{"id":"1c9cd3f9-314c-4ab4-b023-6ff099b637b4","arxiv_id":"2501.08076","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A catalogue of 291 hard X-ray sources in the Lockman Hole, interpreted as candidate reflection-dominated Compton-thick AGN, with density estimates and a consistency check against a one-megasecond XMM-Newton spectrum.","lead":"Using X-ray data from the SRG/eROSITA telescope, this paper picked out 291 unusually hard X-ray sources in the Lockman Hole sky region, about 5 percent of all distant galaxies detected there, and called them candidate heavily obscured black holes. The catalogue gives observatories a target list for follow-up checks that could reveal a rare hidden stage of black hole growth.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'Compton-thick' label is not directly supported by the fitted column densities: even the 1 Ms XMM-Newton prototype yields NH≈1.4e22 and Fe EW≈0.19 keV, so the hard spectra could equally be intrinsic rather than reflection-dominated CT AGN.","rationale":"The paper is best read as a catalogue of hard-spectrum X-ray sources selected from eROSITA Lockman Hole data, with a careful but ultimately speculative physical interpretation. The reader's CONDITIONAL verdict is appropriate. My stress-testing identifies a more specific vulnerability than the reader's low-count concern: even in the regime of high signal-to-noise (the 1 Ms XMM-Newton observation of srcid 1430), the data do not exhibit the canonical signatures of a Compton-thick AGN. The line-of-sight column density is NH ≈ 1.4×10^22 cm^-2 and the iron line EW is only 0.19 keV. The paper's UXCLUMPY model reconciles these with a Compton-thick inner ring, but this geometry is inferred, not directly measured. The hard continuum (Γ ≈ 0.4–1.0) is the sole selection signature, and it is not unique to reflection-dominated CT AGN. The stacking analysis in §4 does mitigate the low-count noise concern for Category 1 sources: the stacked spectra remain hard and unabsorbed. However, a systematic background hard tail at faint fluxes could in principle affect both individual and stacked spectra, although the contrast with Category 2 stacks (Γ≈1.9) makes a global background bias less plausible. I therefore do not recommend REJECT: the catalogue is honestly presented, the methods are transparent, and the authors flag the NH puzzle. But the density lower limit (0.35 deg^-2) should be understood as a density of hard-spectrum candidates, not confirmed CT AGN, until broadband follow-up shows a reflection hump or strong iron line. The proposed test—a model-comparison on the existing XMM data—would settle whether reflection is actually required. If it is not, the term 'CT AGN candidates' is premature and the paper would need to be revised to 'hard-spectrum AGN candidates.'","tokens_in":25088,"tokens_out":16443,"duration_ms":168237,"concrete_test":"Re-fit the public 1.04 Ms XMM-Newton spectrum of srcid 1430 (Section 4.3) with an absorbed power law, a power law plus Compton reflection (pexrav or mytorus), and the UXCLUMPY model, comparing C-statistics to assess whether the reflection component is statistically required. If the reflection normalization is consistent with zero (ΔC-stat < 9 for the extra parameters), then a hard intrinsic power law describes the data as well as the reflection-dominated CT model, and the 'CT AGN candidate' interpretation is not unique. As a secondary check, stack the eROSITA spectra of all Category 1 sources and measure the Fe Kα line; a reflection-dominated spectrum should show EW ≳ 0.5 keV.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that the 291 hard-spectrum sources constitute a sample of Compton-thick AGN candidates—depends on interpreting their hard, absorption-free X-ray spectra as reflection-dominated emission from Compton-thick material. The paper's most direct evidence for this interpretation is the UXCLUMPY spectral modelling (Table 2). However, even for the prototype Category 1 source srcid 1430 (SRGe J105348.6+573032) with a 1.04 Ms XMM-Newton observation (Section 4.3), the fitted line-of-sight column density is NH = (1.43 ± 0.06) × 10^22 cm^-2, two orders of magnitude below the Compton-thick threshold of 10^24 cm^-2, and the measured Fe Kα equivalent width is 0.19^+0.08_-0.10 keV, about one fifth of the ~1 keV expected in a reflection-dominated spectrum. The combined Category 1 eROSITA spectra give NH = (2.0 ± 0.2) × 10^22 cm^-2. The paper proposes a plausible but speculative resolution: a Compton-thick inner ring with covering factor CTKcover > 0.49 and a transparent outer torus (Section 5.2). Yet the hard observed photon indices (Γ ≈ 0.4–1.0 in Table A1) are equally consistent with intrinsically hard, unobscured AGN, and the paper does not test whether a simple hard power law with no reflection component fits the data comparably. Because the 'reflection-dominated CT AGN' classification underpins the Category 1 definition and the sky-density lower limit in Section 5.4, this ambiguity propagates to the paper's headline density. The authors explicitly acknowledge the low-NH puzzle (Section 5.2), so the issue is a limitation of the evidence, not an internal inconsistency.