{"id":"cc350460-2d1c-4dc5-9755-b0fbda4f3ba5","arxiv_id":"2602.22384","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Using a new deblended Herschel catalog, the authors show that UV-to-mid-infrared SED fits recover the star-forming main sequence but under-predict far-infrared luminosity in low-mass galaxies when fixed dust templates are used.","lead":"A new catalog combines sharp Hubble images with blurry Herschel far-infrared maps to measure dust emission from 41,000 distant galaxies. Comparing fits with and without the far-infrared data shows that the star-forming main sequence is unchanged, but that fixed mid-infrared-to-total-infrared conversions under-estimate dust emission in low-mass galaxies.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"SFMS validation is partly circular because Herschel photometry for faint sources is extracted using Prospector UV-MIR flux priors; the paper's claim that data dominate lacks a quantitative test.","rationale":"The reader's weakest assumption correctly identifies the circular dependency between the Herschel photometric extraction and the UV-MIR models being validated. This is the single most load-bearing concern because it directly undermines the paper's claim that adding FIR data confirms the Leja et al. (2022) SFMS. The paper makes a plausible argument that the data dominate for well-detected sources, and the IR-bright sample is indeed where the F_TIR/F7.7 trend is most secure, but the absence of a quantitative test for the faint, prior-dominated regime leaves the central validation unproven. I agree with the reader's CONDITIONAL verdict: the paper is transparent, the catalog and methods are valuable, and the concern is testable rather than fatal. No change to the verdict is needed; a mock-injection or flat-prior re-extraction test would settle the issue.","tokens_in":38507,"tokens_out":7037,"duration_ms":66521,"concrete_test":"Run mock-injection experiments with the standard T-PHOT pipeline: insert artificial sources with known FIR SEDs (some matching the Prospector prior, some deliberately deviating in U_min and q_PAH) into the real Herschel maps at random positions, recover them with the same Gaussian priors, and compare input vs. recovered fluxes as a function of SNR (0.5–5) and SED shape. Then re-run the Prospector emulator fits using the mock-perturbed photometry and measure the resulting SFMS offset and F_TIR/F7.7 trend. If the recovered fluxes for SNR<3 sources are systematically pulled to the prior by >0.1 dex, or if the SFMS offset changes by >0.1 dex when using flat/weak priors, the SFMS validation is circular and the F_TIR/F7.7 claim needs re-evaluation at low mass.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central validation claim—that adding Herschel FIR data leaves the Prospector SFMS unchanged (0.1±0.07 dex offset)—is weakened by the fact that the Herschel photometry is not independent of the UV-MIR models being tested. As described in §2.2.1, T-PHOT uses Gaussian flux priors derived from Prospector UV-MIR fits. For the >90% of the mass-limited sample that are Herschel upper limits (SNR<3), the data are uninformative and the extracted fluxes will be pulled toward the priors by construction. Re-fitting these 'Herschel-constrained' fluxes with the same Prospector model therefore trivially reproduces the UV-MIR outputs, including the SFMS ridge. The external photometric validation in §3 (Figure 1) applies mainly to bright, detected sources and does not cover the faint regime where the priors control the solution. The paper asserts in §2.2.3 that sources with S/N<1 are 'driven toward fluxes consistent with the image noise rather than the prior model values,' but no test is shown to verify that the resulting flux posteriors are not biased toward the prior. The F_TIR/F7.7 result for the IR-bright subsample is less vulnerable because those sources have strong detections, but the same prior issue could still affect some bands (e.g., SPIRE350, where the error correction is largest), and the χ²<3 cut in §6 may select objects for which the deblending happened to work. Thus the load-bearing concern is that the SFMS agreement is not an independent confirmation of Leja et al. (2022), but rather a consequence of the prior-dominated extraction; a quantitative test of prior-induced bias is missing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents 3D-Herschel, a new 0.3–350 micron photometric catalog built by deblending Herschel imaging in the CANDELS/3D-HST fields with T-PHOT, using Prospector UV-MIR fits as flux priors. The authors fit 41,387 galaxies at 0.5<z<2.5 with a 17-parameter Prospector emulator, comparing fits with and without Herschel FIR data. They report that UV-MIR-only fits recover stellar masses, SFRs, and the star-forming main sequence for the majority of objects (SFMS agreement of 0.1±0.07 dex), but that fixed IR templates in UV-MIR fits give colder dust temperatures (18.4 K vs. 25.4 K) and underestimate F_TIR/F7.7 by ~0.2 dex at low stellar masses. They conclude that MIR-to-IR