{"id":"d1b554ba-e7ad-4c56-a428-35d986a30a6e","arxiv_id":"2501.12520","paper_version":3,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"The authors present 121 BL Lac-type blazar candidates selected from eROSITA, WISE, and radio data, eight of which are newly classified as blazars.","lead":"This paper uses the eROSITA X-ray survey, WISE infrared data, and radio catalogs to identify 121 candidate TeV-emitting blazars of the BL Lac type. The list, including eight objects never before classified as blazars, gives gamma-ray telescopes a prioritized set of targets for follow-up observations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline 121-candidate list is not reproducible from the paper as written: the abstract and Section 4 say 121, while Figure 3, Table 1, and Figure 4 say 135, and the alpha_IR-X cut is stated as >=0 in Section 2 but <=1.0 in criterion V.","rationale":"The reader's weakest_assumption (alpha_R-IR threshold not independently validated) is a legitimate secondary concern about purity, but the more load-bearing problem is that the paper's own text cannot fix the actual selection. A candidate catalog is only as good as its definition; a 14-object discrepancy in the headline number and contradictory alpha_IR-X bounds mean the central claim is not currently checkable. This is not a question of interpretation or external calibration; it is an internal inconsistency. I therefore agree with the CONDITIONAL verdict rather than moving it: if the public code produces one self-consistent set of criteria and counts, the catalog may be sound, and the alpha_R-IR calibration issue can then be tested with a held-out validation sample. The concrete test above is cheap because the code and catalogs are public, and it directly determines whether the 121-candidate claim survives.","tokens_in":21444,"tokens_out":15811,"duration_ms":162635,"concrete_test":"Run the publicly available THC-CAT code (https://github.com/cassiemetzger/THC-CAT) from raw eRASS1, AllWISE, NVSS, SUMSS, and FIRST inputs, applying criteria I-IX exactly as written and also with alpha_IR-X >= 0 instead of <= 1.0; compare the final count and the row count of the machine-readable Table 1 against 121 and 135. If neither choice reproduces the published list, inspect the 'External process to apply infrared cuts' step in Figure 3 to identify the missing criterion.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is the candidate list, so the selection pipeline must be unambiguous. It is not. The abstract and Section 4 report 121 THCs, but Figure 3's caption says 'The resulting outcome of 135 objects', Table 1's caption says 'The complete list of 135 objects', and Figure 4's caption says 'The 135 candidate sources identified in our study'. Figure 3 also routes candidates through an undefined 'External process to apply infrared cuts' that is not among criteria I-IX. Separately, the infrared-to-X-ray spectral-index selection is stated inconsistently: Section 2 says candidates must have alpha_IR-X >= 0, while criterion V (Section 3) and Figure 2 require alpha_IR-X <= 1.0. These are different cuts; the <=1.0 version excludes the known TeV HBLs TXS 1515-273, PKS 0301-243, PKS 1440-389, and 1ES 1215+303 listed in Table A1 with alpha_IR-X > 1. If the pipeline actually used one bound, then either the published count or the stated criterion is wrong; if it used neither, the list was not produced by the described algorithm. Until this is reconciled, the 121-candidate claim cannot be independently checked.