Quantum Self-Calibrating Force Sensor

• The sensor integrates a quantum or SI-traceable reference for in situ calibration, establishing an absolute force scale during operation.
• It employs techniques such as atom interferometry, optomechanics, and NV-center mapping to convert measured signals into calibrated force gradients.
• Field and laboratory implementations demonstrate enhanced precision, portability, and real-time correction of offsets, drifts, and systematic errors.

A self-calibrating quantum force-gradient sensor is a sensor architecture in which the transduction between an external force field and the reported observable is referenced in situ to a quantum or SI-traceable standard, so that offsets, drift, scale factors, or transfer functions are determined during operation rather than by a separate ex situ calibration. In current implementations, this role is played by an absolute atom gravimeter that nightly calibrates mobile spring gravimeters in a $24\ \mathrm{km}^2$ field survey, radiation pressure that provides a traceable reference force for an optomechanical AFM sensor, Bloch-frequency metrology in a trapped-atom optical lattice, a magnetically insensitive interferometer channel in a spinor-BEC gravimeter and magnetic gradiometer, direct NV-center mapping of an MFM tip stray field, or bias-field-controlled gap tuning in a Meissner-levitated microsphere platform [(Shettell et al., 11 Feb 2026); (Melcher et al., 2014); (Balland et al., 2023); (Hardman et al., 2016); (Sakar et al., 2021); (Ren et al., 14 Feb 2026)].

1. Defining concept and scope

The defining feature of self-calibration is that the sensor contains, or is continuously tied to, a reference quantity whose value is set by a quantum phase relation, a fundamental constant, or an SI-traceable actuator. In atom-interferometric gravimetry, the absolute reference is the interferometer phase itself, with $\Delta \phi = k_{\mathrm{eff}}\cdot g \cdot T^2$. In trapped-atom lattice sensing, the force is obtained directly from the Bloch frequency through $F = h\nu_B/d$. In optomechanical force sensing, radiation pressure generates the reference force. In quantum-calibrated magnetic force microscopy, the calibration input is the measured tip stray field of a single NV center rather than an assumed tip model [(Shettell et al., 11 Feb 2026); (Balland et al., 2023); (Melcher et al., 2014); (Sakar et al., 2021)].

Within this class, “force-gradient” does not denote a single readout topology. Some devices measure gradients directly through differential phase or frequency shifts. Others first produce an absolutely referenced scalar force or acceleration map and then compute gradients by spatial differencing. The field-deployable hybrid gravimetry study is of the latter type: it uses a single atomic gravimeter as a drift-free backbone and derives gradients from spatial differences of calibrated gravity values. Dedicated atom gradiometers, by contrast, directly encode the gradient in a differential phase such as $$\Delta \phi \approx k_{\mathrm{eff}}\cdot \Gamma \cdot L \cdot T^2$$ (Shettell et al., 11 Feb 2026).

A related point is that self-calibration is not synonymous with model-free sensing. Each implementation retains a forward model: mechanical susceptibility for optomechanical sensors, Wannier–Stark or Ramsey phase models for trapped atoms, magnetostatic reciprocity for MFM, Breit–Rabi response for spinor interferometry, or levitation-potential models for Meissner platforms. Self-calibration shifts the most error-prone part of the metrology chain from external standards or phenomenological assumptions toward observables that are measured locally and repeatedly during the experiment [(Melcher et al., 2014); (Hardman et al., 2016); (Sakar et al., 2021); (Ren et al., 14 Feb 2026)].

2. Physical principles of self-calibration

The underlying principle is a closed calibration loop between a measured frequency or phase and the target force observable. In the optomechanical AFM sensor, the mechanical element is treated as a single-degree-of-freedom harmonic oscillator with susceptibility

$$\chi(\omega) = \frac{1}{k - m\omega^2 + i m\gamma\omega}, \qquad \gamma = \omega_0/Q.$$

Radiation pressure provides the reference actuation,

$F_{\mathrm{rad}} = \frac{2RP}{c}\cos\theta,$

and steady-state ring-up at resonance yields the stiffness

$k = \frac{2PRQ}{cA}.$

Because $P$ is traceable to a calibrated power meter, the stiffness and hence the force-gradient scale in FM-AFM become absolute rather than empirical. The same device uses

$$\Delta f = -\frac{f_0}{2k}\,\frac{\partial F}{\partial z}$$

to convert frequency shift into force gradient (Melcher et al., 2014).

