How Singapore’s Lutetium Atomic Clock Reached 10⁻¹⁹ Accuracy

A clock is only as accurate as its reference frequency—and as the experimenters’ accounting for everything that can shift it. At Singapore’s Centre for Quantum Technologies, Murray Barrett’s team developed a lutetium-ion optical clock whose reported fractional frequency uncertainty is near 1 × 10⁻¹⁹. Published in Nature on 23 September 2026, the result is described by the team as the most accurate yet reported for an optical atomic clock.306

How the team built the reference

The clock uses a single charged lutetium-176 ion. A laser probes an optical transition at 848 nanometres, providing the frequency reference for the clock. That fast optical oscillation offers a finely divided measure of time, but it does not guarantee accuracy: electromagnetic fields and thermal radiation can still change the frequency the experiment is trying to measure.72521

Barrett’s group had been working toward this result for years. Its 2015 research plan identified hyperfine averaging as a way to make lutetium’s transition suitable for a clock. Rather than relying on one internal atomic state, the method combines measurements across related states so that some state-dependent shifts cancel. The group subsequently demonstrated hyperfine averaging using dynamic decoupling in a multi-ion lutetium clock.29152

Other disturbances required separate attention. In 2019, the researchers measured and suppressed frequency differences among ions in a three-ion clock, including effects associated with electric quadrupole moments. Research on lutetium also examines shifts caused by thermal radiation and electric-field gradients. Those results help explain the development path; they should not be mistaken for a claim that lutetium is insensitive to every environmental change.1821

What the two record figures measure

The approximately 1 × 10⁻¹⁹ figure describes reported fractional frequency uncertainty for the clock reference. The team also compared two independently built single-ion clocks, reporting a comparison uncertainty of 5.7 × 10⁻¹⁹. The comparison checks whether the references agree; its larger uncertainty is not a second way of stating the individual clock’s accuracy.6524

The group has used correlation spectroscopy to compare lutetium frequency references. By reducing the effect of interrogation-laser noise on a comparison, the technique makes a small difference between clocks easier to resolve. It does not, on its own, rule out a systematic shift shared by both clocks. A 200-hour comparison is described in the question’s account of the result, but the supplied source excerpts do not independently establish that duration or its full uncertainty budget.2628

The familiar claim that the clock would gain or lose less than a second over 260 billion years is an illustration, not an observation made over that span. At a constant fractional error of 1 × 10⁻¹⁹, the arithmetic gives roughly one second in 317 billion years; at 1.2 × 10⁻¹⁹, it gives roughly 264 billion years. Reports use both uncertainty figures, so the analogy depends on which is meant.64

Why the result matters—and what remains

A more accurate optical reference could contribute to a future redefinition of the second, but a result from one laboratory cannot alone establish reproducibility across laboratories or determine which atomic transition should become the standard. The supplied evidence does not confirm a formal timetable for such a change.3024

Accuracy at this scale also makes location part of the measurement. Near Earth’s surface, the gravitational frequency shift across a one-millimetre height difference is roughly 1.1 × 10⁻¹⁹. Comparing clocks at different sites therefore requires knowledge of their gravitational-potential difference, not just confidence in the clocks themselves. Making a transportable version poses another test: it would have to retain control of the ion, laser and environmental shifts—and demonstrate its uncertainty again after moving.521