operation sequentially. A quantum computer can do truly parallel processing.
The superconducting flux qubit technology used by D-Wave is one of three promising approaches to quantum computing often put on the table by researchers, Williams said. Research at NIST has taken a different direction.
'The state of the art really is in ion traps,' he said.
The difficulty in making any system work is controlling the gateways between qubits so that you get accurate results. This is the work done by transistors and integrated circuits in digital technology. The state of a qubit is so delicate that errors tend to creep in, and these multiply as the number of qubits used increases.
If you can't control the noise from errors, you can't scale, Williams said. 'If you can't scale, you're dead in the water.'
The ion trap technology being investigated by NIST has been pushed close to 99 percent accuracy, Williams said. This is not good enough for practical computing, but is better than what has been accomplished by researchers working with superconducting flux qubits. 'None of them have been able to push the things past about 90 percent,' he said.
Some limited uses for quantum computations might be easier to achieve. Both the Air Force Office of Strategic Research and the Defense Advanced Research Projects Agency are pursuing quantum simulators.
'With 100 qubits, I could simulate something you could not possibly compute,' even with a relatively high error rate, Williams said. If the process fails, you can repeat it until you get a good answer. Even these machines are a decade away from practical applications, he said.D-Wave's offering
D-Wave claims its processor, called a thermally assisted adiabatic quantum computer, can perform adequately even in the presence of error-induced noise. It uses programmable gateways that can be tuned to favor certain relationships between qubit states.
'The D-Wave processor's natural physical evolution drives it into the solution of this difficult problem, similarly to the way a ball rolling down a hill is driven into the lowest point,' explains a paper prepared by the company. 'If the physical system made up of the various devices on the D-Wave processor can reach its minimum energy state, then simply reading out the states of the memory elements provides the exact solution to the problem.'
D-Wave's is not the first quantum computer built. Demonstrations have been made with smaller models. The difference is that D-Wave has a working application to run on its computer, one designed to do highly constrained database searches that are impractical on a traditional computer.
The company plans to drive the market for its computer by giving developers free access to it so they can write their own programs and conduct research.
Because the programs actually run on a digital computer and feed data to the quantum processor, it is not difficult to port an application to the new computer, Martin said.
The computer as it now stands fills about 7 feet of a standard 19-inch rack and will always be a server room machine, Martin said.
'We might shrink it down to 5 feet, but it's not going to be much smaller than that,' he said. 'It's not going to be a desktop machine.'
One thing Martin and other researchers agree on is that there will always be a market for digital computers.
'Quantum computers are probably not a replacement ever for digital computers,' said IBM's DiVincenzo. 'It won't be a replacement for a desktop or a laptop, and it doesn't improve anything that you do on a traditional computer.'