{"id":"239a72a3-0b78-4f26-88ec-2c2f0b0fb475","arxiv_id":"1908.02291","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Merging galaxies show a roughly four-times higher detection rate of neutral hydrogen absorption toward their centers than non-mergers, with absorption seen in 84% of the sample and in 100% of post-merger systems.","lead":"By studying 38 merging galaxies with radio telescopes, astronomers find that about 84% show cold hydrogen gas absorbing radio waves near their centers, roughly four times the rate in non-mergers. The result supports the idea that mergers push gas inward, and suggests this gas survives even after the galaxies fully combine.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Post-merger 100% detection claim rests on 13 objects and lacks a non-merger control matched for radio power and redshift; the main merger-versus-non-merger contrast is robust, but the evolutionary interpretation is not.","rationale":"The reader's weakest assumption focuses on the 7 missing mergers out of 45. That is a legitimate completeness concern, but it is not the most load-bearing: even in the extreme case where all 7 missing sources are non-detections, the detection rate falls from 84% (32/38) to 71% (32/45), still a factor of ~3 above the 23% non-merger rate. The reader also flags the post-merger subsample as small, and the paper itself acknowledges the stage-to-stage differences are not statistically significant. However, the reader does not identify the lack of a radio-power-matched control for the evolutionary claim as a distinct load-bearing weakness. The paper's multi-wavelength and correlation analyses are appropriately cautious. Therefore I agree with the conditional verdict but differ in emphasis: the core merger-versus-non-merger claim is solid, while the post-merger survival narrative is an over-interpretation of a small, possibly selection-biased subsample. A concrete test involving a matched control and a completeness check would settle whether that narrative survives.","tokens_in":18767,"tokens_out":1763,"duration_ms":16573,"concrete_test":"Construct a mass- and redshift-matched non-merger control from Maccagni et al. (2017) restricted to sources with P1.4 comparable to the merger sample, and recompute the incidence of HI absorption in three merger stages versus that control. If the post-merger rate remains 100% while the matched control rate is ~23%, the survival claim gains support; if the control rate rises to ~80% when matched in radio power, the evolutionary claim would be weakened. Additionally, check the 7 missing mergers by comparing their z, P1.4, and M* distributions to the 38 studied; if the missing systems are biased toward higher P1.4 or red mergers, redo the detection rate including them as non-detections.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central statistical claim, an 84 +/- 15% detection rate (32/38) versus 23 +/- 3% for the Maccagni et al. (2017) non-merger sample, is supported by the data and by sensitivity and mass-matching checks. The weaker link is the paper's headline interpretation that HI survives the coalescence period, based on a 100% detection rate in the 13 post-mergers. This claim has no statistically significant contrast: the paper itself states that differences between merger stages are not significant in two-sided KS tests. The 100% post-merger rate (13/13) is formally compatible with the 76 +/- 14% ongoing-merger rate (using the quoted 12/12? No: Table A1 shows actual counts are uncertain; the text quotes 13, 12, 13 per stage, with 100% detection in post-mergers). The comparison also lacks the non-merger control at matched radio power and redshift: the evolution of detection rate from pre- to post-merger is compared across samples that differ in redshift and P1.4, and the non-merger sample of Maccagni et al. (2017) is not matched in radio power. Since the sample is selected for S1.4 > 20 mJy, and the non-merger comparison sample may be drawn from a heterogeneous set, the interpretation that the merger stage progression drives the detection rate rather than selection or radio-spectral properties is not secured. Missing objects: 7 of 45 radio-loud mergers without HI measurements are assumed not to preferentially lack HI; the paper does not compare SDSS/FIRST properties of the 38 with the 7, so a bias in the detection rate cannot be excluded. The main claim is robust against this because even 32/45 = 71% remains ~3x the non-merger rate, but the post-merger interpretation would be more fragile.