{"id":"ba59ade6-ea71-4849-9db2-b890a0a74868","arxiv_id":"2505.02687","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Cluster galaxies at redshift about 1.1 to 1.4 show lower IR-to-radio ratios than field galaxies at roughly 2 to 3 sigma, suggesting a possible environment-linked radio excess.","lead":"The paper compares infrared-to-radio light ratios of galaxies in 11 massive clusters at redshifts 1 to 1.8 with field galaxies, using new VLA 3 GHz images. It finds a marginal difference, strongest around z 1.1 to 1.4, hinting that cluster environments may alter star formation or radio emission as quenching begins.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The KS-based cluster/field offset may be an artifact of pooling template-derived radio upper limits for 51 non-detected galaxies with detections, since those limits assume the IR-radio correlation and the cluster images are shallower than GOODS-S.","rationale":"The paper is an honest, modest observational study with a genuinely new 11-cluster VLA dataset, and the authors explicitly hedge their conclusions. However, the central statistical claim depends on the KS comparison between cluster and field q distributions, and that comparison is underspecified at exactly the load-bearing point: whether and how q lower limits from 51 radio-undetected cluster members enter the test. The method used to construct those limits, a Rieke et al. (2009) SFG template, assumes the IR-radio correlation that the paper is trying to test, so including the resulting values as point data would be partly circular. In addition, the quoted GOODS-S depths (0.75 microJy/beam at 3 GHz, 20 microJy at 24 micron) are not equivalent to the cluster depths (4-10 microJy/beam at 3 GHz, 36-156 microJy at 24 micron), so the field sample will be less censored and the cluster sample will miss radio-faint galaxies, both of which can shift the apparent q distributions. The p-values quoted (0.01-0.04) are marginal enough that a proper treatment of censoring or depth could easily move them above the conventional 0.05 threshold. The physical interpretation of a low-redshift radio excess may be correct, but the current analysis does not yet establish it. The reader's weakest assumption identifies the same issue, and the appropriate remedy is to require the authors to specify and re-run the tests as described in concrete_test. No change from the reader's conditional verdict is needed; the paper should not be accepted until the censoring and template treatment are clarified.","tokens_in":18251,"tokens_out":6456,"duration_ms":72815,"concrete_test":"Re-run the comparison with the censoring made explicit: (i) run the KS tests using only the 78 radio detections; (ii) treat the 51 non-detections as q lower limits derived from the local 3-sigma RMS (not from Rieke et al. templates) and compare the cluster and field distributions with a survival-analysis two-sample test (e.g., Peto-Peto or log-rank) that handles censoring; (iii) repeat (i) and (ii) with the GOODS-S sample restricted to the same 3 GHz RMS and 24 micron depth as the cluster images. If the full-sample and low-redshift q24/q160 p-values rise above roughly 0.1 under (i)-(iii), the claimed cluster/field offset is an artifact of censoring, template assumptions, or depth mismatch.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's headline result is a KS test difference (q24 p=0.02, q160 p=0.01) between cluster and GOODS-S field q distributions. For 51/129 cluster members the 3 GHz radio flux is not detected; Section 2.3 says these 'upper limits' were computed by scaling the IR-detected flux and reading the predicted radio flux from Rieke et al. (2009) SFG templates. Those templates assume the very IR-radio correlation being tested. Section 4 then describes the KS tests without stating whether these censored q lower limits are included as point values, and Figs. 4-5 histograms appear to include them. Pooling lower limits with detections in a two-sample KS test is invalid: the test requires uncensored observations, and template-derived values can only pull the cluster distribution toward the assumed template q. Independently, the claimed depth equivalence in Section 2.4 is not supported by the quoted numbers: GOODS-S 3 GHz RMS is 0.75 microJy/beam and 24 micron 5-sigma is 20 microJy, versus cluster RMS 4-10 microJy/beam (Table 1) and 24 micron 3-sigma depths of 36-156 microJy (Section 2.1). A shallower cluster sample censors more radio-faint, high-q galaxies, which can produce a spurious low-q offset. Because the quoted p-values all sit near the 2-3 sigma threshold, either effect is sufficient to erase the claimed environmental signal.