Microwave-free vector magnetometry and crystal orientation determination with nitrogen-vacancy centers using Bayesian inference
Paper details
Section titled “Paper details”| Field | Value |
|---|---|
| Journal | Communications Physics |
| Published | 2026-07-06 |
| Authors | Hilario Espinós, Omkar Dhungel, Arne Wickenbrock, Dmitry Budker, Ricardo Puebla |
| DOI | 10.1038/s42005-026-02736-y |
| Citations | 1 |
Bayesian inversion of a microwave-free photoluminescence map reconstructed the laboratory-z component of an external magnetic field as 1.165 ± 0.002 mT. The same method recovered the diamond orientation from noisy optical data without prior crystallographic alignment or microwave excitation. The authors present the work as proof of concept for near-zero-field vector magnetometry.
Cross-relaxation encodes field direction
Section titled “Cross-relaxation encodes field direction”The negatively charged NV center is an electronic spin-1 system with zero-field splitting D = 2π × 2.87 GHz. The photoluminescence dips arise from cross-relaxation between NV centers with different crystallographic orientations. For an axis n_i, the spin spectrum depends on B_parallel = B · n_i and the transverse field, but not the transverse azimuth. When |B · n_i| = |B · n_j|, transition energies coincide under field reversal. Dipole-coupled flip-flop processes then redistribute population from the bright m_s = 0 state into the darker m_s = ±1 manifold and lower PL.
The tetrahedral NV geometry converts these conditions into nine planes in magnetic-field space. Six are symmetry planes for one orientation pair. Three are antisymmetry planes where two pairs meet. The ideal model gives an antisymmetry crossing twice the contrast weight of a single crossing. The corresponding weights are 1 and 2.
Scanning bias-field strength and sample rotation produces a two-dimensional PL map with dips at these conditions. The Bayesian method fits the full map rather than tracking isolated resonance peaks.
Bayesian inversion uses the full PL map
Section titled “Bayesian inversion uses the full PL map”The forward model is PL_model = 1 − C Σ_i w_i L(δ_i; Γ). Here C sets contrast, Γ is the resonance linewidth, w_i weights each resonance, and δ_i is a linear deviation from its condition. The parameter set contains the external field, crystal orientation, Γ, and w_i. Linewidth and weights can come from independent calibration or enter the inference as unknowns.
The likelihood treats measurement noise as independent and approximately Gaussian. It propagates uncertainty in bias field and rotation angle into the total variance. A closed-form resonance model avoids repeated Hamiltonian diagonalization during likelihood evaluation. This reduces the computation needed for Bayesian sampling.
The posterior retains discrete NV symmetries as separate high-probability regions. It reports uncertainty and ambiguity rather than forcing one fitted orientation or field solution.
Experimental conditions
Section titled “Experimental conditions”Wide-field fluorescence microscopy recorded PL through a photodiode after 515 nm excitation.
| Item | Experimental condition |
|---|---|
| Excitation | 515 nm Toptica iBeam smart laser |
| Optics | 50× Mitutoyo long-working-distance objective, NA = 0.45 |
| Detector | Photodiode through the same objective |
| Diamond | (111)-oriented HPHT-grown single-crystal diamond |
| Nitrogen and NV content | Estimated substitutional nitrogen of ~100 ppm; estimated NV concentration >5 ppm |
| Processing | 3 MeV electron irradiation at 10^18 cm^-2; annealing at 1050 °C for 2 h |
Helmholtz coils generated an arbitrarily directed field. ODMR calibrated the coils, but the authors did not require it for the microwave-free sensing method.
For orientation determination, a 1 mT transverse field rotated in the xy plane while the axial field varied. The proof of concept rotated the field rather than the diamond. The authors state that these protocols are equivalent at the model level.
The map contained 154 bias-field values and 72 rotation angles. The likelihood used σ_noise = 0.0018, σ_Bbias = 1 µT, and σ_φ = 1°. The method also applies to CVD-grown and arbitrarily cut diamonds when cross-relaxation features remain measurable. Such features have been observed in ensemble samples down to ~0.3 ppm NV density.
Crystal orientation recovery
Section titled “Crystal orientation recovery”Orientation was parameterized by α about laboratory z, β about the diamond [−1 1 0] axis, and ζ about [111]. The search used α in [0, 2π), β in [0, arccos(1/√3)], and ζ in [0, 2π/3). These limits reduce the 24 equivalent configurations to 8.
The posterior contained two well-localized maxima related by inversion symmetry.
| Angle | First solution | Inversion-related solution |
|---|---|---|
| α | 4.7587 ± 0.0008 rad | 0.6832 ± 0.0009 rad |
| β | 0.2342 ± 0.0003 rad | 0.2705 ± 0.0003 rad |
| ζ | 0.4775 ± 0.0007 rad | 1.6452 ± 0.0007 rad |
Both parameter sets reproduced the measured PL pattern. Orientation precision was limited mainly by systematic effects rather than posterior statistical precision.
