{"id":"908f1f04-6326-4ef5-a33c-b49a4a7a4ac5","arxiv_id":"2502.02239","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A catalogue of 29 morphological parameters for 12,075 eRASS1 clusters, with new slosh and multipole-magnitude shape parameters and simulations that quantify how noise, PSF and selection bias the measurements.","lead":"This paper measures 29 X-ray morphology parameters for more than 12,000 galaxy clusters from the first eROSITA all-sky survey, including two newly introduced forward-modelled shape parameters called slosh and multipole magnitudes. It also maps how survey depth, detection thresholds and cluster concentration bias what gets detected, and provides a public catalogue of the measurements.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Unresolved AGN are omitted from the simulations that set the concentration bias and selection functions; they can also directly inflate measured c500, so the 0.3 dex offset and the high-z selection threshold are quantitatively insecure.","rationale":"The reader's weakest_assumption identifies the same load-bearing concern: the simulation suite used to quantify parameter biases and selection functions omits AGN, and the authors explicitly acknowledge that adding a point-source population would likely introduce additional bias, particularly at low luminosity and high redshift. My stress-test agrees with this assessment and sharpens it: the omission affects not only the selection-function interpretation but also the measured concentration values themselves, because an unresolved central AGN adds flux within the 0.1 R500 aperture used for c500. The matched-sample comparisons in Section 7.1 provide independent evidence of a residual ~0.1 dex eRASS1 offset relative to XMM-Newton and Chandra, with the more carefully matched eRASS:4 analysis recovering 1:1 agreement. This supports the view that part of the 0.3 dex whole-sample difference is a measurement systematic rather than a pure selection effect. However, the paper is otherwise careful: the bias discussion is thorough, the eRASS:4/Chandra comparison is a strong internal check, the authors flag their own simulation limitations, and the qualitative direction of the effect is unlikely to reverse. A conditional verdict remains appropriate, pending a quantitative test of the AGN sensitivity. I therefore do not recommend changing the reader's verdict; the concern strengthens the conditionality rather than overturning the paper.","tokens_in":46082,"tokens_out":9795,"duration_ms":104133,"concrete_test":"Extend the Section 4 simulation pipeline by injecting into each simulated cluster a central point source drawn from a realistic AGN population (e.g., an AGN fraction of 10-30% with AGN-to-cluster flux ratios between 0.01 and 0.3, calibrated to eRASS1 point-source detections), rerun the eSASS detection and MBProj2D measurement steps, and compare the recovered c500 bias and detection fraction versus redshift and luminosity to the AGN-free curves in Figs. 12 and 18. If the median injected c500 bias or the z=0.4 detection threshold shifts by more than about 0.1 dex, the 0.3 dex population offset and the quantitative selection functions require revision; if the shift is below 0.05 dex, the omission is benign.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central quantitative claim is that eRASS1 clusters have median c500 about 0.3 dex higher than SPT, Planck ESZ, and eFEDS samples, and that this reflects an X-ray selection bias toward concentrated clusters as a function of redshift and luminosity. The quantitative selection curves in Fig. 12 and the bias curves in Fig. 18 are computed from simulations that explicitly contain no AGN (Section 4), and Section 6.11 concedes that adding a point-source population would likely add bias, especially for low-luminosity and high-redshift objects. This is not just a limitation of the selection-function modeling: an unresolved central AGN adds flux inside the 0.1 R500 aperture, directly increasing the measured concentration. With eROSITA's ~30 arcsec survey PSF, faint AGN in cluster cores cannot be resolved or masked, so the eRASS1 c500 values themselves may be systematically inflated relative to Chandra/XMM-Newton measurements, which resolve and remove point sources. The matched-sample comparison in Section 7.1 already shows a residual ~0.1 dex offset between eRASS1 and XMM/Chandra for the same clusters, with the eRASS:4 analysis using matched procedures showing 1:1 agreement; unresolved AGN are a plausible contributor to exactly this offset. Because the claimed 0.3 dex population difference is the sum of a ~0.1 dex measurement offset and a ~0.2 dex selection effect, and because the selection curves that support the interpretation are computed without point sources, the quantitative magnitude of the central claim is not yet securely established. The direction of the effect is likely preserved, so this is a serious quantitative caveat rather than a fatal flaw.