{"id":"bbcc2518-d31b-4a21-9440-670306717d4d","arxiv_id":"2607.24490","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.5,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"A ³He nuclear spin oscillator with artificial feedback reaches ~4×10⁻¹⁹ eV energy sensitivity in one second, about five times better than GNOME alkali magnetometers.","lead":"A lab built a self-oscillating ³He nuclear spin sensor and measured roughly five times better energy sensitivity to exotic nuclear spin couplings than the alkali magnetometers used in the GNOME dark-matter network. It is a concrete hardware step for global searches for ultralight bosonic dark matter, though co-magnetometers already outperform it.","discovery_kind":"extension","skeptic_critique":{"model":"moonshotai/kimi-k3","headline":"The δB≈3 pT \"one-second sensitivity\" is a frequency-fit noise floor at 1-s binning, but the oscillator's 4 mHz linewidth implies a response time of tens of seconds — the paper never demonstrates sensitivity to a magnetic transient actually lasting ~1 s, which is the signal class the GNOME comparison","rationale":"The reader correctly localized the weakness to the δB≈3 pT figure but framed it as a metrology issue (single run, no uncertainty, no cross-check). That is a real but secondary concern: even with pristine error bars, the figure answers the wrong question unless the device's bandwidth is established. The load-bearing condition for the strongest claim is that \"sensitivity in 1 s of averaging\" equals \"sensitivity to a 1-s exotic-field transient,\" and the paper's own data (4 mHz linewidth; only a 40 s FWHM pulse demonstrated) suggest the response time is tens of seconds. This is a correctness-of-comparison concern, not a consensus disagreement, and it is testable with equipment and a pipeline the authors already have: one calibrated fast-pulse run or one numerical integration of their own Bloch model settles it. I keep the verdict CONDITIONAL (same level as the reader) but the condition should be expanded: in addition to uncertainty quantification of the Allan-deviation point, the paper needs a bandwidth/transfer-function measurement before the 5× headline can be read as applying to GNOME-relevant transient signals. If the fast-pulse test shows full response at 1 s, the concern dissolves and only the reader's error-bar condition remains; if it shows strong attenuation, the claim survives only in weakened, slow-signal form. Credit where due: the apparatus, analysis pipeline, and model are described concretely enough that this check is straightforward, and the authors are transparent about the more sensitive co-magnetometer alternative, so the stakes of the claim are modest.","tokens_in":10122,"tokens_out":3082,"duration_ms":100544,"concrete_test":"Measure the transient transfer function directly: apply calibrated z-directed pulses (same coils as Fig. 3) with known amplitude ~100 pT and FWHM of 1 s, 5 s, and 40 s, then run the identical 1-s binning/sinusoidal-fit pipeline used for Fig. 3b and compare recovered peak ΔB to applied amplitude. Alternatively (analytically), integrate Eqs. (7)–(9) with the fitted parameters from Fig. 2b, adding B_0(t)=B_0+δB·L(t) for Lorentzians of those widths. If the recovered 1-s-pulse amplitude is attenuated by more than ~30% relative to the 40-s pulse, the \"one-second energy sensitivity\" in Eq. 15 does not describe sensitivity to 1-s exotic transients, and the factor-of-five claim must be restated for slow (≳100 s) signals — where the K-magnetometer comparison figure (quoted at 1 s) would also need recalculation at matched duration.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The factor-of-five claim (Abstract; §5, Eqs. 15–16) rests on equating \"sensitivity in one second of averaging\" with sensitivity to a one-second-duration exotic-field transient. These are not the same for a feedback-driven oscillator. The Allan deviation in Fig. 4b characterizes the stability of the fitted oscillation frequency when the field is (assumed) static; it says nothing about the device's transfer function for a time-varying B_z. Two internal facts indicate the bandwidth may be far below 1 Hz: (1) the measured FWHM linewidth is ≈4 mHz (§4, Fig. 4a), corresponding to a natural response time of order 1/(π·Δf) ≈ 80 s; (2) the only pulse-response validation (Fig. 