{"id":"6b50bfdc-8b1e-46cb-ba38-1fb0af16bb37","arxiv_id":"2411.15314","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Radio AGN hosts have low star-forming fractions mostly because they are massive galaxies, and once matched in mass and redshift they closely track normal galaxies out to z ~ 3.","lead":"Galaxies that host a radio-loud active black hole are often thought to be 'red and dead', but this study finds that is mainly because they are massive, not because the black hole quenches star formation. Using deep survey data across three sky fields, it maps how star formation in these hosts compares to normal galaxies from redshift 0 to 3.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Radio-excess selection depends on SFR through 24 µm flux, so the headline fSF and ΔMS results may be biased; the paper only removes this bias in §3.3, not in Figures 7–8.","rationale":"The reader's weakest assumption (mass completeness via Pozzetti) is reasonable, but it is not the point where the central claim is least secure: the mass limits are computed separately for star-forming and quiescent galaxies, the Mstar-matched comparison is designed to remove the mass dependence, and most radio AGNs lie above the adopted limits. A more direct threat is the SFR-dependence of the radio-excess selection itself. The q24 and qIRRC criteria use the ratio of IR/FIR to radio luminosity; for a fixed radio luminosity, a higher-SFR host has a larger denominator in q24=log(S24/S1.4), making it harder to satisfy the radio-excess threshold. The authors acknowledge this in Section 3.3 and construct an (Mstar,z)-dependent S1.4 threshold above which the selection is SFR-independent, but the headline fSF and ΔMS results in Figures 7 and 8 are not restricted to this unbiased subsample. If the bias is strong, it would lower the measured fSF and push ΔMS negative preferentially at low z and high Mstar, exactly the regime where the paper claims suppressed star formation. The proposed test—recomputing the headline maps with the Section 3.3 subsample—uses only quantities already in Table 2 and would settle the matter. I therefore keep the conditional verdict unchanged, with this check as a condition.","tokens_in":28789,"tokens_out":9355,"duration_ms":95134,"concrete_test":"Recompute Figure 7 (fSF vs z and matched-galaxy fSF) and Figure 8 (ΔMS vs z and vs M*,z) using only the 869 radio AGNs (416 for ΔMS) above the (M*,z)-dependent S1.4GHz threshold defined in Section 3.3, with the same M*/z matching. If the residual fSF gap and the negative low-z/high-M* ΔMS shrink to ≲5% and ≲0.1 dex respectively, the claimed suppression is largely a selection artifact; if they persist, the conclusions are robust. A complementary check is to simulate the q24 selection function by injecting AGN radio luminosities into MS-galaxy SEDs and computing recovery fractions as a function of ΔMS, z, and M*, then correct Figs. 7–8.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing untested input is the SFR dependence of the radio-AGN selection itself. Both the Zhu et al. (2023) q24 selection (Sec. 2.1) and the new qIRRC selection (Sec. 2.2) use a radio-excess criterion: at fixed 1.4 GHz luminosity, a host with higher SFR has higher 24 µm/FIR emission, hence a higher q24/qIR, and is less likely to be classified as a radio AGN. The authors explicitly recognize this in Sec. 3.3 and define an (M*,z)-dependent S1.4GHz threshold above which selection no longer depends on SFR. However, the headline fSF comparison (Fig. 7) and the ΔMS analysis (Fig. 8) are made on the full samples, not on the SFR-independent subsample. If the selection preferentially removes star-forming hosts, the measured fSF gap and the negative ΔMS in the low-z/high-M* regime are overestimated. This is not an internal inconsistency, but it means the central claim about the size of the AGN-suppression effect is not yet secure.