{"id":"19d2e786-d8d0-4e15-a09a-c89ebcfc0fe9","arxiv_id":"2609.09266","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"Intermediate-luminosity quasars at z~6-7 have systematically lower black hole masses and higher Eddington ratios than luminosity-matched quasars at z~1.3, with a significant super-Eddington fraction.","lead":"This paper presents measurements of black hole masses and accretion rates for 27 quasars at redshifts between 6 and 7, using new JWST and Subaru spectroscopy. It finds that these early-universe quasars have lower masses and higher Eddington ratios than similar-luminosity quasars at lower redshift, with up to 22% accreting near or above the Eddington limit.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"The 27-object sample mixes 12 Hβ-based and 15 MgII-based virial masses while the z≈1.3 control is MgII-only; without an overlap calibration, the −0.4 dex offset and the super-Eddington tail may be partly inter-estimator systematics rather than redshift evolution.","rationale":"The reader's weakest assumption identifies the global virial-estimator uncertainty (R_BLR–L and f at z≈6), which is real and transparently discussed. The more load-bearing issue for this particular claim is narrower: the high-z distribution is constructed from two different estimators while the low-z control uses only one, and the paper itself notes that no object has both lines. A systematic offset between the Hβ and MgII recipes of order 0.3–0.5 dex would be comparable to the claimed −0.35 dex median mass offset and would directly affect the three MgII-based super-Eddington objects. The existing systematics tests do not close this gap, because the noise-degradation test only tests MgII at fixed quality, the Woo et al. 2026 test remaps both estimators but does not calibrate one against the other, and the Hα/Hβ comparison is internal to the JWST sample. The proposed stratified re-analysis is inexpensive and decisive: if the MgII-only subsample shows the same offset and tail, acceptance stands; if not, the central comparison should be reported separately for the two estimators or conditioned on a measured Hβ–MgII cross-calibration at the SHELLQs luminosity. For that reason I would move the verdict from ACCEPT to CONDITIONAL rather than reject the paper, whose data and analysis are otherwise careful and transparent.","tokens_in":47687,"tokens_out":6118,"duration_ms":64291,"concrete_test":"Recompute the §4.2 Monte Carlo luminosity-matching analysis twice: once for the 12 Hβ-based SHELLQs objects and once for the 15 MgII-based SHELLQs objects, each compared with the same luminosity-matched MgII-only z≈1.3 SDSS control using the same prescriptions (Eqs. 7–9). If the MgII-only high-z subsample alone reproduces the Δlog M_BH ≈ −0.35 dex offset and the super-Eddington tail, the combined result is robust to estimator mixing. If the offset and tail instead weaken or disappear in the MgII-only subsample, the headline result is at least partly a Hβ/MgII cross-calibration artifact and should be reported separately or corrected with a measured relative calibration.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central comparison in §4.1–4.2 combines two single-epoch mass estimators into one high-redshift distribution: 12 Hβ-based masses from Eq. 5 (JWST targets) and 15 MgII-based masses from Eq. 7 (MOIRCS plus previous SHELLQs measurements). The z≈1.3 SDSS control is MgII-only, and the paper states explicitly that none of the 27 objects has both Hβ and MgII measurements, noting that individual high-redshift objects can show Hβ–MgII offsets up to ≈0.5 dex. The claimed high-redshift versus low-redshift median offset is −0.35 dex, comparable to that inter-estimator range. All three fiducial super-Eddington objects (J0921+0007, J1217+0131, J0859+0052) are MgII-based, so the Eddington-ratio tail is directly sensitive to the MgII calibration. If, at the SHELLQs luminosities, the MgII recipe is systematically offset relative to the Hβ recipe, the mixed-sample median mass and the super-Eddington fraction could shift substantially without any change in the accretion physics. This is distinct from the general 0.4 dex systematic in the R_BLR–L and f calibration: it is an internal, checkable cross-calibration between the two estimator families used to build the same sample. The paper's other systematics checks (noise degradation, iron template choice, Woo et al. 2026 recalibration) do not directly address this Hβ–MgII mixing.