{"id":"ea4bc8e3-d138-48e9-8281-e9aefdf16c89","arxiv_id":"2505.24395","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A combined PSF-scattering correction and exponential profile fitting pipeline estimates intra-halo light fractions in galaxy groups, and mock tests show the PSF correction is essential to avoid overestimates at low IHL fractions.","lead":"This paper presents two new techniques: one removes telescope point-spread-function scattered light from images, and the other fits an exponential model to measure intra-halo light around galaxy groups. Tests on 5,440 mock images show the PSF correction avoids large overestimates of faint group light, and the method is applied to one real group.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Mock IHL is injected after PSF convolution, so the validation omits PSF blurring of the target; the unconvolved fitting model may bias the real G400138 IHL fractions.","rationale":"I read the paper's central claim as: PSF-scattered flux removal significantly improves IHL recovery in mocks and yields unbiased fractions for G400138. The strongest support is the controlled comparison in Sec 5.2. However, the mock injection order in Sec 4 creates an environment in which the IHL is not convolved by the PSF, while real data always are. The IHL estimator does not convolve the model with the PSF either. This is an internal consistency issue: the data model used for validation differs from the data model of the real observations in a way that directly affects the fitted parameters. The reader's concern about the circular exponential assumption is real but is partially mitigated by the Sec 5.3 elliptical test and by the paper's explicit caveats; the missing PSF convolution is not mentioned anywhere. I therefore regard this as the most load-bearing concern. The proposed test is straightforward because ProFit supports PSF convolution and the mock-generation pipeline is described in enough detail to reorder the injection. If the test shows similar biases, the paper's claims stand largely unchanged; if not, the G400138 fractions and the reported recovery accuracy require revision. I maintain the CONDITIONAL verdict because the methodological comparison (with vs without PSF correction) may still hold, but the specific numerical conclusions need verification.","tokens_in":27133,"tokens_out":9160,"duration_ms":110265,"concrete_test":"Re-run the mock suite with the IHL injected before the PSF convolution step so the final images contain a PSF-convolved IHL, and fit with a PSF-convolved Sérsic model (ProFit accepts a PSF image). Compare median Flux_inj/Flux_fitted and Reff_inj/Reff_fitted at f_IHL=0.01, 0.1, and 0.5 against Figs 9 and 11. If the biases remain within roughly a factor of two of the quoted values, the validation is robust; if they shift substantially, the headline recovery numbers and the G400138 f_IHL values should be re-derived. Also refit G400138 with the PSF-convolved model and recompute f_IHL from Eq. (2)/(3) to check whether the medians move by more than the quoted uncertainties.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Sec 4, the mock construction first convolves the HSC cutouts with the PSF and adds noise, and only then injects the IHL: 'We then convolve the HSC cutouts ... with our ... extended PSF model ... Finally, we add image noise ... we inject the corresponding IHL component into each mock group.' The injected IHL is therefore not PSF-convolved. Correspondingly, the IHL estimator in Sec 3 Steps v-vi fits a circular Sérsic n=1 template with no PSF convolution step, and ProFit is not given a PSF image for these fits. As a result, the mocks only test recovery of a sharp, unconvolved IHL contaminated by source-scattered light; they do not test the PSF effect on the IHL itself. In real HSC data the diffuse IHL is convolved with the PSF, so an unconvolved exponential model is misspecified. The quoted recovery biases (Sec 5.2: flux 9.6% low, R_eff 12.2% low at f_IHL=0.01) and the G400138 fractions (Sec 6) may therefore be optimistic or biased, not because of the exponential-profile choice the reader flagged, but because the model and mocks omit a basic physical process. The paper's caveats in Sec 2.3 address oversubtraction of source PSF wings, not convolution of the target component.