{"id":"17d5e06c-4d3c-4eab-943f-935a030ae840","arxiv_id":"2608.08203","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A self-consistent Hα luminosity function and dust-corrected star formation rate density from z=1.2 to 6.6, measured from JWST medium-band photometry of 4,101 galaxies.","lead":"Using JWST medium-band images from three survey programs, the authors measure the Hα emission line in about 4,100 galaxies to build a star formation census from redshift 1.2 to 6.6, reaching into the epoch of reionization. The result gives a consistent picture of how fast galaxies formed stars across cosmic time, and hints that the star formation rate density may be higher than UV-based estimates, a claim the authors mark as tentative.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 0.2–0.3 dex SFRD excess at z~1–2 rests on a one-parameter dust correction; the paper's own Sec. 5.1 says changing the attenuation law or the 0.44 ratio shifts SFRD by ~0.2 dex, enough to erase the claimed tension.","rationale":"The reader correctly identifies the dust attenuation prescription as the weakest step, and I agree that the observed LF is the solid part of the paper: photometric redshifts are validated against spectroscopy, line fluxes against NIRSpec prisms, and completeness by image simulations. The reported tension with UV/IR SFRDs, however, is a dust-corrected normalization and therefore inherits every assumption in Eq. (2). The paper itself concedes in Sec. 5.1 that plausible changes to the attenuation law or the 0.44 ratio shift the SFRD by ~0.2 dex, which is the same magnitude as the claimed excess. I would not move the verdict to REJECT because the authors label the result tentative and because the observed LF and its evolution are independently supported. I also considered the field-overdensity issue (Sec. 5.2) and the use of LF-derived priors for the SFR function fits (Sec. 4.3); both are real secondary worries, but the dust prescription is more load-bearing because it globally scales every luminosity. I marked partial rather than full agreement with the reader because the reader's summary states A_Hα = 0.44 A_V, whereas the paper's phrase 'stellar to nebular attenuation ratio of 0.44' corresponds to the Calzetti color-excess ratio, making A_Hα roughly 1.9–2.3 A_V; the direction matters for interpreting the tension, even though the underlying single-parameter vulnerability is the same.","tokens_in":20319,"tokens_out":10574,"duration_ms":102612,"concrete_test":"Use the existing CANUCS NIRSpec prism spectra to measure the Balmer decrement Hβ/Hα for sources with useful line detections, derive per-object A_Hα, and recompute the z=1.5 and z=2.0 SFRD by replacing Eq. (2) with these measured attenuations. If the recomputed SFRD moves by more than ~0.15 dex, or if the prism-derived A_Hα values disagree systematically with the fixed Calzetti scaling across the sample, the headline 0.2–0.3 dex excess is a dust-prescription artifact rather than a robust measurement.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is not the observed LF (which is carefully built and validated), but the headline statement that the dust-corrected SFRD is 0.2–0.3 dex above UV/IR estimates and 'eclipsing both.' That normalization is set by Eq. (2), which converts every observed Hα luminosity using a single stellar A_V from DENSEBASIS SED fits plus a fixed Calzetti attenuation law and a fixed stellar-to-nebular ratio (0.44). No rest-frame MIR/FIR data enter per-galaxy; the same scaling is applied at all redshifts. Section 5.1 explicitly states that systematic changes to the attenuation law or the 0.44 ratio shift the SFRD by ~0.2 dex, i.e. exactly the size of the reported excess. The 0.44 is the stellar-to-nebular color-excess ratio of Calzetti, so A_Hα is roughly 1.9–2.3 A_V rather than 0.44 A_V, but the vulnerability is unchanged: one scalar per galaxy, calibrated locally, carries the entire high-normalization result. The per-object correction also flattens the faint end (Sec. 4.3), so the SFR function shape is partly an artifact of the dust prescription. The observed LF can stand while the 'eclipsing both' SFRD claim falls.