{"id":"c5686994-369c-4253-b3a1-b18654a5888d","arxiv_id":"2607.24412","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Joint Bayesian fit of uGMRT Ω_HI and CHIME 21-cm power spectrum to a three-parameter HI–halo mass relation predicts a z≈1 HIMF with excess high-mass HI galaxies relative to prior work.","lead":"Researchers jointly fit two z≈1 radio observations to map how neutral hydrogen sits inside dark-matter halos, then predict the HI mass function. The result suggests more massive HI-rich galaxies at cosmic noon than earlier surveys and simulations implied.","discovery_kind":"extension","skeptic_critique":{"model":"moonshotai/kimi-k3","headline":"The HIMF is a model-mapped derivative of two integral/clustering observables; the zero-scatter deterministic HIHM in eq. 1 can control the high-mass excess more than the data do.","rationale":"I agree with the reader that the decisive soft spot is the assumed HI–halo mapping rather than the Bayesian machinery. The inference setup is transparent: emcee over (logα,β,logvc0), Gaussian likelihood, CHIME covariance, and a clear prior table. The simulations and normalization scaling are standard for this pipeline, and the authors are appropriately explicit that the HIMF follows from binning halos populated through eq. 1. That transparency strengthens the paper, but it also exposes the limitation: Ω_HI is one number and the CHIME PS is a small set of correlated bandpowers, so they cannot independently determine the full functional shape or stochasticity of M_HI(M_h). The most consequential place this bites is exactly the headline result, the high-mass HIMF excess over CH24 and hydro simulations, because rare-object abundances are exponentially sensitive to scatter and cutoff location. This is not an accusation of inconsistency; the authors scope the result to the model and show CH24’s parameters outside their posterior. My stress test therefore recommends keeping CONDITIONAL: the claim is credible as a conditional prediction of this HIHM class, but the decisive robustness check is to show the high-mass tail survives scatter/assembly-bias/generalized HIHM freedom while still fitting the same two observables.","tokens_in":9081,"tokens_out":2247,"duration_ms":87566,"concrete_test":"Refit the same uGMRT Ω_HI and CHIME P(k) covariance after adding a lognormal scatter σ_HI=0.2,0.35,0.5 dex in M_HI at fixed M_h (and, in a second run, one extra high-mass-slope parameter multiplying eq. 1 for M_h>M_cut). Propagate each posterior to the HIMF exactly as in §2–3. If the 68/95% HIMF band at M_HI≳3×10^10 M⊙ shifts enough to include CH24/TNG/EAGLE/SIMBA, or if σ_HI is unconstrained while the tail changes by >1σ, the high-mass excess is an artifact of the deterministic HIHM rather than a robust inference.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central prediction in Fig. 3 is not measured directly: §2 populates FoF halos with the Padmanabhan et al. (2017) form (eq. 1), assigns each halo’s HI the host peculiar velocity (HC method), and obtains the HIMF by binning the resulting M_HI values. The likelihood (eq. 4) constrains essentially Ω_HI and eight CHIME bandpowers over 0.4≲k≲1.5 h/Mpc. Those data can fix combinations of (α,β,vc0), but the high-mass HIMF tail sits on the exponentially steep part of the halo mass function, where modest scatter or assembly-bias in M_HI at fixed M_h is strongly amplified. A deterministic HIHM transfers the HMF cutoff almost one-to-one into the HIMF; adding lognormal scatter would up-scatter abundant lower-mass halos into rare high-M_HI bins and down-scatter massive halos, changing the M_HI≳3×10^10 M⊙ abundance even if Ω_HI and P(k) remain fit. The paper’s own evidence of model sensitivity is that the CH24-fitting HIHM point lies outside the 95% contour while still using the same functional family, and the lowest-k CHIME bin is ∼3σ from the best fit, hinting at missing stochastic/scale-dependent freedom. Thus the “larger abundance than CH24/hydro” claim is plausible within this restricted HIHM class but should