EQUS researchers’ low-overhead surface-code approaches to implementing logic gates with qubits protected by error-correcting codes, implemented in Google’s quantum processor in 2024.
Functional quantum computers will require fault-tolerant architectures based on quantum error-correction codes to mitigate the high levels of noise present in quantum hardware.
The idea is to arrange qubits in such a way that if some break or go wrong while performing a computation, they can be corrected using an appropriate code, with the
most common correction architecture being a 2D-mapped family of codes known as a surface code.
A surface code uses adjacent qubits in a 2D grid to detect and fix errors.
But conventional surface codes come with a high ‘overhead’ – meaning a dauntingly large number of physical qubits and gates would be needed to achieve the required performance.
In 2021 an EQUS team at the University of Sydney made a small tweak to the conventional surface code – a local change of basis, or a rotation of every second qubit.
“By flipping half of the qubits in our design, we found we could effectively double our ability to suppress errors,” says Pablo Bonilla, who was a 21-year-old undergraduate student at the outset of the work.
They showed that architectures based on the new code, which they named the XZZX surface code, have much higher memory thresholds and lower overheads than those based on conventional surface codes.
The XZZX code exploits the common structures in the noise affecting physical qubits. It can be tailored to fix the types of errors we know are most likely to occur in any particular physical qubit-hosting platform, whether that be silicon or trapped ions or superconducting plasmons.
In other words, XZZX can “correct very efficiently for the dominant source of errors, and spend fewer resources on correcting very rare errors,” in the words of team supervisor Prof Stephen Bartlett.
The far more scalable surface code can also continue to counter quantum decoherence as more qubits and gates are added to the system.
“What’s great about this design is that we can effectively retrofit it to the surface codes being developed across the industry,” Stephen Bartlett said.
“We are optimistic that this work will help the industry build better experimental devices.”
Indeed the potential of the new code was immediately apparent, supported by excellent publicity efforts by the team behind the code. Pablo Bonilla and supervisor Dr Ben Brown’s communication efforts were recognised by the EQUS award for Best Contribution to Public Debate in 2021, featuring in an ABC News story, with the associated YouTube video viewed over 100,000 times.
Amazon Web Services soon announced plans to implement XZZX in its quantum computing labs.
And in 2024 quantum computing giant Google implemented a variation on the XZZX code to reduce a very common type of noise in their 105-qubit superconducting quantum computing processor ‘Willow’, improving noise reduction.
With this fault-tolerant architecture in place, Google claims Willow features the first processor to achieve ‘below threshold’ quantum error correction, i.e., the ability to achieve fewer errors while scaling up the number of qubits.