Null-Axis Magnetometry

• Null-axis magnetometry is a sensor technique that uses a defined null condition at zero or near-zero magnetic fields to infer external field perturbations with high sensitivity.
• It leverages diverse platforms such as SiC defects, NV centers in diamond, and atomic vapors using methods like optical pumping, microwave polarization, and closed-loop compensation for vector field reconstruction.
• These methods enable low-interference measurements critical for applications in biomagnetism, quantum materials, and dark matter searches, achieving sensitivities ranging from fT/√Hz to µT/√Hz.

Null-axis magnetometry designates a class of magnetic-field sensing methods in which the sensor is operated at zero bias field, near zero field, or at a geometry-defined null response, and the unknown field is inferred from the perturbation of that null condition or from the compensation required to restore it. In the recent literature, this concept spans all-optical vector magnetometry with axial $S=3/2$ color centers in 4H-SiC, zero-bias and near-zero-bias nitrogen-vacancy (NV) protocols in diamond, and atomic magnetometers based on Hanle resonances, parametric resonance, spin alignment, and free-induction decay (Likhachev et al., 2024, Münzhuber et al., 2017, Zheng et al., 2018, Gal et al., 2021, Meraki et al., 2023, Mehta et al., 2024, Dawson et al., 2024). A common motivation is the measurement of magnetic fields without imposing a large external bias field or microwave environment that would interfere with the sample under study, although the underlying mechanisms differ substantially across platforms.

1. Terminology and operating logic

The defining feature of null-axis operation is not a single Hamiltonian or transduction channel, but the imposition of a reference condition in which the sensor response is maximally symmetric, spectroscopically centered, or intentionally suppressed along one axis. In the SiC implementation, the magnetometer is explicitly based on an external magnetic field cancellation scheme that maintains a local region of zero magnetic field at the site of optical excitation of spin centers, so that the level-anticrossing spectrum observed in zero external magnetic field becomes the “reference spectrum” (Likhachev et al., 2024). In single-beam zero-field optically pumped magnetometers, the same logic appears as continuous closed-loop nulling of the static fields along all three axes so that the atoms remain on the peak of the zero-field Hanle resonance (Dawson et al., 2024). In the dark-photon search of NASDUCK′, null-axis refers instead to a geometry-defined null response along one axis, which is used as a noise reference channel rather than as a signal axis (Barir et al., 25 Feb 2026).

The cited literature shows that null-axis magnetometry should not be identified with the absence of internal structure. Zero-field and near-zero-field response is typically made possible by intrinsic fine structure, hyperfine coupling, optical orientation or alignment, polarization-selective excitation, RF dressing, or symmetry-based channel separation. In SiC, fine-structure and hyperfine-structure level anticrossings provide the readout basis; in NV centers, strain, hyperfine coupling, and microwave polarization determine whether first-order sensitivity is recovered at zero field; in atomic magnetometers, the controlled preparation of orientation and alignment determines which field components remain visible at zero field (Likhachev et al., 2024, Zheng et al., 2018, Gal et al., 2021).

2. Solid-state defect realizations

A particularly explicit null-axis protocol was demonstrated for the axial V2 center in 4H-SiC. The center is described by the isotropic spin Hamiltonian

$$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$

with $S=3/2$, $g \simeq 2.00$, $D/h \simeq 35\ \text{MHz}$, and $A/h \simeq 9\ \text{MHz}$ for a neighboring second-shell $^{29}$Si nucleus. In zero external field the fine structure splits the $m_S=\pm 3/2$ and $m_S=\pm 1/2$ manifolds by $2D$, while small longitudinal fields produce two ground-state level anticrossings, LAC1 at $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$0 and LAC2 at $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$1. Hyperfine coupling further splits each LAC into doublets separated by $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$2. Under continuous-wave 785 nm excitation, optical pumping drives the V2 center into the $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$3 subspace, and at the LAC points the mixing of electronic and nuclear eigenstates produces a sharp drop in photoluminescence. The reported sensitivity is $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$4 for the longitudinal $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$5 component in a sensing volume $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$6 at 300 K, with $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$7 for the perpendicular components, and vector magnetometry was stable from 300 K up to at least 500 K (Likhachev et al., 2024).