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a search for reflection-dominated Compton-thick AGN (CT AGN) candidates in the SRG/eROSITA Lockman Hole survey, selecting sources whose 90% upper bound on the photon index from a simple power-law fit lies below a fiducial threshold of Gamma0 = 1.3 in the 0.3-8 keV band. From 6528 extragalactic sources, 291 candidates are found, of which 81 have no significant intrinsic absorption (Category 1), 49 show significant absorption (Category 2), and 161 lack reliable redshifts (Category 3). The authors publish bright (37 sources) and faint (254 sources) catalogues, perform stacking analysis to look for redshift evolution, fit combined eROSITA spectra with the UXCLUMPY model, and compare one Category 1 source (srcid 1430) with ~1 Ms of XMM-Newton data. They estimate the fraction and sky density of reflection-dominated CT AGN candidates, finding a lower limit on the sky density of 0.35 deg^-2 at a limiting flux of 1.5e-14 erg/s/cm^2.","tokens_in":25494,"tokens_out":4620,"duration_ms":44754,"significance":"The paper's main product is a well-defined, carefully selected catalogue of hard-spectrum X-ray sources, built with a transparent methodology (W-statistic fitting, 90% confidence intervals, a bright subsample, and XMM-Newton cross-checks). If the interpretation that these sources are reflection-dominated CT AGN is correct, the sample would be among the first of its kind from eROSITA and would provide a valuable target list for follow-up at higher energies. The authors are honest about the limitations of their data and propose a physically motivated, albeit speculative, model to reconcile the low fitted column densities with a CT interpretation. However, the direct evidence for the CT nature is weak: the UXCLUMPY fits yield NH ~ 1e22 cm^-2, the Fe K-alpha equivalent width in the prototype is only ~0.2 keV, and the Category 1 definition relies on non-detection of absorption rather than a positive detection of a reflection component. The significance of the paper therefore rests heavily on the interpretation, which is not yet secure.","major_comments":[{"comment":"The definition of Category 1 as 'reflection-dominated CT AGN candidates' is based on the absence of a statistically significant intrinsic NH in phabs*zphabs*zpowerlw fits. As the authors themselves state in Sect. 3.2, the NH upper limits for many Category 1 sources are unconstraining, often in the 1e22-1e23 cm^-2 range, and adding intrinsic absorption to faint sources increases the degeneracy and the photon-index error. Non-detection of absorption at these low count levels is therefore not equivalent to a reflection-dominated spectrum. The paper should assess the probability that a normal (non-CT) AGN with an intrinsically hard photon index, or an absorbed source with poor photon statistics, would satisfy the Category 1 criteria. Without such a contamination estimate, the physical interpretation of the Category 1 sample and the density lower limit derived from it in Sect. 5.4 are not robust.","section":"Sect. 4.1, Table 2"},{"comment":"The deep XMM-Newton spectrum of the prototype Category 1 source srcid 1430, used as an anchor for the physical interpretation, yields NH = (1.43 ± 0.06) × 10^22 cm^-2 and an Fe K-alpha equivalent width of 0.19 keV (Sect. 4.3), both far below the canonical values expected for reflection-dominated CT AGN (NH > 1e24 cm^-2, EW ~ 1 keV). The UXCLUMPY fits to the combined eROSITA Category 1 spectra also give NH = (2.0 ± 0.2) × 10^22 cm^-2 (Table 2). The paper proposes an inner Compton-thick ring with a transparent outer torus to explain these values, but this scenario is explicitly speculative (Sect. 5.2). The manuscript should directly test whether a simple hard power law (without any reflection component) can describe the available spectra as well as UXCLUMPY, and report the statistical comparison. Without such a test, the assignment of the 'Compton-thick' label is not supported by the data.","section":"Sect. 5.2"},{"comment":"The lower limit on the sky density of reflection-dominated CT AGN (0.35 deg^-2) is computed using the number of Category 1 sources above the flux threshold. Because the Category 1 selection does not establish that the sources are reflection-dominated (see comments above), this number is not a lower limit on the CT AGN density. The paper should either