{"id":"cce6a193-291a-4c8f-afd1-340b31c8e29b","arxiv_id":"2507.05380","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"About 21% of hard X-ray selected broad-line AGN show double-peaked broad H-alpha lines, and these objects tend to have higher black hole masses and lower Eddington ratios.","lead":"A census of 343 hard X-ray selected active galaxies finds that about 21% of those with broad optical lines show double-peaked profiles, likely from rotating accretion disk gas. The same galaxies tend to host heavier black holes accreting at lower rates, which could bias common black hole mass measurements.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 21% DPE fraction hinges on a subjective visual reclassification that changes 30% of algorithmic DPEs; without a reproducibility test, the headline and all population differences are not robust.","rationale":"The paper is a valuable first systematic census of DPEs in a hard X-ray selected sample, with careful fitting and multi-wavelength comparison. However, the central quantitative claim—21% prevalence—is not determined by the stated algorithm alone; the algorithm gives 30%, and the visual reclassification is a large, non-transparent correction. The reader identified this as the weakest assumption, and I agree. The mass comparison is based on only 18 DPEs with sigma*, which additionally raises a selection-bias question (Are those 18 representative?), but the classification issue is upstream and affects all 71 DPEs. Thus the single most load-bearing concern is the reproducibility of the DPE/non-DPE split. If an independent blinded review reproduces the 71/272 split, the conclusions are well-supported; if not, the headline fraction and all differences could shift. My recommendation remains CONDITIONAL: the paper should be accepted only if the visual classification is made reproducible (e.g., inter-rater agreement, clear decision rules, or release of individual classifications), since the current dependence on human judgment makes the core number fragile.","tokens_in":34016,"tokens_out":5306,"duration_ms":61713,"concrete_test":"Have two independent, blinded classifiers re-examine the 343 continuum-subtracted H-alpha fits using the same visual criteria (shoulder presence, red/blue asymmetry, outflow-like profiles), and compute inter-rater agreement (Cohen's kappa) and each classifier's DPE fraction. If the fraction varies by more than +/-5 percentage points or kappa <0.7, the 21% rate is not robust and the population comparisons should be re-run with a pre-registered algorithmic classifier.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—71 DPEs (21%) and the associated mass/Eddington differences—rests on the classification in Section 2. A disk model is fit to all 343 objects, and objects are classified as DPEs if i>14 deg, sigma>600 km/s, and xi1<1200. This yields 102 DPEs (30%). The authors then visually inspect the fits and reassign 46 of these to non-DPE (outflows) and add 15 non-DPEs to DPEs, netting 71 (21%). Thus the headline fraction depends on an undocumented, subjective step that removes nearly half of the algorithmic DPEs. The thresholds themselves are inherited from variability-selected ZTF AGN (Ward et al. 2024) and are not re-calibrated for hard X-ray selected BASS spectra, where S/N, outflow properties, and host galaxy contamination differ. Because all subsequent comparisons (masses, Eddington ratios, X-ray luminosity, morphology) use this split, any irreproducibility in the visual step propagates to every population claim. The paper acknowledges missed low-inclination DPEs, but the dominant uncertainty is the large manual reclassification, which changes the DPE fraction by 9 percentage points.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a systematic search for double-peaked emitters (DPEs) among 343 hard X-ray selected broad-line AGN from the BASS survey, fitting the Hα profile with the Chen & Halpern (1989) circular disk model. The authors report 71 DPEs, corresponding to about 21% of the sample, and provide best-fit disk parameters for these objects. They compare DPEs with the remaining 272 broad-line AGN across a range of multiwavelength properties, finding that DPEs have higher stellar-velocity-dispersion-based black hole masses by about 0.4 dex, lower Eddington ratios by about 0.3 dex, a preference for elliptical hosts, higher X-ray luminosities, and higher [O I]/Hα ratios, while showing no significant differences in optical and mid-IR variability, WISE colors, obscuration, αox, or radio luminosity