In trapped-atom interferometry near a surface, the calibration is established by the optical lattice spacing $d=\lambda_l/2$ and the Bloch relation

$\Delta \phi = k_{\mathrm{eff}}\cdot g \cdot T^2$0

The force follows from

$\Delta \phi = k_{\mathrm{eff}}\cdot g \cdot T^2$1

while local gradients can be obtained as

$\Delta \phi = k_{\mathrm{eff}}\cdot g \cdot T^2$2

Because the force scale is tied to the lattice wavelength and Planck’s constant, no mechanical spring constant enters the calibration. The $\Delta \phi = k_{\mathrm{eff}}\cdot g \cdot T^2$3 Ramsey protocol removes dependence on the hyperfine clock frequency and its light shifts, leaving $\Delta \phi = k_{\mathrm{eff}}\cdot g \cdot T^2$4 as the quantity encoding the external force (Balland et al., 2023).

In multi-state atom interferometry with a spinor BEC, self-calibration takes a different form. Three interferometers run simultaneously on the $\Delta \phi = k_{\mathrm{eff}}\cdot g \cdot T^2$5 Zeeman sublevels. The $\Delta \phi = k_{\mathrm{eff}}\cdot g \cdot T^2$6 channel is magnetically insensitive to first order and provides the inertial scale factor

$\Delta \phi = k_{\mathrm{eff}}\cdot g \cdot T^2$7

The differential phases of the $\Delta \phi = k_{\mathrm{eff}}\cdot g \cdot T^2$8 channels then isolate the magnetic-field gradient without requiring an external calibration of the scale factor, because all three channels share the same $\Delta \phi = k_{\mathrm{eff}}\cdot g \cdot T^2$9 and $F = h\nu_B/d$0 (Hardman et al., 2016).

In quantum-calibrated MFM, the relevant principle is reciprocity rather than a direct force standard. The measured MFM observable is linked to the sample stray field through a transfer function constructed from the NV-measured tip field. For FM-AFM,

$F = h\nu_B/d$1

and the tip transfer function is computed from the directly measured $F = h\nu_B/d$2 rather than from monopole or dipole assumptions. This replaces a model-dependent tip calibration with a quantum-measured transfer kernel (Sakar et al., 2021).

3. Major architectural realizations

A field-deployable gravimetric realization uses a compact atomic gravimeter as an on-site absolute reference and two Scintrex CG6 Autograv spring gravimeters as mobile heads. The spring gravimeters are returned each evening to a fixed location adjacent to the atomic container, where all instruments record one-minute data under common conditions and the nightly offset and linear drift are fitted against the tide-corrected atomic time series. The method yields asynchronous cross-comparison of relative measurements acquired on different days by different instruments (Shettell et al., 11 Feb 2026).

An optomechanical realization employs a laser-machined fused-silica parallelogram flexure with an integrated proof mass and two fiber cavities: an actuation cavity for radiation-pressure drive and a Fabry–Perot readout cavity for interferometric displacement measurement. At room temperature and high vacuum at $F = h\nu_B/d$3, the device reached $F = h\nu_B/d$4, stiffness near $F = h\nu_B/d$5 from two independent calibrations, and force resolution of approximately $F = h\nu_B/d$6 (Melcher et al., 2014).

A local trapped-atom realization uses laser-cooled $F = h\nu_B/d$7 atoms in a shallow, blue-detuned vertical optical lattice retroreflected from a silica dielectric mirror. The lattice spacing is $F = h\nu_B/d$8, the vertical cloud radius after transport is $F = h\nu_B/d$9, and the atom–surface distance is calibrated by controlled contact with the surface and atom-loss fitting. Measurements are performed over $$\Delta \phi \approx k_{\mathrm{eff}}\cdot \Gamma \cdot L \cdot T^2$$0–$$\Delta \phi \approx k_{\mathrm{eff}}\cdot \Gamma \cdot L \cdot T^2$$1, enabling local force and gradient measurements without mechanical calibration (Balland et al., 2023).

