{"id":"357b64c4-2860-41f1-96a1-7e3418c429cf","arxiv_id":"2608.06815","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"An optical magnetometer using EPR-entangled probe light and variational readout demonstrates 2.1 ± 0.3 dB sensitivity below the standard quantum limit near 7 kHz in a room-temperature cesium vapor.","lead":"A Copenhagen physics team built an optical atomic magnetometer whose sensitivity dips below the standard quantum limit at acoustic frequencies around 7 kHz, using entangled light and a variational readout technique. This is the first demonstration of quantum-enhanced magnetometry in the low-frequency range relevant for biomedical and geomagnetic sensing.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Sub-SQL margin is model-reconstructed: the SQL reference and the hybrid sensitivity share the same Lorentzian spin model, so known lineshape residuals (Voigt discrepancy) could bias the 2.1 dB margin.","rationale":"The reader's weakest assumption correctly identifies that both the conditional spectrum and the SQL reference are model-reconstructed from the same calibrated linear-response model, so the sub-SQL margin is not directly measured. My analysis confirms this: the ratio S_cond/S_SQL cancels ρ_S and A_B but retains χ_S, Γ_S, and the detection-angle factor, all of which are outputs of the Lorentzian fit. The paper explicitly flags a systematic near-resonance discrepancy reduced by a Voigt profile and uses a different cell for SQL validation, both of which are unquantified sources of systematic uncertainty. The proposed Voigt re-analysis is a concrete, feasible check that would settle whether the 2.1 dB margin survives model-form changes. Given these unresolved model dependencies, the conditional verdict is appropriate; the paper should provide this re-analysis or otherwise demonstrate that the margin is insensitive to the Lorentzian assumption and the cell swap. I agree with the reader's assessment and recommend no change to the conditional verdict.","tokens_in":25522,"tokens_out":10923,"duration_ms":97059,"concrete_test":"Re-analyze the same spin-noise and RF-driven spectra using the Voigt-corrected susceptibility described in Supplementary Sec. XI, re-fit all shared parameters (Γ_S, γ_S, n_S, A_B, broadband terms) to these spectra, and recompute both the SQL reference and the hybrid conditional sensitivity at Ω/2π ≈ 7 kHz with φ_EPR = 130° and θ_mag = −55°. If the maximum sub-SQL suppression under the Voigt model falls below 1.8 dB (the reported value minus 1σ), the Lorentzian lineshape assumption is load-bearing and the headline margin is not robust.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—2.1±0.3 dB below the SQL near 7 kHz—is not a directly measured quantity. The conditional spectrum is converted to field sensitivity via the fitted transfer function, and the SQL is computed from Eq. (8) using the same spin-susceptibility model (χ_S, ρ_S) and fitted parameters Γ_S, γ_S, n_S. In the ratio S_cond/S_SQL, ρ_S and A_B cancel, but the margin still depends on χ_S, Γ_S, and cosθ_mag, all of which inherit the Lorentzian lineshape assumption. The paper itself reports a systematic deviation between Lorentzian fits and data within ±500 Hz of resonance, reduced by a Voigt profile (Supplementary Sec. XI), and acknowledges off-resonant modelling is complicated by residual magnetic-field inhomogeneities and adjacent Zeeman manifolds (Methods). If the true susceptibility at ~3.7 kHz detuning differs from the Lorentzian—e.g., Voigt wings fall faster, reducing χ_S and hence lowering the SQL—the inferred margin shrinks. The quoted 0.3 dB uncertainty comes from a parametric bootstrap over fit covariance only and does not include model-form uncertainty from the Lorentzian versus Voigt choice or from the cell-swapped SQL validation. Because the paper's headline is a claim of first sub-SQL magnetometry in the low-acoustic regime, this model dependence is load-bearing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a room-temperature optical atomic magnetometer operating at Larmor frequencies near 10 kHz and claims to surpass the standard quantum limit (SQL) by combining variational readout with EPR-entangled probe conditioning. The central result is a hybrid configuration (θ_mag ≈ −55°, φ_EPR ≈ 130°) that allegedly achieves 2.1 ± 0.3 dB suppression of the noise-equivalent magnetic-field sensitivity below the SQL around 7 kHz (Fig. 4d and inset). The authors first characterize the noise budget and validate near-SQL operation at 59 kHz, then demonstrate variational readout, EPR conditioning, and the hybrid combination, supported by a linear-response model and extensive calibration.","tokens_in":25860,"tokens_out":4264,"duration_ms":38053,"significance":"If the central claim is sound, this would be the first demonstration of sub-SQL optical magnetometry in the low-acoustic frequency range, which is relevant for biomedical and geomagnetic sensing. The paper is strong in experimental methodology: it includes