{"id":"48340af4-f736-4c79-89e4-ae19f5c7e679","arxiv_id":"2412.00188","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A semi-analytic galaxy formation model predicts that the star formation rate surface density main sequence peaks in slope near redshift 3 and flattens by today, with massive galaxies driving the change.","lead":"Using a computer simulation of galaxy formation, this paper follows how the rate of new stars per square kiloparsec of galaxy area evolves from the early universe to today. It finds that this rate falls steeply over cosmic time, and that the steepest dependence on galaxy mass occurred around redshift 3, providing a simple formula for comparing with telescope data.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central slopes and their evolution hinge on using the SAM's StellarHalfLightRadius as the observed half-light radius; a mass-dependent size bias would shift the Sigma_SFRMS slope, and no independent size validation is provided.","rationale":"The reader's weakest assumption correctly identifies the most load-bearing concern: the simulated effective radius StellarHalfLightRadius is used as R_e in Sigma_SFR, and the paper provides no validation of this model size against the observed half-light radii of the same galaxy populations. Because Sigma_SFR scales as R_e^{-2}, any mass-dependent size error directly changes the slope of Sigma_SFRMS, not just its normalization, and therefore affects the claimed slope peak at z~2.9 and the decomposition into SFR and size contributions. This is a genuine correctness risk, not merely a comparison nuisance. The reader's CONDITIONAL verdict is therefore appropriate: the paper's physically motivated and internally consistent predictions are plausible, but the unvalidated size proxy leaves room for substantial quantitative error. A secondary concern is that the simulated 'main sequence' may not apply the same star-forming selection as the observed samples; this could also affect slopes, but the size proxy is more fundamental because it enters every Sigma_SFR value. The proposed test, recomputing Table 2 with an empirical size correction, would settle whether the central slope evolution survives. No evidence of circular fitting or other methodological misconduct was found; the issue is an unvalidated input assumption, not a hidden use of the observed Sigma_SFR data.","tokens_in":12550,"tokens_out":9187,"duration_ms":86953,"concrete_test":"Compare L-Galaxies2020 StellarHalfLightRadius against observed stellar-mass--half-light-radius relations in the same redshift bins used in Table 2, e.g., van der Wel et al. (2014) at z<3, Calabro et al. (2024) at z~5 and 7.5, and the JADES/CRISTAL samples at z>10. If the model R_e deviates by more than ~0.2 dex at fixed M* in a mass-dependent way, replace StellarHalfLightRadius with the empirically calibrated size relation, recompute Sigma_SFR for each snapshot, and refit the slopes in Table 2. If the z~2.9 peak and the subsequent decline to z=0 do not survive this correction, the central claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The key claim is the evolution of the Sigma_SFRMS slope from a peak of 0.709 at z~2.9 to 0.085 at z=0, with the interpretation that SFR decline, not size growth, drives it. But Sigma_SFR is defined as SFR/(2*pi*R_e^2) (Section 1), and the paper uses L-Galaxies2020's StellarHalfLightRadius as R_e (Section 2.1) without validating this model size against the observed half-light radii of the comparison samples. Since R_e enters squared, a modest mass-dependent systematic offset of 0.2 dex in log R_e changes log Sigma_SFR by 0.4 dex and changes the fitted slope by -2 * d(log R_e)/d(log M*). The claimed agreement at z=0, 1, 2, 5, and 7.5, the slope peak near z~2.9, and the attribution to SFR versus size evolution all depend on the model's size-mass relation being accurate across this full redshift range. The statement in Section 2.3 that 'we consistently used the same size estimates in the simulated data and observational results' conflates using the same operational definition with having validated model sizes; it does not remove the bias. Without an independent size check, the central quantitative claims are not yet secure.