{"id":"fdb01981-c62a-4ed2-9731-a0a56f25d15d","arxiv_id":"2603.02670","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"The z=0 HI column-density distribution function has been measured for the first time down to 10^17.8 cm^-2 from 1-kpc-resolution 21-cm images, showing modest evolution since z~3 at LLS-like densities.","lead":"Using 21-cm emission maps from two surveys, this paper measures how much sky area is covered by hydrogen gas at different column densities in the present-day universe, reaching a hundred times deeper than previous work. It provides the first local-universe benchmark for the faint gas that feeds galaxies, against which higher-redshift absorption measurements and simulations can be compared.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No completeness correction for the 50%-incomplete log N_HI=17.8 bin; the faint-end f(N_HI), length incidence, and f_cov may be understated by up to ~0.3 dex.","rationale":"Good faith: the paper is a careful, systematic effort, with mock tests and detailed error accounting. The high-column regime (19.8–21.3) is likely robust and the consistency with Z05 is a real check. The concern is not about the existence of deep HI imaging, but about the statistical completeness of the cosmic CDDF at the very column density that defines the paper's headline. The reader identified sample representativeness as the weakest assumption; the N_HI,lim incompleteness is a concrete, checkable instance of that failure that can be tested without new observations. If the test confirms a >0.1 dex shift at 17.8, then the abstract's 'first z=0 CDDF extending down to 17.8' should be qualified as a lower limit or corrected; the paper remains valuable but the central faint-end claim is not yet established. I keep the reader's CONDITIONAL disposition because the issue is addressable with existing data and the high-column results are unaffected.","tokens_in":38858,"tokens_out":18843,"duration_ms":182496,"concrete_test":"For each column-density bin, split the main sample into galaxies with N_HI,lim at least 0.1 dex below the bin center (detected) and those with N_HI,lim above it (unmeasured). Compute the HIMF weight fraction of the detected subset per M_HI bin. Recompute f(N_HI) at 17.8 and 18.0 using only detected galaxies, rescaling each M_HI bin by the inverse detected-weight fraction (i.e., assuming unmeasured galaxies have the same mean A(N) as detected galaxies in the same bin). If the corrected values differ from Table 3 by more than 0.1 dex, the published faint-end CDDF, ℓ(10^17.8), and f_cov are incomplete and must be re-derived or explicitly presented as lower limits.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4.1 reports that the median and 90th-percentile N_HI,lim of the 70-galaxy main sample are 10^17.8 and 10^18.0 cm^-2. Equation (1) sums A_i(log N_HI) over all galaxies with no term for the fraction of galaxies whose images can actually reach a given column-density bin. A galaxy with N_HI,lim > 17.8 contributes A_i = 0 at log N_HI=17.8, so at that column roughly half the sample contributes zero area. No completeness correction is described in §4.2 or §4.3. The resulting f(N_HI) at 17.8 is therefore a lower limit, not a complete cosmic CDDF. This propagates directly into the quoted length incidence ℓ(10^17.8)=0.267±0.033, the inferred f_cov~0.7, and the claim of weak z=0–3 evolution at 17.8<log N_HI<19.2; if non-detected galaxies have similar low-column areas to detected ones, log f is understated by up to ~0.3 dex, far above the quoted 0.04–0.06 dex errors. This is distinct from, and potentially larger than, the acknowledged HI-rich sample bias.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper combines FEASTS (FAST + interferometry) and MHONGOOSE 21-cm images to construct the z=0 HI column density distribution function f(N_HI) down to log N_HI = 17.8 cm^-2 at ~1 kpc resolution. Galaxy areas are counted in column-density bins and weighted using the HIMF and the inverse occupation of M_HI bins, following Zwaan et al. (2005). The authors report a Schechter-function fit over log N_HI = 17.8–22, compare with z~3 absorber-based measurements and with OWLS/TNG50 simulations, derive length incidence and impact-parameter distributions, and estimate covering fractions including f_cov≈0.7 for LLS-like gas and f_cov≈0.006 within the virial radius of Milky-Way-like galaxies. The main claims are that this is the first z=0 measurement at Lyman-limit columns, that f(N_HI) has declined modestly since z~3 at 19.2<log N_HI<21 while showing weaker evolution at 17.8–19.2, and that TNG50 overpredicts extended low-column HI around Milky-Way-like galaxies.","tokens_in":39272,"tokens_out":6089,"duration_ms":56562,"significance":"If the central measurement survives the completeness and representativeness concerns below, this would be a genuinely new and valuable anchor: the first z=0 f(N_HI) extending to