{"id":"75f6b22f-959c-4679-8ff0-42dffed04061","arxiv_id":"2504.15346","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"At cosmic noon, central dust attenuation flattens galaxy light profiles, making massive dusty galaxies appear about 30 percent larger in rest-optical than in rest-near-infrared sizes.","lead":"Using JWST imaging and spectroscopy of 141 galaxies at cosmic noon, this study links dust attenuation to the sizes and shapes of galaxies. It finds that massive dusty galaxies appear about 30 percent larger in rest-optical light than in rest-near-infrared light, because central dust flattens the optical light profile.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Size-ratio trend in Fig. 11 is not controlled for redshift-dependent rest-frame wavelengths; the F150W/F444W baseline shifts with z, so the Av correlation may be partly a redshift artifact.","rationale":"The reader's weakest assumption is the dust model (fixed n1 = -1, Eb tied to n2), which mainly affects the secondary claim about attenuation-law diversity and the exact Av values. The headline claim, however, is the wavelength-dependent size gradient. The size ratio itself is directly measured, so the dust-model concern is one step removed from the central result. A more direct threat to the causal interpretation is the uncontrolled redshift dependence in Figure 11. Because F150W and F444W are fixed observed bands, the rest-frame wavelengths they probe vary by roughly a factor of two across the sample's redshift range; F150W moves from ~0.5 micron at the low-z end to ~0.33 micron at the high-z end, and F444W moves from ~1.6 micron to ~1.0 micron. Since dust attenuation is stronger at shorter wavelengths, and since stellar population gradients alone can produce size ratios of the observed magnitude, the correlation between size ratio and Av could be induced by the redshift distribution. This concern is concrete, testable with the existing data, and does not require redoing the SED fits. It does not overturn the paper, but it does mean the central claim is conditional on controlling for this effect, matching the reader's CONDITIONAL verdict.","tokens_in":32330,"tokens_out":5229,"duration_ms":51137,"concrete_test":"Restrict the Figure 11 analysis to narrow redshift bins (e.g., 1.7 < z < 2.3, 2.3 < z < 2.9, 2.9 < z < 3.5). Within each bin, compute the Spearman rank correlation between log(RF150W/RF444W) and Av with bootstrap uncertainties (1000 resamples), and fit log(RF150W/RF444W) as a bivariate function of Av and z. If the correlation is not significant at 95% confidence in each bin, or if the bivariate fit attributes the trend primarily to z rather than Av, the dust-driven interpretation of the size gradient would not be supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that massive galaxies appear ~30% larger in rest-optical than rest-NIR because central dust flattens optical light profiles—rests on the correlation between log(RF150W/RF444W) and Av in Figure 11 (Section 6.4). The load-bearing step is the causal attribution, and it is insecure because Figure 11 pools galaxies across the full redshift range 1.7 < z < 3.5 without binning or controlling for redshift. The rest-frame coverage of the two filters shifts with z: F150W probes 0.48–0.63 micron at z = 1.7 but 0.29–0.38 micron at z = 3.5, while F444W probes 1.4–1.85 micron at z = 1.7 and 0.84–1.11 micron at z = 3.5. Dust attenuation is stronger at shorter rest wavelengths, so even a fixed dust column produces a larger size ratio at higher z. If the massive, high-Av galaxies in Figure 11 are preferentially at higher redshift, the trend with Av could be a redshift artifact rather than evidence for central dust flattening. In addition, stellar population gradients (old centers, young outskirts) generate wavelength-dependent sizes in the same direction without any dust. The paper reports no Spearman coefficient, fit, or significance for Figure 11, so the visual trend is hard to evaluate quantitatively.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents a study of 141 Blue Jay galaxies at 1.7<z<3.5, combining JWST/NIRSpec R~1000 spectroscopy, NIRCam photometry, and HST photometry. Stellar populations and dust attenuation are modeled with Prospector using non-parametric SFHs and a two-component dust model: a birth-cloud power-law with fixed slope n1=-1 (Eq. 1) and an ISM attenuation law with a UV-bump strength tied to the ISM slope via Eb=0.85-1.9n2 (Eq. 4). The authors derive effective attenuation laws, report correlations between attenuation parameters and stellar mass, SFR, and surface densities, and compare stellar-continuum attenuation with Balmer-decrement nebular attenuation. Morphological fits with Pysersic in F150W, F356W, and F444W