{"id":"c9bf86c5-0b96-4253-bb3f-afff20688751","arxiv_id":"2507.22453","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Simulated AGN contamination at COSMOS-like depths shows the Gini coefficient is the most robust morphology measure up to z~2, with concentration parameters shifting significantly above 25% AGN fraction and magnitude 23.","lead":"This paper simulates what happens to standard measures of galaxy shape and light concentration when a bright galactic nucleus (an AGN) is added and the galaxy is pushed to the depth and redshift of the COSMOS survey, up to z~2. The practical message: the Gini coefficient stays reliable, while concentration indices shift strongly once the AGN contributes more than 25% of the light and the galaxy is fainter than magnitude 23.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Section 3.1 says simulated COSMOS-like galaxies were convolved with the SDSS PSF; if this is not a wording artifact, the headline thresholds and stability ranking are calibrated at ~1.4 arcsec resolution rather than HST/ACS ~0.09 arcsec, so the central 'COSMOS-like' claim is not yet supported.","rationale":"The paper is a calibration study: it injects AGN light into local templates and measures how six morphological parameters respond under simulated COSMOS depth and resolution. The measurements are transparent, and the qualitative trends (spirals more affected, Gini stable) are plausible and consistent with the authors' earlier z~0 work [38], which supports a conditional reading. However, the central quantitative contribution, the thresholds of AGN >25% and mag >23, and the parameter ranking, inherits every detail of the simulation. Section 3.1's PSF sentence is where the 'COSMOS-like' condition is least secure. If the SDSS PSF was used instead of the HST/ACS PSF, the simulated images have more than an order of magnitude worse resolution than real COSMOS, so the abstract and conclusion numbers are not calibrated to the stated survey. This is a more immediate problem than the local-template assumption flagged by the reader: it concerns the internal validity of the synthetic observations, not only the external generalization to real z~2 structures. The reader's template concern is real and remains relevant, but the PSF issue should be checked first because it can be settled from the public galSVM code and a single rerun. I keep the overall verdict at CONDITIONAL (no change from the reader), but the conditions should explicitly include verifying the PSF adopted in Section 3.1 and re-deriving the headline thresholds with the ACS PSF. If the PSF is confirmed wrong and the thresholds shift, the appropriate verdict would be REJECT or UNVERDICTED for the quantitative claims as currently stated.","tokens_in":53739,"tokens_out":10326,"duration_ms":106723,"concrete_test":"Inspect the galSVM configuration used in Section 3.1 (or rerun the pipeline) and measure the FWHM of a simulated point source / unresolved AGN in the output images. If the FWHM is ~1.4 arcsec (SDSS seeing) rather than ~0.09 arcsec (HST/ACS F814W), the PSF treatment is wrong. Then rerun one representative magnitude bin (e.g., mag_hz < 23, all AGN fractions) with the COSMOS/ACS PSF; if the Var values in Figures 4 or 6 or the AGN >25% threshold shift by more than ~10%, the paper's quantitative claims are not robust to the resolution treatment.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim depends on the simulated images representing COSMOS observations. Section 3.1 states: 'All local galaxies have been re-sampled with a pixel scale of 0.03 arcsec/pix corresponding to COSMOS and convolved with their point spread function (PSF) obtained from the SDSS survey for each galaxy to match the same spatial resolution.' The target data are HST/ACS F814W images with a PSF of ~0.09 arcsec FWHM, while the SDSS PSF is ~1.4 arcsec FWHM at 0.396 arcsec/pix. Convolving with the SDSS PSF after rebinning to 0.03 arcsec/pix leaves the simulated galaxies at ground-based seeing rather than HST resolution. This is not cosmetic: the six parameters are non-parametric light-distribution measures, and the exact thresholds (AGN >25%, mag >23) plus the GINI > M20 > CCON > CABR ranking are precisely the quantities most sensitive to PSF size. A broad PSF smears the unresolved AGN and dilutes the central light excess, miscalibrating the AGN effect, while also artificially lowering concentration at faint magnitudes and shifting the mag >23 boundary. Because the same PSF is applied to all simulated images, internal comparisons stay self-consistent, but the external label 'COSMOS-like conditions' fails if the ACS PSF was not substituted. This should be resolved before the quantitative degradation map is used.