Lyman-limit columns, with roughly 100 times better sensitivity than the earlier Z05 measurement at high column densities. The paper is unusually careful in several respects: it quantifies distance uncertainties, galaxy stochastic sampling, inclination bootstrapping, HIMF systematics, resolution dependence, self-absorption, and it validates the new full-HI combination procedure with mock tests. The impact-parameter constraints and the TNG50 comparison are novel and provide testable predictions for simulations and for absorber–galaxy association studies. However, the absolute normalization and the headline low-column-density claims currently rest on an uncorrected completeness effect at the quoted 17.8 limit and on an unquantified selection bias toward HI-rich systems, so the significance is real but not yet established at the claimed precision.","major_comments":[{"comment":"Equation (1) sums A_i(log N_HI) over galaxies with no completeness term, yet §4.1 states that the median and 90th-percentile N_HI,lim are 10^17.8 and 10^18.0 cm^-2. At log N_HI = 17.75–17.95, roughly half (and at 17.95 still a substantial fraction) of the sample contributes zero area by construction because N_HI,lim exceeds the bin. The Table 3 entries at 17.75–17.95, the direct length incidence ℓ(10^17.8)=0.267±0.033, and the weak-evolution statement at 17.8<log N_HI<19.2 are therefore lower limits, not complete measurements. If non-detected galaxies have similar low-column areas to detected ones, log f(N_HI) would be understated by up to ~0.3 dex, far above the quoted 0.04–0.06 dex errors. The statement in §5 that the points below the dashed line are 'less reliable' is not a substitute for a completeness correction. The authors should either add an explicit completeness weighting as a","section":"§4.1–§4.2, Table 3"},{"comment":"The HIMF-weighted construction assumes that the measured A_i(M_HI) are representative of all galaxies at fixed HI mass. Both surveys are explicitly biased toward HI-rich, star-forming galaxies: FEASTS selects f_HI>50 Jy km/s and MHONGOOSE targets star-forming systems. The error budget in §4.3 quantifies stochastic sampling among the observed galaxies but not selection against HI-poor galaxies at fixed M_HI. If HI-rich galaxies have more extended low-column HI, the absolute f(N_HI), the inferred evolution relative to z~3, and the impact-parameter distributions are all biased in a direction not captured by the quoted errors, and the shift could exceed the 0.075 dex HIMF systematic. The paper acknowledges the bias in §2.4 and gives an upper-limit interpretation in §4.4.1, but the central claim of a cosmic z=0 f(N_HI) requires either a quantitative correction or a clear statement that the re","section":"§2.4, §4.4.1, §6.3.3"},{"comment":"The quoted covering fraction of ~0.7 and the inference f_cloud = 0.675±0.096 rest on integrating the best-fit Schechter function below the reliable limit to log N_HI = 17.5. The paper's own Table 3 shows a flattening below log N_HI = 17.8, and while §6.2.1 considers a forced-flattening test, the quoted uncertainty on f_cloud does not propagate the extrapolation uncertainty. Because the abstract lifts out 'a covering fraction of ~0.7', the extrapolation-dependent nature of this number should be quantified in the error budget or the claim should be reframed as an extrapolation-based estimate rather than a direct measurement.","section":"§6.2.1, §6.2"}],"minor_comments":[{"comment":"The main sample is defined in §2.4 as 40 FEASTS + 30 MHONGOOSE = 70 galaxies, but the Summary in §7 says 'a sample of 65 galaxies'. Please reconcile.","section":"§2.4 vs §7"},{"comment":"The text in §2.2.5 says the final FEASTS sample has 40 galaxies, while the Appendix text introducing Table 2 says 'the finally analyzed FEASTS sample of 35 galaxies'. The table caption also states 40 galaxies. This inconsistency should be fixed.","section":"Appendix A, Table 2"},{"comment":"The table's first bin is centered at log N_HI = 17.75 rather than at 17.8, and the abstract states the limit as 10^17.8. Please clarify whether the quoted limit is the first bin center, the bin edge, or the median N_HI,lim, and define the binning consistently.","section":"Table 3"}],"recommendation":"major_revision","confidential_remarks":"The completeness issue at log N_HI ≈ 17.8 is the main reason for major revision; it is concrete and fixable within the manuscript's scope, either by adding a completeness weight or by moving the reliable limit to log N_HI = 18.0 and adjusting the claims. The representativeness bias is more difficult but at least needs an explicit quantitative bound. I do not see grounds for rejection: the method is sound, the systematics work is careful, and the paper's caveats are mostly explicit. The main concern is that the headline 'first measurement down to 17.8' is currently stronger than the data support."