{"id":"bfa1cf61-0e06-4159-84ef-39b92aa4c2d0","arxiv_id":"2504.18820","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"In JADES galaxies at 1<z<7, offsets from the mass-metallicity relation are negatively correlated with offsets from the mass-size relation in four metallicity diagnostics, including for z>3.","lead":"This paper finds that among JWST-observed galaxies at redshifts 1 to 7, more compact galaxies tend to have higher gas-phase metallicities, using four different metallicity measurements. The result extends a well-established local relation into the early universe and suggests that gravitational potential, not just star formation, shaped early galaxy chemistry.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Residual correlation may be driven by shared redshift trends if the unpublished mass–size relation (Eq. 10) mis-models size evolution; a partial-correlation or re-fit test is needed.","rationale":"The reader's weakest assumption correctly identifies the unpublished, uncertainty-free mass–size relation as the key quantitative weak point. I partially agree: the deeper issue is the possibility of a spurious residual–residual correlation induced by shared redshift dependence. The paper itself flags this risk in Section 3.2, where it writes that fitting the MZR over a wide redshift range 'may introduce potential biases' and that 'unaccounted-for redshift evolution in metallicity may contribute additional scatter', but then asserts that correcting size evolution makes artificial correlations unlikely. That assertion is precisely what needs testing. The two residuals are not independent of redshift by construction: Δlog Re depends on the adopted α, β, k, and Δlog(O/H) depends on the choice of a redshift-independent MZR fit. If either relation mis-models the true redshift evolution, the Spearman p-values can be significant even without an intrinsic size–metallicity relation. Other concerns (no SFR control, F444W approximation for z>3.5) are real but secondary; the partial-correlation test directly addresses the central statistical inference. The recommended decision remains conditional acceptance, pending this robustness check, so the reader's verdict is unchanged.","tokens_in":22277,"tokens_out":7472,"duration_ms":76882,"concrete_test":"Recompute Δlog(O/H) and Δlog Re after (1) refitting the mass–size relation (Eq. 10) using the 1,123 JADES galaxies with 1 µm sizes from Section 2.3, with bootstrap uncertainties, and (2) rerunning the Spearman test on the residuals with redshift partialled out (e.g., partial Spearman correlation, or fitting Δlog(O/H) = a + b ΔlogRe + c log(1+z)). If the four tracers no longer all give p<0.01 (or p<0.05 for z>3), the claimed correlation is not robust. As an independent check, repeat using published mass–size relations (e.g., van der Wel et al. 2014; Mowla et al. 2019) and using a single observed band (F444W) for the z>3.5 subset where the rest-frame 1 µm size is approximated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim rests on a Spearman correlation between two residuals: Δlog(O/H), the offset from a MZR fitted without a redshift term (§3.1), and Δlog Re, the offset from the mass–size relation (Eq. 10). Both residuals will retain any redshift trend not fully removed by their respective fitted relations. The mass–size relation used to define Δlog Re is taken from an unpublished companion paper (Song et al. in prep) with coefficients α=0.162, β=−0.614, k=−0.964 and, as the reader notes, no uncertainties. If the true size evolution differs—for example, if β is off by even 0.2 in log(1+z), which is plausible given the scatter of published high-z mass–size relations—then Δlog Re will be systematically correlated with redshift across the 1<z<7 sample. The MZR residual is also potentially redshift-dependent: Section 3.2 explicitly acknowledges that fitting the MZR over a wide redshift range 'may introduce potential biases' and that 'unaccounted-for redshift evolution in metallicity may contribute additional scatter', but then asserts without proof that correcting size evolution makes artificial correlations unlikely. If both residuals share a common monotonic redshift component, the Spearman test can reach p<0.01 even when no intrinsic size–metallicity relation exists at fixed stellar mass and redshift. This is the most load-bearing threat to the central claim, and it is directly testable.