{"id":"ca7e466e-5d79-4ff5-8a23-2a7e29680566","arxiv_id":"2610.19030","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"The stellar-to-halo mass ratio for massive galaxies peaks at z~1 and declines afterward, indicating that massive galaxy formation becomes less efficient below z~1, with halos growing but central galaxies remaining quiescent.","lead":"This paper combines three independent galaxy survey fields covering 9.3 square degrees to measure how the most massive galaxies populate their dark matter halos from redshift 0.25 to 4. The analysis reveals that galaxy formation became less efficient after z~1, with massive halos continuing to grow while their central galaxies stopped adding stars, likely due to black hole feedback.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"The N(z) model is validated only by varying a smoothing width; a direct empirical check against the ~100k spectroscopic redshifts could reveal shape discrepancies that propagate into the central SHMR evolution result.","rationale":"The reader correctly identifies the N(z) shape and its degeneracy with sigma_logM as the weakest assumption. The paper acknowledges this and takes steps to mitigate it, but does not exploit the available spectroscopic-redshift sample to empirically validate the N(z) shape. A direct spec-z check is the most concrete way to settle whether the N(z) uncertainty actually biases the central SHMR evolution result. Because the authors have already performed internal tests (varying Gaussian width) and the low-z bins are likely less affected, the concern does not warrant a verdict change (still CONDITIONAL), but it strengthens the condition: the authors should demonstrate that the spec-z N(z) is consistent with their model. The reader asked for a uniform-prior test on sigma_logM; the spec-z check is complementary and more fundamental. Hence partial agreement: the reader identifies the right area, but this test targets the root cause rather than a consequence.","tokens_in":52439,"tokens_out":3344,"duration_ms":36615,"concrete_test":"Construct an empirical N(z) for each redshift and stellar-mass bin using only the galaxies with spectroscopic redshifts (34,105 COSMOS, 49,053 XMM-LSS, 16,942 ECDF-S) by stacking their delta-function spec-z PDFs. Compare the empirical N(z) shape (mean, width, 10th/90th percentile tails) to the photo-z based N(z) used in the analysis. If the empirical N(z) differs by more than the Gaussian smoothing width [0.04(1+z)] in the wings or by more than 0.03(1+z) in the mean, re-fit the HOD model (Section 4.4) using that empirical N(z) for the affected bins. Then recompute M_peak_h and M_peak_star at z<1 (Figure 15) and check whether the increase in M_peak_h or the constancy of M_peak_star shifts by more than the reported 1-sigma posterior uncertainties.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim — that for the most massive galaxies M_peak_h increases from z=1 to z=0 while M_peak_star remains constant — depends critically on the shape of N(z) (Sections 4.3, 6.4.1, Figure 15). The authors acknowledge (Section 4.3) that misrepresentations in N(z) wings can cause 20–50% differences in omega(theta) and that at z>1.5 sigma_logM hits the prior boundary due to degeneracy with N(z). To mitigate this, they convolve N(z) with a Gaussian of width 0.04(1+z) and adopt a Gaussian prior on sigma_logM. However, they test only the smoothing width (0.02–0.05) and do not validate the N(z) shape against their extensive spectroscopic redshift sample (34k COSMOS, 49k XMM-LSS, 17k ECDF-S; Section 2.3). Using spec-z to construct an empirical N(z) — by stacking the delta-function PDFs — would directly test whether the photo-z based N(z) systematically over- or under-broadens the distribution. If the empirical N(z) has, e.g., narrower wings or a different mean, the HOD parameters — especially sigma_logM — and hence the peak SHMR evolution could shift. The low-z bins (z<1) are not immune, because even modest N(z) errors can bias the inferred M_peak_h and M_peak_star enough to affect the claimed trend (Section 4.3). A spec-z validation is the most direct way to assess whether the N(z) uncertainty is under control.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"This paper presents a galaxy clustering analysis using optical and near-infrared data from UltraVISTA and VIDEO across the COSMOS, XMM-LSS, and ECDF-S fields (total area 9.3 deg²) to constrain the halo occupation distribution (HOD) and stellar-to-halo mass relation (SHMR) for stellar-mass-selected galaxies over 0.25<z<4. The authors measure angular two-point correlation functions with photometric-redshift probability weighting, fit a five-parameter HOD model via MCMC, and derive the SHMR, satellite fractions, and galaxy bias. The central claim is that for the most massive galaxies the peak halo mass M_peak_h increases from z=1 to z=0 while the peak stellar mass M_peak_star remains approximately constant, implying suppressed stellar-mass growth in massive haloes at z<1, which they attribute to AGN feedback.","tokens_in":52909,"tokens_out":2323,"duration_ms":25942,"significance":"If the central result holds, it provides an observational constraint on the growth of massive galaxies and their haloes at late cosmic times, with implications for feedback models in galaxy formation