Phase-continuous comparison of three all-optical time scales over 20 days

Abijith S. Kowligy; Alejandra L. Collopy; Arman Cing\"oz; Ben Lewis; Dahyeon Lee; Jeffrey A. Sherman; Jonathan D. Roslund; Jun Ye; Kai Zhou; Kyungtae Kim

REVIEW 2 major objections 3 minor 3 references

Phase-continuous comparison of three all-optical time scales over 20 days

T0 review · 2 major / 3 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Three all-optical time scales, steered by a strontium clock, kept <1e-16 instability and <100 ps total difference over 20+ days.

desk verdict A strong direct 20-day comparison of three optical-flywheel time scales; the headline numbers are measured and credible, but the simulation-based claim to beat AT1 is fitted from the same data. read the letter →

arxiv 2608.00910 v1 pith:T3WH7IO4 submitted 2026-08-02 physics.atom-ph physics.optics

Dahyeon Lee , Kyungtae Kim , Zoey Z. Hu , Ben Lewis , William Warfield , Kai Zhou , Alejandra L. Collopy , Jeffrey A. Sherman

show 6 more authors

Abijith S. Kowligy Parth B. Patel Jonathan D. Roslund Arman Cingöz Martin M. Boyd Jun Ye

This is my paper · ORCID

classification physics.atom-phphysics.optics

keywords all-opticaltimescaleopticalflywheelcryogenicsiliconcavityiodineclockstrontiumlatticehydrogenmaserDickeffectfrequencycomb

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that a time scale can be kept entirely in the optical domain by steering optical flywheels with a strontium optical lattice clock, removing the hydrogen-maser bottleneck that currently limits timekeeping. Three such time scales, based on two cryogenic silicon cavities and one commercial iodine clock, ran continuously for over 20 days and were compared phase-continuously. They reached relative fractional instability below 1e-16 after a few days of averaging and a total time difference under 100 ps. This matters because current time scales need weeks of averaging to reach performance that optical flywheels reach in days, and the result points toward a practical path for realizing the future optical second.

What carries the argument

The optical flywheel: a continuously running laser oscillator whose short-term stability is orders of magnitude better than a hydrogen maser, so that a time scale can bridge gaps in optical-clock operation without a large Dick noise penalty. This paper uses two cryogenic silicon cavities (Si3 and Si6) and an iodine optical clock (VA21), each steered by a Sr optical lattice clock. During Sr-off gaps, steering uses linear drift extrapolation for Si3 (slope from the past 10 hours), a zero-drift assumption for Si6, and a Kalman filter for VA21. Optical frequency combs convert each flywheel to the rf domain for phase-continuous comparison.

What would settle it

Run the same three-flywheel comparison with an additional continuously operating optical frequency standard that is independent of the Sr clock; if the absolute time error of any all-optical time scale exceeds the pairwise differences reported here, the pairwise comparison is not representative of true timekeeping error.

Watch

Extended reading notes

Core claim

The central claim is that replacing hydrogen masers with optical flywheel oscillators removes the main bottleneck in time scale performance. The paper realizes three independent all-optical time scales—two from cryogenic silicon cavities and one from a commercial iodine optical clock—each steered by a high-uptime 87Sr optical lattice clock, and compares them phase-continuously for 27 days. The measured relative fractional instability reaches below 1e-16 after a few days of averaging, and the peak-to-peak time difference between pairs stays around 200 ps, ending below 100 ps. The authors also verify that conversion of the optical signals to 100 MHz rf via optical frequency combs adds less tha

Load-bearing premise

During gaps when the Sr clock is off, the steering algorithm assumes the silicon cavities' frequency drift stays equal to its average over the past 10 hours (Si3) or zero (Si6); if the true drift changes on shorter timescales, the predicted frequency is wrong and the reported roughly 20 ps per 6-hour gap (hence <100 ps total) is an underestimate.