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper uses NIRCam medium-band photometry from the CANUCS, JWST in Technicolor, and JUMPS surveys to measure the Hα luminosity function over 1.25 < z < 6.6, and then derives dust-corrected star formation rate functions and the cosmic star formation rate density. The photometric redshifts are validated against spectroscopy (outlier fraction 1.7%), line fluxes are validated against NIRSpec prism measurements (0.04 dex offset, 0.23 dex scatter), and completeness is assessed with full image simulations. The observed luminosity function is the primary result. The dust-corrected SFRD is reported to be 0.2–0.3 dex higher than UV-based estimates, with a peak at z ~ 2 and a decline toward z = 6, but this elevation is explicitly described as tentative and dependent on the adopted dust attenuation prescription.","tokens_in":20559,"tokens_out":5234,"duration_ms":52358,"significance":"If the observed luminosity function stands, this is a valuable, self-consistent measurement that bridges cosmic noon and the epoch of reionization, extending Hα studies to fainter luminosities over a wide redshift range with a uniform methodology. The validation against spectroscopy and image simulations is a clear strength, and the public availability of the data adds to its usefulness. However, the headline claim of an elevated, dust-corrected SFRD that 'eclipses' both UV and IR measurements is not established at the same level as the observed LF: it rests on a single per-galaxy dust correction, on priors derived from the observed LF when fitting SFR functions, and on fields that are strongly affected by overdensities. The paper's own error analysis shows that plausible changes to the dust prescription shift the SFRD by about 0.2 dex, the same size as the claimed excess. The manuscript is therefore more compelling as a measurement of the observed Hα LF and its evolution than as a definitive new SFRD normalization.","major_comments":[{"comment":"The entire dust-corrected SFRD ladder is set by Eq. (2) using a single stellar A_V per galaxy, a fixed Calzetti attenuation curve, and the sentence 'We assume a stellar to nebular attenuation ratio of 0.44.' This sentence is ambiguous and potentially misapplied: Calzetti's 0.44 is the ratio of stellar to nebular color excess, E(B-V)_star/E(B-V)_gas, which translates to A_Hα ≈ 1.9–2.3 A_V for the Calzetti curve, not A_Hα = 0.44 A_V as written. The authors should state the exact formula used and verify the numerical implementation. More importantly, Sec. 5.1 itself states that changing the attenuation law or the 0.44 ratio shifts the SFRD by ~0.2 dex, which is exactly the size of the reported excess over UV measurements. Without per-galaxy MIR/FIR constraints, the abstract's claim of eclipsing both UV and IR estimates is not supported by the presented evidence; the paper should present the SFRD with a systematic uncertainty band from alternative dust prescriptions and frame the excess as tentative.","section":"Sec. 5.1, Eq. (2)"},{"comment":"The SFR function fits use 'a prior on the SFR function parameters based on the observed luminosity function priors' because flat priors led to degenerate fits. This introduces a form of circularity: the dust-corrected SFR functions, and hence the integrated SFRD, are not independent of the observed LF shape. The authors should quantify how the fitted SFR functions and the resulting SFRD change under different prior choices, or present the SFRD as an integral of the data (with completeness corrections) rather than relying on the prior-influenced Schechter parameters.","section":"Sec. 4.3"},{"comment":"Field-to-field variations dominate the SFRD in several redshift bins: individual fields contribute >40% of the total SFRD at z ≈ 2, 3, 5, and 6, and the text identifies specific overdensities in MACS0417, MACS1149, and A370. This means that the high-redshift SFRD points and the claimed plateau in the obscured fraction at z > 4 are not robust against cosmic variance, which the formal error bars do not capture. The authors should show the SFRD computed after excluding the overdense fields or otherwise propagate the field-to-field scatter into the quoted uncertainties and into the comparison with UV/IR measurements.","section":"Sec. 5.2, Fig. 8"},{"comment":"The paper notes that the per-object dust correction shifts galaxies from low to high luminosity and thereby flattens the faint-end slope. This is a selection effect introduced by the correction itself, so the resulting dust-corrected LF/SFR function slope is not an independent physical measurement. The text should either demonstrate that the flattening is not an artifact of the A_Hα prescription or clearly separate the observed LF slope evolution from the dust-corrected slope evolution when discussing trends with redshift.","section":"Sec. 4.3, Fig. 4"}],"minor_comments":[{"comment":"The phrase 'though eclipsing both' is colloquial and imprecise; it should be replaced by a quantitative statement, e.g., 'higher by 0.2–0.3 dex than typical UV and IR estimates, with systematic uncertainties of similar magnitude.'","section":"Abstract"},{"comment":"There is a typo in the caption: 'Comprison' should be 'Comparison.'","section":"Fig. 2 caption"},{"comment":"The JUMPS survey is referred to as the 'JWST Ultimate Medium-Band Photometric Survey' in the abstract and as the 'JWST Ultimate Photometric Survey' (listed as 'JWSTUltimate Photometric Survey') in Sec. 2; please make the acronym definition