{"id":"e98dd895-aee5-43c7-a3c1-27f0b90a94ac","arxiv_id":"2507.11492","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Clumps in low-redshift analogues of high-redshift galaxies are a few hundred parsecs across, and simulated cosmic-noon observations show that clustering and limited resolution inflate their apparent sizes.","lead":"Using adaptive-optics observations of 18 low-redshift stand-ins for distant star-forming galaxies, researchers identified 84 compact star-forming clumps. By artificially moving these galaxies to cosmic-noon distance, they show that clump clustering and blurring inflate apparent clump sizes, helping explain why distant clumps look larger than they may really be.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The artificial-redshift simulation is too underdescribed to support the central size-inflation claim; in particular, noise injection and clump-selection thresholds at z=2.2 are unspecified, so Figure 8's uniform radius increase may be a simulation artifact.","rationale":"The paper's aim is to show that clump clustering and limited resolution inflate the apparent sizes of high-redshift clumps, using low-redshift analogues artificially redshifted to z=2.2. The decisive evidence is Figure 8, which reports that all 16 matched clumps have larger radii at z=2.2 even after a quadratic FWHM correction. The reader's weakest assumption identifies the fidelity of this simulation as the key vulnerability, and I agree. My stress-test sharpens that concern: the simulation description (Section 2.1) omits how noise is added, how the S/N=6 threshold and FellWalker parameters are scaled, and how clump matching is performed. Without these details, the simulated cubes may not reproduce the noise-limited conditions of actual OSIRIS observations, and the measured size increase could be an artifact of rescaling fluxes while leaving the noise floor unchanged. I considered other issues—the lack of radii error bars in Table 3, the unresolved-clump upper limits, and the dependence on unpublished 'Gonçalves et al. in prep' for the Schmidt-Kennicutt uncertainty—but these are secondary: they affect quantitative interpretations of gas masses or the significance of individual sizes, not the core existence of the clustering-induced inflation. A targeted re-simulation with explicit sky-noise injection and identical detection thresholds would settle whether the effect survives realistic noise. Until then, the conditional verdict is appropriate: the claim is plausible and consistent with prior work, but the paper's own test is not yet fully convincing.","tokens_in":20882,"tokens_out":5932,"duration_ms":73874,"concrete_test":"Re-run the Section 2.1 simulation on at least three LBAs (e.g., 143417, 092600, 210358) with explicit noise injection: after rescaling to z=2.2, add Gaussian noise per spaxel with variance matched to an OSIRIS sky background and exposure time, then apply the same S/N=6 spaxel cut and the FellWalker parameters of Table 2; measure the 16 matched clump radii. If the median radius increase over the low-z values disappears or reverses, the clustering-size conclusion is not robust to realistic noise.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim—that clump clustering and low resolution make high-z clumps appear larger—rests entirely on the artificial redshift simulation of Section 2.1. The manuscript says only that datacubes were 'spatially smoothed' at the OSIRIS angular resolution and spaxel scale and that fluxes were rescaled using Hα; it does not state how noise was injected, whether the S/N=6 spaxel cut and FellWalker thresholds (Table 2) were re-evaluated on the degraded cubes, or how the '16 corresponding clumps' in Figure 8 were matched between z≈0.2 and z=2.2. Real OSIRIS observations at z~2 are sky-noise limited with roughly constant per-spaxel noise; if the simulation rescales fluxes without adding a corresponding sky background, the simulated high-z cubes have artificially high S/N, which changes which clumps are detected and how their Gaussian-fitted FWHMs behave after the 'quadratic correction.' Because the 16/17 surviving clumps and their size increase are the only direct evidence for the headline conclusion, an under-specified simulation leaves the claim unsupported. The qualitative conclusion may still be true (Fisher et al. 2017a find a similar effect in HST data), but the paper's own test is not yet convincing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents OSIRIS/Keck Pa-alpha IFU observations of 18 Lyman Break Analogs at z~0.1-0.2, identifies 84 star-forming clumps using the FellWalker algorithm followed by Gaussian profile fits, and measures clump radii (a few hundred pc), velocity shear (~12 km/s), velocity dispersion (~70 km/s), gas masses, and dynamical masses. The authors then artificially redshift the datacubes to z=2.2 and report that clump radii increase even after correcting for the larger FWHM, leading to the central claim that high-redshift kpc-scale clump sizes are inflated by clump clustering and limited resolution. The paper also compares the low- and simulated-high-redshift clump properties with literature samples and argues that LBAs are valuable laboratories for studying clumpy star formation at cosmic noon.","tokens_in":21180,"tokens_out":5396,"duration_ms":61569,"significance":"If