A simultaneous gravimetric and magnetic-gradiometric realization uses a Bose–Einstein condensate of $$\Delta \phi \approx k_{\mathrm{eff}}\cdot \Gamma \cdot L \cdot T^2$$2 in a vertical Mach–Zehnder interferometer based on Bragg transitions. Because the Bragg pulses act identically on the three Zeeman states, the $$\Delta \phi \approx k_{\mathrm{eff}}\cdot \Gamma \cdot L \cdot T^2$$3 channel calibrates the inertial response while the $$\Delta \phi \approx k_{\mathrm{eff}}\cdot \Gamma \cdot L \cdot T^2$$4 channels yield the magnetic gradient. The reported asymptotic gravimetric precision was $$\Delta \phi \approx k_{\mathrm{eff}}\cdot \Gamma \cdot L \cdot T^2$$5, and the magnetic-field gradient precision was $$\Delta \phi \approx k_{\mathrm{eff}}\cdot \Gamma \cdot L \cdot T^2$$6 (Hardman et al., 2016).

A quantum-calibrated magnetic-force realization uses a single NV center in diamond to map the two-dimensional stray field distribution of an MFM tip at a fixed standoff of $$\Delta \phi \approx k_{\mathrm{eff}}\cdot \Gamma \cdot L \cdot T^2$$7. From the measured tip field and the cantilever parameters, an instrument calibration function is derived, allowing raw MFM phase images to be converted into quantum-calibrated stray-field maps. The method quantitatively recovered sample stray fields up to approximately $$\Delta \phi \approx k_{\mathrm{eff}}\cdot \Gamma \cdot L \cdot T^2$$8 (Sakar et al., 2021).

A cryogenic near-field realization has been proposed in which a ferromagnetic microsphere is Meissner-levitated above a type-I superconducting Pb plane and read out by a SQUID-coupled flux-tunable microwave resonator. A uniform bias field $$\Delta \phi \approx k_{\mathrm{eff}}\cdot \Gamma \cdot L \cdot T^2$$9 reproducibly tunes the equilibrium gap without mechanical approach, and the force gradient is encoded as a resonance-frequency shift tracked by a PLL. The proposal projects force sensitivities of $$\chi(\omega) = \frac{1}{k - m\omega^2 + i m\gamma\omega}, \qquad \gamma = \omega_0/Q.$$0 at millikelvin temperatures and gives the SQL-related scaling $$\chi(\omega) = \frac{1}{k - m\omega^2 + i m\gamma\omega}, \qquad \gamma = \omega_0/Q.$$1 (Ren et al., 14 Feb 2026).

4. Field-deployable hybrid gravimetry as a self-calibrating force-gradient sensor

The most explicit large-area field realization is the hybrid quantum-enabled gravimetry deployment over remote, densely forested tropical terrain on an island off Singapore. The survey covered $$\chi(\omega) = \frac{1}{k - m\omega^2 + i m\gamma\omega}, \qquad \gamma = \omega_0/Q.$$2 over seven days of field measurement from 6–12 October 2023, after a setup phase from 2–5 October. Station spacing was at least $$\chi(\omega) = \frac{1}{k - m\omega^2 + i m\gamma\omega}, \qquad \gamma = \omega_0/Q.$$3 along predefined survey axes, and the atomic gravimeter operated continuously in an air-conditioned cargo container on mechanically stable, level ground, with a generator providing continuous power. The container footprint was approximately $$\chi(\omega) = \frac{1}{k - m\omega^2 + i m\gamma\omega}, \qquad \gamma = \omega_0/Q.$$4 (Shettell et al., 11 Feb 2026).