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis is the first z=0 HI column density distribution function reaching log N_HI=17.8, built from FEASTS and MHONGOOSE imaging at ~1 kpc resolution. That alone fills a real gap: prior 21-cm work stopped at DLA-like columns (log N_HI>19.8). The paper also delivers length incidences, impact-parameter distributions, and a TNG50 comparison, all areas where the local universe was missing a direct anchor. The systematic-error accounting is unusually thorough — distance errors, galaxy bootstrap, inclination sampling, HIMF choice, resolution tests, and mock-based tests of the FAST+interferometric combination. Given the data in hand, the measurement is credible and the paper is a serious piece of work.\n\nThe soft spots are real, though. The biggest is the incompleteness of the faintest bin. The median N_HI,lim is 10^17.8, so at that column roughly half the sample contributes zero area. Equation (1) has no completeness term, and no correction is described in §4.2–4.3. The point at 17.8 is therefore a lower limit, not a complete cosmic CDDF. The quoted length incidence ℓ(10^17.8)=0.267±0.033, the f_cov~0.7, and the claimed weak evolution between 17.8<log N_HI<19.2 all rest on that bin. If undetected galaxies have low-column areas comparable to detected ones, log f could be understated by up to ~0.3 dex — larger than the quoted 0.04–0.06 dex errors. The paper acknowledges the 50% completeness in §4.1 but doesn't propagate it.\n\nSecond, the sample is deliberately HI-rich and star-forming, and the paper says this directly in §2.4. But the central f(N_HI) is a volume-weighted measurement assuming within-bin representativeness at fixed M_HI. If HI-rich galaxies have more extended low-column gas, the faint end and all the impact-parameter statistics are biased. The FEASTS-vs-MHONGOOSE comparison in §4.4.1 suggests 0.1–0.2 dex systematics, but the authors treat that as an upper limit rather than an error term.\n\nThird, the abstract's 'covering fraction of ~0.7' overstates what is actually one of three possibilities in §6.2.1, and it's an f_cloud inferred from the length-incidence extrapolation below the reliable limit. The low-N_HI length incidence depends on Schechter-function extrapolation to 17.5. There are also minor internal inconsistencies in sample size (70 vs 65 galaxies; 40 vs 35 in the FEASTS table).\n\nNone of this kills the paper. The measurement is plausible, the caveats are mostly discussed, and the central result — a first local CDDF in the LLS regime — will be used by the field. But the faint-end claims need reframing as lower limits or a completeness correction, and the sample-bias caveat should be folded into the error budget. I'd send it to a serious referee and require major revision before publication. It deserves that referee time.","headline":"First z=0 HI CDDF to LLS columns: valuable, but the faintest bin is incomplete and the HI-rich sample bias isn't corrected; needs revision before publication.","tokens_in":39882,"tokens_out":3543,"would_cite":true,"duration_ms":31587,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The first z=0 measurement of the HI column density distribution function reaches log N_HI = 17.8 cm^-2, showing weak evolution in faint gas and a decline in denser gas since z~3.","keywords":["galaxy evolution","neutral hydrogen","HI column density distribution","Lyman limit systems","21-cm imaging","circumgalactic medium","HI mass function","cosmic gas"],"falsifier":"A concrete test: measure the extent of low-column HI around galaxies selected without an HI-rich bias and recompute f(N_HI) with those radial profiles at the same 17.8 cm^-2 depth and roughly 1 kpc resolution; if the faint end at 17.8-19.2 changes by more than about 0.075 dex (the HIMF systematic), the claimed z=0 distribution and weak-evolution conclusion are refuted.","tokens_in":38756,"feed_emoji":"🌌","tokens_out":8627,"duration_ms":76616,"temperature":0.7,"pith_summary":"The paper presents the first measurement at z=0 of the HI column density distribution function f(N_HI) down to log N_HI = 17.8 cm^-2, the Lyman-limit regime, using 21-cm emission maps at roughly 1 kpc resolution. It claims that this distribution is lower than at z~3 by 0.1-0.4 dex for log N_HI between 19.2 and 21, but comparable at 17.8-19.2, implying only weak evolution in the faintest gas. It also claims that gas at these densities lies closer to galaxies than absorption-line surveys suggest, and that a leading cosmological simulation over-predicts distant low-column gas around Milky-Way-like galaxies. These results matter because they supply the first local anchor for models of gas accretion, outflows, and galaxy evolution.","feed_headline":"First z=0 census of faint hydrogen gas reaches Lyman-limit densities","feed_subtitle":"21-cm maps show how much sky is covered by atomic gas down to Lyman-limit densities and how it changed since redshift 3","key_machinery":"The