{"id":"c27fb530-bbcd-45ea-85fc-6f5587d8e32a","arxiv_id":"1908.09673","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"In a sample of 86 hyperluminous quasars, 24 percent show C IV broad absorption lines, roughly double the fraction at lower luminosities, with faster outflows and two candidates at 0.15c.","lead":"This paper measures the fraction of quasars with broad ultraviolet absorption lines (BALs) in a sample of 86 hyperluminous quasars, finding 24 percent with the standard C IV-based index, about twice the fraction in less luminous samples. The result suggests that the most luminous quasars produce more and faster outflows, which may matter for how these systems affect their host galaxies.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed intrinsic BAL excess depends on a mean-SNR correction and unmatched comparisons; it may not survive a matched automated re-analysis.","rationale":"The reader identified sample selection and SNR as the weakest assumption, and I agree that the SNR correction is the most load-bearing step. My concern sharpens this: the 2% correction is derived from a mean-SNR comparison, not a matched analysis, and the visual versus automated classification difference is an additional uncontrolled channel. The paper's own statement that G09 found an SNR dependence, and the two newly classified BALs with redshift/continuum issues, show that methodology can change classifications. I do not find a formal error or evidence of bad faith; the measurement is carefully documented and the uncertainties are honestly reported. However, the central claim that the excess is intrinsic to the hyperluminous regime requires a common-pipeline, SNR-binned comparison to rule out bias. Since the reader's CONDITIONAL verdict already reflects this uncertainty, my recommendation is UNCHANGED. If the proposed re-analysis were performed and the excess disappeared, the verdict should move to REJECT or UNVERDICTED, but on current evidence CONDITIONAL remains appropriate.","tokens_in":22240,"tokens_out":5514,"duration_ms":62743,"concrete_test":"Re-run one identical automated BAL-detection pipeline (e.g., the G09/Paris code) on the 86 WISSH SDSS spectra and on a redshift- and SNR-matched random subset of G09 spectra. Then compare BI>0 fractions using the same continuum-fitting algorithm, and bin both samples by 1500-1600 Å SNR to compute per-bin G09 BAL fractions; apply those binwise efficiencies to the WISSH SNR distribution. If the automated WISSH fraction is consistent with <18% or the binwise SNR-corrected prediction reaches about 24%, the claimed luminosity-driven excess is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that the 24% C IV BI>0 fraction is intrinsic to Lbol>1e47 quasars (Sec. 3.1). This requires that the comparison with G09 is controlled for all observational biases. The paper only corrects for the mean SNR difference (Sec. 3.2), adopting a 2% shift from G09. That control is too coarse: (i) G09's SNR relation is an average over lower-luminosity SDSS quasars and is applied to the mean SNRs (9 vs 18), not to the full WISSH SNR distribution; if the detection efficiency rises steeply above SNR~15, the correction could be several times larger. (ii) WISSH spectra were visually inspected with hand-fitted IRAF splines, while G09 used an automated algorithm, so a systematic difference in continuum placement can inflate BI>0 counts without a physical cause. (iii) The WISSH sample is WISE/SDSS-selected; if hyperluminous selection preferentially includes reddened sightlines, the BAL fraction is biased high relative to the optical quasar samples. The two newly identified BALs (0414+06, 1210+17) show that redshift/continuum issues do change classifications. Without a matched, automated, SNR-binned analysis, the 9% excess over the largest literature value is not securely attributable to luminosity.