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript uses multiwavelength data in the W-CDF-S, ELAIS-S1, and XMM-LSS fields to construct two radio-AGN samples (the Zhu et al. 2023 sample and a new qIRRC-based sample) and two star-forming galaxy definitions (UVJ and nSFR). It computes the star-forming fraction fSF and the offset ΔMS from the star-forming main sequence as functions of stellar mass and redshift up to z≈3. After mass-matching radio AGNs to control galaxies, the authors conclude that the low fSF of radio AGNs is primarily a stellar-mass effect, and that star-forming radio AGNs are suppressed relative to the main sequence only in low-z, high-M* hosts, while being on or above the main sequence at higher redshifts or lower M*. The paper includes multiple robustness checks using alternative AGN selections and star-forming galaxy definitions.","tokens_in":29016,"tokens_out":10168,"duration_ms":103223,"significance":"If the conclusion holds, the finding that the apparent quiescence of radio-AGN hosts is mostly a mass-selection effect, with only a modest residual suppression that changes sign with z and M*, would sharpen current understanding of radio-mode feedback and host-galaxy evolution. The paper's strengths are its large multi-field samples, explicit mass-completeness treatment, M*-matched controls, and several internal cross-checks: two radio-AGN selections, two star-forming galaxy definitions, and an S1.4GHz flux-cut test. It also provides a radio-AGN catalog. The main open concern is that the headline fSF and ΔMS results are not yet shown to be free of the SFR-dependent selection effect that the paper itself identifies in Section 3.3.","major_comments":[{"comment":"The radio-AGN selection criteria in both samples are SFR-dependent: at fixed S1.4GHz, a host with higher SFR has higher S24µm (or LIR), hence a larger q24/qIR and is less likely to pass the radio-excess threshold. This biases the full-sample fSF low and ΔMS low, especially in the low-z/high-M* bins where the claimed suppression is largest. Section 3.3 identifies this effect and removes it with an (M*, z)-dependent S1.4GHz threshold, but that threshold is applied only to the L1.4GHz-split analysis in Figures 9-10, not to the headline fSF and ΔMS comparisons in Figures 7-8 and 11. Please recompute the main fSF and ΔMS results using only sources above the Section 3.3 threshold, or otherwise quantify the size of this selection bias; as written, the magnitude of the residual 'AGN suppression' component of the central claim is not yet secure.","section":"Sec. 2.1, Sec. 2.2, Sec. 3.3; Figs. 7-8"},{"comment":"The M*-matched control uses bins of ±0.1 dex in M* and ±0.075(1+z) in z. Because fSF declines steeply with M* and evolves rapidly with z, a systematic placement of AGNs at the high-M* or high-z edge of their matching bin would make the matched-galaxy fSF too high and hence mimic a small residual AGN effect. The paper checks narrower M* and z bin widths for the ΔMS analysis in the footnote to Section 3.2, but it does not report an equivalent test for the fSF control in Figure 7. Please report the median M* and z offsets of the matched pairs for the fSF analysis and verify that the conclusions are unchanged with narrower matching bins, for example ±0.05 dex or ±0.02 dex in M*.","section":"Sec. 3.1"}],"minor_comments":[{"comment":"The text says the new selection is missing only 11 of the 1718 Zhu et al. (2023) radio AGNs 'as we aim to select more AGNs while keeping the original ones'; please clarify why these 11 are not recovered by the looser criterion.","section":"Sec. 2.2"},{"comment":"The caption states the comparison with Leja et al. (2022) and Popesso et al. (2023) spans z=0-2, while the figure panels show z=0.5-3.0; please correct the redshift range.","section":"Fig. 5 caption"},{"comment":"The sentence 'Figure 6 shows the relationship between SFR and M* at six redshifts' appears to refer to the main-sequence plot in Figure 5; please fix the cross-reference.","section":"Appendix A.1"},{"comment":"Several cited works are given as arXiv preprints (e.g., Igo et al. 2024, Wang et al. 2024); please update to the published versions if they have appeared by the time of the revision.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The main requested revision is concrete and lies within the authors' existing methodology: use the Section 3.3 flux-limited subsample to recompute the headline fSF and ΔMS maps, and add the matching-width test for fSF. If those checks confirm the current trends, the paper will be a solid contribution. I see no reason for rejection; the issue is incomplete validation of a load-bearing point rather than an internal contradiction."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper is a serious piece of work: it gives the first population-scale fSF and ΔMS for radio AGN hosts out to z~3 in W-CDF-S, ELAIS-S1, and XMM-LSS, with a new 6,766-object IRRC-selected sample alongside the higher-purity Zhu et al. sample. The core result—that the low fSF of radio AGNs is mostly a reflection of their stellar mass distribution, not an independent quenching effect—is well supported by the mass-matched control comparison, and the same trend appears with two radio AGN selections, two star-forming definitions, and after applying a 1.4 GHz flux cut. That consistency is the paper's main strength.