3b) uses a Lorentzian pulse with 40 s FWHM — safely inside the slow regime where the oscillator can adiabatically track. In a nuclear spin maser with artificial feedback, an external field step must pull the sustained oscillation through the feedback loop; for perturbations fast compared with the narrowed linewidth, the response is attenuated and phase-shifted. If the transfer function at 1-s duration suppresses the signal by even a factor of a few, then δB≈3 pT is not the relevant figure for 1-s exotic-coupling pulses, and the comparison to the K magnetometer's \"≈0.1 pT in one second\" (an open-loop SERF device with kHz-scale bandwidth, where 1-s averaging genuinely means 1-s signal sensitivity) is not like-for-like. The numbers in Eqs. 15–16 are internally consistent (γδB with γ=32.434 Hz/µT gives 4.0×10⁻¹⁹ eV; g_S μ_B δB/q with q=6 gives 1.9×10⁻¹⁸ eV; ratio 4.8), so the arithmetic is not the issue — the mapping from a static-noise figure to a transient-sensitivity figure is. This is distinct from, and more basic than, the reader's concern about missing error bars: even a perfectly characterized 3 pT Allan deviation at τ=1 s would not rescue the claim if the bandwidth is inadequate.","agreement_with_reader":"partial"},"referee_report":{"model":"moonshotai/kimi-k3","summary":"The manuscript reports the construction and characterization of a nuclear spin oscillator (spin maser) based on 3He, using hybrid K/Rb spin-exchange optical pumping in a shielded vapor cell, with the 3He precession read out via Faraday rotation on potassium and sustained by analog feedback to a transverse field coil. The authors demonstrate self-sustained oscillation at 8.87 Hz in a 273 nT leading field, show qualitative agreement of the startup dynamics with Bloch-equation numerics (Fig. 2), recover an applied 40 s FWHM Lorentzian Bz pulse of 100 pT amplitude (Fig. 3), and characterize the device with a power spectral density (≈4 mHz FWHM linewidth) and an Allan deviation reaching ≈3 pT at 1 s averaging (Fig. 4). From δB≈3 pT they derive an energy sensitivity δE≈4×10⁻¹⁹ eV at 1 s (Eq. 15), roughly five times better than the potassium SERF magnetometers used in GNOME (Eq. 16), and they candidly note that noble-gas comagnetometers remain roughly two orders of magnitude better still. The claimed application is transient exotic-spin-coupling searches of the GNOME type.","tokens_in":10495,"tokens_out":5484,"duration_ms":179832,"significance":"If the sensitivity figure holds, this is a credible small-scale demonstration of a new detector class for the GNOME program: the first 3He-based feedback nuclear spin oscillator with hybrid K/Rb pumping, with exceptionally long transverse coherence (T₂* up to 1500 s) and a genuinely calibration-free frequency-to-field conversion via f = γB (Eq. 1), which is a real methodological strength for cross-network comparisons. The recovery of a GNOME-style Lorentzian pulse (Fig. 3b) is a concrete, falsifiable benchmark, and the honest placement of the device relative to comagnetometers (§5, final paragraph) is to the authors' credit. The energy-sensitivity arithmetic in Eqs. (15)–(16) is internally consistent. The impact is incremental rather than transformative — the authors themselves point to comagnetometers as the better long-term GNOME detector — but a 3He maser with neutron-dominated coupling (Eq. 14) fills a distinct niche and the result merits publication once the quantitative sensitivity claim is properly substantiated.","major_comments":[{"comment":"The headline claim (Abstract; §5, Eqs. 15–16) that the device has ≈5× higher sensitivity 'to exotic spin couplings' than GNOME alkali magnetometers equates the 1-s Allan-deviation floor of the fitted frequency (Fig. 4b, measured under a static field) with sensitivity to a transient of ~1 s duration. These need not coincide for a feedback-narrowed oscillator: the measured linewidth is ≈4 mHz (§4, Fig. 4a), and the only transient-response validation uses a Lorentzian pulse of 40 s FWHM (Fig. 3b) — firmly in the adiabatic regime. The GNOME domain-wall signals the paper targets include crossing times of order seconds, where feedback-loop response (gain α, phase Δφ) can attenuate or phase-shift the signal. This may well be answerable — the precession frequency tracks B instantaneously and the 1-s-bin fits may already capture