{"id":"c49f050c-cf57-430c-bc82-17c16e267606","arxiv_id":"2504.21524","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A diamagnetically stabilized levitated magnet, read out optically, achieves 32 fT/√Hz magnetic field sensitivity at room temperature and in ambient magnetic fields.","lead":"A levitated magnet, held up by magnetic forces and a graphite stabilizer, acts as a room-temperature magnetometer with a measured sensitivity of 32 femtotesla per square-root hertz at about 305 hertz. The device combines ferromagnetic spin physics with optical readout to match SQUID and atomic magnetometer performance without cryogenics.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Absolute sensitivity is anchored entirely by the uncalibrated 10 pT drive-field amplitude; any offset in that amplitude propagates linearly into the quoted 32 fT/√Hz.","rationale":"The reader's weakest_assumption identifies exactly the point on which the entire absolute sensitivity scale rests. The stress-test pass confirms this is the most load-bearing concern about the central claim: every step from the response curve to the noise spectrum conversion uses the 10 pT drive amplitude as the reference, and no independent calibration is reported. The concern is concrete and quantitatively important—a 2× error in the drive field changes the quoted sensitivity by 2×, which would move the result from a several-orders-of-magnitude improvement over Ref. [27] to merely a large improvement, and it would alter the comparison with SQUID and SERF magnetometers. The paper does contain internal evidence that reduces the risk: the independently estimated Johnson noise floor (30 fT/√Hz) and the atomic-magnetometer validation mentioned in the text are consistent with the 32 fT/√Hz result. If the drive-field calibration were off by a large factor, the measured total noise would deviate from the sum of independently estimated sources. However, this consistency is only indirect: the atomic-magnetometer validation is not explicitly linked to the drive-field scale, and the quoted uncertainty does not include the calibration systematics. The missing Supplementary Text (referred to for the full frequency-range spectra and detailed noise analysis) further prevents full verification. This is not an internal inconsistency or a claim outside consensus; it is a missing calibration step that is standard for precision magnetometry and is fully addressable. The reader's conditional verdict remains appropriate, so no verdict change is recommended.","tokens_in":8330,"tokens_out":9774,"duration_ms":110863,"concrete_test":"Independently calibrate the field coils in situ: place a calibrated fluxgate sensor or a commercial atomic magnetometer at the sensor-magnet position (with the levitated magnet removed), drive the coils at 305 Hz with the same current used in the experiment, and record the actual alternating-field amplitude. Then recompute the response factor and the on-resonance sensitivity from the measured voltage noise using this independently measured field instead of the nominal 10 pT. If the resulting sensitivity changes by more than ~20%, the 32±3 fT/√Hz headline and the associated SQUID/SERF comparison require revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of 32±3 fT/√Hz at ~305 Hz is obtained by dividing the voltage noise spectrum by the response factor determined from a 10 pT alternating drive field. In 'Experimental Setup', the paper states: \"we apply an alternating magnetic field along the x-axis with an amplitude of 10 pT using a set of field coils placed inside the magnetic shielding.\" No calibration procedure, uncertainty, or in-situ verification of this field amplitude is given. The response curve in Fig. 2 and the conversion of voltage noise to magnetic noise both scale linearly with this assumed amplitude. If the actual field at the sensor magnet differs—due to coil constant error, frequency-dependent eddy-current screening by the magnetic shield or the EGPG, or field inhomogeneity over the magnet—the reported sensitivity changes proportionally. The '±3 fT/√Hz' quoted is only the statistical spread over five 100-s datasets, not a systematic calibration uncertainty. The paper mentions that the Johnson-noise estimate was validated with a commercial atomic magnetometer, but it never connects that validation to the drive-field calibration; the atomic-magnetometer measurement could in principle be used as an absolute reference, yet the text does not describe such a cross-check. Thus the headline number is not traceable to an absolute magnetic-field standard, and the comparison to SQUIDs and SERF magnetometers rests on this single unverified scale.