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript uses the L-Galaxies2020 semi-analytic model to predict the evolution of the star-formation-rate surface density Sigma_SFR and its main sequence (Sigma_SFRMS) from z=12 to z=0. The authors define Sigma_SFR = SFR/(2*pi*R_e^2), compare the SAM predictions with literature observations at z=0, 1, 2, 5, and 7.5 and with individual galaxies at z>10, fit the evolution of the slope and normalization of the Sigma_SFRMS, and attribute the slope decline from 0.709 at z~2.9 to 0.085 at z=0 to a faster SFR decline in the most massive galaxies, with a smaller contribution from size growth. They also provide a redshift-dependent fitting formula, Eq. (3).","tokens_in":12799,"tokens_out":5893,"duration_ms":49542,"significance":"If the assumed size proxy is accurate, this is a useful first SAM-based prediction of the Sigma_SFRMS evolution and it provides a compact, easy-to-use fitting formula. The analysis is transparent in important ways: the mass-bin definitions, sample sizes, and fit coefficients are tabulated, and the comparisons with observed data are clearly described. The CSFRD comparison in Appendix A is also a valuable consistency check. However, the quantitative agreement with observations is weaker than the abstract suggests, and the central slope evolution depends on an unvalidated half-light-radius proxy, so the significance of the specific quantitative claims is currently conditional.","major_comments":[{"comment":"The central quantitative results use the SAM's StellarHalfLightRadius as the effective radius R_e entering Sigma_SFR = SFR/(2*pi*R_e^2), but the manuscript does not validate the simulated size-mass relation against the observed half-light radii of the comparison samples. Since R_e enters squared, a mass-dependent systematic offset of delta dex in log R_e shifts log Sigma_SFR by -2*delta and changes the fitted slope by -2*d(delta)/dlogM*. The claimed slope peak of 0.709 at z~2.9, the decline to 0.085 at z=0, and the agreement at z=0, 1, 2, 5, and 7.5 are therefore not secure. The sentence in §2.3 that \"we consistently used the same size estimates in the simulated data and observational results\" describes an operational definition, not an accuracy validation. Please add a quantitative comparison of the SAM size-mass relation with observed R_e at the relevant redshifts, such as the JWST samples in Calabrò et al. (2024), and re-derive the slope evolution after applying any measured size correction, or explicitly state the magnitude of the residual size systematic and its effect on the slopes.","section":"§2.1, §2.3, Fig. 2, Table 2"},{"comment":"The abstract's claim that the simulated Sigma_SFRMS \"agrees with the observed one at z=0, 1, 2, 5, and 7.5\" is overstated relative to the body of the paper. In §3.2 the z=2 simulated normalization is about 0.5 dex higher than the Salim et al. (2023) relation, and at z=5 the simulated low-mass end is about 0.5 dex below the Calabrò et al. (2024) relation. The abstract should be rephrased to say agreement within about 0.5 dex with these offsets acknowledged, or the analysis must show that these offsets are within the combined systematic uncertainties.","section":"Abstract and §3.2"},{"comment":"The claim that the Sigma_SFRMS is already present at z~11 is weakly supported. The z=11.5 fit is based on 42 galaxies with slope 0.95 ± 0.655, and the z=12.5 fit on 7 galaxies with slope 0.637 ± 1.593; the combined z=10.8 fit uses 220 galaxies and yields slope 0.65 ± 0.26. These uncertainties are large enough that a well-defined main-sequence slope at z>10 is not established. Please either restrict the claim to z ≲ 10.8 or present a robustness test, such as a bootstrap or a fixed-slope comparison, demonstrating that the high-z slope is meaningful.","section":"Table 2 and §3.2"},{"comment":"The causal attribution of the slope evolution would benefit from an explicit decomposition. Since log Sigma_SFR = log SFR - 2*log R_e - const, the per-mass-bin change between z~4 and z=0 should be expressed as Δlog Sigma_SFR = Δlog SFR - 2*Δlog R_e. Using the quoted values, the SFR drop dominates, with Δlog SFR ranging from about -1.9 to -1.2 dex while 2*Δlog R_e ranges from about 0.95 to 0.83 