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes JADES DR3 galaxies at 1 < z < 7 to test whether galaxy size, as a proxy for gravitational potential, correlates with gas-phase metallicity at fixed stellar mass. Using four strong-line metallicity calibrations (N2S2Hα, R23, N2, O3N2), the authors define Δlog(O/H) as the residual from a fitted mass–metallicity relation and Δlog Re as the residual from a mass–size relation that includes a redshift term. They report a significant negative Spearman correlation between these two residuals for all four calibrators (p < 0.01), and for the z > 3 subsample (p < 0.05). They compare the slope of this relation with MaNGA and TNG50 results and interpret the finding as evidence that gravitational potential regulates metallicity in the early universe.","tokens_in":22596,"tokens_out":4933,"duration_ms":46624,"significance":"If the central correlation is robust, this would be among the first evidence at 1 < z < 7 that galaxy size, independent of stellar mass and redshift, plays a role in setting gas-phase metallicity. The paper's strengths include the use of four independent metallicity calibrators, the public JADES DR3 sample, a dedicated z > 3 subsample, and quantitative comparison with MaNGA and TNG50. The consistency across calibrators is a genuine point in favor of the result. However, the central claim depends on an externally calibrated mass–size relation that is not yet published and is used without uncertainties, and the residual-based test has not been shown to be robust against shared redshift trends in the two residuals. These issues are load-bearing for the main conclusion and require additional analysis.","major_comments":[{"comment":"The definition of Δlog Re uses coefficients α = 0.162, β = −0.614, k = −0.964 from an unpublished companion paper (Song et al. in prep) with no reported uncertainties. Every value of Δlog Re in the analysis inherits these coefficients, so an error in β, the log(1+z) coefficient, would make Δlog Re systematically correlated with redshift. Because Δlog(O/H) is the residual from a MZR fit that includes no redshift term, both residuals could share a common monotonic redshift trend, and the Spearman test could reach p < 0.01 even in the absence of an intrinsic size–metallicity relation at fixed mass and redshift. Please report the uncertainties on these coefficients, re-fit the mass–size relation within the JADES sample, or repeat the analysis with published mass–size relations at these redshifts to show that the correlation is not an artifact of the adopted calibration.","section":"Section 3.2, Eq. (10)"},{"comment":"The text acknowledges that fitting the MZR over a wide redshift range 'may introduce potential biases' and that unaccounted-for redshift evolution in metallicity may contribute additional scatter, but then asserts without proof that correcting for size evolution makes artificial correlations unlikely. This is not a substitute for a quantitative test. A partial Spearman correlation controlling for redshift, or an analysis in which both the MZR and the mass–size relation include explicit redshift terms, is needed to establish that the reported Δlog(O/H)–Δlog Re correlation is not driven by residual redshift trends shared by both variables.","section":"Section 3.2, paragraph following Fig. 3"},{"comment":"The z > 3 test reuses the full-sample mass–size relation and MZR rather than re-fitting them to the high-redshift subsample. This narrows the redshift range and reduces the shared-trend concern, but it does not remove the dependence on the external mass–size coefficients in Eq. (10). The text should state this explicitly and discuss what is and is not tested by the z > 3 subsample.","section":"Appendix B"}],"minor_comments":[{"comment":"The Spearman r values are displayed as positive (e.g., r = 0.4257 for N2S2Hα, r = 0.3767 for N2) while the text describes a negative correlation; please clarify whether these are absolute values or specify the sign convention in the caption.","section":"Figure 3"},{"comment":"The caption reads 'mass-extinction diagram'; this should be 'mass-excitation (MEx) diagram'.","section":"Figure 1"},{"comment":"The heading contains the typo 'metalllicity', and the text contains 'correpsonding'; these should be corrected.","section":"Section 3.2"},{"comment":"Please state explicitly that the z > 3 analysis uses the full-sample MZR and mass–size relation rather than re-fitting them; the current wording, 'consistent with those described in Section 3.2', leaves this ambiguous.","section":"Appendix B"},{"comment":"The final sentence 'at very early universe' is ungrammatical; suggest 'in the very early universe' or 'at very early times'.