{"id":"914e427d-cf35-440d-8d0f-969da1d5dcca","arxiv_id":"2009.13212","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A single multipass cell with two oppositely pumped rubidium ensembles directly subtracts Faraday rotation and achieves 10.1 fT/cm/√Hz gradiometer sensitivity.","lead":"A new magnetic gradiometer uses a single glass cell with two rubidium atom clouds polarized in opposite directions, so their signals cancel out and only the magnetic difference remains. It measures gradient sensitivity of about 10 femtotesla per centimeter per square-root hertz and also reduces atomic diffusion noise that limits other multipass sensors.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Sensitivity claim rests on the Eq. (1) FID model; footnote 24's admitted non-exponential decay makes Δν identifiability in the four-parameter fit unproven.","rationale":"The paper is technically strong: direct subtraction is demonstrated with >98% cancellation, the fitted frequency difference tracks the applied gradient with the nominal coil calibration, and the spin-noise spectrum is matched by a parameter-free theory using an independently measured T2. I considered the diffusion claim as an alternative weak point, but the parameter-free agreement and the independent low-polarization T2 measurement make it well supported. The most load-bearing assumption is therefore the generative model for the differential FID. Eq. (1) imposes an exponential decay; footnote 24 explicitly admits that the high-polarization decay can be non-exponential. Because the headline sensitivity is derived from the scatter of the fitted Δν, a misspecified envelope could either bias the estimated gradient (partially checked by the linear calibration) or inflate the scatter through coupling to Δφ, ΔT2, and Δd0 (not checked). The claimed independent identifiability follows from the linear independence of e^{-t/T2}cos, t e^{-t/T2}cos, e^{-t/T2}sin, t e^{-t/T2}sin in the small-difference expansion, but the actual fit is to Eq. (1) at large φ0 with a possibly non-exponential decay, and no numerical identifiability or error-propagation study is supplied. This does not invalidate the central result, but it makes the headline sensitivity conditional on a numerical test. The reader's weakest assumption identifies the same issue, so I agree; since the reader already returned CONDITIONAL and this concern does not move the verdict, I recommend UNCHANGED.","tokens_in":9127,"tokens_out":18715,"duration_ms":173385,"concrete_test":"Run a Monte Carlo of the Section III fit with the reported parameters (νL=96 kHz, φ0≈3.5 rad, T2≈0.55 ms, Tm=2 ms). Generate differential FIDs with the exponential Eq. (1) and with non-exponential decay f(t)=e^{-(t/T2)^β} for β=0.7–0.9, adding ~1% shot-to-shot jitter in Δφ, ΔT2, Δd0; refit each with Eq. (1). If the recovered Δν is biased by more than 10% of the true value, or if the non-magnetic jitter inflates the scatter beyond the pure-magnetic-noise comparison (3.6 vs 14.2 fT/√Hz), the reported 10.1 fT/cm/√Hz is inflated. Also compute the Fisher information matrix of {Δν, Δφ, ΔT2, Δd0} and its condition number; a large Δν–Δd0 correlation would show the 2.2 mHz scatter is not uniquely attributable to the magnetic gradient.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The measured 10.1 fT/cm/√Hz rests on the third fit in Section III, which assumes Eq. (1) is the correct generative model for both ensembles and that {Δν, Δφ, ΔT2, Δd0} are independently identifiable. Footnote 24 concedes that at high polarization the FID decay 'can be non-exponential due to partial spin-exchange relaxation suppression,' while Eq. (1) forces an exponential envelope. For the direct differential signal, the Δν term is proportional to the true (possibly non-exponential) decay envelope times t·cos(ωt); if the envelope is actually non-exponential, the fitted exponential model can couple Δν to ΔT2 and Δd0. The paper's assertion that all four variables 'can be determined independently' is plausible from the ideal basis functions in Eq. (2), but it is not demonstrated: there is no Monte Carlo, no covariance or Fisher-matrix analysis, and no uncertainty is quoted on the 2.2 mHz scatter or on the headline 10.1 fT/cm/√Hz. If the decay shape is non-exponential and fluctuates shot to shot (pump power, RF amplitude, temperature), σ_Δν could be inflated by non-magnetic parameter fluctuations, making the reported sensitivity an overestimate of true magnetic-gradient sensitivity. The calibration slope agreement with the nominal coil constant checks linearity but cannot detect such inflation or a constant multiplicative bias absorbed by the fitted 0.3 nT/cm offset.