{"id":"597b4d8b-e297-486b-9e83-aee7fc374571","arxiv_id":"2502.03679","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"In IllustrisTNG, the size-based differential clustering of galaxies is caused mainly by satellite galaxies, while central galaxies alone show almost no such signal.","lead":"The paper uses the IllustrisTNG computer simulations to ask why some galaxies of the same mass are bigger than others. It finds that the simulated difference in how strongly big and small galaxies cluster is driven by satellite galaxies, not by halo spin or shape, matching a proposed explanation for real observations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Satellite attribution rests on under-resolved TNG300 satellites; the paper's own Secs. 4.2-4.3 say they may not be reliable, so the central claim is not yet established.","rationale":"The paper makes a clear, testable claim: removing satellites eliminates the differential clustering signal in TNG300, which is presented as evidence that observed size-dependent clustering is a satellite effect rather than a direct size-halo property link. The central-only test is a clean approach, and the paper is commendably explicit about its caveats. However, the same caveats flag the load-bearing assumption: low-mass satellites in TNG300 may be under-resolved. The paper's own Sec. 4.2 and Sec. 4.3 state that satellite treatment may be inaccurate due to limited resolution, which directly affects the sizes and survival of the very galaxies that are claimed to drive the signal. If a resolution cut changes the result, the conclusion is numerical rather than physical. This does not require rejecting the paper; it is a condition that must be checked before the satellite attribution can be accepted. The reader's weakest assumption identifies the same concern, so I agree. The conditional verdict remains appropriate; no change in verdict is needed, but the condition should be the resolution test.","tokens_in":19113,"tokens_out":6585,"duration_ms":60943,"concrete_test":"Repeat the differential clustering measurement in TNG300 (bottom row of Fig. 5) after applying a resolution cut: keep only satellites with at least 100 dark matter particles in the subfind subhalo and at least 25 stellar particles within the stellar half-mass radius. If the large-to-small ratio w_p(large)/w_p(small) changes by more than the jackknife error relative to the fiducial result, or if the central-only result becomes inconsistent with the full-sample result, the satellite attribution is resolution-dependent and the central claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's key claim (Abstract; Sec. 4.2) is that 'nearly all' of the size-based differential clustering signal in TNG300 arises from satellite galaxies, based on Fig. 8, which shows that the large/small ratio is close to unity for centrals only. For this interpretation to hold, the sizes and abundances of low-mass satellites in TNG300 must be reliable. The paper itself states the opposite: Sec. 4.2 says 'the treatment of satellite galaxies in IllustrisTNG may be inaccurate in some respects', and Sec. 4.3 says 'satellite galaxies in the lowest stellar mass bins ... may have too few particles, or the simulation may have too low a spatial resolution, to accurately simulate important environmental processes such as tidal and ram pressure stripping.' The lowest stellar mass bin (log M* ~ 9.75-10.25) corresponds to subhalos with only ~10^2-10^3 dark matter particles after stripping, and stellar half-mass radii measured from as few as ~100 stellar particles (the sample cut is m* >= 100 m_baryon). If these satellites are artificially small or disrupted (or artificially retained), the satellite fraction difference between large and small galaxies - the physical mechanism driving the differential clustering - will be wrong. Therefore the central conclusion that the observed differential clustering is a satellite effect rather than evidence for size-halo property correlations is not robust to resolution effects.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper investigates correlations between the ratio of galaxy size to host halo virial radius (r_gal/R_vir) and halo spin, concentration, and formation time using the IllustrisTNG suite (TNG50, TNG100, TNG300) together with matched dark-matter-only simulations from which halo properties are measured. It reports essentially no correlation with spin at z=0, a positive correlation with NFW concentration only for c_NFW >~16, and a positive correlation with formation time. The paper then measures the projected two-point correlation function of large versus small galaxies in stellar mass bins in TNG300 and compares directly with SDSS measurements from Behroozi et al. (2022). The simulation qualitatively reproduces the observed scale-dependent differential clustering, but the signal nearly disappears when only central galaxies are considered, leading the authors to conclude that the differential clustering is driven primarily by satellite galaxies rather than by correlations between galaxy size and secondary halo properties.","tokens_in":19415,"tokens_out":5577,"duration_ms":53273,"significance":"If the satellite-driven interpretation is correct, the paper resolves the tension raised by Behroozi et al. (2022): size-dependent clustering need not imply a direct physical link between galaxy size and halo spin or concentration, and