{"id":"99c144c3-0fa3-4706-b23d-1663b36d5668","arxiv_id":"1908.06775","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Among UV-bright red-sequence galaxies at z about 0.06 to 0.4, the fraction classified as UV upturn increases with stellar mass and rises then falls with redshift, with the mass trend concentrated in retired/passive systems.","lead":"The paper measures how often red-sequence galaxies with strong ultraviolet emission show the 'UV upturn' feature, using public survey data from GAMA, GALEX, and SDSS. It finds the fraction of such galaxies rises with stellar mass and appears to rise then fall with redshift, with the clearest signal in quiescent 'retired/passive' galaxies.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"GALEX FUV detection limit is not modelled; the claimed rise in UV-upturn fraction with redshift may be a selection artefact, and Sec. 5.1 itself says the post-peak decline is not secure.","rationale":"The central measurement is a new population-level estimate, and the Bayesian logistic treatment is a reasonable and standard choice. The mass trend is supported by the retired/passive subset and by the CDF shift, and the authors are transparent about caveats in Appendix A2.2. However, the redshift part of the headline claim rests on the assumption that FUV detection does not differentially select the outcome variable. That assumption is violated by construction: UV upturn systems are defined by bluer FUV-NUV, i.e. brighter FUV at fixed red-sequence optical colour, so the MIS FUV limit removes UV weak systems preferentially as redshift grows. The model's inclusion of log M* addresses a luminosity/distance bias in the predictor, not a selection on the response. The Appendix C volume cut (M_r <= -22) also does not remove FUV selection. The paper's own Sec. 5.1 admits the post-peak decline is not statistically identifiable, making the abstract's 'in-fall can be clearly seen' an overstatement. This is the load-bearing soft spot because the up-rise/infall is one of the two headline results and is exactly the shape a pure FUV-selection effect would produce. A completeness or 1/Vmax re-analysis can settle it with public GAMA/GALEX data. The reader's conditionality already captures this concern, so I would keep the CONDITIONAL verdict and require the completeness check before any stronger claim is made.","tokens_in":24419,"tokens_out":8265,"duration_ms":94686,"concrete_test":"Compute for every final-sample galaxy the GALEX MIS FUV detection limit (FUV_AB about 22.7) and use the k-corrections shown in Fig. A1 to derive the maximum redshift at which that galaxy would still have been detected. Re-run the logistic regression with 1/Vmax weights, or on the volume-complete subset with absolute FUV bright enough to be detected at z=0.35. If the z=0.06-0.25 rise and the z about 0.25-0.35 decline shrink to within the 95% credible interval of a flat trend, the headline redshift evolution is a selection artefact. A simpler cross-check is to repeat the analysis using only objects with FUV brighter than 22 AB; if the peak position moves or vanishes, detection bias is driving the trend.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing premise is that the logistic regression estimates the true fraction of UV upturn among UV-bright red-sequence galaxies. This requires that entry into the sample does not depend on UV class in a redshift- or mass-dependent way. It does: the final sample requires FUV detection in GALEX MIS, and UV upturn galaxies are defined by bluer FUV-NUV, hence brighter FUV at fixed r and NUV. At higher redshift, the FUV flux limit therefore preferentially removes UV weak galaxies, inflating the apparent upturn fraction. Including log M* in the regression does not correct for this, because the selection is in FUV apparent magnitude, not in log M*. The Appendix C volume-limited test cuts on M_r, not on absolute FUV, so it does not remove the bias. Internal support for this reading: Sec. 5.1 states that after the inferred peak 'the credible intervals widen considerably, and one cannot safely affirm whether the probability decreases, increases, or plateaus', while the abstract and conclusions assert an infall 'can be clearly seen'. The uncorrected internal-extinction boundary migration estimated in Appendix A2.2 is a second, smaller systematic of the same kind: it changes class fractions by tens of per cent, and the paper does not demonstrate that this migration is symmetric in redshift and mass.