{"id":"877db56f-bce1-407d-bab9-e00a35509ae5","arxiv_id":"2508.21812","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A JAX-based Hamiltonian Monte Carlo pipeline infers local IGM damping wing statistics from quasar spectra and maps them to global reionization constraints with precision comparable to direct inference.","lead":"This paper presents a Bayesian inference framework that measures local properties of the neutral hydrogen gas in front of distant quasars during the epoch of reionization. It demonstrates on mock spectra that the framework recovers the gas column density and the distance to the first neutral patch, and can convert these local measurements into global reionization timing constraints.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Local precision and equivalence claims depend on the Kist+25b topology-invariance of the toy-built likelihood; this paper neither verifies it nor avoids circular mock validation, so the quoted 0.69 dex/31.4 cMpc may not transfer to real quasars.","rationale":"The paper's strongest claim is that local damping-wing statistics can be inferred precisely and that global constraints retain or improve statistical fidelity. The weakest link is the asserted reionization-model independence of the likelihood, which is cited from a companion paper rather than established here. If that invariance is false, the toy-based likelihood is incorrect and both the local and converted global constraints fail. The circularity of mock ensemble (ii), which generates mocks from the same toy prescription used to build the likelihood, makes the quoted precision numbers less convincing: they measure self-consistency of the toy model, not the fidelity of the method on realistic topologies. The coverage tests on realistic mocks are the manuscript's best direct evidence, but the need for explicit posterior broadening signals that the approximate likelihood is not fully faithful; coverage correction can hide a misshapen likelihood rather than validate it. This is not a fatal flaw: the companion paper may indeed demonstrate the required invariance, and the method is presented as a framework with clear caveats. But the central claim is conditional on that unpublished result and on an external re-derivation of the likelihood from realistic topologies. The reader's verdict of CONDITIONAL already captures this correctly, so no verdict change is needed; the concrete test above is the step that would settle whether the concern actually lands.","tokens_in":32407,"tokens_out":4664,"duration_ms":61541,"concrete_test":"Recompute the IGM transmission likelihood directly from the realistic semi-numerical 21cmFAST sightlines, binning by the same (log N_HI^DW, r_patch) grid used for mock ensemble (ii): log N_HI^DW in [20.1, 20.7] and r_patch in [2, 14] cMpc, at tQ = 10^4, 10^6, 10^8 yr, with at least ~800 sightlines per bin. Compare the resulting mean vectors and covariance matrices against those used in the paper's toy-based likelihood via a quantitative metric (e.g., KL divergence or chi2 distance on held-out spectra). If the realistic and toy likelihoods differ significantly, rerun the precision and coverage calculations with the realistic likelihood; if the 0.69 dex and 31.4 cMpc numbers shift by more than the quoted error bars, the headline precision claims are not transferable to real quasar spectra.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claims—0.69 dex precision on log N_HI^DW, 31.4 cMpc on r_patch, and improved statistical fidelity for global constraints—all rest on the topology-independence of the IGM transmission likelihood in the local (log N_HI^DW, r_patch) parameterization. The likelihood (Eq. 11) uses Gaussian means/covariances estimated from synthetic toy xHI skewers (Section 2.2.1). The only justification that this toy prescription reproduces the mean and variance of realistic reionization topologies is inherited from Kist et al. (2025b), which is cited but not demonstrated or independently verified in this manuscript. This is load-bearing because the likelihood is the entire bridge from local summary statistics to observed spectra; if the invariance fails, every inferred constraint is invalid. Moreover, mock ensemble (ii), the source of the headline precision numbers, draws its IGM transmission profiles from the same toy prescription used to build the likelihood (Section 3.4). The validation is therefore partially circular: the pipeline is tested on data generated by the same approximate model it assumes. Coverage tests on realistic-topology mocks (ensemble (i), Figure 6) provide some internal evidence, but those tests are passed only after an explicit posterior-broadening correction (Section 3.5). Such a correction can compensate for an overconfident or misshapen likelihood without establishing that the shape is correct; it does not rescue a biased mean or a wrong covariance structure. Thus the key open question is whether the toy-based likelihood is actually statistically equivalent to the realistic-topology likelihood in the high-density, short-r_patch region where the quoted precision is measured.