{"id":"58b35b92-44ef-4ea7-9787-4d38f321fd04","arxiv_id":"2512.13835","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A Bayesian analysis of cross-relaxation photoluminescence maps determines NV-crystal orientation and magnetic-field vectors without microwave driving or field alignment.","lead":"This paper shows how to measure both the direction of a magnetic field and the orientation of a diamond crystal without using microwaves, by analyzing the crystal's glow with Bayesian statistics. If it works as claimed, it could make smaller, simpler quantum magnetic sensors for uses where microwaves are disruptive.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Uncertainty quantification is circular: sigma_noise is fixed to the residual of the same model-data fit, so reported posterior widths do not capture the acknowledged 10^-3-10^-2 systematic PL mismatches; the ODMR cross-check is only 'compatible' with no quantitative comparison.","rationale":"Reading in good faith: the paper demonstrates a genuinely useful extension of cross-relaxation magnetometry and provides a real experimental validation with an ODMR cross-check. The analytical resonance-plane model is elegant, and the symmetry reduction is well motivated. However, the most load-bearing assumption is that Eq. (8) is accurate enough that the Bayesian posterior widths are meaningful. The paper's own systematic-error analysis shows residuals an order of magnitude above shot noise, and sigma_noise is set to those same residuals, creating a circular calibration. This does not necessarily invalidate the point estimates (the ODMR agreement is encouraging), but it does undermine the advertised uncertainty quantification. The reader's weakest_assumption identifies exactly this model specification issue; my concern sharpens it by focusing on the consequence for the central claim. The proposed coverage test would settle whether the reported credible intervals are trustworthy. No independent verification of the orientation estimate exists, so the end-to-end field comparison to ODMR is the only external anchor; making that comparison quantitative is essential. The verdict remains CONDITIONAL: the method is plausible and partially validated, but the model form and uncertainty calibration must be specified and tested before the central claims can be fully accepted.","tokens_in":13956,"tokens_out":4928,"duration_ms":48244,"concrete_test":"Acquire a validation set of at least 10 PL maps at known fields set by calibrated currents/ODMR, with the same crystal orientation. For each map, run the published inference and record whether the ODMR value lies inside the 95% credible interval (using the reported posterior covariance). If coverage is significantly below 95% (e.g., fewer than 8 of 10), the sigma_noise-residual calibration understates systematic error and the uncertainty claims require revision. Also report the functional form and fitted values of L, C, Gamma, and w_i used to generate the simulated maps.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that the posterior distributions in Figs. 3(c) and 5(b) faithfully quantify uncertainty. This fails if the likelihood model is misspecified. Equation (8) postulates PL = 1 - C * sum_i w_i L(delta_i; Gamma), with L, C, and Gamma never specified (neither functional form nor fitted values). In the experiments, sigma_noise is set to the standard deviation of the residuals between this model and the very data being fit (Section 'Analysis of systematic errors'). Because the residuals are dominated by known systematic effects--polarization-dependent excitation, inhomogeneous illumination, misalignment, and an unmodeled shoulder attributed to 13C hyperfine structure--they are spatially correlated and not i.i.d. Gaussian. Setting sigma to their RMS forces chi^2/dof ~ 1 by construction, but does not propagate the systematic error into the posterior. The quoted uncertainties (e.g., b_z = 1.165(2) mT, sub-mrad angles) are therefore effective fit precisions, not true accuracies. The orientation estimate has no independent validation, and the field check against ODMR is described only as 'compatible' without giving ODMR error bars or a statistical distance; a bias of several times the reported 2 uT would still be 'compatible' in a loose sense. Since both the orientation and magnetometry demonstrations depend entirely on Eq. (8), the load-bearing assumption is that the additive fixed-position dip model is accurate to within sigma_noise. The paper itself documents residuals at the 10^-3-10^-2 level--an order of magnitude above shot noise--so the claimed 'full posterior distributions that quantify uncertainties' are not demonstrated.