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents a simulation study of how unresolved AGN light affects six non-parametric morphological parameters (CABR, GINI, CCON, M20, ASYM, SMOOTH) for galaxies observed in conditions mimicking the COSMOS survey up to z~2. Starting from 2251 local non-AGN SDSS galaxies with visual morphological classifications, the authors add AGN point sources contributing 5-75% of total flux, redshift and dim the galaxies to five magnitude/redshift limits matching COSMOS, insert them into real COSMOS backgrounds, and measure the parameters with galSVM. Changes are quantified by a mean absolute relative deviation (called variance) with a 20% significance threshold. The main results are that combined magnitude/redshift/AGN effects have the largest impact; spiral galaxies are more affected than early types; GINI is the most stable parameter, followed by M20, CCON and CABR; significant changes in concentration parameters typically occur for AGN contributions above 25% and magnitudes above 23; and the CABR-CCON and CABR-ASYM diagrams are the most effective for classifying AGN hosts.","tokens_in":53854,"tokens_out":10523,"duration_ms":108934,"significance":"If the results are correct, the paper would give survey users a quantitative guide to which morphological parameters and diagram combinations can be trusted under AGN contamination and survey depth/resolution effects up to z~2. The experimental design is a strength: the simulations are described carefully, with rest-frame band matching to reduce K-correction effects, real COSMOS backgrounds for noise, and a full grid of five magnitude/redshift limits and five AGN fractions. The ranking of parameter stability is a useful, falsifiable output. However, the significance is conditional on the fidelity of the simulations to real COSMOS observations; the PSF calibration and the use of purely local structural templates are the main uncertainties. The paper also overstates the universality of its 'all concentration parameters' conclusion, since GINI remains stable for early-type galaxies in the authors' own measurements.","major_comments":[{"comment":"The simulations are labeled 'COSMOS-like,' but the galaxies are convolved with 'their point spread function (PSF) obtained from the SDSS survey' (Section 3.1), and the AGN is modeled with an SDSS Moffat PSF (Section 3.2). The target data are HST/ACS F814W images with a PSF of about 0.09 arcsec FWHM, whereas the SDSS PSF is about 1.4 arcsec FWHM. After rebinning to 0.03 arcsec/pix, convolving with the SDSS PSF leaves the simulated images at ground-based seeing rather than HST resolution. Because all six parameters are non-parametric light-distribution measures, the reported thresholds (AGN contribution >25%, mag>23) and the stability ranking (GINI > M20 > CCON > CABR) are sensitive to PSF size; a broad PSF smears the unresolved AGN and dilutes the central light excess, potentially miscalibrating the AGN effect and shifting the magnitude boundary. The authors should either re-run the simulations with the appropriate ACS PSF or demonstrate quantitatively that the conclusions are unchanged with a narrower PSF; otherwise, the external claim of 'COSMOS-like conditions' is not supported.","section":"Section 3.1 and 3.2"},{"comment":"The only structural templates are local z~0 galaxies, randomly scaled in brightness to match the COSMOS magnitude and redshift distributions (Section 3.1). This assumes no evolution in the intrinsic light distribution of galaxies: no compactness evolution, no clumpy star-forming structures, and no increasing irregularity fraction. The paper does not validate this assumption against observed high-redshift galaxies. If real z~2 galaxies are structurally different, the per-parameter variances and the thresholds would shift in actual COSMOS data. The authors should add a validation test using observed high-redshift galaxies (e.g., from COSMOS or CANDELS) or clearly restrict the conclusions to 'local templates observed at higher redshift' and temper the claims about COSMOS-like conditions at z~2.","section":"Section 3.1 and Section 5"},{"comment":"The statement in Section 6 that 'all four concentration parameters will undergo significant changes of >20% for >25% of the added AGN' and the abstract's 'all the concentration parameters change significantly if the AGN contribution is >25%' are contradicted by the authors' own measurements for GINI in early-type galaxies. Appendix A2 shows Var<20% for GINI in early types at all magnitude limits even at 75% AGN, and Section 4.3 states GINI shows 'variance of <20% in almost all cases, regardless of morphological type.' The universal claim should be qualified by morphological type and parameter, or the threshold should be revised; as written, this is an internal inconsistency in the paper's central claim.","section":"Abstract and Section 6"},{"comment":"The variance values in Eq. (1) are point estimates of mean absolute relative deviation, but the paper provides no uncertainties (e.g., bootstrap or jackknife) for these values or for the derived parameter ranking. The ranking GINI > M20 > CCON > CABR is based on point estimates; in the total sample at 25% AGN, CABR (42%) and GINI (41%) are nearly identical (Appendices A1 and A2), so the ordering is not clearly significant without error bars. The authors should provide confidence intervals or a significance test for the ranking and for the 20% threshold crossings, or at least demonstrate that the ordering is stable across subsets and bootstrap resamples.","section":"Section 