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The main obstacle is the unpublished mass–size relation of Song et al. (in prep). If the coefficients cannot be released in this manuscript, the authors should either derive the relation internally from the JADES sample or show that the result is unchanged under plausible published alternatives. The residual-redshift concern is directly testable and should be addressed before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version. The paper extends the local Δlog(O/H)-Δlog Re anticorrelation to 1<z<7 using JADES DR3, with four metallicity calibrators giving consistent Spearman p<0.01. That consistency is the strongest thing in the paper, and the z>3 subsample makes the result more than a low-z tail effect. It is a genuine new empirical result, even though the residual method is standard.\n\nWhat the paper does well: the sample selection is careful — AGN removal with MEx plus X-ray cross-match, rest-frame 1µm sizes, PSF homogenization, and four independent strong-line calibrators. The comparison with MaNGA and TNG50 is useful and shows the slope is in the right ballpark. The authors also flag the MZR fitting caveat in Section 3.2, which is honest.\n\nSoft spots, in order. First, the definition of Δlog Re uses Eq. 10, a mass-size relation taken from Song et al. (in prep) with no uncertainties. This is load-bearing, because every galaxy's compactness residual comes from that relation. The stress-test worry is legitimate: if the true β for size evolution is more negative than -0.6, high-z galaxies will have positive ΔRe, and because the MZR is fitted without a redshift term, they will also tend to have negative ΔO/H. That combination yields a spurious negative correlation. The paper's response — that correcting size evolution makes artificial correlations unlikely — is an assertion, not a test. They need to either fit the MSR internally to this sample, propagate its errors, or do a partial Spearman correlation controlling for redshift. Narrow redshift bins would also work. The z>3 test is a step in the right direction but 3<z<7 is still a wide baseline.\n\nSecond, SFR is not controlled in the main test. The paper cites Cuestas et al. (2025) for no SFR dependence at z~2.3, but that's a different sample; since CIGALE SFRs are already in hand, a quick check would settle it. Third, the abstract says p-values 'much less than 0.01' when two tracers are around 0.006–0.008. That's a minor overstatement.\n\nThis is not a fatal flaw. The consistency across four calibrators and the z>3 signal are real evidence, and the concerns are directly addressable. I would send this to peer review, but the referee should demand the redshift-control test and error bars on the mass-size relation. If those hold up, this is a solid contribution to the high-redshift scaling-relation literature. Worth a reading-group slot for anyone working on JWST metallicity.","headline":"Solid first census of the size-metallicity relation at 1<z<7, but the residual correlation depends on an unpublished mass-size relation and needs a redshift-control test before I'd fully trust it.","tokens_in":23161,"tokens_out":4104,"would_cite":false,"duration_ms":41677,"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":"At fixed mass and redshift, more compact galaxies at 1<z<7 are more metal-rich, a trend seen with four independent metallicity calibrations.","keywords":["gas-phase metallicity","mass-metallicity relation","mass-size relation","galaxy size","gravitational potential","JADES","high-redshift galaxies","strong-line metallicity diagnostics"],"falsifier":"Re-fit the mass-size relation to the same galaxies with full uncertainty propagation (or adopt an independent size-mass calibration from the same redshift range), recompute $\\Delta\\log R_e$, and check whether the Spearman p-values remain below 0.01 for all four metallicity tracers; if the anticorrelation disappears, the central claim fails. A complementary check is to measure stellar velocity dispersions and see whether the size-metallicity correlation vanishes once the actual potential depth is controlled for.","tokens_in":22093,"feed_emoji":"🔭","tokens_out":11182,"duration_ms":100402,"temperature":0.7,"pith_summary":"Gas-phase