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a microwave-free vector magnetometry and crystal orientation determination method for NV ensembles. The PL signal under a tunable axial bias field and sample rotation is modeled as one minus a weighted sum of symmetric lineshape dips located at the nine cross-relaxation resonance planes (Eq. 8). Bayesian inference with a Gaussian likelihood is used to invert experimental PL maps for the crystal orientation (Euler angles restricted by NV symmetry) and for the external field vector (b_z, b_perp, phi_0). The authors demonstrate orientation determination from a single PL map and field reconstruction with reported posterior uncertainties, with an ODMR cross-check. The paper also analyzes systematic errors and discusses symmetry-induced degeneracies and a fundamental global-rotation ambiguity.","tokens_in":14363,"tokens_out":4148,"duration_ms":39040,"significance":"If the method performs as claimed, it would be a useful addition to NV magnetometry: it removes microwave delivery, relaxes the alignment constraints of earlier cross-relaxation schemes, and produces full posterior distributions that explicitly exhibit discrete degeneracies. The analytical resonance-condition model and the symmetry reduction to a fundamental domain are clever and the experimental maps in Figs. 3 and 5 match the simulated maps convincingly. The main value is the combination of a closed-form forward model with Bayesian inversion for near-zero-field, RF-free vector sensing. However, the central uncertainty-quantification claim is currently not supported because the forward model is underspecified and the noise scale is taken from the residuals of the very fit being assessed. The method is promising, but the manuscript needs to close the model and provide a non-circular noise estimate before the posterior widths can be interpreted as accuracies.","major_comments":[{"comment":"Equation (8) is the foundation of the entire inference, but it is not closed: the lineshape function L, contrast C, linewidth Gamma, and weights w_i are never specified, and no fitted or fixed values are reported. The paper states only that L is a symmetric lineshape with linewidth Gamma and that w_i = 1 or 2 in the ideal case. Because every posterior and MAP value in Figs. 3 and 5 depends on Eq. (8), the reader cannot reproduce the analysis or judge whether the chosen L/C/Gamma are physically reasonable. The authors should provide the full model, including the functional form of L, the values (or priors and posteriors) of C and Gamma, and the precise definition of each delta_i for all nine resonance conditions. Without this, the 'analytical model' is not actually specified.","section":"METHOD, Eq. (8)"},{"comment":"The noise parameter sigma_noise = 0.0018 is set to the standard deviation of the residuals between the modeled and measured PL signals, i.e., the residuals of the same dataset being fit. This forces chi^2/dof ~ 1 by construction and does not propagate the documented systematic effects (amplitude differences of order 10^-3 to 10^-2, polarization-dependent excitation, inhomogeneous illumination, field misalignment, and the unmodeled 13C shoulder) into the posterior. Consequently the quoted uncertainties, e.g., b_z = 1.165(2) mT and sub-mrad orientation angles, are effective fit precisions, not true accuracies. To support the paper's central claim that the posterior distributions 'quantify uncertainties', the authors should either estimate sigma_noise from independent repeated measurements or held-out data, or include the acknowledged systematics as additional nuisance parameters/covariance","section":"Analysis of systematic errors"},{"comment":"The field reconstruction is validated only by the statement that the MAP values are 'compatible with independent ODMR measurements'. No ODMR values, error bars, or quantitative difference are given. Since the magnetometry demonstration is a central experimental claim, the authors should report the ODMR result and the discrepancy in units of combined uncertainty (e.g., (b_ODMR - b_MAP)/sigma_comb). Without this, a bias several times the reported 2 microtesla uncertainty would still be loosely 'compatible'.","section":"Application to magnetometry"}],"minor_comments":[{"comment":"Typo: 'that is poses challenges' should be 'that poses challenges'.","section":"Introduction"},{"comment":"The symmetry nomenclature is nonstandard: the orientation-preserving tetrahedral rotation group is usually denoted T (order 12), while T_d is the full tetrahedral group including improper operations (order 24). The text's 'proper tetrahedral group T_d' followed by 'proper octahedral group O_h' is confusing and should be corrected.","section":"Background"},{"comment":"The set of delta_i is written informally as 'delta = {B_x^s - B_y^s, B_x^s - B_z^s, ...}'. Please enumerate all nine resonance conditions explicitly, including the anti-symmetry planes, so that the model is unambiguous.","section":"METHOD, Eq. (8)"},{"comment":"Equation (10) should differentiate the model prediction s_model with respect to x_i, not just s. Equation (11) omits the Gaussian normalization prefactor (2*pi*sigma^2)^(-N/2); if only posterior ratios are used this is harmless, but it should be stated.","section":"METHOD, Eqs. (10)-(11)"},{"comment":"The priors used for alpha, beta, zeta (and later for b_z, b_perp, phi_0) are not stated. Please specify them explicitly, along with the sampling algorithm (e.g., MCMC type), number of samples, and convergence checks. This is important for reproducibility of the reported posterior widths.","section":"Application to orientation determination"},{"comment":"The posterior panels in Figs. 3(c) and 5(b) lack colormap/contour level definitions and do not indicate whether they are normalized histograms or kernel density estimates. Please clarify the visualization so that the reported MAP values and credible intervals can be interpreted.","section":"Figures 3 and 5"}],"recommendation":"major_revision","confidential_remarks":"The paper does not appear to have any integrity issue; the experimental data and simulated maps seem internally consistent. The main problem is that the forward model is underspecified and the uncertainty calibration is circular, both of which are fixable with a revised manuscript that provides the missing model details, independent noise estimation, and quantitative ODMR comparison. I would be willing to review a revised version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this paper does something new. It extends cross-relaxation NV magnetometry to arbitrary crystal orientations and field directions, and it demonstrates—experimentally—that Bayesian inference on photoluminescence maps can recover both the diamond orientation and an applied field vector without microwaves. The core idea is solid, and the experiments support it. The main problems are that the model is underspecified and the reported posterior widths are not real uncertainties yet.