Editorial extensions

If this is right

Maser-free time scales are realistic now: with roughly 70% Sr uptime, three independent all-optical time scales reached sub-1e-16 instability in days rather than weeks.
A cryogenic silicon cavity steered only one hour per day would still outperform a hydrogen maser steered twelve hours a day, according to the paper's simulation.
All-optical time scales can be downconverted to conventional 100 MHz rf signals with less than 2 ps added timing error, so existing rf infrastructure can distribute the benefit.
The same steering algorithms and flywheel hardware can be adopted by national timing laboratories to improve UTC and TAI once optical frequency standards become more continuous.
Commercially available iodine clocks and fiber lasers make the flywheel stack realistic outside specialized laboratories.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If flywheel drift during gaps is the limiting error, then improving drift prediction (e.g., integrating cavity temperature or using a second flywheel as a drift reference) should reduce the roughly 20 ps per 6-hour gap in proportion to prediction error, not to gap length alone.
Because the reported comparison is pairwise, the <100 ps total difference is a strict bound on each time scale's error only if the three flywheels' noises are independent; correlated errors such as common Sr steering effects would be invisible.
A direct test of the gap-error model would be to add a fourth, continuously operating optical clock and measure each all-optical time scale's absolute time error through a 10+ hour gap; the Si3 prediction error should equal the extrapolation error of its 10-hour average drift.
The approach transfers naturally to optical clock networks: pairing each optical clock with an optical flywheel and comparing over fiber links would let the network realize a distributed all-optical time scale without masers.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Here is the take: this is the first multi-flywheel all-optical time scale comparison running phase-continuously for 20+ days. Two cryogenic silicon cavities and a commercial iodine clock, all steered by one Sr lattice, reach sub-1e-16 relative instability after a few days and accumulate <100 ps mutual time difference. The central measurement is direct, internally consistent, and looks solid. The optical-to-rf conversion fidelity check with a residual slope consistent with zero is a nice, honest addition, and the cycle-slip tracking in post-processing is the kind of detail that makes the result believable.

What's new: Milner et al. demonstrated a single optical-carrier timescale; this paper scales it to three parallel optical flywheels, includes a turnkey iodine clock, and runs the comparison long enough to make the stability statement interesting. The 20-day phase-continuous record, with the gap-length distribution and time-difference correlation, is real evidence.

The soft spots are real but contained. First, the claim that each of the three time scales individually beats NIST AT1 is not measured; it comes from a simulation whose noise parameters were estimated from the same overlapping Allan deviations that the simulation then reproduces (Extended Data Fig. 2 and Methods). The agreement between simulated and measured relative time differences is a consistency check, not an independent validation. That overstates the strength of the 'outperforms AT1' claim. Second, all three time scales share the same Sr steering reference, so pairwise comparisons cancel common-mode Sr errors. The quoted <10^-16 is a relative statement; absolute accuracy of any single all-optical time scale is not demonstrated. The paper is mostly careful about saying 'when compared with each other,' but the forward-looking part leans on the unvalidated simulation.

One more minor concern: the Si3 extrapolation uses the prior 10-hour average drift. If the drift changes on shorter timescales, the per-gap prediction error grows. In this data set that behavior would show up in the measured time differences, so the <100 ps final value is not an underestimate for this run; it just doesn't generalize automatically to other operating conditions.

Who this is for: time/frequency metrology people, especially those thinking about optical redefinition of the second and UTC realization. It deserves a serious referee; the measured comparison is a milestone even if the simulation claims need to be tempered or independently verified.

Referee Report

2 major / 3 minor

Summary. The paper reports a 27-day phase-continuous comparison of three all-optical time scales, each generated by steering an optical flywheel (two cryogenic silicon cavities and one iodine clock) with a shared Sr optical lattice clock. The authors measure pairwise time differences and overlapping Allan deviations, reporting relative instability below 1e-16 after a few days of averaging and a total time difference below 100 ps over 20 days. They also verify the fidelity of optical-to-rf conversion with a residual slope consistent with zero. A simulation based on measured flywheel noise characteristics is used to project individual time-scale stability and to compare with the NIST AT1 maser ensemble.

Significance. The central measurement is a significant technical milestone. It demonstrates that optical flywheels can sustain continuously operating time scales with sub-nanosecond accumulated errors over 20 days, and that pairwise relative stability reaches the low-1e-16 regime. The direct, phase-continuous comparison is a strong result, and the explicit correction of comb cycle slips and the verification of optical-to-rf conversion fidelity add credibility. However, the broader claim of superiority over AT1 rests on a simulation whose noise parameters are fitted from the same data, and the common Sr reference means the pairwise measurements do not constrain individual time-scale stability. The paper would be strengthened by out-of-sample validation or a more careful qualification of the simulated comparison.

major comments (2)

[Methods (Simulation of all-optical time scales); Fig. 4(c)] The noise parameters for Si3, Si6, and VA21 are estimated from the drift-removed overlapping Allan deviations of the same flywheels (Fig. 2c, Extended Data Fig. 2). The agreement between simulation and measurement in Fig. 4(b) is therefore a consistency check, not an out-of-sample validation. The claim that all three time scales outperform AT1 (Fig. 4c) rests on the unverified assumption that these noise models are representative outside the fitted data. Please provide an out-of-sample validation (e.g., fit on a subset of the 7-month record and test on the 20-day campaign) or explicitly reframe Fig. 4(c) as a model-based projection rather than a measurement.
[Main text, Fig. 4] Because all three time scales are steered by the same Sr optical frequency standard, the pairwise comparisons in Fig. 4(a,b) cannot detect common-mode errors in the Sr reference. The individual time-scale stability curves in Fig. 4(c) rely on the simulation's implicit assumption that the Sr reference contributes negligibly, which is not tested by this experiment. This limitation should be stated explicitly in the main text, and the statement 'All three time scales perform better than NIST AT1' should be qualified accordingly.

