{"id":"205962c2-a87e-49d0-80a5-0b937cca1bd6","arxiv_id":"2505.20396","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"A redshift-limited dark matter-neutrino interaction reproduces the observed preference for non-zero coupling in CMB and cosmic shear data and evades small-scale structure bounds.","lead":"Dark matter particles that interact with neutrinos mainly during a narrow cosmic window around redshift 100,000 can explain hints from CMB and weak lensing data while dodging other constraints. The paper shows the data prefer such a short-lived interaction and builds a particle physics model with a resonance to realize it.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The abstract's >3σ preference is not supported by the paper's own statistics: Table I reports a 95% lower bound (~2σ) in a log-prior that excludes uνDM=0, and Appendix A admits a 2σ preference; the Δχ² gain over the constant-νDM model is only ~2.3–2.6σ.","rationale":"The reader's weakest assumption was the ad hoc redshift-window parametrization and the unconstrained Δz tail. My concern is related but distinct: even accepting the window parametrization, the quantitative significance quoted in the abstract ('>3σ') is not supported by the statistics reported in the paper. This is a load-bearing issue because the >3σ claim is the central observational result. The paper's own appendix says '2σ level,' and the comparison against the constant-νDM model (which already fits the data better than ΛCDM) yields only about 2.5σ. The 95% lower bound on uνDM is not a valid exclusion of zero because of the log-prior support and boundary issues. I therefore flag this as the most important concern. However, the paper's broader point—that redshift-localized νDM interactions can mimic CMB+lensing hints while evading small-scale bounds—may still hold with a properly quantified significance, and the model-building section is a useful proof of concept. The reader's conditional verdict (pending chains and subhalo verification) remains appropriate, but the condition should be extended to include a statistically rigorous significance assessment. Hence I do not change the verdict, but the revision must address the >3σ versus 2σ inconsistency.","tokens_in":12833,"tokens_out":8452,"duration_ms":86484,"concrete_test":"Re-derive the significance with a proper nested model comparison: run the same MCMC chains but treat ΛCDM as a separate model with uνDM=0 (e.g., via a mixed prior with a point mass at zero), and compute the Bayes factor or Savage–Dickey density ratio between the window model and ΛCDM, as well as between the window model and the constant-uνDM model. Also compute the profile-likelihood ratio for the window parameters, accounting for the boundary at uνDM=0. If the resulting preference falls below 3σ in both comparisons, the abstract's claim must be downgraded.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of a '>3σ preference for non-zero uνDM' is not established by the quoted statistics. In the main text (Section II), the 3σ claim is justified by the statement that the 95% credible lower bound on uνDM is log10 uνDM > -3.65 (Table I). This is logically insufficient: a 95% lower bound corresponds to roughly 1.65σ for a one-sided interval, and because the prior is uniform in log10 uνDM over [-8,0], the value uνDM=0 lies outside the prior support (at log10 uνDM → -∞). Thus the posterior interval cannot, by itself, exclude the ΛCDM limit. Appendix A states explicitly that the 1D posterior 'exhibits a long tail, indicating a preference for non-zero values at the 2σ level,' contradicting the abstract's '>3σ'. Furthermore, the benchmark BP1 improves χ² over ΛCDM by Δχ²=-16.6 with three extra parameters; compared with the 'uνDM=const' model, which already improves by Δχ²=-7.06 with one extra parameter, BP1 gains only Δχ²=-9.54 for two additional degrees of freedom, corresponding to roughly 2.3–2.6σ, not >3σ. The boundary/undefined-parameter problem at uνDM=0 (zmin and Δz are unconstrained when uνDM=0) makes the naive Δχ² interpretation additionally unreliable. Consequently, the headline statistical preference is overstated, and the paper's central observational claim is not yet robustly quantified.