{"id":"747ddd22-782a-41f1-85e1-e6a78145b01c","arxiv_id":"2412.09526","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"In FLAMINGO simulations, gas density is biased low by up to about 8% on gigaparsec scales due to star formation, while gas velocities are unbiased on large scales and AGN feedback shapes outflows to tens of virial radii.","lead":"This paper uses the FLAMINGO simulations to show that cosmic gas is less clustered than dark matter by 4-5% on enormous scales because star formation removes gas from dense regions, while gas velocities match dark matter on those scales. It also shows that AGN feedback flings gas out to ten times a halo's radius, and that different feedback models leave different outflow signatures that future kSZ observations could detect.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The z≈1 gas-bias peak of ~8% is measured at k=0.01 h/Mpc, essentially the fundamental mode of the 681 h^-1 Mpc box, with no error bars or large-box check at that redshift; the quantitative central claim is not yet secured against sample variance.","rationale":"After reading the paper in good faith, I find the principal mechanism for the large-scale gas bias—star formation removing gas from clustered haloes—convincing. The no-cooling run has no bias, models with a lower stellar mass function show a smaller bias, and the bias is nearly independent of AGN feedback strength, all of which point to star formation rather than feedback as the cause. The z=0 result is backed by the L=2800 Mpc convergence check. The 'gas velocities unbiased' part is also supported by the velocity-divergence ratio and by the velocity-swap test in Fig. 5, which shows the large-scale pairwise-velocity suppression is a density-weighting effect. I do not think the jet-run convergence problem flagged by the reader is the most load-bearing issue for the central claim: the large-scale gas bias is obtained from thermal runs, and the jet-specific comparison mainly affects the outflow-implementation results in Section 4.2. The weakest spot in the central claim is statistical: the z≈1 8% amplitude is measured at the largest scales of a single 681 h^-1 Mpc realization, where cosmic variance is severe, and no error bars are given anywhere in Figs. 1, 2, or 4. Because the abstract advertises the 8% number, this needs a variance estimate or a large-box check at that redshift. This does not overturn the paper's qualitative conclusions, so I keep the reader's CONDITIONAL verdict rather than moving to ACCEPT or REJECT.","tokens_in":25850,"tokens_out":11085,"duration_ms":110252,"concrete_test":"Compute the gas-to-gravity-only power ratio at k=0.01 h/Mpc at z≈1 in the L=2800 Mpc FLAMINGO run (same resolution and subgrid model as the fiducial run) and compare it with the L=681 Mpc value. If the large box does not reproduce the ≈8% suppression to within ~2%, the z≈1 peak is likely a finite-volume/sample-variance artifact and the abstract should be revised to a z=0-only amplitude. If the z≈1 snapshot of the large box is unavailable, split the 681 Mpc box into octants and bootstrap the ratio at k=0.01 h/Mpc to quantify the mode-count uncertainty.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3.1.1 (right panel of Fig. 1) presents the redshift evolution of the gas bias at k=0.01 h/Mpc and the abstract quotes a maximum suppression of ≈8% at z≈1. In the L=681 h^-1 Mpc box, the fundamental mode is k_f≈0.0092 h/Mpc, so the z≈1 point is estimated from only a handful of modes in a single realization. At z=0 the 5% suppression is checked with the L=2800 Mpc run (thin green line in Fig. 1, middle panel), but no analogous large-box verification is shown at z≈1, and the paper provides no error bars for any power-spectrum or velocity ratio. A ~8% difference at one mode with roughly 30-40% cosmic variance is not statistically secure by itself; the convergence in the larger box at z=0 does not automatically validate the z≈1 amplitude, since the mode count and growth factors differ. This is load-bearing because the strongest claim includes the 8% peak, not just the existence of a low-z bias.