{"id":"70398f3f-231c-415e-8615-9a86ac5f6f0e","arxiv_id":"2506.24128","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"More than 85% of z>5.9 quasars in massive host galaxies are UV-faint, indicating that UV-bright quasars are extreme outliers in black hole mass.","lead":"Using 190 quasars with ALMA [CII] observations, the authors find that more than 85% of quasars in massive host galaxies at z>5.9 are UV-faint rather than UV-bright. The result implies that the ultrabright quasars that dominate studies of early supermassive black holes are rare high-mass outliers, not representative systems.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The '>85% UV-faint' headline is an LF-weighted quantity, and the paper's own Matsuoka et al. (2018c) LF alternative implies a faint fraction near 83%, below the headline; the claim needs LF-robustness error bars.","rationale":"Good faith: the paper offers a valuable new census — 46 UV-faint QSOs with ALMA [CII] from CISTERN is a major increase, and the cross-checks using L_IR and line FWHM support the interpretation that the 13 [CII]-bright UV-faint QSOs resemble UV-bright hosts. The conditional verdict from the reader is reasonable; my stress-test does not overturn it. However, I find the selection-bias assumption (the reader's weakest assumption) less load-bearing than the LF weighting: because the faint end of the UV LF dominates the counts, even a factor of several overestimate in the 21% massive-host fraction would leave the UV-faint fraction above 85%. The more fragile step is the convolution with the assumed LF: the paper's own alternative LF (Matsuoka) reduces the faint/bright excess from 29 to 24, which translates to ~83% UV-faint, below the headline. The paper should report the cumulative fraction under each LF and propagate LF uncertainties into the 85%/3% numbers. The MBH extrapolation in §4 is explicitly hedged and does not affect the central census claim. CISTERN data release (Bouwens et al. 2025, in prep) should also allow direct tests of the UV-faint [CII] detection fraction against parent samples.","tokens_in":25288,"tokens_out":9836,"duration_ms":107329,"concrete_test":"Recompute the cumulative fraction in Fig. 3 by multiplying the §3.2 massive-host fractions (0.82, 0.48, 0.38, 0.28, 0.21 for the five UV bins) with the Matsuoka et al. (2018c) and Willott et al. (2010c) z~6 QSO LFs. If the resulting UV-faint (M>-24.5) fraction drops below 85% for Matsuoka, revise the headline to '~80-90%' with LF-driven uncertainty and report the cumulative fraction for each LF.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim is not the directly observed fraction of massive-host QSOs that are UV-faint (13/63 ≈ 21% in the sample), but the volume-density-weighted fraction obtained by multiplying the z~6 QSO UV LF (Schindler et al. 2023) by the observed massive-host fraction in each UV bin. The paper reports that the alternative Matsuoka et al. (2018c) LF reduces the excess of UV-faint over UV-bright [CII]-luminous QSOs from 29× to 24×. Reproducing the implication: with the intermediate/bright ratio of ~5 evident from Fig. 3 (15% cumulatively brightward of −24.5, 3% brightward of −26), a 24× excess gives a UV-faint fraction of 24/(24+5) ≈ 83%, below the '>85%' headline. The paper quotes only the excess ratio for the alternative LFs, not the resulting cumulative fraction, and gives no error bars on the 85%/3% numbers. Thus the headline is a model-dependent extrapolation and would fail for the lower envelope of plausible LF choices.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper uses new ALMA/CISTERN [CII] 158 μm observations together with archival measurements to assemble a sample of 190 z>5.9 QSOs spanning M_UV from -22 to -28. Defining a 'massive host galaxy' by L[CII] > 1.8e9 Lsun (the median [CII] luminosity of the UV-bright subsample), the authors identify 61 massive-host QSOs, including 13 that are UV-faint (M_UV,AB > -24.5). Combining the observed fraction of [CII]-luminous QSOs in each UV-luminosity bin with the Schindler et al. (2023) z~6 QSO UV luminosity function, they derive cumulative UV luminosity distributions for QSOs in massive hosts, reporting that only ~15% are brighter than M_UV=-24.5, ~3% are brighter than M_UV=-26, and hence that >85% are UV-faint. They also derive [CII] luminosity functions for UV-bright and UV-faint QSOs, finding a ~29x excess of UV-faint over UV-bright volume density at fixed host mass, and they extrapolate