Quantum sensing for gravity cartography

The study addresses the challenge of creating practical, high-resolution gravity maps (gravity cartography) for subsurface imaging in real-world environments. Traditional gravimetry and atom interferometer-based gravimeters are limited by the need to average out micro-seismic vibrations, leading to long measurement times that preclude metre-scale spatial resolution surveys. The authors aim to overcome this limitation by developing a robust quantum gravity gradient sensor capable of suppressing dominant noise sources and systematic effects to enable fast, high-resolution detection of underground features. The work is motivated by applications in geophysics, civil engineering, archaeology, and environmental monitoring, where improved subsurface information can reduce risk and inform decision making.

Atom interferometry has provided highly sensitive measurements of gravity and fundamental constants, and has been demonstrated in diverse environments (laboratory, volcanoes, mountains, airborne, marine, and space). Conventional devices use light-pulse Raman transitions to create Mach–Zehnder-like matter-wave interferometers whose phase is proportional to local gravity. However, their performance for mapping is fundamentally limited by micro-seismic vibration averaging, restricting practical survey resolution. Early gravity gradiometers based on coupled atom interferometers established the principle of differential phase extraction to suppress common-mode noise. Prior studies also highlight impacts of laser systematics, magnetic fields, transverse motion, and environmental effects (buildings, topography) on gravity gradient measurements, motivating designs that enhance robustness and noise immunity for field deployment.

The authors implement a field-deployable cold-atom gravity gradiometer in an 'hourglass' configuration. Two counter-oriented single-beam MOTs produce vertically separated 87Rb atom clouds with a 1 m baseline aligned to sense the vertical gravity gradient component Gzz. Common Raman beams drive simultaneous light-pulse atom interferometers in both ensembles. Differential operation suppresses primary noise (vibration, micro-seismic), systematic shifts (tilt), and optical path length changes. The single-beam MOTs provide common laser intensity noise, stabilizing cloud temperatures to within a few hundred nK and reducing center-of-mass motion relative to six-beam MOTs; baseline variations are below 75 ppm (<0.1 E systematic). Practical design choices include a compact, robust optical delivery without off-axis beams, enabling long-term field alignment stability; compact magnetic shielding (25 dB) to suppress external fields; and an all-fiber laser system allowing independent control and reversal of counter-propagating Raman beams. Interleaving opposite Raman directions suppresses systematics from residual magnetic fields and parasitic Raman transitions without requiring a phase lock. Measurement sequence: Each MOT is loaded for 1–1.5 s, followed by sub-Doppler cooling to the microkelvin regime. Approximately 10^4 atoms participate per interferometer. The interferometry sequence applies three light pulses separated by time τ, and populations in the two ground states are detected via fluorescence at a typical 0.7 Hz measurement rate. Differential phases are extracted by ellipse (Lissajous) fitting of upper versus lower interferometer outputs to yield the gradiometric phase. Validation was performed by moving calibrated test masses near the sensor, producing a measured gravity gradient change of (205 ± 13.1) E, consistent with a modeled 202 E. Field survey: An 8.5 m line with 0.5 m station spacing was measured above a pre-existing 2 m × 2 m multi-utility tunnel (~0.2 m concrete wall) situated under a roadway between multistory buildings. A forward model of an air/soil contrast infinite cuboid void with local buildings and terrain was constructed using CAD plans, ground-penetrating radar cross-checks, and topographic scans, predicting a ~150 E peak anomaly (≈17.5 mrad interferometer phase). The gradiometer data were compared to the site model. Bayesian inference: A data-driven Bayesian framework, assuming a buried cuboid anomaly a priori, integrates gradiometer measurements with site/geophysical parameter estimates to infer position, depth, and cross-sectional area. Soil density for a void relative to surrounding soil was modeled with a Gaussian prior (mean −1.80 g cm^-3, SD 0.10 g cm^-3). The probability of excavation (POE) metric summarizes anomaly likelihood. An alternative inference, assuming known tunnel geometry and incorporating site topography, estimates soil density from the gradiometer data.