simulations. The study is among the largest-area deep-field clustering analyses at these redshifts, and the combined three-field approach reduces cosmic variance by a factor of ~2 compared to single-field studies. The data and methodology are described in sufficient detail to allow reproduction.","major_comments":[{"comment":"The paper's central claim—that M_peak_h increases from z=1→0 while M_peak_star remains constant—depends critically on the shape of the redshift distribution N(z). The authors acknowledge (Section 4.3) that misrepresentations in N(z) wings can cause 20–50% differences in ω(θ) and that at z>1.5 the HOD parameter σ_logM hits its upper prior boundary due to degeneracy with N(z). They test the Gaussian smoothing width (0.02–0.05(1+z)) but do not validate the N(z) shape against their extensive spectroscopic-redshift sample (34k COSMOS, 49k XMM-LSS, 17k ECDF-S; Section 2.3). Because the spec-z sample exists and the authors already adjust PDFs for those galaxies, a direct empirical N(z) from the spec-z galaxies (e.g., by stacking their delta-function PDFs) would test whether the photo-z based N(z) systematically over- or under-broadens the distribution. If the empirical N(z) has different wings, the inferred HOD parameters and the evolution of M_peak_h and M_peak_star (Figure 15) could shift. This validation is needed to confirm that the N(z) uncertainty is under control for the central result.","section":"§4.3, §6.4.1, Figure 15","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Many of the reduced χ² values in Table A1 are substantially larger than 1 (e.g., 3.1, 3.5, 4.9, 6.4), indicating that the best-fit HOD models do not provide an adequate statistical description of the angular correlation function and number density data within the jackknife covariance errors. The authors do not discuss these large reduced χ² values. High reduced χ² can arise from underestimated covariance (e.g., from the jackknife with limited patches), model misspecification, or unmodelled systematics (e.g., N(z) uncertainties). The authors should discuss the origin of these high values and consider whether the covariance matrix is converged (e.g., by comparing to mock-based or analytic covariances) or whether additional freedom (e.g., a more flexible HOD) is needed to fit the data.","section":"Table A1, §4.4","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The authors note that stellar mass thresholds follow the higher completeness limit of XMM-LSS/ECDF-S fields. It would be helpful to state explicitly whether this choice excludes the deepest COSMOS data, and how the results change if COSMOS-only lower-mass bins are modelled separately.","section":"§2.5.2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The normalisation of DD, DR, and RR pairs is described textually but it would be clearer to write the weighted version of the Landy-Szalay estimator explicitly with the weights w_i.","section":"§3.1, Equation (4)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The spikes in the N(z) distributions are attributed to spectroscopic redshifts, but the figure does not distinguish between spec-z and photo-z contributions. Adding a comparison N(z) constructed only from spec-z galaxies (where available) would directly address the major comment above.","section":"§4.3, Figure 5","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The model curves for the three fields are overplotted as solid lines with different field colours. Because the curves are very similar, a separate panel showing the combined fit with the data points colour-coded by field would improve readability.","section":"§5.1, Figures 6–7","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The SHMR is computed by summing contributions from threshold samples. The authors should clarify why the median stellar mass of the bin is used for M_th*,i rather than, e.g., the mean or the lower edge, and quantify the sensitivity of the SHMR peak to this choice.","section":"§6.4, Equation (24)","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid clustering analysis with a clear central claim. The main concerns—N(z) validation against spec-z and high reduced χ² values—are addressable with existing data and further analysis. If the authors can demonstrate that the N(z) shape does not bias the peak SHMR evolution, and can explain the high reduced χ² values, the paper would be suitable for MNRAS. The scope and methodology are appropriate for the journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"The real advance here is the area: 9.3 deg2 across three independent fields gives them roughly 5x the volume of previous single-field studies, which matters for rare massive galaxies at z>2. The analysis is thorough—photo-z PDF weighting, cosmic variance treatment, jackknife covariances, MCMC convergence checks. The galaxy bias and satellite fraction trends all look consistent with the literature, so the machinery is sound.