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper presents rest-frame near-UV/optical spectroscopy of 21 SHELLQs quasars at 6.07 <= z <= 6.90: 12 from JWST/NIRSpec (Balmer lines) and nine from Subaru/MOIRCS (MgII). Combining these with earlier SHELLQs work yields 27 objects with single-epoch virial black hole masses. The authors report a median log(M_BH/M_sun) = 8.6 and median log(L_bol/L_Edd) = -0.63, and, relative to luminosity-matched z ~ 1.3 SDSS quasars, median offsets of -0.35 dex in M_BH and +0.23 dex in Eddington ratio, with 11% of the high-redshift sample near or above Eddington (22% with the Woo et al. 2026 calibration). They interpret this as evidence for more efficient SMBH growth at z ~ 6-7, extending downsizing to the intermediate-luminosity regime.","tokens_in":48086,"tokens_out":10511,"duration_ms":94150,"significance":"The sample is valuable: it occupies the poorly explored log L_bol ~ 46-47 regime between luminous high-redshift quasars and faint JWST AGN. The paper is careful in several specific ways: two MgII iron templates, a noise-degradation test on SDSS spectra, an H-alpha versus H-beta mass comparison (Fig. 13), Monte Carlo resampling, and a machine-readable table of spectral measurements. If the offsets are astrophysical, the result provides an important demographic anchor for theories of early SMBH growth. At present, however, the headline comparison mixes two single-epoch estimators in the high-redshift sample while the control is MgII-only, and the paper's own text says that none of the 27 objects has both lines and that individual H-beta-MgII offsets can reach ~0.5 dex. This is comparable to the claimed -0.35 dex offset, so the central comparison needs an explicit cross-estimator treatment before the evolutionary conclusion can be taken at face value.","major_comments":[{"comment":"This is the load-bearing point. The high-redshift distribution is built from 12 H-beta-based masses (Eq. 5) and 15 MgII-based masses (Eq. 7), while the z ~ 1.3 control is MgII-only (Section 4.2). The paper explicitly notes in Section 4 that no object in the sample has both H-beta and MgII and that individual high-redshift objects can show inter-estimator offsets up to ~0.5 dex, which is the same order as the reported median Delta log M_BH = -0.35 dex. Moreover, all three objects defining the fiducial super-Eddington fraction are MgII-based. The existing H-alpha-H-beta comparison (Fig. 13) does not constrain the H-beta-MgII relation. Please add: (i) the separate MgII-only and H-beta-only SHELLQs distributions compared with the same luminosity-matched control; (ii) a quantitative propagation of an inter-estimator offset taken from the literature or from SDSS quasars with both lines measured, including the effect on the super-Eddington fraction; and (iii) a statement of whether the redshift offset survives each estimator considered alone. This is a checkable internal calibration issue rather than the general 0.4 dex systematic discussed for the local R_BLR-L and f calibrations.","section":"Section 4.1-4.2, Eqs. (5), (7)"}],"minor_comments":[{"comment":"The noise-degradation test states that the +/-19% MgII FWHM scatter propagates to +/-0.17 dex through Eq. (5), but Eq. (5) is the H-beta-based mass recipe; the MOIRCS MgII masses are computed with Eq. (7). The equation reference should be corrected and the propagation redone with the actual MgII recipe.","section":"Section 4.3"},{"comment":"The abstract and Section 5 quote the mass range as 7.2 < log(M_BH/M_sun) < 9.4, whereas Table 6 gives log(M_BH/M_sun) = 6.97 for J0217-0208 and Section 4.1 states 7.0 <= log(M_BH/M_sun) <= 9.5; these values should be reconciled after deciding how to treat the marginal classification of J0217-0208.","section":"Abstract, Section 4.1, Table 6"},{"comment":"The sentence 'This result suggests that