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper estimates the incidence of UV upturn among UV-bright red-sequence galaxies (RSGs) selected from GAMA-DR3 with GALEX MIS and SDSS photometry, using the Yi et al. (2011) colour criteria and WHAN emission-line classifications. A Bayesian logistic regression models the probability that a UV-bright RSG is classed as UV upturn as a quadratic function of redshift and stellar mass, with a second version stratified by emission-line class. The main claims are that the UV-upturn fraction rises with stellar mass, particularly for retired/passive systems, and that for retired/passive systems the fraction rises with redshift up to z ~ 0.20-0.25 and then declines.","tokens_in":24613,"tokens_out":2827,"duration_ms":33003,"significance":"If the trends are real, the paper provides a statistically framed measurement of how UV-upturn incidence depends on stellar mass and redshift among quiescent UV-bright galaxies, complementing cluster-based studies such as Ali et al. (2018c). The analysis is transparent in its use of public GAMA/GALEX/SDSS data, makes the statistical model explicit, and includes supplementary material: a volume-limited subsample and a discussion of internal-extinction effects. These strengths make the paper a potentially useful contribution to the study of evolved stellar populations in early-type galaxies. However, the central claims rest on the assumption that the FUV-detection-based sample is not biased by UV class in a redshift- or mass-dependent way, and this assumption is not tested; the paper itself also hedges the post-peak decline in Sec. 5.1 while asserting it clearly in the abstract and conclusions.","major_comments":[{"comment":"The sample requires FUV detection in GALEX MIS, while the UV-upturn class is defined by bluer (FUV-NUV) and hence brighter FUV at fixed r and NUV. At higher redshift, the apparent FUV flux limit removes UV-weak galaxies preferentially, which would inflate the estimated UV-upturn fraction. Including log M* in the logistic regression does not correct for this because the selection is in FUV apparent magnitude, not in stellar mass. The volume-limited test in Appendix C cuts on M_r rather than absolute FUV magnitude, so it does not remove the bias. The paper should either model the FUV detection probability as a function of redshift, magnitude, and colour, or restrict the analysis to a region of absolute FUV magnitude where the sample is complete, and then re-fit the logistic model.","section":"§2.4, §5.1, Appendix C"},{"comment":"There is an internal inconsistency about the post-peak redshift trend. Section 5.1 states that 'one cannot safely affirm whether the probability decreases, increases, or plateaus' after the peak, but the abstract and the conclusions state that an in-fall 'can be clearly seen'. The conclusions should be reworded to match the stated statistical uncertainty, or the authors should provide an additional test (e.g., a comparison of model evidence for a declining versus a plateauing trend) that supports the stronger claim.","section":"§5.1 versus Summary & Conclusions, point 5"},{"comment":"Internal extinction is not corrected, and the paper's own estimate is that roughly 30 per cent of UV-weak systems would migrate to another UV class and about 6 per cent of UV-upturn systems would migrate out if extinction were accounted for. Since the UV class is the binary response variable in the logistic regression, this level of classification migration can directly change the fitted fraction surface. The paper does not show that the migration is symmetric in redshift and stellar mass; if it is not, the claimed mass and redshift trends could be partly driven by misclassification. A robustness test using extinction-corrected classifications, or explicitly propagating the classification uncertainty into the regression, is needed before the central trends can be considered secure.","section":"Appendix A2.2, Eq. (A2)"}],"minor_comments":[{"comment":"There is a typo: 'restrictions on the quality ofz measurements ware taken into account' should read 'were taken into account'.","section":"§2.1"},{"comment":"The text says the WHAN diagram 'segragates' galaxies into five groups; 'segregates' is meant.","section":"§3.1"},{"comment":"The phrase 'one must consider the a few