Münschhuber et al. realized zero-bias vector magnetometry with an ensemble of NV centers in diamond by combining polarization-selective optical excitation with polarization-selective microwave driving. The ground-state Hamiltonian was written as

$$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$8

with $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$9 and $S=3/2$0 on the MHz scale. The anisotropy of the NV electric dipole moments under linearly polarized 532 nm excitation reduces the ambiguity of the optically detected magnetic resonances, while appropriately polarized microwaves suppress transitions from selected crystallographic orientations. The method allows full vector reconstruction of small (less $S=3/2$1) magnetic fields without an external bias field, with a reported sensitivity of $S=3/2$2 for 200 ms integration and $S=3/2$3 frequency resolution (Münzhuber et al., 2017). A different zero-field NV route employs circularly polarized microwaves to excite only one of the two overlapping $S=3/2$4 transitions. In a $S=3/2$5C-enriched HPHT crystal, that restored a first-order $S=3/2$6 dependence and yielded a noise floor of $S=3/2$7, with a photon-shot-noise limited floor of $S=3/2$8 not reached because of laser-intensity and electronic noise (Zheng et al., 2018).

Near-zero-field microwave-free NV magnetometry has also been based on NV–NV cross-relaxation. Dhungel et al. studied zero-field cross-relaxation features in samples with NV concentrations from $S=3/2$9 to $g \simeq 2.00$0. No microwave or radio-frequency fields are applied; the signal is a dip in fluorescence when differently oriented NV ensembles become resonant. For the S2 sample, the photon-shot-noise limit was estimated as $g \simeq 2.00$1, whereas the measured total-noise sensitivity was $g \simeq 2.00$2, with dynamic range extending over at least $g \simeq 2.00$3 around zero (Dhungel et al., 2023). Wang et al. addressed a different near-surface zero-field limitation by exploiting a first-shell $g \simeq 2.00$4C hyperfine coupling of $g \simeq 2.00$5, which acts as an effective built-in bias of $g \simeq 2.00$6. In their shallow-NV scheme, the hyperfine bias breaks the clock-transition condition, suppresses surface-induced charge noise, enhances the dc magnetic sensitivity by a factor of 22, and enables dual-frequency relaxometry at $g \simeq 2.00$7 under zero external field (Wang et al., 2021).

3. Atomic-vapor implementations

In atomic magnetometry, null-axis operation is closely tied to the preparation of atomic orientation, alignment, or both. Le Gal et al. extended the three-step approach for elliptically polarized pumping in metastable $g \simeq 2.00$8He. In their irreducible-tensor description, elliptical pumping simultaneously creates a longitudinal orientation and a transverse alignment, so that Hanle resonances appear for all three components $g \simeq 2.00$9, $D/h \simeq 35\ \text{MHz}$0, and $D/h \simeq 35\ \text{MHz}$1. They then developed parametric-resonance magnetometry with one and two RF fields. In the two-RF configuration, a fast field $D/h \simeq 35\ \text{MHz}$2 and a slow field $D/h \simeq 35\ \text{MHz}$3 produce signals at $D/h \simeq 35\ \text{MHz}$4, $D/h \simeq 35\ \text{MHz}$5, and $D/h \simeq 35\ \text{MHz}$6 that are sensitive to $D/h \simeq 35\ \text{MHz}$7, $D/h \simeq 35\ \text{MHz}$8, and $D/h \simeq 35\ \text{MHz}$9, respectively. Their reported optimal compromise for equal sensitivity to $A/h \simeq 9\ \text{MHz}$0 and $A/h \simeq 9\ \text{MHz}$1 is $A/h \simeq 9\ \text{MHz}$2, and they identified a region in $A/h \simeq 9\ \text{MHz}$3 space where $A/h \simeq 9\ \text{MHz}$4, termed the isotropic condition (Gal et al., 2021).

A purely alignment-based zero-field optical magnetometer was demonstrated in cesium. In that system, linearly polarized resonant light prepares a spin-aligned $A/h \simeq 9\ \text{MHz}$5 ground-state manifold, and the signal is the polarization rotation of the transmitted probe measured by balanced polarimetry. The steady-state alignment yields a first-order response

$A/h \simeq 9\ \text{MHz}$6

so that the detected signal is dispersive in $A/h \simeq 9\ \text{MHz}$7 and, to first order, independent of $A/h \simeq 9\ \text{MHz}$8 and $A/h \simeq 9\ \text{MHz}$9. Nulling is achieved with three orthogonal compensation coils and sequential sweeps of the three axes; after 2–3 iterations the total field reaches $^{29}$0 to $^{29}$1 along all axes. The measured $^{29}$2 bandwidth was $^{29}$3, the noise floor was $^{29}$4 at 1 Hz and $^{29}$5 in the 10–100 Hz band, and the instrument detected a synthetic cardiac signal with effective $^{29}$6 for a single, unaveraged beat (Meraki et al., 2023).