distributions. The paper discusses these results in the context of disk-wind models of the broad-line region and the possible bias introduced by DPEs in virial black hole mass estimates.","tokens_in":34348,"tokens_out":4900,"duration_ms":59541,"significance":"If the classification is robust, this is one of the largest systematically selected samples of DPEs, and the hard X-ray selection reduces viewing-angle biases that affect optically selected samples. The claimed DPE fraction of about 21% is substantially higher than the 3.6% found in SDSS quasar samples and would strengthen the case that disk emission contributes to a significant fraction of broad-line AGN, with direct implications for virial mass measurements. The paper also provides best-fit disk parameters for 71 objects and assembles a homogeneous multiwavelength comparison, which is valuable. Notable strengths are the reproducible MCMC fitting procedure, the use of a well-characterized parent sample, and the consistency checks against previously known DPEs. However, the central classification rests on a subjective visual reclassification step that changes the sample by 31 objects, and the headline fraction is quoted without uncertainty; these issues need to be addressed before the population conclusions can be fully trusted.","major_comments":[{"comment":"The classification procedure as described in Section 2 is not fully reproducible: after the automated threshold selection yields 102 DPEs (30%), the authors 'visually inspected the disk profile fits' and reassigned 46 DPE candidates to the non-DPE class and 15 non-DPEs to the DPE class, resulting in 71 DPEs (21%). This manual step changes the headline fraction by 9 percentage points, yet no criteria, number of inspectors, inter-rater agreement, or uncertainty on the final 21% are provided. Because every subsequent population comparison in Sections 4–7 uses this split, the manual step is load-bearing. Please make the visual step reproducible by specifying quantitative decision rules, providing a catalog of the 61 reassigned objects with the reasons for each decision, and reporting a measure of inter-rater reliability. In addition, report the DPE fraction with a confidence interval, e.g., 71/343 = 20.7% with a binomial 95% CI, rather than the current uncertainty-free '~21%'.","section":"Section 2"},{"comment":"The thresholds i>14°, σ>600 km/s, and ξ1<1200 are adopted unchanged from Ward et al. (2024), a variability-selected ZTF sample, but are applied here to a hard X-ray selected BASS sample with different signal-to-noise, host galaxy contamination, and outflow properties. The transferability of these thresholds is not tested. In particular, the visual reassignment of 46 objects as outflows rather than disks indicates that the disk model can absorb asymmetric outflow profiles, and Section 7 notes that a large fraction of BASS Sy1.9 and Sy1 objects show [O III] outflow wings. Please demonstrate that the DPE fraction is stable under plausible variations of the thresholds (for example, i>12–20°, σ>400–800 km/s, ξ1<800–1600), or recalibrate the thresholds on a BASS-based validation subset. Without this, the 21% fraction and the derived population differences may depend sensitively on the adopted cuts.","section":"Section 2, Figure 1"},{"comment":"The abstract states that DPEs have 'lower Eddington ratios by ~0.3 dex', but the supporting KS test in Table 3 for log L/LEdd gives p=0.075 with only 17 DPEs and 81 non-DPEs, which is not significant at the conventional 0.05 level. The word 'confirm' in Section 5 is therefore too strong. Similarly, the mass difference of ~0.4 dex is based on only 18 DPEs (p=0.015) and is evaluated in a table with many comparisons, so the risk of false positives is nontrivial. Please rephrase these claims as trends or marginal results, report confidence intervals on the median differences, and consider a multiple-comparison correction or a note on the number of tests performed.","section":"Abstract, Section 5, Table 3"},{"comment":"The parent sample is reduced from 742 AGN to 343 broad-line AGN by excluding 196 Type 2 objects, 179 objects without Hα coverage or with continuum artifacts, and 24 objects with low-S/N broad lines. The paper does not compare the excluded objects with the included sample in terms of redshift, luminosity, or line properties. If the excluded objects preferentially lack Hα coverage at certain redshifts or have systematically weaker broad lines, the estimated DPE fraction could be biased. Please provide a comparison