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a direct magnetic gradiometer built from two oppositely polarized 87Rb ensembles inside a single V-shaped multipass cell. A far-detuned probe undergoes multiple passes through both ensembles, and the Faraday rotations from the two ensembles subtract intrinsically, yielding a direct differential signal with cancellation of more than 98% of two individual rotations that each exceed 3.5 rad. The authors introduce two analysis methods for extracting the frequency difference Δν (a nonlinear four-parameter fit and a real-time FFT/matched-filter method) and report a measured gradiometer sensitivity of 10.1 fT/cm/√Hz with a 1.4 cm baseline. They also measure the spin-noise spectrum of the unpolarized ensemble and compare it with a parameter-free theoretical prediction that includes atomic diffusion, finding nearly Lorentzian line shapes and a slower diffusion-induced decay than in earlier multipass-cell work. The central claims are the genuine differential sensitivity and the diffusion advantage of the overlapping-beam geometry.","tokens_in":9390,"tokens_out":3521,"duration_ms":32948,"significance":"If the measured sensitivity is validated, the result is significant because it demonstrates a single-cell, single-output direct gradiometer with fT/cm-scale sensitivity, avoiding the signal-processing complications of large optical rotations and providing real-time common-mode rejection. The spin-noise measurement is also valuable: the reported overlap of multiple probe passes with a uniform beam is claimed to reduce diffusion broadening and to preserve the Lorentzian spin-noise line shape, which is relevant for future spin-squeezing experiments. The paper includes a parameter-free comparison with the theory of Ref. [23] and a calibration against a known gradient coil, and the agreement of the spin-noise spectrum with theory is a clear strength. The main weaknesses are statistical: the identifiability of the four-parameter fit is asserted rather than demonstrated, and no uncertainty is quoted on the headline sensitivity.","major_comments":[{"comment":"The extraction of the gradient signal rests on the assumed model in Eq. (1), whose four residual parameters {Δν, Δφ, ΔT2, Δd0} are asserted to be independently determinable without supporting evidence. Footnote 24 explicitly concedes that at high initial polarization the FID decay can be non-exponential due to partial spin-exchange relaxation suppression, while Eq. (1) enforces an exponential envelope. Under a misspecified envelope, the Δν term is not guaranteed to be orthogonal to ΔT2 and Δd0, and the reported 2.2 mHz scatter could be inflated by shot-to-shot fluctuation of non-magnetic parameters; the paper provides no covariance matrix, Fisher-matrix analysis, Monte Carlo study, or noise-injection test. Please add such an identifiability analysis and quote an uncertainty (including number of repetitions and standard error) on σ_Δν and on the headline 10.1 fT/cm/√Hz. This is load-bearing because the central claim is the measured sensitivity.","section":"Section III, Eq. (1) and third fit"},{"comment":"The linear calibration against the nominal coil constant verifies the slope but does not validate the noise floor: a constant multiplicative gain error or a non-magnetic broadening of the fit residual would leave the calibration slope unchanged. The inset scatter is shown without error bars, and the number of repetitions, the standard error of σ_Δν, and the stationarity of the scatter over time are not reported. Please report these so that the reader can judge whether 10.1 fT/cm/√Hz is a stable, unbiased estimate.","section":"Section III, Fig. 3, calibration"},{"comment":"The claim of 'excellent agreement' between the measured spin-noise spectrum and the parameter-free prediction from Eq. (17) of Ref. [23] is presented visually only; no quantitative goodness-of-fit statistic or residual analysis is given. Because this agreement is one of the paper's two central results, please provide a numerical comparison (e.g., normalized residuals in the peak region) and state the uncertainties of the input parameters (T2, D, w0) that enter the prediction.","section":"Section IV, spin-noise comparison"}],"minor_comments":[{"comment":"The argument of the trigonometric functions is written as 2πν_L + d0; it should be 2πν_L t + d0. This typo obscures the derivation of the matched filter.","section":"Section III, Eq. (2)"},{"comment":"The phrase 'femtotesla projected sensitivity' appears without a definition; clarify whether it refers to the 3.6 fT/√Hz unpolarized measurement or to an extrapolation.","section":"Abstract and Conclusions"},{"comment":"The values of Δφ, ΔT2, and Δd0 obtained from the fits are not reported; reporting them would help assess the balance of the two arms and the validity of Eq. (2).","section":"Section III, third fit"},{"comment":"The inset labels 'top' and 'bottom' for the individual signals are not visible in the printed version; consider adding arrows or using separate panels.","section":"Fig. 2"}],"recommendation":"major_revision","confidential_remarks":"The paper is well within the scope of the journal and the experimental work appears carefully done. The main concern is statistical rather than conceptual: the four-parameter fit identifiability and the absence of uncertainties are fixable with additional analysis and should be addressed before publication. The parameter-free spin-noise comparison is a strong positive element that reduces circularity concerns."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a well-executed experimental paper from a strong group. The new piece is the single V-shaped multipass cell with opposite circular pumping and overlapping beams, giving direct subtraction of Faraday rotations that individually exceed 3.5 rad. That is a real simplification over two-cell or two-output gradiometers, and the diffusion advantage—where overlapping beams yield a nearly Lorentzian spin noise spectrum—is the most convincing part of the work. The spin noise spectrum matches the parameter-free prediction from Ref. [23], and the measured rms rotation agrees with theory. That part is solid and reproducible.\n\nThe soft spot is the headline sensitivity. The 10.1 fT/cm/√Hz comes from a four-parameter fit {Δν, Δφ, ΔT2, Δd0} to Eq. (1), which assumes exponential FID decay. The paper itself admits in footnote 24 that high polarization can make the decay non-exponential. If the decay shape is not exactly as modeled, Δν could couple to the other fit parameters, and the quoted 2.2 mHz scatter has no error bar. The authors assert that the four variables are independently determined, but they give no covariance matrix or Monte Carlo check. So the sensitivity number is a plausible demonstration, not a rigorously characterized result. The 'femtotesla projected sensitivity' in the conclusions is also unsupported—there is no argument for what limits it.\n\nThat said, these are missing supporting details, not load-bearing flaws. The linear calibration against the known coil constant checks the overall gradient response, and the diffusion result stands on its own. The citation pattern is fair: the self-citations to Refs. [16,20,23] are directly relevant, and Ref. [23] provides the parameter-free theory that validates the noise measurement. Nothing here feels circular.\n\nWho gets value from this paper? Researchers in atomic magnetometry, especially those working on gradiometers and spin noise in multipass cells. It deserves a serious referee. My recommendation: send it to peer review. The referee should ask for an uncertainty on the sensitivity, a test or justification of the exponential decay assumption, and a clarification of the projected sensitivity. But this is a useful, honest paper that should be in the literature.\n\nI'd bring it to reading group and would cite it if I were working in this area.","headline":"A solid single-cell direct gradiometer with competitive sensitivity and a well-supported diffusion advantage; the headline number lacks an uncertainty analysis, but the core physics is credible.","tokens_in":9937,"tokens_out":1926,"would_cite":true,"duration_ms":19269,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":[],"pacs":["32.10.