{"id":"79d8b954-eb80-4a30-845b-8c35f8d46ae8","arxiv_id":"2501.08936","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"The authors demonstrate compact sealed 3He NMR cells with reduced T1 relaxation for cryogenic magnetometry, and project relative field sensitivities of 10^-11 to 10^-7 over 4 to 300 K and 0.1 to 7 T.","lead":"Sealed, high-pressure 3He gas cells are shown to work as NMR magnetic field probes from room temperature down to 4.2 Kelvin, with added oxygen or silica gel cutting the slow spin relaxation that previously made such sensors impractical. The paper maps the expected single-pulse relative field sensitivity across temperature and field, claiming 10^-11 to 10^-7 for fields above 0.1 Tesla.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The low-field/high-temperature edge of the claimed sensitivity range rests on unvalidated T2*-vs-B and Q scalings; a second-field test is required.","rationale":"The paper's experimental core is credible: seven sealed cells, T1 reduction across a wide temperature range, and reproducible SNR data. The strongest claim, however, is not the engineering but the projected sensitivity range in Fig. 12, which is obtained by scaling the single 7.05 T measurement to 0.1 T using Eq. 15. The reader's weakest_assumption correctly names this scaling law. My stress-test sharpens it: the T2* scaling is the most consequential term, because the CRLB depends steeply on T2*, and constant-relative-inhomogeneity is not a safe assumption when moving to low fields where absolute gradients, sample susceptibility, and external field fluctuations dominate. A modest failure of that assumption changes the projected sensitivity by orders of magnitude, so the headline '10^-11 < δB/B < 10^-7 in a single pulse' is not yet supported across the full B–T plane. The paper itself warrants this caution in Sec. 4.3 by noting that absolute field measurement requires spherical cells and susceptibility corrections. A single additional measurement at a lower field would resolve whether the projection is trustworthy. Because the reader already assigned CONDITIONAL, my concern does not move the verdict; it strengthens the same condition.","tokens_in":18102,"tokens_out":10334,"duration_ms":105267,"concrete_test":"Mount cell #4 (30 bar 3He, 5 bar O2) in a second, independently characterized magnet at a lower field, e.g. a 1.5 T MRI or calibration magnet at 300 K, using the same RF coil and a matched circuit with Q ≈ 100. Measure SNR, T2*, and Larmor frequency from a single π/2 pulse and compare with Eq. 15 using the measured Q. If SNR deviates by more than 2×, or if T2* is not consistent with the B0/B scaling (e.g. remains 1–5 ms rather than about 4.7 ms), the Fig. 12 sensitivity map is not supported at the low-field/high-temperature corner. If an absolute field claim is retained, repeat in a spherical cell against a current-regulated reference magnet to verify field accuracy.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4.3 derives Fig. 12 from Eq. 15 using several extrapolation assumptions, none checked outside the single 7.05 T dataset. The most load-bearing is assumption (b): T2* = (T2*)0 (B0/B), i.e. a constant 0.7 ppm relative inhomogeneity. At 0.1 T this gives T2* ≈ 70 ms instead of the ≈ 1 ms measured at 7.05 T. Since the CRLB frequency variance scales as 1/(SNR^2 * T2*^3) for optimal acquisition (Eqs. 5 and 6 with t = 3T2*), a field-independent absolute inhomogeneity of similar magnitude would shorten T2* by roughly 70× and degrade δB/B by roughly 600× at 0.1 T, moving the high-temperature edge from about 10^-7 to well above 10^-4. Assumption (a), Q(B,T) = 100, is also unvalidated; the measured Q0 = 46 required resistive spoiling to stabilize it, and Q at 3 MHz is not demonstrated. Thus the single-pulse part of the claim at low B is an optimistic projection, not a measurement. Separately, Sec. 4.3 concedes that absolute field determination requires spherical cells and susceptibility corrections, so the abstract's 'precision magnetometry' should be read as relative field monitoring until those are implemented.