{"id":"5ad60bbe-f467-4d6a-9815-8d615fe161ec","arxiv_id":"2504.15348","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"21-cm power spectrum forecasts show HERA and CMB-S4 combined can constrain the neutrino self-coupling G_eff to about 10 percent across strong, moderate, and mild interaction models, with futuristic lunar arrays reaching 1 to 5 percent.","lead":"This paper forecasts how future 21-cm radio observations, combined with CMB measurements, could pin down the strength of hypothetical self-interactions among neutrinos. It finds that HERA plus CMB-S4 could measure the coupling to about 8 to 14 percent, while a futuristic lunar array could reach 1 to 5 percent precision.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"HERA-only sensitivity to the moderate self-interaction model depends on the popII-only astrophysical scenario; adding popIII stars shifts the MIν 2σ constraint from 16% to 176% (Table IV), so robustness of the source model is the key untested assumption.","rationale":"I agree with the reader's identification of the semi-analytic astrophysical chain as the weakest link, and I sharpen it to a specific, testable fragility: the HERA-only constraint for the MIν model changes by an order of magnitude when popIII stars are included, despite the paper listing popII-only as the baseline. This is not an ad hominem or a disagreement with the field's consensus; it is an internal sensitivity already visible in Table IV. The paper is transparent about this dependence, which is creditworthy, and the combined CMB+21cm claim appears more robust across the two source scenarios, which is why I would not move the verdict away from CONDITIONAL. The concrete test of swapping the halo mass function is feasible with the public code and would directly establish whether the forecasted sensitivities are artifacts of the fiducial astrophysical assumptions. The reader's verdict of CONDITIONAL remains appropriate pending this robustness check.","tokens_in":37077,"tokens_out":13074,"duration_ms":127504,"concrete_test":"Modify the public 21cmFirstCLASS code to replace the Sheth-Tormen halo mass function (Eq. 15) with an alternative high-redshift fit such as Tinker et al. (2008), and rerun the Fisher forecast for the MIν model under the popII+popIII scenario. If the HERA-only 2σ constraint on log10Geff is no longer below the CMB-S4 value of 66%, the claim that HERA improves on CMB for moderate couplings fails; if the HERA+CMB-S4 combined constraint shifts by more than a factor of two, the degeneracy-breaking result is also not robust.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central forecast is built on the semi-analytic chain from the linear matter power spectrum to the 21-cm signal: the Sheth-Tormen halo mass function (Eq. 15) combined with the popII/popIII star-formation model (Eqs. 7-16) in 21cmFirstCLASS. This chain is not validated for the high-redshift, low-mass halos (Mh ~ 1e8-1e11 Msun) that carry the neutrino signature, and the paper's own Table IV reveals a strong sensitivity to the source model: for the moderate-coupling MIν model, the HERA-only 2σ constraint on log10Geff degrades from 16% (popII-only) to 176% (popII+popIII), moving from a factor-of-four improvement over CMB-S4 to a bound far weaker than CMB-S4's 66%. The claim that HERA can improve on CMB constraints for moderate couplings therefore rests on the assumption that popIII stars in mini-halos do not contribute at cosmic dawn, an assumption that is theoretically uncertain and possibly contradicted by recent JWST candidate popIII galaxies. More generally, an error in the high-z HMF shape or in the mapping from halos to Lyα and X-ray emissivities would directly bias the Fisher derivatives dΔ21^2/dGeff, either manufacturing or erasing the scale-dependent signature the forecast uses. The combined HERA+CMB-S4 constraints are less sensitive to this choice (MIν: 31% vs 66% for CMB-S4 alone), so the degeneracy-breaking claim is more robust, but it inherits the same unvalidated astrophysical mapping.