-f","07.55.Ge","42.50.Lc","32.80.Bx"],"model":"deepseek-v4-flash","headline":"A single multipass cell measures magnetic gradients at 10.1 fT/cm/√Hz by cancelling two 3.5-radian Faraday rotations.","keywords":["optical magnetometry","magnetic gradiometer","multipass cell","Faraday rotation","spin noise","rubidium atomic vapor","free induction decay","common-mode noise rejection"],"falsifier":"Measure the same gradiometer signal while sweeping the pump intensity from low to high polarization, and compare the fitted Δν to the coil-calibrated gradient; if Δν shifts systematically as the decay shape changes, the four-parameter model is not the right null hypothesis for the background. Separately, record the spin noise spectrum with the probe beams forced to non-overlapping positions at unchanged optical depth: a clear broadening to the diffusion-limited multi-Lorentzian shape would confirm that the overlap, not just the multipass gain, is responsible for the near-Lorentzian line.","tokens_in":8917,"feed_emoji":"🧲","tokens_out":5883,"duration_ms":49911,"temperature":0.7,"pith_summary":"This paper claims that a direct magnetic gradiometer can be built from two rubidium ensembles inside a single multipass cell, with the differential signal obtained by optical subtraction rather than by combining separate sensor outputs. Each ensemble produces a Faraday rotation of more than 3.5 radians, yet because the ensembles are pumped with opposite circular polarization the two rotations nearly cancel, leaving a small signal that tracks the magnetic-field gradient. The authors report a measured gradiometer sensitivity of 10.1 fT/cm/√Hz with a 1.4 cm baseline, and they argue that the V-shaped overlapping-beam geometry reduces atomic-diffusion effects on spin noise, in parameter-free agreement with theory. If correct, the approach offers a compact path to femtotesla-level gradient sensing with common-mode noise rejection and a route toward spin-squeezing enhancement.","feed_headline":"Two canceling rotations reach 10.1 fT/cm/√Hz","feed_subtitle":"A single multipass cell subtracts two Faraday signals directly; overlapping beams reduce diffusion noise.","key_machinery":"The load-bearing mechanism is the direct-subtraction free-induction signal of Eq. (1): two sinusoids with Larmor frequency νL, opposite initial phases, and individual amplitudes φ0, phases d0, and relaxation times T2, one shifted by +Δν/2 and the other by −Δν/2. Because both arms share the same probe beam and polarimeter, their rotations—each exceeding 3.5 rad—cancel to better than 98%, and the residual signal's envelope grows in proportion to the frequency difference Δν. The V-shaped multipass cell makes this work by directing 60 probe passes through two oppositely pumped ensembles while keeping the beams overlapping over a uniform 3.6 mm diameter region, which is what suppresses diffusion decorrelation. The extraction is carried by two analysis tools: a nonlinear fit with four independent residual parameters, and a matched-filter FFT that reads Δν from the imaginary Fourier component in real time.","core_discovery":"The paper's central claim is that a single V-shaped multipass cell containing two 87Rb ensembles can act as a direct gradiometer: pumping the ensembles with opposite circular polarization makes their large paramagnetic Faraday rotations subtract intrinsically, producing a near-zero baseline signal whose time evolution encodes the frequency difference Δν between the two arms caused by the magnetic gradient. With this arrangement the authors measure a gradiometer sensitivity of 10.1 fT/cm/√Hz for a 1.4 cm baseline, and they show that the residual spin noise spectrum is nearly Lorentzian. The fast-decaying diffusion component is suppressed because the probe beams overlap over a uniform wide region, so the spin-noise correlation function matches the theoretical prediction Cth_d(t)=1/(1+4tD/$w0^{2}$) with no free parameters. The paper further demonstrates that two analysis methods—a four-parameter nonlinear fit and a real-time matched-filter FFT—extract Δν at similar sensitivity, with the unpolarized-ensemble noise floor approaching the fundamental 2.7 fT/√Hz level from simulation.","pith_inferences":["An immediate testable extension would be to vary the pump power so the individual decays become visibly non-exponential and check whether the fitted Δν drifts; a drift would indicate the four-parameter model misses polarization-dependent decay.","If the diffusion suppression holds at higher atomic density, the same overlapping-beam geometry could serve as a platform for QND spin squeezing in a dense vapor, since the probe would address a single uniform ensemble rather than many independent beamlets.","In an unshielded environment the direct subtraction should reject broadband common-mode noise at the analog level, but the limiting factor would likely be how well the two arms' gains and phases remain matched at high frequencies, which the paper does not test."],"forward_implications":["A single-cell direct gradiometer can reach 10.1 fT/cm/√Hz with a 1.4 cm baseline, removing the need to combine two separate