\n\nThe soft spots are real but not fatal. First, the new sample's quoted purity and completeness come from a fitted 0.3 dex shift in the IRRC normalization and a symmetry assumption for the star-forming galaxy distribution; neither is independently validated. Second, the mass-completeness limits from VIDEO Ks via Pozzetti et al. could be biased if the Ks selection behaves differently for quiescent vs star-forming galaxies, and the whole mass-matched argument depends on those limits. The reader flagged this; it's a legitimate concern.\n\nThe stress-test about SFR-dependent selection also lands. Both radio AGN selections use the radio-IR ratio, so at fixed 1.4 GHz luminosity, a higher-SFR host is harder to classify as a radio AGN. That means Figures 7 and 8, which use the full sample, may overestimate the fSF gap and the negative ΔMS. The authors know this—Section 3.3 builds an (M*,z)-dependent luminosity threshold above which selection is SFR-independent—but they don't apply that cut to the headline figures. The good news is that the bias works in the direction of making the mass-matching conclusion conservative: if you are missing star-forming hosts, the true fSF of radio AGNs is even closer to the galaxy value. And the Section 3.3 analysis for the bright subsample shows the same qualitative pattern, so the central claims probably survive. But the reader can't tell from Figures 7-8 how large the bias is.\n\nOverall, this is a paper that deserves a serious referee. It is not a definitive settling of the question—the selection bias and SED systematics prevent that—but it is a large, careful, well-documented step. I'd recommend sending it to review with a request to move the SFR-independent analysis into the main figures, or at least quantify the bias on the measured fSF and ΔMS, and to add an independent check on the new sample's purity (e.g., stacking or a radio spectral index diagnostic). The reader's conditional verdict is about right.","headline":"A careful, large-sample study of radio AGN star-forming fractions to z~3; the mass-driven quenching claim holds up, but the selection bias in the headline figures means the suppression amplitude is not yet secure.","tokens_in":29603,"tokens_out":3489,"would_cite":true,"duration_ms":32539,"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 paper argues that radio AGNs appear star-formation-suppressed mainly because they live in massive, quiescent galaxies, and that on the star-forming main sequence they are suppressed only at low redshift and high stellar mass.","keywords":["radio active galactic nuclei","star formation","main sequence","star-forming fraction","mass completeness","infrared-radio correlation","galaxy evolution","AGN feedback"],"falsifier":"Measure the star-formation rates of the same radio-AGN hosts and the same mass-matched control galaxies with an independent, template-free tracer such as molecular-gas (CO) emission or deep dust-continuum stacking; if the SED-based pattern — below the MS at low $z$ and high $M_\\star$, on or above it at high $z$ and low $M_\\star$ — is not reproduced, the conclusions rest on SED-template systematics rather than real differences in star formation.","tokens_in":28558,"feed_emoji":"📡","tokens_out":11866,"duration_ms":105806,"temperature":0.7,"pith_summary":"This paper investigates whether radio-emitting active galactic nuclei (radio AGNs) suppress or enhance star formation in their host galaxies, a central question for how supermassive black holes shape galaxy evolution. Using radio and multiwavelength data in three deep sky fields, the authors compare the star-forming fraction ($f_{\\rm SF}$, the fraction of star-forming galaxies among all galaxies) of radio AGNs with that of normal galaxies, and then with control galaxies matched in stellar mass ($M_\\star$) and redshift. Their central finding is that the low $f_{\\rm SF}$ of radio AGNs at $z\\lesssim 2$ is mostly a mass-selection effect: radio AGNs preferentially live in massive galaxies, which are quiescent anyway, and once $M_\\star$ and $z$ are matched the $f_{\\rm SF}$ deficit shrinks to roughly 10% or less. For the star-forming subset, they find that radio AGNs sit below the star-forming