dex, but this calculation is not presented. In addition, the sentence \"the change in Re also increases ΣSFR\" is directionally ambiguous: an increase in R_e decreases Sigma_SFR. Please correct the wording and add the numerical decomposition.","section":"§3.3"}],"minor_comments":[{"comment":"There is a typo in the abstract: \"Telesescope\" should be \"Telescope\".","section":"Abstract"},{"comment":"The text contains formatting artifacts such as \"di fferences\" and \"di ffuse\" that should be corrected to \"differences\" and \"diffuse\".","section":"§2.3"},{"comment":"The last row of Table 2, labeled z=10.8, appears after the z=12.5 row; the ordering should be chronological or the combined-fit row should be clearly separated with a note.","section":"Table 2"},{"comment":"Equation (3) has a spacing typo in the term \"ΣS FR\"; it should be \"Σ_SFR\".","section":"Eq. (3)"}],"recommendation":"major_revision","confidential_remarks":"The paper is a Letter, and the required size validation may require a non-negligible addition. If the authors cannot provide a comparison of simulated and observed sizes, the claims should be reframed as SAM predictions contingent on the size proxy. I would not recommend rejection because the internal analysis is transparent and the fitting formula is useful, but the abstract overstates the agreement and the central slope evolution is not yet secure without the size check."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is worth engaging with. It is the first SAM-based, mass-resolved look at the Sigma_SFR main sequence, and it hands observers a fitting formula (Eq 3) plus a concrete claim: the slope peaks around z~3 at ~0.7 and flattens to ~0.08 by z=0. The analysis is transparent — definitions, mass bins, fits are all clearly laid out — and the authors do not hide the rough edges: they note the 0.5 dex offset at z=2, the disagreement at the low-mass end at z=5, and the enormous slope errors at z>10. The CSFRD breakdown by mass bin is a nice addition.\n\nThe main soft spot is the size assumption. Sigma_SFR uses R_e squared, and the model's StellarHalfLightRadius is taken at face value as the observed half-light radius, with no independent validation against the same samples used for comparison. A modest mass-dependent size bias would shift both the normalization and the slope of the predicted main sequence. The statement that they 'consistently used the same size estimates' addresses the operational definition, not the accuracy. This does not sink the paper — the broad trend of Sigma_SFR rising with redshift and the slope peaking near z~3 are likely robust — but the agreement claims in the abstract are stronger than the evidence supports. The abstract says the simulated relation 'agrees' at z=0,1,2,5,7.5 when the text itself documents 0.5 dex offsets at z=2 and z=5. That should be toned down.\n\nA secondary gripe: Eq 3 is a polynomial fit to the model's own slopes and normalizations, with no propagated uncertainties. It is an interpolation, not an independent prediction, which is fine, but the paper should say so explicitly. No code or catalogs are provided, which limits reproducibility for a letter-length paper.\n\nOverall: the central model-based claims are probably correct, the observational comparisons are mostly fair, and the paper gives the community a testable formula. It deserves a serious referee. The main revision I would ask for is a caveat section on the size validation, plus a more measured abstract. If that is done, this is a solid contribution to the growing Sigma_SFR literature.","headline":"A useful and novel SAM-based Sigma_SFRMS fitting formula, but the size assumption needs an explicit caveat before the agreement claims can be fully trusted.","tokens_in":13452,"tokens_out":2610,"would_cite":true,"duration_ms":23807,"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":"A semi-analytic simulation traces the star formation rate surface density of galaxies from z=12 to z=0, finding a ~3.5 dex decline and a main-sequence slope that peaks at 0.709±0.005 near z≈2.9 before falling to 0.085±0.003 at z=0.","keywords":["star formation rate surface density","main sequence","semi-analytic model","L-Galaxies2020","galaxy