detailed calibration of the spin transfer function, cross-validation against observed ponderomotive squeezing, RF-driven response fits, and a Monte Carlo bootstrap for statistical uncertainty. The ratio S_meas/S_SQL cancels the transduction factor A_B and the spin susceptibility ρ_S(Ω), which reduces some calibration dependence. However, the headline margin is model-reconstructed rather than directly measured, and the SQL reference itself is derived from the same Lorentzian spin model used to fit the data. The paper's own supplementary material reports systematic Lorentzian residuals and improved Voigt fits, making the model-form sensitivity of the 2.1 dB claim a load-bearing issue.","major_comments":[{"comment":"The sub-SQL margin is not a directly measured quantity. Both the conditional spectrum and the SQL reference in Eq. (8) are computed from the same Lorentzian susceptibility model, with parameters Γ_S, γ_S, n_S, and A_B extracted by fitting the same spin-noise data. In the ratio S_meas/S_SQL, A_B and ρ_S cancel, but the ratio still depends on χ_S, Γ_S, and cos(θ_mag), all of which inherit the Lorentzian lineshape assumption. Supplementary Sec. XI explicitly reports a systematic deviation between Lorentzian fits and data within ±500 Hz of resonance, reduced by a Voigt model, and Methods acknowledges that off-resonant modelling is complicated by residual magnetic-field inhomogeneities and adjacent Zeeman manifolds. Because the 2.1 dB claim is evaluated at ~3.7 kHz detuning, where Voigt wings can differ from Lorentzian tails, a Voigt correction could shift the inferred SQL and reduce the margin. I ask the authors to quantify the model-form uncertainty by recomputing the margin with the Voigt susceptibility and/or by extracting the SQL directly from the data in a way that does not assume the same lineshape model.","section":"Methods, 'BENCHMARKING AGAINST THE STANDARD QUANTUM LIMIT'; Eq. (8); Supplementary Sec. XI"},{"comment":"The SQL validation was performed on a different Cs cell of identical geometry, because the original cell's atomic density had degraded. The text asserts that this does not affect the results since the SQL depends only on the cell's own parameters, which is logically correct if the model is fully universal. However, the validation is intended to test the theoretical SQL model against experimental data; without validation on the same cell used for the 2.1 dB claim, the transfer of the model relies on the unverified assumption that the measurement cell has no additional inhomogeneous broadening or other lineshape anomalies beyond the fitted Lorentzian parameters. Please provide explicit evidence that the measurement cell's noise spectra are consistent with the same model form (e.g., by showing the fit residuals for the measurement cell in the off-resonant window), or state this as a caveat on the claim.","section":"Methods, 'BENCHMARKING AGAINST THE STANDARD QUANTUM LIMIT'"},{"comment":"The quoted ±0.3 dB uncertainty on the sub-SQL margin is obtained from a parametric Monte Carlo bootstrap over the fit covariance matrix only. This does not include systematic contributions from model-form uncertainty (Lorentzian versus Voigt), from the cell-swapped SQL validation, or from the ±10% model-data agreement in the off-resonant window that the authors themselves cite. Since the systematic residuals are comparable in size to the quoted statistical error, the reported uncertainty likely understates the true uncertainty of the 2.1 dB margin. Please provide a more conservative uncertainty estimate that includes these systematic contributions, or give a quantitative argument for why they are negligible.","section":"Methods, 'UNCERTAINTY ANALYSIS AND MODEL UNCERTAINTY'"}],"minor_comments":[{"comment":"The abstract claims 'we demonstrate such sensitivity' and 'we demonstrate overcoming the limit'; because the sub-SQL sensitivity is reconstructed from a fitted model rather than measured directly as a raw noise trace, consider phrasing such as 'we report a reconstructed sensitivity that exceeds the SQL' to avoid overstatement.","section":"Abstract and Introduction"},{"comment":"The inset is described as 'conditional variance normalized to the SQL' but the vertical axis label is not defined in the caption; please state explicitly that the normalization uses the SQL from Eq. (8) with the calibrated parameters, and clarify whether the spin thermal noise and optical losses are included in the numerator.","section":"Fig. 4d caption and inset"},{"comment":"The sentence 'the choice of calibration cell does not affect the results' is too strong given that the validation is meant to test the model against experiment; consider rephrasing to explain why cell-specific parameters are sufficient for transferring the validation.","section":"Methods, 'BENCHMARKING AGAINST THE STANDARD QUANTUM LIMIT'"},{"comment":"The terms 'beyond-SQL' and 'sub-SQL' are used interchangeably; please define the preferred term once in the introduction and use it consistently.