{"id":"a0c86dde-c08b-45d8-b102-7023a63fd1d4","arxiv_id":"2412.06894","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"At z=4-6, the dust-obscured star formation rate density is log rho = -1.52, implying that about 66% of cosmic star formation was obscured by dust, based on eight spectroscopically confirmed dusty galaxies.","lead":"Using JWST and ALMA observations along 25 quasar sightlines, this paper finds eight dusty star-forming galaxies at z=4-6 and uses them to measure how much cosmic star formation was hidden by dust at that early epoch. It concludes that about two-thirds of star formation at redshift roughly 5 was dust-obscured, a higher fraction than several earlier estimates suggested.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The L_IR scale set by one ALMA 1.2-mm point and an assumed dust temperature is the pivotal assumption; the paper's own Tdust=30-50 K bracket (Section 5.1) shifts rho_SFR,IR by ~0.12 dex, comparable to the quoted errors, so the absolute SFRD and the 66% obscured fraction remain conditional on the…","rationale":"Read in good faith, the paper's logic is coherent: spectroscopic redshifts are secured for the ALMA-selected DSFGs (six with two or more lines), ALMA completeness is tested with injection simulations, cosmic variance is treated explicitly, and the z~5.0-5.3 redshift gap is acknowledged. The weakest point is the absolute L_IR scale, because the rest-frame far-IR is sampled at one wavelength and CIGALE's energy-balance assumption is not directly testable with current data. However, the paper already quantifies the Tdust sensitivity in Section 5.1 and Figure 11, and within Tdust=30-50 K the SFRD stays within roughly 0.12 dex of the benchmark, so the qualitative conclusion that a large fraction of star formation is obscured is probably robust. This is not a fatal objection, and it does not change the reader's conditional verdict: the paper should be published with the far-IR SED caveat kept prominent and ideally tested with multi-band ALMA continuum data.","tokens_in":39879,"tokens_out":10464,"duration_ms":125214,"concrete_test":"Obtain ALMA Band 7 (0.87 mm) continuum observations of the eight ASPIRE DSFGs at depth comparable to the 1.2-mm signal-to-noise, and jointly fit each 1.2+0.87 mm SED with Tdust and beta free. Rebuild the IRLF and SFRD using the resulting L_IR values; if the sample median L_IR shifts by more than about 0.15 dex relative to the CIGALE benchmark in Section 3.2, the headline log rho_SFR,IR = -1.52 and the 66% obscured fraction should be revised accordingly, whereas a shift below 0.1 dex would validate the energy-balance assumption.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claims, log rho_SFR,IR = -1.52 and a 66% obscured fraction, inherit almost entirely from the L_IR of eight sources. Section 3.2 derives L_IR with CIGALE from four photometric points, of which only the ALMA 1.2-mm point constrains the rest-frame far-IR; because the dust SED shape is otherwise fixed by the prior Tdust=30-50 K, beta=1.8 model, L_IR is essentially a one-band bolometric correction. Section 5.1 tests Tdust=30, 40, and 50 K: the 50 K case raises the SFRD by 0.12 dex, and the benchmark is equivalent to about 33 K, so the explored range is comparable to the quoted +0.14/-0.13 dex uncertainty. The quoted errors therefore absorb part of the thermal prior, but not the full systematic if the true SED is colder than 30 K or if patchy dust breaks energy balance, as in HDF850.1; the paper itself notes J0244m5008.C03 is poorly fit in F115W. An L_IR shift of about 0.2-0.3 dex would move the obscured fraction close to 50%, directly altering the 'majority obscured' headline. The paper is transparent about the caveat, but the numerical values are not independently verified.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a measurement of the dust-obscured cosmic star formation rate density (SFRD) at z=4-6 using the ASPIRE JWST Cycle-1 and ALMA Cycle-9 surveys along 25 quasar sightlines. Eight dusty star-forming galaxies (DSFGs) are identified at z=4-6 through NIRCam WFSS detections of Ha or [O III], and their infrared luminosities are derived from CIGALE energy-balance SED fitting using three NIRCam bands plus a single ALMA 1.2 mm continuum point. From these sources the authors construct a 1/Vmax infrared luminosity function (IRLF), fit it with a double power law, integrate to obtain log rho_SFR,IR = -1.52+0.14/-0.13 (M_sun/yr/Mpc^3), and conclude that 66 +/- 7% of cosmic star formation at z~5 is obscured by dust. The paper also reports a flattened faint-end IRLF slope alpha = 0.59+0.39/-0.45 and discusses systematic effects from dust temperature, the bright-end