consistent.","section":"Sec. 2"},{"comment":"The reference entries for Harikane et al. (2023) and Larson et al. (2023) appear malformed with titles and bibcodes embedded in an unusual way; they should be formatted consistently with the journal style.","section":"References"},{"comment":"In the F430M bin (z ≈ 5.53) the fit returns N = 34 and α = −0.46 ± 0.46, which is essentially unconstrained; this should be explicitly flagged in the table or the fit should be omitted from the comparison.","section":"Table A.1"},{"comment":"The abstract states a redshift range of 1.2 < z < 6.6, while Table 1 lists bins starting at 1.25 < z < 1.75 and Sec. 3.1 gives 1.35 < z < 6.57; please harmonize these numbers.","section":"Sec. 1 / Table 1"}],"recommendation":"major_revision","confidential_remarks":"The observed luminosity function work is solid and valuable, and the paper is appropriate for A&A. My main concern is the gap between the strong abstract claim about the SFRD and the paper's own admission that the dust prescription shifts the result by the size of the claimed effect. The revision should reframe the headline around the observed LF and present the dust-corrected SFRD as a tentative, systematically uncertain result. I do not see a need to reject the paper—the methodology and data are worth publishing—but the central claim needs to be made robust or explicitly deflated."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The real product of this paper is the observed Hα luminosity function, and it is good. Combining CANUCS, Technicolor, and JUMPS into one self-consistent framework across z=1.25-6.6 is genuinely new, and the validation is serious: photometric redshifts against spectroscopy (1.7% outliers), line fluxes against NIRSpec prisms (0.04 dex offset, 0.23 dex scatter), and completeness from full image simulations. The faint end is probed more than an order of magnitude deeper than ground-based work, and the comparison with grism surveys at z>4 is sensible. That part should be published, and I would cite it.\n\nThe soft spot is exactly where the stress-test note lands. The headline claim—dust-corrected SFRD 0.2-0.3 dex above UV/IR estimates, “eclipsing both”—rests on a single dust prescription applied to every galaxy: Eq. (2) with A_Hα = 0.44 A_V, Calzetti attenuation, and a stellar A_V from DENSEBASIS, with no rest-frame MIR/FIR input. The authors themselves say in Sec. 5.1 that changing the attenuation law or the 0.44 ratio shifts the SFRD by ~0.2 dex, which is the same size as the reported excess. So the observed LF is robust, but the normalization of the dust-corrected SFRD is not yet a measurement at that precision. The abstract oversells it; the discussion is more honest.\n\nTwo smaller issues. First, cosmic variance: the paper shows individual fields contribute >40% of the SFRD in several bins, and overdensities in MACS0417 and A370 drive the bright end at z~2 and z~6. That is handled with jackknife errors, but the central values remain vulnerable. Second, the SFR function fits use priors derived from the same survey's LF fits before integrating to SFRD. That is mild circularity, worth flagging but not disqualifying.\n\nVerdict: send to peer review. The observed LF is a solid, reproducible contribution, and the SFRD tension is worth airing even if it turns out to be systematics. Ask the authors to either add a systematic error bar on the dust correction and quote the SFRD as tentative, or pull the “eclipsing both” language from the abstract. The paper is honest on the inside; the packaging should catch up.","headline":"The observed Hα luminosity function is carefully built and worth having; the dust-corrected SFRD excess is a one-parameter dust prescription away from disappearing.","tokens_in":21357,"tokens_out":1536,"would_cite":true,"duration_ms":16920,"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 single JWST medium-band dataset now measures the Hα luminosity function from z = 1.25 to 6.6 and finds the dust-corrected cosmic star formation rate density runs 0.2–0.3 dex above UV-based estimates at cosmic noon.","keywords":["Hα luminosity function","star formation rate density","JWST NIRCam medium-band photometry","emission-line galaxies","cosmic noon","epoch of reionization","dust attenuation","photometric redshifts"],"falsifier":"Measure star formation for the same galaxies in the $z = 1.25$–$2.25$ bins from rest-frame mid- or far-infrared luminosity (MIRI photometry or ALMA continuum): if the resulting star formation rate density tracks the UV-based estimates instead of sitting 0.2–0.3 dex higher, then the $A_{\\mathrm{H}\\alpha} = 0.44\\,A_V$ rescaling over-corrects the sample. A second test is to rerun the $z \\approx 6$ bin on a wider-area survey to see whether the excess over grism measurements survives with the Abell 370 overdensity the