{"id":"db9bfaa8-e349-463d-8f06-906353a67918","arxiv_id":"1908.11380","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"In the EAGLE and IllustrisTNG simulations, galaxies hosted by haloes with low circumgalactic gas fractions are preferentially quenched and spheroidal, and the authors attribute this to AGN feedback expelling the gas.","lead":"This paper compares two major computer simulations of galaxy formation and finds that in both, galaxies with less gas in their outskirts are more likely to be dead and round. The authors conclude that black hole feedback blows away this gas, cutting off the fuel for new stars.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Title-level 'expulsion of efficiently-cooling gas' is not directly demonstrated in TNG; Fig. 3's cooling-time contrast is density-degenerate, and Section 4.1 concedes no fluid-element tracking was done.","rationale":"The correlational core of the paper is solid: the residual correlations in EAGLE and TNG are strong, use public simulations, and the cross-code comparison is meaningful despite shared BH seeding and growth prescriptions. I agree with the reader that the weak point is the causal, mechanism-specific interpretation, and I would keep the CONDITIONAL verdict unless the authors add direct evidence or soften the abstract and title. My concern is more pointed than 'correlation is not causation': the specific phrase 'expulsion of efficiently-cooling gas' has two unverified parts in TNG -- that gas is expelled (crosses r200 outward) and that the expelled gas is preferentially the fast-cooling component. Fig. 3 is the in-paper evidence most suggestive of the latter, but it is degenerate with overall density: CGM-poor haloes have lower density by definition, so their cooling times are longer even if expulsion is cooling-time-agnostic. The authors concede this in Section 4.1. Consequently, the 'efficiently-cooling' selectivity, which is what makes the mechanism distinctive, is imported from O19 for EAGLE and not independently established here for TNG. The claim of 'consensus' between simulations is therefore stronger than the analyses in this paper warrant. An AGN-off control run or high-cadence fluid tracking in TNG would settle the mechanism; absent that, the title and abstract should be softened from asserting consensus on expulsion to saying the results are consistent with it. Since the reader's recommended remedies are exactly these, no verdict adjustment is needed beyond the existing CONDITIONAL.","tokens_in":35996,"tokens_out":5638,"duration_ms":62632,"concrete_test":"Run the O19 snipshot analysis on TNG-100 (or TNG-50 if higher-cadence outputs exist): identify kinetic AGN feedback events, trace the gas cells that subsequently cross r200 outward, and compare the pre-event tcool and entropy distributions of ejected cells against those of cells that remain inside r200. Also check whether the fCGM decline after feedback events is actually due to outward crossing of r200 or to heating in place. If ejected cells are not preferentially drawn from the short-tcool tail, or if heating without crossing accounts for the decline, the central mechanism-specific claim is falsified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central, title-level claim is that AGN feedback quenches galaxies and drives morphological evolution by expelling efficiently-cooling CGM gas. The strongest direct evidence cited is the companion paper O19, which used high-cadence EAGLE snipshots; this paper does not reproduce that analysis for TNG. Section 4.1 explicitly concedes that 'an explicit demonstration that feedback preferentially expels rapidly-cooling gas would require the detailed tracking of fluid elements with high temporal resolution, which is beyond the scope of this study.' The Fig. 3 comparison of cooling-time CDFs for CGM-rich versus CGM-poor stacks cannot fill that gap: because fCGM is lower in the poor stacks, their gas is at lower density and therefore has longer tcool by construction (since Λ is proportional to n_H^2), a degeneracy the authors themselves note. Thus the 'efficiently-cooling' selectivity -- the part of the claim that distinguishes expulsion from undifferentiated mass loss, preventative heating, or consumption of gas by star formation -- rests on EAGLE-only snipshots and is merely assumed for TNG. The abstract's 'consensus from these state-of-the-art simulations' therefore overstates the evidence. This is a correctness risk for the causal mechanism, not for the correlational results, which are robust and clearly presented.