machinery is the construction of full-HI images: sensitive single-dish 21-cm data supply the large-scale diffuse gas, while interferometric data supply the small-scale structure; the two are combined via feathering with a simulated single-dish beam, then corrected for beam variation and cleaned. The column density distribution f(N_HI) is built by counting area in 0.1-dex bins from these ~1-kpc-resolution maps, weighting each of the 70 galaxies by the HI mass function at its HI mass and by the inverse sampling density in its mass bin. A Schechter function fit (power law plus exponential cutoff) characterizes the result and is integrated to obtain incidence and covering fractions.","core_discovery":"The central discovery is a measured f(N_HI) — the sky area per logarithmic column-density bin — at z=0, extending two orders of magnitude deeper in column density than earlier 21-cm studies. The shape is well fit by a Schechter function with log N*_HI = 21.26 and slope beta = 1.15, and it matches earlier DLA-regime measurements while extending sensitivity by about 100 times. The paper then uses this function to derive a covering fraction f_cov ~0.7 for Lyman-limit gas when averaged over 1-kpc pixels, and a covering fraction ~0.006 within the virial radius of Milky-Way-like galaxies. The key comparative claims are a 0.1-0.4 dex decline at 19.2 < log N_HI < 21 since z~3, and much smaller impac","pith_inferences":["Beyond the paper: because the largest systematic is galaxy sampling within HI mass bins, a larger, less HI-biased sample would directly test whether the faint end of f(N_HI) and the claimed weak evolution survive.","Beyond the paper: the b/r_001-N_HI relation could be inverted to estimate HI masses of Lyman-limit-system host galaxies from impact parameter and column density alone, converting absorption surveys into mass measurements.","Beyond the paper: the discrepancy between 21-cm and absorption impact parameters predicts that future sensitive surveys will find a population of low-redshift Lyman-limit systems with no bright nearby galaxy, tracing intergalactic or group gas rather than individual halos."],"forward_implications":["The z=0 f(N_HI) gives simulations and analytic models a direct local data point for Lyman-limit gas, replacing extrapolations from higher redshift.","Weak evolution at 17.8-19.2 means the faint atomic gas reservoir around galaxies has been roughly stable since z~3, while the 0.1-0.4 dex decline at higher column densities indicates the denser HI reservoir has thinned.","The low impact parameters at fixed N_HI imply that low-redshift absorber surveys are missing many of the actual galaxy hosts; the gas may be found much closer to galaxies than previously thought.","The over-prediction of distant low-column HI in cosmological simulations points to a need to revise how cool gas or HI post-processing is handled.","The covering-fraction result (0.006 for Milky-Way-like galaxies) quantifies how little Lyman-limit gas typical halos contain, informing searches for gas accretion."],"fun_headline_variants":["Deep 21-cm census maps atomic gas down to Lyman-limit densities","Atomic gas covers 70% of sky at 1-kpc scale, new survey finds","HI distribution at z=0 shows weak evolution since z~3 for faint gas","100x more sensitive 21-cm survey probes faintest atomic gas yet"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The whole census rests on assuming that the galaxies in the two surveys — both biased toward HI-rich, star-forming systems — represent all galaxies of the same HI mass; if HI-poor galaxies at fixed M_HI have less extended faint gas, the faint end of f(N_HI), the evolution offsets, and the impact-parameter distributions all shift.","fun_headline_variants_meta":{"raw":{"variants":["Deep 21-cm census maps atomic gas down to Lyman-limit densities","Atomic gas covers 70% of sky at 1-kpc scale, new survey finds","HI distribution at z=0 shows weak evolution since z~3 for faint gas","100x more sensitive 21-cm survey probes faintest atomic gas yet"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000427,"raw_usage":{"total_tokens":2159,"prompt_tokens":1020,"completion_tokens":1139,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":764,"completion_tokens_details":{"reasoning_tokens":1051}},"tokens_in":764,"tokens_out":1139,"duration_ms":10202,"temperature":1.0,"reasoning_tokens":1051,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T19:17:37.592080+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete test: measure the extent of low-column HI around galaxies selected without an HI-rich bias and recompute f(N_HI) with those radial profiles at the same 17.8 cm^-2 depth and roughly 1 kpc resolution; if the faint end at 17.8-19.2 changes by more than about 0.075 dex (the HIMF systematic), the claimed z=0 distribution and weak-evolution conclusion are refuted.","supporting_citations":[],"review_version":1}