4.1, Eq. (1)"}],"minor_comments":[{"comment":"The phrase 'moved all the galaxies to lower magnitudes (higher redshifts)' should be 'fainter magnitudes (higher redshifts)' to avoid confusion with the magnitude scale.","section":"Abstract"},{"comment":"The caption contains the typo 'mag_hzmag_hz < 24'; it should read 'mag_hz < 24'.","section":"Figure 4 caption"},{"comment":"Reference [60] is cited as the source for the AB magnitude system, but the AB system is defined by Oke and Gunn (1983); Parekh et al. (2015) is not the standard reference for this definition.","section":"References"},{"comment":"The quantity in Eq. (1) is called 'variance' but is actually a mean absolute relative deviation; consider renaming it to avoid confusion with the statistical variance.","section":"Eq. (1)"},{"comment":"In the CABR bullet, the sentence 'the variance in the parameter is significant in all magnitude and redshift conditions when the contribution of the AGN is equal to or larger than 25%, 50%, and 75%, respectively' is ambiguous; please specify which morphological types correspond to which threshold.","section":"Section 4.1"},{"comment":"The received/revised dates contain the typo '22 July 202'; it should be '22 July 2025'.","section":"Header"}],"recommendation":"major_revision","confidential_remarks":"To the editor: The PSF issue in Section 3.1 is the most serious concern and, in my view, should be resolved before acceptance; it is not a stylistic matter. The paper would be significantly strengthened by a validation against high-redshift data, but that may be beyond the scope; at minimum, the conclusions should be restricted to the local-template scenario. The internal inconsistency regarding GINI and 'all four concentration parameters' should be corrected. I recommend major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this paper extends the authors' z~0 AGN morphology study to COSMOS-like conditions out to z~2. The genuinely new deliverable is a variance grid: five magnitude/redshift cuts crossed with five AGN fractions, with per-parameter variance values for six non-parametric morphological measures. The headline result—GINI most stable, then M20, CCON, CABR—is consistent across the grid and matches their earlier work at z~0.\n\nThe methodology is mostly transparent: they use real COSMOS backgrounds, measure in rest-frame SDSS bands to handle K-correction, and the variance statistic is emergent rather than fitted, so circularity is low. The practical guidance (concentration parameters change substantially above roughly 25% AGN contribution and magnitude 23) is useful for survey planning.\n\nThe soft spots are real but not necessarily fatal. First, the local template assumption: using z~0 non-AGN galaxy structures as stand-ins for z~2 galaxy structures assumes no strong evolution in compactness, clumpiness, or irregularity fraction. The thresholds would shift if real high-z galaxies are structurally different. Second, there are no error bars on the variance values, and the fraction of galaxies with measurable parameters drops to 60–70% at the faintest, most AGN-dominated conditions (much lower for ASYM and SMOOTH), so the variance grid is not equally well-sampled everywhere.\n\nThe bigger concern is the PSF handling. Section 3.1 says the local galaxies were \"convolved with their point spread function (PSF) obtained from the SDSS survey.\" That suggests the simulated images are at ground-based seeing, not HST/ACS ~0.09 arcsec resolution. If that is correct, the \"COSMOS-like\" label is misleading and the exact thresholds could be miscalibrated, because a broad PSF smears the AGN and dilutes the central concentration signal. It might be a wording artifact, but it needs explicit clarification. I would want a side-by-side comparison of simulated and real COSMOS images at matched magnitude/redshift to confirm the effective resolution actually matches.\n\nAlso, no simulated images or code are shipped, only links to public tools and catalogs, so reproduction requires re-implementation. The abstract also slightly overstates diagram effectiveness by dropping the \"up to magnitude 22 and AGN fraction 25%\" qualifiers from Section 6.\n\nIn sum: the paper deserves a serious referee. The qualitative ranking is likely robust, but the quantitative thresholds rest on the PSF assumption and the local-template assumption. A referee should ask for a clear statement of the exact PSF used, a validation of the simulated resolution against actual COSMOS data, and ideally a check using an observed high-redshift sample. I would bring it to a reading group once the PSF question is settled; for now I would not cite it without that resolution.","headline":"Useful z~2 extension of the AGN-morphology calibration grid, but the PSF handling in the simulation needs to be verified before trusting the quantitative thresholds.","tokens_in":54637,"tokens_out":4057,"would_cite":false,"duration_ms":44532,"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":"This paper claims that in COSMOS-like conditions up to $z\\sim2$, the four concentration-type morphological parameters change significantly when the AGN contributes more than 25% of the total