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a room-temperature magnetometer based on a diamagnetically stabilized, magnetically levitated permanent-magnet disk (LeMaMa). A lifting magnet provides the magnetic force, an epoxy-glued pyrolytic graphite (EGPG) plate provides diamagnetic stabilization, and torsional oscillations of the disk are read out by an optical lever onto a quadrant photodiode. Applying a 10 pT alternating field from in-shield coils, the authors measure a response curve with resonance at f0 ≈ 304.84 Hz and Q ≈ 1.2 × 10^4, convert the voltage noise spectrum to a magnetic noise spectrum, and report a sensitivity of 32 ± 3 fT/√Hz at resonance (five 100-s datasets). They present a noise budget identifying Johnson noise from conducting surroundings (~30 fT/√Hz), air collisions (≤20), vibration (~10), and detection noise (~2), claim validation of the Johnson-noise estimate with a commercial atomic magnetometer, and demonstrate resonance tuning between 260 and 318 Hz by varying the bias field (1.42–2.12 mT). On this basis they argue the sensor is competitive with SQUID and SERF magnetometers while operating at room temperature and in mT-range fields, and discuss applications in bio-magnetometry and searches for exotic spin-dependent forces.","tokens_in":8590,"tokens_out":13126,"duration_ms":132297,"significance":"If the headline number survives scrutiny, this is a notable experimental advance: a room-temperature, optically read levitated magnet reaching 32 fT/√Hz at ~305 Hz is roughly four orders of magnitude better than the previous levitated reflection magnetometer (370 pT/√Hz, Ref. 27) and sits in the same range as SQUID and SERF devices, with the distinctive feature of operation in a mT-level bias field. Credit where due: the sensitivity is measured, not reverse-engineered—the response curve is fitted to a driven harmonic-oscillator model and the noise is converted by that same curve; the spin-correlation benefit is taken from prior theory (Refs. 24, 25) and is not fitted to the data; the noise decomposition is explicit; the frequency tunability (260–318 Hz) is directly demonstrated; and the projected 0.1 fT/√Hz limit gives a concrete falsifiable target. The principal barrier to accepting the quantitative claim is traceability of the absolute field scale, which rests on a single stated 10 pT drive amplitude without calibration uncertainty. The paper is well suited to this journal and will interest the levitodynamics and quantum-sensing communities.","major_comments":[{"comment":"The absolute scale of the headline sensitivity is anchored entirely by the 10 pT amplitude of the drive field, and that amplitude is never calibrated. In the 'Experimental Setup' section the text states 'we apply an alternating magnetic field along the x-axis with an amplitude of 10 pT using a set of field coils placed inside the magnetic shielding', but no coil geometry, current-to-field constant, in-situ verification, or uncertainty is provided. Since the voltage-noise spectrum is converted to a magnetic-noise spectrum by dividing by the fitted response curve A/sqrt((f^2 - f0^2)^2 + gamma^2 f^2) whose amplitude A is set by this 10 pT drive, any fractional error in the field actually seen by the 410-um-radius sensor magnet (eddy-current screening by the shield or the EGPG, field inhomogeneity across the magnet, current-setting error) propagates linearly into the quoted 32 ± 3 fT/√Hz and therefore into the comparison with SQUID and SERF sensors. A second internal ambiguity is that the drive is quoted as an 'amplitude' while the recorded output is described as 'root-mean-square (RMS)'; the peak-versus-RMS convention must be stated explicitly and used consistently in the noise conversion, because a √2 discrepancy is comparable to the reported statistical uncertainty. Please supply a calibration procedure with a quantified uncertainty, ideally cross-checked against the commercial atomic magnetometer already