of the redshift, X-ray luminosity, and host galaxy properties of the 343 included objects with the 203 objects excluded after the Type 2 cut, or otherwise argue that the exclusion is independent of DPE status.","section":"Section 2"},{"comment":"The highly significant difference in [O I]/Hα ratios (p=2×10⁻⁵ in Table 3) is used in Section 7 as evidence for a lower ionization state in DPEs. However, Section 4 itself notes that 'for DPEs, the dip in the center of the Hα and Hβ broad line profiles can mean that the line ratios are inflated if the broad lines are modeled as Gaussians.' Since the narrow-line measurements from Oh et al. (2022) were obtained with Gaussian fits, this systematic effect may produce the observed difference rather than a physical difference in ionization. Please correct the narrow-line measurements using the disk-plus-Gaussian decomposition presented here, or explicitly downgrade the [O I]/Hα claim to a tentative result pending such a correction.","section":"Section 4, Section 7"}],"minor_comments":[{"comment":"The caption of Table 1 states 'Properties of the 70 DPE candidates' while the text and abstract consistently report 71 DPEs; please correct this inconsistency.","section":"Table 1"},{"comment":"The non-DPE median standard deviations for SIIb/Hα, NIIb/Hα, and OIIIb/Hβ are reported as 329.63, 11737.19, and 4715.37, which are implausibly large and likely reflect a formatting or data error; these entries should be checked and corrected.","section":"Table 3"},{"comment":"The text says that for BAT 744 no adequate model was found and its disk parameters are not reported, but it is unclear whether BAT 744 is counted among the 71 DPEs and, if so, how its non-detection of disk parameters affects the population statistics; please clarify.","section":"Section 2"},{"comment":"In the first paragraph of Section 7, the 19% detection rate among optically variable AGN is attributed to 'Ward et al. 2021', but the relevant reference appears to be Ward et al. (2024), which is cited earlier in the paper; please verify and correct the citation.","section":"Section 7"},{"comment":"In the caption of Figure 1, the DPE region is described as 'to the right of the vertical dashed lines' for the left and center panels and 'to the left' for the right panel; this is correct but could be made clearer by labeling the shaded DPE region directly in each panel.","section":"Figure 1"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the useful news: this is the first systematic DPE census in a hard X-ray selected AGN sample, and the paper does a lot of careful multi-wavelength legwork. The 21% lower-limit fraction is consistent with the ZTF variability-selected rate and the older radio-loud rate, which gives me some confidence that the number is in the right ballpark. The disk parameter tables and the comparisons across optical, IR, X-ray, and radio are genuinely valuable.\n\nThe main soft spot is the classification. The automated fit yields 30% DPE candidates, then visual inspection moves 61 objects, netting 21%. That visual step is not reproducible on the information given—no inter-rater test, no clear decision rule beyond \"outflows causing asymmetries.\" The authors call the 21% a lower limit, which is fair, but the 9-point swing means the headline number is a human judgment call. I'd like to see a blinded reclassification, or at least a table of the ambiguous objects. The thresholds from Ward et al. (2024) are arguably transferable, but that is an assumption.\n\nSecond, the Eddington ratio claim. In the abstract and conclusions they say DPEs have lower Eddington ratios by ~0.3 dex, but the KS test in Table 3 gives p=0.075. That is not enough to claim a confirmed difference at the usual 0.05 level. The mass difference (p=0.015) is on firmer ground, and the host galaxy and X-ray luminosity differences are more secure.\n\nMinor: the exclusions of 203 objects for no Hα coverage or low S/N are reasonable but not tested for bias. No code or data release is mentioned, which will slow follow-up. Also, the paper's statement that DPEs do not segregate in Lbol vs MBH is fine, but it is based on only 18 DPEs with dispersion-based masses.