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents two techniques for measuring intra-halo light in galaxy groups and clusters: a ProFound-based routine that rescales source fluxes and subtracts PSF-scattered flux outside source segments (Sec. 2), and a ProFit-based MCMC fit of a circular exponential (Sérsic n=1) IHL model to the masked, PSF-corrected image (Sec. 3). The methods are tested on 5440 HSC-SSP PDR3 mock images of GAMA groups with injected circular exponential IHL components at fractions f_IHL = 0.01-0.5. The mocks show that without PSF correction the fitted IHL flux can be ~100 times too high and the effective radius ~10 times too large at f_IHL=0.01, while after correction the bias is reduced to ~9.6% in flux and ~12.2% in Reff at f_IHL=0.01 (Sec. 5.2). The pipeline is then applied to the real GAMA group G400138 in HSC-UD g,r,i, yielding median PSF-corrected fractions f_g,IHL~0.19, f_r,IHL~0.08, f_i,IHL~0.06 (Sec. 6).","tokens_in":27437,"tokens_out":9596,"duration_ms":123478,"significance":"If the central claim holds, the proposed combination of PSF-scattered-light removal and single-component exponential fitting offers a computationally efficient, automatable route to IHL measurements in large deep surveys, and the 5440-mock controlled test is a useful resource. The paper is transparent about several caveats, including the up to 4% oversubtraction of intrinsic source flux in Sec. 2.3, unreliable recovery at f_IHL=0.01-0.05 in Sec. 5.2, and the limitations of a circular exponential model in Sec. 5.3. The use of publicly available software (ProFound, ProFit) and a detailed mock-generation recipe supports reproducibility. However, the external validity of the recovery biases and of the real G400138 measurement is limited by two modelling gaps identified in the major comments.","major_comments":[{"comment":"The mock validation does not include the PSF convolution of the target IHL. In Sec. 4, the construction sequence is: convolve the HSC cutouts (with galaxy segments and injected point sources) with the extended PSF model, add Gaussian noise, and only then 'inject the corresponding IHL component into each mock group'. In Sec. 3 Step v, ProFit fits an unconvolved circular Sérsic n=1 template and is not given a PSF image. The mocks therefore test recovery of a sharp, unconvolved IHL that is contaminated by PSF-scattered source flux, not the PSF-convolved IHL that is present in real HSC data. The quoted recovery biases in Sec. 5.2 (9.6% low flux and 12.2% low Reff at f_IHL=0.01) and the G400138 fractions in Sec. 6 are thus not directly transferable to real data, because the target component is misspecified there. I recommend adding a mock variant in which the IHL is injected before the PSF convolution (and/or fitting with a PSF-convolved ProFit model) and rerunning the recovery and G400138 analyses.","section":"Sec. 4 / Sec. 3 Step v"},{"comment":"The recovery test is internally consistent but does not validate the exponential model: the injected IHL is a circular exponential (Sec. 4, 'we select a circular exponential model') and the fitting template is the same functional form (Sec. 3 Step v). The one morphological robustness test in Sec. 5.3 varies only ellipticity (Arat=0.5, theta=45 deg) while keeping an exponential radial profile. Appendix B claims 'supporting evidence' for an exponential model by fitting an exponential to the SB-limit IHL of G400138 and extrapolating that same fit into the core; this is not an independent test. The real-data f_IHL values in Sec. 6 should therefore be presented as conditional on the exponential-profile assumption, or the authors should add tests with Sérsic n != 1, substructure, or a realistic simulated IHL morphology to quantify the resulting bias.","section":"Sec. 3 / Sec. 5.3 / Appendix B"},{"comment":"The uncertainties quoted for the real G400138 fractions are fitting uncertainties only; the three methods (ProFit fixed, ProFit free, SB-extrapolated) share the same PSF-correction pipeline and the same exponential-model assumption, so their spread does not encompass the dominant systematics discussed in Secs. 2.3 and 5.3. The reported medians f_g,IHL~0.19, f_r,IHL~0.08, f_i,IHL~0.06 and their asymmetric ranges would be more robust with an explicit systematic error term from the PSF oversubtraction (up to 4% source flux) and from the model-choice sensitivity.","section":"Sec. 6 / Table 2"}],"minor_comments":[{"comment":"There is a typo: 'profile profile' should be 'profile'.","section":"Sec. 2.3"},{"comment":"The chronological order of noise addition and IHL injection is described ambiguously: 'Finally, we add image noise...' appears before the paragraph describing the IHL injection. Please clarify whether the noise is added before or after the IHL injection, since this matters for interpreting the mock construction and the PSF-convolution issue raised above.","section":"Sec. 4"},{"comment":"The conversions from log-ratios to percentages (9.6% low flux and 12.2% low Reff at f_IHL=0.01) are not exactly consistent with the displayed log values; please verify the arithmetic or the quoted log values.","section":"Sec. 