{"id":"f6fffe96-3579-4dee-96ab-c4213eb4c3f9","arxiv_id":"2607.25800","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Stacked FAST spectra show cluster galaxies have less neutral hydrogen than field galaxies out to 5R200, including blue star-forming members.","lead":"Using stacked 21-cm spectra from FAST, this study measures hydrogen gas in galaxies around 36 galaxy clusters out to five times the virial radius, finding gas levels already depressed far outside the clusters. The result suggests galaxies lose their cold gas fuel before they ever reach the dense cluster core.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 5R200 field offset rests on an unmatched external baseline; a matched DESI+FASHI field control is needed before claiming HI depletion beyond the virial radius.","rationale":"I read the paper in good faith and take the central claim to be that HI depletion is already present at 5R200 relative to field galaxies, which the authors use as direct evidence for pre-processing in cluster outskirts. The internal radial trends are built from a consistent sample and pipeline, so they are not the main risk. The vulnerable step is the external comparison: the field baseline is taken from a different survey and, in the color-resolved analysis, from an HI-selected sample. Because the cluster stacks include non-detections and apply a blending correction that the field sample does not receive, the offset at large radius may reflect selection rather than environment. This concern is concrete and testable. The reader identified the phase-space membership boundary as the weakest assumption; that is certainly a real uncertainty, but it is less fatal to the central claim because interloper contamination would generally dilute an observed deficiency, not create it. A more directly load-bearing issue is the unmatched field baseline. My recommendation of UNCHANGED reflects that the paper is already CONDITIONAL in the reader's verdict and this concern strengthens, rather than replaces, the conditions under which the central claim should be accepted.","tokens_in":31791,"tokens_out":3874,"duration_ms":41934,"concrete_test":"Construct a field control sample from DESI galaxies satisfying the same r<20 mag and stellar-mass cuts, located at projected distance >10R200 from any AXES cluster and not within a comparably massive group, and matched to the cluster sample in redshift and M* (and g-r when testing color). Run the identical FASHI extraction, HISS stacking, 6-arcmin aperture, and blending-combination pipeline on this control. If the 4.5–5.5R200 cluster bins remain ≥0.3 dex below the matched control at the same stellar mass and color, the pre-processing claim survives; if they become consistent, the 5R200 field offset is a selection/normalization artifact.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central new claim — that M_HI/M* remains below field values even at 4.5–5.5R200 — is supported in §4.1 and Fig. 5 by comparison with H. Guo et al. (2021), an ALFALFA stacking analysis with a different survey footprint, depth, angular resolution, and target-selection function. This is not a matched control. The color-resolved claim in §4.2 and Fig. 7 is even more exposed: cluster members are mass-selected from DESI and include non-detections, while the field reference is the FASHI HI-detected sample, which is by construction gas-rich. The blending correction (§3.2) further rescales cluster averages by factors that are not applied to the field baseline, and the quoted 5R200 offset is only ~1σ. If the field reference is biased high relative to a comparably selected, comparably processed control, the apparent 'pre-processing at 5R200' could be a survey-selection or normalization artifact. The internal radial decline (Figs. 5, 10) is credible and would not be erased, but it does not by itself establish 'lower than field at 5R200' — which is exactly the headline claim. The load-bearing weak point is therefore not the membership boundary (real as that is) but the lack of a matched field control applied through the same stacking and blending pipeline.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper stacks FAST/FASHI 21-cm spectra at the positions of DESI/SGA spectroscopic member galaxies of 36 AXES X-ray clusters at z<0.07, measuring average M_HI and M_HI/M* in three stellar-mass bins and six projected radial bins out to 5R200. The authors report a monotonic decline of M_HI and M_HI/M* toward cluster centers, interpret the turnover near ~2R200 as the splashback/shock radius, and claim that M_HI/M* remains below the field value even at 4.5–5.5R200, which they attribute to pre-processing in filaments and infalling groups. They further report that the HI deficiency persists at fixed g−r color, that relaxed and disturbed clusters show no significant difference, and that integrated HI masses are broadly consistent with TNG simulations. The internal radial trend is based on a large, uniformly selected sample with