{"id":"2849ba49-9c3d-493e-88a3-9add27dfac2f","arxiv_id":"2412.16440","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"Simulations seeded from IllustrisTNG show CGM accretion supplies 20-50% of star-forming gas in dwarf galaxies and drives episodic SMBH accretion near 10% Eddington at z=2.","lead":"New hydrodynamical simulations of six dwarf galaxies, seeded from the IllustrisTNG cosmological run, trace how gas from the circumgalactic medium feeds star formation at redshifts 0, 1, and 2. The simulations report that CGM gas contributes 20 to 50 percent of the star-forming gas and that high-redshift dwarf black holes can accrete in bursts near 10 percent of the Eddington rate.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed redshift trends are not supported by the design: six handpicked halos, each evolved only 1.5 Gyr, with higher-z halos systematically more massive and more filament-fed, so the z=2 results may be selection rather than cosmic epoch.","rationale":"Good-faith reading: the paper presents a novel, technically involved simulation campaign: TNG50-1 initial conditions, ~100x particle splitting, GIZMO MFM hydrodynamics, and Lagrangian gas tracking to compute where the gas in dwarf disks at z=0,1,2 came from. This is a legitimate and useful approach, and the paper deserves credit for using realistic cosmological environments rather than idealized disks and for comparing with FIRE/TNG results. The particle-tracking method in Section 4.2 is appropriate for reservoir-origin fractions, and the SMBH duty-cycle finding is interesting. However, the headline claim of 'coevolution ... across cosmic time' is not established by the actual experiment. Six halos, two per redshift, evolved for 1.5 Gyr from fixed initial redshifts, is a cross-section, not a time sequence; the redshift axis is entangled with halo mass, environment, and assembly state (Table 1: z2b 9.63e10 Msun filament-nexus vs z0a 2.84e10 isolated). The same 20-50%/40-70% numbers and the ~10% Eddington episodes could plausibly be reproduced by selecting massive, gas-rich, merger-prone dwarfs at any redshift. The reader identified the same weakest assumption; the conditional verdict is appropriate. The cleanest resolution is a mass-matched TNG50-1 population analysis or a long-duration descendant run, either of which would separate epoch from selection. No code/data release and no convergence tests add to the uncertainty but are secondary to the sampling issue.","tokens_in":19217,"tokens_out":9670,"duration_ms":83905,"concrete_test":"From the TNG50-1 snapshot catalogs, select at z=0, 1, and 2 all halos in a narrow virial-mass bin (e.g., 4-6e10 Msun) that are star-forming and contain a central SMBH, and compute their CGM mass fractions, gas inflow rates across R_vir, and SMBH Eddington-ratio distributions. If the redshift trends reported in Figures 6, 10, and 16 persist at fixed mass, the z=2 emphasis is an epoch effect; if they weaken or reverse, the six-halo result is selection. A direct complement is to evolve one z=2 halo for ~10 Gyr to z=0 with the same GIZMO setup and compare its final CGM and SFR to the z=0 snapshot runs, testing the 'coevolution across cosmic time' claim as a trajectory.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central quantitative claims - CGM supplies 20-50% of star-forming gas and 40-70% of disk mass, strongest at z=2, and SMBH accretion episodes near 10% Eddington at z=2 - rest on comparing six halos selected from TNG50-1 at z=0, 1, and 2, each evolved for only 1.5 Gyr (Section 2.2.1). No halo is followed from z=2 to z=0, so 'across cosmic time' is inferred from a cross-sectional comparison rather than an evolutionary sequence. The comparison is not controlled: from Table 1, z2b has M_vir = 9.63e10 Msun in a filament nexus, z2a is undergoing a minor merger, while z0a is isolated with M_vir = 2.84e10. The two z=2 halos are the most massive and most filament-fed of the six, so the reported enhancement in CGM accretion, SFR, and Eddington episodes at z=2 could track halo mass or gas supply rather than redshift. The selection in Section 2.2.1 for gas-rich, SMBH-bearing, non-merging halos further biases toward systems that will show strong accretion. The 1.5 Gyr runtime is ~10 dynamical times but less than one-sixth of the z=2-to-z=0 interval, so the paper cannot directly test coevolution of a given dwarf with its CGM. Section 4.5 flags subgrid and UV-background limitations but not this sample/redshift-confounding issue.