NPL and PTB Compare Two Ytterbium-Ion Optical Clocks Over a 2220 km Fiber Link at 4.8e-18 Uncertainty, arXiv 21 October 2026: The First International Comparison Under the CCTF's 5e-18 Bar, TRL 5
Quentir Defense Monitor
Evidence-based insights for quantum defense and security. Published by Quentir Systems LLC · October 22, 2026.
On October 21, 2026, a team of 29 physicists from the United Kingdom, Germany and France posted a preprint reporting that two single-ion optical clocks, one at the National Physical Laboratory in Teddington and one at the Physikalisch-Technische Bundesanstalt in Braunschweig, agree to 1.0 part in a billion billion, with an uncertainty of 4.8 parts in a billion billion, after being compared across 2220 kilometers of optical fiber. The preprint by Alexandra Tofful, Martin Filzinger and colleagues calls this the first international clock comparison with a fractional uncertainty below 5 × 10⁻¹⁸, and the number matters because the committee that will propose a new definition of the second wrote in 2025 that nobody had reached it. This post reads the preprint for what it establishes, what it leaves open, and what a defense reader should take from a result that is, on its face, a metrology milestone.
The defense relevance sits in the link rather than in the clocks. A force that loses satellite timing needs a terrestrial reference it can trust, and a national time scale built on optical clocks is only as good as the proof that two such clocks, built by different teams, actually tell the same time. The preprint supplies that proof for one species over one continental fiber path, and it also shows, in its own uncertainty budget, where the next limit lies. The fiber contributes less than one part in 10⁻¹⁸. The gravity field between Teddington and Braunschweig contributes three and a half times more.
What NPL and PTB measured from March 18 to April 12, 2025: two ytterbium-171 ion clocks, three French relay nodes and 12.6 days of data
Both clocks interrogate the same transition in a single trapped ytterbium-171 ion, the electric octupole line between the ²S₁/₂ (F=0) and ²F₇/₂ (F=3) levels, which is one of the transitions on the CIPM list of secondary representations of the second. The hardware is independent. NPL holds its ion in an end-cap trap and cancels the probe-light shift with an interleaved dual-servo scheme that alternates between two pulse powers. PTB uses a conventional ring-electrode Paul trap and Rabi-controlled hyper-Ramsey spectroscopy. The point of the exercise is that two groups who made different engineering choices arrive at the same frequency, which is what the redefinition process requires and what earlier same-species comparisons between other laboratories had failed to show.
The two clocks never left their buildings. Each was measured against a 1542 nm transfer laser through a single branch of a frequency comb, and the transfer light traveled the 2220 km between the two institutes over the European research fiber network, through three nodes in France at the Laboratoire de Physique des Lasers, the Laboratoire Temps Espace in Paris and the University of Strasbourg. Optical path-length stabilization on every segment kept the comparison phase coherent. The PTB to Strasbourg and Strasbourg to Paris segments each carried a systematic uncertainty below 10⁻¹⁹. The Paris to NPL segment, which NPL describes on its own site as a 760 km fiber link to Paris and which crosses the Channel, was the weakest piece at 9 × 10⁻¹⁹.
The campaign ran from March 18 to April 12, 2025. In those 24.7 days the team collected 1.09 × 10⁶ seconds of usable data, or 12.6 days. The NPL clock was available 92 percent of the time and the PTB clock 86 percent, and the combined uptime of both clocks and the whole link was 51 percent. The frequency ratio came out at 1.000 000 000 000 000 001 0(48), a difference of 1.0 × 10⁻¹⁸ with a total uncertainty of 4.8 × 10⁻¹⁸, combining a statistical part of 1.6 × 10⁻¹⁸ and a systematic part of 4.6 × 10⁻¹⁸. The combined systematic budget of the two clocks themselves is 2.8 × 10⁻¹⁸, dominated by the black-body radiation shift, the electric quadrupole shift and the second-order Doppler shift. The single largest term in the whole comparison, 3.5 × 10⁻¹⁸, is the correction for the relativistic redshift between the two sites, which depends on the difference in gravity potential and on how that difference moves with tides and groundwater. The authors put the equivalent height uncertainty between the two clocks at just over 3 centimeters.