IRLF, and AGN contamination.","tokens_in":40315,"tokens_out":12859,"duration_ms":139179,"significance":"If correct, this is a valuable direct measurement: it provides spectroscopic redshifts for DSFGs at z=4-6, a redshift range where photometric redshifts are notoriously unreliable, and uses 25 independent sightlines to suppress cosmic variance. The resulting SFRD is higher than ASPECS-based estimates and close to ALPINE/ALCS values, supporting a picture in which a majority of star formation at z~5 is obscured. The paper is transparent about its main assumptions, includes detailed completeness simulations, a lensing correction, and a systematic exploration of dust temperature, and it places the result in the context of existing literature. However, the headline claims of a 'spectroscopically complete census' and the 66% obscured fraction rest on two assumptions that are acknowledged but not fully quantified in the quoted uncertainties: the coverage gap in NIRCam WFSS at z~5.0-5.3, and the energy-balance SED modeling from a single far-IR photometric point. These issues, if unaddressed, weaken the precision of the central result even though the measurement itself is credible.","major_comments":[{"comment":"The title and abstract claim a 'spectroscopically complete' DSFG sample at z=4-6, but the text in Section 4.2 explicitly states that 'the full ASPIRE DSFG sample is not spectroscopically complete across all redshifts' because of the NIRCam F356W gap at z~5.0-5.3 where neither Ha nor [O III] falls in the grism bandpass. The correction for this 'redshift desert' is described only as an argument that the true number density 'should be close to our measurement, or just slightly higher' rather than a quantitative correction. The 1/Vmax method in Eq. (2) is not specified as to whether Vmax excludes the redshift gap; if Vmax includes the gap, the LF is underestimated, and if it excludes it, the SFRD integral over z=4-6 assumes the LF is identical in the gap. The authors should state precisely how Vmax is computed over the gap and provide a quantitative upper/lower bound on the IRLF and SFRD from this assumption. Without this, the word 'complete' in the title overstates the actual selection.","section":"Section 4.2, Section 4.3, Eq. (2)"},{"comment":"The central SFRD value and the 66% obscured fraction depend on L_IR values derived from CIGALE energy-balance fitting with only one ALMA 1.2 mm point constraining the rest-frame far-IR SED. Section 5.1 tests dust temperature variations and finds a 0.12 dex shift for Tdust=50 K, which is comparable to the quoted +0.14/-0.13 dex error, and it also tests beta and lambda_thick variations. However, this does not bound the energy-balance assumption itself, which the paper identifies as questionable: J0244m5008.C03 is poorly fit in F115W, and HDF850.1 is cited as a case where UV photon leakage breaks the energy-balance picture. The quoted error bars in Section 4.4 appear to be purely statistical/MCMC and do not include a systematic term for energy-balance violations. The authors should either estimate the magnitude of this systematic (e.g., by fitting L_IR directly from the ALMA flux with a range of dust temperatures and beta values and comparing SFRD, or by explicitly modeling a UV-leakage component for J0244m5008.C03) or clearly state that the headline rho_SFR,IR value is conditional on the energy-balance assumption. This is load-bearing because a ~0.2 dex shift in L_IR would move the obscured fraction toward 50% and change the paper's main qualitative conclusion.","section":"Section 3.2, Section 5.1, Section 4.4"},{"comment":"The SFRD integral is not determined purely by the ASPIRE sample: the bright end of the IRLF (L > L* ~ 10^12.6 L_sun) is constrained by literature measurements that are not spectroscopically complete, with their uncertainties artificially increased by sqrt(2) to account for photometric redshifts. The contribution from ULIRGs (10^12-10^13 L_sun) to the obscured SFRD is 44 +/- 14%, so the quoted SFRD accuracy depends partly on the assumed literature LF. The authors should state explicitly what fraction of the quoted uncertainty on log rho_SFR,IR comes from the literature-anchored bright end versus the ASPIRE data alone, and report an ASPIRE-only SFRD if possible. This would make the contribution of the new data to the headline result more transparent.","section":"Section 5.2, Table 3"}],"minor_comments":[{"comment":"The abstract's 'spectroscopically complete DSFG sample at z=4-6' should be qualified, e.g., 'complete for DSFGs with Ha or [O III] in the F356W bandpass' or 'complete over the redshift intervals z=4.0-5.0 and z=5.3-6.0', to match the statement in