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents six high-resolution GIZMO simulations of dwarf galaxies, initialized at redshifts z=0, 1, and 2 from TNG50-1 halos of virial mass around 2-9.6 × 10^10 M_sun, with particle splitting to ~600 M_sun baryon resolution. The authors track CGM/IGM accretion, star formation, metal enrichment, and SMBH accretion over 1.5 Gyr, and report quantitative claims that CGM accretion provides 20%-50% of star-forming gas and 40%-70% of disk gas mass, with the strongest effects at z=2, plus episodic SMBH accretion reaching ~10% of Eddington at z=2. The paper also compares the simulated dwarfs with their TNG50-1 counterparts and discusses implications for JWST observations.","tokens_in":19522,"tokens_out":3451,"duration_ms":31581,"significance":"If the quantitative claims hold, this would be a useful step toward connecting realistic cosmological initial conditions with resolved multiphase CGM physics in dwarfs. The strengths of the paper include the use of TNG50-1 initial conditions rather than idealized disks, the super-Lagrangian refinement to reach ~600 M_sun baryon mass resolution, explicit particle tracking to quantify gas provenance, and the attention to multiphase temperature/density structure and metal distribution. The episodic SMBH accretion finding at z=2 is interesting and potentially relevant to early BH growth. However, the central redshift-evolution claims rest on a small, handpicked sample with a cross-sectional design, and the paper does not currently provide a resolution or convergence study that would support the quantitative fractions.","major_comments":[{"comment":"The design confounds redshift with halo mass and environment. Table 1 shows that the z=2 halos are the most massive of the sample (z2b: M_vir=9.63e10 M_sun in a filament nexus; z2a: 6.30e10 M_sun undergoing a minor merger), while z0a has M_vir=2.84e10 M_sun and is isolated. The selection criteria also require gas-rich, SMBH-bearing, non-major-merging halos, which biases the sample toward systems with strong gas supply. Therefore the reported z=2 enhancements in CGM accretion, SFR, and Eddington episodes in Sections 3.2, 3.5, and 3.6 could largely reflect halo mass, environment, or selection rather than cosmic epoch. I request either a matched-mass comparison across redshifts, a larger statistical sample, or a clear quantitative decomposition of the redshift versus mass/environment dependence; without this, the abstract's \"across cosmic time\" framing is not supported.","section":"Section 2.2.1 and Table 1"},{"comment":"Each halo is evolved for only 1.5 Gyr from its initial redshift, and no halo is followed from z=2 to z=0. The paper therefore presents a cross-sectional comparison of six snapshots, not an evolutionary sequence of the same dwarf and its CGM. The phrase \"coevolution across cosmic time\" in the title and abstract overstates what the simulations can constrain. The authors should either reframe the claims as a comparison of initial conditions at different epochs, or extend the runs along the merger tree, or at least explicitly state that no individual system is followed across the full redshift range.","section":"Section 2.2.1"},{"comment":"Section 4.5 states that \"the convergence of these models [star formation and SMBH accretion subgrid models] carefully verified,\" but no convergence or resolution study appears anywhere in the manuscript. Since the central quantitative claims (20-50% CGM contribution to star-forming gas, 40-70% to disk mass, and ~10% Eddington accretion episodes) depend on resolved gas structure and subgrid feedback, a quantitative resolution test is load-bearing. If the verification exists, it should be shown (e.g., a factor-of-two mass resolution comparison); if not, the sentence should be corrected to describe the current state as unverified.","section":"Section 4.5"},{"comment":"The mass accretion rates in Figure 6 are computed as the time derivative of enclosed mass within <0.2 R_vir and R_vir, which mixes inflow and outflow and is not a direct measurement of the accretion rate onto the galaxy or halo. The text interprets positive and negative values as net inflow and outflow, but this is not equivalent to the particle-tracking accretion rates in Section 4.2. The two definitions should be clearly distinguished, and the claims about \"accretion rates\" should specify which definition is being used, particularly when discussing peaks and anti-correlations.","section":"Section 3.2 and Figure 6"},{"comment":"Table 3 compares the simulated z=1 and z=2 dwarfs with