minor comments (3)

[Steering description (main text)] The concern that the reported ~20 ps per 6-hour gap is an underestimate because of drift extrapolation does not land: the time differences in Fig. 4(a) are directly measured, so they already reflect the actual drift behavior. However, the paper should distinguish this empirical average from a worst-case bound; a brief discussion of the distribution of gap errors would be helpful.
[Fig. 5(a)] The text mentions a gap in the rf comparison during MJD 61055-61060 due to a 'numerical precision issue' (Ref. 37). Please clarify what is meant by this issue and confirm that the optical comparison remains continuous for the full 20 days.
[Abstract / Introduction] Minor grammar: 'superior short-term stability than hydrogen masers' should be 'superior short-term stability to hydrogen masers' or 'better than'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: headline instability and time-difference claims are direct measurements; the simulation is explicitly labeled and not used to generate headline results.

full rationale

The paper's central quantitative claims—sub-10^-16 relative instability after a few days and <100 ps total time difference over 20 days—are obtained from direct, phase-continuous pairwise measurements of the three time scales (Fig. 4a,b), not from a fit. The steering algorithm description is explicit, and the gap-dependent accumulated time differences are measured, not derived. The only potentially questioned step is the simulation of individual time-scale stabilities in Fig. 4c, which uses noise parameters 'estimated from the drift-removed overlapping Allan deviations of each flywheel' (Methods). This is a forward model, not a circular reduction: the individual time-scale stability is not fitted from the pairwise time differences; it is computed by simulating flywheel noise, applying the same steering algorithm, and masking with the actual Sr uptime pattern. The pairwise measurements are used only to check consistency ('The validity of the simulation was confirmed by reconstructing the relative time differences between pairs of simulated time scales'), which is a reasonable in-sample model validation, not a fitting of the target quantity. The paper explicitly labels Fig. 4c as 'simulated stability,' so it does not present a model output as a measurement. The claim that all three time scales outperform AT1 is a model-based projection, and the Methods acknowledge the limitation ('the true timing error of individual time scales cannot be measured without a superior time scale'). These are correctness-risk concerns, not circularity. Self-citations to prior work (e.g., Ref. 10) provide context but are not load-bearing for the measured results. Therefore no circular step meets the required evidentiary bar.

Assumptions & free parameters 9 free parameters · 4 assumptions · 0 invented entities

The central claim of safe long-term operation with <100 ps error rests on (i) fitted noise models used in simulations that underpin the AT1 comparison, (ii) steering design choices (extrapolation windows, update intervals), and (iii) the assumption that the continuous phase comparison chains (combs, phase meters, fiber links) introduce no uncorrected systematic error. No new physical entities are invented.

free parameters (9)

Si3 flicker frequency noise = 3.5e-17 (at 1 s)
Estimated from drift-removed overlapping Allan deviation of Si3 (Fig. 2c) for simulation.
Si3 random walk frequency noise = 3e-18 (at 1 s)
Estimated from Si3 Allan deviation for simulation.
Si6 flicker frequency noise = 6e-17 (at 1 s)
Estimated from Si6 Allan deviation for simulation.
Si6 random walk frequency noise = 6.5e-18 (at 1 s)
Estimated from Si6 Allan deviation for simulation.
VA21 white frequency noise = 4e-14 (at 1 s)
Estimated from VA21 Allan deviation for simulation.
VA21 flicker frequency noise = 3e-16 (at 1 s)
Estimated from VA21 Allan deviation for simulation.
VA21 random walk frequency noise = 2e-18 (at 1 s)
Estimated from VA21 Allan deviation for simulation.
Si3 drift extrapolation window = 10 hours
Steering parameter: drift slope averaged over past 10 hours for gap prediction.
Steering measurement averaging interval = 1 minute
Steering parameter: chosen to reach the silicon cavity flicker floor.

assumptions (4)

domain assumption Flywheel frequency noise is stationary and can be represented as white + flicker + random-walk noise
Assumed in the simulation (Methods, 'Simulation of all-optical time scales'). If the noise is non-stationary or has additional components, the simulated time scale stability and AT1 comparison are invalid.
domain assumption The Sr lattice clock provides an exact frequency reference during its uptime
Used to steer all three flywheels. Any systematic error is common to all three time scales and does not affect their relative comparison, but it would affect the absolute accuracy of a single time scale.
domain assumption The comparison chains preserve phase continuity except for explicitly corrected cycle slips
Required for the <100 ps total time difference claim; the paper states all measurements except Sr are continuous, and comb cycle slips are tracked and corrected (Methods, 'Correction of frequency comb cycle slips').
domain assumption The steering algorithms (linear extrapolation for Si3/Si6, Kalman filter for VA21) are appropriate and correctly implemented
The accumulated time difference depends on the prediction error during gaps; a suboptimal filter would degrade the reported performance.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Phase-continuous comparison of three all-optical time scales over 20 days." pith.science (2026). https://pith.science/paper/T3WH7IO4