\n\nThe central result—that for the most massive galaxies M_peak_h increases from z=1 to z=0 while M_peak_star stays flat—is physically interesting and consistent with AGN feedback expectations. But the soft spot is the N(z) treatment. They test only the Gaussian smoothing width (0.02–0.05) and rely on a Gaussian prior on sigma_logM to break the degeneracy. The stress test has a point: they have ~100k spectroscopic redshifts across the three fields and could stack the delta-function spec-z PDFs to check whether their photo-z N(z) has the right shape. That would be a direct empirical check, not just a width variation. If the wings are off, the central evolution claim could shift. The authors themselves note (Section 4.3) that N(z) wings can cause 20–50% differences in omega(theta), and that sigma_logM hits the prior boundary at z>1.5—that’s a flag.\n\nOther concerns: the reduced chi-squared values are high (2–5 in many bins), which means the model isn’t capturing the data well even after accounting for covariance. And the satellite fraction at z>2.5 is effectively zero; the HOD posteriors show the satellite parameters are basically prior-dominated there. The high-redshift results rest almost entirely on the central galaxy modelling.\n\nThis is a solid data paper that deserves review. The main conclusion is likely correct, but the referee should ask for a direct spec-z N(z) comparison and a discussion of how the sigma_logM prior influences the peak M_h evolution. If those are addressed, this will be a useful reference.\n\nYes, I’d send it to a serious referee. The work is honest and the dataset is genuinely valuable.","headline":"Largest combined deep field analysis to date of the galaxy-halo connection at high mass and high redshift; central claim is plausible but N(z) validation is half-done.","tokens_in":53437,"tokens_out":1293,"would_cite":true,"duration_ms":14069,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek/deepseek-v4-flash","headline":"Massive galaxies stop growing in stars while their dark-matter haloes keep growing below redshift 1.","keywords":["galaxy evolution","dark matter haloes","halo occupation distribution","galaxy clustering","stellar-to-halo mass relation","active galactic nuclei feedback","photometric redshifts","cosmic variance"],"falsifier":"If deep spectroscopic follow-up of a representative sample of massive galaxies at z~0.5 showed that their stellar masses have increased significantly since z~1 (i.e., that M_*^peak does evolve), the central claim would be contradicted. Alternatively, if the satellite fraction at fixed halo mass were found to be constant rather than increasing from z=1 to z=0, the hierarchical-growth interpretation would weaken.","tokens_in":53953,"feed_emoji":"🌌","tokens_out":2835,"duration_ms":32521,"temperature":0.7,"pith_summary":"This paper combines optical and near-infrared data from three extragalactic fields covering 9.3 square degrees to study how massive galaxies populate dark matter haloes from redshift 0.25 to 4. By measuring the angular clustering of stellar-mass-selected galaxies and fitting a five-parameter halo occupation distribution model, the authors obtain the stellar-to-halo mass relation and its evolution. They find that for the most massive galaxies, the halo mass at which the stellar-to-halo mass ratio peaks increases from z=1 to z=0, while the corresponding peak stellar mass stays nearly constant. This indicates that in the last half of cosmic time, the stellar mass growth of the most massive central galaxies has become inefficient while their host haloes continue to grow hierarchically. The authors interpret this as evidence that feedback from active galactic nuclei – likely low-power radio jets – has shut down star formation in these galaxies.","feed_headline":"Massive galaxies' stars stall while halos keep growing","feed_subtitle":"A 9.3-sq-deg survey finds that below redshift 1, the most massive central galaxies stop adding stars as their dark-matter halos grow…","key_machinery":"The five-parameter halo occupation distribution (HOD) model of Zheng et al. (2005), describing central and satellite galaxy occupation as a function of halo mass. The model is fitted to the angular two-point correlation function measured with the Landy–Szalay estimator, using a Markov chain Monte Carlo method that simultaneously fits the number density and clustering in each of three independent survey fields. The redshift distribution N(z) is constructed from photometric-redshift likelihoods with a magnitude-dependent prior and a Gaussian broadening of 0.04(1+z). The model projects the 3D galaxy power spectrum into the angular correlation function via the Limber approximation, allowing the extraction of the stellar-to-halo mass relation and derived quantities such as galaxy bias, satellite fraction, and peak halo/ stellar masses.","core_discovery":"Over the redshift range z=0.25 to z=4, the stellar-to-halo mass ratio for the most massive galaxy population peaks at z~1. Below z=1, the peak halo mass M_h^peak increases (from ~$10^{12}$.5 to ~$10^{13}$ M_sun in the highest stellar-mass bin), while the peak stellar mass M_*^peak in those haloes remains approximately constant at ~5×$10^{10}$ M_sun. This divergence implies that from z=1 to today, massive haloes grow through hierarchical accretion but their central galaxies do not build up additional stars. The satellite fraction also rises steadily since z~1, consistent with haloes growing via accretion of smaller systems whose central galaxies become satellites. The combined evidence points to a mechanism that suppresses star formation in the most massive central galaxies at late times, and