Section 4.2.","section":"Abstract and Section 4.2"},{"comment":"For J0244m5008.C03, the poorly fitted F115W point is noted in Section 5.1 but not flagged in Table 2 or Figure 5; adding a note or a different symbol for this source would help the reader see the energy-balance issue directly.","section":"Section 3.2, Table 2"},{"comment":"The sentence describing the correction for the redshift desert gives no formula or magnitude; a short description of the assumed correction (e.g., 'we divide by the fractional redshift coverage' or 'we add a 10% systematic') would clarify the method.","section":"Section 4.2"},{"comment":"The two single-line redshifts (conf=2) are well argued, but the caption of Figure 4 does not mention the confidence level; adding 'conf=2' in the caption would avoid confusion with the six secure redshifts.","section":"Section 3.1, Figure 4"},{"comment":"The authors quote the Tdust=50 K shift as 0.12 dex but do not quote the resulting SFRD for Tdust=30 K; reporting both bounds in Figure 11 or in the text would make the direction and magnitude of the SED systematic clearer.","section":"Section 5.1"}],"recommendation":"major_revision","confidential_remarks":"This is a strong observational paper with a transparent discussion of caveats. The main issues are the mismatch between the 'spectroscopically complete' claim and the actual redshift coverage, and the absence of a quantitative systematic error for the energy-balance SED assumption. Both are fixable in revision without new observations, by adding the relevant calculations and qualifying the headline claims. Given the small sample size (8 sources) and the reliance on one-band far-IR SEDs, I would support publication after these revisions, but the current version overstates the precision and completeness of the measurement."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is the first spectroscopically complete census of z=4-6 dusty star-forming galaxies, and the IRLF/SFRD measurement is a genuine step forward. The eight-source sample is small, but the completeness argument is the careful part: 25 independent sightlines kill cosmic variance, ALMA detections are >6.5 sigma, JWST grism gives secure redshifts for six sources and well-argued single-line redshifts for two, and the lensing correction for J0109m3047.C02 is handled properly. The flattened faint-end slope (alpha=0.59) and the obscured SFRD of log rho=-1.52 dex are new numbers that will be cited.\n\nThe soft spot is exactly where the stress-test puts it. L_IR comes from CIGALE energy-balance fitting with four photometric points, of which only the ALMA 1.2-mm point constrains the rest-frame far-IR. The paper is transparent about this (Section 3.2: \"we have to rely on the energy balance assumption\") and tests Tdust=30-50 K, finding only a 0.12 dex shift in SFRD. That is reassuring but not fully conclusive: the quoted errors (+0.14/-0.13) are comparable to that shift, and the benchmark SED is about 33 K, so colder-than-30 K SEDs or patchy dust that leaks UV could push the scale down. The paper itself notes that J0244m5008.C03 is poorly fit in F115W, which is a patchy-geometry warning sign. Still, I don't think this is a load-bearing flaw; it is a standard one-band bolometric correction with a stated systematic, and the authors explicitly bound the plausible warm-dust range.\n\nThe \"spectroscopically complete\" claim is slightly oversold because of the redshift desert at z=5.0-5.3 where H-alpha and [O III] fall outside the F356W grism coverage. The authors acknowledge this in Section 4.2 and argue the true number density is \"close to or slightly higher.\" That is fair, but the title says complete.\n\nThe citation pattern is clean, and the comparison to ALPINE/ASPECS/ALCS is well handled. The ASPECS discrepancy is plausibly cosmic variance, and the authors give a concrete calculation showing a 37% chance of zero detections in the HUDF. That is good reasoning.\n\nThis paper is for anyone working on high-redshift galaxy evolution, particularly the obscured star-formation history. It deserves serious refereeing: the measurement will be a reference point regardless of small-number statistics, and the caveats are stated rather than hidden. A referee should push for a quantitative statement of the Tdust-prior systematic in the SFRD error budget, but I would not block on it. I would bring it to reading group and would cite it.","headline":"First spectroscopically complete IR luminosity function at z=4-6: small sample, careful completeness, real result, but the absolute SFRD scale still leans on one ALMA band and a dust-temperature