not be read as data-driven until stochastic/alternative HIHM mappings are shown to preserve it.","agreement_with_reader":"agree"},"referee_report":{"model":"moonshotai/kimi-k3","summary":"The authors constrain the three parameters of the Padmanabhan et al. (2017) HI mass–halo mass (HIHM) relation at z≈1 by jointly fitting, within an N-body simulation framework, two existing measurements: the uGMRT stacking estimate Ω_HI = (4.5±1.1)×10^{-4} at z≈1.06 and the eight CHIME 21-cm autocorrelation bandpowers at z≈1.16 spanning 0.4 ≲ k ≲ 1.5 h/Mpc. HI is painted onto FoF halos deterministically via eq. (1) (with halo peculiar velocities assigned by the HC method), and an MCMC over (log α, β, log v_c0) yields closed, strongly correlated posteriors. Using the inferred posterior, they predict the z≈1 HI mass function: nearly flat below ~3.6×10^9 M_⊙, sharply declining above, with ~90% of the HI in M_HI ∈ [2×10^9, 4×10^11] M_⊙, and a larger abundance of high-mass (M_HI ≳ 3×10^10 M_⊙) HI galaxies than inferred by Chowdhury et al. (2024) (CH24) or predicted by TNG100/EAGLE/SIMBA.","tokens_in":9417,"tokens_out":3282,"duration_ms":109592,"significance":"If the result holds, this is a useful and timely contribution: direct HIMF measurements are unavailable beyond z≈0.1, and this work provides an independent z≈1 estimate anchored to actual 21-cm data rather than to hydrodynamical assumptions. Strengths that deserve explicit credit: (i) the joint likelihood correctly incorporates the published CHIME covariance and the uGMRT Ω_HI error in a single Bayesian framework; (ii) the posteriors are closed and the parameter correlations are displayed, so the inference is checkable; (iii) the use of 50 independent realizations and a documented interpolation/scaling procedure makes the forward model reproducible; (iv) the predicted HIMF is a genuinely falsifiable output — it can be tested by future CHIME/MeerKAT/SKA-Mid measurements and by direct HIMF estimates. The main significance caveat is that the most newsworthy element (the high-mass excess over CH24 and hydro simulations) is currently conditional on a rigid, zero-scatter HIHM mapping, and the paper itself contains hints (the excluded CH24-fitting point; the 3σ low-k residual) that the restricted model class is doing some of the work.","major_comments":[{"comment":"The paper's headline result — a larger abundance of M_HI ≳ 3×10^10 M_⊙ galaxies than CH24 and hydrodynamical simulations (Fig. 3, §4) — is obtained by binning M_HI values assigned through the deterministic, zero-scatter HIHM of eq. (1). The likelihood (eq. 4) constrains only Ω_HI and eight CHIME bandpowers over 0.4 ≲ k ≲ 1.5 h/Mpc; these data fix combinations of (α, β, v_c0) but do not directly probe the steep high-mass tail, which sits on the exponential part of the halo mass function. With a deterministic mapping, the HMF cutoff transfers almost one-to-one into the HIMF; even modest lognormal scatter in M_HI at fixed M_h (or assembly bias) would up-scatter abundant lower-mass halos into rare high-M_HI bins and could materially change the claimed excess while leaving Ω_HI and P(k) fits essentially intact. The authors' own Fig. 2 hints at model-class sensitivity: the CH24-fitting HIHM po","section":"§3–4, Fig. 3, eq. (1)"},{"comment":"The authors note that the smallest-k CHIME bin lies ~3σ from the best-fit PS (§3, Fig. 1), but the point is not pursued. With only eight bandpowers, a 3σ residual in the lowest-k bin is a meaningful goodness-of-fit concern: it may indicate missing scale-dependent freedom in the model (e.g., redshift-space distortion treatment, the HC peculiar-velocity assignment, or scale dependence the rigid 3-parameter HIHM cannot absorb). At minimum, a reduced χ² for the best fit, a discussion of whether the misfit is a statistical outlier given the covariance, and its effect on the posterior (e.g., refitting excluding that bin) should be reported.","section":"§3, Fig. 1"},{"comment":"The HIMF in Fig. 3 is plotted down to M_HI = 10^7 M_⊙, yet the halo catalog has a minimum