Laser with $7\times10^{-18}$ stability

Ben Lewis, Zoey Z. Hu, Dahyeon Lee, Kyungtae Kim, William Warfield, Kai Zhou, Daniele Nicolodi, Thomas Legero

Oct 21, 2026arXiv:2610.30532v1

physics.optics

Scorecard· 16/16

8.0/10 impact

Sets a new record for laser stability (7×10⁻¹⁸) with convincing thermal-noise-limited evidence and direct clock validation, giving a benchmark and enabling tool for the optical clock community, though the concept is incremental.

Abstract

We report a laser with record frequency stability, based on a 21 cm cryogenic silicon cavity. For best short-term stability, the cavity operates at 6 K to minimize Brownian thermal noise and achieves a fractional stability of $7\times10^{-18}$ between 2 to 50 s of averaging time and 2.5 mHz linewidth. For best long-term stability the cavity operates at 16 K, where silicon has a zero-crossing of thermal expansion. Slightly increased thermal noise limits the performance to $9\times10^{-18}$ , consistent with the expected temperature scaling, but the long-term performance is improved. The laser then has a linewidth of 2.9 mHz and long-term frequency drift less than 4 μHz s $^{- 1}$ . The laser is used to enhance the performance of a state-of-the-art Sr lattice clock, operating with Rabi and Ramsey interrogation times beyond 10 s.

AI Impact Assessments

(1 model)

Paper type: Experimental/tool-system paper in precision metrology (ultrastable lasers).

Core Contribution

The paper reports a cryogenic silicon cavity laser (Si7, 21 cm) with a fractional frequency instability of 7×10⁻¹⁸ at 6 K and 9×10⁻¹⁸ at 16 K, a new record, flat between about 2 and 50 s. Linewidths are 2.5 and 2.9 mHz, and long-term drift is below 4 µHz/s. The advance comes from two design changes to the earlier JILA/PTB 21 cm silicon cavity:

Operating at 16 K (the second zero-crossing of silicon's thermal expansion) or at the 6 K cryostat base temperature, which lowers Brownian coating noise.

Enlarging the mode from 480 to 850 µm by using a 10.2 m radius-of-curvature mirror, which reduces thermal noise further.

Together these cut thermal noise three- to fivefold. Enabling engineering includes:

A custom low-vibration closed-cycle cryostat with a circulating-helium loop.

Two-stage vibration isolation, including an active vibration isolation table.

Erbium thermal masses, which have high heat capacity at 16 K, giving a roughly 4-day thermal time constant.

A synthetic-FM-triplet PDH scheme for RAM control.

Tiltmeter feedforward.

The laser is demonstrated on a Sr lattice clock with Ramsey dark times up to 12 s and a 12.4 s Rabi pulse.

Methodological Rigor

The stability measurement uses a three-cornered hat against two other silicon cavities (Si3, Si6) with Ekstrom–Koppang error bars. An independent check comes from the Sr clock for long-term behaviour. This is the standard approach, and the cross-validation is good.

The thermal-noise attribution is persuasive:

The floor is flat across 2–50 s, as expected for flicker frequency noise.

Noise drops with temperature from 16 K to 6 K.

The implied coating loss of 3.3×10⁻⁴ matches independent literature values at 16 K and 4 K.

A full noise budget covers vibration, RAM, PDH, detection and temperature.

Weaknesses:

The 6→16 K improvement is slightly smaller than the √T scaling predicts, which is explained only by invoking coating loss.

Excess noise beyond 100 s is unexplained, with candidate causes of vacuum pressure and intracavity power heating that cannot be measured.

Si3 and Si6 showed uncanceled fiber noise below 10 s, so the reference-cavity contributions are less certain. The three-cornered hat then relies on assumptions about uncorrelated noise.

Drift is reported from only about 3 weeks of data, with a 45 Hz jump after an interruption that was manually removed. A 2-day maintenance interruption is needed every two months because of helium contamination.

The 6 K long-term performance is not shown by the Sr clock.

Potential Impact

Optical clocks. Clocks with 10⁻¹⁸–10⁻¹⁹ uncertainty are limited in averaging time by the Dick effect and laser coherence. A laser this stable permits 10+ s interrogation, approaching the 170 s Sr lifetime, and up to 10× faster averaging in the Boulder network. This directly speeds up clock comparisons, systematic evaluation and frequency-ratio measurements.

Fundamental physics. Dark-matter searches, Lorentz and position invariance tests, gravitational-wave detection at mHz frequencies and equivalence-principle tests are all laser-noise limited.

Timekeeping. The drift of about 10⁻²⁰/s, comparable to hydrogen masers, supports flywheel oscillators for all-optical time scales and the redefinition of the second.

Design template. The paper demonstrates a validated path (low temperature + larger mode + low vibration) toward 10⁻¹⁸. It points to crystalline coatings for roughly 2×10⁻¹⁸, and the authors cite it as support for ambitious proposals such as a lunar silicon cavity.

Direct industrial translation is limited. The system is a complex, expensive, bespoke laboratory apparatus, and the audience is the national-metrology-lab and frontier-AMO community.

Timeliness

The field is in the "laser stability is the bottleneck" regime. Clock systematics have reached 10⁻¹⁸–10⁻¹⁹, and competing efforts include room-temperature long cavities at 4×10⁻¹⁷, a 4 K 10 cm cavity at 4×10⁻¹⁷, and crystalline-coated silicon at 2.5×10⁻¹⁷. A result well below 10⁻¹⁷ is timely, and it was likely the benchmark the community awaited. Redefinition of the SI second also raises the demand for such oscillators.

Strengths

A clear record: roughly 3× better than the previous best (2.5×10⁻¹⁷ with crystalline mirrors) and about 10⁻¹⁷ below prior cavity results.

Strong evidence that the device is thermal-noise limited, with the noise budget closed above a few seconds.

Two operating modes (6 K for short-term stability, 16 K for long-term stability) with a clear trade-off.

Low drift only 9 months after cooldown, whereas 124 K cavities took years to settle. This suggests a temperature-dependent drift mechanism.

Validation through actual clock spectroscopy, not just beat notes.

Practical engineering lessons (erbium thermal mass, circulating helium, feedforward) that are broadly reusable.

Limitations

The record is incremental in concept. It combines known ingredients (cryogenic Si, large mode, low-T) with excellent engineering, so conceptual novelty is modest.

Thermal noise is still set by amorphous coatings. The next step (crystalline coatings) is not demonstrated here, and it was previously shown with excess noise and photothermal issues.

Reproducibility is limited by the need for custom cryostats, ultra-low-vibration infrastructure and multiple reference cavities of comparable quality, and there are no engineering drawings.

The excess noise beyond 100 s is unresolved, and there is only a short drift record.

Operational interruptions (helium contamination every 2 months, hysteresis at the 10 Hz level) are a practical drawback.

Overall

This is a high-quality, record-setting metrology paper from the leading groups (JILA, PTB). It is not conceptually revolutionary, but it sets the new benchmark in ultrastable lasers and materially enables improved optical clocks and fundamental-physics searches. It is likely to be widely cited as the state-of-the-art reference. Its impact is concentrated in the precision-measurement community, with real downstream value in clock networks, timekeeping and tests of fundamental physics.

Rating:8/ 10

Significance 8Rigor 7.5Novelty 6Clarity 7.5

Generated Oct 23, 2026

Comparison History (0)

No comparisons yet.