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper measures 29 morphological parameters for 12,075 clusters in the eRASS1 catalogue, introducing two forward-modelled PSF-aware parameters (slosh and multipole magnitudes). It uses MCMC and bootstrap uncertainty propagation, simulations without AGN to quantify parameter biases and the selection function, and matched subsamples against XMM-Newton, Chandra, and eFEDS. The central empirical claim is that eRASS1 clusters are systematically more concentrated than SPT, Planck ESZ, and eFEDS samples, with median log c500 around 0.3 dex higher, and that the concentration distribution depends on redshift and luminosity because the X-ray selection preferentially detects concentrated clusters.","tokens_in":46459,"tokens_out":5201,"duration_ms":52245,"significance":"If correct, this is an important reference catalogue and a clear demonstration of how X-ray selection shapes morphological distributions. The paper's strengths are the scale of the catalogue, the explicit treatment of the PSF through forward-modelled parameters, the detailed uncertainty propagation, and the external validation using matched samples, including 1:1 agreement between eRASS:4 and Chandra when the analysis procedures are matched. However, the quantitative central claim depends on simulations that contain no AGN, a limitation stated in Section 6.11, and on a residual 0.1 dex offset between eRASS1 and XMM/Chandra for the same clusters. These issues make the quantitative selection and bias curves insecure, although the qualitative direction of the selection effect is plausible.","major_comments":[{"comment":"The simulations used to set the quantitative selection functions and bias corrections contain no AGN, and Section 6.11 concedes that adding a point-source population would likely add bias, especially for low-luminosity, high-redshift objects. Since an unresolved central AGN adds flux inside the 0.1 R500 aperture and eROSITA's roughly 30 arcsec survey PSF cannot resolve faint core point sources, the measured eRASS1 c500 values themselves may be inflated. This matters because the claimed 0.3 dex population offset is the sum of roughly a 0.1 dex measurement offset seen in Section 7.1 and a roughly 0.2 dex selection effect; if unresolved AGN contribute to the measurement offset, the quantitative population comparison and the selection curves in Fig. 12 are not yet secure. Please quantify this by adding point-source populations to the simulations or by testing c500 against known central AGN indicators in the eRASS1 sample.","section":"§4, §6.11, Figs. 12 and 18"},{"comment":"The matched-sample comparisons show a median eRASS1 concentration about 0.1 dex higher than XMM-Newton or Chandra for the same clusters, while the matched eRASS:4 analysis gives a 1:1 relation. This residual offset is not explained, and Section 7.1 lists unresolved point sources as a possible cause. The conclusion that cuts on exposure, counts, detection likelihood, and extension likelihood do not remove the 0.3 dex offset is based on observed distributions that retain this offset; the Abstract's statement that the population difference is a selection effect is therefore stronger than the current evidence supports. The authors should either correct the eRASS1 values for the inferred measurement offset before comparing populations, or demonstrate explicitly that the offset is not AGN-related.","section":"§7.1 and §7.2"},{"comment":"The Gaussian mixture model is trained on the 175 brightest clusters and then applied to the full sample, including those same 175 objects, to produce the reported roughly 25% disturbed fractions. This training-on-test overlap, without cross-validation, makes the Dshape and Dcomb fractions and the two-component amplitudes in Table 5 difficult to interpret. A cross-validated or independent-sample test is needed before the disturbance classification is used as a catalogue product.","section":"§8 and Table 5"},{"comment":"The simulated detection pipeline is image-based rather than the photon-based pipeline used for the real catalogue, and the authors note it is less sensitive near the detection threshold. This could shift the high-redshift selection thresholds in Fig. 12, in particular the claimed loss of clusters with c80−800 of about −0.2 at z = 0.4, and should be quantified so that the quoted threshold values are not taken as final.","section":"§4 and Fig. 12"}],"minor_comments":[{"comment":"In the Introduction, 'parameters are are affected by the signal to noise' contains a duplicated word.","section":"§1"},{"comment":"The text uses 'GGM' in several places, for example 'the GGM model does not show strong evidence', where 'GMM' is intended.","section":"§8"},{"comment":"The phrase 'For eFEDS1, we show distributions' appears to be a typo for 'For eFEDS'.","section":"§7.2"},{"comment":"The Type column codes (M, I, V, F, P) are not defined in the table itself; adding a footnote would substantially improve readability for catalogue users.","section":"Table 1"}],"recommendation":"major_revision","confidential_remarks":"The paper is honest about its limitations and the core catalogue will be useful, but the main quantitative claim and the disturbance score need the additional work described in the major comments. This is a request for more evidence rather than a rejection; the authors should be able to address the AGN and cross-validation points with targeted simulations and refits."