mass of 1.089×10^9 M_⊙ (10 particles), and eq. (1) with the best-fit Mcut ≈ 4×10^11 M_⊙ exponentially suppresses M_HI for low-mass halos. It is unclear how the flat plateau at ~7×10^-3 Mpc^-3 dex^-1 for M_HI ≲ 3.6×10^9 M_⊙ arises from resolved halos, and the text's statement that the plateau 'corresponds to the HI contribution of the halos below the low-mass cutoff' (§4) is confusing given the exp(-Mcut/M_h) suppression in eq. (1). Please clarify the origin of the plateau, indicate which part of the plotted HIMF is resolution-limited, and state the halo-mass (or M_HI) completeness limit explicitly on the figure.","section":"§3, Fig. 3; §2"},{"comment":"The [150.08 Mpc]^3 particle-mesh simulation (§2) is used to model bandpowers down to k ≈ 0.4 h/Mpc, where the box contains few independent modes, and up to k ≈ 1.5 h/Mpc, where PM force resolution and the absence of subhalo/satellite structure (HC method assigns each halo a single peculiar velocity, suppressing fingers-of-god) may matter. Fifty realizations mitigate mean estimation but not systematic resolution effects on P(k). Please quantify: (i) the grid/Nyquist scale and the PM resolution relative to k = 1.5 h/Mpc; (ii) the expected impact of neglecting intra-halo velocity dispersion on the redshift-space PS over the fitted k-range; (iii) whether sample variance in the simulated Ω_HI(θ) should enter the likelihood alongside the observational ΔΩ_HI.","section":"§2"}],"minor_comments":[{"comment":"The uGMRT Ω_HI input (Chowdhury et al. 2020) is derived by stacking blue, star-forming galaxies; any residual correction from that sample to the full galaxy population propagates directly into the HIHM normalization. A sentence on this systematic, and its size relative to the ±1.1×10^-4 statistical error, would help.","section":"§1–2"},{"comment":"The likelihood is evaluated at a single simulation redshift z = 1, while the two observables have different effective redshifts (uGMRT z ≈ 1.06, CHIME z ≈ 1.16, with the CHIME band spanning z = 1.01–1.34). Please comment on the evolution of the HIHM/HIMF across this interval and whether it is negligible.","section":"§2, eq. (4)"},{"comment":"The α-dependence is obtained by scaling Ω_HI ∝ α and P(k) ∝ α² from a single α_fix = 0.09 grid (§2). This is exact only if shot noise scales the same way as the clustering term; since both scale with ΣM_HI² at fixed halo population this appears to hold, but it should be stated explicitly, and the accuracy of the linear interpolation on the 11×11 (β, log v_c0) grid should be quantified.","section":"§2"},{"comment":"The comparison with CH24 (Fig. 3, §4) should note that the CH24 HIMF is itself an indirect estimate specific to star-forming galaxies; the phrase 'earlier observations' in the abstract overstates the directness of that comparison.","section":"§4"},{"comment":"Fig. 3 caption mentions a purple dotted curve ('fits CH24') that is discussed in §4 but is not clearly identifiable in the figure description; please check color/style consistency between text, caption, and figure.","section":"Fig. 3"},{"comment":"Citation details: Gibbon et al. (2015) is cited for MeerKAT 21-cm PS detections, which appears to be an instrumentation paper — please verify; Gupta et al. (2017) lacks volume/page; several arXiv-only entries (CHIME 2025, 2026; Paul et al. 2023) may have since appeared in journals.","section":"References"},{"comment":"Notation: 'M_HI' loses its subscript in a few places in §1 ('relateM HI the HI mass...'); units for M_HI are missing in the abstract's low-mass bound as typeset; the units of v_c0 in Table 1 should be stated in the column header.","section":"§1, Table 1"}],"recommendation":"major_revision","confidential_remarks":"The simulations, HIHM family, and much of the methodology are shared with the authors' companion forecast paper (Chhabra & Bharadwaj 2025, arXiv:2508.19126); the present work is nonetheless a distinct, data-constrained result and the overlap is disclosed in §2. The novelty is the joint use of the new CHIME PS with uGMRT stacking to pin the z≈1 HIHM; this is timely and within journal scope. My main reservation — that the headline high-mass HIMF excess is partly a property of the assumed deterministic HIHM mapping rather than of the data — is detailed in the major comments and is, in my view, addressable within the scope of the manuscript."