Optical clocks at sea

The study addresses whether compact, robust optical clocks can deliver maser-level timekeeping performance outside the laboratory, specifically on moving maritime platforms. The authors motivate the work by the growing reliance on Global Navigation Satellite System (GNSS) timing and the need for resilient, high-precision time sources for navigation autonomy, geophysical monitoring, distributed sensing, and quantum networks. They highlight advances in fieldable optical timekeeping and long-distance optical time transfer that suggest the feasibility of picosecond-level global synchronization. Molecular iodine has a long legacy as an optical frequency standard, including recognized length standards and early optical clock demonstrations, and is compatible with mature 1,064 and 1,550 nm laser technologies. The authors propose integrated iodine-based optical clocks that prioritize robust lasers and vapour-cell spectroscopy to achieve a favorable balance of size, weight, power, environmental insensitivity, and performance, and they test these clocks against national time standards and at sea.

The paper situates its work within prior developments: GNSS provides global nanosecond-level synchronization based on distributed microwave atomic clocks. Laboratory optical clocks have achieved fractional inaccuracies below 10−18, and optical time transfer now supports femtosecond-level performance over long distances. Iodine transitions have served as optical frequency and length standards for decades, underpinning early optical clock demonstrations and being considered for space missions. Transportable optical clocks based on neutral atoms and ions have shown high performance but face challenges in size, complexity, and environmental sensitivity for field deployment. Related compact approaches include two-photon transitions in alkali vapours. Active hydrogen masers are the field benchmark for long holdover but require large environmental chambers to stabilize temperature and humidity. This work extends iodine’s role by demonstrating fully integrated, fieldable iodine optical clocks with performance comparable to masers and validated operation in a maritime environment.

System architecture: The clocks are integrated in 3U, 19-inch rackmount chassis (≈35 l per clock), producing 100 MHz, 10 MHz, and 1 pulse-per-second outputs, and auxiliary optical outputs at 1,064 nm (clock laser) and 1,550 nm (comb). Physics packages (iodine spectrometer, laser system, erbium-fibre frequency comb) and control electronics were designed in-house to minimize SWaP. FPGA-based controllers implement digital locks (laser and comb), residual amplitude modulation (RAM) suppression, and pump/probe power stabilization. Each system consumes ~85 W (excluding external PSU) and weighs ~26 kg.

Clock variants: Two identical high-performance systems (PICKLES, EPIC) target short-term instability <1×10−13/√τ; a third (VIPER) targets <5×10−13/√τ using a smaller spectrometer and simplified laser, reducing physics package volume by ~50% and power by ~5 W.

Spectrometer: Modulation transfer spectroscopy interrogates the a10 hyperfine component of the I2 R(56) 32–0 transition at 532 nm with multi-pass pump/probe beams for high SNR. RAM is sensed and actively cancelled; pump/probe powers are stabilized. PICKLES/EPIC use traditional vapour cells with cold fingers stabilized at −5 °C; self-collisional shift ≈ −2×10−12 per °C necessitates millikelvin stabilization achieved via TECs (≤1 mK over >10 days). VIPER uses a starved iodine cell (fixed gaseous iodine ~4 Pa at 0 °C), eliminating the cold finger and reducing temperature coefficient by ≥10× (out-of-loop fluctuations ≤10 mK during at-sea trials). Optical benches with bonded, alignment-free optics are enclosed by temperature-stabilized thermal shields (≤1 mK over >10 days). PICKLES/EPIC spectrometers (~2.5 l) include single-layer magnetic shields; VIPER’s smaller (~1 l) spectrometer omits magnetic shielding. Measured sensitivities: second-order Zeeman <1×10−15 G−2; linear Zeeman <1×10−14 G−1 for fields parallel to propagation due to imperfect polarization.

Laser system: A low-noise 1,064 nm fiberized source is split; one path to the comb, the other frequency-doubled to 532 nm then split into pump/probe. An AOM offsets the pump by 200 MHz to avoid coherent backgrounds, and is frequency-modulated (typ. 300 kHz with ~1 MHz deviation) and amplitude-modulated to cancel RAM; drive tones are digitally synthesized by the FPGA. Light shift ≈ 1×10−12 per mW implies that ~0.1% power stability suffices to limit shifts to ~1×10−15. RAM-induced frequency shifts are ~1×10−15 per ppm; RAM is stabilized to ppm levels. The assembled laser occupies ~1 l.