issues' contains a stray article and should be edited.","section":"§6.2"},{"comment":"The bar charts display fractions without error bars or credible intervals. Given the small counts in the highest-redshift bins, adding Poisson or binomial error bars would help the reader judge the significance of the apparent trends.","section":"Fig. 6 and Fig. 7"},{"comment":"The text says the volume-limited subsample shows 'very similar trends' to Figs. 9 and 10, but it would be useful to state explicitly whether the peak redshift and the mass-dependence slope are consistent within the credible intervals, rather than only visually similar.","section":"Appendix C"}],"recommendation":"major_revision","confidential_remarks":"The FUV detection-selection issue is the main risk to the paper's central claims. It is not a circularity problem, but a missing completeness model for a selection variable that is correlated with the response variable. The internal inconsistency between Sec. 5.1 and the abstract/conclusions should be fixed by the authors. The paper's transparency about its caveats is a positive feature, but the caveats currently undermine the strength of the stated conclusions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know: this is a measurement of how common UV upturn is among UV-bright red-sequence galaxies as a function of stellar mass and redshift, using public GAMA, GALEX MIS, and SDSS data. The genuinely new piece is that it covers field galaxies over a continuous redshift range (0.06-0.4) and stratifies by WHAN emission-line class, whereas most prior work targeted clusters. The Bayesian logistic regression is standard, externally anchored to the Yi et al. (2011) cuts, and there is no circularity.\n\nThe paper does several things well. It builds the sample carefully, handles flags, k-corrections, and foreground extinction, and is unusually honest about the internal extinction problem: Appendix A2.2 quantifies the boundary migration, estimating a ~30% depletion of the UV weak class, and then tests against dust-to-stellar mass ratios. That is real credit. The volume-limited appendix is a sensible sanity check. The mass trend in retired/passive systems, which is the most robust piece, is consistent with the simple CDFs: UV upturn hosts are systematically more massive.\n\nNow the soft spots. The headline redshift trend is not supported by the model's own credible intervals. The text of Section 5.1 says that after z~0.25 one cannot safely say whether the probability decreases, increases, or plateaus. The abstract and conclusions claim an infall 'can be clearly seen.' That mismatch is real and should be fixed.\n\nThe stronger concern is selection. The sample requires FUV detection in GALEX MIS, and UV upturn galaxies are defined by bluer FUV-NUV, hence brighter FUV at fixed r and NUV. At higher redshift the FUV flux limit preferentially removes UV weak galaxies and inflates the apparent upturn fraction. Including log M* in the regression does not cure this, because the selection is in apparent FUV magnitude, not in stellar mass. The volume-limited cut on M_r does not remove the bias either. That is a load-bearing flaw for the redshift dependence, and the stress-test note is right to focus on it.\n\nThe internal extinction migration is a smaller but similar problem. The authors estimate tens of percent changes to the class fractions, but they do not propagate those boundary shifts through the fitted mass and redshift trends. If the migration is asymmetric in z or mass, the joint trends shift. Also, several per-class subsamples are tiny for a six-parameter logistic model; the wide credible intervals honestly show this, but the paper should report posterior checks or model comparison.\n\nBottom line: this is a real measurement, mostly well analyzed, but the abstract overstates a non-secure feature and the selection function is not handled. It deserves peer review, not desk rejection. I would send it to a referee and ask for a selection treatment or at least a much more cautious redshift claim.","headline":"A plausible but not secure measurement of UV upturn prevalence among field red-sequence galaxies; the mass trend is the robust part, while the claimed redshift infall and uncorrected selection need work.","tokens_in":757,"tokens_out":975,"would_cite":false,"duration_ms":28574,"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":"The proportion