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"Using the FLAMINGO simulation suite, the paper compares gas density and velocity statistics in full-physics runs against gravity-only runs, quantifying how galaxy formation and AGN feedback modify the cosmic gas field. The central claims are that on large scales (k ≲ 0.01 h/Mpc) gas is anti-biased relative to dark matter — a roughly 5% power-spectrum suppression at z=0, rising to about 8% near z≈1 — while gas velocities are unbiased on those scales; the bias is interpreted as a consequence of star formation removing gas from the densest, most clustered regions. On smaller scales, feedback suppresses gas clustering, potential flows, and pairwise velocities, and the paper associates these effects with AGN-driven outflows around group and cluster haloes, highlighting differences between thermal and jet AGN feedback implementations. The paper closes with observational implications for tSZ, kSZ, and weak-lensing analyses.","tokens_in":26185,"tokens_out":5723,"duration_ms":55352,"significance":"If the main results hold, this is a valuable quantitative mapping of gas density and velocity bias in a large, observationally calibrated hydrodynamical suite, directly relevant to kSZ/tSZ and lensing analyses. The use of the non-radiative control run, the z=0 large-box convergence check, the decomposition of pairwise velocities into density and velocity contributions, and the cross-comparison with BAHAMAS are clear strengths, and the quantities studied are not calibration targets of the FLAMINGO suite, so the conclusions have independent content. However, the headline 8% peak at z≈1 is not yet secured by error estimates or a large-volume check, and the thermal-versus-jet comparison is weakened by the admitted non-convergence of the jet simulations.","major_comments":[{"comment":"The abstract and Sec. 3.1.1 report a maximum gas bias of about 8% at z≈1, measured at k=0.01 h/Mpc. In the L=681 h^-1 Mpc box this wavenumber is essentially the fundamental mode (k_f≈0.0092 h/Mpc), so the z≈1 point is estimated from a handful of modes in a single realization, and no error bar is provided. The z=0 value is checked with the L=2800 Mpc run, but no analogous check is shown at z≈1. Because this peak is a headline quantitative claim, please add mode-count or jackknife error bars, or better, a large-volume or multi-realization estimate at z≈1, and adjust the abstract and summary statements if the peak is not robust.","section":"Sec. 3.1.1 / Fig. 1 (right panel); abstract"},{"comment":"Footnote 1 states that the two jet runs have large-scale power spectra that do not converge and attributes this to an unexplained behaviour, yet raw jet data are displayed in Figs. 2, 4, and 5 and are used in the thermal-versus-jet outflow comparison of Fig. 9. The Section 4.2 conclusion that the outflow radial profile and halo-mass scaling depend on the AGN feedback implementation rests on these runs. The admitted artifact therefore needs to be corrected, shown not to affect the outflow statistics, or the jet-related conclusions must be explicitly qualified as provisional.","section":"Sec. 4.2 / Fig. 9; Footnote 1"},{"comment":"The pairwise-velocity ratio R_omega is quoted as converging to a constant 2–3% bias on large scales, and Fig. 5 uses this to conclude that the density field, not the velocity field, produces the bias. No uncertainties are reported for R_omega or for S_delta_delta and S_theta_theta in Figs. 1–2, so it is not possible to tell whether the large-scale velocity contribution is consistent with zero or with a few-percent velocity bias. Please report errors, for example jackknife over subvolumes, at least for the representative models shown in Fig. 4.","section":"Sec. 3.3 / Figs. 4 and 5"}],"minor_comments":[{"comment":"Writing the power-spectrum suppression as 'b2≈5%' is potentially confusing because the plotted quantity is P_gas/P_DM = b^2; a 5% suppression corresponds to a linear bias b≈0.975. Please define the notation explicitly.","section":"Abstract and Sec. 3.1.1"},{"comment":"The L=2800 Mpc run used in the middle panel of Fig. 1 is mentioned only in the caption; please specify its box size, resolution, calibration model, and the corresponding run name in Sec. 2.","section":"Sec. 2 and Fig. 1 caption"},{"comment":"The error bars in Fig. 9 are standard deviations over 100 haloes; for comparing model differences, the uncertainty in the mean would be more appropriate, since the standard deviation is dominated by halo-to-halo scatter.","section":"Fig. 9"},{"comment":"The notation 'm_gas^2/m_b^2' in Fig. A.1 is confusing: the dotted curves appear to show (M_gas/M_bar)^2 as a function of redshift, while the text says 'total gas mass relative to the total baryon mass squared'. Please make the normalization explicit and consistent in the caption and text.","section":"Appendix A"},{"comment":"The text states that the velocity-divergence power spectrum is converged only for k < 0.6 h/Mpc, but Fig. 2 appears to extend to larger k; please indicate in the caption which parts of the curves should be regarded as converged.","section":"Sec. 3.2 and Fig. 2"}],"recommendation":"major_revision","confidential_remarks":"The main risk is not circularity, since the gas bias and outflow properties are not calibration targets of FLAMINGO. The two load-bearing issues are the statistical robustness of the z≈1 gas-bias peak and the interpretation of the jet runs in view of the admitted large-scale non-convergence. I would ask the authors to add uncertainties, to provide a z>0 large-volume or mode-count based estimate, and to make the jet comparisons provisional unless the convergence issue is addressed. With those changes the paper would be a solid contribution to A&A."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Main take: the large-scale anti-bias result is probably real, but the headline 8% at z≈1 is not yet secured, and the jet/thermal comparison rests on runs the authors admit are not converged. Still, this is a solid, useful paper that deserves a referee.\n\nWhat's new: the FLAMINGO-scale quantification of gas density and velocity statistics, the decomposition of the pairwise-velocity bias into density versus velocity contributions via velocity swapping, and the thermal-versus-jet outflow comparison. The internal checks are good: the L=2800 Mpc run confirms the z=0 bias, the non-radiative run behaves as expected, and the linear-theory residual matches CLASS. The authors also cite prior work (Shaw, Park, Kuruvilla, Kwan) honestly and do not oversell novelty.\n\nSoft spots: the z≈1 8% peak is measured at k=0.01 h/Mpc, essentially the fundamental mode of the 681 Mpc box. There are no error bars and no large-box check at that redshift, so that single point carries maybe 30-40% cosmic variance. The z=0 convergence check does not automatically validate the z≈1 amplitude; this should be flagged as a caveat, not treated as a fatal flaw, because the existence of a low-z bias is well supported. The jet runs have an unexplained large-scale non-convergence (footnote 1), and the authors display raw data anyway. That weakens the quantitative jet-versus-thermal outflow comparison in Section 4.2 and Figure 9; they need to address this before publication. Minor: power-spectrum and velocity ratios lack error bars throughout, and the outflow selection criteria are admittedly sensitive to choices, though the trends are robust.\n\nWho this is for: people modeling kSZ, tSZ, pairwise velocities, or baryonic effects on large-scale structure. They will get a useful reference for gas bias amplitudes and a clear physical interpretation. I would send it to a serious referee, with my own verdict conditional on the jet convergence and error-bar issues being addressed.","headline":"Solid and useful quantification of the large-scale gas anti-bias in FLAMINGO, but the headline 8% peak at z≈1 and the jet-versus-thermal comparison need more work before they are trusted.","tokens_in":26641,"tokens_out":1672,"would_cite":true,"duration_ms":16270,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["98.80.-k","95.35.+d"],"model":"deepseek-v4-flash","headline":"Gas underclusters dark matter by up to 8 percent, even on gigaparsec scales.","keywords":["gas density field","gas velocity field","large-scale structure","baryonic feedback","AGN feedback","power spectrum","kinetic Sunyaev-Zeldovich effect","FLAMINGO simulation"],"falsifier":"Measure the gas momentum field around stacked group-scale haloes with kinetic Sunyaev-Zeldovich observations at separations above about 10 $h^{-1}$ Mpc: if no 5–8% gas power suppression peaking near z≈1 appears, or if a higher-resolution resimulation removes the suppression at k=0.01 h/Mpc, the star-formation-depletion explanation for the large-scale gas bias would be falsified.","tokens_in":25661,"feed_emoji":"🌌","tokens_out":6381,"duration_ms":62907,"temperature":0.7,"pith_summary":"Using the FLAMINGO cosmological-hydrodynamical