Frequency comb: An all-fibre erbium comb at 1,560 nm operates at 200 MHz repetition. The pulse train is amplified and spectrally broadened to enable self-referencing (f0) and a branch to 1,064 nm for optical beat. Typical f0 linewidth ≈250 kHz (beat SNR ~45 dB in 300 kHz RBW); optical beat SNR ~40 dB. f0 is stabilized via pump current; the optical beat is stabilized using an intracavity PZT with ~200 kHz bandwidth. Timing jitter between two identical combs supports clock readout of ≈4×10−17 at 1 s. The comb optics package occupies ~0.5 l. RF division provides 100 MHz, 10 MHz, and 1 PPS outputs from the repetition rate.

Autonomous operation: Software autonomously brings the system from OFF to fully locked, including temperature regulation, ML-assisted identification of the correct hyperfine manifold, laser centring and lock, RAM and power servos, comb f0 lock to a pre-calibrated value, tooth identification, and comb lock to the clock laser. The controller monitors faults and re-establishes locks as needed.

NIST laboratory evaluation: In April 2022, PICKLES and EPIC were shipped to NIST Boulder and operated on an optical table without special environmental shielding (lab temperature-stabilized and in regular use). The 10 MHz outputs were compared to a 5 MHz output from NIST active hydrogen maser ST05 (low drift ~3×10−17/day) using a Microchip 53100A phase noise analyzer in a three-cornered hat (TCH) configuration; the 1,064 nm optical beat between PICKLES and EPIC was also recorded. The clocks ran autonomously for 34 days. Allan deviation curves combine TCH-extracted 1–1,000 s instabilities with direct comparisons to ST05 beyond 1,000 s. Phase noise for a single clock inferred from pairwise measurements: at 1 Hz offset, 10 MHz ≈ −124 dBc/Hz and 100 MHz ≈ −105 dBc/Hz; white phase noise floors near −170 dBc/Hz (10 MHz) and −155 dBc/Hz (100 MHz). Optical phase noise at 1,064 nm at 1 Hz offset ≈ +23 dBc/Hz with a servo bump near 10 kHz.

At-sea (RIMPAC 2022) evaluation: Three clocks (PICKLES, EPIC, VIPER) plus three 1U PSUs, three control laptops, UPS, three frequency counters, and a 53100A were installed in an open server rack (total 23U) inside a Conex container craned onto the deck of HMNZS Aotearoa. The system operated at sea for ~3 weeks with one VIPER restart due to an external PSU software fault. Environmental conditions inside the Conex included daily 2–3 °C temperature swings and 4–5% RH variations; the rack was in front of a cycling A/C unit. Ship dynamics: pitch ±1.5° at ±1.2°/s, roll ±6° at ±3°/s, surge/sway/heave accelerations up to ±0.4/±1.5/±1.2 m s−2, vertical RMS vibration ~0.03 m s−2 (1–100 Hz). The Earth’s field projection varied by ±270 mG along the route. The 100 MHz outputs were intercompared in TCH; pairwise optical beats (1,064 nm) were also collected.

Long-term drift characterization: Relative PICKLES-EPIC drift measured at NIST, at sea, and later at the authors’ facility follows a consistent trend (~−2×10−15/day), indicating drift dominated by vapour-cell changes (e.g., helium permeation). VIPER exhibited reproducibility <2×10−16 after a 4 h off state in retrace tests.