the size-inflation result holds, it would help resolve a long-standing discrepancy between AO- and lensing-resolved clump sizes and seeing-limited high-redshift measurements, and it would strengthen the case for using LBAs as local laboratories for cosmic noon studies. The study has concrete strengths: a moderately large AO sample, a clearly described two-step clump-finding procedure, and an explicit attempt to simulate the observational bias rather than only speculate about it. However, the central claim rests on an underspecified simulation, and the absence of uncertainties on radii and masses limits the quantitative force of the conclusions. The paper is a useful empirical contribution but requires revision before its headline conclusion is fully supported.","major_comments":[{"comment":"The artificial-redshift simulation is described only as spatial smoothing to the OSIRIS resolution and spaxel scale, flux rescaling via Halpha, and a 'quadratic correction' for FWHM. The manuscript does not state how noise was injected, whether sky-noise-limited conditions appropriate to z~2 were simulated, how the S/N=6 spaxel threshold and the FellWalker parameters in Table 2 were re-evaluated on the degraded cubes, or how the 16 corresponding clumps in Figure 8 were matched to the low-z clumps. Because the central claim that high-z clump sizes are inflated rests on this simulation, please provide the full recipe and a quantitative matched-sample test (e.g., sign test or bootstrap on radius ratios) rather than the deterministic statement that all matched clumps increased in radius.","section":"§2.1, Figure 8"},{"comment":"Clump radii in Table 3 are quoted without uncertainties, and Figure 8 has no error bars or statistical test on the radius increase. The radius is defined from Gaussian fits along two axes, so fitting errors and PSF calibration errors should be propagated; without them, the claim that all 16 clumps show a size increase at z=2.2 cannot be assessed against measurement noise.","section":"Table 3, Figure 8"},{"comment":"Gas masses are derived by applying the Kennicutt (1998) Schmidt-Kennicutt relation, calibrated on galaxy-aperture scales, to individual clump apertures. The text acknowledges this assumption 'is likely untrue,' but the resulting Mgas values are then compared directly with Mdyn in Figure 5 and used to infer feedback contributions, with no error bars or systematic offset quoted. Please either provide a clump-scale calibration, propagate the stated ~0.3 dex systematic, or explicitly reframe the Mgas-Mdyn comparison as an order-of-magnitude check.","section":"§4.3, Eq. (2), Figure 5"}],"minor_comments":[{"comment":"There is a typo, 'positiong,' and the formula for R should be written with an explicit square root, e.g., R = 0.5 sqrt(FWHM_x FWHM_y), to avoid ambiguity.","section":"§3"},{"comment":"The column headers 'Ref f logM' are unclear; please separate the effective-radius and stellar-mass columns with definitions in the table note.","section":"Table 1"},{"comment":"Figure 3 reports 31 resolved clumps and Figure 4 reports 77 clumps, while Table 3 and the text state 38 resolved and 84 total; please explain the exclusions (e.g., missing kinematics or non-detections) in the captions or text.","section":"Figures 3 and 4"},{"comment":"The text says the simulated high-z cubes contain 17 clumps, while Figure 8 and the discussion refer to 16 corresponding clumps; please make the counts and the matching rule consistent.","section":"§5.1, Figure 8"},{"comment":"The normalization of the histograms 'by the number of data in each bin' is not defined; please state whether the y-axis is a fraction per bin and give the bin width.","section":"Figure 7"},{"comment":"In the Toomre parameter expression, the factor A is not clearly defined; please specify A=3.36 for stellar disks and A=pi for gas disks and define kappa as the epicyclic frequency in the text.","section":"§5.2, Eq. (4)"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of a galaxy-evolution journal and the artificial-redshift experiment is a worthwhile idea. My concern is not with the qualitative conclusion, which is plausible and supported by prior work such as Fisher et al. (2017a), but with the fact that the manuscript's own test is not yet described well enough to carry the claim. I therefore recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The useful thing here is the dataset: 84 Pa-alpha clumps in 18 LBAs observed with Keck/OSIRIS and AO, with measured sizes, velocity dispersions, shears, and dynamical masses. That census is new and will get cited. The low-z measurements are handled honestly—clumps come out at a few hundred parsecs, dynamically hot (sigma ~ 70 km/s, Vs ~ 12 km/s), with dynamical masses typically above the Schmidt-Kennicutt gas masses. The literature comparison (Fisher, Livermore, Lenkic, Wisnioski, the lensed samples) is thorough, and the authors are upfront that applying S-K at clump scale is 'likely untrue,' which is more candor than you usually get.\n\nThe central claim—clustering plus limited resolution inflates high-z clump sizes—is not new. Fisher et al. (2017a) already showed it, and the paper says so. The new contribution is a Pa-alpha/AO version of that test on LBAs, plus the new census. That is a legitimate extension, not a breakthrough.