reframe this quantity as the density of 'hard-spectrum, unabsorbed sources' or provide evidence that the Category 1 population is dominated by true CT AGN, for example by showing that reflection models are statistically required over simple power-law models in the stacked spectra. As presented, the lower limit in Sect. 5.4 inherits the assumption that the absence of absorption implies reflection dominance, which is not justified.","section":"Sect. 3.2, Sect. 5.4"}],"minor_comments":[{"comment":"The table header spells the UXCLUMPY parameter as 'CTKover' while the text and Fig. 7 caption use 'CTKcover'; please make the notation consistent.","section":"Table 2"},{"comment":"There is a typo in the sentence 'Indeed, it is is well known, that broadband spectral modelling...' — 'is is' should be 'is'.","section":"Sect. 5.2"},{"comment":"The text states 'the area of the survey is 28.65deg' and later 'deg^2'; please use deg^2 consistently for the area units.","section":"Sect. 5.4"},{"comment":"The sentence 'For 81 sources, no statistically significant intrinsic absorption was detected, suggesting that the data do not require absorption turnover' is slightly ambiguous; 'do not require' would be clearer as 'do not require an absorption turnover'.","section":"Sect. 3.2"},{"comment":"The statement that catalogues 'will be made publicly available via the VizieR system after the publication of this work' is fine for a preprint, but the paper should clarify whether the catalogue is provided as supplementary material at submission or only after acceptance.","section":"Data Availability"}],"recommendation":"major_revision","confidential_remarks":"The catalogue and selection methodology are valuable and the paper is generally careful with statistics. The main concern is terminological overreach: the 'Compton-thick' and 'reflection-dominated' labels are used in the title, abstract, and conclusions despite the fitted column densities being two orders of magnitude below the CT threshold and the absence of a positive reflection detection. The authors do acknowledge this in Sect. 5.2, but the framing of the central results still overstates the evidence. I would recommend asking the authors to either (a) moderate the terminology (e.g., 'hard-spectrum AGN candidates' instead of 'Compton-thick AGN candidates') throughout the title/abstract, or (b) add a dedicated test that discriminates reflection from intrinsically hard power-law models, even in a stacked spectrum. The current version is defensible as a candidate selection, but the physical claims need either stronger evidence or more cautious wording."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is best read as a catalogue paper, not a Compton-thick confirmation. The 291 hard-spectrum sources are a genuinely new sample from the eROSITA Lockman Hole field, and the authors are careful about selection, errors, and their own limitations. But the data do not actually show Compton-thick absorption: the UXCLUMPY fits give NH ~ 1-2 x 10^22 cm^-2, and the one source with a 1 Msec XMM observation has Fe K alpha EW of 0.19 keV, about a fifth of the ~1 keV expected in reflection-dominated CT AGN. The hard spectra could just as well be intrinsically hard, unobscured AGN; the paper doesn't test that alternative.\n\nThat's the main soft spot, and the authors acknowledge it (Sect. 5.2). They propose a speculative geometry — an inner Compton-thick ring with covering fraction >0.49 and a transparent outer torus — that is consistent but not required by the spectra. The Category 1 classification also rests on non-detection of absorption in low-count spectra, which the authors note is often unconstraining. So the density lower limit in Sect. 5.4 inherits that uncertainty.\n\nWhat the paper does well: the selection method is transparent and reproducible (Gamma < 1.3 at 90% confidence, W-stat fits, bright subsample of 37 objects with >100 counts). The XMM-Newton cross-check for one source is a genuine external anchor, even if it doesn't settle the CT question. The stacking analysis and redshift-binned fits are reasonable. The catalogue itself — 291 objects with photometric/spectroscopic redshifts and spectral parameters — will be useful for follow-up and for census work, independent of whether every source is actually CT.\n\nI'd send it to peer review; it's a solid catalogue with an overreaching but explicitly hedged interpretation. A referee should ask the authors to add a simple test of an intrinsically hard power law against the reflection model, and to soften the language from \"CT AGN candidates\" to \"hard-spectrum candidates requiring follow-up\" unless they can justify the interpretation more directly.