metallicity records how many generations of stars have enriched a galaxy's gas, and at low redshift it is known to rise with stellar mass. This paper asks whether the underlying driver, the depth of a galaxy's gravitational potential as traced by its physical size, already operated at 1<z<7, when the universe was roughly one to six billion years old. Combining near-infrared imaging and medium-resolution spectroscopy from the JADES deep fields, the authors measure oxygen abundances with four independent strong-line calibrations and compare each galaxy's offset from the mass-metallicity relation, $\\Delta\\log(\\mathrm{O/H})$, with its offset from the mass-size relation, $\\Delta\\log R_e$. They find a significant negative correlation: at fixed stellar mass and redshift, more compact galaxies are more metal-rich, with Spearman p-values below 0.01 for all four tracers and below 0.05 for the $z>3$ subsample. Read as the paper's authors read it, this is the first evidence that gravitational potential regulates gas-phase metallicity towards cosmic dawn, meaning the material-exchange processes that set local galaxy chemical abundances were already active in the early universe.","feed_headline":"Smaller galaxies at 1<z<7 were systematically metal-rich","feed_subtitle":"Four metallicity tracers tie compactness to oxygen abundance, implicating gravity in early metal cycling.","key_machinery":"The carrying object is the residual pair $(\\Delta\\log(\\mathrm{O/H}), \\Delta\\log R_e)$. The metallicity residual comes from the fitted mass-metallicity relation for each of four calibrations, and the size residual is defined as $\\Delta\\log R_e = \\log R_e - \\log R_{\\rm model}$ with $\\log R_{\\rm model} = \\alpha\\log(M_*/M_\\odot)+\\beta\\log(1+z)+k$, using coefficients $\\alpha=0.162$, $\\beta=-0.614$, $k=-0.964$ from a companion analysis. Sizes are measured with Sérsic fits in JWST and HST bands and homogenized to rest-frame 1 μm, where the light traces older stellar populations and hence better approximates the underlying mass distribution and gravitational potential. The argument works by cross-checking the same anticorrelation across four independent metallicity diagnostics, repeating it for the $z>3$ subsample, and comparing the slope against low-redshift observations and cosmological simulations, so that a single calibration's systematics cannot produce the signal.","core_discovery":"The central claim is that the residual compactness of a high-redshift galaxy, defined by its deviation from the mass-size relation at the corresponding redshift, is significantly negatively correlated with its deviation from the mass-metallicity relation. The paper demonstrates this with 75-97 galaxies at 1<z<7 from the JADES survey, using four strong-line metallicity estimators (N2S2Hα, R23, N2, and O3N2): compact galaxies are enriched in oxygen at fixed stellar mass and redshift. For the full sample the Spearman p-values are below 0.01 for every tracer, and for the smaller $z>3$ subsample (21-34 galaxies) they are below 0.05. The authors also quantify the effect by fitting a modified mass-metallicity relation of the form $12+\\log(\\mathrm{O/H}) = k(\\log(M_*/M_\\odot)+\\beta\\,\\Delta\\log R_e)+b$, finding $\\beta$ between $-0.4$ and $-1.1$, broadly consistent with low-redshift measurements and with cosmological simulations. They conclude that gravitational potential, traced by size, plays a key role in regulating gas-phase metallicity and metal retention in the early universe.","pith_inferences":["A natural extension is to split the sample by stellar mass: the gas-regulator picture predicts the size-metallicity anticorrelation should be strongest at low masses, where outflows remove a larger fraction of the metal budget.","The gravitational-potential interpretation predicts that directly measured stellar velocity dispersions should correlate with metallicity at fixed mass and redshift; if dispersion explains the trend and size does not add information, the radius-based interpretation would be weakened.","The result's dependence on the companion mass-size relation could be tested by re-fitting $\\Delta\\log R_e$ with a relation derived from this sample with full uncertainty propagation, or by using an independent size-mass calibration in the same redshift range.","At $z>5$ the sample contains only a handful of galaxies; the same analysis on future data releases would show whether the anticorrelation steepens or flattens