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents a search for new TeV-emitting high-synchrotron-peaked BL Lac (HBL) candidates built on the eRASS1 X-ray catalog. The selection pipeline (Section 3, criteria I-IX and Figure 3) requires WISE detections with magnitude limits and colors near the locus of the 56 known TeV HBLs (a color box, a 0.21-mag distance to the fitted line, and a brightness cut), an X-ray counterpart in eRASS1 within 5 arcseconds, a point-like X-ray morphology, |b| ≥ 10 degrees, a radio counterpart in NVSS/FIRST/SUMSS/RADIO, and effective spectral-index cuts in the αR-IR versus αIR-X plane (αR-IR ≤ 0.43, and an αIR-X bound that the text states inconsistently). The final list is ranked by 0.2-2.3 keV X-ray flux for TeV follow-up. The paper claims 121 candidates (with 23 lacking ROSAT 2RXS counterparts and 11 not previously identified as blazars in 4LAC/BZCAT/3HSP, 8 of which are genuinely new), but the figure and table captions state 135 objects. The written pipeline, the figures, and the tables disagree on several counts, so the exact algorithm and the candidate list must be reconciled before the central claim can be verified.","tokens_in":21797,"tokens_out":26627,"duration_ms":224731,"significance":"This is a candidate-list paper with real value if the list is clean: only 56 TeV HBLs are currently known, so a flux-ranked sample of roughly 120 new candidates, including 23 objects below the ROSAT sensitivity and 8 objects not previously classified as blazars, is a useful resource for IACT follow-up with H.E.S.S., VERITAS, MAGIC, and CTAO. The paper earns credit for shipping the full selection code publicly (GitHub), stating the nine criteria explicitly, and providing falsifiable predictions that pointed TeV observations can confirm or reject. The radio-to-infrared spectral-index criterion (αR-IR ≤ 0.43) is a sensible, physically motivated addition to the established WISE-color plus X-ray selection of Massaro et al. (2011, 2013), Arsioli et al. (2015), and Chang et al. (2017), and the cross-checks against 3HSP, 4LAC, BZCAT, 2FHL, 3FHL, and 1CGH give useful context on how much of the list is genuinely new. The main risk is that the thresholds are calibrated on the known TeV HBL sample without a quantified contamination estimate, so the classification purity of the list is not yet demonstrated.","major_comments":[{"comment":"Section 2 states the infrared-to-X-ray selection as αIR-X ≥ 0 and asserts that four known TeV HBLs in the eRASS1 footprint (PKS 1440-389, 1ES 1215+303, TXS 1515-273, PKS 0301-243) have αIR-X < 0, and Section 3's narrative text repeats 'αIR-X ≥ 0 are selected (V)'. Enumerated criterion V and the Figure 2 caption, however, require αIR-X ≤ 1.0, and Table A1 lists exactly these four sources with αIR-X = 1.05-1.18, i.e., greater than 1, not less than 0. The candidate table itself corroborates the ≤ 1.0 version (all listed αIR-X values are ≤ 1.00). The Section 2 discussion therefore contains a sign error and a wrong cut direction, and because a reader implementing the stated ≥ 0 cut would retain the four known TeV HBLs and obtain a different candidate list, the published algorithm does not uniquely determine the result.","section":"Section 2 vs. Section 3 criterion V, Figure 2, Table A1"},{"comment":"The headline count is stated inconsistently: the abstract, Section 4 ('We have identified 121 TeV HBL Candidates'), and Section 5 ('121 promising objects') report 121 candidates, whereas the Figure 3 caption ('The resulting outcome of 135 objects'), the Table 1 caption ('The complete list of 135 objects'), and the Figure 4 caption ('The 135 candidate sources identified in our study') report 135. Because the candidate list is the central result of the paper, the authors must determine which number the public code and the machine-readable table actually yield, make every statement and caption agree with it, and re-check the derived quantities that depend on the total (e.g., the 19% ROSAT-undetected fraction in Section 5).","section":"Abstract, Section 4, Section 5 vs. Figure 3, Table 1, Figure 4 captions"},{"comment":"Figure 3 does not describe the same pipeline as criteria I-IX. The flow chart contains a box labeled 'External process to apply infrared cuts' (between 137 and 121 sources) that is never defined in the text, a TeVCAT-exclusion step that is absent from the criteria list, and an 'is extragalactic' condition folded into the extended-source filter that likewise has no written criterion. The step-by-step counts also disagree with