used for the Johnson-noise validation, or report the sensitivity with the associated systematic scale error.","section":"Experimental Setup; Fig. 2"},{"comment":"The quoted error of ±3 fT/√Hz is only the statistical spread over five 100-s datasets, so the error budget of the headline number is incomplete. Systematic contributions that should be propagated include the drive-field calibration discussed above, the fitted response amplitude A, the photodiode voltage calibration, and the stated dimensional tolerances (±2 um on radius and thickness), the last of which enter the resonance model f0 = sqrt(MB/I)/2π through the moment of inertia. I do not see circularity here: the sensitivity is measured, and the noise decomposition in the 'System sensitivity studies' section is explanatory rather than used to force the result. However, two supporting claims need hardened evidence: the statement that the Johnson-noise estimate is 'validated by measurements using a commercial atomic magnetometer' and the Monte-Carlo air-collision estimate of 'below 20 fT/√Hz' are described only verbally, with the details deferred to a 'Supplementary Text' that is not included in the posted preprint; these data must be part of the submitted manuscript so that the claimed noise floor near resonance can be checked.","section":"System sensitivity studies"}],"minor_comments":[{"comment":"The translational-mode peak frequencies are given as 3, 5, and 15 Hz for the x, y, and z axes in the main text but as 3, 4, and 10 Hz in the Fig. 3 caption; please reconcile the two sets of numbers.","section":"System sensitivity studies; Fig. 3 caption"},{"comment":"The color coding used for the noise components differs between the text ('solid purple line', 'green dashed line', 'black dotted line') and the Fig. 3 caption ('red line (rescaled from the actual noise background)', 'yellow line (determined by fitting)', 'black-dashed line'); unify the notation so the reader can match each component to its curve.","section":"System sensitivity studies; Fig. 3 caption"},{"comment":"When the quoted near-resonance components are combined in quadrature (30 fT/√Hz Johnson, ~20 fT/√Hz air collision, ~10 fT/√Hz vibration, ~2 fT/√Hz detection), the total is about 37 fT/√Hz, which is above the measured 32 ± 3 fT/√Hz; please state whether the component estimates are intended as upper bounds or are expected to add with some correlation.","section":"System sensitivity studies"},{"comment":"The claim of operation 'under Earth's magnetic field' and 'in an ambient environment' is not directly demonstrated, because all measurements are taken inside a four-layer magnetic shield; since the DC bias from the lifting magnet (1.92 mT) already exceeds the Earth's field, the relevant question is whether the shield is needed only for AC interference, and the text should say so explicitly or report an unshielded test.","section":"Abstract and Discussion"},{"comment":"Reference [34] is incomplete ('Measurement systems: application and design' with no authors, venue, or year), and the reference list mixes entries with and without DOIs; please normalize the bibliography.","section":"References"},{"comment":"The green response points in Fig. 2 are not described as single measurements or averages, and no error bars are shown; a short statement would let the reader judge the quality of the fit and the accuracy of the response amplitude A.","section":"Fig. 2"}],"recommendation":"major_revision","confidential_remarks":"To the editor: The experiment is impressive and, in my view, the manuscript's central difficulty is not the physics but the traceability of the absolute field scale. The fix is well within the scope of a revision: report the coil constant and the in-situ calibration of the 10 pT drive (the commercial atomic magnetometer already used for the Johnson-noise validation is the natural reference), fix the peak-versus-RMS convention, and make the supplement with the validation data available. The citations to Refs. 24–25 are appropriate, and the self-citation pattern (Budker is a coauthor) does not trouble me. If the calibration question is answered, this would be a strong contribution for this journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Ji et al. report a diamagnetically stabilized levitated magnet (LeMaMa) that hits 32 