@misc{pith2026260800910,
  author       = {Pith},
  title        = {Pith review of: Phase-continuous comparison of three all-optical time scales over 20 days},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/T3WH7IO4}},
  note         = {Machine review of arXiv:2608.00910}
}

abstract

Optical frequency standards have progressed rapidly over the past two decades, leading to the anticipated redefinition of the SI second by an optical frequency. However, time scales have not yet significantly improved despite this development because they are still fully reliant on rf flywheel oscillators, mostly hydrogen masers, which impose a performance limit related to incompletely sampled noise known as the Dick effect. To best benefit from the exceptional stability and accuracy of optical frequency standards, time scales must employ optical flywheels with orders-of-magnitude better short-term (<$10^4$ s) stability than masers. Here, we introduce three optical flywheel oscillators (two cryogenic silicon cavities and one iodine optical clock) with superior short-term stability than hydrogen masers and long-term stability on par with masers. Steering each optical flywheel with a high-uptime Sr optical frequency standard generates three parallel all-optical time scales with continuous operation over >20 days. When compared with each other, these all-optical time scales achieve <$10^{-16}$ relative instability after just a few days of averaging. During typical steering gaps of ~6 hours, the accumulated time difference is ~20 ps, leading to the total time difference of <100 ps over the full measurement period. With the proliferation of long-distance optical fiber links and commercialization of optical flywheels and frequency standards, we anticipate all-optical time scales to be the future of timekeeping.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

3 extracted references · 3 canonical work pages

[2]

M., Luiten, A

Fortier, T. M., Luiten, A. N. & Margolis, H. S. Optical atomic clocks: defining the future of time and frequency metrology. Optica 13, 143–163 (2026). 3. Dimarcq, N. et al. Roadmap towards the redefinition of the second. Metrologia 61, 012001 (2024). 4. Riehle, F. Towards a redefinition of the second based on optical atomic clocks. Comptes Rendus Phys. 16...

work page 2026
[22]

Roslund, J. D. et al. Optical Clocks at Sea. Nature 628, 736–740 (2024). 23. Herman, D. et al. Femtosecond Timekeeping: Slip-Free Clockwork for Optical Timescales. Phys. Rev. Appl. 9, 044002 (2018). 24. Sinclair, L. C. et al. Invited Article: A compact optically coherent fiber frequency comb. Rev. Sci. Instrum. 86, 081301 (2015). 25. Lezius, M. et al. Spa...

work page 2024
[41]

D., Triano, T

Caldwell, E. D., Triano, T. M. & Sinclair, L. C. High-precision optical time and frequency transfer. Adv. Opt. Photonics 17, 375 (2025). 42. Hoghooghi, N. et al. Ultrastable optical frequency transfer and attosecond timing in deployed multicore fiber. Optica 12, 894 (2025). 43. Lepek, A. & Walls, F. L. Cross correlation analysis improves time domain measu...

work page arXiv 2025

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [2]

M., Luiten, A

Fortier, T. M., Luiten, A. N. & Margolis, H. S. Optical atomic clocks: defining the future of time and frequency metrology. Optica 13, 143–163 (2026). 3. Dimarcq, N. et al. Roadmap towards the redefinition of the second. Metrologia 61, 012001 (2024). 4. Riehle, F. Towards a redefinition of the second based on optical atomic clocks. Comptes Rendus Phys. 16...

work page 2026

[2] [22]

Roslund, J. D. et al. Optical Clocks at Sea. Nature 628, 736–740 (2024). 23. Herman, D. et al. Femtosecond Timekeeping: Slip-Free Clockwork for Optical Timescales. Phys. Rev. Appl. 9, 044002 (2018). 24. Sinclair, L. C. et al. Invited Article: A compact optically coherent fiber frequency comb. Rev. Sci. Instrum. 86, 081301 (2015). 25. Lezius, M. et al. Spa...

work page 2024

[3] [41]

D., Triano, T

Caldwell, E. D., Triano, T. M. & Sinclair, L. C. High-precision optical time and frequency transfer. Adv. Opt. Photonics 17, 375 (2025). 42. Hoghooghi, N. et al. Ultrastable optical frequency transfer and attosecond timing in deployed multicore fiber. Optica 12, 894 (2025). 43. Lepek, A. & Walls, F. L. Cross correlation analysis improves time domain measu...

work page arXiv 2025