the tables and text: Section 4 says 18 sources were discarded for lacking a radio counterpart, while Table A3 lists 36 entries and Figure 3 removes 36 (177→141); and Figure 3 removes 41 sources at the 'not extended and extragalactic' step (218→177) while Table A2 lists only 31 extended sources. The criteria list, Figure 3, and the appendix tables must be brought into agreement so that the selection can be reproduced from the paper alone.","section":"Section 3, Figure 3, Section 4, Table A3"},{"comment":"The radio-to-infrared cut αR-IR ≤ 0.43 is calibrated on the distribution of known TeV HBLs, ROMA-BZCAT BL Lacs, and FSRQs shown in Figure 4, but the paper provides no quantitative estimate of how many non-HBL sources pass the full set of criteria I-IX. Since the central claim is that the final list consists of TeV-emitting HBL candidates, the contamination rate directly affects the usefulness of the list for follow-up planning. I request a quantitative contamination test — for example, running the complete selection on the spectroscopically confirmed FSRQ population from ROMA-BZCAT or 4LAC and reporting the pass rate, and stating how many final candidates have spectroscopic classifications inconsistent with HBLs — in addition to the cross-checks against 3HSP, BZCAT, and 4LAC already provided in Section 5.","section":"Section 3, criterion IX, Figure 4"}],"minor_comments":[{"comment":"The text refers to 'THC J0825+016', but Table 4 lists J0824+016; the two designations must be aligned.","section":"Section 5 vs. Table 4"},{"comment":"The listed αIR-X = 0.24 disagrees with the value 0.75 in Table 1 for the same source with identical fluxes; one of the two entries is a typo.","section":"Table 3, row J0244-583"},{"comment":"The abstract's '11 that have not been previously associated with blazars' should be qualified, since Section 5 explains that 3 of the 11 appear in BROS, KDEBLLACS/WIBRaLS2, and the eROSITA blazar catalog, leaving 8 genuinely new candidates.","section":"Abstract vs. Section 5"},{"comment":"'a radio counterpart could no be identified' contains a typo ('no' should be 'not').","section":"Section 4, first paragraph"},{"comment":"'the our selection criterion' contains a duplicated article.","section":"Figure 2 caption"},{"comment":"SUMSS fluxes are at 843 MHz while NVSS and FIRST fluxes are at 1.4 GHz, and the text says the radio flux is 'preferentially measured at νR = 1.4 GHz' without stating whether νR is adjusted for SUMSS-only sources; this should be stated explicitly, as a fixed νR = 1.4 GHz introduces a small systematic offset in αR-IR for those objects.","section":"Equation 2 and Table 1 notes"},{"comment":"The argument that the eRASS1 chance-coincidence rate follows from the eFEDS rate with 'approximately 10 times' lower exposure should be made explicit (the rate scales with source density, not exposure directly), so that the claim of fewer than one expected spurious candidate is checkable.","section":"Section 4, coincident-rate estimate"},{"comment":"The 'distance of 0.21 from the best-fit line' is not defined; please state whether it is the perpendicular Euclidean distance in the color-color plane.","section":"Section 2, criterion III"},{"comment":"The statement that three candidates appear in the eROSITA blazar catalog relies on 'Hämmerich et al. (in prep)'; if that catalog is not yet public, this should be flagged or replaced with an available reference.","section":"Section 5, Hämmerich et al. (in prep)"}],"recommendation":"major_revision","confidential_remarks":"To the editor: the manuscript shows clear signs of a revision in which the text, figures, and tables were not synchronized (the 121/135 discrepancy, the sign error surrounding the αIR-X cut in Section 2, and the undefined 'External process' box in Figure 3 all point this way). The public code likely implements the intended pipeline, and the referee's recommendation is to require a full consistency pass in which the authors re-run the code, report the exact count and the number removed at each step, and align the criteria list, Figure 3, Tables 1/A2/A3, and the abstract. I would also ask the editor to make the contamination quantification for criterion IX part of the revision, since the classification claim depends on it. The scientific approach itself is standard for this field, and the candidate list, once pinned down, should be a serviceable resource for the TeV community."