Why 4.8e-18 clears the bar CCTF Recommendation 24-3 said no one had reached, and what the April 2026 seven-clock campaign left open
The Consultative Committee for Time and Frequency maintains a roadmap toward replacing the cesium definition of the second with an optical one, with a proposal due at the 29th General Conference on Weights and Measures in 2030. Its Recommendation 24-3, adopted at the 24th session in 2025, lists the criteria that still stand in the way. It records "noticeable inconsistencies in some comparisons of optical frequency standards," notes that no consensus exists on a single species versus an ensemble, and states that "a comparison uncertainty below 5E-18 between clocks of the same type developed by different laboratories has not yet been achieved." It adds that the criterion "could realistically be fulfilled by 2030."
The new preprint is the first published measurement that meets that line, and it arrives eighteen months after the data were taken. The prior best for this pair came from a larger campaign, posted in April 2026 by Marco Pizzocaro and 43 co-authors, that compared seven optical clocks at four institutes over the same European fiber network and reported the NPL and PTB ytterbium-ion clocks in agreement at 7.7 × 10⁻¹⁸, the first international verification of two independently built optical clocks below one part in 10¹⁷. That campaign was limited by statistics. Before the fiber network existed, the same two clocks had been compared by satellite in the ROCIT project of 2022 at 2.8 × 10⁻¹⁶, roughly sixty times worse than today's number. The progression from satellite to fiber is the story a timing engineer should notice: a continental fiber link now compares clocks at a level a satellite link cannot approach.
What the preprint does not claim deserves equal weight. One pair of clocks of one species met the bar for one month. The CCTF criteria also ask for regular contributions of optical clock data to International Atomic Time with uncertainties below those of the cesium fountains, and the recommendation says plainly that this is not yet happening and that cesium fountains remain critical through and beyond the redefinition. A 51 percent combined uptime is a research campaign, and the authors say so by listing the sustainability of links and clocks as unfinished work. They also offer the alternative route: carry one clock to the other site so the gravity term drops out, which is what transportable optical clocks are being built to do.
Quantum pillar: networking (network time synchronization). Use posture: defensive. Technology readiness: TRL 5 of 9. Two laboratory clocks were compared over a real 2220 km deployed telecom fiber path across three countries for 25 days, which is a validation in a realistic setting of the link and the clocks together, well short of a continuously operating timing service with the uptime a network would require.
What a fiber-linked clock pair means for GNSS-independent timing: the UK framework's ten measures and a 3 centimeter gravity problem
Timing is the quiet dependency of modern forces. Radar networks, encrypted links, data centers and power grids all take their time from satellite navigation receivers, and the United Kingdom's own estimate, in the Position, Navigation and Timing resilience framework the Department for Science, Innovation and Technology published on October 18, 2023, is that a 24-hour loss of satellite timing would cost the economy £1.4 billion and a seven-day loss £7.6 billion. The framework's ten measures include a proposal for a national timing center that would give the UK a resilient and sovereign terrestrial time scale, UTC(NPL), "including sovereign components and optical clocks," a Ministry of Defense time proposal for a system of last resort, a terrestrial eLoran navigation system, holdover clocks and resilient receiver chips. The same push funded the six quantum sensing and navigation firms, among them a satellite-clock project and a rail-positioning project, that the Monitor read in the £50 million competition announced on October 13, 2026.
The connection to the preprint is direct. A sovereign time scale anchored on optical clocks needs two things the paper tests. It needs clocks that agree with clocks elsewhere, so that UTC(NPL) stays tied to international time when satellite links are degraded, and it needs a terrestrial channel that carries the reference without degrading it. The preprint shows that the second requirement is now met at a level far beyond any military need: a link uncertainty of 9 × 10⁻¹⁹ on the Channel crossing corresponds to a timing error that would take centuries to reach a nanosecond. Fiber time transfer at that quality means the limit on a continental timing network is no longer the fiber. It is knowledge of the gravity field, and that is a geodesy problem with a 3 centimeter answer that gets worse as groundwater and tides move the potential under each building.
