Quantum memories gain protection without a measure-and-react loop
Two newly published experiments keep quantum information alive longer by building correction into controlled physical operations.
Reading tools for text size, copying, printing and your reading list

Protecting quantum information often involves repeatedly checking for signs of an error, then deciding how to respond. Two experiments published in Nature Physics on November 16 demonstrate a different approach: controlled operations remove errors without that repeated measurement-and-feedback loop. Both results were already public in September 2025 preprints; today’s milestone is journal publication.[1][2][3][4]
In Zhongchu Ni and colleagues’ experiment, information is encoded in microwave photons inside a superconducting cavity. Photon loss threatens that stored information. Their correction process transfers the error information to a helper qubit, then resets the helper through a deliberately engineered interaction with its surroundings. The protected memory lasts about two milliseconds, 18% longer than the best benchmark within their system.[1][3]
Lida Sun and colleagues report a separate experiment using controlled drives and dissipation. Their protected reaches 1.04 times the coherence time of their system’s best physical qubit. That crosses “break-even”: the protected information outlasts the strongest physical-qubit reference in that apparatus.[2][4]
What counts as break-even?
In Sun’s bosonic experiment the protected qubit used photon-number states in a microwave cavity. In the September 2025 author version, fitting 20 correction rounds gave 1.627 milliseconds of coherence, against 1.565 milliseconds for the best physical reference, encoded in the same cavity. The ratio is about 1.04. The authors give the fit values without a numerical error bar on that ratio in the accompanying results text.[2][4]
Li and colleagues instead protected information stored in internal levels of a calcium ion against changing magnetic fields. Their April 2025 author version reports a lifetime of 11.6 ± 1.9 milliseconds under deliberately added low-frequency noise, versus 0.9 ± 0.1 milliseconds for the physical reference. Under ambient noise the fitted lifetimes were 60.9 ± 33.0 and 26.0 ± 5.4 milliseconds, respectively. Their journal paper appeared October 23.[5][6]
Each team tested storage against its own best physical reference under its chosen noise conditions. In the ion experiment, changing the noise changed both the lifetime and the confidence in the fit. The larger number of milliseconds cannot tell a reader which machine would run a calculation more reliably.[4][6]
Keep your reading trail
Keep up with these topics
Explore the subjects covered in this story.