yield size-mass relations and a wavelength-dependent size ratio RF150W/RF444W; the central claim is that massive galaxies (M* >~1e10 Msun) appear about 30% larger in the rest-optical than in the rest-NIR because central dust flattens the optical light profile (Section 6.4, Figure 11).","tokens_in":32650,"tokens_out":8394,"duration_ms":81000,"significance":"If the size-gradient result holds, it is a valuable JWST-era measurement linking dust geometry to observed morphology at cosmic noon. The paper has real strengths: the size ratios are directly measured on 87-93 galaxies, the comparison with Balmer-decrement attenuation provides an independent dust tracer, and the authors are unusually candid about which trends are imposed by the model rather than required by the data (Section 5). The attenuation-law diversity and the mass-dust relation are interesting and well-contextualized with the literature. However, the headline causal claim that dust drives the 30% size excess is not yet backed by a controlled statistical analysis, and several quantitative statements rest on correlation coefficients reported without uncertainties.","major_comments":[{"comment":"The central size-gradient claim is not supported by a controlled statistical analysis. Figure 11 pools galaxies over the full range 1.7<z<3.5, while the rest-frame coverage of F150W and F444W shifts by a factor of about 1.7 across this range: F150W probes ~0.55 micron at z=1.7 but ~0.33 micron at z=3.5, and F444W probes ~1.6 micron at z=1.7 but ~1.0 micron at z=3.5. Because dust attenuation is stronger at bluer rest-frame wavelengths, and because stellar population gradients across the Balmer/4000 A break can also produce wavelength-dependent sizes, a trend of RF150W/RF444W with Av could arise partly from redshift-dependent wavelength sampling even with no change in central dust geometry. The paper reports no Spearman coefficient, fit, or significance for Figure 11, and the '~30%' statement is based on a visual trend. Please add redshift-binned versions of Figure 11 or a partial-correlation analysis controlling for redshift, quote correlation statistics with bootstrap uncertainties, check whether the massive high-Av galaxies preferentially lie at high z, and, if possible, use the Balmer-decrement attenuation as an independent dust indicator for the 67 galaxies with reliable measurements.","section":"Section 6.4, Figure 11"},{"comment":"The paper explicitly states in Section 5 that the trends of attenuation-law slope and UV-bump strength with SFR are 'by construction': the birth-cloud slope is fixed at n1=-1, and the ISM bump strength is tied to the ISM slope by Eb=0.85-1.9n2. Nevertheless, the abstract and conclusions (item i) present these as empirical findings of diversity in the attenuation law. As written, this portion of the headline result is a property of the fitting assumptions rather than an independent measurement. Please state this model-imposed nature in the abstract and conclusions, and perform a robustness check, for example leaving n1 free in a subset of galaxies or fitting the bump strength independently, to show whether the data actually require the reported variation. This matters for the derived Av values that are used in the size-gradient interpretation.","section":"Section 5, Eqs. (1) and (4)"},{"comment":"The causal attribution of the size ratio trend to central dust is not unique. Stellar population gradients (old centres with young outskirts) produce the same sign of wavelength-dependent size ratio without any dust, and the paper invokes exactly this mechanism for low-mass galaxies in the same section. Since high-Av massive galaxies likely have different SFHs from low-Av systems, the observed correlation could be partly due to stellar population gradients that correlate with Av. A quantitative test would be to compare rest-optical and rest-NIR Sersic indices or to use spatially resolved SED information to separate dust from stellar population effects; at minimum, the discussion should quantify how much of the 30% effect could be explained by stellar population gradients alone.","section":"Section 6.4"},{"comment":"Many Spearman rank coefficients are quoted without uncertainties (e.g., r=0.532 and r=0.667 in Figure 6; the r values in Figures 7, 12, and 13), and claims such as 'the surface densities correlate more strongly than the global quantities' are based on comparing coefficients from 16-36 galaxies per bin without error bars. Please provide bootstrap or posterior uncertainties for all quoted correlation coefficients and temper comparative statements unless the difference is statistically significant. This is partly a presentation issue, but it affects the quantitative conclusions about which galaxy properties