5.2"},{"comment":"There are typos: 'uncertanties' should be 'uncertainties' and 'accross' should be 'across'.","section":"Sec. 5.2 / Fig. 10"},{"comment":"In the software section, 'Robotham 023a' should be 'Robotham 2023a'.","section":"Software / References"}],"recommendation":"major_revision","confidential_remarks":"The PSF-convolution omission in the mock construction is the main substantive issue; it is fixable within the current scope by injecting IHL before the PSF convolution or by supplying a PSF to ProFit. The circularity of the exponential-model choice is partially acknowledged, but the Appendix B claim should be softened or supported independently. The paper is otherwise carefully written and the PSF-correction concept is promising; I would support a major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Main takeaway: this is a genuinely useful methods paper for the LSB/IHL community, but the mock validation has a blind spot that matters more than the authors' own caveats: the injected IHL is not PSF-convolved, and the fitting model is unconvolved, so the recovery tests don't exercise the PSF effect on the component being measured.\n\nWhat is new: the specific pipeline has a clean design. The PSF-scattered flux removal step rescales each detected source by C = flux_obs/flux_conv, then subtracts a reconvolved scattered-light image from the surroundings. The second step fits a circular exponential IHL with an MCMC optimiser in ProFit. The combination of these two steps is not in the cited literature. On top of that, the mock campaign is large (5440 images) and carefully described, and the authors are honest about known limitations: up to 4% oversubtraction of intrinsic source flux, unreliable recovery at f_IHL = 0.01–0.05, and some pathological cases flagged in Appendix A. Without PSF correction, the fitted flux is overestimated by up to a factor of 100 and the effective radius by a factor of 10 at f_IHL = 0.01; after correction, the recovered parameters are within ~10–12% at that fraction. The qualitative point—PSF-scattered source light severely contaminates IHL measurements—is convincingly demonstrated.\n\nThe main soft spot, which I think is load-bearing, is the stress-test concern. Section 4 says the cutouts are convolved with the PSF and noise is added before the IHL component is injected. Section 3 fits a circular exponential Sersic n=1 model with no PSF convolution step in ProFit. So the mocks only validate recovery of a sharp IHL contaminated by source-scattered light; they do not validate recovery of a PSF-convolved IHL. In real HSC data, the IHL itself is convolved, so an unconvolved exponential model is misspecified. The quoted recovery biases (flux 9.6% low, R_eff 12.2% low at f_IHL = 0.01) may not transfer to the real G400138 measurement. Section 2.3's caveats address oversubtraction of source PSF wings, not convolution of the target component. This is not fatal to the PSF-removal technique itself, but it is load-bearing for the absolute accuracy claims and for the real-data fractions.\n\nThe other soft spots are milder. The model choice is circularly validated: a circular exponential is injected and fitted, and Appendix B fits an exponential to the same group's IHL defined by that fit, so it is not an independent test of the exponential assumption. The elliptical test in Sec. 5.3 does show flexibility, but only for that one shape. The abstract says \"statistically robust\" while the paper itself says recovery is unreliable at f_IHL = 0.01–0.05; that claim should be tempered. Reproducibility is somewhat limited because the WAVES catalogue is not yet public, though the code and HSC images are available.\n\nWho this is for: anyone working on intra-halo/intra-cluster light, low-surface-brightness science, or PSF correction in wide surveys. This deserves a serious referee. I would send it to review with a request to rerun or supplement the mocks with a version where the IHL is PSF-convolved before fitting, and to fit with a PSF-convolved model when applying to real data. After that revision, it would be a solid methods contribution worth citing.","headline":"Useful PSF-correction pipeline for IHL work, but the mock validation omits PSF convolution of the injected IHL, so the real-galaxy fractions are less certain than the abstract implies.","tokens_in":27991,"tokens_out":4337,"would_cite":true,"duration_ms":58468,"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":"Removing telescope-scattered light first shrinks a 100-fold error in faint galaxy light to under 10%.","keywords":["galaxies: evolution","galaxies: haloes","galaxies: clusters: intracluster medium","instrumentation: detectors","methods: data analysis","intra-halo