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents a fully Bayesian HMC inference framework for quasar Lyα damping wings, in which the IGM absorption is parameterized by two 'local' summary statistics—the Lorentzian-weighted HI column density log N_HI^DW and the distance r_patch to the first neutral patch in the pre-quasar IGM—together with the quasar lifetime log t_Q. The IGM transmission likelihood is built from a Gaussian mean/covariance model estimated from synthetic toy xHI skewers (Sec. 2.2.1), justified by a topology-invariance claim inherited from the companion paper Kist et al. (2025b). The pipeline jointly reconstructs the quasar continuum using PCA coefficients. The authors perform coverage tests on realistic-topology mocks, apply a posterior-broadening correction, and report precision from a second mock ensemble: 0.69 dex on log N_HI^DW and 31.4 cMpc on r_patch for strong damping-wing cases. They show that converting local constraints to global (⟨x_HI⟩, log t_Q) yields precision comparable to direct global inference and improved coverage.","tokens_in":32802,"tokens_out":5947,"duration_ms":74091,"significance":"If the topology-invariance assumption is correct, this is a valuable and timely framework: it cleanly separates inference of local absorption geometry from reionization-model assumptions, which can be folded in later as priors via Eq. (10). The joint continuum–IGM inference, the use of realistic mock ensembles, and the systematic coverage-testing procedure are clear strengths and represent good statistical practice. The paper is well written and the methodology is reproducible in principle. However, the headline local precision figures are obtained from mocks generated with the same toy prescription that defines the likelihood, and the central topology-invariance property is only cited, not demonstrated here. The coverage correction is a necessary safeguard but cannot by itself justify the transfer of the quoted precision to realistic reionization topologies. The contribution is therefore potentially significant, but the central empirical claims require additional validation before they can be accepted.","major_comments":[{"comment":"The quoted local precision values in the abstract and §5 (0.69 dex on log N_HI^DW and 31.4 cMpc on r_patch) are derived from mock ensemble (ii), whose IGM transmission profiles are simulated with the same toy xHI prescription used to build the likelihood (§2.2.1). This is a circular validation: it tests self-consistency with the assumed model, not accuracy on realistic data. The realistic-topology ensemble (i) is used only for coverage tests (§4.2, Fig. 6), where the local parameter posteriors are overconfident and require HDR relabeling (§3.5). Such a correction can rescale the reported credible intervals but cannot repair a biased mean or a wrong covariance structure. The manuscript should validate the local inference on realistic-topology sightlines with known true (log N_HI^DW, r_patch) values, or otherwise demonstrate that the toy-based likelihood transfers to realistic topologies.","section":"§3.4, §4.3.1"},{"comment":"The entire likelihood rests on the claim that the mean and variance of IGM transmission at fixed (log N_HI^DW, r_patch) are independent of reionization topology. This is cited to Kist et al. (2025b) but is not tested in this paper. Because Eq. (11) uses toy-profile means and covariances, a failure of this invariance would invalidate all inferred constraints, not just the quoted precision. Please include a direct comparison of toy versus realistic transmission-profile means and covariances at matched local parameter values, or a sensitivity analysis that explores plausible topology differences and their impact on the inferred posteriors.","section":"§2.2, Eq. (11)"},{"comment":"The coverage-correction procedure guarantees by construction that the relabeled HDRs contain the truth with the stated frequency, but it does not change the ordering of probability density: it is a monotone relabeling of the original posterior. If the Gaussian likelihood is overconfident in some regions and underconfident in others, the correction cannot fix the shape of the posterior. The black coverage curve in Fig. 6 shows C_alpha < alpha for the local parameters, indicating systematic overconfidence. The nearly perfect coverage of the converted (⟨x_HI⟩, log t_Q) marginals (green