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper investigates whether the infrared-radio correlation parameter q differs between galaxies in massive clusters at 1 < z < 1.8 and field galaxies. Using 24 and 160 micron photometry from Spitzer/Herschel and new VLA 3 GHz imaging for 129 cluster members from 11 clusters, the authors compute q24 and q160 and compare their distributions with GOODS-S field galaxies via two-sample Kolmogorov-Smirnov tests. They report marginal differences for the full sample (p=0.02 for q24, p=0.01 for q160), find that this difference is driven by the low-redshift subsample (1<z<1.37), and interpret it as evidence for an environment-induced radio excess at the onset of widespread cluster quenching. They also report no significant q differences with projected clustercentric radius and no differences between AGN hosts and non-AGN star-forming galaxies. The conclusions are heavily hedged and the authors call for larger samples.","tokens_in":18548,"tokens_out":6154,"duration_ms":68808,"significance":"If the claimed environmental offset in q is real, this would be one of the first multi-cluster results extending the local radio-excess phenomenon to z~1-1.4, with implications for how cluster environments modify the star-formation and radio emission of galaxies during the quenching epoch. The paper's strengths include the homogeneous treatment of 11 clusters, use of new VLA data, multi-wavelength AGN identification, and comparison against a well-studied field sample. However, the central statistical evidence is only at the 2-2.5 sigma level, and the analysis has unresolved issues with censored data and depth mismatches that could plausibly manufacture the reported offsets. The result is therefore interesting but not yet established.","major_comments":[{"comment":"The KS tests in §4.1 do not state how the 51 radio-undetected cluster galaxies are treated. §2.3 explains that for these galaxies the radio flux 'upper limits' were computed by scaling IR-detected fluxes and reading predicted radio fluxes from Rieke et al. (2009) SFG templates, which encode the very IR-radio correlation under study. The histograms in Figs. 4 and 5 appear to include these template-derived lower limits as if they were measured point values. Two-sample KS tests require uncensored observations, and pooling model-dependent lower limits with detections can artificially pull the cluster q distribution toward the assumed template and can only bias the test. Please state explicitly the number of detections and limits entering each KS test, rerun the comparison using only the 78 secure radio detections, and (ideally) apply a censored-data test such as the Peto-Prentice or log-rank test that treats the lower limits as censored values.","section":"§2.3, §4.1"},{"comment":"The claim in §2.4 that the GOODS-S depth is 'equivalent to that of our data' is contradicted by the numbers quoted in the manuscript itself: the GOODS-S 3 GHz image has RMS 0.75 microJy/beam at the pointing center, while the cluster images have measured RMS values of 4-10 microJy/beam (Table 1); GOODS-S 24 micron reaches 5 sigma at 20 microJy, whereas the cluster 24 micron depths are 3 sigma at 36-156 microJy (§2.1). The cluster sample is therefore shallower in both bands, and the radio non-detections preferentially remove radio-faint, high-q galaxies from the cluster distribution. This censoring alone can produce a spurious low-q offset between cluster and field. The comparison should be repeated using a GOODS-S subsample that is flux-limited to match the cluster depths, or the analysis should explicitly model the selection function.","section":"§2.4, Table 1"},{"comment":"The statistical reporting is internally inconsistent and some statements overstate the results. In §4.2, KS p-values of 0.28 and 0.35 are described as 'low p values' that are 'highly suggestive of the null hypothesis'; these are high p-values (they fail to reject the null, which is not the same as evidence for the null), and the phrasing should be corrected. In §4.3, the sentence 'we tested if the q values that plotted as a function of projected radius from cluster center are different in two sub-samples' is confused and does not match the section title 'q values as a function of redshift'; please rewrite to state clearly which comparison is being made. Finally, §6 says 'We reject the null hypothesis ... at the 95% and 99% significance level', which is stronger than the marginal p-values (0.02 and 0.01) and contradicts the careful hedging elsewhere in the paper; this sentence should be softened to match the reported significance.","section":"§4.2, §4.3, §6"},{"comment":"The paper does not report