of UV upturn galaxies among red-sequence galaxies rises with stellar mass and peaks near z≈0.25.","keywords":["UV upturn","red-sequence galaxies","Bayesian logistic regression","WHAN diagram","GALEX","GAMA survey","galaxy evolution","stellar mass"],"falsifier":"Re-run the same sample and model after applying the paper's own internal-extinction corrections ($A_{\\rm FUV}=2.536A_V$, $A_{\\rm NUV}=2.045A_V$, $A_r=0.8695A_V$, with $A_V\\approx0.2$ for UV weak systems and lower for UV upturn systems) and after adding a term for GALEX FUV detection probability. If the $z\\approx0.25$ peak and the rising mass trend in retired/passive galaxies disappear or reverse under these corrections, the central claim fails; if they persist, it would be corroborated.","tokens_in":24156,"feed_emoji":"🔭","tokens_out":7608,"duration_ms":66684,"temperature":0.7,"pith_summary":"This paper tries to establish how common the ultraviolet (UV) upturn is among red-sequence galaxies that are bright enough to be seen in the far-UV, and how that incidence changes with redshift, stellar mass, and emission-line classification. Using a matched GALEX/SDSS/GAMA sample spanning z=0.06 to 0.40, the authors define UV-bright red-sequence galaxies by the adopted colour cuts and model the probability of hosting a UV upturn with a Bayesian logistic regression. They find that among retired/passive galaxies the fraction of UV upturn hosts rises from z≈0.06 to z≈0.20–0.25 and then falls convincingly to z≈0.35, and that the fraction rises with stellar mass across the sampled range. The result matters because it turns the UV upturn from a property of individual ellipticals into a population statistic that changes on gigayear timescales and with galaxy mass, giving stellar-population models a target to explain.","feed_headline":"UV upturn peaks near z=0.25, rises with mass","feed_subtitle":"Bayesian model of 506 UV-bright red-sequence galaxies shows retired/passive systems drive both trends.","key_machinery":"The analysis rests on a Bayesian logistic regression for the binary outcome 'UV upturn versus UV weak'. The logit of the probability is modelled as a second-degree polynomial in redshift and stellar mass, with a hierarchical term for WHAN emission-line classes, so the dependence on $z$ and $\\log M_\\star$ is estimated jointly rather than in bins. The other load-bearing piece is the adopted colour-class cut: $(NUV-r)>5.4$ picks UV-bright red-sequence galaxies, and $(FUV-NUV)<0.9$ with $(FUV-r)<6.6$ separates UV upturn hosts from UV weak ones. Stratifying by WHAN classes (star-forming, strong and weak AGN, retired/passive, unclassified) is what lets the paper attribute the overall trends to quiescent galaxies.","core_discovery":"The central claim is that the probability that a UV-bright red-sequence galaxy manifests the UV upturn is not constant: it depends on both stellar mass and redshift, and the cleanest, most statistically reliable trend is in retired/passive (quiescent, lineless) systems. In those systems the upturn fraction increases with stellar mass over roughly $\\log M_\\star \\approx 10$–$11.5$ and, in redshift, rises to a maximum around $z\\approx0.20$–$0.25$ before declining out to $z\\approx0.35$; the paper states that this decline is clearly seen. Broadly, the same shape appears in the runs without emission-line stratification, while star-forming and unclassified galaxies show weaker or less certain trends, and weak/strong AGN classes are too scarce and boundary-affected to support a claim. The authors also report that the photometric criteria they use are robust against bona fide AGN contamination, and that even within a volume-limited subsample the same qualitative trends survive.","pith_inferences":["Beyond the paper, one testable extension is to replace stellar mass with central velocity dispersion in the same logistic model; if the mass trend is really about the depth of the potential well, the dispersion version should show a tighter or steeper relation than seen here.","The redshift peak could partly reflect GALEX MIS depth and large UV k-corrections; a direct check is to repeat the analysis with a sample from the deeper GALEX DIS fields, where FUV detectability is higher at fixed $z$ and mass.","The analysis only covers UV-bright red-sequence galaxies. The same Bayesian machinery applied to the full red sequence, including FUV non-detections modelled through survival