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents sealed, high-pressure gaseous 3He cells as NMR probes for field monitoring from 4 K to 300 K at B > 0.1 T. Quartz cells (Vs = 0.126 cm3) are filled via helium permeation to 30 bar (100 bar projected), and three strategies reduce the long T1 of thermally polarized 3He: 5 bar of O2 released from SrO2, silica gel with high surface-to-volume ratio, and Gd3+-loaded silica gel. At 7.05 T the 30 bar cell yields SNR = 70 at 300 K and 3320 at 4.2 K; SNR scales linearly with pressure and as T^-0.94, and cells are stable over two months. T1(T) is mapped across the O2 gas/liquid/solid transitions (about 0.4 s in the gas phase, about 20 min near 30 K, about 10 s at 4.2 K), while T2* stays within a factor of about 2 over the full range; the silica gel data are described by a BPP model. Using the Cramér-Rao lower bound (Eq. 5), single-pulse sensitivities are computed from measured FIDs (Fig. 11), and Eq. 15 extrapolates from the single 7.05 T anchor to 0.1 T ≤ B ≤ 7 T, assuming constant Q = 100, SNR ∝ B^2/T, and T2* ∝ B0/B at fixed 0.7 ppm relative inhomogeneity. This extrapolation generates the headline range 10^-11 < δB/B < 10^-7 in Fig. 12.","tokens_in":18464,"tokens_out":24455,"duration_ms":239496,"significance":"If the headline claim holds, the work closes a genuine gap: no compact NMR field probe currently operates continuously from 4 K to 300 K, where liquid/solid 1H probes fail. The measured SNR improvement of about 47 at cryogenic temperature at 7.05 T, the linear pressure dependence, the two-month stability, and the T1 engineering across the O2 phase transitions are concrete, reproducible experimental results; the silica gel/Gd3+ BPP analysis is a useful addition to the 3He relaxation literature. The paper is transparent about its limits: the CRLB is labeled approximate for the non-Lorentzian 4.2 K FID, the absolute-field caveat is stated in Section 4.3, and the five extrapolation assumptions behind Eq. 15 are listed explicitly. The Fig. 12 envelope is a crisp, falsifiable prediction at 0.1–1 T, which makes a low-field test the natural route to validating the central claim.","major_comments":[{"comment":"The headline range 10^-11 < δB/B < 10^-7 in the abstract and Section 4.3 is a CRLB extrapolation, not a measured result: every experimental input is anchored at the single field B0 = 7.05 T. Assumption (b) of Eq. (15), T2* = (T2*)0·(B0/B) with a constant relative inhomogeneity of 0.7 ppm, is the most load-bearing. At 0.1 T it gives T2* ≈ 70 ms instead of the ≈1 ms measured at 7.05 T, and because the CRLB frequency variance scales as 1/(SNR²·T2*³) for t ≈ 3T2* (Eqs. 5 and 6), δB/B ∝ 1/(SNR·T2*^{3/2}). If the absolute inhomogeneity rather than the relative one were field-independent, T2* would stay ≈1 ms at 0.1 T and the low-field edge of Fig. 12 would degrade by roughly 600×, moving the high-temperature edge at 0.1 T from ≈10^-7 toward 10^-4. The paper contains no measurement below 7.05 T that distinguishes these scalings. I recommend either (i) adding a low-field (0.1–1 T) SNR and T2* measurement with the same cells, or (ii) presenting Fig. 12 explicitly as a projection, with the T2*(B) assumption varied to show the sensitivity of the envelope, and rewording the abstract claim accordingly.","section":"Sec. 4.3, Eq. (15), Fig. 12"},{"comment":"Assumptions (a) and (c) of Eq. (15) are mutually inconsistent. The text states that the quadratic B-dependence of the SNR results from Eq. 3, noting that in the derivation of that equation the quality factor Q exhibits a linear dependence on ω0, yet assumption (a) simultaneously fixes Q(B,T) = 100. From Eq. (3), a truly constant Q implies SNR ∝ (Q·ω0)^{1/2}·M0 ∝ B^{3/2}/T, not B²/T; the (B/B0)² term in Eq. (15) already encodes Q ∝ ω0, and the additional factor Q/Q0 = 100/46 double-counts the same effect. At 0.1 T the two self-consistent readings of Eq. (3) differ by roughly an order of magnitude in SNR and hence in δB/B (a factor of about 2 to 20 depending on which alternative is adopted), which is comparable to the width of the claimed sensitivity range. Moreover, Q0 = 46 was only stabilized at 228 MHz by resistive spoiling (Sec. 3.3); a Q of 100 at the 3 MHz Larmor frequency of 0.1 T requires a redesigned probe and is not demonstrated. Please choose one consistent Q model, correct Eq. (15) and Fig. 12 accordingly, and state the Q(B) assumption as a design target rather than a measured quantity.","section":"Sec. 4.3, Eq. (15), assumptions (a) and (c)"},{"comment":"The