their original TNG50-1 counterparts at \"the same epoch,\" noting that the runs correspond to z≈2→1.3 and z≈1→0.4. This comparison is potentially informative, but the large differences in SFR (e.g., z1a: 0.36 vs 0.0078 M_sun/yr) are attributed to resolution, cooling, and subgrid modeling without a demonstration that the initial conditions at the start of the runs are actually consistent with TNG50-1. Please clarify whether the comparison is between the end of the 1.5 Gyr evolution and the TNG halo at the same physical time, and discuss whether the differences may partly reflect the non-cosmological re-initialization (for example, the lack of incoming cosmological accretion during the run).","section":"Section 4.4 and Table 3"}],"minor_comments":[{"comment":"The sentence \"the initial conditions of our simulations taking the dwarf galaxies of 2-5 × 10^10 M_sun from the realistic cosmology simulations, IllustrisTNG\" is grammatically unclear and should be rewritten for clarity.","section":"Abstract"},{"comment":"In the sentence listing warm-phase metal fractions, the text says \"20% (z2a) and 14% (z2a)\" but the second value should presumably refer to z2b; please correct the label.","section":"Section 4.3"},{"comment":"The formula for the disk-wind energy is partially garbled as \"0.5 ˙MwV 2 w\"; use proper typesetting for the equation.","section":"Section 2.1.4"},{"comment":"There are several LaTeX artifacts in names, such as \"V oit\" (should be \"Voit\"), \"V oronoi\" (should be \"Voronoi\"), and the running header \"T UNG AND CHEN\" (should be \"TUNG AND CHEN\"). These should be cleaned before publication.","section":"Throughout"},{"comment":"The claim \"which ensures the convergence of the star formation\" is unsupported by any convergence test in the paper; either add a reference to a published resolution study or soften the statement.","section":"Section 2.2.2"},{"comment":"The table columns contain many zeros with inconsistent formatting (e.g., \"0.0\" vs \"0.000\"). Please standardize the decimal notation.","section":"Table 2"}],"recommendation":"major_revision","confidential_remarks":"This is an interesting exploratory study with realistic initial conditions and useful particle-tracking diagnostics, but the central redshift-evolution claim is not currently supported by the sample design. The paper would benefit from either a matched-mass comparison, a longer evolution along the TNG merger tree, or a clear restatement of the results as a comparison of six selected snapshots rather than a claim about coevolution across cosmic time. I would be willing to revisit after major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Useful setup, concrete numbers, but the redshift trend is overclaimed. The genuinely new bit is taking dwarf halos (2–5e10 Msun) from TNG50-1, splitting particles by ~100x in GIZMO, and using particle tagging to separate gas that started in the galaxy, CGM, or IGM. That gives numbers people will quote: CGM supplies 20–50% of the star-forming gas and 40–70% of the disk mass; z=2 SMBHs show accretion peaks near 10% Eddington. Those are useful observables for JWST work.\n\nThe soft spot is the inference, not the simulation. The title says 'Across Cosmic Time,' but no halo is followed from z=2 to z=0. Each of six halos (two per redshift) runs 1.5 Gyr from a snapshot. The z=2 halos are the most massive and best-fed—z2b is 9.6e10 Msun in a filament nexus, z2a is mid-merger—and the selection criteria favored gas-rich, SMBH-bearing, non-major-merger systems. So the reported redshift trend is entangled with halo mass and gas supply. The 20–50% and 40–70% numbers may survive as trends with gas availability, but not as shown as trends with cosmic epoch.\n\nSmaller issues: no convergence test despite the claim in Section 4.5; Table 3 likely has a decimal error (z1b Mgas ~ 1.6e6 Msun is inconsistent with the accretion rates in Table 2); no code or data. The qualitative picture—cold accretion dominant, high-z dwarfs burstier with episodic SMBH growth—is consistent with previous work, and the comparison to Hafen et al. is fair. No circularity; the fractions are emergent.