light"],"falsifier":"Take a set of galaxy-group mocks like those used here but inject an elliptical or substructured IHL (e.g., axial ratio 0.5 with a tidal stream), run the full PSF-correction-plus-exponential-fit pipeline, and check whether the recovered flux and radius deviate from the injected values by more than the paper's quoted ~10%/12% biases at f_IHL = 0.01; any substantially larger deviation would falsify the claim that the method gives unbiased IHL measurements for realistic morphologies.","tokens_in":26950,"feed_emoji":"🔭","tokens_out":12377,"duration_ms":123363,"temperature":0.7,"pith_summary":"The paper argues that measurements of intra-halo light (IHL) — the diffuse stellar component around galaxy groups and clusters — are severely biased unless telescope-scattered light from the point spread function (PSF) is subtracted before any modelling is done. It presents two techniques that together aim to make IHL measurements robust: a PSF-scattered flux removal step, and an MCMC fit of a circular exponential profile to the remaining diffuse light. The claim is supported with 5440 mock Hyper Suprime-Cam observations of GAMA groups with injected IHL fractions from 1% to 50%. Without the PSF correction, the fitted IHL flux is overestimated by up to a factor of 100 and the effective radius by a factor of 10 at the faintest injected fraction; after correction the fitted flux is 9.6% low and the effective radius 12.2% low. Applied to the real GAMA group G400138, the method yields median IHL fractions of about 19% in g, 8% in r, and 6% in i, a redward decline consistent with earlier work.","feed_headline":"Scattered-light cleanup shrinks faint-light error from 100x to 10%","feed_subtitle":"Removing telescope-scattered light first cuts the fitted flux bias to under 10 percent in mock tests.","key_machinery":"The machinery is a two-stage image-processing pipeline. Stage one estimates the intrinsic flux of every detected source by comparing its observed flux with the flux remaining inside its segmentation map after a PSF convolution, rescales each source by the ratio of those fluxes, then convolves the rescaled image with the PSF model and subtracts it from the original image outside the source segments — this yields an image whose diffuse background is nominally free of scattered light. Stage two masks all detected sources and fits the remaining light with ProFit's circular exponential Sérsic template (n = 1, axial ratio 1, position angle 0), leaving only magnitude and effective radius as free parameters, optimised with the Highlander genetic/MCMC algorithm. The same template is used to inject mock IHL components, whose effective radii are tied to halo mass via the r_IHL–M200 relation from Proctor et al. 2024. The comparison between injected and fitted magnitude and effective radius across 5440 mocks, with and without stage one, is what carries the argument.","core_discovery":"The paper's central claim is that removing the PSF-scattered flux before fitting an exponential model is not optional for IHL measurements — it is the step that turns order-of-magnitude errors into near-percent-level ones. In the authors' mock suite, without PSF correction the fitted IHL flux at an injected fraction f_IHL = 0.01 is on average ~100 times larger than the injected value and the effective radius ~10 times larger; after applying their PSF-correction pipeline, the same mock fits come in 9.6% low in flux and 12.2% low in radius at f_IHL = 0.01, with the bias shrinking to 3.5% and 0.28% at f_IHL = 0.5. The authors interpret the remaining small deficits as a known oversubtraction caused by the correction itself. For the real group G400138, using the same PSF-corrected pipeline, they report median intra-halo light fractions of f_g,IHL ~ 0.19, f_r,IHL ~ 0.08, and f_i,IHL ~ 0.06, which lie in the range of previous measurements and reproduce the trend of lower IHL fractions at redder wavelengths.","pith_inferences":["Inference: because the mocks inject exactly the functional form the fitter assumes, the quoted 9.6% and 12.2% biases are best-case; real IHL with ellipticity or tidal substructure will recover less accurately, as the paper's own elliptical-mock test shows roughly half the flux missed at f_IHL = 0.5.","Inference: the pipeline's remaining bias is a systematic flux deficit, so a simple empirical correction factor derived from the mock suite could produce unbiased f_IHL values in real applications, at the cost of added scatter.","Inference: the r_IHL–M200 relation used to set mock sizes is simulation-based; the same pipeline applied to a large real sample with independent mass estimates could test that relation observationally.","Inference: applied to upcoming deep