best predict dust attenuation.","section":"Sections 5, 6.1, and 7"}],"minor_comments":[{"comment":"In the description of the Drude profile, 'Delta lambda is is the FWHM' contains a duplicated 'is'.","section":"Section 3.2"},{"comment":"The caption appears to print 'n1 = 1' and 'n2 = 0.7' without the minus signs that the text and the panel labels use (n1=-1, n2=-0.7). Please correct the caption typography.","section":"Figure 1 caption"},{"comment":"The phrase 'star dust geometry' should be 'star-dust geometry' for consistency with the main text.","section":"Abstract"},{"comment":"The word 'anomolous' is a typo for 'anomalous'.","section":"Section 7"},{"comment":"The SFR-bin labels in Figure 3 are garbled in the text version (e.g., '10 2.3<SFR [yr 1]<100.6'); please ensure the rendered labels display the intended powers of ten.","section":"Figure 3"},{"comment":"The sample-flow numbers (141 total, 137 with attenuation laws, 104/98 with morphological fits, 93/87 star-forming with morphology, 67 with Balmer decrements) are scattered across the text; a single sample-selection table or flowchart would make the analysis much easier to follow.","section":"Sections 2 and 3.5"}],"recommendation":"major_revision","confidential_remarks":"The authors are transparent about the model-imposed nature of several trends, which I view favorably. The main load-bearing issue is the lack of redshift control and statistical quantification in Figure 11; this is fixable within the scope of the manuscript and does not appear to warrant rejection. The paper fits the journal's scope well."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a solid, mostly honest paper from the Blue Jay group. The genuinely new result is the wavelength-dependent size gradient: massive galaxies at 1.7<z<3.5 appear ~30% larger in rest-optical (F150W) than rest-NIR (F444W), and the authors tie this to dust attenuation. Suess et al. 2022 saw the size ratio but didn't connect it to dust, so the attribution is new. They also show the attenuation law depends on Av and stellar population properties, and that surface densities correlate more tightly with Av than global quantities. The Balmer decrement comparison gives an independent check on the SED-derived attenuation.\n\nWhat's good: the data are high quality, the modeling is standard (Prospector + Pysersic), and the paper is transparent about where the model drives the answer. They explicitly say that the slope/bump correlations with SFR in Fig. 6 are by construction, which is exactly the kind of admission you want to see. The size-mass plots in Fig 10 are binned in redshift, which shows the rest-optical vs rest-NIR difference persists within bins, so the effect is not just a broad redshift artifact.\n\nSoft spots, in proportion:\n\n- The central causal claim in Fig. 11—that Av drives the size ratio—is not well controlled for redshift. Figure 11 pools all redshifts and no Spearman coefficient is given. Since F150W and F444W probe bluer rest wavelengths at higher z, a fixed dust column yields a larger size ratio at high z. If the massive, high-Av points are preferentially at high z, the Av trend could be partly a redshift artifact. The authors bin by z elsewhere but not here. This is the biggest weakness, but it's addressable, not fatal: the Fig. 10 redshift-binned slopes suggest the size gradient exists at fixed z.\n\n- The attenuation-law shape vs SFR trends are partly built in: n1 is fixed to -1 and Eb is tied to n2. So the diversity of attenuation laws is narrower than it looks. They admit this, but it means some of the Section 5 claims are model outputs, not new measurements.\n\n- Many Spearman coefficients are quoted without uncertainties, and sample sizes in some bins are small (8-21 galaxies). Minor.\n\n- No code or data release is mentioned; for a survey paper that's a smaller issue, but it would help to make the morphological and SED fits public.\n\nOverall: a thoughtful, careful paper with a new result that will be cited. The size-gradient claim needs a redshift-controlled version of Fig. 11 to be fully convincing, but the data and analysis are solid enough to deserve peer review. I'd send it to a referee and ask for that extra control, plus error bars on the Spearman statistics. Recommend engaging.","headline":"A careful JWST study that makes a plausible case for dust-driven rest-optical size inflation at cosmic noon, but the load-bearing Av correlation in Fig. 11 lacks redshift control and is partly model-built.","tokens_in":33234,"tokens_out":2875,"would_cite":true,"duration_ms":26631,"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":"Massive dusty galaxies at cosmic noon appear about 30 percent larger in rest-optical than rest-NIR light because central dust