curve) may arise because the probabilistic mapping from local to global parameters is highly non-injective and smooths out local errors. The statement in §5 that the method removes 'the need for coverage corrections' for global constraints is not supported, since the local stage itself relies on the correctio","section":"§3.5, Eqs. (14)–(16)"}],"minor_comments":[{"comment":"The caption reads 'Contours correspond to the 68 %and 68 %percentile regions.' Presumably the second should be a different level (e.g., 95%). Please correct.","section":"Figure 2 caption"},{"comment":"The normalization notation N(u_min, u_max) in Eqs. (2) and (3) is used before being clearly defined; a short verbal definition or distinct symbol would improve readability.","section":"§2.2.1"},{"comment":"The phrase 'more nearby neutral patches are also the greatest contributors' is grammatically awkward; consider 'nearer neutral patches'.","section":"§2.2.2"},{"comment":"Several key references are 'in prep.' (Kist et al. 2025a; Hennawi et al. 2025a; Euclid Collaboration: Yang et al. 2025). If these are central to the current work, their availability should be clarified, and the topology-invariance result in Kist et al. (2025b) should be explicitly flagged as a premise of the likelihood construction.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper's central claims depend heavily on Kist et al. (2025b) for the topology-invariance property. If that paper is not yet accepted or publicly available, the editor should weigh the extent to which this manuscript can stand alone. The manuscript also cites several 'in prep' works; this makes it harder to evaluate the novelty and the provenance of the continuum model. The fit to MNRAS is otherwise good, but the validation gap described in the major comments should be addressed before acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is the first inference of the local damping-wing summary statistics (Lorentzian-weighted HI column density and distance to the first neutral patch) and a clean probabilistic bridge from those local constraints to global reionization parameters. The framework is well-built: HMC sampling, faithful continuum marginalization, coverage tests, and a principled posterior-broadening correction. The claim that converted global constraints match direct-inference precision while improving statistical fidelity is supported by the internal tests in Figures 6 and 8. The authors also deserve credit for being explicit about the Gaussian likelihood approximation and for flagging simulation-based inference as the future fix.\n\nThe soft spots are real but not fatal. The load-bearing assumption is the topology-independence of the IGM transmission likelihood in the local parameterization. That is inherited from Kist et al. (2025b), which is cited but not demonstrated here. If that invariance fails, the whole likelihood is wrong. The paper does not attempt to verify it directly, and the companion paper is not yet published, so a reader cannot check the central pillar. The precision numbers in the abstract (0.69 dex, 31.4 cMpc) come from mock ensemble (ii), which draws its transmission profiles from the same toy prescription used to build the likelihood. That is circular for the point estimates. The coverage tests on realistic-topology mocks (ensemble i) provide some external validation, but the posterior-broadening correction can fix overconfidence without correcting a biased mean or a wrong covariance shape. So the quoted precision may not transfer exactly to real quasars.\n\nThere is also no code or data release beyond \"on reasonable request,\" which is a practical limitation for a methods paper.\n\nOverall, the central logic holds up conditionally. This is a serious, honest methods paper, and the caveats are the kind that peer review can address. I would send it to review, asking the authors to either include the invariance demonstration or make the companion paper available, and to release the pipeline code and mock data. The precision claims should be hedged as coming from toy-model mocks until the invariance is independently established.","headline":"Careful Bayesian machinery for a genuinely new observable, but the headline precision numbers rest on an invariance result from an unpublished companion paper and a partly circular mock test — worth refereeing, with requests for evidence.","tokens_in":33280,"tokens_out":1784,"would_cite":false,"duration_ms":23593,"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":"A fully Bayesian inference framework extracts local, reionization-model-independent damping-wing parameters from quasar spectra and maps them to global reionization constraints without losing precision.","keywords":["quasar damping