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The central claim is plausible but rests on a model-reconstructed SQL reference with known lineshape residuals. The authors have the tools to address this—Voigt-model recomputation, measurement-cell validation, and systematic uncertainty propagation—so major revision rather than rejection is appropriate. I would encourage the editor to request these robustness checks before considering the paper for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper reports the first sub-SQL optical magnetometer in the low-acoustic band: Larmor at 10.7 kHz, sensitivity 2.1±0.3 dB below the SQL around 7 kHz, using a hybrid of variational readout and EPR conditioning. The EPR-only version gives 1.4±0.2 dB. If the claim holds, it closes a real gap, since earlier quantum-enhanced magnetometry ran at higher RF Larmor frequencies.\n\nThe new element is the hybrid combination itself. Variational readout and EPR conditioning are established separately, but putting them together on a room-temperature Cs magnetometer and showing the enhancement window can be shifted a few kHz below resonance is a solid experimental achievement. The calibration is unusually careful: multi-tone RF transfer functions, power sweeps, cross-validation against observed ponderomotive squeezing, a cell-swap SQL benchmark, and a Monte Carlo bootstrap for the uncertainty. The supplementary model is explicit enough to rebuild.\n\nThe soft spot is the one the stress-test flags. The sub-SQL margin is not a directly measured number. The conditional sensitivity and the SQL reference are computed from the same fitted spin-susceptibility model. In the ratio, ρ_S and A_B cancel, but the margin still inherits any bias in χ_S, Γ_S, and the Lorentzian lineshape. The paper discloses a Voigt discrepancy near resonance and some off-resonant modelling complications. I think the specific worry about the 7 kHz window is overstated, because the Voigt issue is reported within ±500 Hz of resonance, not at the 3.7 kHz detuning where the enhancement sits. But the broader point holds: the 0.3 dB uncertainty is a parametric bootstrap, not a model-form uncertainty. If the off-resonant susceptibility is off by a few percent, the margin shifts by a comparable amount.\n\nThis is not a disqualifying problem. The authors enforced a ±10% model-data agreement in the analysis window, used an independent cell for the SQL validation, and canceled the worst-calibrated factors. The honest reading is that this is a well-executed experiment whose central quantitative claim is model-dependent in a way the text discloses but does not quantify. I'd want data and code released and a sensitivity analysis over the Lorentzian vs Voigt choice before quoting 2.1 dB as established.\n\nThe paper deserves a serious referee. The experimental effort is substantial, the novelty is real, and the claim is significant. Send it to review, with a request for a model-form uncertainty estimate.\n\nRecommendation: engage; accept for review with that condition.","headline":"A careful, potentially first demonstration of sub-SQL optical magnetometry at acoustic frequencies, with the caveat that the headline 2.1 dB margin is reconstructed from the same model that defines the SQL.","tokens_in":26418,"tokens_out":3696,"would_cite":true,"duration_ms":31157,"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":"A room-temperature optical magnetometer reaches 2.1 ± 0.3 dB below the standard quantum limit around 7 kHz by combining variational readout with EPR conditioning.","keywords":["optical magnetometry","standard quantum limit","Einstein-Podolsky-Rosen entanglement","variational readout","ponderomotive squeezing","quantum backaction","acoustic frequency sensing","Faraday magnetometer"],"falsifier":"Measure the hybrid-conditioned photocurrent at the 7 kHz sideband while sweeping probe power, and compare the noise floor against an SQL computed from parameters fixed independently of the fitted noise partition, for example from the RF-driven spin response alone. If the model misassigns technical or inhomogeneous-broadening noise to backaction or imprecision, the inferred SQL and the reported 2.1 dB margin would shift by an amount comparable to the misassigned noise.","tokens_in":25333,"feed_emoji":"🧲","tokens_out":12214,"duration_ms":90847,"temperature":0.7,"pith_summary":"The paper reports a room-temperature optical atomic magnetometer whose noise-equivalent magnetic-field sensitivity falls below the standard quantum limit (SQL) in the low-acoustic frequency band. The central claim is that combining variational readout, detecting the collective spin at a rotated quadrature $\\theta_{\\mathrm{mag}} \\approx -55^\\circ$, with EPR conditioning, using one arm of an entangled two-mode optical state to subtract correlated noise at phase $\\phi_{\\mathrm{EPR}} \\approx 130^\\circ$, yields $2.1 \\pm 0.3$ dB of suppression below the SQL near 7 kHz at a Larmor frequency of 10.7 kHz. This matters