the Hearin et al. (2019) picture is supported. The paper provides the first direct measurement of size-based differential clustering in a cosmological hydrodynamic simulation, and it makes falsifiable high-redshift predictions for forthcoming surveys. Notable strengths are the use of halo properties from matched dark-matter-only runs to avoid baryonic contamination of halo parameters, direct comparison with observational clustering measurements, and the use of jackknife error estimates. The results are emergent measurements from an existing simulation suite and are not obtained by fitting model parameters to the target clustering signal, so circularity is not a concern.","major_comments":[{"comment":"The central claim that 'nearly all' of the size-based differential clustering arises from satellites is not robust to the resolution limitations that the paper itself acknowledges. The lowest stellar mass bin (9.75 < log M*/M_sun < 10.25) includes satellites whose stellar masses correspond to roughly 100 stellar particles, and Sec. 4.3 states that such systems may have too few particles or too low spatial resolution to accurately simulate tidal and ram pressure stripping. The central-only clustering in Fig. 8 is a null measurement with large jackknife errors, so the wording that the signal has 'almost entirely disappeared' is stronger than the data support. The authors should demonstrate resolution robustness, for example by recomputing the all-galaxy and central-only clustering in TNG100 (which has higher mass resolution, albeit smaller volume) or by removing the lowest mass bin and showing that the satellite-driven conclusion is unchanged.","section":"Sec. 4.3; Fig. 8"},{"comment":"The critical concentration threshold c ~ 16 is identified post hoc from the same data and is not supported by any statistical test. The paper does not report confidence intervals on the r_gal/R_vir versus c_NFW relation in the high-concentration regime, nor does it test whether the change in slope at c ~ 16 is significant rather than a binning artifact. The subsequent interpretation in Sec. 4.1 that the upturn is driven by stripped or splashback halos with non-monotonic mass accretion histories is plausible but not quantitatively demonstrated. The authors should add a quantitative significance analysis, e.g., Spearman correlation coefficients in concentration bins with bootstrap errors, and state clearly whether the threshold is a fitted parameter or a descriptive summary of the data.","section":"Sec. 3.2; Fig. 3"}],"minor_comments":[{"comment":"The summation over line-of-sight separations in the estimator for w_p(r) does not explicitly include the bin width Delta pi; as written, the projected correlation function is missing the usual multiplicative factor and the units are not transparent.","section":"Sec. 2.4, Eq. (2)"},{"comment":"The phrase 'at lower correlations log c_NFW <~ 1.2' should read 'at lower concentrations'.","section":"Sec. 3.2, first paragraph"},{"comment":"There is a typo in the list of B22 models: 'dicfferent' should be 'different'.","section":"Sec. 4.2"},{"comment":"The summary statement that 'central galaxies with larger radii tend to be hosted by halos of larger mass and higher concentration' is not fully consistent with Fig. 10, where at z=0-1 the concentration difference between large and small centrals appears only in the highest stellar mass bin.","section":"Sec. 5, conclusions"},{"comment":"The abstract says 'galaxy formation history and environment ... plays an important role'; the verb should agree with the plural compound subject ('play'). Also, the spelling 'bijective' with a special character should be rendered as 'bijective' for accessibility.","section":"Abstract and Sec. 2.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is timely and well written, and the central interpretation is interesting. However, the headline conclusion rests on the satellite population in TNG300 at the lowest stellar masses, where resolution is a serious concern and the paper's own caveats undercut the claim. I would want to see a resolution test (e.g., using TNG100 or removing the lowest mass bin) before accepting the satellite-driven interpretation. The post hoc c~16 threshold should also be backed by a formal significance test. These are fixable within the scope of the paper, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: this paper does something no one has done before—directly measuring size-based differential clustering in a full cosmological hydrodynamic simulation (TNG300) and comparing it to the SDSS/Behroozi et al. measurement. The central finding is that nearly all of the signal comes from satellite galaxies; when you restrict to centrals, the large/small ratio goes to unity. That is a clean, testable statement and it supports the earlier Hearin et al. suggestion, so the novelty is partly anticipated, but the direct measurement is new and valuable.\n\nThe paper also does careful work on the halo property side: it measures halo spin, concentration, and formation time from matched dark-matter-only runs, avoiding baryonic contamination, across a wider dynamic range than the zoom-in studies. The lack of spin correlation and the concentration threshold are clearly presented, and the authors offer a plausible physical explanation for the high-concentration upturn (stripped or splashback centrals). The c~16 threshold does look post hoc, but they flag it and it is a minor issue.