the enhanced fraction of rapidly accreting SMBHs at high redshift contributes to the observed difference from the low-redshift population' appears twice in Section 4.2; one copy should be removed.","section":"Section 4.2"},{"comment":"Given that selection functions and completeness corrections are deferred to a future paper, the word 'representative' in the abstract and Section 5 is stronger than the current sample allows; I suggest softening it or adding a caveat about the incomplete near-IR follow-up within SHELLQs.","section":"Abstract and Section 5"},{"comment":"The abstract quotes Delta log M_BH = -0.4 and Delta log L_bol/L_Edd = +0.2, while Section 4.2 reports -0.35 and +0.23; the values and their rounding should be made consistent.","section":"Abstract and Section 4.2"}],"recommendation":"major_revision","confidential_remarks":"For the editor: I agree with the reader's positive evaluation of the data quality and the value of the sample, but I do not agree that the central comparison is ready for acceptance in its current form. The mixed H-beta/MgII estimator issue is not a generic systematic caveat: the claimed median offset is of the same size as the paper's own stated H-beta-MgII scatter, and the super-Eddington tail is entirely MgII-based. The requested split-by-estimator and inter-calibration analyses are well within the scope of the manuscript and should be included. I see no citation or scope problems, and I do not regard the issue as grounds for rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"The paper is a workmanlike study that does what it sets out to do: it presents new JWST and Subaru spectroscopy for 21 intermediate-luminosity z~6 quasars and combines these with earlier SHELLQs measurements to construct a 27-object sample with black hole mass estimates. The spectral fitting is thorough, the systematic checks are reasonable, and the commitment to making line measurements public is appreciated. The central finding – that the high-redshift sample has systematically lower masses and higher Eddington ratios than a luminosity-matched z~1.3 control – is interesting and consistent with the anti-hierarchical growth picture.\n\nThe soft spot is the one the stress-test note flags: the 27-object sample mixes 12 Hβ-based masses from JWST with 15 MgII-based masses from MOIRCS and earlier work, while the low-redshift control is MgII-only. The paper acknowledges that no object in the sample has both Hβ and MgII, and that individual objects can show Hβ–MgII offsets of ~0.5 dex. The median offset they report between the high- and low-redshift samples is -0.35 dex – comparable to that inter-estimator range. The three fiducial super-Eddington objects are all MgII-based. This means that some fraction of the claimed offset and the super-Eddington tail could reflect systematic differences between the two mass recipes rather than genuine redshift evolution.\n\nThe paper does not directly address this issue. The noise-degradation test and iron template checks are fine, but they don't do a cross-calibration. The Woo et al. 2026 recalibration that they apply in Section 4.4 shifts both estimators, but it doesn't remove the ambiguity of mixing them.\n\nThis is not a fatal flaw. The authors are transparent about the limitation, and the trends they report are plausible and broadly consistent with earlier work on more luminous quasars. But the specific numbers – the -0.4 dex offset and the 11-22% super-Eddington fraction – should be interpreted with the estimator-mixing caveat firmly in mind. A reader who cites these values should add that caveat.