the sample sizes of the GOODS-S field comparison in each redshift bin or the number of AGN excluded from each q24 subsample. Without these counts the reader cannot evaluate whether the KS p-values are driven by small-number statistics or by asymmetric sample sizes. Please provide N for the full sample and for every redshift bin, for both the cluster and field samples, and for both q24 and q160, and state how many of the cluster sample points in each bin are detections versus lower limits.","section":"§4.1"}],"minor_comments":[{"comment":"The sentence 'Since the estimated depths of the radio images are in the range of 2.0-7.0 microJy, we found that our observed data did not reach the expected RMS values' is confusing because the measured RMS values in Table 1 are 4-10 microJy/beam; please clarify the expected versus measured depths.","section":"§2.3"},{"comment":"The caption states that sources detected only in the IR are lower limits and are shown with 'upper gray arrows'; if these are lower limits in q, the arrows should point downward rather than upward, or the caption should be reworded to avoid confusion.","section":"Fig. 2 caption"},{"comment":"There is a duplicated phrase in the first sentence: 'we considered the variations in the local density by plotting in Fig. 6 the IR-radio the IR-radio correlation'; please remove the repetition.","section":"§4.2"},{"comment":"The table caption says 'number of detected sources', but the column sums to 129, the total number of cluster members in the sample; clarify whether this column lists the number of cluster members per cluster or the number of radio detections.","section":"Table 1"},{"comment":"The sentence 'One hundred and eleven sources are detected at 24 and 160 micron, respectively' is ambiguous; specify whether these are 111 sources detected in each band or 111 sources detected in both bands.","section":"§2.5"},{"comment":"Conclusion 6 states that the radio excess scenario 'causes the lower correlation values in galaxy clusters', which is stronger than the marginal KS evidence and the 'consistent with' language used elsewhere; please rephrase to 'is consistent with' or similar.","section":"§7, conclusion 6"}],"recommendation":"major_revision","confidential_remarks":"The manuscript addresses a timely question and brings a useful new dataset, but the central claim rests on KS tests that currently pool model-dependent lower limits with detections and on an 'equivalent depth' assertion contradicted by the numbers in the paper. These issues are fixable: the authors can rerun the comparison on detections only, apply a censored-data test, and match depths by restricting the field sample. Given the marginal p-values, I expect that a careful reanalysis may weaken the claimed offset, but the paper should not be accepted in its current form. The paper is within scope for A&A."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: this is the first multi-cluster, homogeneously analyzed IR-radio correlation study at z 1-1.8, using new VLA 3 GHz data. That is a real step up from the single-cluster studies (Randriamampandry et al. 2015; Kalita et al. 2021). The paper is honest about the modest significance (p ~ 0.01-0.04, 2-3 sigma) and explicitly asks for larger samples. Good.\n\nThe problem is the censoring. 51 of 129 cluster members have no 3 GHz detection. Section 2.3 says their radio fluxes were predicted from Rieke et al. 2009 SFG templates, which assume the very IR-radio correlation under test. The paper never states whether these lower limits enter the KS tests against GOODS-S. The histograms in Figs. 4-5 suggest they do. A two-sample KS test requires uncensored observations; pooling lower limits as point values is invalid and can inflate the apparent offset. The authors need to either restrict the KS tests to detections-only, or use proper survival analysis that handles censoring without template assumptions.\n\nSecond issue: the claimed depth equivalence with GOODS-S does not hold. Cluster images have RMS 4-10 microJy/beam (Table 1); GOODS-S is 0.75 microJy/beam. That is a 5-13x depth difference. A shallower radio image will miss faint radio sources (high-q galaxies), shifting the cluster distribution to lower q. This alone could produce the observed offset. The paper needs to demonstrate that the offset survives when matching the samples in radio luminosity or when applying a consistent detection threshold.\n\nThe direct flux ratios for the 78 detections are not circular—they are genuine measurements compared against an external field sample. But the template-derived 'upper limits' are circular by construction, and if they are in the KS test, the headline result is suspect. I would not take the environmental radio excess claim seriously until the authors redo the analysis with detections only and with proper censoring treatment, and/or correct the depth issue.