sensor outputs for common-mode rejection.","Cancellation of >3.5 rad rotations keeps the polarimeter signal within a small range, avoiding the signal wrap-around that would otherwise limit multipass Faraday measurements.","The real-time matched-filter FFT extracts Δν from the imaginary Fourier component at sensitivity close to the offline nonlinear fit, enabling shot-by-shot gradient readout.","Because diffusion affects the spin-noise correlation only weakly, the noise spectrum is nearly Lorentzian, and a quantum-noise-limited version could preserve spin-squeezing correlations for longer."],"supporting_citations":[{"why":"supplies the prior multipass-cell magnetometry baseline and the diffusion-limited spin-noise behavior the present design improves on","marker":"[16]"},{"why":"documents large optical rotations in multipass cells and the wrap-around limitation this gradiometer avoids","marker":"[20]"},{"why":"derives the analytic diffusion correlation function used for the parameter-free spin-noise spectrum","marker":"[23]"},{"why":"justifies the matched-filter window for the real-time FFT analysis method","marker":"[26]"},{"why":"provides the Faraday-rotation spin-noise spectroscopy method used to measure the intrinsic noise peak","marker":"[28]"},{"why":"supplies the theoretical optical-rotation noise estimate used to compare with the measured spin noise","marker":"[29]"}],"fun_headline_variants":["Single-cell gradiometer hits 10.1 fT/cm/√Hz","Opposite spins cancel: 10.1 fT/cm/√Hz gradiometer","Multipass cell cuts diffusion, hits 10.1 fT/cm/√Hz","One cell, two Rb clouds: direct gradiometer at 10.1 fT/cm/√Hz"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reported sensitivity depends on the assumption that the free-induction decay in each arm follows the two-sine model of Eq. (1), so that all differences between the arms can be captured by four residual parameters; the paper's own footnote admits the decay can become non-exponential at high polarization, in which case the extracted frequency difference could be biased.","fun_headline_variants_meta":{"raw":{"variants":["Single-cell gradiometer hits 10.1 fT/cm/√Hz","Opposite spins cancel: 10.1 fT/cm/√Hz gradiometer","Multipass cell cuts diffusion, hits 10.1 fT/cm/√Hz","One cell, two Rb clouds: direct gradiometer at 10.1 fT/cm/√Hz"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000316,"raw_usage":{"total_tokens":1758,"prompt_tokens":881,"completion_tokens":877,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":497,"completion_tokens_details":{"reasoning_tokens":780}},"tokens_in":497,"tokens_out":877,"duration_ms":7652,"temperature":1.0,"reasoning_tokens":780,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-27T22:08:54.089344+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the same gradiometer signal while sweeping the pump intensity from low to high polarization, and compare the fitted Δν to the coil-calibrated gradient; if Δν shifts systematically as the decay shape changes, the four-parameter model is not the right null hypothesis for the background. Separately, record the spin noise spectrum with the probe beams forced to non-overlapping positions at unchanged optical depth: a clear broadening to the diffusion-limited multi-Lorentzian shape would confirm that the overlap, not just the multipass gain, is responsible for the near-Lorentzian line.","supporting_citations":[{"cited_title":"Portable magnetometry for detection of biomagnetism in ambient environments,","cited_arxiv_id":null,"evidence_quote":"supplies the prior multipass-cell magnetometry baseline and the diffusion-limited spin-noise behavior the present design improves on"},{"cited_title":"Cav- ity enhanced atomic magnetometry,","cited_arxiv_id":null,"evidence_quote":"derives the analytic diffusion correlation function used for the parameter-free spin-noise spectrum"},{"cited_title":"Furthermore, Eq","cited_arxiv_id":null,"evidence_quote":"justifies the matched-filter window for the real-time FFT analysis method"},{"cited_title":"Equivalence of the time-domain matched filter and the spectral-domain matched filter in one-dimensional nmr spectroscopy,","cited_arxiv_id":null,"evidence_quote":"provides the Faraday-rotation spin-noise spectroscopy method used to measure the intrinsic noise peak"},{"cited_title":"Simple dense-pattern optical multipass cells,","cited_arxiv_id":null,"evidence_quote":"supplies the theoretical optical-rotation noise estimate used to compare with the measured spin noise"}],"review_version":1}