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents Fisher-matrix forecasts for constraints on the neutrino self-coupling strength G_eff from 21-cm power-spectrum measurements, alone and in combination with CMB-S4. The 21-cm signal is modeled with 21cmFirstCLASS, initialized with transfer functions from the modified Boltzmann code nuCLASS, for three fiducial models: strong (SIν), moderate (MIν), and mild (mIν) self-interactions. For cosmic dawn, the forecast uses HERA design sensitivity and two astrophysical scenarios (popII-only and popII+popIII); for the dark ages, it uses two lunar interferometer configurations (Lunar D and LRA1). The main quantitative claims are that HERA at design sensitivity can improve on CMB-only constraints for moderate couplings, that HERA+CMB-S4 breaks the M_ν–G_eff degeneracy and reaches ~8–14% (2σ) on log10(G_eff/MeV^-2), and that LRA1+CMB-S4 reaches ~1–4%.","tokens_in":37440,"tokens_out":3198,"duration_ms":33962,"significance":"If the forecasts hold, the paper would be the first systematic demonstration that 21-cm observations can probe neutrino self-interactions, complementing CMB and LSS probes and opening a new observational window on beyond-Standard-Model neutrino physics. The analysis is technically careful in several respects: the nuCLASS code is made public, the CMB part is validated against the CMB-S4 Science Book and against the earlier forecast of Ref. [32], the CMB settings are conservative (single channel, ℓ<3000), and the Fisher formalism is applied consistently. The main weakness is that the central HERA claims are conditional on an unvalidated semi-analytic astrophysical model, and the paper's own Table IV shows that the HERA-only results change by an order of magnitude when popIII stars are added. The dark-ages forecasts additionally rely on an acknowledged linear-evolution approximation at z>35; the paper argues this is subdominant, but does not provide a quantitative robustness test.","major_comments":[{"comment":"The paper's headline claim that HERA can improve upon existing constraints on G_eff is not robust to the treatment of popIII stars. In Table IV, the HERA-only 2σ constraint on log10(G_eff/MeV^-2) for the moderate-interaction model MIν degrades from 16% (popII-only) to 176% (popII+popIII), while for SIν it degrades from 32% to 72%; for mIν it improves from 112% to 78%. Since the MIν and SIν cases are the ones that support the abstract's statement that HERA 'can improve upon existing constraints' and be sensitive 'beyond the reach of current and future CMB experiments', this statement should be qualified as conditional on the popII-only astrophysical baseline. I would ask the authors to either (a) state explicitly that the HERA-only claim applies only to the baseline scenario and is not robust when popIII sources are included, or (b) provide a physically motivated argument, backed by external data, for why the popII-only scenario is the relevant fiducial for HERA's observational window.","section":"§V.2, Table IV"},{"comment":"The forecast chain from the linear matter power spectrum to the 21-cm signal is built on the Sheth-Tormen halo mass function (Eq. 15), the EPS-based collapse fraction (Eq. 12), and the popII/popIII star-formation prescriptions (Eqs. 7–16). This chain is applied to halos in the range M_h ~ 10^8–10^11 M_sun at z ~ 13–30, where neither the Sheth-Tormen fitting function nor the assumed mapping from halo mass to Lyα and X-ray emissivities has been validated against simulations or observations. Because the Fisher derivatives dΔ21^2/dG_eff are computed through this chain, an error in the high-redshift HMF shape or in the SFR mapping could directly bias the forecasted sensitivity, either manufacturing or erasing the scale-dependent signature. The paper itself notes (footnote 5) that stochasticity in galaxy emissivities may be non-negligible. I request a robustness test against at least one alternative HMF (e.g., a Tinker-type mass function or a simulation-calibrated fit) and an alternative SFR prescription, with the resulting shifts in the Table IV constraints reported.","section":"§III.A, Eqs. (7)–(16), (15)"},{"comment":"The dark-ages forecasts for Lunar D and LRA1 use 21cmFirstCLASS runs in which the matter density field is evolved linearly at z>35, neglecting the mode coupling induced by the baryon–dark-matter relative velocity. The paper acknowledges this approximation and cites Ref. [127] for a ~10% effect in ΛCDM at k≲0.01 Mpc^-1, arguing that it is subdominant for the scales and redshifts considered. However, the mIν model has enhanced small-scale power (Fig. 2), which, after mode coupling, could feed power to the larger scales probed by Lunar D; the paper states this 'may slightly boost the SNR' but does not quantify it. Since the percent-level LRA1 claims are a headline result, I would like to see either a quantitative estimate of this effect for each neutrino model or a clear statement that the LRA1 constraints at k>1.5 Mpc^-1 are unaffected by the linear-evolution approximation because they use the high-resolution box.","section":"§V.3, Sec. IV"}],"minor_comments":[{"comment":"The definition of the smoothing scale appears to have a typo: R = [3M_h/(4πρ̄_m,0)]^{-3} should presumably be R = [3M_h/(4πρ̄_m,0)]^{1/3}, the comoving radius of the top-hat window.","section":"Eq. (13)"},{"comment":"The