manuscript's own Section 4.3 states that for an absolute determination of the magnetic field further requirements must be met, namely spherical cells and susceptibility corrections, neither of which is implemented, and Section 4.2 notes that below 105 K the condensed oxygen layer produces demagnetization fields that will be important for absolute field measurements. In this situation, the abstract's phrase precision magnetometry of magnetic fields B > 0.1 T overstates what is demonstrated: the quoted 10^-11 < δB/B < 10^-7 range applies to relative field changes (monitoring, shimming, field mapping), not to absolute field values. The introduction's framing, which emphasizes the capability of determining absolute field values and the 4×10^-8 absolute accuracy of water-probe calibration, invites the stronger reading. Please scope the title/abstract claim explicitly to relative field measurements, or implement the spherical-cell/susceptibility steps needed for the absolute claim.","section":"Sec. 4.3, first paragraph; abstract"},{"comment":"The measured sensitivity values in Fig. 11 inherit a systematic uncertainty that is not displayed or quantified. The 4.2 K FID is visibly non-exponential, with the paper itself reporting signal distortions produced by first- and higher-order magnetic field gradients due to insufficient shimming and magnetic susceptibility mismatch, and T2* = (2.2 ± 0.7) ms (Sec. 4.1), while the CRLB (Eq. 5) is derived for an exponentially damped sinusoid; the paper acknowledges in Sec. 2 that the CRLB is only an approximate value in this case. Since σf ∝ 1/(T2*)^{3/2}, the ±0.7 ms uncertainty alone spans a factor of about 2.7 in δB/B at 4.2 K, before any line-shape-model error from the C(κ) correction. Please report error bands on the Fig. 11 points using the T2* envelope and, ideally, a Monte Carlo CRLB estimate on the actual FID envelope, and state the line-shape systematic in the quoted sensitivity values.","section":"Sec. 4.1, Fig. 5; Sec. 4.3, Fig. 11"}],"minor_comments":[{"comment":"The abstract claims sensor readout rates of order (Hz), but the field-monitoring curve in Fig. 11 uses 1/TR = 0.1 Hz, and near 30 K the measured T1 ≈ 20 min (Fig. 8) prevents even 0.1 Hz operation without substantial Ernst-angle sensitivity loss; please state the achievable rate more precisely (order 0.1 Hz, with regions of slower operation).","section":"Abstract; Sec. 4.3, Fig. 11"},{"comment":"The Gd3+ loading is described only through precursor masses; no final concentration (e.g., mmol Gd per gram of silica gel) is reported, so the claim that desired T1 times can be set by adjusting the Gd3+ concentration is not reproducible from the text, and the BPP fit parameters are given only for the pure silica gel sample.","section":"Sec. 3.2, ad iii; Fig. 10"},{"comment":"The input parameters for Fig. 11 and Eq. (15) are scattered across the text and figures; in particular, the IFW 300 K value SNR0 = 63 is only inferable from the ratio R = 70/63 = 1.11 quoted in Sec. 4.1. A short table listing SNR0, fBW, Q, T2*, C(κ), p, and the T1 values used for Fig. 11 would materially improve reproducibility.","section":"Sec. 4.3; Sec. 4.1"},{"comment":"For the cylindrical cells used here, the gas-phase O2 contributes a temperature-dependent susceptibility shift that is only discussed qualitatively; since χv ∝ 1/T, this implies a temperature-dependent field offset that is relevant to interpreting δB/B when T changes during a monitoring run. A quantitative estimate, or an explicit statement that the effect is below the quoted sensitivity, would strengthen the relative-field interpretation.","section":"Sec. 4.2"},{"comment":"Copyediting is needed in several places: the title renders as gaseous3He, the abstract exponents appear as 10-11 and 10-7 without superscripts, Sec. 4.3 contains the typo remains largely umaffected, and the parenthesis balance in Eq. (14) is off. None of these affect the physics.