fT/√Hz at a ~305 Hz torsional resonance under room temperature and a ~1.9 mT bias field. That is a genuine record for levitated magnetometry at ambient conditions, about four orders of magnitude better than the cryogenic levitated-mirror magnetometer of Jiang 2020. The new combination is real: diamagnetic stabilization with an epoxy-glued pyrolytic graphite plate (EGPG) to cut eddy-current dissipation and Johnson noise, plus optical lever readout of the magnet's rotation. The authors also demonstrate tuning the resonance from 260 to 318 Hz by varying the bias field, and they give a careful noise budget with air-collision, vibrational, and detection noise contributions.\n\nThe paper does a lot right. The noise decomposition is physically sensible, and the claim that Johnson noise from the conductive plate drops from 110 to 30 fT/√Hz when the graphite is powdered is exactly the kind of testable design improvement that makes the result credible. The statistical uncertainty on the sensitivity is honestly reported as the spread over five 100-s datasets.\n\nThe soft spot is the one the reader flagged: the absolute sensitivity scale is set by a 10 pT calibration field whose amplitude is asserted without stated uncertainty or in-situ verification. The response curve in Fig. 2 and the conversion of voltage noise to magnetic noise both scale linearly with that amplitude. If the field at the magnet differs from 10 pT—due to coil constant error, eddy-current screening by the shield, or field inhomogeneity—the 32 fT/√Hz changes proportionally. The ±3 fT/√Hz is statistical only. The authors mention validating the Johnson-noise estimate with a commercial atomic magnetometer, but never connect that to the drive-field calibration, which is a missed opportunity to close the loop. This is not a reason to reject; it is a missing measurement that the authors can supply. But until they do, the comparison to SQUIDs and SERF magnetometers is not fully traceable.\n\nTwo smaller points. The supplementary text is referenced several times for the noise decomposition, but it is not included in the preprint, so the referee cannot check those details. And the 'ambient environment' framing should be qualified: the sensor sits in a vacuum chamber inside a four-layer magnetic shield, and it is a resonant detector, not a broadband one. The authors are honest about the resonant character, and the tuning demonstration helps, but the abstract's comparison to SQUID/SERF needs that caveat.\n\nBottom line: this is a clever, well-executed experiment with a plausible result. The calibration gap is the main obstacle, and it is addressable. I would send it to peer review with a request for calibration details and uncertainty, and ideally a data release of the noise spectra and response curves.","headline":"Promising room-temperature levitated magnetometer with a real sensitivity record, but the absolute scale depends on an uncalibrated calibration coil; referee it and ask for the missing calibration details.","tokens_in":9116,"tokens_out":3709,"would_cite":true,"duration_ms":38374,"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, magnetically levitated micro-magnet achieves a magnetic-field sensitivity of $32\\pm 3\\,\\mathrm{fT}/\\sqrt{\\mathrm{Hz}}$ at its torsional resonance.","keywords":["levitated magnetometer","femtotesla sensitivity","diamagnetic levitation","torsional resonator","Johnson noise","optical lever","room-temperature magnetometry","ambient environment"],"falsifier":"Drive the calibration coils with the same current while measuring the field at the sensor location with an independently calibrated magnetometer, then recompute the response curve using the measured amplitude; if it is not 10 pT, the quoted $32\\,\\mathrm{fT}/\\sqrt{\\mathrm{Hz}}$ changes by the same factor.","tokens_in":8148,"feed_emoji":"🧲","tokens_out":11141,"duration_ms":100443,"temperature":0.7,"pith_summary":"The paper claims that a tiny ferromagnetic disk, levitated at room temperature by magnetic forces and stabilized by a diamagnetic graphite layer, can detect magnetic fields at $32\\pm 3\\,\\mathrm{fT}/\\sqrt{\\mathrm{Hz}}$ on its torsional resonance. The point of the claim is that femtotesla-class magnetometry, usually the territory of cryogenic SQUIDs or heated atomic-vapor cells, can work in an