\n\nThe paper is a solid contribution that deserves to be published after the authors either add a cross-calibration check (e.g., using objects with both Hβ and MgII from the literature at comparable luminosities) or explicitly restate their conclusions with the caveat that the mixing of estimators contributes systematic uncertainty of up to ~0.3-0.4 dex to the sample comparison. I would send this out for peer review.","headline":"Solid data and careful spectral fitting, but the key population comparison mixes two different mass estimators without cross-calibration, so the claimed -0.4 dex offset and super-Eddington fraction should be taken with a grain of salt.","tokens_in":48709,"tokens_out":1108,"would_cite":false,"duration_ms":11856,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek/deepseek-v4-flash","headline":"Intermediate-luminosity quasars at z~6-7 have systematically smaller black holes and higher accretion rates than their low-redshift counterparts, with up to 22% accreting at super-Eddington rates.","keywords":["quasars","high redshift","supermassive black holes","Eddington ratio","SHELLQs","virial masses","JWST","Subaru"],"falsifier":"If reverberation mapping of any z~6 quasar in this mass-luminosity range showed a BLR radius that deviates from the local R_BLR–L relation by more than 0.4 dex, the mass estimates and thus the inferred Eddington ratios would be systematically wrong, potentially eliminating the apparent super-Eddington excess.","tokens_in":49029,"feed_emoji":"🕳️","tokens_out":2975,"duration_ms":27080,"temperature":0.7,"pith_summary":"The paper assembles a sample of 27 quasars at redshift 6 to 7 from the SHELLQs project, using JWST and Subaru spectroscopy to measure black hole masses via broad emission lines. It finds a median mass of about 400 million solar masses and a median Eddington ratio of 0.23. Compared to luminosity-matched quasars at z~1.3, the high-redshift sample shows black hole masses 0.4 dex lower and Eddington ratios 0.2 dex higher. Moreover, 11-22% of these early quasars accrete at or above the Eddington limit, suggesting that rapid accretion is common during the first billion years. This extends the anti-hierarchical (downsizing) growth scenario to a more representative population.","feed_headline":"Up to 22% of z~6 quasars accrete at super-Eddington rates","feed_subtitle":"A sample of 27 early quasars shows a rapid accretion phase is common, supporting anti-hierarchical black hole growth.","key_machinery":"The single-epoch virial black hole mass estimator, which uses the width of broad emission lines (Balmer lines from JWST/NIRSpec; MgII from Subaru/MOIRCS) as a proxy for virial velocity, and the continuum luminosity to infer the broad-line region radius via the locally calibrated R_BLR–luminosity relation (R_BLR ∝ $L^{0}$.5). The geometric factor f is assumed to be the same as in the local universe.","core_discovery":"The central discovery is that intermediate-luminosity quasars at 6<z<7 — a regime bridging luminous quasars and faint JWST AGN — have a median black hole mass of log(M_BH/M⊙)=8.6 and median Eddington ratio log(L_bol/L_Edd)=-0.63. When matched in bolometric luminosity to SDSS quasars at z≈1.3, the high-redshift sample shows median offsets of Δlog M_BH = -0.4 dex and Δlog L_bol/L_Edd = +0.2 dex. Furthermore, 11% (rising to 22% under an Eddington-ratio-dependent mass calibration) of the high-redshift quasars accrete near or above the Eddington limit, indicating that a significant fraction of early supermassive black holes undergo a rapid growth phase.","pith_inferences":["If the SHELLQs sample is representative, then the intrinsic black hole mass function at z~6 is skewed toward masses below 10^9 M⊙, and the space density of lower-mass active black holes is higher than previously inferred from luminous quasar surveys alone.","The elevated Balmer decrements and red optical continua seen in a subset of objects (J0844−0132 and J1146−0005) suggest a possible connection to the 'Little Red Dot' population, implying an evolutionary sequence from dust-obscured rapid growth to unobscured quasar phases.","A clean test of the paper's central result would be a systematic reverberation mapping campaign of a few z~6 quasars using JWST to directly measure the BLR size and check the R_BLR–luminosity relation at high redshift.","The result that 22% are super-Eddington under the Woo+2026 calibration is sensitive to the assumed f-factor evolution; future ALMA or JWST observations of host galaxy dynamics could provide independent BH mass checks via the M-sigma relation."],"forward_implications":["A significant fraction of supermassive black holes in the reionization epoch grow through near- or super-Eddington accretion