an M_BH-M_UV relation from 34 [CII]-luminous QSOs to estimate a median log M_BH ~8.1 for UV-faint QSOs in massive hosts at z~6.","tokens_in":25535,"tokens_out":8520,"duration_ms":90688,"significance":"The CISTERN program provides a major increase in [CII] coverage of UV-faint QSOs (5x and 6x in the two faint bins), and this is the first quantitative attempt to characterize the UV luminosity distribution of QSOs in massive host galaxies at z>5.9. The qualitative conclusion—that UV-bright QSOs are rare outliers among QSOs in massive hosts—is important and will influence discussions of SMBH growth at early epochs. The [CII] luminosity functions in Table 3 and Figure 4 are new and potentially useful. The main weaknesses are that the headline fractions are volume-density-weighted estimates whose uncertainties are not propagated, and the BH-mass conclusions rest on a large extrapolation with very few direct anchors. With appropriate uncertainty treatment and reframing, the results are publishable in A&A.","major_comments":[{"comment":"The headline fractions (15%, 3%, >85%) are quoted without uncertainties, yet they are not directly measured but are obtained by multiplying the Schindler et al. (2023) UV LF by the observed [CII]-luminous fractions in five UV bins. The paper reports alternative LF determinations only as excess ratios in §3.3 (24x for Matsuoka et al. 2018c and 33x for Willott et al. 2010c), not as the corresponding cumulative UV-faint fractions. Using the Matsuoka et al. (2018c) excess of 24 together with the intermediate-to-bright ratio of ~4 read from Fig. 3 (12% intermediate vs 3% bright) gives a UV-faint fraction of ~83%, below the '>85%' headline. Please propagate the LF uncertainties, report the cumulative fractions for each adopted LF, and state the headline as a range or with a proper error bar.","section":"§3.3, Fig. 3, Table 3"},{"comment":"The representativeness assumption stated in §3.3—'targeting of specific z>6 QSOs was largely a function of the apparent brightness of QSOs and the spectroscopic redshift being known'—is load-bearing for the headline result, but no quantitative selection-bias test is provided. The key input is the observed massive-host fraction of 13/61 ≈ 21% for UV-faint QSOs; if CISTERN's UV-faint targets are biased toward sources with known spectroscopic redshifts or favorable [CII] detectability, this fraction, and hence the >85% result, would be overestimated. A comparison of the CISTERN targets with the parent SHELLQs/wide-area selections, or a sensitivity test that varies the assumed FWHM for non-detections, is needed to bound this effect.","section":"§3.3, Fig. 1"},{"comment":"The 'massive host' threshold L[CII] > 1.8e9 Lsun is defined as the median [CII] luminosity of the UV-bright subsample, and the validation that UV-faint [CII]-bright QSOs live in similar hosts rests on Mdyn and LIR for the [CII]-detected sources only. This does not validate the L[CII]-Mgas zero point or scatter at z~6, and the paper itself cites Kaasinen et al. (2024) as a caution in §1. Since the headline fraction is a ratio of counts above this threshold, the result is sensitive to the threshold choice; a sensitivity test varying the threshold by ±0.3 dex would show how much of the '>85%' conclusion is calibration-dependent.","section":"§3.2, Fig. 2"},{"comment":"The conclusion that M_BH is the dominant driver of UV luminosity rests on Eq. (2), fitted to 34 [CII]-luminous QSOs that are almost all at M_UV < -25, with a single source (J1243+0100) at M_UV = -24.13; the claimed median log M_BH ~ 8.1 for M_UV ~ -23 is an extrapolation over roughly three magnitudes. The stated ±0.4 dex uncertainty reflects the extrapolation and the assumed M_UV = -22 cutoff, not direct measurements. This section should be reframed as a model-dependent extrapolation, and the '15x more massive' claim in the abstract should carry the same explicit caveat.","section":"§4.2, Eq. (2), Fig. 5"}],"minor_comments":[{"comment":"The caption contains the duplicated phrase 'duty cycle cycle'; the sentence beginning 'As such' is also grammatically incomplete.","section":"Fig. 6 caption"},{"comment":"The table header appears as 'RightM UV' and the FWHM column does not explicitly state its units; please clean up the table formatting.","section":"Table A.1"},{"comment":"When the alternative LFs are discussed, the text quotes only the excess ratios; please also give the resulting cumulative UV-faint fractions or explicitly state that they are not computed.","section":"§3.3"},{"comment":"The abstract says 'recent QSO luminosity functions (LFs)' without naming Schindler et al. (2023); adding the citation at first use would help the reader locate the primary input.