• The integrated iodine optical clocks achieved maser-class performance in compact 3U rackmount systems (~35 l, ~85 W, 26 kg), operating autonomously.
• At NIST (34-day run): short-term instabilities (TCH) were ~5×10−14/√τ (PICKLES) and ~6×10−14/√τ (EPIC), outperforming the reference maser’s short-term performance. Fractional instability <5×10−15 after 1×10^5 s averaging, corresponding to <300 ps timing error after 1 day. Raw instability remained below that of a Microchip 5071A caesium beam clock for ~5.5 days; dedrifted iodine performance beat caesium at all observed timescales.
• Long-term drift against UTC(NIST): ~2×10−15/day (PICKLES) and ~4×10−15/day (EPIC), consistent with expected vapour-cell behaviour and an order of magnitude lower than typical space-qualified Rb standards after prolonged operation.
• Phase noise: single-clock L(f) at 1 Hz offset of ~−124 dBc/Hz (10 MHz) and ~−105 dBc/Hz (100 MHz), with low white phase noise floors (≈−170/−155 dBc/Hz), highlighting advantages of optical frequency division. Optical phase noise at 1,064 nm at 1 Hz offset ≈ +23 dBc/Hz.
• At sea (RIMPAC 2022, ~3 weeks; 3 clocks, 20 days continuous operation as stated in abstract): short-term instabilities matched NIST results up to ~1,000 s despite motion and vibration. The PICKLES–EPIC pair achieved ~8×10−15 instability at 1×10^5 s without drift correction, implying ~400 ps 24 h holdover. Clocks showed immunity to dominant ship-motion bands (~0.1 Hz); medium-timescale instability correlated with diurnal temperature swings in the Conex.
• VIPER (simplified design) showed short-term ~1.3×10−13/√τ and a diurnal instability peaking near 4×10−14 at ~4×10^4 s, yet averaged to ~2.5×10−14 after 1 day; it lacked magnetic shielding but remained stable under varying geomagnetic projection.
• Observed broad Allan deviation plateau near 2×10^4 s in PICKLES (≈4×10−15 peak) attributed to RAM coupling via a spurious etalon; addressed in EPIC build.
• Next-generation system improvements: short-term instability reduced to ~2×10−14/√τ; SWaP reduced to ~30 l, ~20 kg, ~70 W; external PSU eliminated.

The findings demonstrate that iodine-based optical clocks can deliver maser-class long-term stability and superior short-term phase noise and instability in a compact, robust package suitable for deployment on moving platforms. By emphasizing a mature laser technology at 1,064/1,550 nm and vapour-cell spectroscopy with careful control of RAM, power, and temperature, the systems maintain sub-nanosecond holdover over days and exhibit insensitivity to ship motion and environmental variations. The NIST comparison validates absolute performance against UTC(NIST) and establishes drifts in the low 10−15/day range, while the at-sea tests show that the same performance persists under realistic maritime conditions without special environmental chambers. This addresses the central question of whether high-performance optical timekeeping can be fielded outside the lab and specifically at sea. The results are significant for resilient navigation (A-PNT), redundancy for national timekeeping, and enabling picosecond-level network synchronization when combined with optical time transfer. Compared to active hydrogen masers, the iodine clocks achieve similar multiday holdover with smaller volume and no dedicated environmental chamber, while outperforming masers in short-term phase noise. Their robustness and autonomy indicate strong potential for distributed optical timing networks and integration into operational systems.

The authors present fully integrated, iodine-based optical clocks that combine a compact 3U rackmount form factor with maser-class long-term stability and superior short-term phase noise, validated both against UTC(NIST) over 34 days and in a demanding maritime environment over weeks. The clocks maintained <300 ps/day timing error and achieved ~8×10−15 instability at 10^5 s at sea, with drifts of only a few 10−15/day. The systems operated autonomously without special environmental conditioning, demonstrating readiness for real-world deployment. This constitutes, to the authors’ knowledge, the highest-performing sea-based optical clocks to date and a key step toward widespread adoption of optical timekeeping and future optical timing networks. Future directions include continued SWaP reductions, further suppression and characterization of environmental sensitivities (e.g., RAM and temperature coupling), extended multiclok at-sea campaigns with independent references, integration with optical time-transfer links for end-to-end network demonstrations, and long-term drift mitigation via improved cell engineering and materials.

• At-sea testing colocated three clocks without an independent, external reference (e.g., caesium beam or GNSS-disciplined reference), leaving a possibility of residual common-mode environmental correlations, though three-way comparisons and differing designs (VIPER) mitigate this risk.
• Medium-timescale instabilities were influenced by diurnal temperature swings and limited air-conditioner capacity in the Conex; environmental control was not optimized.
• PICKLES exhibited a plateau around 2×10^4 s attributed to RAM coupling via a spurious etalon; while mitigated in EPIC, residual RAM sensitivity remains a potential limitation.
• VIPER, using a simplified spectrometer and lacking magnetic shielding, showed larger diurnal temperature sensitivity; while performance remained strong, it indicates trade-offs with reduced thermal/magnetic control.
• Long-term drift at the few 10−15/day level, likely from vapour-cell changes (e.g., helium permeation), limits absolute accuracy over months to years without calibration or drift compensation; not as accurate as state-of-the-art trapped-ion/neutral optical clocks.
• Laboratory phase-noise/instability comparisons rely on specific instruments and configurations; different setups may introduce measurement biases if not carefully controlled.