Dead-zone suppression constitutes another major atomic branch of null-axis magnetometry. The Mehta–Samanta–Grewal scheme uses a single laser beam with equal strength of linear- and circular-polarization components and amplitude modulation at a low-duty cycle. In $^{29}$7Rb, this simultaneously pumps first- and second-order harmonics, so that the free-induction-decay signal contains the Larmor component $^{29}$8 and/or the alignment component $^{29}$9 depending on field direction. The reported result is that the amplitude of the FID signal does not go to zero for any magnetic field direction, with sensitivity in the range of $m_S=\pm 3/2$0 in all directions (Mehta et al., 2024). Dawson et al. then generalized zero-field OPM operation to triaxial closed-loop control in rubidium. Their compact single-beam sensor uses dual-axis magnetic modulation while providing static-field information for all three axes, achieving a bandwidth of 380 Hz with sensitivities of $m_S=\pm 3/2$1 across both transverse axes and $m_S=\pm 3/2$2 along the beam axis (Dawson et al., 2024).

4. Reconstruction protocols and nulling procedures

Two reconstruction strategies dominate the literature. The first is direct compensation to a reference null condition. In the SiC protocol, three orthogonal pairs of Helmholtz coils generate compensating fields $m_S=\pm 3/2$3, $m_S=\pm 3/2$4, and $m_S=\pm 3/2$5, with coil currents related to field components by calibration constants $m_S=\pm 3/2$6, $m_S=\pm 3/2$7, and $m_S=\pm 3/2$8. The unknown external field is nulled in three steps: first, $m_S=\pm 3/2$9 is adjusted until LAC1 returns to its reference zero-field position; second, the net perpendicular compensation field $m_S=\pm 1/2$0 is rotated in the $m_S=\pm 1/2$1 plane until the zero-field LAC spectrum exactly reproduces the reference; and third, the compensation currents themselves define the field components: $m_S=\pm 1/2$2 The total magnitude and spherical angles then follow as

$m_S=\pm 1/2$3

$m_S=\pm 1/2$4

The same work also provides spectral landmarks for fine compensation of the perpendicular field: satellite line “5” disappears for $m_S=\pm 1/2$5, $m_S=\pm 1/2$6 passes through unity at $m_S=\pm 1/2$7, the hyperfine doublet in the second LAC vanishes for $m_S=\pm 1/2$8, and all hyperfine structure collapses above $m_S=\pm 1/2$9 (Likhachev et al., 2024).

The second strategy is projection-based vector reconstruction. In the zero-bias NV method of Münschhuber et al., one first isolates an NV orientation $2D$0 by a suitable combination of laser polarization and microwave polarization. The measured resonances satisfy

$2D$1

so that three non-coplanar orientations yield three scalar equations $2D$2, which are assembled as $2D$3 and inverted as $2D$4. The residual signs produce an 8-fold mirror ambiguity, which the authors state can be lifted by continuity arguments, by applying a small known reference field, or by using circularly polarized microwaves to distinguish $2D$5 from $2D$6 (Münzhuber et al., 2017).

Atomic implementations often encode the vector field directly in demodulation channels. In the dressed-atom framework of Le Gal et al., the two-RF configuration yields separate outputs at $2D$7, $2D$8, and $2D$9, sensitive to $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$00, $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$01, and $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$02, respectively (Gal et al., 2021). In Dawson et al., dual transverse modulation produces demodulated signals at $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$03, $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$04, and $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$05, which serve as the error signals for three PID loops and thereby suppress cross-axis projection error in real time (Dawson et al., 2024). The cesium alignment magnetometer instead uses iterative sweep-and-compensate nulling: $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$06 is used to adjust $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$07, then $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$08 adjusts $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$09, and finally $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$10 adjusts $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$11, until the total field is nulled (Meraki et al., 2023).