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper explores the cosmological consequences of dark matter-neutrino interactions whose strength is non-zero only within a narrow redshift interval, motivated by hints from CMB and weak-lensing data. Using a toy model with three new parameters (interaction strength uνDM, start redshift zmin, and duration Δz), the authors run MCMC fits to Planck, ACT DR4, DES Y3 cosmic shear, and BAO data, and report a preference for an interaction window at z ~ 10^4-10^5. They then present a neutrino-portal dark matter model with a resonantly enhanced scattering cross section, including a time-varying DM mass to suppress the interaction at high redshifts, and show that benchmark points can reproduce the transfer-function features while partially evading small-scale structure bounds. The abstract claims a >3σ preference for non-zero uνDM, but the body and appendix report only 2-2.6σ evidence.","tokens_in":13425,"tokens_out":5271,"duration_ms":51647,"significance":"If substantiated, the proposal that νDM interactions are enhanced only during a specific epoch would offer a way to reconcile CMB and cosmic-shear hints with otherwise stringent astrophysical and small-scale structure constraints, and would motivate resonant-scattering and DM-mass-variation physics in the dark sector. The paper uses public codes (CLASS, MontePython) and standard likelihoods, applies conservative small-scale weak-lensing masks, and explicitly reports transfer functions and subhalo mass functions for its benchmarks. However, the central statistical claim is overstated relative to the paper's own numbers, and the model-comparison framework has a boundary problem because zmin and Δz are undefined at uνDM=0. The core idea is interesting, but the evidence as presented does not support the headline significance.","major_comments":[{"comment":"The paper's headline claim of a '>3σ preference for a non-zero interaction' is not supported by the quoted statistics. In Section II the claim is justified by the 95% lower bound log10 uνDM > -3.65 from Table I, but in a uniform prior over log10 uνDM ∈ [-8,0] this is at most a one-sided ~2σ statement, and the null value uνDM=0 lies outside the prior support (at log10 uνDM → -∞), so the interval cannot by itself exclude ΛCDM. Appendix A explicitly states that the posterior 'indicates a preference for non-zero values at the 2σ level.' Comparing BP1 to the constant-uνDM model in Table I gives Δχ² = -9.54 for two additional degrees of freedom, corresponding to roughly 2.3-2.6σ, not >3σ. The abstract and main text must be corrected, and the evidence should be quantified with a model-comparison statistic that handles the boundary at u=0.","section":"Abstract; Sec. II; Appendix A; Table I"},{"comment":"The three-parameter toy model has a parameter-identification problem at uνDM=0: when the coupling vanishes, zmin and Δz drop out of the likelihood, so the posterior volume in those directions is not data-defined. The reported credible regions and χ² differences therefore do not constitute a fully valid model comparison between ΛCDM and the redshift-window model. The authors should report a profile likelihood over (zmin, Δz) with uνDM treated properly, or compute a Bayes factor with a prior that includes the null model, before claiming any preference for the window scenario.","section":"Sec. II; Table II; Fig. 2"},{"comment":"The neutrino-portal benchmarks are presented without full posteriors or error bars for the portal parameters, and the model is effectively engineered to reproduce the redshift window preferred by the toy-model fit: the resonance position and the time-varying DM mass are chosen to produce an enhancement around z ~ 10^4-10^5. The quoted χ² values (e.g., -12, -5) are not accompanied by a parameter-counting comparison to ΛCDM or to the toy model, so the reader cannot judge whether this is an actual preference or an existence proof. The paper should clearly label these benchmarks as illustrative and state which parameters are fixed rather than fitted.","section":"Sec. III; Fig. 1; Fig. 3"},{"comment":"The claim that the localized-interaction model evades small-scale structure constraints relies on (i) deliberately masking weak-lensing scales at k ≳ a few h/Mpc, which leaves the Δz posterior tail unconstrained, and (ii) converting the Milky Way satellite bound into a hard prior (Δz ≲ 10^6) rather than a likelihood. The extended Press-Schechter calculation in Sec. II is an approximation that has not been validated with N-body simulations for the specific transfer-function shapes produced by these models. The authors should include a sensitivity check of their conclusions to the assumed Δz prior and