There is a number in this story that does a strange thing to your sense of scale. The clock built at Singapore’s Centre for Quantum Technologies is accurate enough that it would need more than 260 billion years to drift off by a single second.
The universe is about 13.8 billion years old. You could run this clock from the Big Bang to now, nineteen times over, and it would still not have lost that second.
That is the headline, and it is genuinely true. It is also not the part that matters most. The part that matters is a committee in France, a definition written in 1967, and a decision that has been scheduled for around 2030.
The short version
- Who: a team led by Murray Barrett at the Centre for Quantum Technologies and the National University of Singapore, published in Nature on September 23, 2026
- The element is lutetium, specifically a single trapped lutetium ion, which is an unusual choice and the whole point of the result
- The number: a fractional frequency uncertainty of 1 part in 1019, measured to 19 decimal places, the lowest reported for any optical atomic clock
- Two clocks were compared and agreed to 5.7 parts in 1019, which the team says is the most precise clock comparison ever made
- The previous record holder was a NIST aluminium ion clock from July 2025, at 5.5 parts in 1019 after about 20 years of development
- Why lutetium: it is remarkably indifferent to temperature and magnetic fields, the two things that wreck other clocks
- Barrett’s claim: it would stay stable from Death Valley to the Antarctic plateau
- The real stake: the SI second is defined by caesium, and the international metrology community is preparing to redefine it using optical clocks, with ratification possible from 2030
What “most accurate” actually means here
Every atomic clock works the same way in principle. You find an atomic transition, which is what happens when an electron hops between two energy levels, and you count the oscillations of the radiation involved. Atoms are identical everywhere, so the oscillation rate is a constant you can build a ruler out of.
Since 1967, the official ruler has been caesium-133. One second is defined as 9,192,631,770 periods of the radiation from a specific transition in that atom. Count to that number and a second has passed. That is not a measurement of the second, it is the definition of it.
The problem is that caesium clocks are no longer the best clocks. The best caesium fountains land somewhere in the low 10-16 range for fractional uncertainty. Optical clocks, which use transitions that oscillate hundreds of thousands of times faster, have been beating that for two decades. Ytterbium and strontium clocks are already reported to the international time authorities as secondary representations of the second.
Which puts metrology in an awkward spot. The definition of the unit is now considerably worse than the instruments measuring it.
The accuracy ladder, in one picture
Fractional uncertainty is the standard measure, and it is simply the fraction of its own rate by which a clock might be off. A watch that gains ten seconds a day is wrong by about one part in ten thousand. The numbers below span sixteen orders of magnitude, which is why a linear chart would be useless.
The orange dot is the interesting one. Caesium sits roughly a thousand times to the left of the best optical clocks, and it is the one that defines the unit. Everything to its right is, technically speaking, measuring itself against a worse ruler.
Why lutetium, and why that is the clever bit
Most of the optical clock field has converged on a handful of species. Ytterbium and strontium dominate the lattice clocks. Aluminium, paired with magnesium as a control partner, powers the NIST quantum logic clock. Lutetium is not on that list, which is exactly why the Singapore result is more than a leaderboard update.
The limiting factor in a very good clock is almost never the atom’s willingness to oscillate steadily. It is the environment. Stray magnetic fields shift energy levels. Thermal radiation from the apparatus itself nudges the transition frequency. A great deal of the engineering in a state-of-the-art clock is spent measuring those perturbations so they can be subtracted, and the residual uncertainty in that subtraction is what ends up in the final number.
Singly ionised lutetium has an atomic structure with unusually low sensitivity to external electromagnetic fields, and it supports three separate clock transitions, which gives the team internal cross-checks most species cannot offer. Barrett’s framing is that the ion barely notices its surroundings. “The properties mean that high accuracy can be achieved even in a wide range of environments,” he said. “The lutetium clock would be stable even if you went from the hottest place recorded on Earth in Death Valley to the coldest place in the Antarctic plateau.”
He is also, to his credit, not hedging on the record. Barrett says he is confident it is “the most accurate clock in the world,” and adds: “in the future, I just don’t see how this clock can be beaten.”
That is a strong claim from a field that has broken its own records roughly every few years for two decades. Worth noting, and worth revisiting.
Two numbers, two different claims. The Singapore team reports 1 part in 1019 for its clock’s own systematic uncertainty, and 5.7 parts in 1019 for how closely two of their clocks agreed. NIST’s aluminium ion clock reported 5.5 parts in 1019. Those are related but not interchangeable figures, and headlines tend to flatten them into one. The useful thing about the comparison number is that it is a check against reality rather than a calculation about a single device.
The actual reason this is news
Record-setting clocks are published fairly regularly. What makes this one load-bearing is a process already underway at the Bureau International des Poids et Mesures, outside Paris.