\n\nThe real soft spot is the satellite attribution. It rests on the lowest-mass satellites in TNG300, exactly the objects the paper itself says may have too few particles and may not resolve tidal/ram-pressure stripping. If their sizes or survival are numerical artifacts, the differential clustering signal would change, possibly enough to weaken the headline claim. This is a genuine limitation, not a manufactured one, and the paper owns it in Sec 4.3. It is not fatal—the signal is qualitatively large and the Hearin mechanism is physically reasonable—but it means the central conclusion is not yet established at the 'robust' level. The central-only clustering is also noisy, as they admit, and the comparison to observations shows quantitative differences for small galaxies that are not fully explained.\n\nNo circularity: the key results are emergent simulation measurements compared to external data; no parameter fitting to the target signal. The citation pattern is fair, leaning on Hearin and B22 appropriately.\n\nWho this is for: anyone working on galaxy sizes, halo properties, or clustering constraints on the galaxy–halo connection. It is an honest, useful paper that deserves serious refereeing. I'd send it out.","headline":"Direct measurement of size-based differential clustering in TNG300, mostly a satellite effect; the interpretation is plausible but rests on the least-resolved satellites.","tokens_in":19875,"tokens_out":2878,"would_cite":true,"duration_ms":26873,"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":"In the TNG simulations, the size-based differential clustering of galaxies is produced mainly by satellite galaxies rather than by correlations between galaxy size and halo spin or concentration.","keywords":["galaxy sizes","dark matter halos","halo spin","halo concentration","satellite galaxies","galaxy clustering","differential clustering","IllustrisTNG"],"falsifier":"Repeat the large-versus-small clustering ratio at $z=0$ for central galaxies only in a higher-resolution cosmological simulation of the same volume, or with satellite sizes recomputed at higher particle resolution: if the signal remains significantly below unity, or if the satellite sizes change enough to alter the ratio, the claim that satellites drive the signal would be overturned.","tokens_in":18941,"feed_emoji":"🌌","tokens_out":7189,"duration_ms":63029,"temperature":0.7,"pith_summary":"This paper asks why galaxies of the same stellar mass come in widely different radial sizes, and whether those differences are imprinted by the dark matter halos that host them. Using the IllustrisTNG suite of cosmological simulations and matching dark-matter-only runs, it looks for correlations between the ratio of galaxy radius to halo virial radius and halo spin, concentration, and formation time. It finds that size correlates with formation time and with concentration only for very concentrated halos, but not with spin at low redshift. The central discovery is that the observed tendency for small galaxies to cluster more strongly than large ones at fixed mass appears in TNG300, yet it is not driven by these halo-size correlations: when satellites are removed, the signal largely disappears. This would resolve a reported tension between observations and simulation-based size-halo relations by pointing to satellite galaxies as the main source of the clustering difference.","feed_headline":"Satellites, not halo spin, drive size-based clustering in TNG","feed_subtitle":"TNG300 finds small galaxies cluster harder mainly because satellites are small, not because halo spin varies with size.","key_machinery":"The load-bearing objects are the ratio $r_{\\rm gal}/R_{\\rm vir}$, the projected two-point correlation function $w_p(r_p)$ computed for large and small galaxy samples, and the decomposition of that clustering into central and satellite contributions. To avoid baryonic contamination of halo properties, the analysis uses halo catalogs from dark-matter-only runs with matched initial conditions, linked to the full-physics galaxies by a bijective matching procedure. The decisive step is recomputing $w_p$ after removing satellite galaxies: the size-based clustering ratio returns near unity, which isolates satellites as the mechanism carrying the signal.","core_discovery":"The paper's central claim is that in the IllustrisTNG300 simulation, the scale-dependent differential clustering of small versus large galaxies at fixed stellar mass is produced almost entirely by satellite galaxies, and is not a direct consequence of correlations between galaxy size and secondary halo properties such as spin or concentration. At $z=0$, TNG reproduces the observed pattern, with large galaxies less clustered than small ones and a stronger signal on small scales, but nearly all of the signal vanishes when the sample is restricted to central galaxies. The correlations the paper does measure, a positive $r_{\\rm gal}/R_{\\rm vir}$-concentration trend above $c \\sim 16$ and a correlation with halo formation time, are confined to central galaxies and cannot account for the clustering pattern. The authors interpret this as support for a simple picture in which a satellite's size is set at the time its halo reached peak mass, so satellites appear small because their halos formed and stopped growing in a denser, earlier universe.","pith_inferences":["If