\n\nStill, the dataset is valuable and the question is timely. The paper deserves peer review, but it needs significant revision. A referee should ask for the detections-only KS test, a statement of how non-detections were treated, and a proper survival analysis.\n\nBottom line: a promising pilot with a useful new dataset, but the central claim is not yet supported. I would engage with it rather than dismiss it.","headline":"First multi-cluster IR-radio study at z~1-1.8 with new VLA data, but the claimed cluster/field offset is not robust until the censored non-detections and depth mismatch are properly handled.","tokens_in":19170,"tokens_out":5215,"would_cite":false,"duration_ms":55026,"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":"Cluster galaxies at $z\\approx1$–$1.4$ show a modest radio excess relative to field galaxies.","keywords":["IR-radio correlation","galaxy clusters","cosmic noon","environmental quenching","radio excess","Kolmogorov-Smirnov test","q24","q160"],"falsifier":"Recompute the two-sample KS tests using only the 78 galaxies with secure 3 GHz detections; if the cluster-field $p$-values rise above $0.05$ and the mean-$q$ difference disappears, the claimed environmental radio excess is produced by the template upper limits rather than by the cluster environment.","tokens_in":17991,"feed_emoji":"📡","tokens_out":10986,"duration_ms":114023,"temperature":0.7,"pith_summary":"This paper asks whether the dense environment of a massive galaxy cluster changes the ratio of infrared to radio emission in star-forming galaxies at $1<z<2$, the era when cluster quenching first turns on. Using 129 member galaxies across 11 massive clusters, with new 3 GHz radio images and infrared data at 24 and 160 $\\mu$m, it computes two versions of the IR-radio correlation parameter, $q_{24}$ and $q_{160}$, and compares their distributions with a matched field sample using Kolmogorov-Smirnov tests. The central claim is that cluster galaxies have a modest but detectable radio excess, meaning a deficit in $q$, relative to field galaxies, at roughly $2\\sigma$–$3\\sigma$ significance, and that the signal is carried by the lower-redshift subsample at $z<1.37$, exactly where widespread environmental quenching begins. If true, the IR-radio correlation becomes an early-warning indicator of environmental transformation in clusters.","feed_headline":"Radio excess marks the moment clusters start quenching galaxies","feed_subtitle":"Cluster galaxies show lower IR-radio ratios than field galaxies at z<1.37, marking the start of quenching.","key_machinery":"The argument runs on two flux-ratio parameters: $q_{24}\\equiv\\log(S_{24}/S_{1.4})$ and $q_{160}\\equiv\\log(S_{160}/S_{6})$, where a lower value means more radio emission per unit infrared emission. The 3 GHz radio fluxes are rescaled to 1.4 and 6 GHz with a fixed spectral index $\\alpha=-0.75$, and the two $q$ distributions are compared across cluster and field samples with two-sample Kolmogorov-Smirnov tests. For the 51 radio-undetected cluster galaxies, radio upper limits are generated from star-forming galaxy templates, and these limits are plotted alongside the 78 detections in the KS comparisons. The redshift split at $z=1.37$, taken from the underlying cluster survey, separates the epoch when quenching begins from the epoch when it does not.","core_discovery":"The paper's central discovery is that the IR-radio correlation parameter in cluster galaxies is systematically lower than in field galaxies at $1<z<1.37$. For the full sample the two-sample Kolmogorov-Smirnov tests return $p=0.02$ for $q_{24}$ and $p=0.01$ for $q_{160}$, both favouring distinct parent distributions; in the low-redshift bin the values are $p=0.03$ and $p=0.04$. Because the mean $q$ values are lower in the clusters, the offset is a radio excess rather than an infrared deficit. The high-redshift bin ($1.37<z<1.8$) shows no significant difference, which the authors read as the environment not yet having imprinted on the galaxies. No difference is found between AGN hosts and non-active galaxies, and all identified AGNs are radio quiet, so the excess is attributed to star-formation-related processes such as ram pressure stripping or galaxy interactions that accompany the onset of cluster quenching.","pith_inferences":["Excluding the 51 template-based upper limits and rerunning the KS tests on the 78 detections would provide a direct test of whether the claimed offset is a real environmental signal or an