\n\nWhat is genuinely good: the symmetry analysis is careful. The authors correctly identify the nine resonance planes, use the tetrahedral/octahedral symmetry to reduce the orientation search space, and handle the inversion degeneracy by reporting two MAP solutions. The analytic model avoids Hamiltonian diagonalization, which makes the inference practical. The simulated PL map reproduces the measured one at the level of dip locations, and the reconstructed field is checked against ODMR. The paper is also honest about known limitations: symmetric rotation axes create ambiguities, and a global rotation about the rotation axis is fundamentally unidentifiable in a single map.\n\nThe soft spots, in proportion: Eq. (8) is the load-bearing model, but the lineshape L, contrast C, and linewidth Γ are never specified—not even fitted values. That is a real reproducibility gap. More importantly, the noise standard deviation used in the likelihood is set to the standard deviation of the residuals of the same fit. That forces the posterior widths to match the fit residuals, but those residuals are dominated by systematic effects (polarization-dependent excitation, inhomogeneous illumination, a 13C shoulder) that the model does not capture. So the quoted uncertainties—sub-mrad angles, microtesla-level field components—are fit precisions, not accuracies. The paper acknowledges this and calls σ_noise an 'effective uncertainty,' but then the abstract's claim of 'full posterior distributions that quantify uncertainties' overstates what is demonstrated. The ODMR cross-check is only described as 'compatible' with no numerical comparison; a bias several times the reported 2 μT would likely still be 'compatible' in that loose sense. Also, the orientation estimate has no independent validation, though the field reconstruction and map reproduction give indirect support.\n\nThese are addressable. Specify the model, release code and data, give the ODMR comparison quantitatively, and either propagate the systematic errors into the posterior or clearly label the uncertainties as conditional on the model. The central method is sound and worth engaging with.\n\nThis paper deserves a serious referee. I would send it to review, with the expectation of a revision. It will be useful to anyone building compact, microwave-free NV magnetometers, and the Bayesian treatment of discrete degeneracies is a nice example for the broader quantum-sensing inference community.","headline":"A genuinely new Bayesian extension of cross-relaxation NV magnetometry that removes the alignment constraint, but the uncertainty budget is not yet credible as reported.","tokens_in":14865,"tokens_out":1485,"would_cite":true,"duration_ms":16194,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that Bayesian inference on photoluminescence maps of cross-relaxation resonances in nitrogen-vacancy centers can determine a diamond's crystal orientation and reconstruct an unknown magnetic field vector, all without micro","keywords":["nitrogen-vacancy centers","vector magnetometry","cross-relaxation","Bayesian inference","photoluminescence","zero-field sensing","diamond","resonance planes"],"falsifier":"Measure a photoluminescence map at a chosen non-symmetric crystal orientation with the external field independently determined by ODMR, then run the Bayesian inference. If the ODMR value lies outside the posterior credible intervals for the field vector (beyond the reported 10⁻³–10⁻² PL residuals) across several field magnitudes and orientations, the additive resonance-plane model is falsified.","tokens_in":13858,"feed_emoji":"💎","tokens_out":3426,"duration_ms":26812,"temperature":0.7,"pith_summary":"This paper claims that a single two-dimensional photoluminescence map — recorded as a diamond sample is rotated in a tunable bias field — contains enough information to determine both the crystal orientation and the vector of an unknown external magnetic field, without any microwave driving. The method relies on cross-relaxation resonances between differently oriented nitrogen-vacancy centers, which appear as sharp dips in the photoluminescence at analytically predicted field configurations. A Bayesian inversion of the map, using a closed-form phenomenological PL model, yields full posterior distributions that quantify uncertainty and naturally exhibit the discrete degeneracies of the NV tetrahedral symmetry. The authors validate the approach experimentally for orientation determination and for reconstructing an unknown field vector, reporting agreement with independent ODMR measurements. If correct, this provides an alignment-free, microwave-free route to compact vector magnetometers.","feed_headline":"Bayesian inference