NIST nudges quantum computers toward commercial viability

PORTLAND, Ore. — Quantum computers can sift through all the possible values of multivariable problems to find the optimal solution in a single machine cycle. The downside is that the values with which they work—qubits, which use superposition to represent both a 1 and a 0 simultaneously—are extremely prone to errors in multistep calculations.

But an ounce of prevention is worth a pound of cure, according to the National Institute of Standards and Technology (NIST), which has demonstrated an error suppression technique that it says makes multistep quantum calculations commercially feasible.

"Quantum computers are inherently prone to errors caused by stray electric or magnetic fields, but theorists have shown that if their error rate is low enough—about one in 10,000—fault-tolerant error correction schemes could enable quantum computations of virtually any length," said John Bollinger, NIST's lead scientist on the project. "We have taken an idea that other people have developed and have experimentally demonstrated that we can use it to suppress the errors in quantum computers enough to meet that one-in-10,000 requirement."

NIST's dynamic decoupling technique, called spin-echo error suppression, uses the echo from error-suppression pulses to realign the "spins" inside the qubits that simultaneously encode the superimposed ones and zeroes. By regularly applying the error-suppressing pulses, spins that are precessing into error are realigned before they get out of phase enough to cause a hard error.

The idea of spin-echo error suppression dates back to 1950, when U.S. physicist Erwin Hahn (currently a professor emeritus at the University of California, Berkeley) discovered that precessing nuclear spin magnetization could be restored to its correct orientation with a 180-degree phased pulse of resonant radio-frequency energy. The technique works because the direction of the precessing spins, which tend slowly to get out of phase with each other, are reversed by the 180-degree pulse, thus rephasing into their original alignment. The technique is called spin echo because there is a lag between the application of the error suppression pulse and the return of its echo with all of the errant spins realigned.

"We are the first group to demonstrate spin-echo refinements that can suppress qubit errors below the one-in-10,000 level," said Bollinger. "We have learned how to adjust the spacing of these pulses to optimize error suppression."

NIST's spin-echo technique requires no prior knowledge regarding the characteristics of the noise to be suppressed. Instead, it uses feedback on how many errors are being made to change the spacing of the spin-echo pulses for optimal error suppression.

NIST applied its error suppressing pulses using a microwave photon at 124 GHz. The qubits were stored in an array of 1,000 supercooled beryllium ions, each representing a single qubit trapped by an electric and magnetic field. Solid-state versions of such arrays could act as the memory banks in future quantum computers. NIST's demonstration showed that the echoes from appropriately timed error suppression pulses can keep the spins of all 1,000 qubits sufficiently aligned to prevent excessive errors.

"Error correction is vitally important for a gate-level approach to quantum computers; they won't work at all without it," said Geordie Rose, founder and CEO of D-Wave Systems Inc. (Vancouver, B.C.), a pioneering quantum computer company.

Because of the error-prone nature of general-purpose, gate-level quantum computer architectures, D-Wave instead chose an adiabatic model, which sacrifices generality for robustness. D-Wave's quantum computer can only solve discrete optimization of multivariable problems. If NIST's spin-echo error suppression method works as advertised, however, D-Wave claims it may switch back to the general-purpose, gate-level approach.

"Our business model takes an agnostic approach to quantum computing technologies," said Rose. "We are very much interested in any technique that advances that work, and NIST has certainly proven itself to be one of the leaders."