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper searches for broad absorption line (BAL) quasars in the WISSH sample of 86 hyperluminous quasars (L_bol > 10^47 erg/s, 1.8 < z < 4.6) using SDSS DR12 spectra. The authors visually identify BAL candidates, fit continua with IRAF splines, and compute the AI, AI1000, and BI indices. They report 21/86 (24±5%) objects with C IV BI>0, which is roughly twice the literature fraction (10-15%), as well as larger mean BI and maximum velocities, and two candidate ultra-fast BAL outflows at ~0.15c. They interpret the excess as evidence that hyperluminous quasars accelerate BAL winds more efficiently, likely through radiation pressure, and estimate kinetic powers that could be relevant for feedback.","tokens_in":22466,"tokens_out":8489,"duration_ms":83831,"significance":"If the reported excess BAL fraction is intrinsic to the hyperluminous regime, it would be an important observational constraint on wind-driving mechanisms and on the luminosity dependence of quasar outflows. The paper uses standard BAL definitions and public SDSS spectra, and the appendix provides spectra of all 38 BAL candidates, which is a useful resource. The two UBO candidates are intriguing. However, the central comparison with lower-luminosity samples is not fully controlled: the analysis relies on a coarse SNR correction, a visual selection method that is not calibrated against the automated algorithms used for the comparison sample, and a parent sample whose selection could bias the BAL fraction. The claimed intrinsic luminosity dependence therefore requires additional verification.","major_comments":[{"comment":"The SNR correction is applied to the mean SNR of the BAL-detected subsamples (9 vs 18 from the Shen et al. 2011 catalogue) rather than to the SNR distributions of the parent quasar samples from which the BAL fractions are measured. G09's SNR-BAL fraction relation is a property of the parent sample; to estimate the bias in the WISSH fraction, one must compare the full WISSH parent sample (86 objects) with the full G09 parent sample in matched SNR bins. As written, the 2% correction is not a valid estimate of the detection-efficiency bias, and the conclusion that SNR can account for only a small part of the excess (Sec. 5, second bullet) is not supported. A matched SNR-binned analysis, ideally applying the same automated BAL search to both samples, is needed.","section":"Sec. 3.2, Fig. 1"},{"comment":"The WISSH BAL selection relies on visual inspection and IRAF spline continuum fits, whereas the G09 comparison sample is classified by an automated algorithm. These methods are not shown to be equivalent; the fact that two WISSH objects (0414+06 and 1210+17) were not found in previous automated catalogues demonstrates that the visual method is more permissive. Without an inter-comparison (e.g., running the visual method on a subsample of G09, or an automated code on WISSH), the observed fraction may be systematically inflated relative to the literature values, undermining the central claim of an intrinsic excess.","section":"Sec. 2 and Sec. 3.1"},{"comment":"The WISSH parent sample is selected using WISE and SDSS photometry, which may preferentially include reddened quasars, and reddened quasars are known to have a higher BAL fraction (as the paper itself notes from Dai et al. 2008). The paper does not test whether the WISSH color distribution or SEDs indicate such a bias. If present, the elevated BAL fraction could be a selection effect rather than a direct consequence of L_bol. The authors should compare the reddening/colors of WISSH to the G09 sample and, if necessary, restrict the comparison to matching color or extinction ranges.","section":"Sec. 3.1 and Sec. 2"}],"minor_comments":[{"comment":"Individual BI, AI, vmin, and vmax values are reported without uncertainties; adding error bars would strengthen the statistical comparisons in Secs. 3.2 and 3.4.","section":"Table 1"},{"comment":"The KS test between the WISSH and G09 BI distributions is based on samples of very different sizes (21 vs 3874) and without accounting for measurement uncertainties; a two-sample test that incorporates errors or a non-parametric approach with proper weighting would be more appropriate.","section":"Sec. 3.2"},{"comment":"There is a typo in the text: 'beetwen' should be 'between'.","section":"Sec. 3.4"},{"comment":"The text and figure caption refer to the WISSH BAL region as 'purple' in one place and 'magenta' in another; the colors should be used consistently.","section":"Fig. 5 and Sec. 3.4"},{"comment":"The identification of the two UBOs as