main sequence (MS) in massive, low-redshift hosts but on or above the MS at high redshift or low stellar mass, so the apparent influence of radio activity on star formation depends strongly on when and where the host galaxy sits.","feed_headline":"Stellar mass, not black hole activity, explains radio AGN quiescence","feed_subtitle":"Radio AGN hosts form stars like normal galaxies once mass is accounted for.","key_machinery":"The load-bearing machinery is the matched-control comparison: for each radio AGN, one hundred reference galaxies are drawn with replacement from within $0.1$ dex in stellar mass and $0.075(1+z)$ in redshift, so the control population shares exactly the two properties that most strongly determine whether a galaxy is star-forming or quiescent. The comparison is built on mass-complete samples, with redshift-dependent mass limits computed from the VIDEO $K_s$-band depth using a standard mass-limit relation and applied separately to star-forming and quiescent galaxies, preventing the easier detectability of star-forming systems from masquerading as an AGN effect. The second mechanism is the star-forming main sequence itself, defined as the median SFR of $UVJ$-selected star-forming galaxies in bins of $M_\\star$ and $z$, against which each AGN's $\\Delta_{\\rm MS}$ is measured. The paper checks the stability of this machinery by repeating the analysis with an nSFR-based star-forming definition and with a looser, more complete radio-AGN selection, finding the same qualitative behavior.","core_discovery":"On the authors' own terms, the discovery is that the apparent quiescence of radio AGNs is primarily demographic. At $z\\approx0-0.5$ only about 11% of radio AGNs are star-forming compared with 76% of galaxies, but when each AGN is compared with galaxies matched within $0.1$ dex in $M_\\star$ and $0.075(1+z)$ in $z$, the gap is only a few to ten percent. The radio-AGN $f_{\\rm SF}$ rises from roughly 10% at $z<0.5$ to about 65% at $z\\approx2-2.5$, mirroring the general galaxy population, and the same trends hold for a second, three-times-larger radio-AGN sample selected through the infrared-radio correlation. For star-forming radio-AGN hosts, the offset from the MS, $\\Delta_{\\rm MS}=\\log({\\rm SFR}_{\\rm AGN}/{\\rm SFR}_{\\rm MS})$, has a global median of $-0.03\\pm0.03$ dex, with a weak positive correlation with redshift and a negative correlation with stellar mass; massive low-redshift hosts fall below the MS while low-mass or high-redshift hosts reach or exceed it. The authors also report that radio luminosity has little influence on $f_{\\rm SF}$ or $\\Delta_{\\rm MS}$ once stellar mass is accounted for, and that radio AGNs identified in X-rays or the mid-infrared differ in SFR from other radio AGNs by only $\\lesssim0.2$ dex.","pith_inferences":["A natural extension the authors leave implicit: weight the reference galaxy population by the observed $M_\\star$-$z$ distribution of radio AGNs and compare population-average SFRs; if the mass-matched result is correct, the weighted galaxy average should match the AGN-host average at every epoch without any need for AGN feedback.","The looser infrared-radio-correlation-selected sample has a 5–20% higher $f_{\\rm SF}$ than the strict radio-excess sample, suggesting that strict radio-excess cuts preferentially retain older, more luminous jets; deeper radio selection reaching fainter jets may therefore find even less evidence of AGN-driven suppression.","The paper's binned maps imply a testable prediction: in bins where AGN hosts and the general population share the same mass distribution, the incidence ratio of radio AGNs in star-forming versus quiescent galaxies should approach unity, and the few bins that already exceed unity at high $z$ and low $M_\\star$ could be pushed to $z>3$ to see whether radio AGNs actually prefer star-forming hosts at t"],"forward_implications":["At $z\\lesssim0.5$, roughly one in ten radio AGNs is star-forming while about three of four galaxies are, but this gap largely disappears after matching stellar mass, so future studies of AGN quenching must control for host mass before attributing low $f_{\\rm SF}$ to feedback.","The radio-AGN $f_{\\rm SF}$ rises steeply with redshift and falls with stellar mass, closely tracking the behavior of normal galaxies, reaching about 65% star-forming at $z\\approx2-2.5$.","Star-forming radio-AGN hosts are not generally suppressed: the global median offset from the MS is $\\Delta_{\\rm