should explicitly acknowledge the approximate nature of the small-scale structure constraints.","section":"Appendix A; Sec. II; Fig. 3"}],"minor_comments":[{"comment":"The word 'schmatically' in the third paragraph should be 'schematically'.","section":"Introduction"},{"comment":"The Thomson cross section σT used to normalize uνDM is not explicitly defined; please give its value or a reference.","section":"Eq. (1)"},{"comment":"Only two parameters (100Ωb h² and 100θs) have quoted uncertainties; the rows for Ωm, ns, τreio, S8, and others should either include 68% intervals or be explicitly labeled as best-fit values.","section":"Table I"},{"comment":"The caption does not specify the parameter values for the three curves; in particular, the brown solid line uses g = 2.3×10^-2, while the actual benchmark uses g = 2.3×10^-4 (footnote 2). Please add a note in the caption to avoid confusion.","section":"Fig. 1 caption"},{"comment":"Reference [70] (Das and Weiner) lists the arXiv identifier astro-ph/0611353, which appears to be from 2006, inconsistent with the 2011 publication year; please verify the identifier.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper presents a creative idea and uses a standard, reproducible MCMC pipeline, but the headline significance is contradicted by the authors' own appendix and by the Δχ² comparisons. I recommend major revision: the abstract and main-text claims need to be brought in line with the 2-2.6σ evidence, and the model-comparison problem at the boundary u=0 should be addressed with a proper statistical treatment. The portal-model section should be rephrased as an illustrative existence proof rather than a fitted model. No concerns about authorship or scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know first: the redshift-window parametrization is new, and the combined ACT+DES analysis is real work. But the headline \">3σ preference\" is contradicted by the paper's own Appendix A and by the Δχ2 numbers. The honest significance is around 2σ to 2.6σ, so treat the abstract with skepticism; the rest of the paper is worth engaging.\n\nWhat's actually new: introducing zmin and Δz to localize νDM interactions, and showing with public CLASS/MontePython and real data that the preferred coupling shifts to larger values and can dodge small-scale bounds. The neutrino portal model with resonant scattering and a time-dependent DM mass is a plausible proof of concept. The paper is transparent about limitations: Appendix A admits a 2σ preference, and the authors flag the need for zoom-in simulations to verify the EPS-based subhalo predictions.\n\nSoft spots: the statistical evidence is not as strong as claimed. A 95% lower bound in a log-prior that excludes uνDM=0 does not exclude the ΛCDM limit; the 1D posterior has a long tail. The benchmark BP1 improves χ2 over the constant-uνDM model by only ~9.5 with two extra parameters, which is ~2.3σ, not >3σ. The toy model is fitted to the same data, and the portal model is engineered to reproduce that fitted window, so the preference is partly by construction. The Δz posterior is poorly constrained because small-scale WL scales are masked, and the small-scale structure constraint rests on an approximate formalism. These are real but fixable issues.\n\nWho should read it: cosmologists and particle phenomenologists working on νDM interactions and dark acoustic oscillations; for them it's a useful demonstration that narrow resonances can change the game. It deserves a serious referee, but the authors should be asked to correct the significance claim, release the chains, and strengthen the small-scale validation before publication.","headline":"A genuinely new redshift-window parametrization with real data work, but the headline >3σ claim does not survive the paper's own statistics; the honest significance is ~2–2.6σ, so the abstract overstates the case.","tokens_in":13785,"tokens_out":1856,"would_cite":false,"duration_ms":18912,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["95.35.+d","98.80.