C IV absorption at ~0.15c is based on the absence of strong absorption between Si IV and C IV and on ionization arguments; supporting constraints from other transitions or photoionization modeling would make this more robust, though the authors are appropriately cautious in their wording.","section":"Sec. 3.3"}],"recommendation":"major_revision","confidential_remarks":"The paper is from an experienced collaboration and uses public data, but the central claim of an intrinsic BAL excess is not yet securely established. The main issues are the flawed application of the SNR correction to the BAL subsample rather than the parent sample, the uncalibrated visual selection, and the potential WISE-selection bias. I recommend requiring a matched automated re-analysis or a clear sensitivity analysis before the claim of a luminosity-driven intrinsic excess is accepted. The paper's conclusions should be tempered accordingly."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is the first BAL census at Lbol > 1e47 erg/s, and the headline number — 24% C IV BI>0 versus 10–15% for lower-luminosity samples — is genuinely new in a regime nobody had probed. I think the excess is real in direction, but the paper does not quite pin down its size: the comparison with G09 is not matched, and the SNR correction is applied to means, not to the full distribution.\n\nWhat the paper does well: the indices are standard (Weymann BI, Hall AI), the spectra are public SDSS DR12 data, and all 38 candidates appear in the appendix with fits and residuals, so the classification is checkable. The authors flag the SNR confound themselves and attempt a correction, and the two UBO candidates (0.15c) are a real addition to a small census. The writing is honest about limits — the UBO fraction comparison is explicitly said to be hampered by statistics.\n\nSoft spots, in order. First, the G09 comparison mixes methods: WISSH was visually classified with hand-fitted splines, G09 used an automated algorithm. That two objects (0414+06, 1210+17) flip classification between pipelines shows this is not a nit; it cuts both ways, and the paper does not quantify how much of the 11-point gap could be method. Second, the SNR issue: even if the mean-based correction under-corrects by a factor of a few, most of the excess survives, so this is not load-bearing — but a matched, SNR-binned re-analysis is needed before quoting 24% as the intrinsic fraction. Third, Table 1 gives no uncertainties on BI, AI, or vmax; propagation from the spectra would be straightforward and should be added.\n\nThe kinetic-power/feedback section is the weakest. It assumes distances (1 pc to 1 kpc), adopts one fiducial column density, and puts the mass at vmax. The authors label it an estimate and give lower limits, so it is not fatal, but the abstract's 'efficient feedback' sentence overstates what the data show.\n\nCitation practice is fair: G09, Allen, Trump, and Dai are all engaged, and the Dai et al. 2MASS comparison supports the luminosity interpretation rather than dodging it. No circularity in the BAL measurement; the WISSH Lbol values come from the team's earlier papers, but the fraction itself is measured independently on public spectra.\n\nBottom line: send it to a serious referee. The right referee request is a matched, automated, SNR-binned re-analysis and trimmed kinetic-power claims. The core measurement is new, reproducible, and likely correct in direction — a data point the AGN-wind community will want.","headline":"A credible first BAL census at Lbol > 1e47 erg/s: the 24% fraction versus ~13% for G09 is likely real in direction, but the comparison is imperfect and the exact size of the excess is not yet pinned down.","tokens_in":23090,"tokens_out":7292,"would_cite":true,"duration_ms":68951,"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":"Hyperluminous quasars launch C IV broad absorption winds almost twice as often as ordinary quasars, and the winds are faster, indicating radiation-pressure acceleration.","keywords":["broad absorption lines","BAL quasars","hyperluminous quasars","C IV outflows","quasar winds","radiation pressure driving","active galactic nuclei feedback","SDSS DR12 spectra"],"falsifier":"Measure the C IV BAL fraction