fast transients — but as written it is an assumption, not a demonstr","section":"§5, Abstract; Fig. 3b vs Fig. 4b"},{"comment":"The value δB ≈ 3 pT at 1 s, which carries the entire quantitative conclusion of §5, is read from the Allan deviation of a single four-hour run at one leading field (273 nT), with no stated uncertainty, no noise budget, and no discussion of what sets the floor (photon shot noise, fit error on 512-point/~9-cycle sinusoid fits, feedback electronics, residual field drift). Please report the value with an uncertainty, show it is reproduced across runs and leading fields, and identify the limiting noise source at 1 s. Relatedly, the Allan upturn beyond ≈400 s is attributed to drift; since the oscillation is feedback-sustained, the stability of the feedback gain α and phase shifter Δφ is a plausible systematic for frequency drift and should be discussed — a phase drift in the feedback loop directly pulls the maser frequency and would masquerade as a field drift.","section":"§4, Fig. 4b; §5"},{"comment":"The abstract states sensitivity 'to exotic spin couplings that is ≈5 times higher' than the GNOME alkali magnetometers, but Eqs. (13)–(14) show the two devices couple to different nucleons: K to the proton (χ ≈ −0.2χ_p), 3He predominantly to the neutron (0.87χ_n). The factor of five is an energy-sensitivity ratio only; for a neutron-coupling search the improvement over K would be far larger than five, and for a proton-coupling search it may be absent or reversed. Since this is the paper's headline sentence, it should be reworded to state precisely what is being compared (energy sensitivity at 1 s averaging), with the coupling-dependent reach stated separately for χ_n and χ_p.","section":"Abstract; §5, Eqs. (13)–(16)"}],"minor_comments":[{"comment":"Inconsistency in the Bloch-model notation: the text below Eq. (5) states that T₁ and the pumping rate G have been combined into an effective T₁* and equilibrium polarization P₀, but Eq. (9) then uses T₁ and G as separate quantities. Please make the two parameterizations consistent or explicitly define the mapping. Also, the numerical values of T₂, α, G, and P₀ used to produce Fig. 2b are never given; since the agreement is only qualitative this is not load-bearing, but reporting them would aid reproducibility.","section":"§2, Eqs. (5) and (9); Fig. 2"},{"comment":"In Eq. (15) the conversion δE = γδB uses γ in Hz/µT; a factor of h is implicit. Please write δE = hγδB or state the unit convention explicitly.","section":"§5, Eq. (15)"},{"comment":"The ≈4 mHz FWHM linewidth is described only as an estimate from the PSD. State how it was extracted (fit function, resolution bandwidth of the 4-hour FFT), since it bears on the response-time discussion requested in the major comments.","section":"§4, Fig. 4a"},{"comment":"Fig. 4b: please verify the correspondence between the left (pT) and right (mHz) axes; at γ = 32.434 Hz/µT, 1 mHz ↔ 30.8 pT, and the two scales as drawn do not appear consistent with this conversion.","section":"Fig. 4b"},{"comment":"Fig. 2: the horizontal time axes appear to run in descending order (3000 → 0 s). If this is intentional it should be explained; otherwise please correct.","section":"Fig. 2"},{"comment":"The 0.1 pT (1 s) potassium-magnetometer figure in Eq. (16) is cited to an Oberlin bachelor's thesis (Ref. [20]). A peer-reviewed GNOME sensitivity reference (e.g., from Refs. [4]–[6] or associated instrument papers) would strengthen the comparison, or the thesis number should be independently justified.","section":"§5, Ref. [20]"},{"comment":"Typos: 'Equastions (7)-(9)' (§4); 'valance protons' → 'valence' (§5); Ref. [12] 'serach' → 'search'; Ref. [15] 'fundmamental' → 'fundamental'; Ref. [20] 'Single Bean SERF Magnetometer' presumably 'Single Beam'.","section":"Various"}],"recommendation":"major_revision","confidential_remarks":"The work is competent, honestly framed in the body, and within scope for an instrumentation/AMO venue. The main risk is the abstract-level 5× sensitivity claim, which currently rests on a single uncharacterized Allan-deviation point and an apples-to-oranges energy-sensitivity comparison across different nucleon couplings; both are fixable with one additional short-pulse measurement (or transfer-function analysis) and revised wording. The citation of the group's own undergraduate thesis (Ref. [20]) for the GNOME comparison number is worth asking the authors to upgrade. I do not see grounds for rejection; the revision requested is well within the apparatus's demonstrated capability."