light and the galaxy is fainter than…","keywords":["Galaxies: high-redshift","Galaxies: formation","Galaxies: fundamental parameters","Galaxies: nuclei","Galaxies: evolution","AGN host galaxies","morphological classification","non-parametric morphology"],"falsifier":"Measure the same six parameters on real, well-resolved AGN host galaxies at $z\\sim1$--$2$ with independently known AGN fractions (for example from spectral decomposition or X-ray luminosity) and check whether the Gini coefficient stays below 20% variance for AGN fractions up to 75% and whether the concentration parameters cross the 20% threshold at AGN fraction about 25% and magnitude about 23.","tokens_in":53351,"feed_emoji":"🔭","tokens_out":5957,"duration_ms":63040,"temperature":0.7,"pith_summary":"This paper tries to establish how much an active galactic nucleus contaminates the six non-parametric morphological measurements used to classify galaxies, and where the practical limits lie when observing galaxies out to $z\\sim2$ in a deep survey like COSMOS. It simulates adding an AGN point source contributing 5% to 75% of the total light to thousands of local galaxies with known visual morphology, then moves them to the magnitudes and redshifts of COSMOS. The central finding is that the concentration parameters change significantly when the AGN contributes more than 25% and the galaxy is fainter than $F814W\\sim23$, and that the Gini coefficient is the most stable parameter throughout, followed by M20, CCON, and CABR. If true, this gives survey designers and morphology users a degradation map: which parameters and diagram combinations can be trusted, up to which depth and AGN fraction.","feed_headline":"Galaxy morphology measures fail past 25% AGN and mag 23","feed_subtitle":"Simulations out to z~2 show the Gini coefficient stays stable while concentration indexes drift.","key_machinery":"The central machinery is a simulation pipeline that takes 2251 local non-AGN galaxies with visual T-type classifications, injects a PSF-convolved AGN point source contributing 5% to 75% of total flux at the galaxy centre, rescales the images to match the magnitude and redshift distributions of five COSMOS F814W-limited samples ($F814W<21$, $22$, $23$, $24$, $25$), and drops them into real COSMOS background images to reproduce depth and resolution. The six parameters are measured with the galSVM code, and the effect is quantified by the variance statistic $\\mathrm{Var}[\\%]$, the average absolute relative change per galaxy, with a change of $\\ge 20\\%$ taken as significant.","core_discovery":"The paper argues that when an unresolved AGN contributes more than 25% of a galaxy's total light and the galaxy is observed at $F814W>23$ ($z>1$), the four commonly used concentration parameters---CABR, CCON, GINI, and M20---shift by more than 20% relative to the uncontaminated local values. Among these, the Gini coefficient is the most stable under both AGN contamination and magnitude/redshift degradation, followed by M20, then CCON, then CABR, the last being the most sensitive to central light. The paper also claims that the most reliable two-dimensional diagnostics for classifying active galaxies at high redshift are the CABR-versus-CCON and CABR-versus-ASYM diagrams, while GINI-versus-M20 fails to separate morphological types under these conditions.","pith_inferences":["The local-template assumption implies the derived thresholds give a conservative picture if real $z\\sim2$ galaxies are intrinsically clumpier or more irregular than local galaxies; in that case, the magnitude where parameters change could be brighter than $F814W\\sim23$.","The stability ranking suggests a practical selection strategy for deep surveys: use GINI as the primary parameter, M20 as a secondary check, and avoid SMOOTH entirely because it is measurable for only a small fraction of sources.","The variance metric is an average absolute relative change, which can overstate instability for parameters with values near zero, so testing ASYM and SMOOTH with an absolute-difference or signal-to-noise-based metric could change their ranking.","Running the same simulation on JWST/NIRCam images rather than HST/ACS F814W would test whether the 25% and $F814W\\sim23$ thresholds hold at longer wavelengths and higher resolution, where high-redshift morphology is now routinely measured."],"forward_implications":["Survey users should treat morphology measurements of AGN hosts at $F814W>23$ with AGN fraction above 25% as unreliable, especially for spiral galaxies.","The Gini coefficient can be used up to $z\\sim2$ with low variance even at high AGN fractions, making it a suitable primary classifier for active galaxies in deep fields.","CABR versus CCON and CABR versus ASYM diagrams separate early- and late-type galaxies up to $F814W<22$ and AGN fractions of 25% or less, while GINI versus M20 is not recommended for active galaxies.","The AGN effect dominates the magnitude/redshift effect when the AGN contributes at least 50% of the light, so nuclear contamination sets the practical depth limit for morphology measurements.","Late spirals are the most affected morphological class, so