The Consultative Committee for Time and Frequency has a published roadmap for redefining the second on an optical transition instead of a caesium one. The earliest the proposal can be presented is 2026. The earliest it can be ratified by the General Conference on Weights and Measures is 2030, and only if a set of mandatory criteria have been met first.
Those criteria are not “build the best clock.” They are harder than that, and they are the reason a single record is not enough.
| What the roadmap requires | Why it is required | Where it stands |
|---|---|---|
| Tie the new unit to the current second | Nothing measured since 1967 can be allowed to break | Ongoing, via caesium comparisons |
| Reproducibility across laboratories | One lab’s clock is a result. Several labs agreeing is a standard | The hard part |
| Frequency ratios between species | Ytterbium, strontium and ion clocks must be measured against each other at the best clocks’ precision | Active work worldwide |
| Contribution to international time | Optical standards already feed into atomic time as secondary representations | Happening now |
| A decision at the CGPM | Redefinition is a formal act, not a publication | 2030 at the earliest |
| Caesium keeps a role | It becomes a secondary standard rather than disappearing | Already planned |
Read against that list, a clock whose defining feature is indifference to its environment is interesting for a specific reason. Reproducibility across laboratories is the criterion most likely to hold the whole thing up, because it means a clock has to give the same answer in a different building with different magnets, different temperatures and different people. An ion that does not care about any of that is a good candidate for passing that test.
What this is actually for
The honest answer to “why do we need this” is that nobody needs to know the time to nineteen decimal places. The applications come from what you can do with a clock that good, which is mostly geophysics and navigation.
General relativity says clocks run slower in stronger gravity. At this precision, that effect is not a curiosity, it is a measurement tool. Two clocks at slightly different heights tick at measurably different rates, which means a network of them can map variations in Earth’s gravitational field, and therefore the distribution of mass underneath them. That is a way of watching water tables, magma and ice sheets move without going anywhere near them.
The more familiar application is already in your pocket and on your wrist. Satellite navigation is a timing problem rather than a positioning one: your device works out where it is by comparing the arrival times of signals from satellites carrying atomic clocks, and an error of a billionth of a second translates to roughly a foot of position error. Every GPS fix a running watch takes during a workout is downstream of atomic timekeeping, and the accuracy you experience is set by the clocks in orbit and the corrections applied to them.
There is also a pattern worth noticing in how precision instruments get reused. Equipment built for one purpose keeps turning out to be a sensor for something else entirely: researchers recently managed to turn a satellite internet constellation into a scanner for the upper atmosphere, using signals that were never intended for science. Clocks at this level invite the same opportunism.
What to watch next
- Independent replication. A lutetium clock built by someone other than the group that championed lutetium is the result that would settle the record
- Frequency ratio measurements. Lutetium against strontium, ytterbium and aluminium, at the precision of the best clocks, is the metrology that feeds the redefinition
- The 2026 proposal window. The roadmap allows a redefinition proposal to be presented from this year. Whether one is, and what it names, is the thing to watch
- Which species the second lands on. It may not be a single transition. Proposals exist for defining the second against a set of optical transitions rather than one
- Transportable clocks. The geodesy applications need clocks that leave the lab, which is where environmental insensitivity stops being elegant and starts being necessary
- Whether Barrett’s “cannot be beaten” holds. The field has a long record of making that sentence age badly
The bottom line
The 260 billion year figure is real and it is the least informative true thing about this clock. What the Singapore team actually did was measure a lutetium ion to a part in ten billion billion, then build a second one and show that the two agreed more closely than any pair of clocks ever has.
The second number is the one a metrologist cares about, because agreement between independent devices is the only evidence that a measurement means anything outside the room it was taken in. That is the criterion the redefinition of the second will be decided on, and it is where the next several years of this field will be spent.
It is also a useful reminder to read the claim rather than the headline. Records in physics are specific, conditional and narrower than they sound, and the gap between what a paper says and what gets reported is where most public confusion about science comes from. It happens with climate research, where a study about long-term planetary trends got widely reported as a prediction it never made, and it happens here too.
Nobody has redefined the second yet. Someone has built a very good argument for how it should be done.
Sources and further reading
- Centre for Quantum Technologies: Singapore scientists build the world’s most accurate atomic clock
- Phys.org: Scientists build the world’s most accurate atomic clock
- NIST: Ion clock sets a new record for the most accurate clock in the world
- BIPM: Frequently asked questions on the redefinition of the second
- Roadmap towards the redefinition of the second (preprint)
- UNILAD Tech: Scientists build the world’s most accurate atomic clock
About this article: GeekBlog covers U.S. technology news, AI, phones, smartwatches and gaming. Every story is written and checked under our Editorial Policy. Spotted a mistake or have a story tip? Contact our editors.