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The new piece here is a joint posterior on the three-parameter Padmanabhan HIHM at z≈1, driven by the recent CHIME auto-PS plus the uGMRT stacking Ω_HI, and the HIMF that falls out of it. That HIMF prefers more high-mass systems than CH24 or the usual hydro runs, and the paper is transparent about how it got there.\n\nWhat they do well is straightforward. The likelihood folds in the CHIME covariance and the Ω_HI error correctly. Posteriors close, correlations are shown, 50 realizations plus the α-scaling/interpolation are documented, and the HIMF is a derived prediction rather than an input. Comparisons to CH24, Padmanabhan+2017, ALFALFA, and the hydro simulations are clear and fair. Within the assumed HIHM family this is a solid, usable constraint for people who need a z≈1 HI distribution for IM forecasts or galaxy-evolution checks.\n\nThe soft spots are real but scoped. The HIMF is not measured; it is the HMF mapped through a deterministic, zero-scatter HIHM. Modest lognormal scatter or assembly bias at fixed M_h would move the high-mass tail even while still fitting Ω_HI and the eight bandpowers, so the claimed excess over CH24/hydro is model-class dependent. Ω_HI comes only from blue star-forming galaxies, the lowest-k CHIME bin sits ~3σ off the best fit, and the box/resolution are modest. None of that breaks the internal logic; it just means the result should be quoted as “inside this HIHM” rather than as a model-independent census.\n\nThis is for people working on 21-cm IM modeling, HIHM calibration, or z~1 galaxy gas content. Math, data handling, and citation pattern look solid. I would send it to referees; the limitations are the ordinary ones for the subfield and are already visible in the text. Worth engaging if you need a concrete z≈1 HIMF prior, with the usual caveats on functional form.","headline":"Clean joint fit of CHIME PS + uGMRT Ω_HI yields a usable z≈1 HIMF posterior inside a standard HIHM model, with a high-mass excess that is real within that model but not yet model-independent.","tokens_in":10512,"tokens_out":590,"would_cite":true,"duration_ms":11767,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"Joint uGMRT stacking and CHIME 21-cm power-spectrum data imply more high-mass HI galaxies at z≈1 than earlier estimates and hydro simulations.","keywords":["HI mass function","21-cm intensity mapping","HI-halo mass relation","neutral hydrogen","z≈1","CHIME","uGMRT","galaxy evolution"],"falsifier":"A direct or stacked measurement of the abundance of z≈1 galaxies with M_HI ≳ 3×10^10 M_⊙ that falls below the paper’s posterior prediction, or a new 21-cm power spectrum whose shape cannot be reproduced by any point inside the reported HIHM posterior.","tokens_in":10252,"feed_emoji":"📡","tokens_out":986,"duration_ms":35560,"temperature":0.7,"pith_summary":"This paper constrains how neutral hydrogen is distributed among galaxies at redshift about one by fitting a simple three-parameter relation between halo mass and HI mass to two independent 21-cm measurements: the cosmic HI density from stacked uGMRT detections of star-forming galaxies, and the CHIME autocorrelation power spectrum. From the resulting posterior it predicts the HI mass function, finding it nearly flat at low masses, falling steeply above a few times 10^9 solar masses, and placing roughly ninety percent of all HI in the window 2×10^9 to 4×10^11 solar masses. The high-mass end is more abundant than an earlier indirect HIMF estimate and than current hydrodynamical simulations. Because this is the epoch when cosmic star formation peaks, a larger reservoir of massive HI systems would change how galaxies grow and how intensity-mapping