maps NV photoluminescence to field and orientation","feed_subtitle":"A single optical PL map yields crystal orientation and field vector, with uncertainties and NV-symmetry degeneracies.","key_machinery":"The load-bearing object is the geometric resonance-plane model: resonance conditions |B·n_i| = |B·n_j| reduce to nine planar surfaces in sample-frame magnetic field space, and the PL signal is modeled as 1 minus a weighted sum of symmetric lineshape dips at those planes (Eq. 8). This closed form replaces Hamiltonian diagonalization, making repeated likelihood evaluations cheap enough for Bayesian inference, and it embeds the NV symmetry as multi-modal posterior peaks rather than as a single best-fit solution.","core_discovery":"The central discovery is that the set of cross-relaxation resonance conditions between NV orientations forms a geometric structure of nine planes in magnetic field space — six symmetry planes where a single pair of NV classes becomes degenerate, and three anti-symmetry planes where two pairs coincide — and that the photoluminescence map is well described by a simple additive superposition of dips located on these planes. Because the plane positions depend linearly on the magnetic field components in the sample frame, the forward model is analytical and fast. The paper shows that inverting this model with Bayesian inference recovers the diamond orientation and the external field vector from e","pith_inferences":["A natural extension is to combine orientation and field estimation into a single joint inference; the paper notes a global rotation ambiguity around the z-axis that would need a second rotation axis or reference field.","The phenomenological lineshape, contrast, and linewidth are not derived from first principles; a predictive microscopic theory of the cross-relaxation signal could improve accuracy beyond the current systematic residuals (10⁻³–10⁻²).","The Bayesian framework's ability to output multi-modal posteriors could be transferred to other NV sensing modalities, such as ODMR-based vector magnetometry, where discrete symmetries create equivalent solutions.","Because the resonance-plane geometry depends only on the tetrahedral axis arrangement, the same inference machinery should apply to other spin-1 defect centers with similar symmetry, not just NV in diamond."],"forward_implications":["A diamond crystal's orientation can be determined from a single photoluminescence map without microwaves or confocal localization, using only the positions of cross-relaxation dips.","An unknown external field vector can be reconstructed with uncertainties quantified as posterior widths; the method was validated against ODMR for b_z ≈ 1.165 mT, b_⊥ ≈ 0.809 mT, φ0 ≈ 0.720 rad.","When the rotation axis is a symmetry axis of the NV tetrahedron, the PL map becomes periodic and the field reconstruction is ambiguous; choosing a non-symmetric rotation axis lifts the ambiguity and makes the field identifiable.","The inference is efficient enough that magnetometry can in principle be performed from a single PL trace (N=1), with uncertainty scaling as 1/√N as more traces are added.","The method opens the way to self-calibrated absolute magnetometry by using intrinsic low-field PL features as field landmarks."],"fun_headline_variants":["Bayesian inference turns NV photoluminescence into vector magnetometry","Crystal orientation and field vector from NV cross-relaxation maps","Microwave-free NV magnetometry via Bayesian inference","Align-free NV sensor: Bayesian fit to PL maps yields field and orientation","Nine-plane geometry unlocks NV vector magnetometry without microwaves"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The inference assumes that the photoluminescence signal is exactly the sum of smooth symmetric dips sitting at the nine analytic resonance-plane positions, with a fixed lineshape, contrast, and linewidth; the paper's own data show systematic residuals of order 10⁻³–10⁻² from polarization-dependent excitation, inhomogeneous illumination, and field misalignment, so if the true signal deviates from this additive model in a parameter-dependent way, the inferred orientation and fi","fun_headline_variants_meta":{"raw":{"variants":["Bayesian inference turns NV photoluminescence into vector magnetometry","Crystal orientation and field vector from NV cross-relaxation maps","Microwave-free NV magnetometry via Bayesian inference","Align-free NV sensor: Bayesian fit to PL maps yields field and orientation","Nine-plane geometry unlocks NV vector magnetometry without microwaves"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000182,"raw_usage":{"total_tokens":1110,"prompt_tokens":668,"completion_tokens":442,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":412,"completion_tokens_details":{"reasoning_tokens":357}},"tokens_in":412,"tokens_out":442,"duration_ms":5508,"temperature":1.0,"reasoning_tokens":357,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T16:20:16.925123+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure a photoluminescence map at a chosen non-symmetric crystal orientation with the external field independently determined by ODMR, then run the Bayesian inference. If the ODMR value lies outside the posterior credible intervals for the field vector (beyond the reported 10⁻³–10⁻² PL residuals) across several field magnitudes and orientations, the additive resonance-plane model is falsified.","supporting_citations":[],"review_version":1}