artifact of censoring and template assumptions.","Deeper VLA observations of the same clusters could convert those upper limits into detections; if the offset persists, it would strengthen the radio-excess interpretation, and if it dissolves, the claim would need revision.","If radio excess and quenching share a common cause, lower-redshift clusters should show progressively larger offsets; this can be checked by adding clusters at $z\\approx1$.","A practical extension: use radio-excess selection from $q$ as a cheap quenching-stage indicator in wide-area cluster surveys that lack deep infrared data."],"forward_implications":["If the offset is real, the IR-radio correlation can serve as an observational tracer of the onset of environmental quenching at $z\\sim1.4$.","At $z\\gtrsim1.4$, cluster and field galaxies share the same $q$ distributions, so environmental effects have not yet altered the radio/IR balance at that epoch.","Because all identified AGNs are radio quiet, the radio excess cannot be blamed on active nuclei and is more likely tied to star-formation-related processes.","The absence of a radial gradient suggests the environmental imprint is not limited to cluster cores but affects the galaxy population across the cluster.","Larger samples will be needed to raise the $\\sim2\\sigma$–$3\\sigma$ result to a definitive detection."],"supporting_citations":[{"why":"Supplies the 11-cluster sample, the star-formation and AGN measurements, and the $z=1.37$ quenching split that structures the analysis.","marker":"Alberts et al. 2016"},{"why":"Provides the $q_{24}$ and $q_{160}$ definitions, the fiducial $q_{160}$ value, and the deep 3 GHz field catalog used as the comparison sample.","marker":"Alberts et al. 2020"},{"why":"Provides the star-forming galaxy SED templates used to compute radio upper limits for the 51 radio-undetected cluster galaxies.","marker":"Rieke et al. 2009"},{"why":"Justifies the $α=-0.75$ spectral index assumption used to scale 3 GHz fluxes to 1.4 and 6 GHz.","marker":"Delvecchio et al. 2021"},{"why":"Provides spectroscopic confirmation and the IRAC/MIPS photometry that define the cluster member sample.","marker":"Brodwin et al. 2013"},{"why":"Supplies the AGN catalog for the field comparison sample so that AGN hosts can be excluded from the $q_{24}$ comparison.","marker":"Lyu et al. 2022"},{"why":"Establishes the local-Universe precedent of a small but significant radio excess in cluster galaxies that this work extends to higher redshift.","marker":"Miller & Owen 2001"}],"fun_headline_variants":["Radio excess in clusters signals onset of quenching at z<1.4","Cluster galaxies show radio excess as quenching begins","Quenching onset imprinted in radio excess of cluster galaxies","Radio excess marks first quenching in massive clusters at cosmic noon","IR-radio offset reveals quenching start in cluster galaxies at z<1.4"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The comparison treats the template-based radio upper limits for the 51 undetected cluster galaxies as data, even though those templates assume the very IR-radio correlation the paper is testing.","fun_headline_variants_meta":{"raw":{"variants":["Radio excess in clusters signals onset of quenching at z<1.4","Cluster galaxies show radio excess as quenching begins","Quenching onset imprinted in radio excess of cluster galaxies","Radio excess marks first quenching in massive clusters at cosmic noon","IR-radio offset reveals quenching start in cluster galaxies at z<1.4"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000499,"raw_usage":{"total_tokens":2516,"prompt_tokens":1094,"completion_tokens":1422,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":710,"completion_tokens_details":{"reasoning_tokens":1346}},"tokens_in":710,"tokens_out":1422,"duration_ms":12484,"temperature":1.0,"reasoning_tokens":1346,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:44:49.363636+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the two-sample KS tests using only the 78 galaxies with secure 3 GHz detections; if the cluster-field $p$-values rise above $0.05$ and the mean-$q$ difference disappears, the claimed environmental radio excess is produced by the template upper limits rather than by the cluster environment.","supporting_citations":[{"cited_title":"H., & Rujopakarn, W","cited_arxiv_id":null,"evidence_quote":"Supplies the AGN catalog for the field comparison sample so that AGN hosts can be excluded from the $q_{24}$ comparison."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the local-Universe precedent of a small but significant radio excess in cluster galaxies that this work extends to higher redshift."}],"review_version":1}