ambient, Earth-field-compatible setup. The authors show that the dominant noise is Johnson magnetic noise from surrounding conductors, and that grinding pyrolytic graphite into epoxy-isolated particles cuts that noise roughly fourfold. If the result stands, it gives biology, chemistry, and fundamental physics a magnetometer that can sit close to a sample without cryogens or magnetic-field suppression.","feed_headline":"A levitated magnet senses 32 fT/√Hz without cryogenics","feed_subtitle":"Diamagnetic stabilization and a 0.2 mm³ magnet bring SQUID-class magnetometry to ambient air.","key_machinery":"The load-bearing object is a torsional oscillator: a magnetized disk levitated in a potential well formed by gravity, the field of a lifting magnet, and the repulsion of a diamagnetic pyrolytic graphite plate. A magnetic field along the $x$-axis exerts a torque that tilts the disk, and a $3.4\\,\\mathrm{m}$ optical lever magnifies the rotation on a quadrant photodiode. The resonant torsional mode at $f_0=\\sqrt{MB/I}/2\\pi$ concentrates the response, while the ferromagnet's strong spin-lattice coupling rapidly averages spin-projection noise; the graphite is ground to powder and epoxy-bonded so eddy currents and the associated Johnson noise are confined to small loops.","core_discovery":"The central result is that a levitated $0.2\\,\\mathrm{mm}^3$ ferromagnetic magnet, read out by an optical lever, reaches a magnetic field sensitivity of $32\\pm 3\\,\\mathrm{fT}/\\sqrt{\\mathrm{Hz}}$ at a torsional resonance near $305\\,\\mathrm{Hz}$ while operating at room temperature, at $0.025\\,\\mathrm{mbar}$ pressure, and in magnetic fields up to the millitesla range. The sensitivity comes from combining high spin density and spin-lattice coupling in the ferromagnet, which suppresses spin-projection noise, with a low-dissipation levitation trap made of a lifting magnet and an epoxy-glued pyrolytic graphite stabilizer. The paper reports that the measured performance is several orders of magnitude better than the earlier reflection-based superconducting levitated magnetometer and comparable to similarly sized SERF and SQUID devices, but without their cryogenic or low-field constraints.","pith_inferences":["The headline sensitivity is a resonant, narrow-band number: the same sensor shows noise near $1\\,\\mathrm{nT}/\\sqrt{\\mathrm{Hz}}$ below 15 Hz and hundreds of $\\mathrm{pT}/\\sqrt{\\mathrm{Hz}}$ between 80 and 200 Hz, so the practical advantage is for signals that can be modulated onto the torsional resonance.","The stated 10 pT calibration amplitude carries no uncertainty and no independent verification at the sensor location; if the true field differs, the quoted sensitivity shifts by the same factor, and a cross-calibration with a commercial atomic magnetometer would settle it.","The trick of granularizing a conductor to shrink eddy-current loops is a general noise-suppression principle that could also be applied to magnetic shields and other conductive structures near any precision magnetometer."],"forward_implications":["A room-temperature sensor runs in Earth's field and in millitesla-range backgrounds, avoiding the shielded low-field environments needed by SERF magnetometers and the cryogens needed by SQUIDs.","The resonance frequency is tunable from about 260 Hz to 318 Hz by changing the bias field along $z$, so the same device can be used for frequency-modulated detection or closed-loop operation.","Because the sensing element is a bare magnet with no cell or window, a signal source can be placed within a millimeter of it, strengthening biomagnetic measurements and short-range searches for spin-dependent forces such as axion-mediated interactions.","The authors estimate that replacing the graphite stabilizer with lower-noise materials, adding high-reflectivity coatings, and using better photodiodes could push sensitivity toward $0.1\\,\\mathrm{fT}/\\sqrt{\\mathrm{Hz}}$, where vibration noise would dominate."],"supporting_citations":[{"why":"Supplies the diamagnetically stabilized magnet levitation scheme used to create the trap for the sensor magnet.","marker":"[30]"},{"why":"Provides the prior superconducting levitated reflection magnetometer whose sensitivity the LeMaMa improves on by several orders of magnitude.","marker":"[27]"},{"why":"Reports a superconducting levitated