Beyond-Ten-Hour Coherence in a Decoherence-Free Trapped-Ion Clock Qubit
Quantum coherence headlines are usually where precision goes to die. This one actually survives contact with the paper. A team led by researchers at Tsinghua University, the Beijing Academy of Quantum Information Sciences, Hefei National Laboratory, and collaborators including the Institute for Basic Science in South Korea reported a trapped-ion logical clock qubit with a fitted coherence time of about 10.5 hours, according to their arXiv preprint and the full-text arXiv HTML version.
The cleaner version matters more than the magical one. This was not a single physical ytterbium ion sitting there heroically coherent for half a day. The physical 171Yb+ ions in the same experiment decohered in about 8.1 seconds, as the paper shows in both the preprint and the HTML text. What lasted much longer was a logical qubit encoded across two ions in a decoherence-free subspace, using the anti-aligned states |01⟩ and |10⟩ so that common-mode magnetic noise and global microwave oscillator phase noise largely cancel out. In quantum, the least interesting version of the claim is often the one everyone wants to tweet.
That distinction is why this looks like a real advance in quantum memory and clock performance, not a general breakthrough in quantum computing. The group built a mixed-species Yb-Ba-Yb chain, with a central 138Ba+ ion providing continuous sympathetic cooling while the two 171Yb+ ions stored the logical qubit, as described in the arXiv HTML methods. The result pushed the dominant limit away from ordinary laboratory slop and toward something narrower and more believable: stochastic ion-exchange hopping under a residual magnetic-field gradient. When your bottleneck gets that specific, it usually means the engineering is real.
The headline number also deserves its own footnote before the field adds a halo. The authors measured the long-lived signal out to 1,600 seconds, saw only modest contrast decay over that window, and then fit an exponential to estimate a coherence time of (3.77 ± 1.09) × 10^4 seconds, or roughly 10.5 hours, according to the preprint. That is standard practice, but it is still an extrapolated lifetime rather than 10 straight hours of raw measurement. The same methods section notes a projected coherence beyond 21 hours if ion hopping were removed, which is interesting engineering direction, not something that happened in the lab on Thursday, per the HTML version.
The larger context makes the result more convincing, not less. This is the latest step in a long, same-lab progression rather than a random arXiv miracle. In 2017, members of this research line reported single-qubit coherence beyond 10 minutes in a trapped-ion system, first in a Nature Photonics paper and in the corresponding arXiv preprint. In 2021, the group reported a single 171Yb+ ion with estimated coherence beyond one hour in Nature Communications, with an open-access version available via PubMed Central. The new result is the logical-memory version of that arc: not better vibes, but better encoding.
There is also a broader field context. A separate 2025 trapped-ion result, published through the American Physical Society, reported coherence above two hours for a decoherence-free-subspace-encoded qubit in a cryogenic trap. So the new 10.5-hour result does not appear from nowhere. What this paper adds is a room-temperature demonstration that passive error protection can move a trapped-ion memory into a different regime without first turning the whole apparatus into a shrine.
That matters most for architectures where memory is a real systems problem. Trapped-ion platforms have long argued that computation, storage, and networking can be separated across zones or nodes, and a long-lived idle qubit is useful if classical coordination, transport, or remote entanglement takes time. The Kihwan Kim lab at Tsinghua has been building toward exactly that picture for years, as its group page makes plain. The novelty here is that the paper turns passive error avoidance from a nice theoretical phrase into a practical way of escaping the noise sources that capped earlier records.
The institutional story is similarly unromantic. This is university and national-lab work, not a startup demo deck: Tsinghua, BAQIS, Hefei National Laboratory, and partner institutions dominate the author list in the arXiv paper. An older Tsinghua University news item on the 2017 record tied that earlier phase of the program to Chinese state research funding, and I did not find a fresh English-language institutional press release for the new result. Fine. The paper is better than a press release anyway.
What this does not show is a trapped-ion computer running for 10 hours, or a shortcut around the rest of fault-tolerant quantum computing. Long memory is one axis. Gate fidelity, routing, readout, throughput, and scale remain rude. But for once the impressive version of the story is also the less inflated one: a two-ion logical qubit outlasted its physical constituents by roughly four orders of magnitude because the researchers encoded the information in a place where the dominant noise had less to hit. Quantum likes to sell destiny. Here it settled for disciplined architecture, and that is why the result is worth taking seriously.