{"id":"fb9d89ef-1141-4681-87c7-d6c2a6573d77","arxiv_id":"2504.15283","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":15,"one_line_summary":"Adding the number density of bright sub-millimetre galaxies as a direct MCMC constraint produces calibrated L-Galaxies models that match SMG counts while remaining consistent only with the lower limits of the observed massive quiescent galaxy density.","lead":"This paper tests nine ways of calibrating the L-Galaxies galaxy formation model and finds one setup that reproduces the number density of dusty star-forming galaxies while still underproducing massive quiescent galaxies. A smart generalist might read it to see how automated parameter fitting can reveal that galaxy formation models contain many nearly equivalent solutions.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Predicted SMG densities and fitted physics hinge on an untested 40% dust-to-metal ratio; a factor-of-two change could shift the 'no HIMF' best-fit parameters.","rationale":"The reader's weakest_assumption identifies the Cochrane scaling relations plus the 40% dust-to-metal assumption as the key risk; I agree. This is the most load-bearing assumption because the paper's central contribution is the addition of nSMG as a calibration constraint, and the only way that constraint enters the likelihood is through a mapping from model cold-gas metals to S870. If that mapping is biased, every downstream conclusion--including the 'no HIMF' model's viability and the claimed high merger efficiencies and halo-mass-independent SMBH accretion--is predicated on a miscalibrated target. The paper acknowledges the assumption in Section 5.4 but asserts it is 'unlikely to significantly affect our results' without a sensitivity test; that assertion is not supported. The degeneracy analysis in Section 4.3 further weakens the physical interpretation, but it does not replace the calibration concern: even a perfect posterior would not rescue the science if the single new observable is systematically offset. The proposed two-point dust-to-metal ratio test is inexpensive relative to the original million-CPU-hour campaign and would directly show whether the best-fit parameters and the 'reasonable match' statement move. Because the paper is otherwise transparent and the framework contribution is genuine, the existing CONDITIONAL verdict should stand; no new rejection is warranted pending this test.","tokens_in":37485,"tokens_out":7389,"duration_ms":67894,"concrete_test":"Re-run the 'no HIMF' MCMC refinement (final 2000 steps) with the dust-to-metal ratio changed from 0.4 to 0.2 and to 0.8, recomputing predicted nSMG at z=2.8 and all observables. If the best-fit alpha_SF,burst or fBH/VBH moves outside the 16-84th percentile ranges shown in Appendix B, or if predicted nSMG shifts by more than a factor of 1.5, the abstract's physical requirements are not robust to this assumption.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The new nSMG constraint is implemented by computing S870 from the Cochrane et al. (2023a) scaling relations applied to L-Galaxies outputs (Section 3.3.3). Since the Henriques et al. (2020) model does not track dust, the dust mass entering S870 is set to 40% of cold-gas metals (Section 5.4). The MCMC likelihood for the only DSFG-related constraint therefore depends directly on this single scalar assumption. If the true dust-to-metal ratio for the massive, high-z population differs by a factor of two--which is within the scatter of the Remy-Ruyer et al. (2014) relation cited in support--the predicted S870 for a given simulated galaxy changes substantially, and the number of objects above the 5.2 mJy threshold shifts. The MCMC would then compensate with different SFRs and dust reservoirs, moving the best-fit parameters in Appendix B, including alpha_SF,burst and the SMBH accretion parameters fBH/VBH that underlie the abstract's 'requires' statements. No sensitivity run or error propagation is provided for this assumption. Moreover, the 'no HIMF' model itself underpredicts nSMG at z=2.8 by more than a factor of 2.5 (Section 4.1.4), so the claimed 'reasonable match' already sits several sigma below the adopted calibration point; the dust assumption is not a minor detail but a load-bearing calibration choice.