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The measured core of this paper — T1 engineering via O2/silica gel/Gd3+, SNR scaling at 7.05 T from 4.2 K to 300 K, and cell stability — is solid and appropriate for the journal. The main risk is that the abstract's 'accessible' sensitivity range is a declared extrapolation from a single field point; a low-field (0.1–1 T) test would settle the matter, but a rewording that explicitly labels Fig. 12 as a projection would also make the claim sound within the manuscript's current scope. I therefore recommend major revision rather than rejection. No concerns about citation practice or novelty disclosure; the limitations are honestly stated in the body, and the main issue is consistency between the body's caveats and the abstract's wording."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the experimental core is solid and worth knowing. Seven sealed high-pressure 3He cells, T1 reduction via SrO2-generated O2, silica gel, and Gd3+ loading, and a full 4.2 K–300 K characterization of T1, T2*, and SNR at 7.05 T. The SNR gain from 70 at 300 K to 3320 at 4.2 K is real, and the two-month reproducibility check on cell #3 is reassuring. This goes beyond Fan et al.'s reservoir-bulb system and the authors' own room-temperature MEOP magnetometer; the sealed-cell approach with demonstrated T1 control is the actual new contribution.\n\nThe soft spot is exactly where the reader's report puts it: the headline '10^-11 < dB/B < 10^-7' range in Fig. 12 is an extrapolation from one field, 7.05 T, using Eq. 15. The assumptions that Q(B,T)=100, SNR scales as B^2, and T2* scales as B0/B with constant 0.7 ppm relative inhomogeneity are stated but not validated at any other field. The stress-test note is right: if absolute field inhomogeneity does not scale with B, T2* at 0.1 T could be tens of times shorter than assumed, pushing the low-field edge up by orders of magnitude. Also, the measured Q0=46 was stabilized by resistive spoiling; Q at 3 MHz is not demonstrated. This is not a fatal flaw, because the experimental data themselves support the sensor's feasibility at the tested field, but the abstract's sensitivity claim overstates what is currently measured.\n\nThe paper is honest about its own limits: Sec. 4.3 explicitly says absolute field determination requires spherical cells and susceptibility corrections. So the claim should be read as relative field monitoring, not absolute precision magnetometry yet.\n\nI'd send this to review. The referee's main job is to require that the sensitivity projection be labeled as a projection, and ideally to ask for a second field point. The experimental work is reproducible, the citation pattern is fine, and the engineering is careful. For a reader in NMR magnetometry or cryogenic field monitoring, it's a useful piece.","headline":"Solid experimental foundation for sealed 3He NMR cells, but the headline sensitivity range is an extrapolation from one field and should be labeled as a projection.","tokens_in":18997,"tokens_out":2222,"would_cite":true,"duration_ms":23867,"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":"Sealed helium-3 gas cells map magnetic fields from 4 K to 300 K.","keywords":["3He NMR magnetometry","cryogenic sensors","thermal nuclear polarization","T1 relaxation engineering","sealed gas cells","permeation filling","Cramér-Rao bound","superconducting magnet monitoring"],"falsifier":"Measure the same 30-bar oxygen-doped cell at a second field strength (e.g., $1$ T) and compare observed SNR and $T_2^*$ with Eq. 15 at the same temperature and pressure; if SNR does not scale as $B^2$ or $T_2^*$ does not scale as $1/B$, the sensitivity contours in Fig. 12 need revision.","tokens_in":17881,"feed_emoji":"🧲","tokens_out":11169,"duration_ms":90821,"temperature":0.7,"pith_summary":"This paper reports a sealed, high-pressure $^3$He gas cell that senses magnetic fields above $0.1$ T by nuclear magnetic resonance, and claims it works at any temperature from $4$ K to $300$ K. To make the sensor practical, the authors add oxygen, silica gel, or gadolinium-doped silica gel to cut the spin-lattice relaxation time $T_1$ from minutes or hours down to seconds, so a fresh thermal polarization is available at readout rates around $0.1$ Hz. Combining measured signals with the Cramér-Rao lower bound, they argue that a single $\\pi/2$-pulse measurement reaches relative field uncertainties between $10^{-11}$ and $10^{-7}$ across the whole $B$-$T$ plane, with signal averaging extending the reach. A sympathetic reader would care because existing cryogenic NMR sensors freeze into solids and lose precision, while optically pumped $^3$He