simulations, this paper compares the gas density and velocity fields with a gravity-only universe. It argues that on large scales the gas is an anti-biased tracer of matter: its clustering is suppressed by about 5% at z=0 and up to 8% at z≈1, while its velocity field is identical to that of dark matter. The suppression is attributed to star formation consuming gas in the densest, most clustered regions, leaving gas preferentially in lower-density environments. On smaller scales, AGN feedback suppresses both gas clustering and infall, and the paper traces this to outflowing bubbles that reach up to about ten virial radii. Establishing this bias matters because cosmological probes that use gas or its baryons as tracers—weak lensing and thermal or kinetic Sunyaev-Zeldovich signals—must account for a non-trivial gas-to-matter mapping.","feed_headline":"Gas underclusters dark matter by up to 8 percent","feed_subtitle":"Star formation strips gas from dense regions; FLAMINGO shows velocities still track dark matter.","key_machinery":"The argument runs on ratio power spectra Sδδ(k)=Pδδ/Pδδ,GrO and Sθθ(k) for gas versus gravity-only runs, evaluated at k=0.01 h/Mpc for the large-scale bias and across k for scale-dependent suppression. To study haloes, the paper stacks spherically averaged density, enclosed-mass, and radial-velocity profiles of cross-matched, isolated haloes, and defines outflow regions kinematically as grid cells where gas has positive radial velocity while dark matter has negative radial velocity. A linear-theory decomposition of pairwise velocities, ω(r)∝∫ k√(Pδδ Pθθ) j1(kr) dk with measured suppression factors inserted, separates density-field effects from velocity-field effects and identifies the density weighting as the source of the large-scale pairwise-velocity bias.","core_discovery":"The paper's central claim is that cosmic gas does not trace the dark matter field even in the linear, gigaparsec regime: the gas mass power spectrum falls short of the gravity-only matter power spectrum by a constant 4–8%, with the deficit peaking near z≈1. The mechanism is that star formation removes mass from the most biased, highest-density sites of the gas field, transforming it into a stellar component that is clustered more strongly than matter and leaving the remaining gas anti-biased. The same simulations show no corresponding large-scale velocity bias: gas velocities agree with dark matter velocities, and the apparent 2–3% suppression of large-scale pairwise velocities is a density-weighting artifact rather than slower infall. On scales k>0.1 h/Mpc, both density and velocity power are suppressed in a way that scales with AGN feedback strength, and the paper attributes this to outflows—gas bubbles with positive radial velocity that reach several to ten virial radii and reshape the gas around group-size haloes.","pith_inferences":["If the large-scale gas bias is as generic as FLAMINGO suggests, kinetic Sunyaev-Zeldovich power-spectrum analyses should include a gas bias parameter on top of the halo bias; the paper notes that pairwise kSZ measurements scale with the square root of the suppression, reducing but not removing the effect.","A direct observational test could search for the predicted peak in gas anti-bias near z≈1 using kSZ or line-intensity maps; a monotonic trend with redshift would challenge the star-formation-depletion mechanism.","The explanation implies that gas bias should track the evolution of the stellar mass function rather than AGN feedback, so future parameter variations that change only star formation or supernova feedback could sharpen the prediction.","The outflow definition used here could be applied to reconstructed velocity maps around groups once kSZ imaging reaches sufficient resolution, turning outflow bubbles into an observable proxy for AGN feedback strength and implementation."],"forward_implications":["Large-scale gas clustering is biased low by roughly 5% at z=0 and up to 8% at z≈1, independent of AGN feedback strength but dependent on the stellar mass function, while total baryons still trace matter.","Gas velocities are unbiased relative to dark matter on large scales, so the large-scale pairwise-velocity suppression seen in gas is a weighting effect of the biased density field rather than slower infall.","On scales k>0.1 h/Mpc, AGN feedback