MS}=-0.03\\pm0.03$ dex, with suppression concentrated in massive, low-redshift systems.","At fixed stellar mass and redshift, the 1.4 GHz radio luminosity has only a minor influence on $f_{\\rm SF}$ or $\\Delta_{\\rm MS}$, so the instantaneous power of the radio jet is not the main controller of the host galaxy's star-formation rate.","Radio AGNs with X-ray or mid-infrared signatures (likely radiative-mode sources) have SFRs within about $0.2$ dex of those without such signatures, implying similar host-galaxy star formation for the two excitation classes at high redshift."],"supporting_citations":[{"why":"Constructs the primary 1,763 radio-AGN sample via radio morphology, flat spectral slope, and radio-excess selection; the paper's main analysis uses this sample.","marker":"Zhu et al. (2023)"},{"why":"Supplies the multiwavelength catalog with CIGALE SED-derived stellar masses, SFRs, photometric redshifts, and AGN flags for all galaxies and radio sources.","marker":"Zou et al. (2022)"},{"why":"Establishes the Ks-band mass-completeness procedure in these fields that the paper follows to build mass-complete galaxy and AGN samples.","marker":"Zou et al. (2024)"},{"why":"Provides the formula used to convert Ks-band magnitudes into per-object stellar mass limits, from which the redshift-binned mass-completeness thresholds are derived.","marker":"Pozzetti et al. (2010)"},{"why":"Calibrates the infrared-radio correlation qIRRC(Mstar,z) used to define the new, larger radio-AGN sample through 2-sigma radio excess.","marker":"Delvecchio et al. (2021)"},{"why":"Supplies the UVJ color cuts and the main-sequence framework used to separate star-forming from quiescent galaxies and to define the MS.","marker":"Whitaker et al. (2015)"},{"why":"Provides the local observational baseline that jet-mode radio AGNs reside in quiescent, massive galaxies, which the paper's mass-matching result reinterprets.","marker":"Heckman & Best (2014)"},{"why":"One of the two recent main-sequence determinations the paper compares against to validate its own MS normalization.","marker":"Leja et al. (2022)"},{"why":"The other recent main-sequence determination used to check that the paper's MS agrees with the literature at z≈0.5–3.","marker":"Popesso et al. (2023)"},{"why":"A recent comparable study finding no strong preference of radio AGNs for low-SFR galaxies; the paper's incidence-ratio results broadly agree with it.","marker":"Igo et al. (2024)"}],"fun_headline_variants":["Radio AGN quiescence traced to stellar mass, not black holes","Star formation in radio AGNs mirrors normal galaxies after mass match","Mass, not black hole activity, sets the star-forming fraction of radio AGNs","Demographics, not black hole feedback, govern radio AGN star formation"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything hinges on the assumption that the $K_s$-band-derived mass-completeness limits and the SED-based stellar masses and star-formation rates are unbiased at every redshift, because a hidden selection or measurement bias would make the mass-matched control sample unrepresentative and the conclusion that stellar mass rather than AGN activity drives the low $f_{\\rm SF}$ would be an artifact.","fun_headline_variants_meta":{"raw":{"variants":["Radio AGN quiescence traced to stellar mass, not black holes","Star formation in radio AGNs mirrors normal galaxies after mass match","Mass, not black hole activity, sets the star-forming fraction of radio AGNs","Demographics, not black hole feedback, govern radio AGN star formation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000783,"raw_usage":{"total_tokens":3590,"prompt_tokens":1214,"completion_tokens":2376,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":830,"completion_tokens_details":{"reasoning_tokens":2297}},"tokens_in":830,"tokens_out":2376,"duration_ms":17547,"temperature":1.0,"reasoning_tokens":2297,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T14:27:09.256841+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the star-formation rates of the same radio-AGN hosts and the same mass-matched control galaxies with an independent, template-free tracer such as molecular-gas (CO) emission or deep dust-continuum stacking; if the SED-based pattern — below the MS at low $z$ and high $M_\\star$, on or above it at high $z$ and low $M_\\star$ — is not reproduced, the conclusions rest on SED-template systematics rather than real differences in star formation.","supporting_citations":[],"review_version":1}