Why More Is Better in Error-Resilient Quantum Computers

Quantum computers are currently plagued by errors, which significantly limits their practical use. However, a new study published in *Nature* reveals that researchers at Google and their collaborators have developed a quantum processor capable of fixing errors faster than they occur, marking a major breakthrough.

At the core of quantum computers are qubits, which are highly susceptible to errors. Presently, quantum computers experience about one error every thousand operations, a far cry from the one-in-10-billion error rates needed for many practical applications, as noted in the study.

To address these high error rates, scientists often rely on spreading quantum computations across multiple redundant qubits. These quantum error correction strategies allow quantum computers to detect and correct mistakes, enabling a group of “physical” qubits to function as one low-error “logical” qubit, which is essential for creating fault-tolerant quantum systems.

"Quantum error correction is the key to unlocking large-scale quantum applications like drug discovery, material design, and enhanced optimization," explains Kevin Satzinger, a research scientist at Google Quantum AI. "This could lead to advancements in areas such as pharmaceuticals and battery technology."

“When I first saw the error rate go down, and go down dramatically, that was the first time I thought to myself, ‘Damn, this is really going to work.’”

— Michael Newman, Google

Quantum error correction schemes, however, are not without limitations. Each strategy is effective only if the hardware’s error rates are low enough to make it beneficial. The error threshold varies depending on the strategy and the types of errors involved. Once this threshold is exceeded, adding more qubits can actually increase the number of errors, rather than reducing them.

"Quantum error correction has been around for nearly 30 years, and the core idea has always been that more error correction should improve error rates," says Michael Newman, a research scientist at Google Quantum AI. "However, this hasn't been true until now."

One of the leading quantum error correction methods being studied is the surface code, where qubits are arranged in a two-dimensional checkerboard pattern, with units of information encoded into sections of this lattice. It provides an error threshold of approximately 0.6 to 1 percent.

Researchers at Google Quantum AI and their collaborators have developed a new quantum computer architecture called Willow, which is capable of quantum error correction below the surface code’s error threshold.

In the study, the team executed surface codes on two Willow quantum processors: one with 72 superconducting transmon qubits—less sensitive to electric charges that can disrupt qubits' quantum properties—and another with 105 qubits.

The researchers discovered that the 105-qubit processor had an error rate of approximately 0.143 percent per error correction cycle, about half the error rate of the 72-qubit processor. For the first time, adding more physical qubits actually reduced error rates, as expected with quantum error correction.

"I've always believed in quantum error correction, but seeing it work is a different experience," says Michael Newman, a research scientist at Google Quantum AI. "When I saw the error rate decrease dramatically, that’s when I thought, 'This is really going to work.'"

Additionally, the 105-qubit processor is the first qubit array to have a longer lifetime than its individual physical qubits, lasting more than twice as long as the best physical qubit. This demonstrates that quantum error correction is enhancing the system as a whole, according to the researchers.

For practical applications, quantum computers need to correct errors during computations. The team notes that their new processors can perform quantum error correction mid-computation, with error-correction cycle times of 1.1 microseconds. Tests conducted over 15 hours show these processors could maintain stability over the extended timescales required for large-scale, fault-tolerant quantum algorithms.

“We have built a system that can scale in principle, but which we must now scale in practice.”

— Kevin Satzinger, Google

To assess Willow’s performance, the researchers used a benchmark known as random circuit sampling. While it has no direct real-world applications, this benchmark is considered the most challenging task for a classical computer that can also be completed by a quantum computer today. The team discovered that the 105-qubit Willow processor could complete the benchmark in under five minutes—an operation that would take today’s fastest supercomputer 10 septillion (or 10^25) years, a span vastly exceeding the age of the universe. (This benchmark is the same one Google used in 2019 to claim its quantum computer achieved quantum supremacy.)

The researchers credit these advancements to several upgrades, including improved fabrication techniques and the application of neural networks to account for device noise. As a result, the qubits were able to stay in superposition for nearly 100 microseconds, about five times longer than Google’s previous quantum processor, Sycamore.

Looking ahead, the team aims to demonstrate their quantum-error-corrected systems performing a quantum computation, according to Satzinger. However, he notes that their research is still far from achieving large-scale quantum applications, as they need to scale up their system to thousands of qubits and further improve the hardware's error rates. "We’ve built a system that can scale in principle, but now we must scale it in practice," Satzinger explains.