\n\nThe weak points are in the simulation and the statistics, and the stress test is roughly right. Section 2.1 is underdescribed: no noise injection, no statement about whether the S/N=6 cut and the FellWalker thresholds were re-run on the degraded cubes, no matching procedure for the 16 'corresponding' clumps. Real z~2 OSIRIS data are sky-noise limited; a flux-rescale-without-sky test can manufacture high S/N and bias the size comparison. So Figure 8, as presented, is not by itself convincing. I would soften the stress test on one point: the size-inflation conclusion does not stand or fall on Figure 8, because Fisher et al. already established the effect in independent HST data. The paper's own test needs work, but the claim has external support.\n\nTwo smaller things. Table 3 radii have no uncertainties, which is awkward for a paper whose headline number is sizes. Figure 8 has no error bars and no statistical test; with 16 clumps all going the same way, a simple sign test would help. And the <0.3 dex bound on the S-K caveat leans on an unpublished self-cited paper (Goncalves et al. in prep); that support should be in hand or public before it anchors an error estimate.\n\nThis is a solid, honest paper with a real new dataset and a conclusion that is probably right but not fully demonstrated by its own test. That is revision territory, not rejection. Send it to review, with a referee who knows Fisher et al. and the OSIRIS/Law et al. pipeline, and ask them to check Sections 2.1 and 5.1 closely.","headline":"New 84-clump LBA census that is worth citing; the redshift-simulation test is underdescribed, but the size-inflation claim leans on Fisher et al. and likely survives revision.","tokens_in":21756,"tokens_out":4832,"would_cite":true,"duration_ms":49236,"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":"Star-forming clumps in nearby Lyman-break analog galaxies have radii of a few hundred parsecs, and apparent kiloparsec-scale clumps at high redshift are inflated by clustering and resolution.","keywords":["Lyman-break analog galaxies","star-forming clumps","clump clustering","artificial redshift simulation","adaptive optics integral field spectroscopy","Pa-alpha emission","clump sizes","cosmic noon galaxies"],"falsifier":"Take one of the 18 LBAs, degrade its cube to z=2.2 with substantially better PSF or with matched real z~2 H-alpha AO observations, identify clumps with the same FellWalker settings, and check whether the radii still exceed the few-hundred-parsec intrinsic values; if they no longer do, the inflation claim is falsified.","tokens_in":20710,"feed_emoji":"🔭","tokens_out":6615,"duration_ms":77178,"temperature":0.7,"pith_summary":"The paper uses 18 nearby Lyman-break analog galaxies, compact starbursts at z~0.1-0.2 that mimic cosmic noon conditions, to measure star-forming clumps at hundreds of parsec resolution. It finds 84 clumps with radii of a few hundred parsecs, low velocity shear (~12 km/s), and high velocity dispersion (~70 km/s). The key test is artificial redshift: when the same datacubes are degraded to z=2.2, clump counts drop from 84 to 17 and every matched clump appears larger even after correcting for resolution. The paper concludes that kpc-scale clump sizes reported in many high-redshift studies are inflated by clump clustering and coarse resolution, not intrinsic to the clumps.","feed_headline":"Clumps in cosmic noon analogues are hundreds of parsecs, not kiloparsecs","feed_subtitle":"Artificial redshift makes the same clumps look bigger, so kpc sizes at high redshift may be resolution artifacts.","key_machinery":"The central mechanism is the artificial redshift simulation, which takes the observed LBA Pa-alpha datacubes and reprojects them to z=2.2 using the same OSIRIS IFU spaxel scale, PSF, noise level, and cosmological surface-brightness dimming as real high-redshift observations. Clumps are identified with the FellWalker clump-finding algorithm, sizes are measured by fitting Gaussians to the spatial flux profiles, and a clump is considered resolved only if its measured radius exceeds 1.2 times the PSF FWHM. This simulation isolates what clustering plus finite resolution do to measured clump sizes, and the comparison between the low-z and simulated high-z measurements is what carries the paper's central conclusion.","core_discovery":"The paper claims that star-forming clumps in cosmic noon galaxies are intrinsically sub-kiloparsec structures, and that the kiloparsec-scale clump sizes seen in many high-redshift surveys are observational artifacts. In the 18 Lyman-break analogs, 84 Pa-alpha clumps are identified with the FellWalker algorithm, of which 38 are resolved; their radii cluster at a few hundred parsecs. When the observed datacubes are artificially redshifted to z=2.2, the number of detectable clumps falls to 17, and all 16 matched clumps show larger radii even after quadratic subtraction of the larger PSF FWHM. The paper interprets this as the clump clustering effect: at lower resolution, blends of several small clumps are detected as one large clump, which also raises the apparent integrated SFR per clump. The measured clump kinematics, with velocity shear much weaker than velocity dispersion, indicate dynamically hot, non-virialized structures, and