wings","intergalactic medium","reionization","Lyman-alpha absorption","neutral hydrogen column density","ionization topology","Bayesian inference","Hamiltonian Monte Carlo"],"falsifier":"Generate mock sightlines at several fixed (log N_DW_HI, r_patch) points from two deliberately different reionization topologies (e.g., inside-out vs outside-in bubbles) using the same density and velocity skewers, and compare the mean and covariance of the Ly-alpha transmission profiles; if the distributions differ beyond the few-percent level claimed, the topology-invariance assumption is falsified. A second, end-to-end check is to run the pipeline on 100+ mock spectra drawn from a realistic topology and test whether the coverage-corrected 68% highest-density regions contain the true local pa","tokens_in":32336,"feed_emoji":"🔭","tokens_out":10117,"duration_ms":103311,"temperature":0.7,"pith_summary":"The paper develops a Bayesian inference pipeline that extracts local, reionization-model-independent information from the Lyman-alpha damping wings of high-redshift quasar spectra. Instead of directly fitting the global neutral fraction, it constrains two pre-quasar summary statistics—the Lorentzian-weighted HI column density and the distance to the first neutral patch—along with the quasar lifetime, then maps these to global constraints through a probabilistic relation. On mock spectra, the pipeline recovers the column density to about 0.7 dex and the patch distance to about 31 cMpc when a damping wing is present, and it preserves the precision of direct global inference while improving statistical fidelity. The paper argues this opens the door to constraining the local ionization topology around quasars and comparing reionization models without redoing the spectral inference.","feed_headline":"Damping wings localize first neutral patch to ~31 cMpc","feed_subtitle":"Model-independent local IGM topology from quasar Ly-alpha spectra, matching global reionization precision.","key_machinery":"The load-bearing object is the Lorentzian-weighted HI column density N_DW_HI: a single number obtained by integrating the pre-quasar neutral hydrogen field with a (v - v_T)^-2 Lorentzian weight over 4 to 104 cMpc, normalized so it is directly proportional to the damping-wing optical depth at a reference velocity. The second statistic, r_patch, is the distance from the quasar to the first neutral patch in the pre-quasar topology, measured after smoothing the neutral fraction field. Together with the quasar lifetime t_Q these three parameters describe the transmission profile. The probabilistic bridge to global reionization parameters is Eq. (10): the topology-informed posterior on (x_HI, t_Q,","core_discovery":"The paper's central claim is that the information in a quasar's Lyman-alpha damping wing can be split cleanly into local quantities that are independent of how reionization actually proceeded, plus a separate statistical mapping that encodes all the model dependence. The local quantities are the Lorentzian-weighted HI column density in front of the quasar and the distance to the first neutral patch in the pre-quasar IGM; the quasar lifetime is added to capture the effect of the quasar's own ionizing radiation. The paper shows how to infer these three parameters jointly with the quasar continuum using a full generative model and Hamiltonian Monte Carlo sampling, and how to convert the resulti","pith_inferences":["A natural extension the paper only gestures at is applying the same local-statistics decomposition to galaxy damping wings, where intervening damped Ly-alpha systems are a major nuisance; the model-independent local likelihood could be combined with an empirical prior on proximate DLAs to separate IGM from absorber contributions.","Because the mapping from local to global parameters is stochastic and non-injective, combining several quasar sightlines hierarchically could constrain the reionization topology itself—e.g., the distribution of bubble sizes—not just the global neutral fraction.","The improved coverage after marginalizing over local parameters suggests that the local-then-global strategy acts as a form of robust marginalization over topology stochasticity; a similar two-step scheme might improve other probes where patchiness dominates the error budget.","A concrete testable extension would be to replace the Gaussian transmission likelihood with simulation-based inference on the same local parameters; if coverage improves further or constraints shift, that would show how much of the remaining overconfidence is due to the Gaussian approximation."],"forward_implications":["Quasar damping-wing analyses can be reported as model-independent local constraints (column density and neutral-patch distance), letting