surveys should be interpreted.","feed_headline":"More high-mass HI galaxies at z≈1 than models predict","feed_subtitle":"Joint uGMRT and CHIME 21-cm data put most neutral gas in one mass window","key_machinery":"The three-parameter HI mass–halo mass (HIHM) relation used to assign HI to dark-matter halos in an N-body simulation; its parameters are fixed by joint Bayesian matching of simulated Ω_HI and the redshift-space 21-cm power spectrum to the two observations, after which the HIMF is read off by binning the assigned HI masses.","core_discovery":"Using the posterior of a three-parameter HI-mass–halo-mass relation jointly constrained by uGMRT Ω_HI and the CHIME 21-cm power spectrum, the predicted z≈1 HI mass function stays nearly constant below ≲3.6×10^9 M_⊙, declines rapidly at higher mass, contains ~90 percent of the HI in M_HI ∈ [2×10^9, 4×10^11] M_⊙, and yields a larger abundance of high-mass HI galaxies than earlier observations and hydrodynamical simulations.","pith_inferences":["The tension with hydro simulations and with the earlier HIMF may point to missing physics in how massive halos retain cold gas at z≈1.","If the high-mass excess is real, 21-cm intensity mapping will be more sensitive to rare, massive hosts than bias models calibrated on steeper HIMFs assume.","Adding the 21-cm bispectrum to the same joint-likelihood pipeline could break remaining parameter degeneracies without requiring a complete galaxy survey."],"forward_implications":["Galaxy-evolution models must allow more massive HI reservoirs near cosmic noon than current hydro simulations supply.","Intensity-mapping forecasts that adopt steeper high-mass HIMFs will under-predict the small-scale 21-cm power.","About ninety percent of the cosmic HI at z≈1 sits in a relatively narrow mass window that future surveys can target preferentially.","Joint analyses with CO, [CII] or [OIII] line-intensity maps can test whether the same high-mass systems dominate the molecular-gas budget.","Tighter CHIME, MeerKAT and SKA-Mid measurements of Ω_HI, the power spectrum and the bispectrum can shrink the HIHM posterior and the HIMF prediction."],"fun_headline_variants":["z≈1 HIMF flat below 3.6e9 M_⊙ then declines; 90% HI in one mass window","Joint uGMRT-CHIME data yield more high-mass HI galaxies than models","HIHM posterior puts most z≈1 neutral gas in 2e9–4e11 M_⊙ range","Predicted z≈1 HI mass function exceeds earlier high-mass abundances","Three-parameter HIHM fit to Ω_HI and CHIME PS constrains z≈1 HIMF"],"cache_read_input_tokens":128,"weakest_assumption_plain":"That a single deterministic three-parameter formula linking each halo’s mass to its HI mass, with no extra scatter or environment dependence, is enough to turn a match to total HI density and the power spectrum into a unique HI mass function.","fun_headline_variants_meta":{"raw":{"variants":["z≈1 HIMF flat below 3.6e9 M_⊙ then declines; 90% HI in one mass window","Joint uGMRT-CHIME data yield more high-mass HI galaxies than models","HIHM posterior puts most z≈1 neutral gas in 2e9–4e11 M_⊙ range","Predicted z≈1 HI mass function exceeds earlier high-mass abundances","Three-parameter HIHM fit to Ω_HI and CHIME PS constrains z≈1 HIMF"]},"model":"grok-4.5","effort":"low","cost_usd":0.004542,"raw_usage":{"total_tokens":1366,"prompt_tokens":859,"num_sources_used":0,"completion_tokens":111,"cost_in_usd_ticks":45424000,"prompt_tokens_details":{"text_tokens":859,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":396,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":859,"tokens_out":111,"duration_ms":7435,"temperature":1.0,"reasoning_tokens":396,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-31T15:31:53.521260+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"A direct or stacked measurement of the abundance of z≈1 galaxies with M_HI ≳ 3×10^10 M_⊙ that falls below the paper’s posterior prediction, or a new 21-cm power spectrum whose shape cannot be reproduced by any point inside the reported HIHM posterior.","supporting_citations":[],"review_version":1}