Optical clocks at sea

The study addresses the challenge of deploying high-performance optical clocks outside laboratories, particularly on mobile platforms such as ships, to enable robust positioning, synchronization, and timing infrastructure beyond reliance on GNSS. Optical timekeeping offers femtosecond-level timing jitter at short times and multiday subnanosecond holdover, and long-distance optical time transfer could enable global picosecond-level synchronization. Molecular iodine (¹²⁷I₂) has a long history as an optical frequency standard with recognized length standards and early optical clock demonstrations, and has been explored for space missions. The authors aim to demonstrate integrated, fieldable iodine-based optical clocks that maintain nanosecond-level timing over days while operating continuously in dynamic, real-world maritime environments.

The paper builds on the legacy of iodine-stabilized lasers as optical frequency standards and length standards, including detailed spectroscopy and absolute frequency atlases at 532 nm. Prior work demonstrated molecular iodine clocks and iodine-based frequency references for space, including flight-like and sounding rocket demonstrations. The broader context includes advances in optical clock networks reaching 10⁻¹⁸-level accuracy, transportable optical clocks for geodesy and relativity tests, and quantum-limited optical time transfer for future satellite links. Commercial and emerging atomic frequency standards (masers, rubidium, caesium) provide benchmarks; masers offer excellent long-term stability but require large environmental chambers, limiting field deployment. Previous compact optical standards based on vapor-cell and two-photon transitions demonstrated potential for robust, compact architectures without laser cooling.

System design: The clocks employ a robust iodine vapor-cell spectrometer (no consumables, no laser cooling, no prestabilization cavity) that is first-order insensitive to platform motion. Mature laser components at 1,064 nm (clock laser) and 1,550 nm (comb outputs) are used. Each integrated 3U, 19-inch rack-mount clock (volume ≈35 l) provides 100 MHz, 10 MHz, and 1 pulse-per-second outputs, plus auxiliary optical outputs at 1,064 nm and 1,550 nm. Physics packages (spectrometer, laser system, frequency comb) were custom-built to reduce SWaP. FPGA-based controllers implement digital locks for the laser and comb, RAM servo control, and stabilized pump/probe powers. Each system uses a commercial 1U power supply and a control laptop, consumes about 85 W (excluding external PSU), and weighs 26 kg. Two high-performance clocks (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 the physics package volume by 50% and power by 5 W. Frequency comb design and electronics are largely identical across units.

Laboratory evaluation at NIST: In April 2022, PICKLES and EPIC were shipped to NIST Boulder and operated on an optical table in a temperature-stabilized but otherwise typical lab environment. The 10 MHz output of each clock was compared to a 5 MHz active hydrogen maser signal (NIST ST05, Symmetricom MHM-2010) using a Microchip 53100A phase noise analyzer in a three-cornered hat (TCH) configuration. ST05 was operated in a separate environmental chamber (environmentally uncorrelated reference). Simultaneously, the 1,064 nm optical beat between PICKLES and EPIC was monitored for cross-validation. Systems were run autonomously and remotely monitored; data collection lasted 34 days. Allan deviation analysis used TCH-derived instability for 1–1,000 s and direct comparison to ST05 for >1,000 s without windowing, dedrifting, or filtering. Linear drift removal was also applied for assessing dedrifted long-term instability.

At-sea deployment (RIMPAC 2022): Three clocks (PICKLES, EPIC, VIPER) were installed in an open server rack with one 1U power supply and control laptop per clock, an uninterruptible power supply, three frequency counters for pairwise beatnotes, and a 53100A phase noise analyzer for 100 MHz comparisons in a TCH configuration (total rack height 23U). The rack was hard-mounted in a Conex cargo container craned onto the deck of HMNZS Aotearoa. After leaving port, the clocks ran continuously for 3 weeks with no user intervention (aside from one VIPER restart due to an external PSU software fault). Environmental conditions in the air-conditioned container had 2–3 °C daily temperature swings and 4–5% RH variation; air conditioning cycled on/off. Ship dynamics included peak pitch ±1.5° at ±1.2° s⁻¹, roll ±6° at ±3° s⁻¹, surge/sway/heave accelerations up to ±0.4/±1.5/±1.2 m s⁻², and vertical RMS vibration 0.03 m s⁻² (1–100 Hz). The Earth’s magnetic field projection varied by ±270 mG along the voyage around the Hawaiian Islands. Performance was assessed via overlapping Allan deviation, TCH analysis for <100 s, direct pairwise instability for longer times, and power spectral density comparisons versus ship motion. Temperature-driven effects were evaluated via the observed diurnal cycle; drift was assessed relative to lab campaigns and environmental chamber tests post-deployment.