conservative resampling-based errors; the headline claim about depletion at 5R200, however, rests on comparisons with external or HI-selected field baselines that are not matched to the cluster sample.","tokens_in":32138,"tokens_out":6360,"duration_ms":67527,"significance":"If firmly established, the claim that cold-gas depletion begins well beyond the virial radius would provide a direct gas-phase counterpart to earlier star-formation suppression observed at 2–5R200 and would strengthen the pre-processing scenario in cluster outskirts. The paper's strengths include the large stacking sample (~170 to >1000 spectra per bin), the use of conservative resampling uncertainties, the public HISS stacking code, and unusually candid discussion of beam-blending, incompleteness, and the approximate nature of the membership boundary. However, the central claim is not yet supported at the quoted significance: the 5R200 offset is described in the text as 1σ, and the field baselines used in Figs. 5 and 7 are not selected or processed in the same way as the cluster sample. These issues are load-bearing and need to be addressed before the paper's main conclusion can be accepted.","major_comments":[{"comment":"The text states that at large cluster-centric distances the average M_HI/M* 'remains 1σ lower' than the Guo et al. (2021) field relation. A 1σ offset is not a statistically significant detection, yet the abstract and §6 conclude that M_HI/M* 'remain lower than field galaxies even at the 5R200' and that this 'provides direct, statistical evidence.' Please report the actual offset, its 1σ uncertainty, and the significance level for the 4.5–5.5R200 bin. If the offset is indeed only 1σ, the headline claim should be softened accordingly.","section":"§4.1, Fig. 5"},{"comment":"The field reference from H. Guo et al. (2021) is an ALFALFA stacking analysis with a different survey footprint, depth, angular resolution, target selection, and spectral extraction pipeline, while the cluster sample is mass-selected from DESI and includes non-detections. The apparent deficit at 5R200 is therefore not a matched comparison. A matched field control drawn from the same DESI/FASHI parent sample, processed through the same extraction, stacking, and blending pipeline, is needed before claiming depletion beyond the virial radius. The internal radial decline would not be erased by such a control, but the normalization offset could change.","section":"§4.1, Fig. 5"},{"comment":"The color-resolved comparison uses FASHI-detected field galaxies, which are by construction HI-selected and therefore gas-rich, while the cluster members are mass-selected and include non-detections. This selection asymmetry alone can produce an apparent deficiency. The statement that this bias is 'modest in the g−r plane' is not quantified, and no matching in stellar mass or redshift is shown. The 0.3–1 dex offsets in Fig. 7 cannot be interpreted as environmental HI deficiency until a mass- and color-matched field sample including non-detections is stacked with the same pipeline.","section":"§4.2, Fig. 7"},{"comment":"The blending correction rescales cluster M_HI by a factor N_stacking/N_total, typically about 0.7 (≈0.15 dex), and up to 0.2 dex for the low-mass bin. The Guo et al. (2021) field baseline has not been subjected to this correction. Because the field reference is not processed through the same blending treatment, a systematic offset of order 0.15–0.2 dex is introduced into the cluster/field comparison. This is comparable to the quoted 1σ offset at 5R200 and must be propagated or removed by applying an identical correction to a matched field sample.","section":"§3.2"},{"comment":"The phase-space membership boundary uses an NFW escape-velocity profile with f_los=0.7 and the Evrard et al. (2008) velocity-dispersion scaling. The paper itself cautions that at R>2R200 the uncertainties grow substantially due to breakdown of spherical symmetry, interlopers, and the assumed mass profile. Since the headline claim concerns the 4.5–5.5R200 bin, the sensitivity of the stacked M_HI to reasonable changes in f_los (e.g., 0.6–0.8), to the concentration–mass relation, and to interloper rejection should be quantified. Without such robustness tests, it is unclear whether the outermost bin mean reflects true members or a mixture whose selection could be HI-correlated.","section":"§2.3, Eq. (2)"}],"minor_comments":[{"comment":"FASHI coverage is quoted as ~4000 deg² in the introduction and ~19,500 deg² in §2.1.3. Please clarify that the smaller number refers to DR1 and the larger to current/DR2 coverage.","section":"Introduction/§2.1.3"},{"comment":"The caption says 'There are systematically less targets in the 5R/R200 bins'; change 'less' to 'fewer' and clarify that the outermost bin corresponds to 4.5–5.5R200.","section":"Fig. 5 caption"},{"comment":"The section title appears as '–Halo Relation Across a Broad Halo Mass Range'; the omitted 'M_HI' should be restored.","section":"§5.4"},{"comment":"The