\n\nWho gets value: people interpreting JWST observations of faint high-z dwarfs, or studying dwarf CGM accretion. I'd send it to review, but with heavy-revision expectations. I wouldn't cite the specific fractions yet.","headline":"Useful new simulation setup and plausible numbers, but the 'across cosmic time' claim rests on a cross-sectional, confounded sample; needs major revision before the quantitative fractions are used.","tokens_in":20125,"tokens_out":6267,"would_cite":false,"duration_ms":49946,"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":"Dwarf galaxies get 20-50% of their star-forming gas from infalling halo gas.","keywords":["dwarf galaxies","circumgalactic medium","gas accretion","redshift evolution","hydrodynamical simulations","supermassive black hole growth","multiphase gas","metal enrichment"],"falsifier":"Run the same simulation pipeline on a larger sample of dwarf halos per redshift, or extend the current runs to several gigayears: if the circumgalactic contribution to star-forming gas and disk mass falls outside 20–50% and 40–70%, or if the $z=2$ black hole outbursts disappear, the claimed redshift dependence would not generalize. A complementary observational check is a JWST survey for accreting supermassive black holes in $z\\sim2$–4 dwarfs, which should find episodic near-10% Eddington accretors if the simulation picture is right.","tokens_in":18956,"feed_emoji":"🌌","tokens_out":12093,"duration_ms":91420,"temperature":0.7,"pith_summary":"The paper claims that the circumgalactic medium—the shell of gas around a dwarf galaxy, extending to the edge of its dark-matter halo—is not a passive reservoir but an active fuel supply whose role grows with cosmic time. Using high-resolution hydrodynamical simulations of six dwarf galaxies drawn from a modern cosmological simulation at $z=2$, $z=1$, and $z=0$, it finds that gas accreted from the circumgalactic medium provides 20–50% of the star-forming gas and 40–70% of the gas mass in the galactic disk. At $z=2$ the infalling gas is even more dominant: circumgalactic and intergalactic gas together make up 60–70% of the halo's gas, driving both vigorous star formation and episodic supermassive black hole accretion near 10% of the Eddington rate. If these numbers hold, dwarf galaxies at early cosmic time grow largely from gas falling in from their surroundings rather than from gas they formed with.","feed_headline":"Dwarf galaxies get 20-50% of star-forming gas from the halo","feed_subtitle":"The same inflow rebuilds 40-70% of the disk and triggers episodic black-hole growth in the early universe.","key_machinery":"The machinery is a set of high-resolution hydrodynamical re-simulations: dwarf-galaxy halos of virial mass $2\\text{–}5\\times10^{10}\\,M_\\odot$ are cut out of the TNG50-1 cosmological volume at $z=0,1,2$ and evolved for 1.5 Gyr with a particle-splitting scheme that raises mass resolution by a factor of about 100, giving baryon cells of roughly $600\\,M_\\odot$ and gas resolution of a few parsecs. The physics package includes metal-line cooling, molecule-based star formation in self-gravitating convergent flows, supernova momentum feedback, and a subgrid black hole accretion model. The key analytic device is particle tagging: each gas parcel is labeled by its initial region and followed to its final position, which turns the simulation output into the quoted percentage contributions from galaxy, circumgalactic medium, and intergalactic medium.","core_discovery":"The central discovery is a quantified, redshift-dependent measure of how dwarf galaxies feed. By tagging each gas parcel according to where it began (galaxy, circumgalactic medium, or intergalactic medium) and following it for 1.5 Gyr, the authors show that the fraction of a dwarf galaxy's star-forming fuel that came from outside the galaxy rises with redshift, so that at $z=2$ gas originally in the galaxy contributes less than 20% of the star-forming gas while circumgalactic and intergalactic inflow supplies the rest. Circumgalactic gas alone rebuilds 40–70% of the disk gas mass at all redshifts in the sample. The same runs show supermassive black holes in $z=2$ dwarfs accreting in episodic bursts whose peaks reach about 10% of the Eddington rate, implying phases of rapid black hole growth in the early universe. The authors conclude that galactic outflows and circumgalactic/intergalactic accretion form a baryon cycle whose importance increases with redshift.","pith_inferences":["Editorial extension: because only two halos are simulated per redshift, the reported trends conflate cosmic epoch with halo mass, environment, and assembly history; re-running the same setup on several halos per redshift would isolate the redshift dependence.","Editorial extension: 1.5 Gyr is short compared with the age of the universe at $z=2$, so the 40–70% disk-replenishment fractions may be a transient response to the chosen initial