the dynamical masses typically exceed gas masses inferred from the Schmidt-Kennicutt relation, pointing to pressure support or feedback.","pith_inferences":["Extending the paper's logic, the same clustering bias likely inflates other morphological measurements at high redshift, such as clump surface densities, clump mass estimates, and the apparent compactness of star-forming regions, not just radii.","If kpc-scale clump sizes are artifacts, models of giant clump migration and bulge growth that assume kpc-scale clumps may need to operate on sub-kiloparsec clumps or invoke a different accretion and coalescence scale.","A testable extension would be to apply the same artificial-redshift pipeline to lensed JWST clump samples at different redshifts, mapping how apparent clump size grows with decreasing resolution and providing a quantitative correction for unresolved high-z surveys.","The prediction that real z~2 clumps observed with 30-40 meter class adaptive optics or gravitational lensing should show few-hundred-parsec radii, matching the simulated appearances, could be checked once such data are available."],"forward_implications":["If cosmic noon clumps are intrinsically sub-kiloparsec, high-redshift surveys without adaptive optics or lensing systematically overestimate clump sizes and the SFR of individual clumps because blended clusters are counted as single objects.","The reduction from 84 clumps at low redshift to 17 at simulated z=2.2 implies that high-redshift samples are detecting clustered groups of clumps, not individual star-forming regions.","The low velocity shear and high velocity dispersion of the resolved clumps imply they are not virialized, with velocity dispersion likely boosted by star formation feedback.","Dynamical masses exceeding gas masses from the Schmidt-Kennicutt relation suggest that clump dynamics include pressure or turbulent support beyond the gas surface density inferred from star formation.","The agreement of LBA clump sizes with lensed few-hundred-parsec samples supports the use of low-redshift analogs as anchors for interpreting cosmic noon observations."],"supporting_citations":[{"why":"Supplies the OSIRIS observations, the Pa-alpha data reduction, and the artificial-redshift simulation prescription used to emulate z=2.2 observations.","marker":"Gonçalves et al. 2010"},{"why":"Defines the Lyman-break analog sample by FUV surface brightness to mimic cosmic noon star-forming galaxies.","marker":"Hoopes et al. 2007"},{"why":"Establishes that LBAs show star-forming clumps and compact morphologies similar to high-redshift galaxies, providing the sample's morphological basis.","marker":"Overzier et al. 2009"},{"why":"Reports the clump clustering effect that systematically increases apparent clump sizes at lower resolution, the effect the paper reproduces in simulation.","marker":"Fisher et al. 2017a"},{"why":"Provides lensed high-redshift clump sizes of a few hundred parsecs and the SFR-size relation used as a comparison baseline.","marker":"Livermore et al. 2015"},{"why":"Large JWST lensed sample giving clump effective radii of 10-700 pc, anchoring the small-size view of clumps at high redshift.","marker":"Claeyssens et al. 2025"},{"why":"Presents the FellWalker clump-finding algorithm used to identify all clumps in the Pa-alpha flux maps.","marker":"Berry 2015"},{"why":"Resolves lensed cosmic noon clumps to few-hundred-parsec scales, providing another intrinsic-size baseline.","marker":"Cava et al. 2018"}],"fun_headline_variants":["Cosmic noon clumps measure hundreds of parsecs, not kiloparsecs","High-redshift kiloparsec clumps are likely resolution artifacts","Clump clustering mimics growth at cosmic noon distances","Real clumps stay small; big ones come from blending","Artificial redshift shows clump sizes are inflated by blur"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The artificial-redshift pipeline produces cubes whose resolution, noise, surface-brightness dimming, and clump-detection behavior match real z=2.2 observations; if it does not, the measured size increase and the clustering conclusion do not follow.","fun_headline_variants_meta":{"raw":{"variants":["Cosmic noon clumps measure hundreds of parsecs, not kiloparsecs","High-redshift kiloparsec clumps are likely resolution artifacts","Clump clustering mimics growth at cosmic noon distances","Real clumps stay small; big ones come from blending","Artificial redshift shows clump sizes are inflated by blur"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000406,"raw_usage":{"total_tokens":2158,"prompt_tokens":1040,"completion_tokens":1118,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":656,"completion_tokens_details":{"reasoning_tokens":1032}},"tokens_in":656,"tokens_out":1118,"duration_ms":10150,"temperature":1.0,"reasoning_tokens":1032,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T17:08:06.270120+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take one of the 18 LBAs, degrade its cube to z=2.2 with substantially better PSF or with matched real z~2 H-alpha AO observations, identify clumps with the same FellWalker settings, and check whether the radii still exceed the few-hundred-parsec intrinsic values; if they no longer do, the inflation claim is falsified.","supporting_citations":[],"review_version":1}