The mobile instruments were leveled to within $$\chi(\omega) = \frac{1}{k - m\omega^2 + i m\gamma\omega}, \qquad \gamma = \omega_0/Q.$$5 arcsec at each station, observed for about $$\chi(\omega) = \frac{1}{k - m\omega^2 + i m\gamma\omega}, \qquad \gamma = \omega_0/Q.$$6 for stability, and repositioned if drift indicated soft or uneven ground. Each station occupation comprised five one-minute samples over five minutes. During overnight periods of approximately 18:00–09:00, both relative gravimeters and the atomic gravimeter recorded continuously at one-minute cadence in a common enclosure for calibration. The field model for a spring gravimeter was

$$\chi(\omega) = \frac{1}{k - m\omega^2 + i m\gamma\omega}, \qquad \gamma = \omega_0/Q.$$7

with nightly linear drift

$$\chi(\omega) = \frac{1}{k - m\omega^2 + i m\gamma\omega}, \qquad \gamma = \omega_0/Q.$$8

Elevation corrections were written as

$$\chi(\omega) = \frac{1}{k - m\omega^2 + i m\gamma\omega}, \qquad \gamma = \omega_0/Q.$$9

with

$F_{\mathrm{rad}} = \frac{2RP}{c}\cos\theta,$0

while terrain correction was neglected because of limited topographic data and low local relief (Shettell et al., 11 Feb 2026).

The atomic reference attained a minimum Allan deviation of approximately $F_{\mathrm{rad}} = \frac{2RP}{c}\cos\theta,$1 in the field, compared with approximately $F_{\mathrm{rad}} = \frac{2RP}{c}\cos\theta,$2 in the laboratory, and the absolute gravity after tide correction showed a slow increase of about $F_{\mathrm{rad}} = \frac{2RP}{c}\cos\theta,$3 over the seven-day campaign, plausibly due to ocean loading. Nightly calibration transferred this evolving baseline into the relative instruments’ corrections. The resulting map, referenced to the atomic gravimeter site, spanned approximately $F_{\mathrm{rad}} = \frac{2RP}{c}\cos\theta,$4 to $F_{\mathrm{rad}} = \frac{2RP}{c}\cos\theta,$5 and revealed a coherent northeast–southwest regional gradient. A repeated path segment measured on different days by both instruments showed temporal consistency, demonstrating asynchronous cross-comparison capability. The campaign comprised 262 field gravity measurements; 59 stations achieved fully converged PPK solutions with $F_{\mathrm{rad}} = \frac{2RP}{c}\cos\theta,$6 vertical uncertainty, corresponding to roughly $F_{\mathrm{rad}} = \frac{2RP}{c}\cos\theta,$7 gravity uncertainty from the free-air correction, and 127 stations had partially converged PPK solutions that were retained for broader trend mapping (Shettell et al., 11 Feb 2026).

In this architecture, the “self-calibrating” property resides not in direct quantum gradiometry but in the daily renewal of offset and drift estimates for each mobile head through the absolute atom-interferometric reference. The gradients are then inferred from spatial differences of calibrated, elevation-corrected $F_{\mathrm{rad}} = \frac{2RP}{c}\cos\theta,$8. This arrangement trades direct differential readout for portability, asynchronous network adjustment, and compatibility with remote surveys (Shettell et al., 11 Feb 2026).

5. Noise sources, systematic limitations, and common misconceptions

A recurrent misconception is that self-calibration eliminates the dominant systematics. The cited implementations do not support that interpretation. In field gravimetry, accuracy still depended on nightly calibration windows, dense-canopy GPS elevation control, and the omission of terrain correction. Residual deviations between neighboring stations after correction were typically $F_{\mathrm{rad}} = \frac{2RP}{c}\cos\theta,$9–$k = \frac{2PRQ}{cA}.$0, attributed largely to imperfect elevation control, minor site effects, and unmodeled environmental factors (Shettell et al., 11 Feb 2026).

In trapped-atom surface-force sensing, the principal systematic was not the force scale but an attractive electrostatic force from Rb atoms adsorbed on the dielectric. The measured force was stronger and longer-ranged than the Casimir–Polder prediction until a model of stray patch-dipole fields was fitted and subtracted. The short-term force sensitivity was $k = \frac{2PRQ}{cA}.$1 at $k = \frac{2PRQ}{cA}.$2, the long-term stability reached $k = \frac{2PRQ}{cA}.$3 after about $k = \frac{2PRQ}{cA}.$4, and dominant per-shot Bloch-frequency noise terms included approximately $k = \frac{2PRQ}{cA}.$5 from vibrations before correction, approximately $k = \frac{2PRQ}{cA}.$6 from quantum projection noise for $k = \frac{2PRQ}{cA}.$7 atoms, and approximately $k = \frac{2PRQ}{cA}.$8 from detection background noise. Distance uncertainty was also strongly amplified because in the retarded regime $k = \frac{2PRQ}{cA}.$9 and $P$0 (Balland et al., 2023).