-k"],"model":"deepseek-v4-flash","headline":"A narrow redshift window around z ~ 10^4–10^5 makes dark-matter–neutrino interactions look real at >3σ.","keywords":["dark matter-neutrino interactions","cosmic shear","cosmic microwave background","resonant scattering","neutrino portal dark matter","small-scale structure","S8 tension","ultralight scalar field"],"falsifier":"If future 21-cm or lensing measurements find no suppression of the matter power spectrum around k ≈ 1 h/Mpc, the window scenario is ruled out; if they find strong suppression at k ≈ 10 h/Mpc, the narrow best-fit BP1 window (which leaves T²(k) ≈ 0.4 at those scales) is ruled out in favor of a wider window or constant interaction. Counts of Milky Way dwarf satellites around M ≈ $10^{8}$ M_⊙, where the model predicts near-ΛCDM abundances, would also adjudicate.","tokens_in":12629,"feed_emoji":"🔭","tokens_out":9627,"duration_ms":92267,"temperature":0.7,"pith_summary":"Recent CMB, weak-lensing, and Lyman-α analyses hint that dark matter may scatter off neutrinos, but the simplest constant-cross-section reading conflicts with dwarf-galaxy and astrophysical bounds. This paper shows that the tension dissolves if the interaction is switched on only in a narrow redshift window, roughly z ~ $10^{4}$–$10^{5}$. In a toy model with constant strength inside [zmin, zmin+Δz] and zero outside, combined ACT and cosmic-shear data prefer a non-zero interaction at >3σ, with best-fit strength uνDM ~ $10^{-3}$ confined to 3.7×$10^{4}$ ≲ z ≲ $10^{6}$. The paper then constructs a neutrino-portal particle model with a resonantly enhanced scattering cross section that peaks in exactly this epoch, and shows it can match the data while evading other constraints. A narrow feature in interaction strength, normally overlooked in cosmological modeling, can therefore leave a wide observable imprint.","feed_headline":"Narrow epoch yields >3σ dark-matter–neutrino signal","feed_subtitle":"The interaction need not be constant: confining it to z≈10^4–10^5 dodges galaxy bounds and fits CMB plus shear data.","key_machinery":"The argument runs through the Boltzmann drag term for the CDM velocity divergence, $\\dot{\\theta}_\\chi = k^2\\psi - H\\theta_\\chi - K_\\chi \\dot{\\mu}_\\chi(\\theta_\\chi - \\theta_\\nu)$ with $K_\\chi = (4/3)\\rho_\\nu/\\rho_\\chi$, where the interaction rate $\\dot{\\mu}_\\chi = a n_\\chi \\sigma_{\\nu\\mathrm{DM}}$ is promoted from a constant to a top-hat function on [zmin, zmin+Δz]. For the particle model, the central identity is the resonance position $E_{\\nu,\\mathrm{res}} \\simeq (m_\\phi^2 - m_\\chi^2)/(2m_\\chi)$ of the s-channel mediator, whose Breit–Wigner shape is smeared by convolving with the Fermi–Dirac neutrino spectrum to give an effective $u_{\\mathrm{eff},\\nu\\mathrm{DM}}(z)$; a field-dependent mass $m_\\chi(\\phi) = m_\\chi(0)(1+g_\\phi\\phi^2/2)$ then shifts the resonance as the ultralight scalar field oscillates, cutting the interaction off at high redshift.","core_discovery":"The central claim is that the observational preference for dark matter–neutrino interactions does not require the interaction to be constant in time; a localized enhancement at z ~ $10^{4}$–$10^{5}$ reproduces the preference and bypasses small-scale-structure bounds. In the windowed toy model the posterior for uνDM shifts to larger values than the constant-σ best fit, peaking near $10^{-2}$.5 with a 95% lower bound above $10^{-3}$.65, and the best-fit window (BP1) improves χ² by 16.6 over ΛCDM. The paper shows that this window lies where the neutrino energy distribution overlaps an s-channel resonance of a light scalar mediator in a neutrino portal model, and that a tiny time-dependent dark matter mass, induced by an ultralight scalar field, can suppress the interaction at z ≳ few×$10^{5}$ to satisfy Milky Way satellite constraints.","pith_inferences":["The paper's window is chosen partly to duck the Milky Way satellite bound; a corollary the authors leave implicit is that any observation probing z ≳ 10^5 with comparable sensitivity would close the loophole, so future experiments that map the pre-reionization epoch are the decisive test.","Because the posterior tail in Δz is driven by masked small scales, the model predicts that unmasked lensing surveys with baryonic-feedback modeling will either sharpen Δz to a narrow value or shift the preferred window to earlier times; this is a testable prediction of the analysis procedure itself.","The time-varying DM mass needed to mimic a top-hat window suggests an alternative, possibly simpler construction: a resonance whose position is set by the temperature of a dark sector bath rather than by a scalar field, which would make ueff(z) a smooth function with a similar peak; such a model could be checked for the same Δχ² while predicting distinct high-z behavior.","The paper's benchmark puts the resonance around neutrino energies of order 1–100 