because earlier sub-SQL optical magnetometry operated at radio-frequency Larmor precession, whereas the kilohertz acoustic band is the range relevant to biomedical and geomagnetic sensing. If the claim holds, engineered quantum correlations can relocate and reshape the sensitivity of a continuous Faraday magnetometer instead of relying on probe-power scaling.","feed_headline":"Magnetometer beats the quantum limit at 7 kHz","feed_subtitle":"Entangled probe light plus a rotated readout suppresses noise 2.1 dB below the standard quantum limit.","key_machinery":"The machinery is a linear-response spin-oscillator model of the Faraday magnetometer. The collective spin is treated as a damped harmonic oscillator with Larmor frequency $\\Omega_S$, susceptibilities $\\chi_S(\\Omega)$ and $\\rho_S(\\Omega)$, readout rate $\\Gamma_S$, and decoherence $\\gamma_S$, coupled to the probe through the quadrature interaction $H_{\\mathrm{eff}}=\\sqrt{\\Gamma_S}\\,X_S X_L-\\sqrt{A_B}\\,B_{\\mathrm{RF}} P_S$. Variational readout rotates the detected light quadrature by $\\theta_{\\mathrm{mag}}$, producing a nonzero imprecision-backaction correlation term proportional to $\\Gamma_S \\mathrm{Re}[\\chi_S(\\Omega)]\\sin(2\\theta_{\\mathrm{mag}})$, i.e., ponderomotive squeezing. EPR conditioning replaces the vacuum mode in the probe's orthogonal polarization with one mode of a two-mode squeezed state and combines the atomic photocurrent with the other mode through a Wiener filter, subtracting correlated noise; the hybrid configuration does both at once. The spin response functions and loss model are common to all configurations, so the configurations differ only in the correlation term $S_{\\mathrm{corr}}(\\Omega)$.","core_discovery":"The discovery, on the paper's own terms, is that the standard quantum limit in continuous Faraday magnetometry is not a hard floor set by probe power but can be exceeded by creating correlations between measurement imprecision and quantum backaction. With the spin precessing at 10.7 kHz, the uncorrelated readout reference gives a peak sensitivity of about $65\\ \\mathrm{fT}/\\sqrt{\\mathrm{Hz}}$; EPR conditioning alone reaches about $47\\ \\mathrm{fT}/\\sqrt{\\mathrm{Hz}}$ and gives $1.4 \\pm 0.2$ dB below the SQL at $\\phi_{\\mathrm{EPR}} = 130^\\circ$, while the hybrid configuration, with the variational readout at $\\theta_{\\mathrm{mag}} \\approx -55^\\circ$, reaches $2.1 \\pm 0.3$ dB below the SQL near 7 kHz. The paper further claims that the depth, central frequency, and bandwidth of the enhancement are tunable through the two phases, and that frequency-dependent optimization of the conditioning phase would extend the enhancement over roughly an 8 kHz window, broadening the 3 dB detection bandwidth from about 8 kHz to about 13 kHz. These results are presented as the first sub-SQL optical magnetometry in the acoustic frequency regime.","pith_inferences":["Because the sub-SQL margin is reconstructed from a fitted model rather than read directly from a single noise trace, a natural test would be to verify the inferred SQL against an independent calibration, for example by injecting a known vacuum-noise-limited probe and checking the model's partition of imprecision versus backaction at 7 kHz.","The same recipe, rotated readout plus an entangled reference channel combined by Wiener filtering, should transfer to other linear Faraday sensors, including solid-state spin ensembles and hot-vapour SERF-type magnetometers, once their susceptibilities are known.","The predicted sensitivity anisotropy at acoustic Larmor frequencies implies that a dual-quadrature or vector readout may outperform any single-axis configuration, a possibility the paper mentions but does not demonstrate experimentally.","A filter cavity or negative-mass reference oscillator implementing the frequency-dependent optimal conditioning phase would convert the demonstrated narrowband enhancement into a broadband one; this is a concrete engineering step suggested by the data."],"forward_implications":["Sub-SQL optical magnetometry is no longer confined to radio-frequency Larmor precession; the same room-temperature apparatus operates below the SQL at about 7 kHz, in the acoustic band relevant to biomagnetic and geomagnetic sensing.","The enhancement depth, central frequency, and bandwidth become controllable via the variational readout angle and the EPR conditioning phase, relaxing the usual peak-sensitivity-versus-bandwidth trade-off.","Frequency-dependent optimization of the conditioning phase would extend quantum enhancement over an approximately 8 kHz window and widen the 3 dB detection bandwidth from roughly 8 kHz to 13 kHz.","Because increasing probe power degrades the collective spin length, quantum correlation engineering becomes the practical route to improved sensitivity instead of power scaling.","The vector magnetometer model predicts a direction- and frequency-dependent quantum-limited sensitivity, suggesting that multi-axis probing or a frequency-dependent