ferromagnetic magnetometer reaching 20 fT/√Hz with SQUID readout, the cryogenic comparison point for the new ambient sensor.","marker":"[20]"},{"why":"Gives the theoretical basis for ferromagnetic spin correlation reducing spin-projection noise, motivating the sensor's spin-lattice relaxation advantage.","marker":"[24]"},{"why":"Shows ferromagnetic torque sensors can surpass the standard energy resolution limit, supporting the claim that thermal noise is not dominant for sensors above roughly 100 µm.","marker":"[25]"},{"why":"Introduces diamagnetic composites that reduce eddy-current dissipation, the basis of the epoxy-glued pyrolytic graphite stabilizer.","marker":"[31]"},{"why":"Provides the Johnson magnetic noise calculation used to identify and quantify the dominant noise from conducting materials.","marker":"[32]"},{"why":"Represents the atomic/SERF magnetometer baseline used for the comparison of sensitivity and operating constraints.","marker":"[36]"},{"why":"Represents the SQUID magnetometer baseline used for the comparison of sensitivity and environmental adaptability.","marker":"[37]"}],"fun_headline_variants":["Levitated magnet hits 32 fT/√Hz at room temperature","SQUID-class sensitivity from a levitated magnet, no cryo","Ambient levitated magnet matches SQUID sensitivity","No cryo: levitated magnet hits 32 fT/√Hz"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"If the field coils actually produce a field stronger or weaker than the stated 10 pT, the quoted sensitivity is off by the same factor, and the paper does not report how that 10 pT value was verified at the sensor magnet.","fun_headline_variants_meta":{"raw":{"variants":["Levitated magnet hits 32 fT/√Hz at room temperature","SQUID-class sensitivity from a levitated magnet, no cryo","Ambient levitated magnet matches SQUID sensitivity","No cryo: levitated magnet hits 32 fT/√Hz"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00055,"raw_usage":{"total_tokens":2597,"prompt_tokens":890,"completion_tokens":1707,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":506,"completion_tokens_details":{"reasoning_tokens":1632}},"tokens_in":506,"tokens_out":1707,"duration_ms":12259,"temperature":1.0,"reasoning_tokens":1632,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:00:50.288161+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Drive the calibration coils with the same current while measuring the field at the sensor location with an independently calibrated magnetometer, then recompute the response curve using the measured amplitude; if it is not 10 pT, the quoted $32\\,\\mathrm{fT}/\\sqrt{\\mathrm{Hz}}$ changes by the same factor.","supporting_citations":[{"cited_title":"Simon, L","cited_arxiv_id":null,"evidence_quote":"Supplies the diamagnetically stabilized magnet levitation scheme used to create the trap for the sensor magnet."},{"cited_title":"Jiang, J","cited_arxiv_id":null,"evidence_quote":"Provides the prior superconducting levitated reflection magnetometer whose sensitivity the LeMaMa improves on by several orders of magnitude."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the theoretical basis for ferromagnetic spin correlation reducing spin-projection noise, motivating the sensor's spin-lattice relaxation advantage."},{"cited_title":"Vinante, et al., Surpassing the Energy Resolution Limit with ferromagnetic torque sensors","cited_arxiv_id":null,"evidence_quote":"Shows ferromagnetic torque sensors can surpass the standard energy resolution limit, supporting the claim that thermal noise is not dominant for sensors above roughly 100 µm."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces diamagnetic composites that reduce eddy-current dissipation, the basis of the epoxy-glued pyrolytic graphite stabilizer."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Johnson magnetic noise calculation used to identify and quantify the dominant noise from conducting materials."},{"cited_title":"Kitching, Chip-scale atomic devices","cited_arxiv_id":null,"evidence_quote":"Represents the atomic/SERF magnetometer baseline used for the comparison of sensitivity and operating constraints."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Represents the SQUID magnetometer baseline used for the comparison of sensitivity and environmental adaptability."}],"review_version":1}