the signal is truly satellite-dominated, size-selected clustering becomes a practical probe of satellite quenching and environmental stripping, not of halo spin, in surveys where color and size are measured together.","Because the conclusion leans on low-mass satellites in TNG300, which may be under-resolved, a natural next test is to repeat the central-only decomposition in a higher-resolution simulation or with subhalo-rich dark-matter-only models; if better-resolved satellites shift either their sizes or their abundance, the size of the satellite contribution could change.","The weak reversal seen for central galaxies at high redshift suggests that size and halo mass are not entirely independent, and this residual correlation could become measurable in larger volumes.","One could test the mechanism directly by predicting, from the same size assignment rule, the satellite size distribution and its dependence on host-centric distance, and comparing it with future wide-field imaging."],"forward_implications":["If the satellite-driven interpretation holds, the observed size-dependent clustering does not require a direct physical link between galaxy radius and halo spin or concentration, so the apparent disagreement between observations and simulation-based size-halo correlations is resolved.","The same mechanism predicts that size-based differential clustering should weaken with cosmic time and reverse in sign by $z \\sim 3$, as the central galaxy population becomes relatively more clustered.","Future measurements of size-dependent clustering from wide-area imaging surveys can test whether the signal indeed comes from the satellite population, since satellites should also show distinctive color and environmental signatures.","Empirical models that assign galaxy sizes using only the halo's peak-mass radius, as in the earlier satellite model, can reproduce the TNG differential clustering without invoking secondary halo properties."],"supporting_citations":[{"why":"Supplies the observational differential-clustering measurements and the empirical spin- and concentration-based models whose tension with observations motivates the paper.","marker":"Behroozi et al. (2022)"},{"why":"Provides the peak-mass-radius satellite interpretation that the paper's central-only test supports.","marker":"Hearin et al. (2019)"},{"why":"Previous zoom-in simulation result correlating galaxy size with halo concentration that the paper does not reproduce; the concentration-based model is built on it.","marker":"Jiang et al. (2019)"},{"why":"The bijective matching procedure used to assign dark-matter-only halo properties to full-physics galaxies.","marker":"Gabrielpillai et al. (2021)"},{"why":"Provides the comparison of TNG clustering with observations, including the point that red satellites may be over-clustered, relevant to how satellite treatment affects the signal.","marker":"Springel et al. (2018)"},{"why":"Establishes that TNG reproduces observed size-mass relations, justifying use of TNG sizes for the clustering analysis.","marker":"Genel et al. (2018)"}],"fun_headline_variants":["Satellites, not halo spin, explain size clustering in TNG","TNG simulation: galaxy clustering by size is satellite-driven","Size-based clustering in IllustrisTNG comes from satellites","Why small galaxies cluster harder: satellites, not halo spin","Differential clustering in TNG is satellite effect, not halo spin"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument depends on the assumption that low-mass satellite galaxies in TNG300 are modeled faithfully enough, in size, abundance, and clustering, that the measured satellite contribution is trustworthy rather than a numerical artifact of limited resolution.","fun_headline_variants_meta":{"raw":{"variants":["Satellites, not halo spin, explain size clustering in TNG","TNG simulation: galaxy clustering by size is satellite-driven","Size-based clustering in IllustrisTNG comes from satellites","Why small galaxies cluster harder: satellites, not halo spin","Differential clustering in TNG is satellite effect, not halo spin"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0003,"raw_usage":{"total_tokens":1789,"prompt_tokens":1058,"completion_tokens":731,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":674,"completion_tokens_details":{"reasoning_tokens":645}},"tokens_in":674,"tokens_out":731,"duration_ms":7174,"temperature":1.0,"reasoning_tokens":645,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T04:06:15.510969+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the large-versus-small clustering ratio at $z=0$ for central galaxies only in a higher-resolution cosmological simulation of the same volume, or with satellite sizes recomputed at higher particle resolution: if the signal remains significantly below unity, or if the satellite sizes change enough to alter the ratio, the claim that satellites drive the signal would be overturned.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the peak-mass-radius satellite interpretation that the paper's central-only test supports."},{"cited_title":"Galaxy Formation in the Santa Cruz semi-analytic model compared with IllustrisTNG -- I. Galaxy scaling relations, dispersions, and residuals at z=0","cited_arxiv_id":"2111.03077","evidence_quote":"The bijective matching procedure used to assign dark-matter-only halo properties to full-physics galaxies."}],"review_version":1}