The race is on to build a universal quantum computer and IBM claims it has leapfrogged the rest of the field with a new breakthrough that paves the way to large-scale systems that can operate reliably.
IBM researchers have developed error-correction techniques that could maintain the integrity of computations performed using qubits, or quantum bits -- the basis of quantum computing. As with conventional computing, isolating and resolving data errors is a key step to building a fully functional quantum computer, said Jay Gambetta, a manager of IBM's quantum computing and information group.
Quantum computing is viewed as a way to advance computing beyond today's PCs and servers. A company called D-Wave Systems has already made a purpose-built quantum computer that is useful for specific tasks. IBM wants to build a larger, "universal" quantum computer that can run a wide range of applications, much like a PC or server.
The complexity of quantum computing systems and the fragility of the interaction among qubits makes error correction important, but also a challenge. Right now, research is focused on limited sets of qubits, but IBM wants to bring error correction to larger systems, Gambetta said.
"Now we're getting to the point where we're putting together and doing practical computing. It will be very exciting over the next few years," Gambetta said.
Quantum computing is one of the ways to advance computing once it becomes physically and economically impractical to build smaller, more powerful chips based on silicon and conventional techniques. Quantum computers use components that are different than those used by conventional computers.
Conventional computers are predictable by nature, using electrical transistors to represent data as ones and zeros. Qubits, on the other hand, use the properties of subatomic particles, harnessing the laws of quantum mechanics to achieve various states. Unlike a conventional bit, which can hold only a one or a zero, a qubit can also hold a one and a zero simultaneously. This technique, called superposition, allows quantum computers to perform multiple calculations in parallel, vastly increasing their processing power relative to conventional computers.
A universal quantum computer might need 100 million qubits, Gambetta estimated. D-Wave's latest quantum computer, the D-Wave Two, built for specific tasks, is a 512-qubit system.
However, it's hard to predict the behavior or state of qubits once they start interacting, or "entangling," as they process data. Under quantum mechanics theory, which examines the interaction and behavior of matter on atomic and subatomic levels, it cannot be predicted whether a qubit will be a one or zero for a specific calculation. In addition, the state of qubits can be upset by something as simple as electromagnetic radiation or matter, which could ultimately disrupt computational cycles and cause data errors.
IBM is designing a quantum computer much like a building, in which the building blocks in this case, qubit arrays -- are structured in a horizontal and vertical format. IBM is now able to detect multiple types of data errors simultaneously in a square array of four qubits, which was not previously possible. In this case, IBM was able to detect bit-flip errors, which involve a switch from a zero state to a one state or vice versa -- which also applies to conventional chips -- and also phase-flip errors, which could be applied to superpositioned qubits.
The error correction was tested on a four-qubit circuit in a lattice structure, which IBM built for the first time. With the ability to detect both bit-flip and phase-flip errors simultaneously, it will now be possible to connect such lattice structures to build a larger computer, which is important when scaling performance.
"Now we need to take that square and show that we can detect [errors] and correct on a larger scale," Gambetta said.
Gambetta declined to predict when a universal quantum computer would come out, saying it's a long journey that many researchers are working toward. IBM last year said it would invest US$3 billion over five years into research that could lead to quantum computers and cognitive systems.
But don't expect a quantum computer on your desk anytime soon. Quantum computers will be likely take on tasks related to large-scale simulations, such as those used for drug design and environmental matters, which could otherwise take years on today's supercomputers. Checking e-mail and browsing the Web will continue on conventional computers even after quantum computers are introduced, Gambetta said.
A paper on the research will appear in the April 29 issue of Nature Communications. The research was partly funded by the US government's Intelligence Advanced Research Projects Activity. IARPA also funds research to develop a new superconductor semiconductor, which is an important component for quantum computers.