conversions depend on stellar mass and that Herschel provides mostly upper limits for distant low-mass galaxies.","tokens_in":38892,"tokens_out":6146,"duration_ms":60230,"significance":"If the results hold, this paper provides a valuable public UV-FIR catalog and one of the largest panchromatic SED-fitting samples at cosmic noon. The mass-dependent F_TIR/F7.7 trend would strengthen JWST/MIRI-based results and caution against fixed MIR-to-IR templates in SED modeling. The paper's strengths include the public data release, multi-field deblending with explicit PSF treatment, Monte Carlo error calibration, external flux validation, and the use of a validated emulator. These assets make the study a useful reference for future FIR and JWST analyses, provided the independence of the FIR photometry from the models being validated is demonstrated.","major_comments":[{"comment":"The central validation claim — that UV-MIR Prospector fits recover the SFMS unchanged when FIR data are added — is weakened by a circularity in the photometric pipeline. T-PHOT fluxes are extracted with Gaussian priors derived directly from Prospector UV-MIR fits (§2.2.1), and the same extracted fluxes are used in §5 and §6.2 to validate those fits and to measure the 0.1±0.07 dex SFMS agreement. For the >90% of the mass-limited sample that are Herschel upper limits (SNR<3), the data cannot dominate the priors, so agreement between with- and without-Herschel fits is partly by construction. The statement in §2.2.3 that S/N<1 sources are driven to noise-consistent fluxes is not supported by a quantitative test. Please add an extraction test with uninformative or no flux priors on a representative subset, and report how the SFMS and IR-bright classifications change.","section":"§2.2.1, §5, §6.2"},{"comment":"The headline F_TIR/F7.7 mass trend is measured on an IR-bright sample of 1118 objects defined after removing ~33% of candidates with chi^2<3 in PACS100/160, often attributed to 'bright interlopers contaminating the photometry of low-mass galaxies.' Because this cut is applied after fitting, it can imprint or exaggerate the low-mass trend. The paper does not report the number or mass distribution of rejected objects, nor whether the trend survives a less aggressive cut or an analysis that includes the rejected objects with inflated uncertainties. Please quantify the selection function and show the robustness of the F_TIR/F7.7 versus stellar mass result to the chi^2 threshold.","section":"§6, Table 6, Figure 9"},{"comment":"The comparison with Shivaei et al. (2024) shows a ~0.28 dex offset in F_TIR/F7.7 at high stellar masses, which the paper attributes to differences in the U_min prior and possible calibration issues. However, the same section reports that model MIPS 24 micron fluxes are ~0.1 dex below observed fluxes, and the paper states this does not fully explain the discrepancy. Since F_TIR/F7.7 is a main physical conclusion, this unresolved systematic should be tested explicitly — for example, by recomputing F7.7 with the observed 24 micron flux as a constraint, or by adopting the Shivaei et al. U_min prior and showing how the mass trend changes.","section":"§7.5, Figure 9"}],"minor_comments":[{"comment":"The running header says 'McNulty et al. 2024' while the paper is dated 2026; update for consistency.","section":"General"},{"comment":"Equation (3) defines R_SB but the text refers to R_MS > 2; use one notation consistently.","section":"Eq. (3) and §6.3"},{"comment":"State explicitly which model (with or without Herschel) is used to compute the chi^2<3 cut, and report how many objects are removed per field and per mass bin.","section":"§6"},{"comment":"The caption says the dashed line marks the Shivaei et al. relation, but §7.5 describes a 0.28 dex offset relative to that relation; clarify what is plotted.","section":"Figure 9"},{"comment":"The bottom-row axis label 'Delta log(SFR IR Bright)' is difficult to parse; use 'Delta log SFR (IR-bright)' or similar.","section":"Figure 8"},{"comment":"The abstract and conclusions state that 3D-Herschel is publicly released, but no URL, DOI, or data-access footnote is provided; add one.","section":"Data availability"}],"recommendation":"major_revision","confidential_remarks":"I see no concerns about novelty or citation practices. The main technical issue is the prior circularity: the FIR photometry used for validation is extracted with priors from the same models being validated, and the paper does not provide an uninformative-prior test. This is fixable within the manuscript's scope but should be addressed before acceptance. A second key point is the poorly characterized chi^2 selection of the IR-bright sample, which affects the main F_TIR/F7.7 result."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, 3D-Herschel is a genuinely useful product: publicly released deblended PACS/SPIRE photometry for 3D-HST/CANDELS, validated against external catalogs, plus 41k Prospector fits with and without FIR data. Second, the paper's central validation claim—that adding Herschel data leaves the SFMS unchanged—is weaker than it looks, because the Herschel fluxes for the majority of sources are not independent of the UV-MIR models being tested.