schemes slow down at low temperature; this design promises one compact probe for the entire range.","feed_headline":"3He gas cell maps magnetic fields from 4 K to 300 K","feed_subtitle":"Thermally polarized NMR probes claim 1e-11 to 1e-7 relative precision in one pulse.","key_machinery":"The load-bearing object is the sealed quartz cell filled by selective helium permeation through quartz at $500\\,^\\circ$C, which allows internal pressures up to $100$ bar without flame-sealing under pressure. Three additive strategies - strontium peroxide decomposition to release about $5$ bar of paramagnetic O$_2$, high-surface-area silica gel, and Gd$^{3+}$-loaded silica gel - shorten $T_1$ to seconds, with $T_1$ minima near $12$ K for the gels and a wall-relaxation regime below $54$ K for oxygen. The argument then runs on the SNR scaling law (Eq. 15), $\\mathrm{SNR} \\propto (B/B_0)^2 (300\\,\\mathrm{K}/T)\\, Q/Q_0\\, p/p_0$, combined with the Cramér-Rao lower bound for frequency estimation of a damped sinusoidal free-induction decay, which turns measured SNR and transverse decay time $T_2^*$ into the relative sensitivity $\\delta B/B$.","core_discovery":"The central claim is that thermally polarized gaseous $^3$He, sealed at up to $100$ bar in small quartz cells and doped to shorten $T_1$, forms a precision magnetometer for $B > 0.1$ T at any temperature $4\\,\\mathrm{K} < T < 300\\,\\mathrm{K}$. The paper demonstrates the enabling steps: a permeation-based filling technique that safely seals cells at high pressure, three recipes for reducing $T_1$ to seconds over wide temperature ranges, and a scaling law (Eq. 15) that extrapolates the measured SNR at $7.05$ T, $30$ bar, and $300$ K to other fields, pressures, and temperatures. From that scaling and the CRLB, the paper derives the headline sensitivity contours: single-pulse relative uncertainties $10^{-11} < \\delta B/B < 10^{-7}$ for SNR $> 1$, with the caveat that these are relative frequency measurements, not absolute field values, since absolute determination would require spherical cells and susceptibility corrections not implemented here.","pith_inferences":["Because SNR scales as $B^2$ while the O$_2$ relaxation route depends mainly on oxygen density, receiver noise (the factor $F$ in Eq. 3) rather than polarization is likely the limiting term at low field; a cryogenic preamplifier could push the $0.1$ T edge below $10^{-7}$, an extension the paper does not test.","The $T_1$ maximum near $30$ K makes the oxygen-doped cell a slow sensor in that window; a silica-gel or Gd-loaded cell, whose $T_1$ minimum sits near $12$ K, would be the better fit there, so the optimal additive depends on the operating temperature.","Once spherical cells and susceptibility matching are added, the same SNR arguments would turn this relative sensor into a candidate primary field standard, since the $^3$He gyromagnetic ratio has been measured directly by Penning-trap work, a step the paper notes but does not take."],"forward_implications":["A single 30-bar cell can monitor field stability and shim superconducting magnets from room temperature down to liquid-helium temperature, with readout every 10 s in the oxygen-doped version.","Accumulating $n$ scans improves $\\delta B/B$ by $\\sqrt{n}$, which the paper claims extends the accessible range to the low-temperature and low-field edges of its sensitivity map.","Because $T_1$ grows with lower gas density, the sensor's refresh rate and sensitivity can be traded off by choosing the filling pressure.","If Eq. 15 holds, the same sealed-cell design should reach about $10^{-7}$ relative precision at $0.1$ T and $300$ K, and better than $10^{-9}$ at $4$ K for fields above about $1$ T."],"supporting_citations":[{"why":"Reservoir-bulb thermally polarized $^3$He NMR at 4.2 K; supplies the $T_2 \\approx 2.6$ s value and the SNR $\\approx 10$ benchmark that this paper's sealed cells improve on.","marker":"[20]"},{"why":"Empirical O$_2$-induced relaxation rate formula (Eq. 9) that lets the authors set $T_1$ by choosing oxygen pressure.","marker":"[35]"},{"why":"Standard SNR expression (Eq. 3); the source of the $Q$, $\\omega_0$, and temperature dependences that Eq. 15 scales.","marker":"[25]"},{"why":"Cramér-Rao lower bound for frequency estimation of a damped sinusoid (Eqs. 5-6), converting