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper examines the connection between the circumgalactic medium (CGM) and the quenching and morphological evolution of central galaxies in the EAGLE and IllustrisTNG cosmological simulations. Using the mass fraction of non-star-forming gas within r200 (fCGM) and residual correlations at fixed halo mass, the authors report that, in both simulations, scatter in fCGM correlates strongly with central black hole mass, specific star formation rate, and kinematic morphology, with the strongest correlations near L* halo masses. They interpret these correlations as evidence that AGN feedback expels efficiently-cooling circumgalactic gas, thereby elevating the CGM cooling time, starving the galaxy of fuel, and facilitating both quenching and morphological transformation. The paper also connects scatter in fCGM to halo concentration and formation time via the cumulative energy injected by efficient feedback relative to halo binding energy. The analysis is correlational; the explicit claim that feedback preferentially expels rapidly-cooling gas is deferred to a companion paper for EAGLE and is not directly demonstrated for IllustrisTNG.","tokens_in":36246,"tokens_out":4130,"duration_ms":42816,"significance":"If the causal interpretation holds, the paper identifies a concrete physical channel linking AGN feedback, CGM thermodynamics, and galaxy quenching, with testable predictions such as the opposite signs of the fCGM versus AGN accretion-rate correlation in EAGLE and TNG. The correlational results are a useful addition to the literature: the analysis is careful in its use of matched dark-matter-only haloes, residual statistics with significance masking, and robustness checks such as measuring fCGM within 0.3 r200. The fCGM and galaxy properties are independent outputs of the simulations, and the CGM was not a calibration target, so the correlations are emergent rather than fitted. However, the paper's central mechanistic claim—that expulsion of efficiently-cooling gas is the crucial causal step—is not established by the evidence presented here, and the abstract's 'consensus' phrasing overstates the direct support for that specific mechanism.","major_comments":[{"comment":"The comparison of cooling-time CDFs between CGM-rich and CGM-poor stacks does not demonstrate that feedback preferentially expels efficiently-cooling gas. Because the radiative cooling time scales approximately as tcool ∝ u/(n_H^2 Λ), and the CGM-poor haloes have lower fCGM and therefore lower characteristic gas densities, their gas has longer cooling times by construction. The authors themselves acknowledge this ('The differing cooling times of the gas-rich and gas-poor samples stem almost entirely from their necessarily different characteristic CGM densities'), so the 'paucity of efficiently-cooling gas' is a density-driven consequence rather than evidence for cooling-time-selective expulsion. The subsequent concession that explicit demonstration would require tracking fluid elements with high temporal resolution confirms that the expulsive, cooling-time-selective mechanism is not directly shown in either simulation; for TNG it is imported from the EAGLE-only companion paper O19. Since this mechanism is the paper's title-level claim, the evidence is insufficient as presented.","section":"Section 4.1, Fig. 3"},{"comment":"The causal direction of the fCGM versus sSFR and morphology correlations is asserted rather than tested. The paper interprets low fCGM as the cause of quenching via starvation, but the correlations are equally consistent with quenching causing low fCGM (e.g., through reduced gas accretion or consumption) or with a common cause such as halo assembly bias. The caveat in Section 3 that the existence of correlations 'does not imply that both correlations necessarily emerge as a direct response to the same physical mechanism' is appropriate, but it is not carried into the abstract or the concluding Section 6, which adopt strong causal language. A concrete test, such as examining whether the fCGM-sSFR correlation persists in simulations with heating-only (preventative) feedback or with AGN feedback disabled, would be needed to substantiate the causal interpretation.","section":"Sections 3, 4, and 6"},{"comment":"The claim of 'consensus from these state-of-the-art simulations that the expulsion of efficiently-cooling gas from the CGM is a crucial step' overstates the evidence. The two simulations show different mechanisms (fixed-efficiency AGN heating in EAGLE versus a calibrated kinetic-mode pivot mass in TNG), different fCGM(M200) relations, and different present-day correlations with BH accretion rate (Fig. 1). Direct evidence for expulsion exists only for EAGLE (in the companion paper O19); for TNG the paper presents solely present-day correlations. The abstract and summary should be softened to 'consistent with' the expulsion scenario, with an explicit reference to the Section 4.1 caveat, unless the authors