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"This is a clean experimental methods paper: they built and ran an artificial-feedback nuclear spin oscillator on ³He, not Xe, and measured its frequency stability. That combination is the actual novelty. Hybrid SEOP (Rb pump, K probe), long T2* (~1500 s free induction), self-sustained oscillation at ~9 Hz, and a direct frequency-to-Bz readout are all executed competently. Startup tracks the Bloch equations at least qualitatively, a 40 s Lorentzian Bz pulse is recovered cleanly in the fitted frequency, and they show PSD linewidth plus Allan deviation. The energy conversion arithmetic (γ δB vs gS μB δB / q) is correct and the citations to the Xe-maser and SEOP literature are appropriate. They also state plainly that alkali–noble-gas co-magnetometers already beat this by orders of magnitude, which is the right level of honesty.\n\nThe soft spots are real but bounded. The δB ≈ 3 pT at 1 s is taken from one multi-hour Allan curve with no uncertainty or noise budget; that is thin for a factor-of-five claim. More importantly, the stress-test point lands: the oscillator linewidth is ~4 mHz and the only demonstrated pulse is 40 s FWHM. For a feedback maser that is not the same thing as open-loop SERF bandwidth. A 1 s exotic transient (the GNOME domain-wall class) may be attenuated by the narrowed response, so equating “Allan floor at τ = 1 s” with “sensitivity to a 1 s signal” is not automatic. The paper never shows the transfer function. Bloch parameters are hand-tuned for a qualitative match only; that is fine because they are not load-bearing for the sensitivity number.\n\nWho it is for: people already inside GNOME or building noble-gas sensors who want a concrete ³He oscillator data point. It does not set a new dark-matter limit and does not claim to. It deserves a serious referee—methods are reproducible enough, data are shown, and the comparison is explicit even if the bandwidth mapping needs tightening. I would engage if I were optimizing the next GNOME station; otherwise it is a useful reference, not a must-read.","headline":"Real ³He feedback oscillator with honest GNOME benchmarking, but the headline 5× claim rests on a static Allan floor that may not map to the 1 s transients GNOME actually cares about.","tokens_in":11605,"tokens_out":569,"would_cite":false,"duration_ms":19776,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"A helium-3 nuclear spin oscillator reaches about five times better energy sensitivity to exotic spin couplings than the alkali magnetometers used in dark-matter searches.","keywords":["nuclear spin oscillator","exotic spin coupling","magnetometry","helium-3","Faraday rotation","spin-exchange optical pumping","Larmor precession"],"falsifier":"An independent, calibrated measurement of the same cell’s magnetic noise floor that finds a one-second sensitivity worse than about 15 picotesla—or an energy sensitivity no better than the alkali comparison value—would overturn the quantitative claim.","tokens_in":11210,"feed_emoji":"🧲","tokens_out":963,"duration_ms":51460,"temperature":0.7,"pith_summary":"This paper builds and characterizes a self-sustaining nuclear spin oscillator that uses polarized helium-3 nuclei as the sensor. A hybrid alkali vapor is optically pumped, the helium-3 is polarized by spin exchange, and Faraday rotation on potassium reads the nuclear precession; that signal is filtered and fed back as a transverse magnetic field so the spins keep oscillating at a frequency strictly proportional to the leading field. From multi-hour data the authors extract a one-second magnetic sensitivity near 3 picotesla, corresponding to an energy sensitivity of roughly 4×10⁻¹⁹ eV—about five times better than the alkali sensors previously fielded for the same class of exotic-physics searches. A reader cares because the longer nuclear coherence and higher density of helium-3, together with a frequency readout that needs no separate calibration, offer a concrete path to weaker spin-dependent dark-matter couplings. The work is presented as a candidate detector for coordinated global