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper uses the MCMC mode of the L-Galaxies semi-analytic model to calibrate 15 free parameters against nine combinations of observational constraints, including a newly implemented constraint on the number density of bright sub-mm galaxies (S870 >= 5.2 mJy at z ~ 2.8), computed from the Cochrane et al. (2023a) scaling relations with an assumed dust-to-metal ratio of 40%. After calibrating each configuration, the authors run the full Millennium volume and compare predictions for stellar mass functions, quiescent fractions, the HIMF, sub-mm number counts, redshift distributions, massive quiescent number densities, cosmic star formation rate density, and SMBH mass functions. They find that configurations including the nSMG constraint reproduce the cumulative S870 number counts within a factor of a few but underpredict the adopted nSMG at z = 2.8 by at least a factor of 2.5; none of the models matches the median high-redshift MQ number density, though some (notably 'no HIMF') reach the observational lower limits. The 'no HIMF' configuration is presented as the best compromise, and the paper argues that it requires high merger-induced star-formation efficiencies and a nearly halo-mass-independent SMBH cold-gas accretion rate. A degeneracy analysis (Section 4.3) shows that models within 1 dex of the best-fit likelihood span a wide range of physical scaling relations.","tokens_in":37900,"tokens_out":7828,"duration_ms":62174,"significance":"The paper's contribution is methodological and empirical: it systematically explores nine constraint configurations within the L-Galaxies MCMC framework and introduces, for the first time in this framework, a sub-mm number density as a direct calibration constraint. The resulting mock catalogues and number-count predictions are falsifiable, and the comparison to observations is transparent. If the central result holds, the 'no HIMF' configuration is a viable L-Galaxies variant that improves the long-standing SMG-MQ tension, and the paper demonstrates the value of robust calibration in exposing degeneracies. The explicit analysis of model degeneracy (Section 4.3) is a strength, as it quantifies how little the current observables constrain the underlying physical scalings. However, the claimed physical requirements are not yet robust because of the sensitivity to the dust-to-metal assumption and the degeneracy among equally good fits.","major_comments":[{"comment":"The nSMG constraint is computed from S870 estimated with the Cochrane et al. (2023a) scaling relations applied to L-Galaxies galaxies, with the dust mass set to 40% of cold-gas metals (Section 3.3.3 and Section 5.4). Because L-Galaxies does not track dust, the likelihood for the only DSFG-related constraint depends directly on this single scalar assumption. The paper asserts that this is unlikely to affect the results, citing Remy-Ruyer et al. (2014), but provides no sensitivity test or error propagation. A factor-of-two change in the dust-to-metal ratio - within the scatter of the cited relation - would shift the predicted S870 and the number of galaxies above 5.2 mJy, and the MCMC would compensate with different best-fit parameters (e.g., alpha_SF,burst and the SMBH growth parameters f_BH and V_BH in Appendix B). The claim that the model 'requires' high merger star-formation efficiencies and halo-mass-independent SMBH accretion is therefore contingent on an unquantified calibration choice. Please add a sensitivity run or explicitly propagate the uncertainty in the dust-to-metal ratio.","section":"3.3.3, 5.4"},{"comment":"The abstract states that the model 'reasonably matches the number density of DSFGs,' but the calibration target itself - nSMG at z=2.8 with S870>=5.2 mJy - is not reproduced: all configurations except 'no low-z SMF, fQ' underpredict this number density by at least a factor of 2.5, and the highlighted 'no HIMF' configuration is among those. The support for the abstract claim comes from the cumulative number counts across flux density (Figure 7), which is a different, weaker statistic, and from the z=2.0 point. Please either soften the abstract and conclusions to state that the model approaches or is within a factor of a few of the observed nSMG, or provide a justification for why a factor-2.5 miss at the adopted calibration point is acceptable.","section":"4.1.4, Figure 6"},{"comment":"The abstract's causal claim that the identified model 'requires high star formation efficiencies in mergers and a null dependency of SMBH cold gas accretion on halo mass' is not supported by the