Despite the progress made in making quantum computers practical, qubit-based systems remain unstable and highly error-prone, and Google may have taken a big step towards fixing it.

With the recently introduced quantum chip called Willow, Google engineers have passed a major milestone in error handling. In particular, they managed to keep a single logical qubit stable enough that failures occurred perhaps once an hour; This is a significant improvement over previous installations that failed every few seconds.

Qubits are the basic building blocks of quantum information. Unlike classical computing bits, which can store 1 or 0, these qubits can store 1, 0, or a superposition of both. This combination is a powerful tool in developing algorithms that can solve problems that would take a classical computer too long to solve.

Unfortunately, qubits are delicate things, their superpositions tend to bond with the environment and lose their mathematical properties. Although today’s systems are reliable enough to provide 99.9 percent reliability, the error rate in practical systems should be close to one in a trillion.

To eliminate errors in these fragile qubits, researchers can distribute a single logical qubit among multiple particles in superposition. But this scaling only works if the additional physical qubits correct errors significantly faster than they create them.

“Willow is the first processor where error-correcting qubits get exponentially better as they grow,” wrote Michael Newman and Kevin Satzinger, researchers on Google’s Quantum AI team.

“Each time we increase our coded qubits from 3×3 to a 5×5 or 7×7 physical qubit lattice, the coded error rate drops by a factor of two.”

Willow has 105 physical qubits, and the combination of its architecture and the error correction algorithms it uses has led to success from a stability perspective, where more qubits mean fewer errors.

This has been a problem since quantum error correction techniques were first introduced in the mid-1990s. Although quantum computing still has a long way to go to be fully realized, large-scale quantum operations are at least possible with this approach.

“This demonstrates the exponential error suppression promised by quantum error correction, an almost 30-year goal for quantum computing and a key element in unlocking large-scale quantum applications,” Newman and Satzinger write.

Stability isn’t Willow’s only advantage: Google claims that one of our fastest supercomputers can complete a particular quantum task in five minutes that would take 10 septillion years (this task was created specifically for quantum computers, but it still shows what’s possible).

Errors will always be present in quantum systems, but what researchers aim to do is make quantum processing rare enough to be practical. This will require better hardware, more qubits and updated algorithms.

“Quantum error correction appears to be working now,” Newman and Satzinger write, “but there is a big difference between the one-in-a-thousand error rate today and the one-in-a-trillion error rate that will be needed tomorrow.”

An earlier unedited version of the work Nature.