conditions; longer runs or comparison with the parent cosmological simulation's own dwarf population would test their stability.","Editorial extension: the subgrid black hole accretion model could exaggerate burstiness; independent runs with different accretion prescriptions would show whether the near-10% Eddington peaks are robust."],"forward_implications":["At $z\\approx2$, most of a dwarf galaxy's star-forming gas is accreted from its surroundings, so models of early galaxy formation must include cosmological inflow rather than relying on gas present at birth.","Because circumgalactic gas replenishes 40–70% of the disk gas mass even at $z=0$, galactic disks should be treated as open systems whose gas content is continuously exchanged with the halo.","Episodic supermassive black hole accretion peaking near 10% of the Eddington rate in $z=2$ dwarfs offers a concrete path to growing massive black holes in the early universe, a signature that the James Webb Space Telescope could detect.","The simulated rise of warm-phase metal fractions in the circumgalactic medium with redshift implies that high-redshift observations of CGM metal lines should see more metal-bearing warm gas than at $z=0$, directly testing the coevolution picture."],"supporting_citations":[{"why":"Supplies the TNG50-1 cosmological simulation volume from which the dwarf-halo initial conditions are extracted.","marker":"Nelson et al. 2019"},{"why":"Describes the TNG50 simulation whose highest-resolution run provides the halos and their baryon content.","marker":"Pillepich et al. 2019"},{"why":"Provides the GIZMO hydrodynamics code and the super-Lagrangian particle splitting used to raise the resolution.","marker":"Hopkins 2015"},{"why":"Defines the meshless-finite-mass hydrodynamics and the cooling, star-formation, and feedback subgrid models adopted in the runs.","marker":"Hopkins et al. 2018a"},{"why":"Supplies the stellar-feedback implementation used to distribute supernova energy and momentum onto surrounding gas.","marker":"Kim et al. 2016"},{"why":"Gives the spherical accretion formula that sets the black hole growth rate in the simulations.","marker":"Bondi 1952"},{"why":"Provides the simulation-based circumgalactic medium baryon fraction at $z=2$ (40–70%) against which the paper checks its own values.","marker":"Hafen et al. 2019"},{"why":"Supplies the observational warm-phase CGM metal fractions at low redshift used to validate the simulated metallicity distribution.","marker":"Zheng et al. 2024"}],"fun_headline_variants":["Dwarf galaxies get most star fuel from outside by z=2","Dwarf galaxies rely on cosmic gas for over 80% of star fuel by z=2","Dwarf galaxies siphon cosmic gas, fueling black hole bursts","Extragalactic gas rebuilds dwarf disks and feeds black holes","Halo gas rebuilds 40-70% of dwarf disk mass"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the six selected dwarf halos—two at each redshift—represent typical dwarf galaxies of their epoch, and that 1.5 Gyr of re-simulation is long enough for the circumgalactic accretion, star formation, and black hole bursts that define the cosmic-time trend to develop fully.","fun_headline_variants_meta":{"raw":{"variants":["Dwarf galaxies get most star fuel from outside by z=2","Dwarf galaxies rely on cosmic gas for over 80% of star fuel by z=2","Dwarf galaxies siphon cosmic gas, fueling black hole bursts","Extragalactic gas rebuilds dwarf disks and feeds black holes","Halo gas rebuilds 40-70% of dwarf disk mass"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001452,"raw_usage":{"total_tokens":5894,"prompt_tokens":1037,"completion_tokens":4857,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":653,"completion_tokens_details":{"reasoning_tokens":4770}},"tokens_in":653,"tokens_out":4857,"duration_ms":31229,"temperature":1.0,"reasoning_tokens":4770,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T10:35:07.256555+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same simulation pipeline on a larger sample of dwarf halos per redshift, or extend the current runs to several gigayears: if the circumgalactic contribution to star-forming gas and disk mass falls outside 20–50% and 40–70%, or if the $z=2$ black hole outbursts disappear, the claimed redshift dependence would not generalize. A complementary observational check is a JWST survey for accreting supermassive black holes in $z\\sim2$–4 dwarfs, which should find episodic near-10% Eddington accretors if the simulation picture is right.","supporting_citations":[],"review_version":1}