\n\nFor me: would cite it as a candidate sample, with the caveat. Not sure I'd bring it to reading group — it's fairly specialized, but if you're working on obscured AGN, it's worth a look.","headline":"Useful catalogue of 291 hard-spectrum sources, but the 'Compton-thick' label is a step beyond what the data support.","tokens_in":26074,"tokens_out":2569,"would_cite":true,"duration_ms":25187,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["95.85.Nv","98.54.Cm"],"model":"deepseek-v4-flash","headline":"The Lockman Hole survey yields 291 candidate Compton-thick AGN, about 5% of its extragalactic X-ray sources.","keywords":["galaxies: active","galaxies: nuclei","X-rays: galaxies","surveys","catalogues","Compton-thick AGN","reflection-dominated","eROSITA"],"falsifier":"Deep hard X-ray observations of the nine brightest Category 1 sources above 10 keV with instruments such as NuSTAR or SRG/ART-XC: if these spectra show no Compton hump near 20–30 keV and no strong neutral iron line at 6.4 keV, the reflection-dominated interpretation would be refuted.","tokens_in":24897,"feed_emoji":"🔭","tokens_out":10855,"duration_ms":88949,"temperature":0.7,"pith_summary":"Using the SRG/eROSITA Lockman Hole survey, the authors search for active galactic nuclei whose X-ray spectra are so hard that the emission is likely dominated by reflection from Compton-thick material rather than direct light. They select sources whose 90% confidence upper bound on the photon index—the spectral slope of the continuum—lies below 1.3, far below the typical value of about 2 for unobscured AGN. This selects 291 candidates, roughly 5% of the field's extragalactic X-ray sources, which the authors present as the eROSITA sample of Compton-thick AGN candidates. Dividing the sample by redshift availability and the significance of intrinsic absorption, they identify 81 sources with no significant absorption as the best reflection-dominated candidates and estimate a lower limit of 0.35 deg$^{-2}$ on their sky density.","feed_headline":"291 Compton-thick AGN candidates found in Lockman Hole","feed_subtitle":"About 5% of the field's X-ray sources are too hard for normal AGN: a hidden population.","key_machinery":"The selection rests on the photon index cut: each source is fitted with an absorbed power-law model ($\\texttt{phabs*powerlaw}$), and those whose 90% upper uncertainty bound on $\\Gamma$ lies below the fiducial value $\\Gamma_0=1.3$ are flagged as CT AGN candidates. The underlying physics is that unsaturated Comptonization in a hot corona, which yields the standard $\\Gamma\\approx2$ spectra of unobscured AGN, cannot produce such shallow slopes; a hard reflection-dominated continuum must come from reprocessing by Compton-thick material. A second fit with $\\texttt{phabs*zphabs*zpowerlw}$ adds intrinsic absorption at the source redshift, allowing separation of Category 1 (no significant absorption) from Category 2 (absorption present) sources. For the bright subsample, combined spectra are then fitted with the UXCLUMPY unification model, whose parameters (covering fraction of a Compton-thick inner ring and angular thickness of a clumpy absorber) provide a geometric picture that distinguishes the two categories.","core_discovery":"The paper identifies 291 extragalactic X-ray sources in the Lockman Hole whose spectra in the 0.3–8 keV band are anomalously hard, with the 90% upper bound on the power-law photon index $\\Gamma$ below $1.3$. Because unobscured type I AGN typically show $\\Gamma \\approx 2$, such shallow slopes are uncharacteristic of direct coronal emission; the authors interpret them as dominated by emission reflected from Compton-thick material with column density $\\gtrsim 10^{24}$ cm$^{-2}$ (i.e. reflection-dominated Compton-thick AGN). Fitting models with and without intrinsic absorption, 81 sources (Category 1) show no statistically significant intrinsic absorption and are designated the best candidates; 49 sources (Category 2) show absorption consistent with mildly obscured AGN; and 161 sources (Category 3) lack reliable redshifts. The authors construct bright (37 sources) and faint (254 sources) catalogues, derive a lower limit of $0.35$ deg$^{-2}$ for the sky density of these candidates, and show that physically motivated UXCLUMPY fits to combined eROSITA spectra are fully consistent with a ~1 Msec XMM-Newton observation of one Category 1 source, the Type 2 galaxy SRGe J105348.6+573032.","pith_inferences":["If the hard-photon-index selection is an efficient proxy for reflection dominance, the implied fraction of heavily obscured AGN among the X-ray-selected population may be higher than some earlier estimates, suggesting that a larger share of supermassive black hole growth is hidden behind Compton-thick material.","The authors note that several Category 1 sources with high X-ray hardness are optically