notation 'M_avg code HI' and 'M_avg corr HI' is hard to parse; define the superscripts/subscripts more clearly and use a consistent typographic convention.","section":"§3.2"},{"comment":"The Stage column contains '—' for five clusters; the text explains the reason, but adding a footnote to the table would improve readability.","section":"Table 1"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses an important question and the internal radial trend appears credible. The main obstacle is the unmatched field baseline: the authors need to produce a matched field control processed through the same stacking and blending pipeline, and to report the actual significance of the 5R200 offset. If the offset disappears under a matched control, the paper's scope should be revised to the radial trend and the simulation comparison, which remain valuable. The color-resolved comparison also needs a non-HI-selected field sample. I see these as feasible within the manuscript's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: the internal radial trend — roughly 0.5 dex decline in stacked HI from 5R200 to the core, with large bin populations — is credible and worth having. The specific claim that M_HI/M* is already below field level at 5R200 rests on a comparison to Guo et al. (2021) ALFALFA stacking that is not matched in selection, depth, resolution, or processing. The stress-test note is right: this is the load-bearing weak point, not the membership boundary. The paper is honest about many systematics, but honesty does not make an unmatched baseline matched.\n\nWhat is new: 36 massive X-ray clusters out to 5R200, roughly an order of magnitude more galaxies per bin than W. Hu et al. (2021), and a color-resolved stacking showing a deficit at fixed g-r. That color result is suggestive, but the field reference is the HI-selected FASHI sample, which is gas-rich by construction, so the offset at fixed color is probably overestimated. The blending correction — about 30% of targets combined, worth 0.15–0.2 dex — is applied to cluster points but not to the field reference. That alone could explain a meaningful part of the reported 5R200 deficit.\n\nSoft spots, in proportion: the decline inside ~2R200 is solid. The “begins well beyond the virial radius” phrasing is stronger than the evidence. The 5R200 offset is about 1σ against a different survey. Membership selection beyond 2R200 is approximate, and the authors admit it. The low-mass incompleteness and blending can mimic the dwarf-galaxy early-onset trend; the authors actually flag this in Section 5.6, so the mass-dependent interpretation is appropriately cautious. The total HI comparison with simulations is a consistency check, not a model test, and they mostly frame it that way.\n\nWho this is for: observers working on environmental quenching and HI stacking. It advances the methodology and provides a useful cluster-member catalog from DESI plus FASHI. It deserves a serious referee, but not acceptance as is. The referee should ask for a matched field control drawn from the same FASHI cubes and the same DESI-based selection, run through the same stacking and blending pipeline, or the central claim should be rephrased to “further decline beyond the virial radius” rather than “lower than field at 5R200.”","headline":"Solid internal radial trend; the beyond-virial claim needs a matched field control before it carries.","tokens_in":32668,"tokens_out":2293,"would_cite":true,"duration_ms":24559,"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":"This paper claims that the average neutral hydrogen fraction of cluster galaxies stays below the field level out to five times the virial radius, implying gas is stripped from galaxies before they enter the cluster core.","keywords":["HI 21-cm","galaxy clusters","environmental quenching","ram-pressure stripping","pre-processing","spectral stacking","FAST","DESI"],"falsifier":"A stacking analysis that splits the 4–5R200 sample into galaxies well inside versus near the assumed escape boundary, or uses high-resolution interferometric data to confirm membership and resolve blends, would settle the claim: if the field-normalized HI deficit vanishes when interlopers are removed, the pre-processing conclusion fails.","tokens_in":31701,"feed_emoji":"📡","tokens_out":2991,"duration_ms":33961,"temperature":0.7,"pith_summary":"The paper stacks 21-cm neutral hydrogen spectra of galaxies around 36 massive clusters to measure the average HI content as a function of cluster-centric distance. It finds that both the average HI mass and the HI-to-stellar-mass ratio decline inward, and that the HI fraction at 5×R200 remains about 0.3–1 dex below field galaxies of the same color and stellar mass. The authors interpret this as statistical evidence that environmental gas removal begins in the cluster outskirts, in groups and filaments, before galaxies reach the dense core. If true, this moves the onset of quenching far earlier than the virial