\n\nOverall, the paper is a solid contribution that should get a serious referee. The core result—DPEs are a substantial minority of X-ray selected broad-line AGN—is likely correct even if the exact fraction varies with the visual step. The authors are honest about limitations. I'd send it to review with a request to quantify the classification uncertainty and soften the Eddington ratio claim.","headline":"First hard X-ray selected census of double-peaked emitters with a plausible 21% fraction, but the classification rests on a subjective visual step and the Eddington ratio claim outruns the p-value.","tokens_in":34898,"tokens_out":3220,"would_cite":true,"duration_ms":38315,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Systematic disk-profile fitting of 343 hard X-ray selected AGN finds 71 double-peaked emitters, about 21%, with heavier black holes and lower Eddington ratios than other broad-line AGN.","keywords":["double-peaked emitters","accretion disks","broad line region","active galactic nuclei","virial black hole masses","Eddington ratio","BASS survey","Hα line profiles"],"falsifier":"If an independent, reproducible spectral classifier—applied to the same 343 continuum-subtracted spectra without human review—recovered a DPE fraction far from 21% (or failed to reproduce the 0.4 dex mass and 0.3 dex Eddington ratio gaps), the paper's central claim would be undercut. A direct observational check is to take multi-epoch rms spectra of a random subset of the 272 non-DPE AGN; if double-peaked variable components appear in fewer than about 5% of them, the inference that many hidden disk emitters contaminate the non-DPE sample would be weakened.","tokens_in":33843,"feed_emoji":"🔭","tokens_out":5110,"duration_ms":52133,"temperature":0.7,"pith_summary":"This paper tries to establish that roughly one in five hard X-ray selected broad-line AGN—71 out of 343—emits its broad Hα line from a rotating accretion disk rather than only from a wind outflow, and that these double-peaked emitters are intrinsically heavier and accreting at lower rates than the rest. The authors fit a circular disk model to every spectrum and report the disk geometry parameters for the 71 objects. If the classification holds, it matters because the standard virial method of measuring black hole masses assumes the broad line width reflects virialized gas; for disk emitters the width depends on inclination and turbulent broadening, so unrecognized disk emission biases mass estimates and could scatter the $M_{\\rm BH}$–$\\sigma_*$ relation. The paper also reports that DPEs are not distinguished by variability, mid-IR colors, obscuration, or changing-look rates, which narrows how they can be found.","feed_headline":"21 percent of X-ray selected AGN show double-peaked broad lines","feed_subtitle":"Disk-shaped emission biases black hole masses; DPEs are heavier and accrete slower (71 of 343 BASS AGN).","key_machinery":"The load-bearing object is the circular accretion-disk line-profile model of Chen & Halpern (1989), which predicts double-peaked Hα profiles from Keplerian rotation within a few hundred to a few thousand gravitational radii. The paper applies this model simultaneously with Gaussians for narrow forbidden lines to continuum-subtracted BASS spectra, allowing a spiral arm of variable amplitude, and separates DPEs from non-DPEs using threshold values of the fitted parameters ($i > 14^\\circ$, $\\sigma > 600$ km s$^{-1}$, $\\xi_1 < 1200$). A final visual inspection reassigns 46 candidates to the non-DPE class and 15 to the DPE class, so the fitted disk profile plus thresholds plus human review is what carries the classification.","core_discovery":"The central claim is that when the Chen & Halpern (1989) circular accretion-disk profile—parameterized by inclination $i$, turbulent broadening $\\sigma$, inner radius $\\xi_1$ in gravitational radii, and emissivity index $q$—is fitted to the broad Hα line of a flux-limited, hard X-ray selected sample of 343 BASS AGN, 71 objects (21%) satisfy the thresholds $i > 14^\\circ$, $\\sigma > 600$ km s$^{-1}$, and $\\xi_1 < 1200$ and are classified as double-peaked emitters. These DPEs have intrinsically higher black hole masses derived from host stellar velocity dispersion, by about 0.4 dex, and lower Eddington ratios, by about 0.3 dex, than the other 272 broad-line AGN; they also prefer elliptical hosts, are more X-ray luminous, and show higher [O I]/Hα narrow-line ratios. The paper argues this implies that disk-dominated emission is a common, not rare, state of the broad line region at low accretion rates, and that unrecognized disk components introduce biases in virial mass estimates for a non-negligible fraction of AGN.","pith_inferences":["If the 21% fraction holds for a hard X-ray selected sample, the true fraction of all broad-line AGN with significant disk emission could be higher, since low-inclination DPEs whose