flattens their optical light profiles.","keywords":["dust attenuation","galaxy morphology","cosmic noon","size-mass relation","JWST","SED fitting","Balmer decrement","rest-frame optical/NIR sizes"],"falsifier":"Dust-correct the rest-optical images for the most massive galaxies (for example by dividing the F150W image by the fitted attenuation map, or by using a Balmer-decrement-based attenuation map) and re-measure half-light radii; if the ~30 percent optical-versus-NIR size excess is caused by central dust, the corrected optical/NIR size ratio should become consistent with unity.","tokens_in":32169,"feed_emoji":"🌌","tokens_out":14465,"duration_ms":115277,"temperature":0.7,"pith_summary":"This paper argues that dust does more than dim galaxies: it changes how their sizes appear, and it must be included when measuring morphology at cosmic noon. Using 141 galaxies at redshifts 1.7–3.5 with JWST spectra and deep imaging, the authors find that the effective dust attenuation law varies widely with stellar mass, star formation rate, and optical extinction. The sharpest result is that massive galaxies (about $10^{10}$ solar masses and above) look roughly 30 percent larger in rest-optical light than in rest-near-infrared light, because dust concentrated in their centres flattens the optical light profile. If correct, single-band size measurements overestimate the true stellar sizes of massive dusty galaxies, and size-mass relations at cosmic noon need dust-aware corrections.","feed_headline":"Central dust makes massive galaxies look 30% larger at cosmic noon","feed_subtitle":"The finding means single-band sizes at cosmic noon can misrepresent the stellar mass distribution.","key_machinery":"The load-bearing machinery is the two-component dust attenuation model combined with flexible attenuation laws and multi-band morphological fitting. A power-law 'birth cloud' component with fixed slope $n_1=-1$ attenuates light from stars younger than 10 Myr and from nebular regions, while an ISM component with a free slope $n_2$ and a 2175 Å UV bump whose strength is tied to $n_2$ attenuates all stars, built on a flexible curve anchored to the local starburst law. The effective attenuation law for each galaxy is recovered by comparing dust-on and dust-free model spectra from the SED fit, so stellar-population, dust, and morphological effects can be separated. Half-light radii come from one-component Sérsic profile fits to NIRCam images in F150W, F356W, and F444W; the flattening of the rest-optical light profile by the fitted central attenuation is what produces the ~30 percent size excess.","core_discovery":"The central discovery is a wavelength-dependent size gradient produced by dust. For the most massive star-forming galaxies in the sample ($M_\\star \\gtrsim 10^{10}\\,M_\\odot$), the half-light radius measured in F150W—which probes rest-optical light—is on average about 30 percent larger than the half-light radius measured in F444W, which probes rest-NIR light. The excess grows with the fitted optical attenuation $A_V$, and the authors interpret it as central dust attenuating the inner regions more strongly, flattening the optical surface-brightness profile and pushing the effective radius outward. Lower-mass galaxies show a wide range of optical-to-NIR size ratios, which the authors attribute to either inside-out growth or central starbursts rather than dust. The paper also establishes that the shape and strength of the attenuation law vary systematically with stellar mass, SFR, and $A_V$; that $A_V$ correlates more tightly with stellar-mass and SFR surface densities than with global quantities; and that nebular attenuation from the Balmer decrement tracks stellar-continuum attenuation among 67 star-forming galaxies.","pith_inferences":["If the size gradient is really produced by central dust, single-band rest-optical sizes systematically overestimate the effective radii of massive dusty galaxies; evolutionary size-mass relations built from such bands would then need dust corrections before being read as structural growth.","Because the fitted attenuation law enforces a fixed birth-cloud slope and a tied bump-slope relation, part of the reported correlation between attenuation-law shape and star formation rate is imposed by the model; a testable next step is to fit steeper or decoupled dust laws and see whether the ~30% size excess survives.","Spatially resolved dust maps (for example Balmer-decrement gradients) for a subset of these galaxies could directly confirm the central dust gradient: if it is real, the dust-corrected optical light profile should recover the flatter, more extended shape that the attenuation model predicts.","The wide optical-to-NIR size ratios of lower-mass galaxies, interpreted