\n\nWhat it does well: the deblending pipeline is carefully described, the Monte Carlo error calibration is a real effort, and the emulator is checked against full Prospector on a thousand galaxies. The catalog comparisons in Figure 1 show reasonable agreement, with acknowledged SPIRE350 offsets. The F_TIR/F7.7 mass trend in the IR-bright subsample is a new quantitative result, directionally consistent with Shivaei et al. 2024 and extended to lower stellar mass. That 0.2 dex underestimate at log M* ~ 9.6 is worth taking seriously.\n\nThe soft spots are real, and the circularity concern lands. In Section 2.2.1, T-PHOT uses Gaussian flux priors derived from Prospector UV-MIR fits. For the >90% of the mass-limited sample that are Herschel upper limits, the data cannot overwhelm those priors; refitting those \"constrained\" fluxes with the same model is close to a tautology. The paper asserts in §2.2.3 that S/N<1 sources are driven by image noise rather than priors, but shows no test. So the SFMS agreement (0.1±0.07 dex) is not an independent confirmation of Leja et al. 2022. That does not kill the paper, but it should change how the SFMS paragraph is read.\n\nThe F_TIR/F7.7 result is less vulnerable because it is restricted to IR-bright sources with SNR>=3 in at least two bands. But two issues remain: the sample is defined after a chi^2<3 cut that removes 33% of the IR-bright candidates, and the flux priors could still affect SPIRE350, where the error correction is largest. The comparison with Shivaei et al. also shows a 0.27 dex offset in L7.7 that is not fully explained. The authors should show the F_TIR/F7.7 trend survives a no-prior extraction test and a less aggressive sample cut.\n\nBottom line: this deserves a serious referee. The catalog alone justifies peer review, and the mass-dependent F_TIR/F7.7 result is important if it holds up. Recommend major revision: add a no-prior photometric extraction test, quantify how much of the SFMS agreement is prior-driven, and report the F_TIR/F7.7 trend without the chi^2 cut. Then it will be a solid contribution.","headline":"Useful new public catalog, but the SFMS validation is partly circular; the mass-dependent F_TIR/F7.7 result is the real news and needs a no-prior test.","tokens_in":39497,"tokens_out":3543,"would_cite":true,"duration_ms":33024,"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":"Adding far-infrared Herschel data to galaxy SED fits leaves the star-forming main sequence unchanged but reveals that mid-infrared-to-total-infrared conversions depend on stellar mass.","keywords":["galaxies: photometry","galaxies: fundamental parameters","infrared: galaxies","dust, extinction","galaxies: star formation","star-forming main sequence","far-infrared","spectral energy distribution fitting"],"falsifier":"Extract the same Herschel imaging with uninformative (flat) flux priors for the IR-bright sample and rerun the fits; if the total-infrared-to-7.7-micron excess at log(M*/M_sun)~9.6 disappears, the mass-dependent offset is an artifact of prior-driven deblending. Independently, JWST/MIRI measurements of 7.7 micron fluxes for the same galaxies would confirm or refute the suppressed PAH strength at low mass.","tokens_in":38386,"feed_emoji":"🔭","tokens_out":8614,"duration_ms":77218,"temperature":0.7,"pith_summary":"This paper builds a new public 0.3-350 micron photometric catalog for 41,387 galaxies and shows that adding Herschel far-infrared constraints to a flexible 17-parameter Bayesian spectral-energy-distribution fit does not shift the star-forming main sequence relative to fits using only ultraviolet-to-mid-infrared data. The average deviation between the two inferred sequences is 0.1±0.07 dex, supporting earlier claims that the low normalization of the sequence is not a side effect of missing far-infrared measurements. However, for the 3.2% of galaxies with strong Herschel detections in at least two bands, the fixed infrared templates used in UV-MIR-only fits prove inadequate: they imply dust temperatures about 7 K colder and, at stellar masses near 10^9.6 solar masses, total-infrared-to-7.7-micron ratios about 0.2 dex lower than fits that include the far-infrared data. The paper concludes that conversions from mid-infrared to total infrared luminosity depend on stellar mass and cautions against mass-independent templates.","feed_headline":"Star-forming main sequence passes far-infrared test","feed_subtitle":"Herschel data confirm galaxy star-forming sequence and expose a mass-dependent infrared conversion flaw.","key_machinery":"The load-bearing piece is a new 0.3-350 micron photometric catalog, 3D-Herschel, built by deblending low-resolution Herschel imaging with forced photometry that uses high-resolution HST positions and UV-MIR-model flux predictions as priors. On top of this catalog, the analysis compares two configurations of a 