classified as Type 1 AGN; this hints that optical and X-ray classifications probe different parts of the circumnuclear geometry, an interpretation not established by the paper itself.","The stacked low-redshift Category 1 spectra have tight upper limits on intrinsic absorption (~$10^{20}$–$10^{21}$ cm$^{-2}$) and very shallow photon indices; future detection of a soft excess or warm absorber in these objects would reveal that their hard continua are not purely reflection-dominated."],"forward_implications":["If the selection is correct, reflection-dominated Compton-thick AGN form a rare subpopulation (~5%) among extragalactic X-ray sources in the Lockman Hole, with a lower limit on sky density of 0.35 deg$^{-2}$ at a limiting flux of $1.5\\times10^{-14}$ erg s$^{-1}$ cm$^{-2}$.","The published bright and faint catalogues serve as target lists for hard X-ray follow-up with observatories such as NuSTAR and SRG/ART-XC to confirm the Compton-thick nature.","Applying the same photon-index cut to the full eROSITA all-sky survey should uncover many more such candidates, enabling a more complete census of heavily obscured supermassive black hole growth.","The UXCLUMPY fitting suggests that Category 1 and Category 2 sources occupy distinct regions of parameter space (high inner-ring covering fraction vs. clumpy absorber thickness), a distinction that can be tested with broadband observations."],"supporting_citations":[{"why":"Established the approach of selecting reflection-dominated CT AGN candidates via anomalously hard photon indices, which the paper adapts to eROSITA data.","marker":"Tozzi et al. 2006"},{"why":"Applied a similar hard photon index selection to XMM-Newton data, providing a precedent for the method and a comparison population.","marker":"Georgantopoulos et al. 2009"},{"why":"Provided the UXCLUMPY spectral model used to fit combined eROSITA and XMM-Newton spectra in this work.","marker":"Buchner et al. 2019"},{"why":"Demonstrated that unsaturated Comptonization produces soft photon indices, supporting the argument that hard spectra require a reflection-dominated origin.","marker":"Sunyaev & Titarchuk 1980"},{"why":"Supplied predicted sky densities of CT AGN at similar flux limits, used for comparison with the measured lower limit.","marker":"Akylas et al. 2012"},{"why":"Provided the X-ray luminosity function and number counts used to estimate the expected sky density of heavily obscured (N_H > 10^25 cm^-2) AGN.","marker":"Ananna et al. 2019"},{"why":"The 4XMM DR13 catalogue used to cross-match eROSITA sources with XMM-Newton observations, enabling spectral comparisons.","marker":"Webb et al. 2020"},{"why":"The mytorus model used to illustrate theoretical spectra of obscured and reflection-dominated AGN in Fig. 1.","marker":"Yaqoob 2012"}],"fun_headline_variants":["291 reflection-dominated AGN candidates found in Lockman Hole","Rare X-ray ghosts: 291 CT AGN candidates in Lockman Hole","Anomalous hard X-rays expose 291 hidden AGN in Lockman Hole","Lockman Hole survey uncovers 291 Compton-thick AGN candidates","About 5% of Lockman Hole X-ray sources are reflection-dominated AGN"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The central assumption is that a 90%-confidence photon index upper bound below 1.3 is produced by reflection from Compton-thick material and not mimicked by other effects such as an absorption turnover, a warm absorber, a soft excess, or low-count statistical fluctuations.","fun_headline_variants_meta":{"raw":{"variants":["291 reflection-dominated AGN candidates found in Lockman Hole","Rare X-ray ghosts: 291 CT AGN candidates in Lockman Hole","Anomalous hard X-rays expose 291 hidden AGN in Lockman Hole","Lockman Hole survey uncovers 291 Compton-thick AGN candidates","About 5% of Lockman Hole X-ray sources are reflection-dominated AGN"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000351,"raw_usage":{"total_tokens":2013,"prompt_tokens":1140,"completion_tokens":873,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":756,"completion_tokens_details":{"reasoning_tokens":774}},"tokens_in":756,"tokens_out":873,"duration_ms":7338,"temperature":1.0,"reasoning_tokens":774,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:30:11.540051+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Deep hard X-ray observations of the nine brightest Category 1 sources above 10 keV with instruments such as NuSTAR or SRG/ART-XC: if these spectra show no Compton hump near 20–30 keV and no strong neutral iron line at 6.4 keV, the reflection-dominated interpretation would be refuted.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Applied a similar hard photon index selection to XMM-Newton data, providing a precedent for the method and a comparison population."}],"review_version":1}