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a solid, carefully done catalogue paper, and the main caveat to keep in mind is exactly the one the authors themselves flag: the simulations that anchor the bias corrections and selection functions contain no AGN. That makes the quantitative size of the headline result (eRASS1 clusters ~0.3 dex more concentrated than SPT/Planck/eFEDS) less secure than it looks, but the direction of the effect is probably right.\n\nWhat's new: the paper measures 29 morphological parameters for 12,247 eRASS1 clusters, making it the largest X-ray morphology catalogue to date. The genuinely new items are the forward-modelled 'slosh' and multipole magnitude parameters, which account for the eROSITA PSF and are validated on an independent SPT-Chandra sample. The matched-sample comparisons against XMM-Newton and Chandra, including an eRASS:4 re-analysis with matched procedures, are reassuring: the same clusters give consistent concentrations when the analysis is aligned, with a residual ~0.1 dex offset in eRASS1 that is plausibly due to unresolved AGN and procedural differences. The bias analysis in Figs. 15-19 is detailed and honestly presented, with per-parameter tables.\n\nWhere it's soft: the omission of point sources in the simulations is not a trivial detail. As the stress-test note points out, an unresolved AGN adds flux inside the 0.1 R500 aperture, directly inflating c500. The paper's own Section 6.11 concedes that adding point sources would likely add bias, especially for faint, distant objects. So the selection curves in Fig. 12 and the bias curves in Fig. 18 should be treated as indicative, not as precise calibrations. The claim that the 0.3 dex offset is largely selection effect rests on these AGN-free simulations; if unresolved AGN are common, part of that offset could be a measurement effect. The direction—X-ray selection favors concentrated clusters—is robust, but the magnitude isn't. No analysis code is released, which limits how much others can re-run the bias pipeline.\n\nVerdict: this deserves a serious referee. It's a reference data product, and the honest caveats mean a competent referee can address the AGN issue without starting from scratch. The GMM disturbance score is descriptive and reasonably validated; the self-classification point doesn't bother me.\n\nFor the reading group: I'd bring it, mainly to get a sense of what eRASS1 morphology can and cannot support yet. I'd cite it for the catalogue itself. My recommendation: send to peer review; ask for a clear statement of the AGN caveat in the abstract and a quantitative bound on how much unresolved point sources could shift the concentration bias, even if only an estimate.","headline":"Solid eRASS1 morphology catalogue with genuinely new forward-modelled parameters; the headline 0.3 dex concentration offset is directionally right but quantitatively depends on AGN-free simulations that the authors honestly flag.","tokens_in":47040,"tokens_out":3130,"would_cite":true,"duration_ms":26253,"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":"The first SRG/eROSITA all-sky morphological catalogue, covering more than twelve thousand optically confirmed galaxy clusters, argues that X-ray selection strongly favors concentrated, likely relaxed clusters, with log concentrations…","keywords":["galaxy clusters","intracluster medium","X-ray surveys","eROSITA","cluster morphology","surface brightness concentration","cluster selection function","cool cores"],"falsifier":"Take the eRASS1 clusters that fall in the deeper eRASS:4 footprint, re-measure $c_{500}$ and $c_{80-800}$ with detected point sources subtracted, and compare the median concentration with the eRASS1 values: if the median drops by roughly 0.3 dex, the claimed selection effect is not sufficient to explain the offset and the central claim fails.","tokens_in":45909,"feed_emoji":"🔭","tokens_out":9346,"duration_ms":83866,"temperature":0.7,"pith_summary":"This paper publishes a catalogue of 29 morphological parameters for more than twelve thousand clusters detected in the first eROSITA all-sky survey, and argues that the survey's cluster sample is strongly shaped by how concentrated each cluster's X-ray surface brightness is. It introduces two forward-modelled parameters, slosh and multipole magnitudes, that fold the telescope point-spread function into the fit, and uses simulations to show that several older image-based parameters such as power ratios, centroid shift, Gini coefficient, and photon asymmetry are badly biased by noise and PSF blurring. The central empirical claim is that eRASS1 clusters have log concentrations roughly 0.3 dex higher than SPT-, Planck ESZ-, or eFEDS-selected samples, meaning the ratio of flux inside $0.1\\,R_{500}$ to flux inside $R_{500}$ is about twice as large, and that cuts on exposure, counts, detection likelihood, or extension likelihood do not