radius.","feed_headline":"Stacking 36 clusters shows gas loss out to 5R200","feed_subtitle":"Even blue, star-forming cluster members carry about half the neutral gas of similar field galaxies.","key_machinery":"The analysis rests on spectral stacking of HI 21-cm emission from the FAST all-sky survey, which recovers average flux from hundreds of galaxies per bin that are individually undetected. Membership is assigned using a projected phase-space boundary from an NFW escape-velocity profile scaled by flos = 0.7 and a velocity-dispersion–mass relation, and a blending correction combines galaxies within the 3-arcmin beam to avoid double-counting.","core_discovery":"Using spectral stacking of FASHI survey data with DESI spectroscopic members, the authors show that the average HI-to-stellar mass ratio of cluster members is suppressed relative to field galaxies at all projected radii out to 5R200, with a decline of roughly 0.5 dex from outskirts to center. The deficit persists when galaxies are split by g−r color, so even optically blue, star-forming cluster members have lower gas fractions than field galaxies of the same color. This is presented as direct statistical evidence that cold gas is stripped from galaxies in the cluster outskirts, likely in infalling groups and filaments, prior to full optical transformation.","pith_inferences":["If the deficit at 5R200 is real, high-resolution HI imaging of galaxies in that region should reveal truncated disks, asymmetric extraplanar gas, or tails pointing away from the cluster center even well outside the virial radius.","The mass-dependent onset of depletion (dwarf galaxies decline earlier than massive ones) could be tested by correcting for blending with interferometric observations; if the trend survives, it points to shallow-potential galaxies being stripped by gentler mechanisms in the outskirts.","The paper's own discussion implies the field-galaxy baseline itself is environment-dependent, so the 'deficit' could partially reflect that the comparison sample occupies an average cosmic density higher than true voids; quantifying this would sharpen the pre-processing claim.","A joint analysis of HI stacking with filament identification (e.g., splitting members by proximity to filament spines) would directly test whether the outskirts deficit is driven by galaxies in filaments rather than those falling in from lower-density regions."],"forward_implications":["The finding implies that HI depletion is not confined to cluster cores but begins in the large-scale feeding structures around clusters.","The observed deficit among blue, star-forming cluster members indicates that gas removal precedes optical color transformation, constraining the timescale of quenching.","The transition radius near 2R200 aligns with the extent of hot X-ray gas, suggesting a common physical boundary where infalling galaxies first encounter stripping conditions.","The total HI content in clusters and their outskirts roughly matches TNG simulations, supporting current models of gas accretion and feedback in massive halos.","The lack of a strong relaxed-versus-disturbed difference suggests that global dynamical state plays a minor role in the average HI content of member galaxies."],"fun_headline_variants":["Cluster gas stripping starts far beyond the virial radius, out to 5R200","Even blue cluster galaxies lose gas before reaching the core","FAST stacking of 36 clusters shows HI loss out to 5R200","Gas depletion in cluster outskirts detected out to 5R200"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"Membership at 2–5R200 is defined by an approximate escape-velocity boundary; if interloper rejection or spectroscopic incompleteness correlates with HI content, the apparent deficit at large radius could be a selection artifact rather than genuine stripping.","fun_headline_variants_meta":{"raw":{"variants":["Cluster gas stripping starts far beyond the virial radius, out to 5R200","Even blue cluster galaxies lose gas before reaching the core","FAST stacking of 36 clusters shows HI loss out to 5R200","Gas depletion in cluster outskirts detected out to 5R200"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000664,"raw_usage":{"total_tokens":2926,"prompt_tokens":857,"completion_tokens":2069,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":601,"completion_tokens_details":{"reasoning_tokens":1991}},"tokens_in":601,"tokens_out":2069,"duration_ms":16452,"temperature":1.0,"reasoning_tokens":1991,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T01:26:13.422641+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A stacking analysis that splits the 4–5R200 sample into galaxies well inside versus near the assumed escape boundary, or uses high-resolution interferometric data to confirm membership and resolve blends, would settle the claim: if the field-normalized HI deficit vanishes when interlopers are removed, the pre-processing conclusion fails.","supporting_citations":[],"review_version":1}