surveys, the method could measure IHL fractions for thousands of groups, but the single-exponential assumption should be diagnosed with residual maps to avoid mistaking unmodelled substructure for background noise."],"forward_implications":["IHL fractions measured from survey images without PSF subtraction are systematically overestimated, with the largest errors at the faintest IHL fractions, so previously published IHL fractions based on PSF-uncorrected images may need revision.","With the correction in place, automated exponential fits recover the injected IHL flux and size to within about 10% at f_IHL = 0.01 and to within a few percent at higher fractions, making the pipeline suitable for large survey samples and stacked IHL analysis.","For the real group G400138, the reported fractions (fg ~ 0.19, fr ~ 0.08, fi ~ 0.06) imply that roughly one-fifth of the g-band light but only about 6% of the i-band light is diffuse intra-halo light, a colour trend consistent with the idea that IHL is built from tidally stripped stellar populations.","The PSF-removal technique is not specific to IHL; it can be applied to any low-surface-brightness measurement in astronomical images, and the mock suite doubles as a calibration of the residual biases."],"supporting_citations":[{"why":"Supplies ProFound source detection and segmentation, used both in the PSF-scattered-flux removal and in masking sources before the IHL fit.","marker":"Robotham et al. 2018"},{"why":"Supplies ProFit, the Bayesian 2D profiling tool used to fit the circular exponential IHL model and to generate mock IHL images.","marker":"Robotham et al. 2017"},{"why":"Presents the extended r-band HSC-SSP PDR3 PSF model used to convolve the mock images and to remove scattered flux in real images.","marker":"Garate-Nuñez et al. 2024"},{"why":"Provides the r_IHL–M200 relation used to set the effective radii of the injected mock IHL components.","marker":"Proctor et al. 2024"},{"why":"Provides the previous f_IHL estimates for G400138 that the paper compares its own measurements against.","marker":"Martínez-Lombilla et al. 2023"},{"why":"Supplies the G3Cv10 galaxy group catalogue defining the mock group configurations and the group member fluxes used in f_IHL.","marker":"Robotham et al. 2011"},{"why":"Supplies the HSC-SSP PDR3 imaging data used to build the mock cutouts and the real-group measurement.","marker":"Aihara et al. 2022"}],"fun_headline_variants":["PSF removal cuts IHL flux bias from 100x to under 10% in mocks","Without PSF cleanup, faint-light flux errors hit 100x in tests","Mock tests: PSF correction shrinks IHL measurement errors to ~4%","New pipeline: PSF first, then fit — IHL bias drops to single digits","For faint intra-halo light, PSF correction isn't optional"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The mock validation assumes real intra-halo light is a smooth circular exponential disc whose size is set by the simulation-based r_IHL–M200 relation, and that the same functional form is the one used for fitting; if real IHL has different shapes or substructure, the quoted recovery biases and the G400138 fractions would not transfer directly.","fun_headline_variants_meta":{"raw":{"variants":["PSF removal cuts IHL flux bias from 100x to under 10% in mocks","Without PSF cleanup, faint-light flux errors hit 100x in tests","Mock tests: PSF correction shrinks IHL measurement errors to ~4%","New pipeline: PSF first, then fit — IHL bias drops to single digits","For faint intra-halo light, PSF correction isn't optional"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000257,"raw_usage":{"total_tokens":1719,"prompt_tokens":1228,"completion_tokens":491,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":844,"completion_tokens_details":{"reasoning_tokens":384}},"tokens_in":844,"tokens_out":491,"duration_ms":5846,"temperature":1.0,"reasoning_tokens":384,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T12:23:14.724940+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a set of galaxy-group mocks like those used here but inject an elliptical or substructured IHL (e.g., axial ratio 0.5 with a tidal stream), run the full PSF-correction-plus-exponential-fit pipeline, and check whether the recovered flux and radius deviate from the injected values by more than the paper's quoted ~10%/12% biases at f_IHL = 0.01; any substantially larger deviation would falsify the claim that the method gives unbiased IHL measurements for realistic morphologies.","supporting_citations":[{"cited_title":"P., Robotham A","cited_arxiv_id":null,"evidence_quote":"Presents the extended r-band HSC-SSP PDR3 PSF model used to convolve the mock images and to remove scattered flux in real images."}],"review_version":1}