paper identifies removed.","tokens_in":20062,"feed_emoji":"🔭","tokens_out":14189,"duration_ms":115375,"temperature":0.7,"pith_summary":"The paper aims to establish the first self-consistent measurement of the $\\mathrm{H}\\alpha$ luminosity function from cosmic noon ($z \\approx 1.25$) into the epoch of reionization ($z = 6.6$), and from it the dust-corrected cosmic star formation rate density across that range. The claim matters because almost all previous $\\mathrm{H}\\alpha$ work stopped at $z \\approx 3$, leaving high-redshift star formation histories to be built from ultraviolet (UV) or infrared tracers that disagree with each other. Tracing $\\mathrm{H}\\alpha$ through eleven NIRCam medium-band filters in three deep surveys (4101 galaxies), the authors recover the expected peak of star formation at $z \\approx 2$ and a decline toward $z = 6$, but with a normalization 0.2–0.3 dex above UV-based estimates and closer to infrared-based ones. If correct, the universe formed more stars at cosmic noon than UV surveys inferred, sharpening a known tension with the measured stellar mass density.","feed_headline":"JWST finds more stars born at cosmic noon than UV surveys saw","feed_subtitle":"Dust-corrected Hα counts from z=1.25 to 6.6 put star formation 0.2–0.3 dex above UV estimates.","key_machinery":"The engine is emission-line boosting of medium-band photometry: for each galaxy a power law is fit in $f_\\lambda$ to four continuum filters bracketing the filter that contains $\\mathrm{H}\\alpha$, and the flux density excess in that filter, multiplied by the filter width, gives the $\\mathrm{H}\\alpha + [\\mathrm{NII}]$ line flux. NIRSpec prism spectroscopy validates the fluxes with no systematic offset and 0.23 dex scatter. The luminosity function is built with the $1/V_\\mathrm{max}$ estimator (Schmidt 1968), corrected for completeness from injected-image simulations, and fit to Schechter functions by unbinned maximum likelihood, giving faint-end slopes $\\alpha \\approx -1.2$ to $-1.6$. The dust-corrected ladder uses $L_{\\mathrm{H}\\alpha,\\mathrm{int}} = L_{\\mathrm{H}\\alpha,\\mathrm{obs}} \\times 10^{0.4 A_{\\mathrm{H}\\alpha}}$ with $A_{\\mathrm{H}\\alpha} = 0.44\\,A_V$ on the Calzetti et al. (2000) curve, $A_V$ from DENSEBASIS SED fits, and the Kennicutt & Evans (2012) conversion $\\log \\mathrm{SFR} = \\log L_{\\mathrm{H}\\alpha} - 41.27$.","core_discovery":"The central claim is that one dataset and one analysis chain can measure the $\\mathrm{H}\\alpha$ luminosity function from $1.25 < z < 6.6$ — from the peak of cosmic star formation into reionization — and that the dust-corrected star formation rate density that follows is higher than previously reported. The observed luminosity functions agree with earlier ground-based, Spitzer, and JWST grism measurements where they overlap, the largest discrepancy being a 0.2–0.5 dex lower normalization at $z \\approx 2.5$ relative to one ground-based survey. After correcting each galaxy for dust with the stellar $A_V$ from SED fitting scaled as $A_{\\mathrm{H}\\alpha} = 0.44\\,A_V$ on the Calzetti et al. (2000) curve, the star formation rate density peaks at $z \\approx 2$, declines toward $z = 6$, and at $z = 1$–$2$ sits 0.2–0.3 dex above UV-based estimates, closer to infrared measurements. The authors call the elevated normalization tentative but statistically robust, identifying the dust attenuation law as the main systematic.","pith_inferences":["A test the paper does not run: use Balmer decrements ($\\mathrm{H}\\alpha/\\mathrm{H}\\beta$) from the existing NIRSpec spectra to check the fixed $0.44$ nebular-to-stellar attenuation ratio per galaxy; a systematic offset would explain the elevated star formation rate density without any new data.","The same medium-band excess machinery can measure the $[\\mathrm{OIII}] + \\mathrm{H}\\beta$ luminosity function in the bluer filters, giving an internal, self-consistent cross-check on the dust correction and on the [NII] subtraction.","Stacking ALMA or MIRI continuum on the $z > 4$ sample would test the paper's unexpected plateau in the obscured fraction, which disagrees with the decline expected from earlier work.","The per-galaxy dust correction mechanically flattens the faint-end slope; re-deriving the luminosity function with a single constant attenuation shift, as earlier surveys did, would separate the dust prescription from a genuine change in the luminosity function shape."],"forward_implications":["The $\\mathrm{H}\\alpha$ luminosity function is now measured in one self-consistent framework from $1.25 < z < 6.6$, with the faint end reaching $\\log L_{\\mathrm{H}\\alpha} \\approx 40.7$ (solar units) at $z \\approx 1.5$ and the shape spanned over almost three orders of magnitude in luminosity.","The dust-corrected star formation rate density peaks at $z \\approx 2$ and declines toward $z = 6$, but its normalization at $z \\approx 1$–$2$ is 0.2–0.3 dex above UV-based estimates and in line with recent infrared measurements.","Comparing observed and dust-corrected star formation rate densities implies an obscured fraction of about 80% at $z = 2$, declining to about 50% at $z = 4$ and staying flat out to $z = 6$.","The faint-end slope stays roughly constant, $\\alpha \\approx -1.2$ to $-1.6$, with no turnover down to $L_{\\mathrm{H}\\alpha} \\approx 10^{40.5}$ erg s$^{-1}$ at $z < 2$.","If the elevated normalization holds, it widens the gap between integrated star formation and the measured stellar mass density, and the paper says this may require a revised initial mass function, attenuation law, or star formation calibration."],"supporting_citations":[{"why":"Supplies the CANUCS/Technicolor mosaics, photometric catalogs, depths, and completeness that form the base dataset for every measurement in the paper.","marker":"Sarrouh & Asada et al. (2026)"},{"why":"Defines the CANUCS survey whose NIRCam flanking-field imaging provides the wide- and medium-band filters used to detect the sample.","marker":"Willott et al. (2022)"},{"why":"EAzY, the SED-fitting code that generates the photometric redshifts whose 1.7% outlier fraction makes assigning Hα to the correct medium-band filter reliable.","marker":"Brammer et al. (2008)"},{"why":"The attenuation curve assumed in Eq. (2); the dust law that, with the 0.44 ratio, sets the normalization of the dust-corrected luminosity function and star formation rate density.","marker":"Calzetti et al. (2000)"},{"why":"DENSEBASIS, the SED code that provides the stellar A_V and masses used for per-galaxy dust corrections and for the [NII] metallicity correction.","marker":"Iyer et al. (2019)"},{"why":"The 1/V_max estimator (Eq. 1) used to build the binned luminosity functions from the flux-limited sample.","marker":"Schmidt (1968)"},{"why":"Together with Hao et al. (2011) and Murphy et al. (2011), supplies the Hα-to-SFR conversion (Eq. 4) that turns dust-corrected luminosities into the star formation rate density.","marker":"Kennicutt & Evans (2012)"},{"why":"The ground-based Hα luminosity function at z ≈ 1–2 that anchors the low-redshift comparison and the reference peak behavior.","marker":"Sobral et al. (2013)"},{"why":"The JWST NIRCam grism Hα luminosity function at 4 < z < 6 against which the medium-band results are checked at high redshift.","marker":"Covelo-Paz et al. (2025)"},{"why":"A second JWST NIRCam grism Hα luminosity function at 4 < z < 6 used for comparison, notably with steeper faint-end slopes.","marker":"Fu et al. (2025)"}],"fun_headline_variants":["JWST Hα survey: cosmic star formation higher than UV estimates","Dust-corrected Hα star formation peaks at z=2, beats UV counts","JWST medium-band Hα reveals more stars at cosmic noon than UV","New Hα luminosity function from JWST: peak at cosmic noon, higher than UV","JWST traces Hα star formation from cosmic noon to reionization"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire dust-corrected star formation rate density ladder rests on one scaling: every galaxy's nebular attenuation is taken to be 0.44 times its stellar V-band attenuation on the Calzetti curve, and the paper states that changing that ratio or curve shifts the result by about 0.2 dex — enough to erase the claimed difference from UV measurements.","fun_headline_variants_meta":{"raw":{"variants":["JWST Hα survey: cosmic star formation higher than UV estimates","Dust-corrected Hα star formation peaks at z=2, beats UV counts","JWST medium-band Hα reveals more stars at cosmic noon than UV","New Hα luminosity function from JWST: peak at cosmic noon, higher than UV","JWST traces Hα star formation from cosmic noon to reionization"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000306,"raw_usage":{"total_tokens":1875,"prompt_tokens":1191,"completion_tokens":684,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":807,"completion_tokens_details":{"reasoning_tokens":582}},"tokens_in":807,"tokens_out":684,"duration_ms":7071,"temperature":1.0,"reasoning_tokens":582,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T00:17:08.407846+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure star formation for the same galaxies in the $z = 1.25$–$2.25$ bins from rest-frame mid- or far-infrared luminosity (MIRI photometry or ALMA continuum): if the resulting star formation rate density tracks the UV-based estimates instead of sitting 0.2–0.3 dex higher, then the $A_{\\mathrm{H}\\alpha} = 0.44\\,A_V$ rescaling over-corrects the sample. A second test is to rerun the $z \\approx 6$ bin on a wider-area survey to see whether the excess over grism measurements survives with the Abell 370 overdensity the paper identifies removed.","supporting_citations":[],"review_version":1}