paper's own degeneracy analysis. Figure 15 shows that parameter sets within 1 dex of the best-fit likelihood produce scaling relations spanning roughly two orders of magnitude, and for three of the four scalings the best-fit model does not occupy the most densely populated region. Thus the quoted physical properties are properties of a single best-fit point, not robust features of the acceptable model ensemble. The paper should either demonstrate that these physical scalings are common to the high-likelihood region or explicitly restrict the abstract and conclusions to the best-fit model.","section":"4.3, Figure 15; Abstract"}],"minor_comments":[{"comment":"The convergence criterion ('no new accepted point with a higher likelihood within the final ~1000 steps') is not a standard convergence diagnostic; please report a quantitative measure such as the Gelman-Rubin statistic or split-chain acceptance diagnostics.","section":"3.5"},{"comment":"The average relative errors quoted in Section 3.2 (~10%, ~25%, ~5%, ~20% for SMFs, f_Q, nSMG, and HIMF) are not consistent with the values in Appendix A, where the f_Q error at z=2.8 is reported as 35% and the nSMG error as ~2%. Please harmonize the numbers.","section":"3.2, Appendix A"},{"comment":"The x-axis of Figure 6 is labelled only 'Configuration' with numbers 0-8; please add tick labels matching Table 1 (e.g., 'baseH20', 'baseL20', etc.) to make the figure self-contained.","section":"Figure 6"},{"comment":"The text says 'To date, five more recent modifications of L-Galaxies have been published' but then lists six references (Yates et al. 2021; Ayromlou et al. 2021; Izquierdo-Villalba et al. 2022; Murphy et al. 2022; Spinoso et al. 2023; Yates et al. 2024). Please correct the count or the reference list.","section":"2"},{"comment":"The phrase 'we further modified L-Galaxies to include this factor in the SfrInst parameter' is vague; please specify exactly how the merger-induced starburst SFR is added to the instantaneous SFR, as this directly affects the S870 estimates.","section":"3.3.3"}],"recommendation":"major_revision","confidential_remarks":"The paper would benefit from reframing as a methods and calibration study; the central 'simultaneous match' claim is stronger than the data support. The lack of a sensitivity analysis for the dust-to-metal ratio is the most serious technical gap; without it, the quoted best-fit parameters and the abstract's causal statements are not robust. I would encourage the editor to request a sensitivity run and a softened abstract as part of the major revision. The 'first time' claim about including SMG number density as a constraint should also be checked carefully against other SAM calibration efforts."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe short version: this paper deserves a serious referee, but the abstract oversells the physical conclusions. What it actually does—adding the SMG number density as an MCMC constraint in L-Galaxies, with a purpose-built merger-tree sampling scheme—is a genuine step forward, and the central calibration result is supported by the figures. Configurations without the SMG constraint underproduce the 870-micron counts by more than two orders of magnitude; with it, they land within a factor of a few, and the 'no HIMF' model also sits near the lower limits of the observed massive-quiescent number density. The comparison across nine constraint sets, the mock catalogues for number counts, and the out-of-sample redshift distributions are all done carefully. Full credit for the degeneracy analysis in Figure 15; it is an honest look at how much the physical scalings vary among similar-likelihood models.\n\nThe soft spots are real, though. The 'no HIMF' model underpredicts nSMG at z=2.8 by a factor of about 2.5, so calling that a 'reasonable match' is generous. More important, the SMG constraint itself is computed from the Cochrane et al. scaling relations assuming 40% of cold-gas metals are dust. The paper states this assumption is unlikely to matter but gives no sensitivity run. The stress-test note is right: if the dust-to-metal ratio is off by a factor of two—well within the scatter of the relation they cite—the predicted S870 for a given galaxy shifts substantially, and the MCMC will compensate by moving the fitted parameters, including the merger SFE and SMBH accretion parameters that anchor the abstract's 'requires' claim. That is a load-bearing calibration choice, not a detail.