For a defense buyer the result also reframes what GNSS-independent timing can honestly promise. An optical clock on its own is a frequency reference, and it becomes a time scale only when it is compared, steered and distributed. The NPL to PTB link demonstrates comparison and distribution over a path with an undersea segment, the kind of infrastructure a national timing authority would lean on if satellites were jammed. It does not demonstrate hardening, redundancy or the 99.999 percent availability a radar network would specify, and the 51 percent figure is the honest measure of that gap.
Who gains, and what stands between a 25-day campaign and a timing service a force could rely on
The posture reading is defensive. The value of a continent-spanning optical clock comparison lies in holding one's own time when the satellite signal is denied or spoofed, and in proving that a sovereign reference is consistent with everyone else's. Nothing in the preprint helps an adversary find, degrade or deceive a system. The same fiber network serves geodesy and fundamental physics, and the authors use the long baseline to tighten a Lorentz-invariance test by a factor of five, but the capability a program office would buy is the defensive one: an auditable chain from an ion in a trap to a time stamp on a network, with no satellite in it.
Three gaps separate the published result from a timing service. The first is continuity. Twelve and a half days of data in twenty-five, with the link and both clocks up together half the time, is adequate for a frequency comparison and inadequate for a reference that has to be there at three in the morning. The second is the gravity term. A 3.5 × 10⁻¹⁸ redshift uncertainty is irrelevant to a radar that needs nanoseconds, but it caps what the link can deliver for the redefinition, and it means every new node on a clock network needs a geodetic survey, which NPL's site puts at 2 to 3 centimeters for the European sites it has measured. The third is the fiber itself. The research network runs on dedicated dark fiber with laser repeater stations at the nodes, and the Channel crossing cost nine times the uncertainty of the German and French segments. A military network would run on commercial or owned fiber with different loss, switching and exposure to cutting, and the preprint says nothing about how the technique behaves there.
What the preprint establishes is specific and solid: between March 18 and April 12, 2025, two independently built ytterbium-ion clocks in two countries agreed at 1.0 × 10⁻¹⁸ with an uncertainty of 4.8 × 10⁻¹⁸ across 2220 km of fiber, the first international comparison to pass the 5 × 10⁻¹⁸ threshold the CCTF had listed as unmet in 2025, with the campaign dates, the uptime and the full uncertainty budget published. What it does not establish is a service. The redefinition of the second, which the CCTF intends to propose in 2030, moves one step closer on the strength of a single pair of clocks. A defense reader should file the result as evidence that the terrestrial fiber backbone of a satellite-free national time scale can now carry a reference better than any clock a force would field, and as a reminder that the clocks on the ends of it still run half the time.
Sources
Primary source: A. Tofful, M. Filzinger, T. Schmidt, A. Amy-Klein, E. Benkler, E. Cantin and 23 co-authors (NPL, PTB, LPL, LTE, University of Strasbourg, Leibniz University Hannover), "International atomic clock comparison achieving 5×10⁻¹⁸ uncertainty," arXiv 2610.29980, posted October 21, 2026. Other material: M. Pizzocaro et al., arXiv 2604.27963 (April 30, 2026); CCTF Recommendation 24-3 (2025), BIPM; UK Department for Science, Innovation and Technology, PNT resilience framework, October 18, 2023; NPL, optical frequency comparison and dissemination. Analysis by Quentir.
- preprint by Alexandra Tofful, Martin Filzinger and colleagues
- a 760 km fiber link to Paris
- Recommendation 24-3, adopted at the 24th session in 2025
- seven optical clocks at four institutes
- Position, Navigation and Timing resilience framework the Department for Science, Innovation and Technology published on October 18, 2023