as inside-out growth or central starbursts, could be separated by measuring rest-UV/optical colour gradients; this would distinguish dust-free structural differences from the dust-driven signal seen at high mass."],"forward_implications":["Size-mass relations in the rest-optical and rest-NIR differ: the rest-optical relation has a positive slope at $1.7<z<3$ that flattens at the massive end, while the rest-NIR relation is nearly flat across $1.7<z<3.5$.","Dust attenuation correlates more strongly with stellar-mass and SFR surface densities than with the global stellar mass and SFR, implying that the concentration of mass and star formation sets the dust column.","The 2175 Å UV-bump strength in the effective attenuation law is weaker than Milky Way/LMC values and cannot lie above the relation built into the model, because birth-cloud dust with no bump dilutes it; this is a modelling consequence, not a direct grain-composition measurement.","Nebular attenuation measured from the Balmer decrement correlates with stellar-continuum $A_V$ (Spearman rank $r\\simeq0.66$), supporting a common dust geometry for stars and gas in highly star-forming systems.","No significant correlation is found between $A_V$ and axis ratio at $1.7<z<3.5$, consistent with clumpy star-dust geometry rather than inclination-driven attenuation at cosmic noon."],"supporting_citations":[{"why":"Defines the two-component dust model (birth clouds around young stars plus diffuse ISM) that the paper adopts to separate attenuation of young and old stellar light.","marker":"Charlot & Fall (2000)"},{"why":"Supplies the flexible ISM attenuation law with free slope and tied UV-bump strength used in the SED fitting.","marker":"Kriek & Conroy (2013)"},{"why":"Provides the parameterisation of the ISM attenuation curve and the Drude-profile 2175 Å bump implemented in the model.","marker":"Noll et al. (2009)"},{"why":"Gives the local starburst attenuation law used as the baseline curve in the ISM model and as a comparison law.","marker":"Calzetti et al. (2000)"},{"why":"Provides the Bayesian SED-based stellar population fits and non-parametric star-formation histories that this paper extends with morphology and attenuation-law analysis.","marker":"Park et al. (2024)"},{"why":"Supplies the Bayesian fitting method used to measure half-light radii and axis ratios from NIRCam imaging.","marker":"Pasha & Miller (2023)"},{"why":"Establishes from simulated galaxies that dust can make light-profile sizes much larger than mass-profile sizes, the effect the paper claims to observe.","marker":"Zhang et al. (2023)"},{"why":"Supplies the rest-optical size-mass relation used as the comparison baseline for the F150W sizes.","marker":"Mowla et al. (2019)"},{"why":"Supplies the rest-NIR size-mass relation used as the comparison baseline for the F444W sizes.","marker":"Martorano et al. (2024)"}],"fun_headline_variants":["Dust inflates massive galaxy sizes by 30% at cosmic noon","Cosmic noon dust makes big galaxies appear 30% larger","Central dust flattens light profiles, swelling massive galaxy sizes","Massive galaxies look 30% bigger in optical due to dust","Dusty cores stretch galaxy sizes by 30% in rest-optical"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the adopted dust model—a fixed birth-cloud slope and a UV-bump strength tied to the ISM slope—describes real attenuation laws, since a different but plausible geometry would change the fitted optical attenuation and with it the size-gradient conclusion.","fun_headline_variants_meta":{"raw":{"variants":["Dust inflates massive galaxy sizes by 30% at cosmic noon","Cosmic noon dust makes big galaxies appear 30% larger","Central dust flattens light profiles, swelling massive galaxy sizes","Massive galaxies look 30% bigger in optical due to dust","Dusty cores stretch galaxy sizes by 30% in rest-optical"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000723,"raw_usage":{"total_tokens":3329,"prompt_tokens":1119,"completion_tokens":2210,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":735,"completion_tokens_details":{"reasoning_tokens":2118}},"tokens_in":735,"tokens_out":2210,"duration_ms":13555,"temperature":1.0,"reasoning_tokens":2118,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T11:28:30.958772+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Dust-correct the rest-optical images for the most massive galaxies (for example by dividing the F150W image by the fitted attenuation map, or by using a Balmer-decrement-based attenuation map) and re-measure half-light radii; if the ~30 percent optical-versus-NIR size excess is caused by central dust, the corrected optical/NIR size ratio should become consistent with unity.","supporting_citations":[],"review_version":1}