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents H I 21-cm absorption observations of nine galaxy mergers obtained with the GMRT and combines them with literature data to build a sample of 38 radio-loud mergers at z <= 0.2. The authors report an H I absorption detection rate of 84 +/- 15% (32/38), about four times the 23 +/- 3% rate found in a 229-object non-merger sample from Maccagni et al. (2017). Two-sided KS tests show significantly different distributions of H I column density (P = 2e-8) and velocity shift (P = 3e-3) between mergers and non-mergers, and these differences persist in a stellar-mass/redshift-matched subsample. The paper also examines correlations with multi-wavelength properties and merger stage, finding tentative trends and a 100% detection rate among 13 post-mergers, which it interprets as evidence that neutral gas survives the coalescence period.","tokens_in":19267,"tokens_out":6962,"duration_ms":67686,"significance":"If the central result holds, the paper makes a strong case that the merger process itself, rather than galaxy stellar mass or AGN radio power, drives elevated cold neutral gas in the nuclear regions of radio-loud galaxies. The statistical treatment is careful in several respects: KS tests are applied to full and matched samples, upper limits on N(H I) are treated as censored data via survival analysis, and the incidence comparison is repeated at different N(H I) sensitivity limits. The new GMRT observations are presented in detail in the appendix. The main caveats concern the completeness of the parent sample and the strength of the evolutionary interpretation drawn from the post-merger subsample.","major_comments":[{"comment":"The statistical sample comprises 38 of the 45 radio-loud mergers; the 7 without H I measurements are excluded without discussion. If all 7 were non-detections, the incidence would be 32/45 = 71%, still well above the non-merger rate but lower than the quoted 84%, and the '~4 times' excess would become '~3 times'. Please either provide the H I limits for these 7 systems or compare their z, P1.4, M*, and merger-stage distributions with the 38 included systems, and state the resulting range of detection rates.","section":"Section 2.1, Table A1"},{"comment":"The quoted factor-of-four excess relative to non-mergers rests on the comparability of the Maccagni et al. (2017) reference sample, but the paper does not demonstrate that this sample satisfies the same selection (for example, S1.4 > 20 mJy and z <= 0.2) or that its P1.4 and redshift distributions match the merger sample. The later mass/redshift matching is performed for only 24 mergers and only for the distribution comparison, not for the incidence comparison. Please report the radio-power and redshift distributions of the full non-merger sample and repeat the incidence comparison for a P1.4- and z-matched control, or state explicitly why this is unnecessary.","section":"Section 3.1, Table 1"},{"comment":"The claim that the 100% detection rate in post-mergers indicates that neutral gas survives the coalescence period and is not yet quenched is not statistically supported as stated. The paper itself notes that differences among merger stages are not significant in KS tests; moreover, the 13 post-mergers lack a comparison control matched in P1.4 and redshift, and the sample is selected on S1.4 > 20 mJy. A Fisher exact test or bootstrap comparison of 13/13 with the ongoing/pre-merger rates and with a matched non-merger control should be reported. Until then, the abstract and conclusions should present the post-merger survival as a tentative interpretation rather than a firm result.","section":"Section 3.3, Fig. 4, Abstract"}],"minor_comments":[{"comment":"The word 'untill' in the first paragraph should be 'until'.","section":"Section 1"},{"comment":"The uncertainty quoted on the 84% detection rate (15%) is not defined; please specify whether it is a binomial or Wilson confidence interval.","section":"Section 3.1"},{"comment":"The median N(H I) and FWHM values for each merger stage are shown in Fig. 4 but are not given numerically in the text; adding them to the text or a small table would improve reproducibility.","section":"Section 3.3, Fig. 4"},{"comment":"The spectral-index column header uses alpha_{0.15}^{1.4}, but the definition is given only in Section 2.1; please restate it in the table note for clarity.","section":"Table A1"},{"comment":"Several LaTeX artifacts such as 'greaterorequalslant' appear in the text; these should be rendered as proper symbols in the final version.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The central statistical result is likely robust and the new data are valuable, but the evolutionary conclusion rests on a small subsample and the missing 7/45 systems need a sensitivity analysis. Both concerns are addressable within the scope of a revision, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe central result is probably right: radio-loud mergers at z<0.2 show HI 21-cm absorption about four times more often than non-mergers, and the statistical machinery behind that claim is careful. The weaker part is the evolutionary narrative—the 100% detection in post-mergers rests on 13 objects, and the authors themselves find no significant difference between merger stages.\n\nWhat's new: nine new GMRT observations, a sample expanded from 10 to 38 mergers, and a systematic look at how detection rate, N(HI), and kinematics depend on separation, stellar mass, redshift, spectral index, and merger stage. The paper does this properly. It uses KS tests with sensitivity cuts, builds a mass- and redshift-matched non-merger sample, treats upper limits with survival analysis, and publishes full tables of properties and spectra. That is reproducible and honest.