remove the offset. If correct, X-ray selection in eRASS1 preferentially finds concentrated, likely relaxed clusters at low redshift while missing the most extreme cool cores at high redshift. The paper also constructs two combined disturbance scores from a two-component Gaussian mixture model and classifies roughly a quarter of bright clusters as disturbed.","feed_headline":"eRASS1 clusters are 0.3 dex more concentrated than other samples","feed_subtitle":"Morphology catalogue for 12,000+ eROSITA clusters shows concentration, not just flux, decides what gets detected","key_machinery":"The load-bearing tool is MBProj2D forward modelling of the X-ray surface brightness: the cluster image is compared with a model that includes the eROSITA PSF, background, point sources and neighbouring clusters, using a density profile with a weak inner-slope prior. Concentration is measured from the model, not the image, as $\\log$ of the ratio of integrated surface brightness in apertures $0.1\\,R_{500}$ and $R_{500}$ ($c_{500}$), or 80 and 800 kpc ($c_{80-800}$). Two new parameters are introduced: slosh, which transforms the radius as $S'(r,\\theta) = A(H)\\,S(r\\,[1 + H\\cos(\\theta+\\theta_0)])$ with $A(H)=(1-H^2)^{3/2}$ to keep the total brightness fixed, and multipole magnitudes $M_m$, which multiply the profile by $[1 + M_m\\sin(m\\theta+\\theta_0)]\\,S(r)$. Simulated clusters placed on eROSITA sky maps and run through the eSASS detection pipeline provide the selection functions and bias curves, and a two-component Gaussian mixture model fit to the bright, high-count clusters defines the $D_{\\rm shape}$ and $D_{\\rm comb}$ disturbance scores.","core_discovery":"Measuring 29 morphological parameters for the 12,075 clusters with usable data from the eRASS1 catalogue, the authors find that the population is systematically more concentrated than clusters selected by the South Pole Telescope, by Planck, or in the deeper eFEDS field: the median log concentration is about 0.3 dex higher, a factor of two in the inner-to-total surface brightness ratio. They show from simulations that concentration controls whether a cluster is detected, with flat-core low-luminosity objects missed at low redshift and very peaked cool cores lost at high redshift because they look point-like; and they demonstrate that the concentration excess survives cuts in exposure time, counts, detection likelihood, and extension likelihood. For the same clusters, concentrations and central densities agree with Chandra and XMM-Newton measurements, and a matched re-analysis of SPT clusters in deeper eRASS:4 data agrees with Chandra at the one-to-one level. Image-based parameters such as power ratios, centroid shift, Gini coefficient, and photon asymmetry are strongly biased by noise and PSF, while the forward-modelled parameters are not. Finally, a two-component Gaussian mixture model applied to the shape parameters and concentration produces a disturbance score; about one quarter of bright clusters are classified as disturbed, and the paper attributes the high relaxed fraction to the survey's sensitivity to concentrated systems.","pith_inferences":["The selection function implies that eRASS1-based scaling relations, for example luminosity or $Y_X$ versus mass, will inherit a redshift- and luminosity-dependent bias because concentrated clusters are preferentially detected at the faint end; the authors do not work out this corollary here.","If the concentration excess is truly selection rather than measurement, then the deeper eRASS:4 survey should show a lower median concentration for the same clusters, an effect that could be checked already with the matched-sample machinery in the paper.","The missing high-redshift extreme cool cores could be recovered by wavelet-based or photon-based detection algorithms, which the authors note may find more flat-profile objects; this is a testable prediction of their selection model.","Because unresolved AGN were absent from the simulations, adding a realistic AGN population would probably shift the quantitative bias curves at low luminosity and high redshift; the direction of the selection effect would likely remain but its amplitude could change."],"forward_implications":["Cosmological analyses using eRASS1 cluster counts will need to include a concentration-dependent selection function, since detection efficiency varies strongly with both redshift and luminosity at fixed concentration.","The published catalogue gives bias and uncertainty tables as a function of redshift and count number, so users can correct or discard image-based morphological parameters rather than comparing them blindly with other surveys.","At low redshift the sample misses low-luminosity groups and clusters with flat surface brightness profiles; at $z \\gtrsim 0.4$ it misses the most concentrated cool cores, including objects like the Phoenix cluster.","The $\\sim$0.3 dex offset between eRASS1 and SPT/Planck/eFEDS concentrations indicates that X-ray-selected samples contain roughly twice the inner flux fraction of SZ-selected samples, so statements about cool-core fractions and