prior.","tokens_in":40925,"tokens_out":2068,"would_cite":true,"duration_ms":23038,"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":"Using 25 independent quasar sightlines, the ASPIRE program measures the dust-obscured star-formation rate density at $z=4$–$6$ and finds that $66\\pm7\\%$ of cosmic star formation at $z\\sim5$ is hidden by dust.","keywords":["dusty star-forming galaxies","infrared luminosity function","obscured star formation rate density","JWST NIRCam grism spectroscopy","ALMA 1.2 mm continuum","z=4-6","cosmic star formation history","cosmic variance"],"falsifier":"Observe the eight ASPIRE DSFGs in a second ALMA band on the Rayleigh-Jeans tail (for example at 2 mm or 870 $\\mu$m) to measure their dust temperatures directly, then recompute $L_{\\rm IR}$ from the measured SEDs; if the resulting obscured star-formation rate density at $z=4$–$6$ differs from $\\log\\rho_{\\rm SFR,IR} = -1.52$ by more than the quoted uncertainty, the energy-balance assumption is the reason.","tokens_in":39696,"feed_emoji":"🌌","tokens_out":7092,"duration_ms":69910,"temperature":0.7,"pith_summary":"This paper tries to settle how much of the star formation happening when the Universe was roughly one billion years old is hidden inside dusty galaxies. Using a survey of 25 independent quasar sightlines with JWST near-infrared grism spectroscopy and ALMA 1.2 mm imaging, the authors identify eight dusty star-forming galaxies at spectroscopic redshifts $z=4$–$6$ and build an infrared luminosity function that is complete down to $L_{\\rm IR}\\sim 2\\times10^{11}\\,L_\\odot$. From that function they derive the dust-obscured star-formation rate density at $z\\sim5$ as $\\log[\\rho_{\\rm SFR,IR}/(M_\\odot\\,{\\rm yr}^{-1}\\,{\\rm Mpc}^{-3})] = -1.52$ and conclude that $66\\pm7\\%$ of all star formation at that epoch is obscured by dust. If correct, earlier deep-field censuses of early star formation missed most of the action because of cosmic variance.","feed_headline":"Two-thirds of star formation at z=5 is dust-obscured","feed_subtitle":"A JWST+ALMA spectroscopic census revises the obscured star-formation rate density at z~5 upward by ~5x.","key_machinery":"The machinery is the pairing of two survey instruments over the same footprints: ALMA 1.2 mm continuum mosaics find the dust emission, and JWST/NIRCam slitless grism spectra in F356W (3.1–4.0 $\\mu$m) catch H$\\alpha$ or [O III] emission lines that pin each source's redshift spectroscopically. Eight such sources at $z=4$–$6$, six of them confirmed by at least a second line, form a sample whose selection is claimed to be spectroscopically complete down to $L_{\\rm IR}\\sim 2\\times10^{11}\\,L_\\odot$. The analysis then applies a $1/V_{\\rm max}$ estimator to build the infrared luminosity function, fits it with a double-power-law model with a free faint-end slope, and integrates the fitted function to obtain the obscured star-formation rate density.","core_discovery":"On its own terms, the paper establishes that the infrared luminosity function at $z=4$–$6$ is flatter at the faint end than previously modeled (power-law slope $\\alpha = 0.59^{+0.39}_{-0.45}$), so luminous and ultra-luminous infrared galaxies ($L_{\\rm IR}\\sim 10^{11}$–$10^{13}\\,L_\\odot$) carry most of the obscured star formation. Integrating this function down to $L_{\\rm IR}=10^{10}\\,L_\\odot$ gives an obscured star-formation rate density about five times the value inferred from the ASPECS/HUDF-based backward modeling and consistent with the ALPINE-based and lensing-cluster measurements. The authors attribute the earlier low values to cosmic variance: the Hubble Ultra Deep Field happens to sit in a void of $z>4$ dusty galaxies, while the 25 independent ASPIRE sightlines average out such fluctuations. They conclude that at $z\\sim5$ the majority of cosmic star formation, $66\\pm7\\%$, is dust-obscured, so the total star-formation rate density at this epoch is close to or above the canonical total-SFRD curve.","pith_inferences":["Editorial inference: the survey targets quasar sightlines, so the eight DSFGs could trace overdense environments around already rare quasars; the authors estimate cosmic variance is small, but a blank-field analog survey would be the clean test of whether this averaged obscured star-formation rate density applies to typical fields.","Editorial inference: the F356W grism has a redshift desert near $z\\approx5.0$–$5.3$ where neither H$\\alpha$ nor [O III] falls in the band; the paper corrects for it statistically, but redder F444W grism data covering that gap would directly test whether a population of $z\\approx5.2$ dusty galaxies is being missed.","Editorial inference: because $L_{\\rm IR}$ is derived from an energy-balance fit with a single