eV at z ~ 10^4–10^5; if that is correct, laboratory searches for MeV-scale mediators with couplings g ~ 10^-4 are near the sensitivity frontier, and null results would constrain the viable parameter region."],"forward_implications":["If the central claim holds, the epoch z ~ 10^4–10^5 becomes a single lever by which νDM interactions can satisfy all current bounds, so targeted CMB and lensing analyses around this epoch are the direct next test.","The windowed model strengthens the preference relative to ΛCDM (Δχ² = −16.6 for BP1) and beats the constant-cross-section model by 2.3σ, so future data that reproduce the window would disfavor the latter.","The intermediate value S8 ≈ 0.8 predicted by the windowed fits lies between the constant-νDM value (0.77) and a recent weak-lensing legacy measurement (0.815), softening the S8 tension without abandoning ΛCDM's cosmological parameters.","In the neutrino portal realization, the same resonance that produces the cosmological signal automatically suppresses the interaction at both high and low neutrino energies, keeping the model below astrophysical bounds from MeV-scale neutrinos and heavy neutral lepton searches.","Adding the 10^-8-level DM mass variation from an ultralight scalar makes the transfer function coincide with the Lyman-α best fit near k ~ 5 h/Mpc while recovering ΛCDM at dwarf-galaxy scales, a shape difference that the paper argues distinguishes it from warm dark matter."],"supporting_citations":[{"why":"Supplies the Lyman-α best-fit uνDM (3.8×10^-6) and the transfer function the benchmark models are compared against.","marker":"[12]"},{"why":"The previous constant-uνDM analysis of CMB, BAO, and cosmic shear whose preference and S8 value this paper extends and contrasts with.","marker":"[16]"},{"why":"Gives the Milky Way dwarf-galaxy bound on νDM scattering that forces suppression at z ≳ few×10^5 and motivates the Δz ≲ 10^6 prior.","marker":"[41]"},{"why":"Provides the momentum-dependent scattering-rate formalism used to convolve the resonance with the neutrino spectrum.","marker":"[44]"},{"why":"DES Y3 cosmic-shear likelihood; one of the two datasets that jointly drive the >3σ preference.","marker":"[46]"},{"why":"Planck 2018 legacy CMB likelihoods included in the analysis.","marker":"[47]"},{"why":"Gives the damped-oscillator solution for the ultralight scalar field used to implement the time-varying DM mass.","marker":"[75]"},{"why":"ACT DR4 CMB likelihood, the other dataset that drives the preference.","marker":"[77]"}],"fun_headline_variants":["Narrow redshift boost: >3σ νDM signal from CMB and shear","Epoch-restricted νDM coupling: >3σ over constant-model fits","z≈10^4–10^5 νDM boost: CMB and shear prefer >3σ","Narrow νDM interaction epoch dodges galaxy bounds","Redshift-limited νDM fits CMB and shear at >3σ"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The >3σ preference is conditional on the ad hoc assumption that νDM interactions are exactly zero outside a finite redshift window and constant inside it; the posterior tail in Δz is unconstrained because small scales are masked, and the physical model must tune a time-varying DM mass to mimic that window.","fun_headline_variants_meta":{"raw":{"variants":["Narrow redshift boost: >3σ νDM signal from CMB and shear","Epoch-restricted νDM coupling: >3σ over constant-model fits","z≈10^4–10^5 νDM boost: CMB and shear prefer >3σ","Narrow νDM interaction epoch dodges galaxy bounds","Redshift-limited νDM fits CMB and shear at >3σ"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001523,"raw_usage":{"total_tokens":6056,"prompt_tokens":860,"completion_tokens":5196,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":476,"completion_tokens_details":{"reasoning_tokens":5094}},"tokens_in":476,"tokens_out":5196,"duration_ms":39018,"temperature":1.0,"reasoning_tokens":5094,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T13:54:36.064295+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"If future 21-cm or lensing measurements find no suppression of the matter power spectrum around k ≈ 1 h/Mpc, the window scenario is ruled out; if they find strong suppression at k ≈ 10 h/Mpc, the narrow best-fit BP1 window (which leaves T²(k) ≈ 0.4 at those scales) is ruled out in favor of a wider window or constant interaction. Counts of Milky Way dwarf satellites around M ≈ $10^{8}$ M_⊙, where the model predicts near-ΛCDM abundances, would also adjudicate.","supporting_citations":[],"review_version":1}