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The scientific core of the paper is valuable and the new CISTERN data represent a real step forward. My main concern for the editor is that the manuscript depends on the unpublished CISTERN survey paper (Bouwens et al. 2025, in prep) for the fundamental measurements; the referee cannot verify source-level fluxes, non-detections, or upper limits. The authors should be asked to include sufficient data in an appendix or to coordinate publication with the CISTERN data paper. In addition, the '>85%' headline is likely to be quoted widely, so it would be prudent to require a conservative phrasing such as 'approximately 80-85% depending on the adopted UV LF' until the LF sensitivity is propagated into the quoted numbers."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. The CISTERN data are genuinely new: they roughly quintuple [CII] coverage of UV-faint z>5.9 QSOs, and the paper uses them to build the first UV-luminosity census of QSOs in massive host galaxies across six magnitudes. The second thing is that the headline \">85% UV-faint\" is a luminosity-function-weighted number, not a raw fraction. It drops below 85% if you adopt the Matsuoka et al. (2018c) LF, as the stress-test note shows.\n\nThe paper does several things well. The sample is large (190 QSOs) and spans a wide UV range. Using [CII] luminosity as a host-mass proxy is reasonable, and the authors validate it against dynamical masses and IR luminosities, which are at least partially independent. The resulting [CII] luminosity functions for UV-bright and UV-faint QSOs are a useful new product. The MBH extrapolation is long, but it is honestly labeled as preliminary and the uncertainties are stated.\n\nThe main soft spot is the missing error bars on the headline fractions. The paper quotes 3% and >85% without uncertainties, and for the alternative LFs it reports only the excess ratio, not the cumulative fraction. Reproducing the stress-test arithmetic: with the Matsuoka LF the excess is 24x rather than 29x, and that translates to a faint fraction near 83%—below the advertised >85%. This is a moderate issue, not a fatal one: every plausible LF gives a strong majority UV-faint, so the qualitative result holds.\n\nTwo other concerns are worth noting. The massive-host threshold is set to the median [CII] luminosity of the UV-bright sample, which gives the definition a whiff of circularity; however, the consistency of M_dyn and L_IR across the bins makes it a defensible, well-posed question. Second, the representativeness of CISTERN is assumed rather than demonstrated—targeting was driven by apparent brightness and known redshift—and the data are not yet public, so the 21% [CII]-luminous fraction for faint QSOs cannot be independently checked. That is a legitimate but testable concern.\n\nWho should read this? Anyone working on z>6 SMBH demographics, QSO host galaxies, or ALMA follow-up of faint quasars. It deserves a serious referee. A referee should ask for error bars on the cumulative fractions, a direct conversion of each alternative LF into cumulative fractions, and robustness tests on the CISTERN selection. With those additions, this becomes a strong paper. My recommendation: send it to review, and require the robustness analysis before acceptance.","headline":"New [CII] data make a plausible case that most z>5.9 QSOs in massive hosts are UV-faint, but the headline fraction is LF-dependent and needs error bars before it is quoted as >85%.","tokens_in":26248,"tokens_out":10503,"would_cite":true,"duration_ms":92053,"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":"More than 85% of massive-host quasars at z>5.9 are UV-faint.","keywords":["high-redshift quasars","z>6","[CII] 158 micron emission","ALMA","host galaxy mass","quasar UV luminosity function","supermassive black holes","QSO demographics"],"falsifier":"Take a UV-complete, spectroscopically confirmed sample of z>5.9 quasars with M_UV,AB between -24.5 and -22, observe all of them with ALMA to a uniform [CII] depth, and measure the fraction with L[CII]>1.8e9 Lsun. If that fraction comes out substantially below 21%, the inferred >85% fraction fails; a particularly clean test is the