17-parameter Bayesian SED fitting framework with nonparametric star-formation history, two-component dust attenuation, and energy-balance dust re-emission: one using only UV-24 micron data with the dust-emission parameters fixed to log-average template colors, and one adding Herschel data with the three dust-emission parameters (polycyclic aromatic hydrocarbon mass fraction, minimum rad","core_discovery":"The paper's central claim is that, at least for the bulk of the mass-complete population, SED fits that stop at 24 microns are not biased in their stellar masses, star-formation rates, or the shape of the star-forming main sequence: adding Herschel photometry moves the inferred SFR ridge by only 0.1±0.07 dex. The far-infrared data do matter, though, for the dust side of the model. When the three dust-emission parameters are allowed to vary, Herschel-constrained fits recover warmer characteristic dust temperatures (about 25 K versus 18.4 K for the fixed-template UV-MIR fits) and a higher total-infrared-to-7.7-micron ratio at low stellar mass (about 0.2 dex at log M* ~ 9.6). This mass dependen","pith_inferences":["Because the Herschel extraction used UV-MIR model predictions as flux priors, the agreement between UV-MIR and UV-FIR fits may be partly inherited from those priors; a rerun with uninformative priors on the IR-bright subsample would test how much of the 0.2 dex total-infrared-to-7.7-micron offset is independent of the priors.","If the mass-dependent trend is driven by PAH destruction in strong radiation fields, the ratio should correlate with metallicity and radiation-field intensity at fixed stellar mass; that correlation is implicit but not directly tested here and is checkable with JWST/MIRI spectra of the same galaxies.","A direct prediction for future PRIMA-class observations is that the mass dependence of the mid-infrared-to-total-infrared conversion should become more pronounced at higher resolution, since Herschel's confusion limit hides the lowest-mass, dustiest systems.","The comparison with a JWST/MIRI study suggests part of the 7.7-micron offset may be a calibration issue; stacking deeper far-infrared data or using MIRI photometry for the same sample would decide whether the mass dependence persists."],"forward_implications":["The Herschel-constrained star-forming main sequence deviates from UV-MIR-only fits by 0.1±0.07 dex at fixed stellar mass, so the low normalization of the sequence is not caused by missing far-infrared data.","Mid-infrared-only SFR and total infrared luminosity estimates using fixed templates will systematically underestimate total infrared luminosity for low-mass obscured galaxies, by about 0.2 dex near log(M*/M_sun)=9.6.","UV-MIR fits with fixed dust parameters imply dust temperatures about 7 K too cold, biasing dust-based interpretations of these galaxies.","The public 3D-Herschel catalog extends forced deblended photometry to 350 microns, allowing future work to include far-infrared constraints in SED fitting.","For galaxies below 10^11 solar masses at z > 1.5, Herschel can only provide upper limits after deblending; next-generation far-infrared telescopes are needed to measure their dust emission."],"fun_headline_variants":["Far-infrared test confirms star-forming main sequence","Adding Herschel data exposes mass-dependent IR bias","Star-forming sequence robust to FIR, but dust models not","MIR templates fail at low mass without Herschel data","Herschel shows UV-only SFRs unaffected, dust conversions biased"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"Everything rests on the assumption that the deblended Herschel fluxes are not shaped by the UV-MIR model predictions used as priors during extraction, because the same UV-MIR fits are then validated against those Herschel fluxes.","fun_headline_variants_meta":{"raw":{"variants":["Far-infrared test confirms star-forming main sequence","Adding Herschel data exposes mass-dependent IR bias","Star-forming sequence robust to FIR, but dust models not","MIR templates fail at low mass without Herschel data","Herschel shows UV-only SFRs unaffected, dust conversions biased"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000178,"raw_usage":{"total_tokens":1264,"prompt_tokens":1004,"completion_tokens":260,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":748,"completion_tokens_details":{"reasoning_tokens":191}},"tokens_in":748,"tokens_out":260,"duration_ms":3102,"temperature":1.0,"reasoning_tokens":191,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T20:42:24.649877+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Extract the same Herschel imaging with uninformative (flat) flux priors for the IR-bright sample and rerun the fits; if the total-infrared-to-7.7-micron excess at log(M*/M_sun)~9.6 disappears, the mass-dependent offset is an artifact of prior-driven deblending. Independently, JWST/MIRI measurements of 7.7 micron fluxes for the same galaxies would confirm or refute the suppressed PAH strength at low mass.","supporting_citations":[],"review_version":1}