shoulders blend into a single peak are missed; the paper itself notes that 21% is a lower limit.","A testable extension is to apply the same fitting and thresholds to spectra with independent reverberation-mapping rms spectra; the rms-based double-peaked components should appear preferentially among the classified DPEs.","The visual reclassification step (46 reassigned out, 15 in) suggests the algorithm's raw thresholds overproduce DPEs in outflow-dominated objects; a fully objective classifier trained on outflow indicators such as [O III] wings could sharpen the selection and change the fraction."],"forward_implications":["At least 21% of hard X-ray selected broad-line AGN have double-peaked broad lines, so disk emission is a common BLR component at low Eddington ratios.","Virial black hole masses for DPEs will be overestimated unless the disk contribution is modeled, because the FWHM depends on inclination and turbulent broadening.","DPE populations with higher BH masses and lower Eddington ratios imply that Eddington ratio, not just viewing angle, regulates the disk-dominated BLR state.","The lack of differences in optical and mid-IR variability, WISE colors, $N_{\\rm H}$, $\\alpha_{\\rm ox}$, and changing-look rate means DPEs cannot be identified by these tracers alone.","Host galaxy morphology differences (46% elliptical versus 31% for an i-band matched control) indicate a connection between black hole mass, accretion state, and host properties."],"supporting_citations":[{"why":"Supplies the circular accretion disk model used to fit every Hα profile in the sample.","marker":"Chen & Halpern 1989"},{"why":"Establishes the parameter thresholds (i>14°, σ>600 km/s, ξ1<1200) and the 19% DPE rate among ZTF variability-selected AGN that the BASS rate is compared to.","marker":"Ward et al. 2024"},{"why":"Provides the classic identification of double-peaked emitters and prior samples with which several BASS DPEs overlap.","marker":"Eracleous & Halpern 1994"},{"why":"Provides the SDSS quasar DPE rate (3.6%) that serves as the low end of the comparison range.","marker":"Strateva et al. 2003"},{"why":"Supplies the BASS broad-line measurements and virial mass estimates used for comparisons.","marker":"Mejía-Restrepo et al. 2022"},{"why":"Supplies the stellar velocity dispersion measurements from which intrinsic BH masses are derived.","marker":"Caglar et al. 2023"},{"why":"Provides the disk-wind model prediction that disk-dominated BLR emission should appear at low accretion rates.","marker":"Elitzur & Ho 2009"},{"why":"Defines the hard X-ray selected parent sample and the BASS AGN catalog used for selection.","marker":"Koss et al. 2022b"}],"fun_headline_variants":["1 in 5 X-ray AGN show double-peaked broad lines","Double-peaked AGN: heavier, slower accretors in hard X-ray sample","BASS: Disk emission marks 21% of broad-line AGN","Disk lines bias black hole masses in 21% of AGN"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The classification rests on the assumption that the disk model plus fixed parameter thresholds and a subjective visual inspection cleanly separates double-peaked disk emitters from AGN whose broad line asymmetries come from outflows.","fun_headline_variants_meta":{"raw":{"variants":["1 in 5 X-ray AGN show double-peaked broad lines","Double-peaked AGN: heavier, slower accretors in hard X-ray sample","BASS: Disk emission marks 21% of broad-line AGN","Disk lines bias black hole masses in 21% of AGN"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000676,"raw_usage":{"total_tokens":3214,"prompt_tokens":1222,"completion_tokens":1992,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":838,"completion_tokens_details":{"reasoning_tokens":1912}},"tokens_in":838,"tokens_out":1992,"duration_ms":18162,"temperature":1.0,"reasoning_tokens":1912,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T19:27:17.827437+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"If an independent, reproducible spectral classifier—applied to the same 343 continuum-subtracted spectra without human review—recovered a DPE fraction far from 21% (or failed to reproduce the 0.4 dex mass and 0.3 dex Eddington ratio gaps), the paper's central claim would be undercut. A direct observational check is to take multi-epoch rms spectra of a random subset of the 272 non-DPE AGN; if double-peaked variable components appear in fewer than about 5% of them, the inference that many hidden disk emitters contaminate the non-DPE sample would be weakened.","supporting_citations":[],"review_version":1}