same measurement on the faintest bin (M_UV,AB>-23.5), where the paper sees 7 massive-host quasars and the assumption about representativeness is most strained.","tokens_in":25069,"feed_emoji":"🔭","tokens_out":6443,"duration_ms":63783,"temperature":0.7,"pith_summary":"This paper tries to establish that the quasars we have studied at the edge of cosmic reionization are not representative of what most supermassive black holes were doing inside massive galaxies. Using the luminosity of the [CII] 158 micron line as a stand-in for host galaxy mass, the authors sort 190 quasars at z>5.9 and find that more than 85% of those in the most massive hosts are UV-faint, with only about 3% as UV-bright as the classic early-universe quasars. Earlier surveys missed this population because they selected quasars by rest-UV brightness; a new ALMA program that targeted faint quasars in [CII] is what makes the census possible. If true, the result reframes early supermassive black hole formation: the bright exemplars are outliers, and the typical active black hole in a massive early galaxy is dim in the UV but embedded in a gas-rich, massive host.","feed_headline":"Over 85% of massive-host quasars at cosmic dawn are UV-faint","feed_subtitle":"New ALMA [CII] census finds faint quasars outnumber bright ones ~29-to-1 in the same host masses.","key_machinery":"The load-bearing instrument is the [CII] 158 micron line, used as a proxy for molecular gas content and hence host galaxy mass, with the threshold L[CII]>1.8e9 Lsun (roughly $10^{10}$.5 Msun of gas) defining a massive host. The analysis converts the observed L[CII] vs M_UV distribution for 190 quasars into volume densities by multiplying published z~6 quasar UV luminosity functions (Schindler et al. 2023) by the measured fraction of [CII]-luminous quasars in each UV-luminosity bin. This two-step weighting is what turns a heterogeneously targeted sample into statements about the relative abundance of UV-bright and UV-faint quasars in massive hosts.","core_discovery":"The central claim is that at z>5.9, quasar UV luminosity is a poor tracer of the underlying host galaxy population. Defining massive hosts by L[CII]>1.8e9 Lsun (the median [CII] luminosity of UV-bright quasars), the paper finds 61 such systems among 190 [CII]-observed quasars, including 13 UV-faint and 7 especially UV-faint ones. From these numbers and published z~6 quasar luminosity functions, the paper infers that only ~15% of massive-host quasars are brighter than M_UV,AB=-24.5 and only ~3% are brighter than -26, so >85% are UV-faint; the volume density of UV-faint quasars at a given host mass is ~29x that of UV-bright ones. The same [CII]-luminous systems show dynamical masses and IR luminosities similar to UV-bright quasars, supporting the claim that the hosts really are comparable. The paper further argues, by extrapolating MBH and Eddington-ratio trends measured for 34 [CII]-luminous quasars, that black hole mass rather than accretion rate is the main driver of the UV spread.","pith_inferences":["Editorial inference: a UV-faint bias could also apply across cosmic time; if so, samples selected on rest-UV flux at any redshift will undercount the active black holes in the most massive galaxies, and the local MBH-host relation may be less evolved than bright-QSO samples suggest.","The paper's mass threshold is defined by the median [CII] luminosity of UV-bright quasars; shifting the threshold to a dynamical-mass or far-infrared-based selection would change the exact percentages, though the authors show dynamical masses are consistent across UV luminosity.","A testable prediction follows from the claim that BH mass drives UV luminosity: JWST spectroscopy of the CISTERN faint quasars should show MBH decreasing by roughly a factor of ~2 per magnitude toward fainter UV, with roughly constant Eddington ratio.","If the missing population of UV-faint, [CII]-bright quasars are dust-obscured type-1 systems, deeper mid-IR photometry (e.g., MIRI) should reveal red continuum slopes for a larger fraction than the ~25% the authors currently estimate."],"forward_implications":["UV-bright z~6 quasars cannot be used alone to census supermassive black hole growth in massive galaxies; the dominant mode is UV-faint.","At fixed host mass, UV-faint quasars outnumber UV-bright ones by roughly 29 times, so models of early black hole growth must reproduce a large population of obscured or low-mass black holes in gas-rich hosts.","The