sentence comparing the forecast uncertainty to Ref. [32] says 'close to our value in Table I'; the relevant entry is in Table IV, not Table I.","section":"§V.1"},{"comment":"In the dark-ages noise formula, the notation V_z and √(V_z k^3) is dimensionally awkward as written; please check the factors and define all symbols (Δln k, f_sky, χ(z)) at first use.","section":"§IV.A.2, Eq. (29)"},{"comment":"The right panel of Fig. 7 is said to come from 'an additional 21cmFirstCLASS simulation' without specifying the box size; Sec. IV gives L=50 Mpc and N_cell=128, but the caption should be self-contained.","section":"Fig. 7 caption and §IV"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of the journal and the central forecasting machinery is sound and transparent. The main issue is not circularity or internal inconsistency, but that the most striking HERA-only claims depend on the popII-only astrophysical scenario, and this dependence is evident in the paper's own Table IV. The authors should be encouraged to add explicit caveats and robustness checks rather than to change the overall direction of the work. The extensive use of the authors' own codes (nuCLASS, 21cmFirstCLASS) is appropriate and the codes are publicly available; I do not see a citation or novelty-disclosure issue."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"I'll skip the throat clearing. The thing to know: this is the first forecast of the 21-cm power spectrum as a probe of self-interacting massive neutrinos, and it's a competent one. The authors validate their CMB pipeline against the CMB-S4 Science Book and Ref. [32], ship nuCLASS, and use public 21cmFirstCLASS, so the machinery is checkable. The headline numbers in Table IV are new: HERA+CMB-S4 would give roughly 8-14% (2-sigma) on log10 Geff for three interaction strengths, and a lunar array like LRA1 would reach about 1-5%.\n\nThe strong part is the joint analysis. Combining HERA with CMB-S4 breaks the M_nu-G_eff degeneracy that plagues the 21-cm signal alone, and those numbers are less sensitive to the astrophysical model. The dark ages forecasts are also worth taking seriously because the signal is nearly free of astrophysics; LRA1's percent-level constraint on G_eff is a clean, if futuristic, result.\n\nThe soft spot is the HERA-only claim, which is more fragile than the abstract implies. The paper shows this itself: for the moderate-interaction model, the HERA-only 2-sigma constraint on log10 G_eff goes from 16% with popII-only stars to 176% when popIII stars are included. That's not a small perturbation. The whole chain from matter power spectrum to halo mass function to star formation rate density relies on the Sheth-Tormen HMF and semi-analytic SFR models that are not validated for the low-mass halos (M_h ~ 1e8-1e11 M_sun) that carry the signal. The authors do flag the linear-evolution approximation at z>35 and the stochasticity of galaxy emissivities, so they aren't hiding it. But the headline 'HERA can improve on existing constraints' should be read with a popII-only asterisk.\n\nMy overall take: the central argument—that 21-cm plus CMB can constrain G_eff better than CMB alone—holds up, especially in the joint and dark-ages cases. The HERA-only sensitivity numbers are conditional on the source model, and the paper says so. That's the right balance for a forecast.\n\nI'd send this to a serious referee rather than desk-reject. The referee should push on the astrophysical robustness, maybe ask for a sensitivity test with a different HMF or star-formation prescription, but the work is transparent and the new-territory claim is real. I'd cite the joint constraints if I were working in this area.","headline":"First 21-cm power-spectrum forecast for neutrino self-interactions; the joint HERA+CMB-S4 numbers are solid, but the HERA-only headline case depends on a popII-only source model that the paper itself shows is fragile.","tokens_in":38068,"tokens_out":2524,"would_cite":true,"duration_ms":23766,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The 21-cm power spectrum can probe neutrino self-interactions beyond the reach of CMB experiments, and combining 21-cm with CMB-S4 breaks the neutrino-mass versus self-coupling degeneracy.","keywords":["21-cm cosmology","neutrino self-interactions","massive neutrinos","cosmic dawn","dark ages","Fisher forecast","halo mass function","CMB-S4"],"falsifier":"Measure the 21-cm power spectrum at $z\\simeq 13$ with HERA at design sensitivity: if the data match $\\Lambda$CDM while the mild-coupling model predicts a visible boost in the abundance of $10^8$–$10^{10}\\,M_\\odot$ halos that shifts the signal, the forecast's central claim is falsified. A weaker but still