in a sample covering $L_{\\rm bol}$ from $10^{46}$ to $10^{48}$ erg/s at $z\\approx2$\\,--\\,4, selected without a luminosity cut and analysed with a blind automated BAL finder at uniform SNR; if the fraction stays near 15\\% across the whole range once selection effects are controlled, the claimed luminosity-driven increase to 24\\% would be refuted. A cheaper check is to degrade the 86 WISSH spectra to SNR of about 9 and re-run the classification: if most of the 21 $BI>0$ objects no longer qualify, the excess fraction would not be robust.","tokens_in":22043,"feed_emoji":"💨","tokens_out":9189,"duration_ms":87387,"temperature":0.7,"pith_summary":"The paper sets out to establish that BAL winds are more common and more powerful in the hyperluminous regime ($L_{\\rm bol}>10^{47}$ erg/s) than in the ordinary quasar population. Using SDSS DR12 spectra of the 86 WISSH quasars, it finds 21 with C IV $BI>0$: an observed fraction of $24\\pm5\\%$, against the 10\\--15\\% found in earlier catalogues, with SNR contributing only about 2\\%. The paper also reports higher average $BI$ (about 4,000 km/s) and maximum velocities (about 17,000 km/s) than previous samples, plus two objects whose C IV absorption implies outflows at about 0.15c. If right, the result means quasar luminosity itself drives faster nuclear winds, most plausibly through radiation pressure, and that the fastest winds can deliver enough kinetic power to affect the host galaxy.","feed_headline":"Brightest quasars drive broad absorption winds twice as often","feed_subtitle":"In the 86 most luminous known quasars, C IV winds appear in 24% of objects and reach speeds near 0.15c.","key_machinery":"The load-bearing objects are the wind-detection indices applied to SDSS spectra: the Balnicity Index ($BI$), the Absorption Index ($AI$), and the intermediate $AI_{1000}$, all integrals of the normalized residual flux $(1 - f(v)/0.9)$ over $0$\\,--\\,$25,000$ km s$^{-1}$ from the C IV peak, with increasingly strict requirements on contiguous trough widths (2,000 km/s for $BI$). The $BI>0$ criterion is the standard that makes the WISSH fraction comparable to earlier catalogues; the $AI$ variants catch weaker and narrower troughs, and the ultra-fast BAL (UBO) classification—C IV absorption blueward of the Si IV peak, implying $v_{\\rm max}>0.1c$—identifies the relativistic tail.","core_discovery":"The paper reports that among 86 WISSH quasars with $L_{\\rm bol}>10^{47}$ erg/s, 21 satisfy the standard C IV Balnicity Index criterion ($BI>0$), an observed fraction of $24\\pm5\\%$—almost twice the 10\\--15\\% typical of previous optically selected samples—and that signal-to-noise differences contribute only about 2 percentage points of this excess. It further finds that WISSH BAL quasars have stronger absorption (mean $BI\\sim4,000$ km/s) and a maximum-velocity distribution distinct from that of the comparison sample ($p<0.015$), with two objects showing C IV absorption at about 0.15c blueward of the Si IV peak. Finally, it estimates that the fastest, densest of these winds carry kinetic power above 0.1\\% of $L_{\\rm bol}$, enough for meaningful feedback onto the host galaxy.","pith_inferences":["A luminosity-unbiased BAL survey from $10^{46}$ to $10^{48}$ erg/s at $z\\approx2$\\,--\\,4 could test whether the fraction rises smoothly with $L_{\\rm bol}$ or switches on above about $10^{47}$ erg/s.","If radiation pressure is the driver, the same data imply a dependence on Eddington ratio rather than luminosity alone; splitting WISSH quasars by Eddington ratio would separate these two drivers.","The two 0.15c C IV outflows are natural candidates for the ultraviolet counterparts of X-ray ultrafast outflows; coordinated X-ray spectroscopy of these two objects would test whether one wind is seen in both bands."],"forward_implications":["BAL wind incidence is luminosity-dependent: at $L_{\\rm bol}>10^{47}$ erg/s about 24\\% of quasars show C IV broad absorption, versus roughly 10\\--15\\% in lower-luminosity samples.","Hyperluminous BAL winds are faster and stronger, with a maximum-velocity distribution that differs from the comparison sample at $p<0.015$ and two objects reaching about 0.15c.","The $L_{\\rm bol}$\\,--\\,$v_{\\rm max}$ relation places