add direct fluid-element tracking or a comparable test for TNG.","section":"Abstract and Section 6"}],"minor_comments":[{"comment":"There is a typographical issue in the sentence 'V olumetric net radiative cooling rates are specified in the publicly-available TNG snapshots' — the word 'Volumetric' has a stray space.","section":"Section 2.4"},{"comment":"The phrase 'the sub-panels here confirm' is vague; specifying 'the lower panels' would help the reader locate the relevant panels.","section":"Section 3, Fig. 1"},{"comment":"The notation ρ′′ is used in the figure captions before it is defined in the main text; define it at first use in the text.","section":"Section 5, Figs. 6 and 7"},{"comment":"The expression for E_AGN in EAGLE includes the contribution from BH seed mass, which is acknowledged, but this limitation is easy to miss; consider stating it explicitly in the text near the equation.","section":"Section 2.5, Eq. (6)"}],"recommendation":"major_revision","confidential_remarks":"The correlational analysis is solid and likely of interest to the MNRAS readership. The main issue is the gap between the title-level mechanistic claim (expulsion of efficiently-cooling gas) and the evidence presented in the paper itself, which is largely correlational and relies on a companion paper for the direct demonstration. The authors could address this either by adding direct evidence (e.g., fluid-element tracking in TNG or a controlled feedback-variant test) or by substantially toning down the causal language in the abstract and summary. The paper is within scope and the results are worth publishing once the claims are brought in line with the evidence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a well-executed simulation comparison that confirms the EAGLE f_CGM–galaxy correlation results in IllustrisTNG and explains why the two simulations differ. Read it for the correlational analysis; the title's causal claim outruns the evidence.\n\nWhat is actually new: D19 and O19 had already reported the f_CGM–M_BH and f_CGM–sSFR correlations in EAGLE, including time-resolved expulsion in the companion paper. This paper adds the TNG confirmation, which is a genuine result—two independent codes with different hydrodynamics and subgrid AGN treatments produce the same scatter correlations at fixed halo mass. It also shows the negative f_CGM–concentration correlation in both simulations, and attributes the differing mass scales to EAGLE's accretion-rate-driven feedback versus TNG's mass-threshold switch to kinetic AGN. The analysis is careful: matched collisionless runs, Spearman rank statistics with significance masks, and a 0.3r200 robustness check. The CGM was not a calibration target in EAGLE, so those correlations are emergent in a meaningful sense.\n\nSoft spots: the paper is honest about the main one. Section 4.1 explicitly concedes that no fluid-element tracking was done, and that the cooling-time contrast in Fig. 3 is consistent with density differences rather than specifically with expulsion of efficiently-cooling gas. Thus the selectively-expulsive mechanism is directly shown only in the EAGLE snipshots of O19; for TNG it is inferred. The abstract's 'consensus from these state-of-the-art simulations' therefore sits slightly ahead of the evidence. A second, subtler issue is that the causal chain depends on the fidelity of subgrid AGN feedback; if the TNG kinetic-mode pivot is a calibrated parameter, the cross-simulation agreement could be more sculpted than the text suggests. Neither flaw undermines the correlational results, which are solid.\n\nWho it is for: this is a useful citable paper for galaxy formation specialists working on CGM–galaxy connections and feedback implementation. It deserves a serious referee—the statistics are sound and the cross-simulation comparison is worth having—but the authors should either soften the causal phrasing in the title and abstract or supply TNG time-resolved expulsion evidence.","headline":"Solid cross-simulation confirmation of the f_CGM correlations with galaxy properties; the causal 'expulsion' mechanism is honestly flagged as inferred, not demonstrated.","tokens_in":36827,"tokens_out":3136,"would_cite":true,"duration_ms":30624,"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":"Two major galaxy simulations agree that AGN feedback quenches galaxies by expelling the circumgalactic gas that would otherwise fuel star formation.","keywords":["circumgalactic medium","AGN feedback","galaxy quenching","galaxy morphology","cosmological hydrodynamical simulations","EAGLE simulation","IllustrisTNG","specific star formation rate"],"falsifier":"Observe $\\sim L^\\star$ galaxies at fixed stellar or halo mass and measure their hot CGM mass fraction (via X-ray surface brightness or