the authors attribute this to radio-mode AGN feedback.","pith_inferences":["The result provides an observational anchor for galaxy-formation simulations: any model that fails to reproduce the decoupling of stellar-mass and halo-mass growth at z<1 may be missing a key feedback process or overestimating the late-time gas accretion rate.","If the suppression is indeed due to AGN feedback, one expects that the most massive central galaxies at z<0.5 should host compact radio cores or show evidence of recent mechanical feedback in X-ray cavities; upcoming surveys with MeerKAT and the SKA can test this directly.","The combined-field approach reduces cosmic variance by a factor ~2 compared to single-field studies; future surveys such as Euclid and Rubin will be able to extend this analysis to even larger volumes and fainter stellar-mass thresholds, testing whether the effect holds for lower-mass galaxies as well."],"forward_implications":["If true, the suppression of stellar-mass growth in massive haloes below z~1 implies that AGN feedback is the primary quenching mechanism for the most massive galaxies in the nearby Universe, consistent with semi-analytic models that couple radio-jet heating to the hot gas halo.","The increasing satellite fraction with decreasing redshift suggests that halo mass assembly at z<1 occurs primarily through the accretion of pre-formed galaxies rather than through in-situ star formation in the central galaxy.","The redshift z~1 marks the epoch at which the star-formation efficiency in massive haloes peaks; above z=1 the efficiency declines because haloes are still assembling rapidly, and below z=1 it declines because star formation shuts off.","The finding motivates targeted radio observations of massive galaxies at z<1 to quantify the duty cycle of low-power radio jets and their correlation with the star-formation history of the central galaxy."],"supporting_citations":[{"why":"Provides the basic five-parameter HOD model that the paper adopts as its theoretical framework.","marker":"Zheng et al. 2005"},{"why":"Demonstrates the bias in N(z) constructed from photometric-redshift likelihoods; the paper follows their method of applying a magnitude-dependent prior.","marker":"Ilbert et al. 2021"},{"why":"A previous clustering+HOD analysis in the COSMOS field that the paper compares its results to and extends with a larger area.","marker":"McCracken et al. 2015"},{"why":"Used a similar weighted-PDF method for photometric redshifts and HOD fitting; the paper reproduces and extends those results to higher masses and redshifts.","marker":"Hatfield et al. 2016"},{"why":"Showed that a single-field COSMOS analysis can shift the inferred M_peak^h by ~0.15 dex; the paper's multi-field approach is designed to reduce this cosmic-variance bias.","marker":"Zaidi et al. 2024"}],"fun_headline_variants":["Dark halos swell as massive galaxies stop making stars","Massive galaxies' stars plateau while halos balloon after z=1","Satellite fraction rise signals stalled star formation in giant galaxies","9.3 deg² survey shows massive centrals star-starved since z=1"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The redshift distribution N(z) of each galaxy sample, constructed from photometric-redshift likelihoods with a magnitude-dependent prior and a Gaussian broadening of width 0.04(1+z), is assumed to be accurate; the authors acknowledge that misrepresentations in the shape of N(z), especially its wings, can cause up to 20–50% differences in the predicted angular correlation function, and that at z>1.5 the HOD parameter sigma_logM hits its upper prior boundary, indicating that the N(z) uncertainty is not fully resolved.","fun_headline_variants_meta":{"raw":{"variants":["Dark halos swell as massive galaxies stop making stars","Massive galaxies' stars plateau while halos balloon after z=1","Satellite fraction rise signals stalled star formation in giant galaxies","9.3 deg² survey shows massive centrals star-starved since z=1"]},"model":"deepseek/deepseek-v4-flash","effort":"low","cost_usd":7.9e-05,"raw_usage":{"cost":7.894488e-05,"is_byok":false,"cost_details":{"upstream_inference_cost":7.894488e-05,"upstream_inference_prompt_cost":4.089176e-05,"upstream_inference_completions_cost":3.805312e-05},"total_tokens":1562,"prompt_tokens":1066,"completion_tokens":496,"prompt_tokens_details":{"audio_tokens":0,"video_tokens":0,"cached_tokens":0,"cache_write_tokens":0},"completion_tokens_details":{"audio_tokens":0,"image_tokens":0,"reasoning_tokens":420}},"tokens_in":1066,"tokens_out":496,"cost_source":"openrouter_usage","duration_ms":6645,"image_pages":[],"temperature":1.0,"billed_attempts":1,"reasoning_tokens":420,"transport_provider":"openrouter","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-10-15T17:18:21.749662+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"falsifier":"If deep spectroscopic follow-up of a representative sample of massive galaxies at z~0.5 showed that their stellar masses have increased significantly since z~1 (i.e., that M_*^peak does evolve), the central claim would be contradicted. Alternatively, if the satellite fraction at fixed halo mass were found to be constant rather than increasing from z=1 to z=0, the hierarchical-growth interpretation would weaken.","supporting_citations":[],"review_version":1}