BAL winds between galaxy-scale molecular/ionized winds and X-ray ultrafast outflows, supporting radiation-pressure launching.","For the highest column densities, BAL kinetic power exceeds 0.1\\% of $L_{\\rm bol}$, enough to affect the host galaxy if coupling is efficient.","Radio properties of hyperluminous BAL quasars match lower-luminosity ones, so the stronger winds do not come with a different jet production rate."],"supporting_citations":[{"why":"Supplies the comparison sample of 4,242 C IV BI>0 BAL quasars, including the BAL fraction, BI distribution, and SNR-dependence baseline.","marker":"G09"},{"why":"Defines the Balnicity Index used throughout the paper to classify BAL quasars.","marker":"Weymann et al. (1991)"},{"why":"Defines the Absorption Index used as a less conservative BAL identifier.","marker":"Hall et al. (2002)"},{"why":"Earlier SDSS BAL catalogue and source of the AI1000 variant and the average covering factor.","marker":"Trump et al. (2006)"},{"why":"Provides bolometric luminosities and SNR measurements for the comparison sample.","marker":"Shen et al. (2011)"},{"why":"Supplies the Lbol-vmax relation and the compiled wind populations against which WISSH BALs are placed.","marker":"Fiore et al. (2017)"},{"why":"Provides the radio-luminosity dependence of BAL fraction used to check WISSH radio properties.","marker":"Shankar et al. (2008)"},{"why":"Companion WISSH work on C IV broad emission line winds, providing redshifts and the [OIII]/weak-[OIII] classification for overlapping objects.","marker":"Vietri et al. (2018)"},{"why":"Supplies the kinetic-energy formula used to estimate BAL wind kinetic power.","marker":"Hamann et al. (2019)"},{"why":"Supplies column-density and spatial-extent constraints used for the feedback estimates.","marker":"Arav et al. (2018)"}],"fun_headline_variants":["Brightest quasars show 24% CIV BAL fraction, double lower-luminosity rate","Hyper-luminous quasars: 24% have CIV BALs, winds up to 0.15c","Brightest quasars: 24% have CIV BALs, speeds up to 0.15c - double rate","Hyper-luminous BAL quasars: double fraction, faster winds, up to 0.15c"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The 24\\% fraction is intrinsic to the hyperluminous regime and not a byproduct of the WISSH sample selection or of inspecting higher signal-to-noise spectra more sensitively than earlier catalogues.","fun_headline_variants_meta":{"raw":{"variants":["Brightest quasars show 24% CIV BAL fraction, double lower-luminosity rate","Hyper-luminous quasars: 24% have CIV BALs, winds up to 0.15c","Brightest quasars: 24% have CIV BALs, speeds up to 0.15c - double rate","Hyper-luminous BAL quasars: double fraction, faster winds, up to 0.15c"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000738,"raw_usage":{"total_tokens":3387,"prompt_tokens":1128,"completion_tokens":2259,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":744,"completion_tokens_details":{"reasoning_tokens":2149}},"tokens_in":744,"tokens_out":2259,"duration_ms":16462,"temperature":1.0,"reasoning_tokens":2149,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:04:34.949681+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the C IV BAL fraction in a sample covering $L_{\\rm bol}$ from $10^{46}$ to $10^{48}$ erg/s at $z\\approx2$\\,--\\,4, selected without a luminosity cut and analysed with a blind automated BAL finder at uniform SNR; if the fraction stays near 15\\% across the whole range once selection effects are controlled, the claimed luminosity-driven increase to 24\\% would be refuted. A cheaper check is to degrade the 86 WISSH spectra to SNR of about 9 and re-run the classification: if most of the 21 $BI>0$ objects no longer qualify, the excess fraction would not be robust.","supporting_citations":[{"cited_title":"R., Hall, P","cited_arxiv_id":null,"evidence_quote":"Earlier SDSS BAL catalogue and source of the AI1000 variant and the average covering factor."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the radio-luminosity dependence of BAL fraction used to check WISSH radio properties."},{"cited_title":"2019, MNRAS, 483, 1808","cited_arxiv_id":null,"evidence_quote":"Supplies the kinetic-energy formula used to estimate BAL wind kinetic power."}],"review_version":1}