Error correction creates qubits that glitch only 5 out of a million times
A quantum chip made by Quantinuum (shown) demonstrated a qubit that made errors only five times out of a million.
Quantinuum
Quantum computers capable of doing a million reliable operations could bring about a new stage of quantum calculation. That “megaquop” regime — as it’s called — isn’t here yet. But scientists just took a significant step toward it.
A quantum computer built by the company Quantinuum has error rates so low that it could perform a million operations with only about five errors, researchers report in a paper posted October 19 on the company’s website and arXiv.org.
“These statistics are consistent with this being a good building block for that megaquop machine,” says quantum physicist David Hayes of Quantinuum, who is based in Broomfield, Colo. In order to reach that regime, however, such computers will need to be scaled up.
Quantinuum’s computer, known as Helios, is made of electrically charged atoms, called ions, hovering over a chip. Those ions serve as the 98 quantum bits, or qubits, of the computer. But, like all qubits, those ions are finicky. Their states are delicate and can flip unintentionally and mess up calculations, like a 0 randomly switching to a 1.
A technique called quantum error correction tackles this by combining multiple error-prone qubits into one, forming a supercharged unit called a logical qubit that is less prone to mistakes. (For those who grew up on 1990s TV, it’s a bit like how Captain Planet is born from the combined powers of the gang of Planeteers.)
Logical qubits come in different sizes, integrating a given number of physical qubits — the ions, in this case. Each time a logical qubit is scaled up, going, say, from seven physical qubits to 19, the error rate drops by a factor called lambda. A previous demonstration of error correction by Google, reported in Nature in 2024, had a lambda of about 2, meaning errors were cut in half each time the logical qubits were scaled up.
Quantinuum’s computer has a lambda of about 10. So errors dropped quickly when the researchers made larger logical qubits. A logical qubit composed of 37 physical qubits had just 5 errors in a million operations — close to what would be needed for the megaquop regime. However, that experiment involved only a single logical qubit. Reaching the megaquop regime could possibly involve around 50 to 100 logical qubits.
The large value of lambda means that achieving a desired error rate takes fewer physical qubits, says Ken Brown, a quantum engineer at Duke University and a scientific adviser for the company Logiqal who was not involved with the research. “That’s a huge savings in the overall number of qubits you would need. So that’s what’s exciting.”