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a useful catalog paper, not a new framework. The genuinely new pieces are the eRASS1 selection, the radio-to-infrared alpha_R-IR cut that separates FSRQ-like contaminants, and 8 blazar classifications not in 4LAC/BZCAT/3HSP, plus 23 candidates missed by ROSAT. The source list, ranked by X-ray flux, is exactly what IACT observers want for follow-up. Credit where due: the selection criteria are enumerated, the code is public, and the discussion is honest about overlap with 3HSP and about sources discarded for extended X-ray or missing radio counterparts. That is the right way to present a candidate list.\n\nThe soft spots are real, though, and one is load-bearing. The abstract and Section 4 say 121 candidates; Figure 3, Table 1, and Figure 4 all say 135. That is not a cosmetic mismatch: the central deliverable is the list itself. The infrared-to-X-ray spectral index selection is also stated inconsistently: Section 2 requires alpha_IR-X >= 0, while criterion V and Figure 2 require alpha_IR-X <= 1.0. Table A1 lists four known TeV HBLs with alpha_IR-X roughly 1.05-1.18, which the text describes as <0. I cannot tell from the paper which cut the pipeline actually used. Since the code is public a referee could settle this quickly, but as written the 121-candidate claim cannot be independently checked. Figure 3 also routes candidates through an 'External process to apply infrared cuts' that is not one of criteria I-IX; that box needs a definition.\n\nThe circularity concern is fair but not disqualifying. The WISE line, 0.21 distance tolerance, magnitude cuts, and alpha_R-IR threshold are tuned to the 56 known TeV HBLs and to known BZCAT populations, with no independent validation sample. That means the candidate list should be treated as provisional, and the quasar-contamination rate is not measured. However, the goal here is to produce targets for follow-up, not a complete census; the ranking by X-ray flux is sensible and the 3HSP overlap gives some confidence. I would not call this a fatal flaw, but it should be stated more carefully.\n\nBottom line: this paper deserves a serious referee. It is a useful, honest catalog paper with two internal inconsistencies that must be fixed before publication. Send it to review, but require the authors to reconcile the counts and the alpha_IR-X definition, and to add an explicit validation statement for the tuned cuts.","headline":"Useful eROSITA-based HBL candidate list that deserves refereeing, but the paper as written cannot be reproduced because the candidate count and the alpha_IR-X cut are stated inconsistently.","tokens_in":22335,"tokens_out":4379,"would_cite":true,"duration_ms":43938,"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":"The deepest wide-area soft-X-ray survey to date yields 121 new TeV-emitting BL Lac blazar candidates.","keywords":["BL Lacertae objects","TeV gamma-ray astronomy","eROSITA","X-ray surveys","blazar candidates","high-synchrotron-peaked blazars","WISE infrared colors","radio-to-infrared spectral index"],"falsifier":"Pointed TeV observations of the X-ray-brightest candidates: if most show no >0.1 TeV emission while optical spectra show broad quasar emission lines, the radio-to-infrared cut is not separating as assumed. A cheaper check is spectra of the eight new candidates, where broad lines would mark them as quasars that slipped through.","tokens_in":21250,"feed_emoji":"🔭","tokens_out":10172,"duration_ms":88575,"temperature":0.7,"pith_summary":"This paper argues that the deepest wide-area soft-X-ray survey yet released, the eROSITA eRASS1 catalog, can