as the universe approaches reionization."],"forward_implications":["At fixed stellar mass and redshift, size acts as an independent lever on gas-phase metallicity, so the mass-metallicity relation is a projection of a relation that also depends on compactness.","Deeper potential wells retain more metals: the fitted $\\beta$ in $x=\\log(M_*/M_\\odot)+\\beta\\,\\Delta\\log R_e$ is between $-0.4$ and $-1.1$, bracketing low-redshift and simulation values and supporting the outflows-suppressed-by-gravity picture.","The anticorrelation persists for galaxies with $z>3$ alone, so the result is not driven by the more numerous lower-redshift galaxies in the sample.","Because all four strong-line diagnostics reproduce the trend despite their systematic differences, the correlation is unlikely to be an artifact of one metallicity calibration.","The consistency with low-redshift data and with cosmological simulations implies that potential-driven metal cycling was already in place a few billion years after the Big Bang rather than emerging only at late times."],"supporting_citations":[{"why":"Supplies the gas-regulator equations used in the introduction to argue that galaxies with deeper potential wells retain more metals, motivating the size-metallicity expectation.","marker":"Wang & Lilly 2021"},{"why":"Establishes the low-redshift $\\Delta\\log(\\mathrm{O/H})$-$\\Delta\\log R_e$ anticorrelation and provides the N2S2Hα calibration plus the MaNGA and TNG50 comparison points used here.","marker":"Ma et al. 2024"},{"why":"Provides the N2S2Hα strong-line metallicity calibration used as one of the four tracers.","marker":"Dopita et al. 2016"},{"why":"Provides the N2 and O3N2 calibrations for gas-phase oxygen abundance used as two of the four tracers.","marker":"Steidel et al. 2014"},{"why":"Gives the R23 upper- and lower-branch calibrations with ionization parameter used for the third tracer.","marker":"Kobulnicky & Kewley 2004"},{"why":"Supplies the N2>-0.8 branch-selection threshold that resolves the R23 degeneracy.","marker":"Kewley & Dopita 2002"},{"why":"Motivates measuring sizes at fixed rest-frame 1 μm because galaxy size varies with rest-frame wavelength and redshift.","marker":"Jia et al. 2024"},{"why":"Supplies the mass-size relation coefficients alpha=0.162, beta=-0.614, k=-0.964 used to compute every Delta log Re.","marker":"Song et al., in prep."},{"why":"Provides the JADES Data Release 3 NIRSpec medium-resolution spectroscopic catalog with the extinction-corrected emission-line measurements.","marker":"D'Eugenio et al. 2024"},{"why":"Documents the JADES survey whose NIRCam imaging and spectra define the sample.","marker":"Eisenstein et al. 2023a,b"}],"fun_headline_variants":["Compact galaxies at 1<z<7 run metal-rich","Size sets metal content in early galaxies","Small early galaxies hold more metals","JADES: compact galaxies are metal-rich at high z"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire analysis leans on a mass-size relation from an unpublished companion paper, whose three coefficients are adopted as fixed numbers without uncertainties, so every compactness residual, and hence the claimed correlation, would shift if that baseline is wrong.","fun_headline_variants_meta":{"raw":{"variants":["Compact galaxies at 1<z<7 run metal-rich","Size sets metal content in early galaxies","Small early galaxies hold more metals","JADES: compact galaxies are metal-rich at high z"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000775,"raw_usage":{"total_tokens":3503,"prompt_tokens":1092,"completion_tokens":2411,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":708,"completion_tokens_details":{"reasoning_tokens":2351}},"tokens_in":708,"tokens_out":2411,"duration_ms":19463,"temperature":1.0,"reasoning_tokens":2351,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:08:56.526982+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-fit the mass-size relation to the same galaxies with full uncertainty propagation (or adopt an independent size-mass calibration from the same redshift range), recompute $\\Delta\\log R_e$, and check whether the Spearman p-values remain below 0.01 for all four metallicity tracers; if the anticorrelation disappears, the central claim fails. A complementary check is to measure stellar velocity dispersions and see whether the size-metallicity correlation vanishes once the actual potential depth is controlled for.","supporting_citations":[],"review_version":1}