episodes, which must be accounted for in models of early black hole formation and growth.","The observed anti-hierarchical trend (lower masses, higher Eddington ratios at higher redshift) extends from the most luminous quasars down to intermediate luminosities, supporting a downsizing picture for black hole accretion over cosmic time.","The systematic offset in black hole mass at fixed luminosity implies that the black hole mass function at z~6 peaks at lower masses than at z~1, with implications for the progenitors of today's most massive black holes.","If the Eddington-ratio-dependent calibration is correct, the super-Eddington fraction doubles, significantly alleviating the tension between the rapid emergence of billion-solar-mass black holes and the limited time available at z>6."],"supporting_citations":[{"why":"Provides the Hβ-based single-epoch black hole mass calibration used for the JWST targets.","marker":"M. Vestergaard & B. M. Peterson (2006)"},{"why":"Provides the MgII-based black hole mass calibration used for the MOIRCS targets.","marker":"M. Vestergaard & P. S. Osmer (2009)"},{"why":"Supplies the local R_BLR–luminosity relation that translates continuum luminosity into BLR radius.","marker":"M. C. Bentz et al. (2013)"},{"why":"Provides the Hα-based mass estimator used as a cross-check and for the Hα-based Eddington ratios.","marker":"J. E. Greene & L. C. Ho (2005)"},{"why":"Quantifies the 0.4 dex systematic uncertainty inherent in single-epoch virial mass estimates, against which offsets are compared.","marker":"Y. Shen (2013)"},{"why":"Defines the luminosity-function sample from which the SHELLQs targets are drawn, establishing the selection baseline.","marker":"Y. Matsuoka et al. (2018c)"},{"why":"Provides the low-redshift SDSS comparison sample with MgII-based mass estimates for the luminosity-matched analysis.","marker":"Q. Wu & Y. Shen (2022)"},{"why":"Introduces the Eddington-ratio-dependent single-epoch calibration used to test the sensitivity of the super-Eddington fraction.","marker":"J.-H. Woo et al. (2026)"}],"fun_headline_variants":["22% of early quasars feed above the Eddington limit","Rapid black hole growth common in z~6 quasars","One in five early quasars go super-Eddington","Super-Eddington accretion found in 22% of z>6 quasars","Early quasars show frequent super-Eddington phases"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The locally calibrated single-epoch black hole mass estimator, which assumes the broad-line region size scales as luminosity to the 0.5 power and uses a geometric factor f that encodes BLR geometry and kinematics, applies without redshift evolution at z~6.","fun_headline_variants_meta":{"raw":{"variants":["22% of early quasars feed above the Eddington limit","Rapid black hole growth common in z~6 quasars","One in five early quasars go super-Eddington","Super-Eddington accretion found in 22% of z>6 quasars","Early quasars show frequent super-Eddington phases"]},"model":"deepseek/deepseek-v4-flash","effort":"low","cost_usd":0.000305,"raw_usage":{"cost":0.00030464,"is_byok":false,"cost_details":{"upstream_inference_cost":0.00030464,"upstream_inference_prompt_cost":0.00016072,"upstream_inference_completions_cost":0.00014392},"total_tokens":1662,"prompt_tokens":1148,"completion_tokens":514,"prompt_tokens_details":{"audio_tokens":0,"video_tokens":0,"cached_tokens":0,"cache_write_tokens":0},"completion_tokens_details":{"audio_tokens":0,"image_tokens":0,"reasoning_tokens":425}},"tokens_in":1148,"tokens_out":514,"cost_source":"openrouter_usage","duration_ms":6333,"image_pages":[],"temperature":1.0,"reasoning_tokens":425,"transport_provider":"openrouter","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-09-10T17:43:00.432451+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"falsifier":"If reverberation mapping of any z~6 quasar in this mass-luminosity range showed a BLR radius that deviates from the local R_BLR–L relation by more than 0.4 dex, the mass estimates and thus the inferred Eddington ratios would be systematically wrong, potentially eliminating the apparent super-Eddington excess.","supporting_citations":[],"review_version":1}