suppresses both gas density and velocity power, with stronger feedback producing stronger suppression.","Outflows, defined as gas with positive radial velocity against infalling dark matter, reach up to roughly ten virial radii, show a biconical structure, and carry more mass and higher velocity as feedback strength increases.","Thermal and jet AGN implementations calibrated to the same observables differ in their outflow footprint: jets redistribute gas to larger radii but are less efficient at lowering central baryon fractions in lower-mass haloes."],"supporting_citations":[{"why":"Supplies the FLAMINGO suite, its resolution, box size, hydrodynamics, and the two AGN feedback implementations used throughout.","marker":"Schaye et al. 2023"},{"why":"Describes the Gaussian-process calibration that fixes feedback strength from cluster gas fractions and stellar mass functions.","marker":"Kugel et al. 2023"},{"why":"Gives the prior finding that gas pairwise velocities are suppressed on large scales, which this paper reinterprets via density weighting.","marker":"Kuruvilla et al. 2020"},{"why":"Provides the BAHAMAS radial-velocity profiles and the no-velocity-bias conclusion that the paper compares against and refines.","marker":"Kwan et al. 2024"},{"why":"Established large-scale gas anti-bias in earlier simulations, which this work confirms on gigaparsec scales.","marker":"Park et al. 2018"},{"why":"Supplies the linear-theory baryon-dark matter clustering offset used to interpret the non-radiative large-scale expectation.","marker":"Angulo et al. 2013"},{"why":"Provides the shot-noise correction formula and halo-model context for the gas power-spectrum analysis.","marker":"Mead et al. 2020"},{"why":"Documents convergence of baryonic suppression and notes the post-processing correction for jet runs, informing the large-box consistency check.","marker":"Schaller et al. 2024"}],"fun_headline_variants":["Gas underclusters dark matter by 8% at large scales","FLAMINGO: gas density suppressed, velocities match dark matter","Star formation strips gas, creating 8% density deficit","Gas velocity field unbiased, density field biased low","AGN feedback drives gas outflows to ten virial radii"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The comparison between thermal and jet AGN feedback assumes the two implementations are numerically converged and differ only in how energy is injected, an assumption the paper itself weakens by admitting that the jet runs' large-scale power spectra do not converge.","fun_headline_variants_meta":{"raw":{"variants":["Gas underclusters dark matter by 8% at large scales","FLAMINGO: gas density suppressed, velocities match dark matter","Star formation strips gas, creating 8% density deficit","Gas velocity field unbiased, density field biased low","AGN feedback drives gas outflows to ten virial radii"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000418,"raw_usage":{"total_tokens":2203,"prompt_tokens":1045,"completion_tokens":1158,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":661,"completion_tokens_details":{"reasoning_tokens":1074}},"tokens_in":661,"tokens_out":1158,"duration_ms":9520,"temperature":1.0,"reasoning_tokens":1074,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T16:57:46.838887+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the gas momentum field around stacked group-scale haloes with kinetic Sunyaev-Zeldovich observations at separations above about 10 $h^{-1}$ Mpc: if no 5–8% gas power suppression peaking near z≈1 appears, or if a higher-resolution resimulation removes the suppression at k=0.01 h/Mpc, the star-formation-depletion explanation for the large-scale gas bias would be falsified.","supporting_citations":[{"cited_title":"G., & Salcido , J","cited_arxiv_id":null,"evidence_quote":"Provides the BAHAMAS radial-velocity profiles and the no-velocity-bias conclusion that the paper compares against and refines."},{"cited_title":"A., & Bond, J","cited_arxiv_id":null,"evidence_quote":"Established large-scale gas anti-bias in earlier simulations, which this work confirms on gigaparsec scales."},{"cited_title":"J., Tr \\\"o ster , T., Heymans , C., Van Waerbeke , L., & McCarthy , I","cited_arxiv_id":null,"evidence_quote":"Provides the shot-noise correction formula and halo-model context for the gas power-spectrum analysis."}],"review_version":1}