evolution","high-redshift galaxies","JWST","cosmic star formation rate density"],"falsifier":"A mass-complete sample of galaxies with measured half-light radii and SFRs at $z\\approx3$ and $z\\approx0$ would settle the central claim: the model predicts a $\\Sigma_{\\rm SFR}{\\rm MS}$ slope of $0.71\\pm0.02$ at $z\\approx3$ and $0.085\\pm0.02$ at $z=0$, along with a ~5000% drop in SFR for $10^{11}\\,M_\\odot$ galaxies. Observing a significantly flatter high-redshift slope or a smaller SFR drop would falsify the explanation.","tokens_in":12290,"feed_emoji":"🌌","tokens_out":4806,"duration_ms":38837,"temperature":0.7,"pith_summary":"This paper uses the L-Galaxies2020 semi-analytic galaxy formation model to trace how the star formation rate per unit area of galaxies, $\\Sigma_{\\rm SFR}$, evolves from $z=12$ to today. It finds that $\\Sigma_{\\rm SFR}$ declines by about 3.5 orders of magnitude over that time, and that the slope of the $\\Sigma_{\\rm SFR}$–stellar mass main sequence rises to a maximum of about 0.71 near $z\\approx3$ before falling to about 0.09 at $z=0$. The paper attributes this slope evolution mainly to a dramatic drop in star formation rate in the most massive galaxies, with a smaller contribution from galaxy size growth. If right, the result explains why JWST sees a steeper and higher $\\Sigma_{\\rm SFR}$ main sequence at high redshift and provides a redshift-dependent fitting formula for the sequence.","feed_headline":"Galaxy star-formation surface density falls 3.5 dex since z=12","feed_subtitle":"A model ties the main-sequence slope peak at z≈3 to massive galaxies' rapid SFR decline.","key_machinery":"The central object is the L-Galaxies2020 semi-analytic model, specifically its predicted effective radius (StellarHalfLightRadius) and SFR, combined through the definition $\\Sigma_{\\rm SFR} = {\\rm SFR}/(2\\pi R_e^2)$. The argument is carried by fitting the slope and normalization of the $\\Sigma_{\\rm SFR}$–$M_*$ main sequence as polynomial functions of redshift (Equations 1 and 2) and combining them into Equation 3, which predicts the sequence at any redshift up to $z\\sim10.8$. The physical mechanism identified is that the most massive galaxies dominate the cosmic star formation rate density below $z\\sim4.5$ and undergo a much larger fractional SFR drop than lower-mass galaxies, which steepens and then flattens the main-sequence slope.","core_discovery":"The central claim is that the star formation rate surface density main sequence ($\\Sigma_{\\rm SFR}{\\rm MS}$) in the L-Galaxies2020 model evolves with a slope that peaks at $0.709\\pm0.005$ at $z\\approx2.9$ and declines to $0.085\\pm0.003$ at $z=0$, while the overall $\\Sigma_{\\rm SFR}$ drops by roughly 3.5 dex from $z=12$ to $z=0$. The decline in slope is driven primarily by a rapid decrease in SFR in the most massive galaxies (about 5000% from $z\\approx4$ to $0$), with a smaller contribution from their size growth; bulge build-up alone is insufficient to explain the evolution. The simulated $\\Sigma_{\\rm SFR}{\\rm MS}$ agrees with observed relations at $z=0,1,2,5,7.5$ and with individual galaxies at $z>10$. The paper also derives a redshift-dependent fitting formula, $\\log(\\Sigma_{\\rm SFR})=y_{\\rm slope}(z)[\\log(M_*/M_\\odot)-9]+y_{\\rm norm}(z)$, with polynomial coefficients given in Equations 1 and 2.","pith_inferences":["If the simulated half-light radii carry even a moderate systematic error, the inferred $\\Sigma_{\\rm SFR}$ normalization changes by about twice that error because $R_e$ enters squared; the agreement with observations is therefore as sensitive to size calibration as to SFR accuracy.","The redshift-dependent fitting formula could be repurposed as an empirical prior for correcting selection effects in surveys that under-detect low-surface-brightness galaxies, an application the paper does not explore.","A direct test of the model's mechanism would be to measure half-light radii and SFRs for a mass-complete sample of massive galaxies at $z\\approx3$ and $z\\approx0$; if the SFR drop of $10^{11}\\,M_\\odot$ galaxies relative to lower-mass galaxies is smaller than the simulated ~5000%, the predicted slope decline would be too steep.","The slope-peak redshift of about 2.9 could