5. Sensitivity, dynamic range, and failure modes

Performance varies widely with platform and readout physics. Reported figures include $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$12 for the longitudinal component and $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$13 for the perpendicular components in the SiC all-optical vector magnetometer (Likhachev et al., 2024); $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$14 sensitivity with 200 ms integration in the polarization-assisted zero-bias NV vector protocol (Münzhuber et al., 2017); $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$15 in zero-field NV ensemble magnetometry with circularly polarized microwaves (Zheng et al., 2018); $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$16 in microwave-free near-zero-field NV cross-relaxation sensing (Dhungel et al., 2023); $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$17 in the 10–100 Hz band for the cesium alignment sensor (Meraki et al., 2023); $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$18 in all directions for the dead-zone-free rubidium FID magnetometer (Mehta et al., 2024); and $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$19 across the transverse axes with $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$20 along the beam axis for the triaxial rubidium OPM (Dawson et al., 2024).

Zero-field operation does not, by itself, guarantee isotropy, linearity, or immunity to technical noise. The NV zero-bias vector protocol has a low-field limit set by strain, $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$21, estimated as $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$22 for $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$23, together with limitations from resonance broadening in large gradients, sign ambiguities, and the need to maintain polarization purity (Münzhuber et al., 2017). The zero-field NV ensemble magnetometer of the $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$24C-enriched HPHT study was limited by laser-intensity and electronic noise rather than by its photon-shot-noise floor (Zheng et al., 2018), while in the cross-relaxation scheme technical noise dominated over the shot-noise estimate (Dhungel et al., 2023). Atomic devices face analogous anisotropy problems: purely circular or purely linear pumping produces dead zones in specific field directions, and single-axis zero-field OPMs are vulnerable to cross-axis projection error unless all axes are sensed and actively nulled (Gal et al., 2021, Mehta et al., 2024, Dawson et al., 2024).

A related practical issue appears in NV Ramsey magnetometry with misaligned bias fields. Oon et al. showed that in $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$25NV centers the absence of a quadrupole moment in the $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$26N nuclear spin leads to pronounced envelope modulation in Ramsey measurements under a misaligned bias field, with significant sensitivity loss if unaddressed. They further showed that double-quantum coherences dramatically suppress these envelope modulations and are resilient to strain heterogeneity and temperature shifts; the paper states that this simplifies zero-field and vector magnetometry configurations where the bias field must bisect NV axes (Oon et al., 2022). This suggests that, in practice, null-axis operation often depends as much on suppressing parasitic couplings and ambiguities as on centering the external field at zero.

6. Applications, scope, and broader significance

The principal application domain is measurement in environments where an external bias field or microwave drive would perturb the target. The zero-field NV ensemble work explicitly identifies zero- and ultralow-field NMR, magnetically shielded-environment sensing, biomagnetism, and searches for exotic physics (Zheng et al., 2018). The microwave-free NV cross-relaxation study points to high-$$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$27 superconductors, zero–ultra-low field NMR, magnetic materials near critical points, living biological tissue, and quantum materials such as spin ices and 2D magnets (Dhungel et al., 2023). The hyperfine-biased shallow-NV method was proposed as a route toward single-molecule magnetic resonance spectroscopy and nanoscale MRI under zero external field (Wang et al., 2021). In atomic systems, the cesium spin-alignment magnetometer demonstrated a synthetic cardiac signal and reported sensitivity and bandwidth consistent with basic magnetocardiography requirements in a shielded room (Meraki et al., 2023).

Null-axis methods have also been extended beyond conventional sensing. NASDUCK′ searched for dark-photon dark matter in the mass range $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$28 (1–500 kHz) using a three-axis magnetometer in a large conductive shielded room. There, a geometry-defined null response along one axis served as a noise reference, and a subtraction procedure reduced the noise floor and improved sensitivity. The reported result was new laboratory limits on the kinetic-mixing parameter $$H = g\mu_B\,\mathbf S\cdot\mathbf B + D\left[S_z^2 - S(S+1)/3\right] + A\,\mathbf I\cdot\mathbf S,$$29, improving previous laboratory bounds by up to three orders of magnitude (Barir et al., 25 Feb 2026).

Taken together, these studies indicate that null-axis magnetometry is best understood as a control-and-readout strategy rather than as a single sensor modality. Its implementations include restoration of a reference spectrum, inversion of zero-bias projections, balanced use of orientation and alignment, dead-zone elimination by mixed multipoles, triaxial closed-loop cancellation, and geometry-defined null channels. The unifying principle is the use of a zero-field or null-response operating point as the metrological reference, with vector information obtained from the way internal spin structure, optical selection rules, or calibrated compensation currents depart from and return to that point.