measurement basis would be needed for optimal broadband operation."],"supporting_citations":[{"why":"Establishes the quantum-noise-limited and entanglement-assisted magnetometry baseline that this work extends into the acoustic regime.","marker":"[9]"},{"why":"Provides the negative-mass reference-frame method for back-action-evading measurement that underlies EPR conditioning.","marker":"[17]"},{"why":"Demonstrates broadband displacement detection with quantum correlations, the optomechanical analogue of variational readout.","marker":"[18]"},{"why":"Confirms continuous force and displacement measurement below the standard quantum limit with correlated light.","marker":"[19]"},{"why":"Shows sub-SQL squeezing in a gravitational-wave detector and supplies the SQL-comparison methodology adapted here.","marker":"[23]"},{"why":"Introduces virtual rigidity, used to explain the frequency shift and sensitivity change from variational readout.","marker":"[34]"},{"why":"Describes the hybrid quantum network for acoustic-frequency sensing whose conditioning scheme this paper applies.","marker":"[37]"},{"why":"Supplies the two-colour high-purity Einstein-Podolsky-Rosen optical source used as the entanglement resource.","marker":"[41]"},{"why":"Characterizes the acoustic-frequency atomic spin oscillator in the quantum regime, providing the atomic platform and its parameters.","marker":"[42]"}],"fun_headline_variants":["Magnetometer beats quantum limit 2.1 dB at 7 kHz","Entangled light lifts magnetometer 2.1 dB below SQL","Acoustic magnetometry exceeds quantum limit by 2.1 dB","Sub-SQL magnetometry at 7 kHz with entangled light"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reported 2.1 dB margin is computed, not read directly: the experimenters fit a model of the atom-light measurement to the same noise spectra they compare against the standard quantum limit, so the claim stands or falls on whether that model correctly divides the measured noise into light-imprecision, backaction, and intrinsic atomic parts near 7 kHz.","fun_headline_variants_meta":{"raw":{"variants":["Magnetometer beats quantum limit 2.1 dB at 7 kHz","Entangled light lifts magnetometer 2.1 dB below SQL","Acoustic magnetometry exceeds quantum limit by 2.1 dB","Sub-SQL magnetometry at 7 kHz with entangled light"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00072,"raw_usage":{"total_tokens":3236,"prompt_tokens":954,"completion_tokens":2282,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":570,"completion_tokens_details":{"reasoning_tokens":2206}},"tokens_in":570,"tokens_out":2282,"duration_ms":15343,"temperature":1.0,"reasoning_tokens":2206,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T14:29:30.022920+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the hybrid-conditioned photocurrent at the 7 kHz sideband while sweeping probe power, and compare the noise floor against an SQL computed from parameters fixed independently of the fitted noise partition, for example from the RF-driven spin response alone. If the model misassigns technical or inhomogeneous-broadening noise to backaction or imprecision, the inferred SQL and the reported 2.1 dB margin would shift by an amount comparable to the misassigned noise.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the quantum-noise-limited and entanglement-assisted magnetometry baseline that this work extends into the acoustic regime."},{"cited_title":"B., Thomas, R","cited_arxiv_id":null,"evidence_quote":"Provides the negative-mass reference-frame method for back-action-evading measurement that underlies EPR conditioning."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrates broadband displacement detection with quantum correlations, the optomechanical analogue of variational readout."},{"cited_title":"& Schliesser, A","cited_arxiv_id":null,"evidence_quote":"Confirms continuous force and displacement measurement below the standard quantum limit with correlated light."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows sub-SQL squeezing in a gravitational-wave detector and supplies the SQL-comparison methodology adapted here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces virtual rigidity, used to explain the frequency shift and sensitivity change from variational readout."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Describes the hybrid quantum network for acoustic-frequency sensing whose conditioning scheme this paper applies."},{"cited_title":"B., Novikov, V ., Kerdoncuff, H.et al.Two-colour high-purity Einstein-Podolsky-Rosen photonic state.Nat","cited_arxiv_id":null,"evidence_quote":"Supplies the two-colour high-purity Einstein-Podolsky-Rosen optical source used as the entanglement resource."},{"cited_title":"B.et al.Acoustic fre- quency atomic spin oscillator in the quantum regime.Nat","cited_arxiv_id":null,"evidence_quote":"Characterizes the acoustic-frequency atomic spin oscillator in the quantum regime, providing the atomic platform and its parameters."}],"review_version":2}