External-field reconstruction
Section titled “External-field reconstruction”With the crystal orientation fixed, the same map was inverted for the external field. The field was parameterized by a transverse magnitude, azimuth, and longitudinal component. Maximum a posteriori estimates were b_z = 1.165 ± 0.002 mT, b_perp = 0.8091 ± 0.0009 mT, and φ_0 = 0.7196 ± 0.0008 rad. These estimates were compatible with independent ODMR measurements.
The choice of rotation axis determined whether the map was unique. A separate symmetry simulation fixed b_z = 1 mT, b_perp = 2 mT, and φ_0 = 0. With laboratory z aligned to [100], the map repeated every π/2 and had mirror symmetries every π/4. Alignment to [110] produced π periodicity and mirror symmetries every π/2. Alignment to [111] produced 2π/3 periodicity and mirror symmetries every π/3. An arbitrary [123] alignment produced no discernible symmetry. The reconstruction experiment used an axis outside these high-symmetry directions, which allowed a unique field vector from one configuration.
Precision, ambiguities, and device design
Section titled “Precision, ambiguities, and device design”The likelihood value σ_noise = 0.0018 represented the standard deviation of model residuals. Measured shot noise was one order lower, so systematic effects dominated the residuals. A trace at φ = 335° reproduced the main dip positions but differed in feature contrast. PL amplitude deviations were of order 10^-3 to 10^-2. A weak shoulder could arise from hyperfine coupling to naturally abundant 13C nuclei.
Linearly polarized excitation gave unequal PL contrast across NV families. Angle-dependent illumination added nonmagnetic modulation, while mechanical shifts changed the field seen by the ensemble. The authors suggest circular polarization to reduce unequal excitation, with isotropic excitation needed for complete averaging. They also suggest synchronizing laser and sample rotation and correcting mechanical misalignment. Temperature and laser power can affect linewidth and contrast, while uniform crossing shifts have little first-order effect. Isotopically enriched 12C would remove the additional spectral structure. Strain or temperature gradients, density variations, and bias-field nonuniformity broaden the effective linewidth.
Joint orientation and field inference remains ambiguous under a global rotation about laboratory z. A second rotation axis or independent reference field can remove that ambiguity. Joint inference also becomes more costly as the number of free parameters grows. Posterior width for b_perp decreased approximately as 1/√N across 1000 repetitions. With orientation known, magnetometry was possible from one PL trace, but orientation required several angles. The MATLAB implementation used 8-core grid evaluation and took about 10 s for a 100 × 100 × 100 grid. Single-trace acquisition took seconds, and compiled or GPU implementations were expected to improve speed. Supplementary Note 3 contains the practical dynamic range and magnetically induced state-mixing analysis.
Conventional ODMR uses microwaves that can heat the diamond, create stray fields, and require electrical connections. Prior cross-relaxation magnetometry had projected sensitivities in the tens of pT/√Hz but required prior crystallographic knowledge and precise field alignment. Earlier 2D PL work characterized control-parameter maps without solving this inverse problem. The posterior preserves discrete NV symmetry ambiguities instead of forcing one physical solution. No measured sensitivity was reported for the new instrument.
For device engineering, the proposed magnetometer removes microwave delivery to the NV ensemble, but it still requires a tunable bias axis and controlled rotation. It also needs a homogeneous sensing volume, controlled excitation, and an axis outside the NV symmetry directions. The supplied paper is an unedited article-in-press manuscript and warns that errors may remain.
Materials and support
Section titled “Materials and support”The paper used HPHT diamond, but the method also applies to CVD-grown diamond when cross-relaxation features are measurable. 6CCVD’s diamond materials catalog includes MPCVD-grown single-crystal plates and wafers up to 125 mm, which provide a candidate material family for a microwave-free NV magnetometer build. NV formation and sensing-volume uniformity must be specified separately, and engineering support can help define those requirements.
View Original Abstract
Abstract Nitrogen-vacancy (NV) centers in diamond provide a solid-state platform for quantum sensing. While optically detected magnetic resonance techniques offer high sensitivity, their reliance on microwaves introduces heating and stray electromagnetic fields that can perturb nearby samples. Optical approaches based on cross-relaxation between differently oriented NV centers remove this constraint but have so far required stringent alignment of the external field with crystallographic axes, restricting their practicality. Here we introduce a general framework for microwave-free vector magnetometry at near-zero field that leverages Bayesian inference to extract both the magnetic field vector and the NV orientation directly from photoluminescence maps. An analytical model of cross-relaxation resonances enables efficient inference under arbitrary field and orientation configurations, while naturally incorporating the discrete degeneracies of the NV symmetry. We experimentally demonstrate robust orientation determination and vector-field reconstruction, establishing a general route toward compact and alignment-free NV magnetometers for practical sensing applications.
Single Crystal Cvd Diamond Plate
View Specs →
Electronic Grade Intrinsic Diamond Epitaxial Wafer H Terminated Superior 100 Mosfet Ready Substrate
View Specs →