different reionization models be folded in afterward without rerunning the spectral fit.","For a given reionization model, converting local constraints yields global neutral-fraction and quasar-lifetime posteriors with the same precision as direct global inference, while the marginal global constraints pass coverage tests better than direct fits.","The framework extracts local topological information—most concretely the distance to the first neutral patch—that global pipelines discard, with precision up to <5 cMpc for short-lived quasars.","Comparing reionization models becomes a matter of comparing the induced priors P_top(log N_DW_HI, r_patch | x_HI), rather than redoing the likelihood calculation.","The main cost is computational: a higher-dimensional parameter space (three astrophysical parameters plus seven continuum coefficients) and a denser grid of simulated transmission profiles."],"supporting_citations":[{"why":"Introduces the two local summary statistics and establishes the reionization-topology-invariance of IGM transmission profiles at fixed local parameters, justifying the toy-skewer likelihood.","marker":"Kist et al. (2025b)"},{"why":"Supplies the fully Bayesian joint continuum-plus-IGM inference framework, the Gaussian likelihood approximation, and the coverage-correction procedure adapted here.","marker":"Hennawi et al. (2025b)"},{"why":"Provides the hybrid method for combining hydrodynamical sightlines with independent neutral-fraction skewers and 1D radiative transfer to simulate IGM transmission profiles.","marker":"Davies et al. (2018)"},{"why":"Provides the one-dimensional radiative transfer model for the quasar's ionizing radiation used to produce post-quasar transmission profiles.","marker":"Davies et al. (2016)"},{"why":"The Nyx cosmological hydrodynamical simulations supply the density, velocity, and temperature skewers used in all mock sightlines.","marker":"Almgren et al. (2013); Lukić et al. (2015)"},{"why":"The modified semi-numerical reionization code generates the realistic topologies that determine the prior P_top(log N_DW_HI, r_patch | x_HI) and the global-to-local conversion.","marker":"Mesinger et al. (2011); Davies & Furlanetto (2022)"},{"why":"Quantifies the error budget from continuum reconstruction and IGM stochasticity, guiding the choice of PCA dimensionality and the comparison of precision values.","marker":"Kist et al. (2025c)"},{"why":"Previously constrained the distance to the first neutral patch for galaxy damping wings, establishing r_patch as a physically meaningful summary statistic in a neighbouring context.","marker":"Mason et al. (2025)"}],"fun_headline_variants":["Damping wings reveal local IGM topology to 31 cMpc","Model-free local HI from quasar Ly-alpha damping wings","HMC recovers quasar damping wing distance to 31 cMpc","Local IGM damping wing constraints at 0.69 dex precision","Quasar damping wings split local vs global reionization info"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The whole pipeline rests on the companion paper's claim that the mean and variance of the Lyman-alpha transmission profile at fixed local parameter values do not depend on how the neutral patches are arranged along the sightline—if that invariance fails, the likelihood built from toy neutral-fraction skewers is invalid.","fun_headline_variants_meta":{"raw":{"variants":["Damping wings reveal local IGM topology to 31 cMpc","Model-free local HI from quasar Ly-alpha damping wings","HMC recovers quasar damping wing distance to 31 cMpc","Local IGM damping wing constraints at 0.69 dex precision","Quasar damping wings split local vs global reionization info"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000237,"raw_usage":{"total_tokens":1394,"prompt_tokens":842,"completion_tokens":552,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":586,"completion_tokens_details":{"reasoning_tokens":475}},"tokens_in":586,"tokens_out":552,"duration_ms":7708,"temperature":1.0,"reasoning_tokens":475,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T13:54:05.873732+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Generate mock sightlines at several fixed (log N_DW_HI, r_patch) points from two deliberately different reionization topologies (e.g., inside-out vs outside-in bubbles) using the same density and velocity skewers, and compare the mean and covariance of the Ly-alpha transmission profiles; if the distributions differ beyond the few-percent level claimed, the topology-invariance assumption is falsified. A second, end-to-end check is to run the pipeline on 100+ mock spectra drawn from a realistic topology and test whether the coverage-corrected 68% highest-density regions contain the true local pa","supporting_citations":[],"review_version":1}