• The quantum gravity gradiometer achieves an average short-term sensitivity of (466 ± 8) μE/√Hz during outdoor operation and a statistical uncertainty of 20 E within 10 min.
• Baseline stability: relative baseline change <75 ppm, implying <0.1 E systematic error from baseline variations.
• Test-mass validation: measured gravity gradient change (205 ± 13.1) E versus modeled 202 E.
• Field survey (0.5 m spacing over 8.5 m) detected a 2 m tunnel with signal-to-noise ratio ~8, consistent with a modeled ~150 E peak anomaly.
• Bayesian inversion localized the tunnel center horizontally at (0.19 ± 0.19) m along the survey line and estimated the depth to center as (1.89 −0.59/+2.3) m.
• Density inference (with geometry known) yielded a near-Gaussian posterior for density contrast with mean −1.80 g cm^-3 and SD 0.15 g cm^-3.
• Static operation performance corresponds to ~1.4 ng uncertainty per individual gravimeter for 20 E gradient uncertainty, surpassing reported performance of commercial survey gravimeters by a factor of ~1.5–4.
• Operationally, the sensor could be repositioned within ~75 s and aligned to vertical within 1 millidegree; with rail/vehicle deployment and mitigated systematics, a 10-point line scan detection at SNR ~3 could be achieved within ~15 min total measurement time.

By suppressing micro-seismic vibration and other dominant noise sources through differential atom interferometry in a robust hourglass gradiometer architecture, the instrument directly addresses the key barrier preventing high-spatial-resolution gravity cartography: long averaging times. The demonstrated 20 E statistical uncertainty in 10 minutes enables submetre-resolution surveys capable of detecting and locating infrastructural voids (e.g., tunnels) even in complex urban environments with confounding signals from buildings and terrain. The Bayesian inference framework translates raw gravity gradient measurements into actionable subsurface information (location, depth, extent, and density), while quantifying uncertainty and highlighting the inherent ambiguity among depth, size, and density typical of potential-field inversions. The results are significant for civil engineering (risk reduction for ground conditions, detection of voids/utilities, sinkhole precursors), archaeology (tombs, chambers, ancient infrastructure), hydrology and environmental monitoring (aquifer mapping, soil moisture), and soil mechanics (compaction mapping). The removal of vibration as a rate-limiting factor allows future sensitivity improvements to translate directly into faster mapping or detection of smaller/deeper targets, further broadening practical impact.

This work demonstrates a practical, field-deployable quantum gravity gradient sensor capable of submetre-resolution gravity cartography. The hourglass cold-atom gradiometer suppresses vibration, laser, thermal, magnetic, and tilt systematics, achieving 20 E statistical uncertainty in 10 minutes and detecting a 2 m tunnel with SNR ~8, with accurate localization and depth estimation via Bayesian inference. The approach offers a new window into the subsurface for applications across engineering, archaeology, and environmental monitoring. Future enhancements—such as large-momentum transfer beam splitters—could deliver a 10–100× sensitivity improvement, enabling faster surveys and detection of smaller/deeper features. Practical instruments with such performance are anticipated within 5–10 years. Further engineering to mitigate residual systematics (e.g., Coriolis) and streamlined deployment (e.g., rail/vehicle mounting) will enhance repeatability and throughput.

• Residual systematic effects (e.g., Coriolis effect) limited survey repeatability in this prototype and require additional mitigation.
• Inference ambiguity between anomaly depth, area, and density is inherent to potential-field data, leading to spread in the POE and parameter posteriors.
• Complex urban environments introduce substantial signals from buildings and local topography that must be modeled; results depend on the accuracy of auxiliary data (CAD plans, GPR, topography).
• Density inference requires a priori knowledge of target geometry and site topography; accuracy depends on prior assumptions.
• Measurement rate (~0.7 Hz) and current sensitivity constrain total survey time; although vibration is suppressed, throughput still depends on operational logistics (station moves, alignment).