decisive test is the CMB-S4 bound on $\\log_{10} G_{\\rm eff}$ for the mild model, which the paper forecasts at 534% uncertainty; a much tighter bound would indicate the 21-cm complementarity is not needed.","tokens_in":36854,"feed_emoji":"📡","tokens_out":12158,"duration_ms":97515,"temperature":0.7,"pith_summary":"This paper asks whether the 21-cm power spectrum—the radio signal from neutral hydrogen during cosmic dawn and the dark ages—can measure the self-interaction strength of massive neutrinos, a parameter that current and planned CMB observations constrain only weakly. The argument is that self-interacting neutrinos imprint scale-dependent bumps and suppressions on the matter power spectrum at $k > 0.1\\,{\\rm Mpc}^{-1}$, which change the abundance of the small and intermediate halos that host the first galaxies, and those changes propagate into the 21-cm signal. Using Fisher forecasts built from a Boltzmann solver that includes neutrino self-interactions and a semi-analytic 21-cm simulation, the paper finds that the HERA array at design sensitivity can improve on existing constraints on the coupling $G_{\\rm eff}$ and reach values of the coupling beyond CMB experiments. The central quantitative result is that a joint HERA+CMB-S4 analysis yields 2-$\\sigma$ uncertainties of roughly 8%–14% on $\\log_{10} G_{\\rm eff}$ for strong, moderate, and mild self-interaction models, while a futuristic lunar array could reach 1%–5%. This matters because recent cosmology hints at a bimodal posterior for $G_{\\rm eff}$, and the neutrino mass–coupling degeneracy is precisely the kind of correlation that joint 21-cm and CMB data can break.","feed_headline":"21-cm + CMB data can pin neutrino self-coupling to ~10%","feed_subtitle":"HERA plus CMB-S4 breaks the neutrino mass–coupling degeneracy; lunar arrays could reach ~1–5 percent.","key_machinery":"The load-bearing chain runs from a four-Fermi effective operator $\\mathcal{L} \\supset \\frac{1}{2}G_{\\rm eff}(\\bar\\nu\\nu)(\\bar\\nu\\nu)$, whose coupling $G_{\\rm eff}$ controls how long neutrinos behave as a fluid instead of free streaming. A modified Boltzmann solver computes the resulting scale-dependent matter power spectrum $P_m(k,z)$; that spectrum enters the Sheth–Tormen halo mass function through the variance $\\sigma^2(M_h,z) = \\int_0^\\infty (dk/k) P_m(k,z) W_R^2(k)$, and the halo abundance determines the star formation rate density that drives the 21-cm signal in a semi-analytic simulation. The Fisher matrix then converts simulated sensitivities into forecast uncertainties. The mechanism that carries the argument is this mapping from small-scale power features to halo-abundance features to 21-cm brightness fluctuations, rather than any single direct observable.","core_discovery":"Self-interacting neutrinos delay the onset of free streaming. Modes that enter the horizon before self-decoupling experience suppressed growth, while modes entering near self-decoupling are boosted, producing a scale-dependent feature in the linear matter power spectrum. The paper shows that this feature propagates through the halo mass function to the star formation rate density and hence to the 21-cm power spectrum, making cosmic dawn an indirect probe of scales $k \\sim 1$–$100\\,{\\rm Mpc}^{-1}$ that CMB observations cannot reach directly. The paper's central claim is that HERA alone outperforms CMB-S4 for moderate and mild self-interactions, CMB-S4 alone is better for the strong case, and adding HERA to CMB-S4 improves the constraint on $\\log_{10} G_{\\rm eff}$ in all three benchmark models, with the joint analysis reaching $\\sim 10\\%$ 2-$\\sigma$ uncertainties. During the dark ages, where astrophysics is absent, the paper forecasts that a large lunar interferometer can reach percent-level constraints on $\\log_{10} G_{\\rm eff}$.","pith_inferences":["If the forecast chain is right, the same halo-abundance route should make cosmic-dawn 21-cm observations sensitive to other models that alter small-scale power, such as warm dark matter or dark matter–baryon interactions; the Fisher setup here is a template for those searches.","A direct test of the mechanism would be to compare the scale-dependent shape of the 21-cm power spectrum at $z\\sim 13$: the moderate- and mild-coupling models predict a boost in halo abundance in the $10^8$–$10^{10}\\,M_\\odot$ range, so the signal should rise faster than in $\\Lambda$CDM even if the overall amplitude is degenerate with astrophysics.","The dark-ages forecast assumes linear evolution of the density field above $z>35$; including the baryon–dark-matter relative-velocity mode coupling, which the paper notes is neglected, could slightly boost the large-scale