Researchers at Google Quantum AI have managed a new milestone in quantum computing, with chips that can correct errors below a “critical threshold”.

The team has published a description of the error-correcting chips in Nature.

“It’s fair to say that this is the most significant demonstration to date of quantum error correction,” says Professor Stephen Bartlett, a quantum physicist and director of the University of Sydney’s Nano Institute, who was not involved with the research.

Quantum computers promise to be an order of magnitude more powerful than classical computers at solving certain problems.

Where classical computers use bits of information to compute, quantum computers use qubits, which can hold and process far more information.

But existing quantum computers are, for now, proof-of-concept: they’re not yet usable for any practical applications. That’s because they’re extremely vulnerable to noise – even tiny fluctuations in the quantum computer’s environment can cause huge numbers of errors in the qubits.

Researchers have been working on ways to manage these errors, and a particularly promising method is “quantum error correction”: spreading information over lots of qubits to identify and compensate for errors.

But adding more qubits also adds more errors. Researchers have been trying to build a quantum computer that can correct more errors than it introduces – this is called the “surface code threshold”.

“Many research teams and companies have been working towards a demonstration of this type for a long time,” Bartlett tells Cosmos.

“It’s all about showing that quantum error correction, something that is key for quantum computing, can actually work in practice.”

Google Quantum AI’s paper represents the first quantum computer to pass this threshold.

The team used their newest generation of quantum chip architecture, called Willow, to build a 72-qubit processor and a 105-qubit processor.

They could run code on the quantum computer for 1 million cycles, over several hours, while error-correcting in real time.

They were able to lower the number of errors while making the computer more complex, which demonstrates that they’ve passed the threshold.

“There have been a number of previous results which have demonstrated aspects of quantum error correction,” says Bartlett.

“However this is the first demonstration of the essential ingredient, showing that if you increase the size of the quantum code, things actually get better.”

“Our results present device performance that, if scaled, could realise the operational requirements of large scale fault-tolerant quantum algorithms,” write the authors in their paper.

Want to understand how quantum science is changing our lives? Issue 105 of Cosmos Magazine dives deep into all things quantum. Subscribe now to get your copy.