In quantum-calibrated MFM, the combined expanded uncertainty for fields around $P$1 was approximately $P$2, dominated by the cantilever spring constant uncertainty of about $P$3 and the quality-factor uncertainty of about $P$4. The NV-based field uncertainty was negligible in that budget. This is a clear example of self-calibration improving the magnetic transfer function while leaving mechanical metrology as the dominant error source (Sakar et al., 2021).

In the optomechanical AFM sensor, the dominant uncertainty contributor was the optical power calibration: $P$5 on $P$6 RMS, or approximately $P$7 relative uncertainty. The added-mass and radiation-pressure methods agreed within about $P$8, but the device remained thermally dominated at $P$9 rather than quantum-backaction limited (Melcher et al., 2014).

The proposed Meissner-levitated microsphere sensor makes the same point in a prospective setting. Its self-calibration relies on reproducible gap tuning and background subtraction, but the performance is explicitly limited by a trade-off between suppression of electrostatic patch potentials through an Au coating and increased eddy-current dissipation. In the model, the Au layer raises the damping rate and therefore the thermal force-noise PSD $$\Delta f = -\frac{f_0}{2k}\,\frac{\partial F}{\partial z}$$0 (Ren et al., 14 Feb 2026).

6. Development trajectories and broader significance

Several development paths recur across the literature. In field gravimetry, proposed improvements include continuous on-site atomic referencing with real-time offset and drift correction via least-squares or Kalman filters, improved elevation control through multi-constellation GNSS and lidar-derived high-resolution DEMs, autonomous tilt control, and deployment of dual-sensor baselines for direct gradient sensing while retaining the atomic base for absolute traceability (Shettell et al., 11 Feb 2026).

In local quantum sensing near surfaces, the trapped-atom platform suggests improvements through superlattices or magnetic gradients for site selectivity and through suppression of stray electrostatics by surface heating to desorb atoms. The same work indicates that the combination of micrometer-scale locality, sub-micrometer distance control, and force calibration via $$\Delta f = -\frac{f_0}{2k}\,\frac{\partial F}{\partial z}$$1 is suitable for precision tests of short-range QED and gravity (Balland et al., 2023).

In optomechanics and cryogenic levitation, the dominant direction is toward quantum-limited readout. The optomechanical AFM work states that higher cavity finesse, lower absorption, low-noise lasers, and cryogenic operation would move the device toward quantum-limited force-gradient sensing. The Meissner-levitated microsphere proposal provides the most explicit SQL analysis and identifies a “mass-assisted” scaling in which larger microspheres require fewer photons to reach the SQL because displacement-to-flux transduction increases rapidly with size [(Melcher et al., 2014); (Ren et al., 14 Feb 2026)].

A separate trajectory concerns portability of calibration itself. Quantum-calibrated MFM introduced a transfer-standard concept in which a quantum-calibrated stray-field map of a Co/Pt multilayer can be distributed and used to calibrate other MFMs without repeating the NV-based tip characterization. This suggests a broader metrological pattern: self-calibration can be either continuously local, as in atomic gravimetry or Bloch-frequency sensing, or encoded into a reusable calibration artifact that preserves traceability to the original quantum reference (Sakar et al., 2021).

Taken together, these results show that a self-calibrating quantum force-gradient sensor is less a single instrument than a design principle. The common structure is a quantum or SI-traceable reference embedded in the sensing loop, an explicit model linking that reference to the force or gradient observable, and a repeated in situ procedure that updates the calibration under the same environmental conditions as the measurement. The practical consequence is not the removal of all systematics, but the relocation of the calibration burden toward observables that can be measured locally, repeatedly, and with traceable uncertainty [(Shettell et al., 11 Feb 2026); (Melcher et al., 2014); (Balland et al., 2023); (Sakar et al., 2021); (Ren et al., 14 Feb 2026)].