\n\nThe abstract's physical statements also go beyond what the degeneracy analysis supports. Their own Figure 15 shows that the best-fit scalings rarely sit in the most populated region of similar-likelihood models, so 'requires high star formation efficiencies in mergers and a null dependency of SMBH cold gas accretion on halo mass' should be read as 'preferred by one configuration under a fixed dust assumption,' not as a robust inference. This is addressable by softening the language and adding a sensitivity test.\n\nThe code and data for the modifications are 'available upon request,' which is weak for a calibration paper whose value depends on reproducibility. A referee should ask for the modified code or at least the specific parameter outputs in machine-readable form.\n\nBottom line: the framework is a real contribution and the paper is worth engaging with, but it needs a careful revision on the dust sensitivity and on how strongly the physical conclusions are phrased. Send it to referees, not to the desk bin.","headline":"A serious calibration framework worth refereeing, but the dust-to-metal assumption is load-bearing and the abstract's physical claims should be softened.","tokens_in":38419,"tokens_out":3418,"would_cite":false,"duration_ms":30721,"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":"One calibrated galaxy-formation model now fits both dusty starbursts and massive quiescent galaxies.","keywords":["semi-analytic models","MCMC calibration","dusty star-forming galaxies","sub-millimetre galaxies","massive quiescent galaxies","L-Galaxies","AGN feedback","merger-induced starbursts"],"falsifier":"Run the identical calibration with sub-mm fluxes computed from a dust-tracking version of L-Galaxies or from radiative-transfer post-processing; if the resulting SMG number-density target moves significantly, the 'no HIMF' parameters are an artifact of the scaling relations. A sharper observational test would be an MQ number-density measurement at z~3-4 whose central value sits clearly above the lower limits used here, since that would rule out the 'no HIMF' compromise.","tokens_in":37345,"feed_emoji":"🌌","tokens_out":11633,"duration_ms":97431,"temperature":0.7,"pith_summary":"This paper asks whether the long-standing mismatch between dusty star-forming galaxies (DSFGs) and massive quiescent galaxies (MQs) in galaxy formation models can be cured by systematic calibration rather than by changing the physics by hand. The authors use the MCMC mode of the L-Galaxies semi-analytic model to fit 15 free parameters against nine different combinations of constraints, adding the number density of bright sub-millimetre galaxies as a direct calibration target for the first time. Among the nine best-fit models, one configuration — 'no HIMF', which drops the local neutral-hydrogen mass function from the constraint set — matches the observed 870-micron number counts and stays consistent with the lower limits on the high-redshift massive quiescent population. The price of that compromise is specific physics: high star-formation efficiency in merger-driven starbursts and supermassive-black-hole growth by cold-gas accretion that does not depend on halo mass. The result matters because it demonstrates that the SMG-MQ tension can be partly absorbed by parameter-space exploration, while also exposing how degenerate the fitted mechanisms are.","feed_headline":"One recalibrated model fits dusty starbursts and massive quiescents","feed_subtitle":"Adding sub-millimetre number counts to the fit lets one L-Galaxies model match both populations.","key_machinery":"The carrying object is the MCMC calibration mode of the L-Galaxies semi-analytic model, which explores 15 free parameters against chosen observables using a representative sample of dark-matter merger trees. Two new pieces are added: a merger-tree sampling procedure that keeps the rare SMG population representative, and a direct SMG number-density constraint built by converting simulated galaxies to 870-micron flux densities with the Cochrane et al. (2023a) scaling relations. The diagnostic that carries the physical interpretation is a set of fitted scaling laws — the merger starburst efficiency $M_{\\star,\\mathrm{burst}}/M_{\\mathrm{cold}}$ vs mass ratio, the SN feedback efficiencies vs $V_{\\mathrm{max}}$, and the SMBH cold-gas accretion fraction $\\Delta M_{\\mathrm{BH,Q}}/M_{\\mathrm{cold}}$ vs $V_{\\mathrm{200c}}$. The 'no HIMF' configuration's signature is a nearly flat SMBH accretion fraction combined with steep merger-starburst efficiency, which together let galaxies build stars and black holes rapidly and then quench.","core_discovery":"The central claim is that the L-Galaxies SAM contains a parameter combination that simultaneously reproduces the observed number density of DSFGs and the observationally-derived lower limits for high-redshift massive quiescent galaxies. This is configuration 'no HIMF', calibrated against the Leja et al. (2020) stellar mass functions and Leja et al. (2022) quiescent fractions at $z=0.4$ and $z=2.8$, plus the AS2UDS SMG number density at $z\\sim2.8$. Its best-fit parameters imply that merger-induced starbursts convert a large fraction of cold gas into stars even in minor mergers, and that the fraction of cold gas accreted by SMBHs during mergers is nearly flat with halo virial velocity, so massive black holes can grow quickly. The paper also claims this fit produces an SMBH mass function at $z=0$ in better agreement with observations than the other configurations, while the fitted models that instead match DSFGs best fail on the quiescent population, and vice versa. The authors frame the result as a step toward resolving the tension, not a unique solution: nearly equally good fits span about two dex in feedback efficiencies.","pith_inferences":["The same calibration recipe could be transferred to other semi-analytic and cosmological simulation pipelines, offering a route to the SMG-MQ compromise without changing the IMF.","The fitted halo-mass-independence of SMBH cold-gas accretion is probably a stand-in for a more physical trigger: the MCMC finds that rapid black-hole growth must accompany rapid merger starbursts, so an explicit coupling between starburst-driven turbulence and black-hole feeding could reproduce the same observable.","Because the 40% dust-to-metal assumption is fixed rather than fitted, repeating the calibration with a dust-evolution model would directly test whether the claimed parameters are an artifact of the dust recipe rather than a required property of galaxy formation.","The representative merger-tree selection method should work for any rare tracer with a defined number density, so similar one-point constraints could bring other rare populations — for example high-redshift quasars or bright Lyman-alpha emitters — into MCMC calibration loops."],"forward_implications":["Adding a single SMG number-density constraint at $z\\sim2.8$ lifts predicted 870-micron number counts from more than two orders of magnitude below observation to within a factor of a few across an order of magnitude in flux density.","When the SMG constraint is included, the model that also matches the massive quiescent lower limits is 'no HIMF'; dropping the $z=0$ HI mass function does not break its HI predictions, so the HIMF appears to be a less critical constraint than previously assumed.","The best-fit physics of the 'no HIMF' model — high merger-driven starburst efficiencies and halo-mass-independent SMBH cold-gas accretion — provides a concrete, testable alternative to top-heavy IMF modifications for producing bright SMGs.","The $z=0$ SMBH mass function is not used in calibration, yet it discriminates between the two SMBH growth modes (peaked vs Schechter-like), so future calibrations that include it should break some of the degeneracies shown in Figure 15."],"supporting_citations":[{"why":"Supplies the scaling relations that convert simulated SFR, stellar mass, dust mass, and redshift into 870-micron flux densities, so it defines the SMG number-density constraint.","marker":"Cochrane et al. (2023a)"},{"why":"Provides the AS2UDS catalogue and the bright-SMG sample from which the SMG number-density calibration target at z~2.8 is measured.","marker":"Dudzevičiūtė et al. (2020)"},{"why":"Replaces the older stellar mass functions with the continuity-model SMFs at z=0.4 and 2.8, driving the higher low-redshift normalization and the CSFRD overprediction.","marker":"Leja et al. (2020)"},{"why":"Supplies the quiescent fractions as a function of stellar mass used as fQ constraints.","marker":"Leja et al. (2022)"},{"why":"Supplies the base L-Galaxies model, its MCMC mode, and the 15 free parameters varied in the calibration.","marker":"Henriques et al. (2020)"},{"why":"Provides the observational compilation of massive quiescent number densities at 3