\n\nSoft spots: first, 7 of 45 radio-loud mergers have no HI measurement, and the paper never compares their properties to the 38 that do. If most of those are non-detections, the 84% drops to 71%—still three times the non-merger rate, so the main claim survives, but the paper should say so. The post-merger interpretation is shakier. It is based on 13/13 detections, the KS tests between stages are not significant, and there is no non-merger control matched in radio power and redshift for that specific comparison. The paper actually admits the stage differences are not statistically significant, which makes the abstract's claim that 100% detection 'indicates' survival past coalescence too strong. The sample also mixes new data with heterogeneous literature measurements, though the matching checks mitigate that.\n\nOverall: this is a solid, incremental observational study. It extends the pilot sample, adds new data, and gives a fair and fairly robust statistical treatment. The main claim will hold; the evolutionary interpretation needs qualification.\n\nSend it to peer review. A good referee should ask for a completeness check on the missing 7 systems and for a softened abstract, but the core result deserves publication in a specialist journal.","headline":"Solid confirmation that radio-loud mergers show ~4x higher HI 21-cm absorption than non-mergers, but the post-merger gas-survival claim is overinterpreted relative to the statistics.","tokens_in":19730,"tokens_out":2684,"would_cite":true,"duration_ms":28277,"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":"Galaxy mergers with radio-loud cores show neutral hydrogen absorption in about 84% of cases, roughly four times the rate in non-mergers.","keywords":["HI 21-cm absorption","galaxy mergers","radio-loud AGN","neutral hydrogen","circumnuclear gas","post-merger galaxies","AGN feedback","multi-wavelength galaxy properties"],"falsifier":"HI 21-cm observations of the seven unobserved radio-loud mergers in the parent sample at comparable sensitivity: if most of those seven turn out to be non-detections, the quoted 84% detection rate falls (for example, if four of seven are non-detections, the rate drops to roughly 71%) and the contrast with the 23% non-merger rate weakens.","tokens_in":18600,"feed_emoji":"📡","tokens_out":8052,"duration_ms":86294,"temperature":0.7,"pith_summary":"Using 38 radio-loud galaxy mergers at redshifts up to 0.2, the paper establishes that neutral atomic hydrogen is almost always present in the circumnuclear regions of merging galaxies: 84 ± 15% of the mergers show HI 21-cm absorption, compared with 23 ± 3% of non-merger radio galaxies. The column densities and velocity shifts of the absorbing gas differ sharply between the two populations, with median HI column density about five times higher and roughly three times more infalling (redshifted) absorption in mergers. These differences survive when the samples are matched in stellar mass and redshift, pointing to the merger process itself as the driver. The paper also finds that the incidence, column density, and line width rise from pre-merger through post-merger stages, with all 13 post-mergers detected, and argues that neutral gas therefore survives the coalescence period rather than being immediately destroyed by nuclear radio activity.","feed_headline":"Merging galaxies harbor neutral gas at ~4x the non-merger rate","feed_subtitle":"In 38 radio mergers, 84% show atomic hydrogen absorption versus 23% of non-mergers, with 100% after coalescence.","key_machinery":"The load-bearing probe is the HI 21-cm absorption line: the spin-flip transition of neutral hydrogen seen in absorption against a compact background radio continuum source, which traces cold gas along the line of sight through the central kiloparsecs. The argument compares this absorption in 38 radio-loud mergers with 229 non-merger radio galaxies, using two-sided Kolmogorov-Smirnov tests on column density, line width, and velocity shift, and a stellar-mass/redshift-matched control to rule out galaxy-mass effects. Conversion of optical depth to column density assumes a spin temperature of 100 K and unity covering factor. Merger stages are assigned with a six-stage classification scheme compressed into pre-merger, ongoing, and post-merger phases, and the evolution along that sequence is the third quantitative axis of the paper.","core_discovery":"The central discovery is that the nuclear neutral hydrogen content of radio-loud galaxies depends strongly on merger state. Among 38 mergers, 32 are detected in HI 21-cm absorption, a detection rate of 84 ± 15%, against 23 ± 3% in a reference sample of 229 non-mergers. A two-sided Kolmogorov-Smirnov test gives $P_{\\mathrm{KS}} = 2\\times10^{-8}$ for the column-density distributions and $P_{\\mathrm{KS}} = 3\\times10^{-3}$ for the absorption velocity shifts; the median $N(\\mathrm{H\\,I})$ is roughly five times higher in mergers, and the fraction of components with $v_{\\mathrm{shift}} \\ge 100$ km s$^{-1}$ is 30 ± 7%, versus 9 ± 4% in non-mergers. Matching the samples in stellar mass and redshift leaves the differences intact. The detection rate climbs from pre-merger through ongoing to post-merger stages and reaches 100% in