high-redshift studies of star-forming AGN hosts need extra caution when interpreting morphological parameters."],"supporting_citations":[{"why":"Predecessor at $z\\sim0$ that supplies the local sample, the AGN injection method, and the 20% variance threshold; this paper extends its findings to COSMOS-like conditions up to $z\\sim2$.","marker":"[38]"},{"why":"Establishes how survey depth and resolution change the six morphological parameters, providing the template-scaling approach and baseline variances that this work builds on.","marker":"[37]"},{"why":"GalSVM code used to measure all six non-parametric morphological parameters of the simulated galaxies.","marker":"[31]"},{"why":"Provides the visual T-type morphological classifications used to define the local early-type, early-spiral, and late-spiral samples.","marker":"[13]"},{"why":"Defines the COSMOS survey whose magnitude and redshift distributions define the simulated COSMOS-like conditions.","marker":"[63]"},{"why":"Supplies the photometric redshifts used to build the COSMOS redshift distribution that galaxies are scaled to match.","marker":"[66]"},{"why":"Introduces the Gini coefficient and M20 moment-of-light definitions whose stability is the main result.","marker":"[34,35]"},{"why":"Introduces the Conselice-Bershady concentration index whose complex behaviour under AGN injection is analysed.","marker":"[36]"},{"why":"Introduces the Abraham concentration index and the asymmetry index, the former being the most affected parameter and the latter used in the recommended diagnostic diagrams.","marker":"[27,28]"}],"fun_headline_variants":["AGN light skews galaxy shape metrics at 25% or more","Gini resists AGN contamination; concentration does not","z~2: AGN breaks concentration measures, Gini holds","Morphology metrics fail with AGN >25% and mag >23","For AGN hosts, use CABR-gini not Gini-M20"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The results assume that local $z\\sim0$ galaxies are faithful structural templates for galaxies at $z\\sim2$, so the intrinsic light distribution of high-redshift galaxies does not evolve in ways the simulation does not include.","fun_headline_variants_meta":{"raw":{"variants":["AGN light skews galaxy shape metrics at 25% or more","Gini resists AGN contamination; concentration does not","z~2: AGN breaks concentration measures, Gini holds","Morphology metrics fail with AGN >25% and mag >23","For AGN hosts, use CABR-gini not Gini-M20"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000297,"raw_usage":{"total_tokens":1732,"prompt_tokens":963,"completion_tokens":769,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":579,"completion_tokens_details":{"reasoning_tokens":675}},"tokens_in":579,"tokens_out":769,"duration_ms":8221,"temperature":1.0,"reasoning_tokens":675,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T11:40:31.401389+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the same six parameters on real, well-resolved AGN host galaxies at $z\\sim1$--$2$ with independently known AGN fractions (for example from spectral decomposition or X-ray luminosity) and check whether the Gini coefficient stays below 20% variance for AGN fractions up to 75% and whether the concentration parameters cross the 20% threshold at AGN fraction about 25% and magnitude about 23.","supporting_citations":[{"cited_title":"Effect of AGN on the morphological properties of their host galaxies in the local Universe","cited_arxiv_id":null,"evidence_quote":"Predecessor at $z\\sim0$ that supplies the local sample, the AGN injection method, and the 20% variance threshold; this paper extends its findings to COSMOS-like conditions up to $z\\sim2$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes how survey depth and resolution change the six morphological parameters, providing the template-scaling approach and baseline variances that this work builds on."},{"cited_title":"A robust morphological classification of high-redshift galaxies using support vector machines on seeing-limited images","cited_arxiv_id":null,"evidence_quote":"GalSVM code used to measure all six non-parametric morphological parameters of the simulated galaxies."},{"cited_title":"A catalog of detailed visual morphological classifications for 14,034 galaxies in the sloan digital sky survey","cited_arxiv_id":null,"evidence_quote":"Provides the visual T-type morphological classifications used to define the local early-type, early-spiral, and late-spiral samples."},{"cited_title":"The cosmic evolution survey (COSMOS): Overview","cited_arxiv_id":null,"evidence_quote":"Defines the COSMOS survey whose magnitude and redshift distributions define the simulated COSMOS-like conditions."},{"cited_title":"Cosmos Photometric Redshifts with 30-Bands for 2-deg 2","cited_arxiv_id":null,"evidence_quote":"Supplies the photometric redshifts used to build the COSMOS redshift distribution that galaxies are scaled to match."},{"cited_title":"Structural and photometric classification of galaxies","cited_arxiv_id":null,"evidence_quote":"Introduces the Conselice-Bershady concentration index whose complex behaviour under AGN injection is analysed."}],"review_version":1}