substantially expand the known population of TeV-emitting BL Lac blazars. Cross-matching infrared and X-ray colors plus archival radio data yields 121 high-synchrotron-peaked BL Lac candidates, including 23 objects too faint for the earlier ROSAT survey and 8 that had never been flagged as blazars. The new piece is a radio-to-infrared spectral index cut that removes quasar-like objects whose infrared and X-ray properties mimic genuine TeV blazars. If pointed TeV observations confirm the candidates, the list gives current and future Cherenkov telescopes a ready queue of targets ranked by X-ray brightness.","feed_headline":"Deep X-ray survey flags 121 TeV-blazar candidates","feed_subtitle":"The eROSITA survey plus a radio-to-infrared color cut finds blazars too faint for ROSAT, ready for Cherenkov follow-up.","key_machinery":"The machinery is a nine-step filtering funnel applied to roughly 747 million AllWISE sources, of which 121 survive to the final list. The load-bearing new ingredient is the radio-to-infrared effective spectral index $\\alpha_{\\rm R-IR} = -\\log(F_R/F_{3.4\\,\\mu\\rm m})/\\log(\\nu_R/\\nu_{3.4\\,\\mu\\rm m})$, required to be at most 0.43: in the plane of $\\alpha_{\\rm R-IR}$ versus $\\alpha_{\\rm IR-X}$, flat-spectrum radio quasars and BL Lacs separate, and the cut rejects quasars whose synchrotron peak is in the infrared but whose hard X-ray spectrum mimics an HBL. The funnel also uses the WISE 3.4–4.6–12 $\\mu$m color-color strip calibrated on the 56 known TeV HBLs (distance no more than 0.21 from the line $y=0.84x-1.05$), a 5″ positional match to eRASS1, $\\alpha_{\\rm IR-X}\\ge 0$, non-extended X-ray emission, $|b|\\ge 10^\\circ$, and a radio counterpart. The surviving sources are ranked by X-ray flux because the paper treats the 0.2–2.3 keV flux as the practical predictor of TeV detectability.","core_discovery":"The central claim is a ranked candidate catalog rather than a detection: 121 high-frequency-peaked BL Lac objects selected from the eRASS1 X-ray source list, AllWISE infrared magnitudes, and NVSS/SUMSS/FIRST radio catalogs, ordered by their 0.2–2.3 keV flux because X-ray and TeV fluxes are empirically correlated in this class. The selection requires the WISE color-color strip occupied by gamma-ray blazars, an infrared-to-X-ray effective spectral index $\\alpha_{\\rm IR-X}\\ge 0$, point-like X-ray morphology, Galactic latitude $|b|\\ge 10^\\circ$, a radio counterpart within 5″, and the new radio-to-infrared cutoff $\\alpha_{\\rm R-IR}\\le 0.43$. Of the 121 candidates, 53 are already identified as BL Lacs in ROMA-BZCAT, 108 overlap the 3HSP catalog, 23 have no ROSAT counterpart, and 8 are new blazar candidates. The paper's case is that the depth and energy coverage of eRASS1, together with the radio-IR cut, is what brings these objects to the surface.","pith_inferences":["Editorial inference: the $(\\alpha_{\\rm R-IR}, \\alpha_{\\rm IR-X})$ plane could become a standard two-index diagnostic for classifying blazar SEDs in any future survey combining radio, infrared, and X-ray data.","Editorial inference: if even a few of the X-ray-brightest candidates are confirmed at TeV energies, it would validate the X-ray-flux-to-TeV-flux correlation in a regime below the ROSAT era and strengthen the case that X-ray depth is the most efficient path to new TeV blazars.","Editorial inference: optical spectroscopy of the eight new candidates is the quickest falsification check, since genuine BL Lacs have featureless optical continua while quasars show broad emission lines.","Editorial inference: applying the $\\alpha_{\\rm R-IR}$ cut to the full set of 56 known TeV HBLs, not just the 21 in the eRASS1 footprint, would give a quantitative false-negative rate for the cut, a number the paper does not report."],"forward_implications":["The 121-candidate list immediately supplies observation targets for VERITAS, MAGIC, HESS, and LST-1, with the X-ray-ranked tables telling observers which northern and southern sources to point at first.","Because 23 candidates