analysis or a detection-probability term, would tell whether the reported trends hold for the whole quiescent population or only the UV-bright tail.","If the in-fall after $z\\approx0.25$ is physical, it would predict that the hot stellar populations responsible for the upturn were more common or hotter at lookback times of roughly 3 Gyr; a stellar-population synthesis model with an evolving fraction of extreme horizontal branch stars could be fit to the observed fractions to quantify that evolution."],"forward_implications":["If the rising-with-mass trend is real, the UV upturn is more frequent in more massive quiescent galaxies, which points toward hot evolved stellar populations becoming more prominent in deep potential wells.","The redshift peak near $z\\approx0.20$–$0.25$ implies the occurrence rate of the UV upturn within UV-bright red-sequence galaxies changes over a lookback time of roughly 2.5–3 Gyr, so it is an evolutionary quantity, not a fixed galaxy property.","The absence of a reliable trend for weak and strong AGN classes means the adopted UV classification is not being driven by AGN activity; any explanation of the UV upturn does not need an AGN component.","The paper's fractions are likely underestimated, not overestimated, if internal extinction is corrected for, because more UV weak systems migrate into the UV upturn class than leave it."],"supporting_citations":[{"why":"Defines the colour cuts that classify UV-bright red-sequence galaxies into UV weak and UV upturn samples.","marker":"Yi et al. (2011)"},{"why":"Supplies the stellar-mass estimates used as the mass covariate in the regression.","marker":"Taylor et al. (2011)"},{"why":"Defines the WHAN diagram classes used to stratify the sample by emission-line type.","marker":"Cid Fernandes et al. (2011)"},{"why":"Computes the k-corrections applied to the SDSS and GALEX photometry before colour classification.","marker":"Blanton & Roweis (2007)"},{"why":"Provides the MagPhys SED-fitting outputs used to estimate dust masses and check internal extinction effects.","marker":"Da Cunha et al. (2008)"},{"why":"Gives the extinction-law coefficients used in the appendix to estimate how internal dust shifts the classification boundaries.","marker":"Werle et al. (2019)"},{"why":"Releases the GAMA-DR3 catalogue and cross-matching tables from which the primary sample is drawn.","marker":"Liske et al. (2015)"},{"why":"Provides the cluster-based redshift evolution of the UV upturn that this paper contrasts with its own field-sample result.","marker":"Ali et al. (2018c)"}],"fun_headline_variants":["UV upturn peaks at z~0.25, climbs with stellar mass","Passive galaxies drive UV upturn peak near z=0.25","Red-sequence UV upturn: peak z~0.25, mass boost","UV upturn fraction: mass up, z~0.25 peak"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim stands or falls on the assumption that the adopted colour cuts, applied to k-corrected GALEX/SDSS photometry without correcting for each galaxy's own dust, correctly separate true UV upturn galaxies from UV weak ones at every redshift and mass, and that FUV detection does not selectively exclude red-sequence galaxies.","fun_headline_variants_meta":{"raw":{"variants":["UV upturn peaks at z~0.25, climbs with stellar mass","Passive galaxies drive UV upturn peak near z=0.25","Red-sequence UV upturn: peak z~0.25, mass boost","UV upturn fraction: mass up, z~0.25 peak"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001236,"raw_usage":{"total_tokens":5135,"prompt_tokens":1066,"completion_tokens":4069,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":682,"completion_tokens_details":{"reasoning_tokens":3988}},"tokens_in":682,"tokens_out":4069,"duration_ms":27107,"temperature":1.0,"reasoning_tokens":3988,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:35:34.903718+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the same sample and model after applying the paper's own internal-extinction corrections ($A_{\\rm FUV}=2.536A_V$, $A_{\\rm NUV}=2.045A_V$, $A_r=0.8695A_V$, with $A_V\\approx0.2$ for UV weak systems and lower for UV upturn systems) and after adding a term for GALEX FUV detection probability. If the $z\\approx0.25$ peak and the rising mass trend in retired/passive galaxies disappear or reverse under these corrections, the central claim fails; if they persist, it would be corroborated.","supporting_citations":[],"review_version":1}