far-infrared photometric point, the quoted star-formation rate density scales roughly as $T_{\\rm dust}^4$; if future high-frequency ALMA observations show these galaxies are warmer than the assumed $\\sim33$ K, the obscured star-formation rate density would rise by roughly 0.1–0.2 dex, strengthening the pa"],"forward_implications":["The obscured star-formation rate density at $z\\sim5$ is about 0.3 dex higher than the canonical total-SFRD curve at that redshift, meaning the broadly used Madau–Dickinson curve may undercount early star formation.","Previous deep single-field millimeter surveys (notably ASPECS in the Hubble Ultra Deep Field) can miss the $z=4$–$6$ dusty population entirely because of cosmic variance; multi-field surveys are necessary at this epoch.","Luminous and ultra-luminous infrared galaxies with $L_{\\rm IR}$ between $10^{11}$ and $10^{13}\\,L_\\odot$ contribute about $81\\%$ of the obscured star-formation rate density at $z=4$–$6$, so future surveys must cover enough volume to sample the bright end of the luminosity function.","Because the faint-end slope of the infrared luminosity function is flat ($\\alpha\\approx0.6$), the integrated obscured star-formation rate density does not depend strongly on the exact low-luminosity cutoff, making the measurement robust to survey depth.","Future JWST NIRCam grism surveys over much larger volumes can spectroscopically constrain the bright end of the $z=4$–$6$ infrared luminosity function, resolving the main remaining uncertainty in the obscured star-formation history."],"supporting_citations":[{"why":"Supplies the ASPECS survey result that found no $z>4$ DSFG in the Hubble Ultra Deep Field, the key comparison the paper attributes to cosmic variance.","marker":"Aravena et al. 2020"},{"why":"Provides the ALPINE-based infrared luminosity function and obscured star-formation rate density that ASPIRE's number count is found to match.","marker":"Gruppioni et al. 2020"},{"why":"Provides the backward-modeled star-formation rate density from millimeter number counts that ASPIRE's measurement exceeds by roughly five times.","marker":"Zavala et al. 2021"},{"why":"Supplies the ALCS lensing-cluster infrared luminosity function and obscured star-formation rate density consistent with ASPIRE, plus the all-redshift 1.2 mm number count used for comparison.","marker":"Fujimoto et al. 2023"},{"why":"Provides H$\\alpha$ detections of $z>5$ dusty galaxies including HDF850.1 and the FRESCO survey volume used to sanity-check the bright-end luminosity function and spectroscopic completeness.","marker":"Sun et al. 2024"},{"why":"Defines the canonical total cosmic star-formation rate density curve used to derive the $66\\%$ obscured fraction at $z\\sim5$.","marker":"Madau & Dickinson 2014"}],"fun_headline_variants":["Dust veils 66% of star formation at z=5","Star formation at z=5: 66% hidden by dust","Obscured star formation dominates at z~5: 2/3 of SFRD","Cosmic variance overturned: obscured SFRD 5x higher at z=4-6","JWST+ALMA census: dust hides most star birth at z=5"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything rests on the infrared luminosities, and those come from an energy-balance model that fits the UV-to-millimeter photometry with only one far-infrared data point; if the dust is patchy or the dust temperature falls outside the assumed 30–50 K range, the luminosity scale and the reported obscured star-formation rate density move with it.","fun_headline_variants_meta":{"raw":{"variants":["Dust veils 66% of star formation at z=5","Star formation at z=5: 66% hidden by dust","Obscured star formation dominates at z~5: 2/3 of SFRD","Cosmic variance overturned: obscured SFRD 5x higher at z=4-6","JWST+ALMA census: dust hides most star birth at z=5"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000631,"raw_usage":{"total_tokens":3074,"prompt_tokens":1268,"completion_tokens":1806,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":884,"completion_tokens_details":{"reasoning_tokens":1699}},"tokens_in":884,"tokens_out":1806,"duration_ms":14771,"temperature":1.0,"reasoning_tokens":1699,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T19:18:35.250550+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Observe the eight ASPIRE DSFGs in a second ALMA band on the Rayleigh-Jeans tail (for example at 2 mm or 870 $\\mu$m) to measure their dust temperatures directly, then recompute $L_{\\rm IR}$ from the measured SEDs; if the resulting obscured star-formation rate density at $z=4$–$6$ differs from $\\log\\rho_{\\rm SFR,IR} = -1.52$ by more than the quoted uncertainty, the energy-balance assumption is the reason.","supporting_citations":[],"review_version":1}