median black hole mass in massive hosts at z~6 is estimated at log10(MBH/Msun)~8.1, consistent with the local MBH-host relation, making the famous UV-bright quasars ~15x more massive than typical.","The [CII] luminosity functions constructed for UV-bright and UV-faint quasars can be directly compared with [CII] surveys of galaxies to constrain quasar lifetimes and duty cycles."],"supporting_citations":[{"why":"Supplies the z~6 quasar UV luminosity function used to convert the observed [CII]-luminous fractions into volume densities.","marker":"Schindler et al. (2023)"},{"why":"Establishes [CII] luminosity as a tracer of molecular gas mass, the foundation for using L[CII] as a host-mass proxy.","marker":"Zanella et al. (2018)"},{"why":"Describes the CISTERN ALMA program that provides [CII] measurements for the UV-faint quasars, the new data enabling the census.","marker":"Bouwens et al. (2025, in prep)"},{"why":"Supplies the literature compilation of [CII] measurements of bright z~6 quasars and the earlier framework for [CII] studies.","marker":"Decarli et al. (2018)"},{"why":"Provides [CII] data for many UV-bright quasars and the methodology used to compute IR luminosities.","marker":"Venemans et al. (2020)"},{"why":"Provides the z~7 quasar UV luminosity function used to test how the ~29x excess changes at higher redshift.","marker":"Matsuoka et al. (2023)"},{"why":"Supplies the FWHM-based dynamical mass fitting formula used to show that UV-faint [CII]-bright quasars have massive hosts.","marker":"Neeleman et al. (2021)"},{"why":"Supplies the local MBH-host scaling relation against which the inferred z~6 black hole masses are compared.","marker":"Kormendy & Ho (2013)"}],"fun_headline_variants":["UV-bright quasars are outliers in massive hosts at z>5.9","Faint outnumber bright quasars 29-to-1 in massive hosts at z~6","Quasar UV brightness misleads on host mass at cosmic dawn","85% of massive-host quasars at z>5.9 are UV-faint after all"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The observed fraction of quasars with a bright [CII] line in each UV-brightness bin is assumed to apply to every z>5.9 quasar in that bin; if the ALMA targets were preferentially drawn toward sources likely to show [CII] (for example, because their redshifts were already known), the 21% massive-host fraction for UV-faint quasars would be overestimated and the headline 85% would shrink.","fun_headline_variants_meta":{"raw":{"variants":["UV-bright quasars are outliers in massive hosts at z>5.9","Faint outnumber bright quasars 29-to-1 in massive hosts at z~6","Quasar UV brightness misleads on host mass at cosmic dawn","85% of massive-host quasars at z>5.9 are UV-faint after all"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000224,"raw_usage":{"total_tokens":1616,"prompt_tokens":1258,"completion_tokens":358,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":874,"completion_tokens_details":{"reasoning_tokens":268}},"tokens_in":874,"tokens_out":358,"duration_ms":4251,"temperature":1.0,"reasoning_tokens":268,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T21:23:40.157006+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a UV-complete, spectroscopically confirmed sample of z>5.9 quasars with M_UV,AB between -24.5 and -22, observe all of them with ALMA to a uniform [CII] depth, and measure the fraction with L[CII]>1.8e9 Lsun. If that fraction comes out substantially below 21%, the inferred >85% fraction fails; a particularly clean test is the same measurement on the faintest bin (M_UV,AB>-23.5), where the paper sees 7 massive-host quasars and the assumption about representativeness is most strained.","supporting_citations":[{"cited_title":"2023, , 943, 67","cited_arxiv_id":null,"evidence_quote":"Supplies the z~6 quasar UV luminosity function used to convert the observed [CII]-luminous fractions into volume densities."},{"cited_title":"P., Walter , F., Neeleman , M., et al","cited_arxiv_id":null,"evidence_quote":"Provides [CII] data for many UV-bright quasars and the methodology used to compute IR luminosities."},{"cited_title":"2023, , 949, L42","cited_arxiv_id":null,"evidence_quote":"Provides the z~7 quasar UV luminosity function used to test how the ~29x excess changes at higher redshift."},{"cited_title":"P., et al","cited_arxiv_id":null,"evidence_quote":"Supplies the FWHM-based dynamical mass fitting formula used to show that UV-faint [CII]-bright quasars have massive hosts."}],"review_version":1}