relaxed fractions must be survey-selection aware.","Around a quarter of bright clusters are classified as disturbed by the combined score, while shape-only and concentration-inclusive scores can disagree for individual clusters; the two scores are provided so users can separate shape disturbance from peaked-core status."],"supporting_citations":[{"why":"Supplies the eRASS1 cluster catalogue of 12,247 objects whose images are analysed here, including redshifts, $R_{500}$ and count information.","marker":"B24"},{"why":"Defines the eRASS1 survey data release and the source-detection pipeline whose outputs the simulations are designed to reproduce.","marker":"Merloni et al. (2024)"},{"why":"Provides the optical confirmation and contamination model that define the main and cosmology subsamples.","marker":"Kluge et al. (2024)"},{"why":"Supplies the XMM-Newton Planck ESZ morphological measurements used as matched-sample and whole-sample concentration and density comparisons.","marker":"Lovisari et al. (2017)"},{"why":"Provides the SPT cluster catalogue from which the high-resolution comparison sample and shape-parameter tests are drawn.","marker":"Bleem et al. (2015)"},{"why":"Supplies the Chandra observations and analysis of SPT clusters used as a cross-instrument consistency check and as training images for the new shape parameters.","marker":"Sanders et al. (2018)"},{"why":"Provides the eFEDS concentration, density, and relaxation measurements that anchor the deeper-survey comparison.","marker":"Ghirardini et al. (2022)"},{"why":"Defines the cuspiness parameter and the density-profile framework on which the central-density and slope measurements are built.","marker":"Vikhlinin et al. (2007)"},{"why":"Provides the Gaussian mixture model implementation used to construct the $D_{\\rm shape}$ and $D_{\\rm comb}$ disturbance scores with measurement-error covariances.","marker":"Melchior & Goulding (2018)"}],"fun_headline_variants":["eRASS1 clusters 0.3 dex more concentrated than SPT, Planck, eFEDS","Survey bias: flat-core clusters missed in eRASS1 morphology catalogue","Concentration controls which clusters eRASS1 detects","New eROSITA morphology catalogue for 12,000+ clusters","Forward-modelled morphology beats PSF bias in eROSITA survey"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quantitative bias curves, reliability thresholds, and selection functions are computed from simulations that contain no AGN, so if unresolved active galactic nuclei are common in eRASS1 clusters the claimed sizes of the concentration bias and high-redshift selection effect could be wrong.","fun_headline_variants_meta":{"raw":{"variants":["eRASS1 clusters 0.3 dex more concentrated than SPT, Planck, eFEDS","Survey bias: flat-core clusters missed in eRASS1 morphology catalogue","Concentration controls which clusters eRASS1 detects","New eROSITA morphology catalogue for 12,000+ clusters","Forward-modelled morphology beats PSF bias in eROSITA survey"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000796,"raw_usage":{"total_tokens":3594,"prompt_tokens":1126,"completion_tokens":2468,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":742,"completion_tokens_details":{"reasoning_tokens":2371}},"tokens_in":742,"tokens_out":2468,"duration_ms":16669,"temperature":1.0,"reasoning_tokens":2371,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T12:49:43.448317+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the eRASS1 clusters that fall in the deeper eRASS:4 footprint, re-measure $c_{500}$ and $c_{80-800}$ with detected point sources subtracted, and compare the median concentration with the eRASS1 values: if the median drops by roughly 0.3 dex, the claimed selection effect is not sufficient to explain the offset and the central claim fails.","supporting_citations":[{"cited_title":"R., Jones, C., et al","cited_arxiv_id":null,"evidence_quote":"Supplies the XMM-Newton Planck ESZ morphological measurements used as matched-sample and whole-sample concentration and density comparisons."},{"cited_title":"E., Stalder, B., de Haan, T., et al","cited_arxiv_id":null,"evidence_quote":"Provides the SPT cluster catalogue from which the high-resolution comparison sample and shape-parameter tests are drawn."},{"cited_title":"S., Fabian, A","cited_arxiv_id":null,"evidence_quote":"Supplies the Chandra observations and analysis of SPT clusters used as a cross-instrument consistency check and as training images for the new shape parameters."},{"cited_title":"E., Bulbul, E., et al","cited_arxiv_id":null,"evidence_quote":"Provides the eFEDS concentration, density, and relaxation measurements that anchor the deeper-survey comparison."},{"cited_title":"R., et al","cited_arxiv_id":null,"evidence_quote":"Defines the cuspiness parameter and the density-profile framework on which the central-density and slope measurements are built."},{"cited_title":"& Goulding, A","cited_arxiv_id":null,"evidence_quote":"Provides the Gaussian mixture model implementation used to construct the $D_{\\rm shape}$ and $D_{\\rm comb}$ disturbance scores with measurement-error covariances."}],"review_version":1}