SNR and $T_2^*$ into magnetic-field sensitivity.","marker":"[29]"},{"why":"Quartz permeation filling technique and the spherical-cell susceptibility discussion adopted here for cell production and for the absolute-field caveat.","marker":"[30]"},{"why":"Earlier $^3$He magnetometer claiming better than $10^{-12}$ using metastability-exchange optical pumping; the performance benchmark this thermal-polarization approach seeks to simplify and extend to cryogenics.","marker":"[9]"},{"why":"Direct Penning-trap measurement of the shielded $^3$He magnetic moment, giving the $\\gamma_{\\mathrm{He}}$ conversion factor that ties measured frequency to field.","marker":"[11]"},{"why":"Temperature dependence of $^3$He surface relaxivity on glass and oxygen surfaces, used to explain the $T_1$ behavior below 105 K.","marker":"[36]"}],"fun_headline_variants":["Gaseous 3He NMR measures fields from 4 K to 300 K","3He gas cell magnetometer: 1e-11 precision at 4 K","Cryogenic field sensing: 3He NMR probes hit 10^-11","Thermally polarized 3He delivers precise B-field maps","3He gas probes: precision magnetometry from 4 K to room temp"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claimed sensitivity range assumes a constant coil quality factor, a constant $0.7$ ppm field inhomogeneity, and an SNR that scales as $B^2/T$ up to $80$ bar and down to $0.1$ T, none of which is verified beyond the single $7.05$ T measurement; absolute field values would also require spherical cells and susceptibility corrections the paper does not implement.","fun_headline_variants_meta":{"raw":{"variants":["Gaseous 3He NMR measures fields from 4 K to 300 K","3He gas cell magnetometer: 1e-11 precision at 4 K","Cryogenic field sensing: 3He NMR probes hit 10^-11","Thermally polarized 3He delivers precise B-field maps","3He gas probes: precision magnetometry from 4 K to room temp"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000975,"raw_usage":{"total_tokens":4115,"prompt_tokens":892,"completion_tokens":3223,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":508,"completion_tokens_details":{"reasoning_tokens":3121}},"tokens_in":508,"tokens_out":3223,"duration_ms":24641,"temperature":1.0,"reasoning_tokens":3121,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:13:58.977463+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the same 30-bar oxygen-doped cell at a second field strength (e.g., $1$ T) and compare observed SNR and $T_2^*$ with Eq. 15 at the same temperature and pressure; if SNR does not scale as $B^2$ or $T_2^*$ does not scale as $1/B$, the sensitivity contours in Fig. 12 need revision.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reservoir-bulb thermally polarized $^3$He NMR at 4.2 K; supplies the $T_2 \\approx 2.6$ s value and the SNR $\\approx 10$ benchmark that this paper's sealed cells improve on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Empirical O$_2$-induced relaxation rate formula (Eq. 9) that lets the authors set $T_1$ by choosing oxygen pressure."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Standard SNR expression (Eq. 3); the source of the $Q$, $\\omega_0$, and temperature dependences that Eq. 15 scales."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Cramér-Rao lower bound for frequency estimation of a damped sinusoid (Eqs. 5-6), converting SNR and $T_2^*$ into magnetic-field sensitivity."},{"cited_title":"Kober, B","cited_arxiv_id":null,"evidence_quote":"Quartz permeation filling technique and the spherical-cell susceptibility discussion adopted here for cell production and for the absolute-field caveat."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Earlier $^3$He magnetometer claiming better than $10^{-12}$ using metastability-exchange optical pumping; the performance benchmark this thermal-polarization approach seeks to simplify and extend to cryogenics."},{"cited_title":"Blaum, Phys","cited_arxiv_id":null,"evidence_quote":"Direct Penning-trap measurement of the shielded $^3$He magnetic moment, giving the $\\gamma_{\\mathrm{He}}$ conversion factor that ties measured frequency to field."},{"cited_title":"Gemmel, W","cited_arxiv_id":null,"evidence_quote":"Temperature dependence of $^3$He surface relaxivity on glass and oxygen surfaces, used to explain the $T_1$ behavior below 105 K."}],"review_version":1}