• Short-term performance at NIST: PICKLES and EPIC achieved short-term instabilities of approximately 5 × 10⁻¹⁴/√τ and 6 × 10⁻¹⁴/√τ, outperforming the reference maser (ST05) at short times. Phase noise at 10 and 100 MHz (via optical frequency division) and at 1,064 nm was excellent and below commercial atomic-disciplined oscillators.
• Long-term stability at NIST: Both clocks reached fractional frequency instabilities <5 × 10⁻¹5 after 100,000 s, corresponding to temporal holdover <300 ps after 1 day. Raw instability was 4 × 10⁻¹⁵ (PICKLES) and 6 × 10⁻¹⁵ (EPIC) after 10⁵ s. With linear drift removal, instability improved to <2 × 10⁻¹⁵ (PICKLES) and <3 × 10⁻¹⁵ (EPIC) at 10⁶ s, maintaining <10⁻¹⁴ for nearly 6 days (PICKLES). Measured drift rates versus UTC(NIST) were 2 × 10⁻¹⁵/day (PICKLES) and 4 × 10⁻¹⁵/day (EPIC).
• Identified features: A plateau in PICKLES’ Allan deviation near 20,000 s (≈7 h) with peak deviation 4 × 10⁻¹⁵ (≈2 Hz optical deviation, ≈2 ppm of the hyperfine line center) suspected from RAM via a spurious etalon; mitigated in EPIC via build modifications.
• At-sea performance: Short-term instability at sea matched lab performance up to ≈1,000 s with no SNR degradation despite vibration and motion. All three clocks were immune to dominant ship motion near 0.1 Hz. Medium-timescale instability was driven by diurnal temperature swings. The PICKLES–EPIC pair showed 8 × 10⁻¹⁵ combined instability at 100,000 s without drift correction (≈400 ps temporal holdover over 24 h), with similar drift rate to NIST measurements. VIPER exhibited 1.3 × 10⁻¹³/√τ short-term instability, a diurnal instability peaking at 4 × 10⁻¹⁴ around 40,000 s, and averaged to 2.5 × 10⁻¹⁴ after 1 day. VIPER lacked magnetic shields yet maintained excellent stability under varying geomagnetic field.
• Operational robustness: Three integrated clocks operated continuously aboard a naval ship in the Pacific Ocean for 20 days, accruing timing errors below 300 ps per day. Performance compares favorably to active hydrogen masers in one-tenth the volume, with better phase noise and short-timescale instability.

The results demonstrate that compact, integrated iodine-based optical clocks can achieve maser-class performance in dynamic, real-world maritime environments without specialized environmental control. By leveraging a robust vapor-cell architecture and optical frequency division, the clocks deliver low phase noise, excellent short-term stability, and multiday subnanosecond holdover. Laboratory measurements against UTC(NIST) validate drift rates at the few ×10⁻¹⁵/day level and instability reaching below 10⁻¹⁴ over many days, while sea trials show performance largely unchanged under ship motion, vibration, variable temperature/humidity, and changing geomagnetic field. These findings address the need for fieldable, mobile optical timekeeping to support navigation, resilient timing infrastructures, remote synchronization of quantum networks, and distributed sensing. Compared with hydrogen masers, the iodine clocks offer similar or better holdover with substantially reduced size and environmental demands, enabling deployment outside specialized labs. The demonstrated immunity to motion-induced perturbations and operation without consumables or laser cooling underscores their suitability for distributed timing networks and mobile platforms.

The study presents fully integrated, compact iodine optical clocks that operate continuously at sea with performance comparable to active hydrogen masers at a fraction of the size and environmental burden. Two high-performance units (PICKLES, EPIC) achieved short-term instabilities near 5–6 × 10⁻¹⁴/√τ, long-term instabilities below 5 × 10⁻¹⁵ at 10⁵ s, and drift rates of 2–4 × 10⁻¹⁵/day. A third, smaller system (VIPER) met relaxed performance goals while maintaining robustness without magnetic shielding. The sea deployment validated environmental insensitivity to motion, vibration, and diurnal conditions, maintaining subnanosecond daily holdover. These results mark a technological step toward widespread adoption of optical timekeeping and future optical timing networks. Future work includes continued SWaP and performance improvements; the next-generation rackmount system has achieved 2 × 10⁻¹⁴/√τ short-term instability with reduced SWaP to 30 l, 20 kg, and 70 W and elimination of the external power supply, further enhancing deployability.

• At-sea testing lacked standard absolute reference clocks (e.g., caesium beam or GPS-disciplined rubidium) for external comparison; three-clock analysis and separate environmental testing were used to bound correlations.
• Potential for correlated environmental sensitivities (temperature, humidity, ship motion, geomagnetic field) among colocated clocks was mitigated by differing designs and analysis but cannot be entirely excluded.
• Temperature-driven medium-timescale instabilities were observed due to limited air-conditioning capacity; a plateau around 10⁵ s in EPIC’s performance was attributed to thermal effects.
• PICKLES exhibited a RAM-related Allan deviation plateau near 20,000 s due to a spurious etalon (addressed in later builds).
• VIPER’s earlier design had higher temperature sensitivity; although stable, it showed more pronounced diurnal features.
• Iodine vapor-cell clocks are not as accurate as state-of-the-art laboratory optical clocks using trapped atoms/ions, though they offer superior fieldability and SWaP.