thermal Sunyaev-Zel'dovich stacking) against specific star formation rate and kinematic morphology; if gas-poor haloes do not preferentially host quenched, spheroidal galaxies, the core correlation fails. Alternatively, distinguish the two simulations' divergent predictions: IllustrisTNG predicts a strong anti-correlation between $f_{\\rm CGM}$ and present-day black-hole accretion rate (AGN luminosity) at $M_{200} \\simeq 10^{12}\\,M_\\odot$, while EAGLE predicts none; high-resolution X-ray observations of CGM around galaxies with known AGN activity would settle which feedback channel is physical.","tokens_in":35747,"feed_emoji":"🌌","tokens_out":5646,"duration_ms":50714,"temperature":0.7,"pith_summary":"Using two leading cosmological hydrodynamical simulations, this paper argues that the expulsion of efficiently-cooling circumgalactic gas by black-hole feedback is a crucial step in shutting down star formation and reshaping the morphology of central galaxies. At a fixed halo mass, the fraction of gas in a halo's circumgalactic medium ($f_{\\rm CGM}$) varies widely, and in both simulations that scatter tracks galaxy behaviour: gas-poor haloes preferentially host quenched, spheroidal galaxies, while gas-rich haloes host star-forming discs. These correlations are strongest near the $L^\\star$ mass scale, where AGN feedback becomes efficient. The paper concludes that feedback raises the CGM cooling time, preventing that gas from accreting to fuel later star formation, and that this mechanism is common to both simulations even though the details of the feedback differ.","feed_headline":"AGN feedback quenches galaxies by expelling their circumgalactic gas","feed_subtitle":"In two independent simulations, gas-poor haloes consistently host quenched, spheroidal galaxies near L* — a clear, testable signature.","key_machinery":"The load-bearing object is the circumgalactic medium mass fraction $f_{\\rm CGM} \\equiv M_{\\rm CGM}/M_{200}$, the mass of non-star-forming gas within $r_{200}$ divided by halo mass, and the scatter about the median $f_{\\rm CGM}$--$M_{200}$ relation at fixed halo mass. The paper measures how residuals in that scatter correlate with residuals in black-hole mass, specific star formation rate, kinematic morphology ($\\kappa_{\\rm co}$, the fraction of stellar kinetic energy in co-rotation), and CGM radiative cooling time $t_{\\rm cool}$. The mechanism carrying the argument is feedback-driven expulsion: AGN feedback ejects the efficiently-cooling component of the CGM, raising the remaining gas's cooling time, so that gas that would have replenished the interstellar medium is no longer available, starving the disc of fuel and leaving it vulnerable to disruption into a spheroid.","core_discovery":"The central claim is that in both EAGLE and IllustrisTNG, the scatter in CGM mass fraction at fixed halo mass is not noise: it encodes whether the central galaxy is star-forming and disc-like or quenched and spheroidal, and this connection is causal, running through AGN feedback. In both simulations, haloes whose central black hole has injected more feedback energy relative to the binding energy of halo baryons have lower $f_{\\rm CGM}$, elevated CGM cooling times, and central galaxies with lower specific star formation rates and less rotational support. The paper further claims that haloes that collapsed early and are more concentrated at fixed mass produce more efficient BH feedback --- in EAGLE because the BH reaches high accretion rates sooner, in IllustrisTNG because it crosses the calibrated pivot mass for switching from thermal to kinetic AGN injection sooner. The shared conclusion is that expulsion of rapidly-cooling CGM gas is a crucial step in producing quenched, early-type galaxies.","pith_inferences":["If the causal chain is right, observable CGM tracers such as O VI and C IV column densities should anticorrelate with quenched fraction at fixed stellar mass, turning $f_{\\rm CGM}$ into a practical observational proxy for the feedback state of a halo.","The two simulations disagree on when and how the expulsion happens (early, accretion-rate-driven in EAGLE; later, kinetic-mode-driven in IllustrisTNG), so the real behaviour may lie between; a simulation that tracks individual gas parcels through feedback events would test the expulsion assumption directly.","Assembly bias could imprint on galaxy morphology through this channel: early-forming haloes would not only be more concentrated but also preferentially host ellipticals, a connection that could be probed in large surveys by stacking gas fractions against large-scale environment.","The morphological link may be indirect --- expulsion mainly removes the fuel that would rebuild discs --- which predicts that morphological transformation lags quenching, and