serve as a diagnostic for when massive galaxies transition from gas-rich star formation to quiescence in any galaxy formation model, not just in this SAM."],"forward_implications":["The fitting formula (Equation 3) predicts the $\\Sigma_{\\rm SFR}{\\rm MS}$ at any redshift up to $z\\sim11$, giving future JWST surveys a direct model-based expectation to test.","The slope peak near $z\\approx3$ coincides with the peak of the cosmic star formation rate density, implying that the high-mass end of the main sequence was evolving fastest at cosmic noon.","The model implies that the decline of the $\\Sigma_{\\rm SFR}{\\rm MS}$ slope toward low redshift is a signature of massive-galaxy quenching, visible even without invoking bulge feedback as the dominant driver.","Individual $z>10$ galaxies should fall within roughly $0.4$ dex of the simulated $\\Sigma_{\\rm SFR}{\\rm MS}$, a testable prediction for the growing sample of spectroscopically confirmed high-redshift galaxies.","The dominance of massive galaxies in the cosmic star formation rate density below $z\\sim4.5$ connects the main-sequence slope evolution to the overall history of star formation in the Universe."],"supporting_citations":[{"why":"Supplies the L-Galaxies2020 semi-analytic model used for all predictions and its calibration.","marker":"Parente et al. (2023)"},{"why":"Provides the base Munich SAM that L-Galaxies2020 extends with dust physics and updated disk instability treatment.","marker":"Henriques et al. (2020)"},{"why":"Defines the observed $\\Sigma_{\\rm SFR}{\\rm MS}$ at $z=0,1,2$, the primary low-redshift comparison.","marker":"Salim et al. (2023)"},{"why":"Provides the JWST high-redshift observations at $z=5$ and $7.5$ and the individual $z>4$ galaxies used for comparison.","marker":"Calabrò et al. (2024)"},{"why":"Provides the H$_2$-surface-density star formation law implemented in the SAM.","marker":"Bigiel et al. (2011)"},{"why":"Gives the reference cosmic star formation rate density that the simulated CSFRD is compared against.","marker":"Madau & Dickinson (2014)"}],"fun_headline_variants":["Star-formation surface density main sequence slope peaks at z≈3","Model: star-formation surface density drops 3.5 dex since z=12","Simulated star-formation surface density relation matches observed z=0-7.5","Massive galaxies' rapid SFR fall drives main-sequence slope decline"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The model's simulated half-light radii match real galaxy sizes at every redshift, even though $\\Sigma_{\\rm SFR}$ depends on the square of size and those sizes are not independently validated against the JWST samples used here.","fun_headline_variants_meta":{"raw":{"variants":["Star-formation surface density main sequence slope peaks at z≈3","Model: star-formation surface density drops 3.5 dex since z=12","Simulated star-formation surface density relation matches observed z=0-7.5","Massive galaxies' rapid SFR fall drives main-sequence slope decline"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001183,"raw_usage":{"total_tokens":5051,"prompt_tokens":1279,"completion_tokens":3772,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":895,"completion_tokens_details":{"reasoning_tokens":3689}},"tokens_in":895,"tokens_out":3772,"duration_ms":23976,"temperature":1.0,"reasoning_tokens":3689,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T05:38:55.604656+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A mass-complete sample of galaxies with measured half-light radii and SFRs at $z\\approx3$ and $z\\approx0$ would settle the central claim: the model predicts a $\\Sigma_{\\rm SFR}{\\rm MS}$ slope of $0.71\\pm0.02$ at $z\\approx3$ and $0.085\\pm0.02$ at $z=0$, along with a ~5000% drop in SFR for $10^{11}\\,M_\\odot$ galaxies. Observing a significantly flatter high-redshift slope or a smaller SFR drop would falsify the explanation.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the base Munich SAM that L-Galaxies2020 extends with dust physics and updated disk instability treatment."},{"cited_title":"2023, , 958, 183","cited_arxiv_id":null,"evidence_quote":"Defines the observed $\\Sigma_{\\rm SFR}{\\rm MS}$ at $z=0,1,2$, the primary low-redshift comparison."}],"review_version":1}