post-mergers, which the paper reads as evidence that circumnuclear neutral gas persists through coalescence and is not yet quenched by the radio-loud nucleus.","pith_inferences":["The tentative anti-correlation of $N(\\mathrm{H\\,I})$ with stellar mass likely tracks gas-rich blue mergers; a sample extending to fainter radio powers and higher redshifts could test whether the mass trend is purely a selection effect.","If the post-merger gas reservoir survives coalescence, the next evolutionary step should be a drop in HI absorption once quasar-driven winds clear the nucleus; searching for HI in radio-loud post-mergers with quasar-like optical spectra would test this sequence.","The fixed spin-temperature and covering-factor values ($T_s=100$ K, $C_f=1$) mean the reported column densities are lower limits if the gas is warm or patchy; resolved 21-cm excitation measurements or comparison with other cold-gas tracers would calibrate the actual column."],"forward_implications":["The mass-matched comparison implies that the high column densities and infall signatures are caused by the merging process itself, not by the stellar mass of the galaxies involved.","Because 100% of the post-mergers show absorption, the neutral gas reservoir that feeds or accompanies nuclear activity survives for roughly a gigayear after coalescence; radio-loud feedback has not yet expelled it.","The incidence of HI absorption rises from pre-merger to post-merger stages, implying that tidal interactions progressively concentrate neutral gas into the nuclear regions.","The absence of strong correlations between HI gas properties and infrared luminosity or star formation rate means that current starburst intensity does not determine how much neutral gas sits at the nucleus.","The higher fraction of redshifted absorption lines in mergers indicates that a substantial part of the nuclear neutral gas is infalling rather than outflowing, which is relevant to black hole feeding."],"supporting_citations":[{"why":"Supplies the pilot sample of ten mergers and the observational methods and literature compilation that define the merger sample.","marker":"Dutta et al. (2018)"},{"why":"Supplies the 229 non-merger radio galaxies and their 23 ± 3% detection rate against which the merger excess is measured.","marker":"Maccagni et al. (2017)"},{"why":"Provides the comparison result of elevated HI emission and atomic gas fraction in post-mergers that supports the survival interpretation.","marker":"Ellison et al. (2018)"},{"why":"Reports 100% HI absorption incidence in coalesced ULIRGs, used as corroboration for the post-merger detection rate.","marker":"Teng et al. (2013)"},{"why":"Defines the six-stage merger classification that the paper compresses into pre-merger, ongoing, and post-merger stages.","marker":"Haan et al. (2011)"},{"why":"Supplies the LIRG/ULIRG luminosity classes and infrared spectral energy distribution method used to characterise the mergers.","marker":"Sanders & Mirabel (1996)"},{"why":"Provides the kcorrect algorithm used to derive stellar masses for the mass-matched control comparison.","marker":"Blanton & Roweis (2007)"},{"why":"Supplies the low-frequency to 1.4 GHz spectral indices used to test whether radio spectral shape tracks HI incidence.","marker":"de Gasperin et al. (2018)"}],"fun_headline_variants":["Merging galaxies show 84% HI absorption rate, 4x non-mergers","Neutral gas persists in 100% of post-merger radio galaxies","HI absorption detection quadruples in merging galaxy centres","Post-merger galaxies retain neutral gas at 100% detection"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim rests on the assumption that the 38 mergers with HI measurements fairly represent all 45 radio-loud mergers in the parent sample; in particular, the seven systems without HI data must not be mostly non-detections.","fun_headline_variants_meta":{"raw":{"variants":["Merging galaxies show 84% HI absorption rate, 4x non-mergers","Neutral gas persists in 100% of post-merger radio galaxies","HI absorption detection quadruples in merging galaxy centres","Post-merger galaxies retain neutral gas at 100% detection"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000386,"raw_usage":{"total_tokens":2077,"prompt_tokens":1018,"completion_tokens":1059,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":634,"completion_tokens_details":{"reasoning_tokens":983}},"tokens_in":634,"tokens_out":1059,"duration_ms":8684,"temperature":1.0,"reasoning_tokens":983,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:48:23.803850+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"HI 21-cm observations of the seven unobserved radio-loud mergers in the parent sample at comparable sensitivity: if most of those seven turn out to be non-detections, the quoted 84% detection rate falls (for example, if four of seven are non-detections, the rate drops to roughly 71%) and the contrast with the 23% non-merger rate weakens.","supporting_citations":[{"cited_title":"H., Veilleux S., Baker A","cited_arxiv_id":null,"evidence_quote":"Reports 100% HI absorption incidence in coalesced ULIRGs, used as corroboration for the post-merger detection rate."}],"review_version":1}