that post-quenching disc regrowth is suppressed."],"forward_implications":["At fixed halo mass, low $f_{\\rm CGM}$ is a marker for quenched, weakly-rotating galaxies and high $f_{\\rm CGM}$ for star-forming discs, with the strongest separation near $L^\\star$.","Haloes that formed early and are intrinsically more concentrated at a given mass are predicted to have lower CGM mass fractions and more quenched centrals, an assembly-bias signature.","IllustrisTNG predicts a strong present-day anti-correlation between $f_{\\rm CGM}$ and black-hole accretion rate (AGN luminosity) near $M_{200} \\simeq 10^{12}\\,M_\\odot$, while EAGLE predicts no such correlation; future X-ray or Sunyaev-Zel'dovich observations of hot CGM can discriminate between these scenarios.","In both simulations the CGM cooling time is systematically longer in gas-poor haloes, and galaxies with longer CGM cooling times have lower specific star formation rates and less rotational support.","The expulsion of efficiently-cooling CGM operates across the whole sampled halo mass range, not only at $L^\\star$."],"supporting_citations":[{"why":"Established the negative correlation between $f_{\\rm CGM}$ and central black-hole mass at fixed halo mass in EAGLE, which this paper extends to IllustrisTNG.","marker":"D19"},{"why":"Used high-cadence snipshot outputs to show that BH feedback episodes expel CGM gas and lower $f_{\\rm CGM}$; this paper imports that expulsive mechanism for EAGLE.","marker":"O19"},{"why":"Showed that AGN feedback becomes the primary self-regulation mechanism once a hot CGM forms, underpinning the feedback-energetics interpretation.","marker":"Bower et al. 2017"},{"why":"Showed in EAGLE that more gas leaves the halo than the galaxy, supporting the view that circumgalactic gas is entrained and expelled by outflows.","marker":"Mitchell et al. 2019"},{"why":"Provided the precedent that AGN feedback preferentially expels low-entropy gas, analogous to the efficiently-cooling-gas expulsion argued here.","marker":"McCarthy et al. 2011"},{"why":"Independently concluded that IllustrisTNG galaxies quench when kinetic-mode BH feedback energy exceeds the binding energy of gas, supporting the kinetic-AGN link.","marker":"Terrazas et al. 2019"},{"why":"Describes the IllustrisTNG kinetic AGN feedback implementation and the calibrated pivot mass, which the paper argues drives the TNG-specific expulsion timing.","marker":"Weinberger et al. 2017"}],"fun_headline_variants":["Gas-poor haloes host quenched spheroids in two simulations","AGN expels halo gas to quench galaxies and shape morphology","Why quenched galaxies sit in gas-poor haloes: AGN feedback","Simulations agree: AGN-driven CGM expulsion quenches galaxies","CGM gas expulsion links AGN feedback to galaxy quenching"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes that the scatter in CGM gas fraction at fixed halo mass is caused by AGN feedback actually expelling gas, rather than being a by-product of galaxy quenching, of the halo's accretion history, or of the particular subgrid feedback recipes; the paper itself notes that it does not track fluid elements to verify the expulsion in either simulation.","fun_headline_variants_meta":{"raw":{"variants":["Gas-poor haloes host quenched spheroids in two simulations","AGN expels halo gas to quench galaxies and shape morphology","Why quenched galaxies sit in gas-poor haloes: AGN feedback","Simulations agree: AGN-driven CGM expulsion quenches galaxies","CGM gas expulsion links AGN feedback to galaxy quenching"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000838,"raw_usage":{"total_tokens":3727,"prompt_tokens":1095,"completion_tokens":2632,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":711,"completion_tokens_details":{"reasoning_tokens":2539}},"tokens_in":711,"tokens_out":2632,"duration_ms":17102,"temperature":1.0,"reasoning_tokens":2539,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T10:16:48.178854+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Observe $\\sim L^\\star$ galaxies at fixed stellar or halo mass and measure their hot CGM mass fraction (via X-ray surface brightness or thermal Sunyaev-Zel'dovich stacking) against specific star formation rate and kinematic morphology; if gas-poor haloes do not preferentially host quenched, spheroidal galaxies, the core correlation fails. Alternatively, distinguish the two simulations' divergent predictions: IllustrisTNG predicts a strong anti-correlation between $f_{\\rm CGM}$ and present-day black-hole accretion rate (AGN luminosity) at $M_{200} \\simeq 10^{12}\\,M_\\odot$, while EAGLE predicts none; high-resolution X-ray observations of CGM around galaxies with known AGN activity would settle which feedback channel is physical.","supporting_citations":[],"review_version":1}