were below the ROSAT detection threshold, the known TeV HBL census of 56 objects is incomplete in exactly the flux range eRASS1 now covers.","Eight candidates with no prior blazar association show that blazar catalogs built on shallower X-ray surveys can miss genuine members of the class.","The same $\\alpha_{\\rm R-IR}\\le 0.43$ cut can be reused on future eROSITA data releases to extend the search to the rest of the sky and to deeper exposure.","The 31 extended-X-ray rejects and 18 radio-undetected rejects are tabulated as secondary follow-up targets, including possible BL Lacs in cluster cores."],"supporting_citations":[{"why":"Supplies the eRASS1 eROSITA catalog: the X-ray counterparts, fluxes, and point-source classifications that drive the search.","marker":"Merloni et al. 2024"},{"why":"Supplies the AllWISE infrared catalog whose 3.4, 4.6, and 12 micron magnitudes define the infrared color cuts.","marker":"Cutri et al. 2013"},{"why":"Supplies NVSS 1.4 GHz fluxes and positions used for the radio counterpart requirement and the alpha_R-IR calculation.","marker":"Condon et al. 1998"},{"why":"Supplies SUMSS 843 MHz fluxes for southern radio counterparts used in the same index.","marker":"Mauch et al. 2013"},{"why":"Supplies FIRST radio fluxes as an additional radio matching catalog for the 5-arcsecond counterpart cut.","marker":"Becker et al. 1995"},{"why":"Establishes the WISE color-color strip where blazars live, the basis of the infrared color selection.","marker":"Massaro et al. 2011"},{"why":"Shows gamma-ray BL Lacs separate from flat-spectrum radio quasars in WISE infrared colors.","marker":"Massaro et al. 2012"},{"why":"Provides the 2WHSP selection method and the wider infrared search region that the paper follows.","marker":"Chang et al. 2017"},{"why":"Provides the 3HSP catalog used as the comparison benchmark, overlapping 108 of the 121 candidates.","marker":"Chang et al. 2019"},{"why":"Provides TeVCAT, the list of 56 TeV-detected HBLs used to calibrate the color and spectral-index cuts.","marker":"Wakely & Horan 2008"}],"fun_headline_variants":["eROSITA finds 121 TeV-blazar candidates","Deep X-ray survey spots 121 TeV-blazar candidates","eROSITA survey uncovers 121 TeV-blazar candidates","Radio-IR cut sharpens eROSITA TeV-blazar search to 121"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole list rests on the assumption that the radio-versus-infrared brightness cut, tuned to known blazars and quasars, separates faint unknown sources of the two kinds just as cleanly; no independent sample is used to measure how many quasar-like objects still pass.","fun_headline_variants_meta":{"raw":{"variants":["eROSITA finds 121 TeV-blazar candidates","Deep X-ray survey spots 121 TeV-blazar candidates","eROSITA survey uncovers 121 TeV-blazar candidates","Radio-IR cut sharpens eROSITA TeV-blazar search to 121"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000608,"raw_usage":{"total_tokens":2899,"prompt_tokens":1077,"completion_tokens":1822,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":693,"completion_tokens_details":{"reasoning_tokens":1747}},"tokens_in":693,"tokens_out":1822,"duration_ms":13311,"temperature":1.0,"reasoning_tokens":1747,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T17:07:05.857988+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Pointed TeV observations of the X-ray-brightest candidates: if most show no >0.1 TeV emission while optical spectra show broad quasar emission lines, the radio-to-infrared cut is not separating as assumed. A cheaper check is spectra of the eight new candidates, where broad lines would mark them as quasars that slipped through.","supporting_citations":[{"cited_title":"J., et al","cited_arxiv_id":null,"evidence_quote":"Supplies SUMSS 843 MHz fluxes for southern radio counterparts used in the same index."},{"cited_title":"P., & Horan, D","cited_arxiv_id":null,"evidence_quote":"Provides TeVCAT, the list of 56 TeV-detected HBLs used to calibrate the color and spectral-index cuts."}],"review_version":1}