signal for the mild-coupling model and improve the lunar-array constraints rather than degrade them.","The paper's claim that CMB-S4 breaks the $M_\\nu$–$G_{\\rm eff}$ degeneracy for the mild model implies that any future small-scale probe with different redshift coverage, for example a galaxy survey at moderate redshift, should show a similar complementarity with 21-cm data."],"forward_implications":["HERA at design sensitivity should improve current constraints on the neutrino self-coupling and probe couplings below the reach of current and future CMB experiments.","For the moderate- and mild-coupling models, HERA alone is forecast to beat CMB-S4; for the strong-coupling model, CMB-S4 is more sensitive, but the pair together is best in all cases.","The joint HERA+CMB-S4 analysis breaks the degeneracy between the sum of neutrino masses and $\\log_{10} G_{\\rm eff}$, yielding 2-sigma uncertainties of 8%, 11%, and 14% for the strong, moderate, and mild models in the baseline astrophysical scenario.","A large lunar radio array probing the dark ages could reach 2-sigma uncertainties of 1%–5% on $\\log_{10} G_{\\rm eff}$, without astrophysical contamination.","Including population III stars in the astrophysical model generally weakens the cosmic-dawn constraints because of additional free parameters, but it improves the mild-coupling case by helping to break the mass–coupling degeneracy."],"supporting_citations":[{"why":"Defines the four-Fermi effective operator and the coupling $G_{\\rm eff}$ that the forecasts target.","marker":"[24]"},{"why":"Gives the Boltzmann-equation treatment of neutrino self-scattering that the modified solver implements.","marker":"[28]"},{"why":"Supplies the strong-coupling benchmark parameters and the bimodal posterior that motivates the SIν fiducial model.","marker":"[48]"},{"why":"The new 21-cm simulation tool that propagates linear fluctuations from the Boltzmann solver into the 21-cm signal.","marker":"[64]"},{"why":"Supplies the semi-analytic framework for the cosmic-dawn and reionization 21-cm evolution.","marker":"[67]"},{"why":"Shows how 21-cm data remove the optical-depth nuisance from CMB parameter inference, the basis of the joint Fisher treatment.","marker":"[45]"},{"why":"Demonstrates the 21-cm mitigation of the optical-depth degeneracy for neutrino masses, which the paper extends to self-interactions.","marker":"[46]"},{"why":"Provides the sensitivity calculator used for the HERA cosmic-dawn forecasts.","marker":"[131]"},{"why":"Supplies the dark-ages noise model and a lunar-array configuration used for the dark-ages forecasts.","marker":"[137]"},{"why":"Gives the large lunar-array specifications that enable the percent-level dark-ages forecasts.","marker":"[144]"}],"fun_headline_variants":["HERA+CMB-S4 breaks neutrino coupling degeneracy to 10%","21-cm + CMB forecasts combine to pin neutrino self-coupling","Lunar 21-cm arrays could reach percent-level neutrino coupling","Cosmic dawn 21-cm signal probes neutrino self-interactions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The forecast collapses if the semi-analytic astrophysical model that turns the matter power spectrum into the 21-cm signal—the halo mass function, star formation efficiencies, and reionization history—does not accurately predict how the abundance of small and intermediate halos changes when the power spectrum is modified.","fun_headline_variants_meta":{"raw":{"variants":["HERA+CMB-S4 breaks neutrino coupling degeneracy to 10%","21-cm + CMB forecasts combine to pin neutrino self-coupling","Lunar 21-cm arrays could reach percent-level neutrino coupling","Cosmic dawn 21-cm signal probes neutrino self-interactions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000523,"raw_usage":{"total_tokens":2601,"prompt_tokens":1088,"completion_tokens":1513,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":704,"completion_tokens_details":{"reasoning_tokens":1434}},"tokens_in":704,"tokens_out":1513,"duration_ms":9390,"temperature":1.0,"reasoning_tokens":1434,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T11:28:41.729269+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the 21-cm power spectrum at $z\\simeq 13$ with HERA at design sensitivity: if the data match $\\Lambda$CDM while the mild-coupling model predicts a visible boost in the abundance of $10^8$–$10^{10}\\,M_\\odot$ halos that shifts the signal, the forecast's central claim is falsified. A weaker but still decisive test is the CMB-S4 bound on $\\log_{10} G_{\\rm eff}$ for the mild model, which the paper forecasts at 534% uncertainty; a much tighter bound would indicate the 21-cm complementarity is not needed.","supporting_citations":[],"review_version":1}