searches for transient exotic spin signals.","feed_headline":"Helium-3 oscillator beats alkali sensors by 5× on exotic spins","feed_subtitle":"Self-sustained Larmor feedback yields roughly 4×10⁻¹⁹ eV energy sensitivity in one second","key_machinery":"The nuclear spin oscillator with artificial feedback: the Faraday-rotation signal proportional to the transverse helium-3 polarization is filtered, phase-shifted, amplified, and returned as a magnetic field By = α Px, locking the spins into self-sustained Larmor oscillation at frequency f = γB. Bloch equations with that feedback term carry the start-up dynamics and steady-state behavior.","core_discovery":"A feedback-driven helium-3 nuclear spin oscillator, polarized by hybrid alkali optical pumping and read out by Faraday rotation, achieves a one-second energy sensitivity of about 4×10⁻¹⁹ eV to spin-dependent exotic fields—approximately five times higher than the energy sensitivity of the potassium alkali magnetometers previously used in the same exotic-physics searches.","pith_inferences":["Because the Allan deviation turns over near 400 s, multi-station coincidence searches that rely on hour-scale stability will still need active drift control or frequent leading-field recalibration.","The same hybrid-pumped cell could host a second noble-gas species, turning the oscillator into a differential pair that cancels ordinary magnetic noise while retaining exotic-signal contrast.","The direct γB frequency relation makes several fielded units natural absolute time-stamps for transient exotic events once they are networked.","If residual gradients, not intrinsic spin noise, still limit T2*, further coil redesign could push the energy sensitivity closer to the comagnetometer regime the authors themselves cite as superior."],"forward_implications":["Frequency readout of exotic spin couplings needs no separate magnetometer scale-factor calibration.","Helium-3’s neutron-dominated nuclear spin content complements alkali sensors that couple mainly to protons.","Demonstrated free-induction decay times up to 1500 s and oscillator linewidths near 4 mHz are longer-lived than typical xenon nuclear oscillators.","Lorentzian magnetic pulses of the type sought in domain-wall dark-matter searches are recovered cleanly in the oscillator frequency record.","The achieved energy sensitivity remains weaker than existing alkali–noble-gas comagnetometers, so the oscillator is a modest rather than leading upgrade path."],"fun_headline_variants":["³He nuclear oscillator beats alkali sensors 5× on exotic spins","Feedback-driven helium-3 spin oscillator gains 5× exotic-field reach","Self-sustained ³He Larmor oscillator tops K magnetometers fivefold","Hybrid-pumped ³He oscillator yields 5× better exotic spin sensitivity","Helium-3 nuclear spin feedback sharpens exotic coupling searches 5×"],"cache_read_input_tokens":128,"weakest_assumption_plain":"The factor-of-five sensitivity claim rests on reading a single multi-hour Allan-deviation curve as roughly 3 picotesla in one second, without a noise budget or an independent calibrated cross-check.","fun_headline_variants_meta":{"raw":{"variants":["³He nuclear oscillator beats alkali sensors 5× on exotic spins","Feedback-driven helium-3 spin oscillator gains 5× exotic-field reach","Self-sustained ³He Larmor oscillator tops K magnetometers fivefold","Hybrid-pumped ³He oscillator yields 5× better exotic spin sensitivity","Helium-3 nuclear spin feedback sharpens exotic coupling searches 5×"]},"model":"grok-4.5","effort":"low","cost_usd":0.004105,"raw_usage":{"total_tokens":1276,"prompt_tokens":779,"num_sources_used":0,"completion_tokens":85,"cost_in_usd_ticks":41048000,"prompt_tokens_details":{"text_tokens":779,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":412,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":779,"tokens_out":85,"duration_ms":7900,"temperature":1.0,"reasoning_tokens":412,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-31T13:20:26.228098+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"An independent, calibrated measurement of the same cell’s magnetic noise floor that finds a one-second sensitivity worse than about 15 picotesla—or an energy sensitivity no better than the alkali comparison value—would overturn the quantitative claim.","supporting_citations":[],"review_version":1}