{"id":"ff966ee9-8778-46a4-9ce6-0897b5385c7d","arxiv_id":"2502.09759","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"Mrk 477's line-of-sight obscuration varies frequently, and across 27 AGN the probability of detecting column-density variability rises from ~20% at time gaps under 10 days to ~60-70% beyond 5 years.","lead":"Fifteen X-ray observations of the Seyfert 2 galaxy Mrk 477 show that the gas blocking our view of its black hole changes on timescales as short as two weeks. Across a 27-source sample, the chance of seeing such obscuration changes grows from about 20% for data taken less than 10 days apart to 60-70% for gaps longer than five years.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The sample-level timescale trend in §5.2 may be driven by Mrk 477's many correlated observation pairs; no source-level uncertainty or exclusion test is given.","rationale":"The authors' main new astrophysical conclusion is the timescale dependence of obscuration variability in a 27-source sample. That conclusion is built from pairwise incompatibility counts in which each source contributes up to many correlated pairs. Mrk 477 alone contributes ~105 of ~350 pairs, and its uneven sampling creates exactly the bin-to-bin spikes seen in Fig. 6 (orange). The paper's own caveat—that the pair-counting approach is invalid for a single source—applies in attenuated form to the full sample, because one source can dominate certain bins. The reader's identified weakest assumption, the time-constant torus reflection parameters, is real but is mitigated by the Appendix B consistency test and by agreement between borus02 and UXCLUMPY; it would bias all epochs, not specifically the timescale trend. The pair-counting issue is unaddressed quantitatively and directly controls the headline trend, so it is the more load-bearing concern. A simple exclusion/weighting test would settle it, and the paper should be required to pass it before the sample-level claim is accepted. The Mrk 477-specific variability result may survive, so the verdict remains conditional rather than reject.","tokens_in":24268,"tokens_out":10171,"duration_ms":108889,"concrete_test":"Recompute the left panel of Fig. 6 excluding Mrk 477, and add source-level jackknife or binomial error bars to the pair fractions in each Δt bin. If the >5 yr bin drops below ~40% or its error bars overlap the <10 d bin, the sample-level timescale trend is not robust to the dominant source's correlated pairs.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing weakness is the statistical construction of the timescale trend (§5.2, Fig. 6), not the tied reflection parameters—the latter is at least tested in Appendix B. The 105 Mrk 477 pairs are roughly one third of all ~350 pairs and are not independent: a single anomalous epoch generates many variable pairs at the same Δt, and the long-timescale bins (>2000 d) are heavily populated by comparisons between the 2010–2015 archival observations and the 2021–2023 monitoring campaigns of this one source. The paper acknowledges this for Mrk 477 alone and states it \"does not appear to dominate\" the full sample, but no quantitative test is provided; Fig. 6 has no error bars and no clustering correction. Since the central new result is the increase from ~20% at Δt<10 d to ~60–70% at >5 yr, the claim would be materially weakened if the trend disappears when Mrk 477 is excluded or when sources rather than pairs are treated as the sampling unit.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a multi-epoch X-ray analysis of the Seyfert 2 galaxy Mrk 477, combining five archival observations (XMM-Newton, NuSTAR, Chandra) with two dedicated monitoring campaigns (five Swift-XRT and five NuSTAR epochs). All 15 epochs are fit simultaneously with two self-consistent torus models (borus02 and UXCLUMPY), with reflection parameters and the photon index tied across epochs while the line-of-sight column density NH,los and a cross-normalization are free per epoch. The authors report NH,los variability between multiple epoch pairs, including a claimed ~14-day change between two Swift-XRT observations, and estimate that 42 of the 105 Mrk 477 observation pairs (40%) are variable at 90% confidence. Combining these 105 pairs with ~245 pairs from 26 other sources analyzed in the authors' earlier papers, they find that the fraction of variable pairs increases from ~20% at Δt < 10 days to ~60–70% at Δt > 5 years. They interpret this timescale trend as evidence against BLR clouds being the sole origin of obscuration variability, unless the clouds are implausibly large.","tokens_in":24545,"tokens_out":6258,"duration_ms":58647,"significance":"If the sample-level timescale trend is robust, it provides a valuable empirical constraint on the origin and location of obscuring material in AGN, complementing small-number eclipsing-event studies such as Markowitz et al. (2014). The Mrk 477 monitoring campaign itself is a useful addition, and the paper has genuine strengths: it uses two independent torus models that cross-check the NH,los measurements, it explicitly tests the tied-reflection-parameter assumption by fitting archival+Swift and NuSTAR-only subsets separately (Appendix B), and it presents the fitted spectral components in detail. The main significance claim, however, is the trend in Fig. 6, and that claim currently rests on a statistical treatment of non-independent observation pairs that needs to be demonstrated to be robust.","major_comments":[{"comment":"The quantity 42/105 = 40% is used both as the observed pair fraction and, after assuming all pairs are independent Bernoulli draws, to compute the probabilities 78% (3 observations) and 95% (4 observations). Pairs sharing an epoch are strongly correlated (a single observation appears in up to 14 pairs), so the effective sample size is far below 105 and the naive binomial variance underestimates the uncertainty. Please provide a clustering-aware estimate (e.g., bootstrap by epoch or by source) or explicitly label these numbers as descriptive fractions without formal probability claims.","section":"§5.2, Eq. (4) and derived probabilities"}],"minor_comments":[{"comment":"Typographical errors: §2.1 heading 'Data Analisys', §2.3 'arund', §1 'as longs as years', §4 'desnsity', §5.4 'perfomed', and the text near Eq. (1) 'Bremmstrahlung' should be corrected.","section":"Throughout"},{"comment":"The binned fractions would be much easier to interpret with binomial error bars (e.g., Wilson intervals), and the Mrk 477-only curve should either be overlaid in a separate panel or have its contribution to each bin indicated so that the reader can see the weighting directly.","section":"Fig. 6"},{"comment":"The notation in the table footnote ('NH,inst.,num.') does not match the table's column headers (e.g., NH,swift1, Cnus3); please align the notation used in the text, the table, and the footnote.","section":"Table 4"},{"comment":"The statement that 'this statistical approach is not valid when looking at a single source' for Mrk 477 is puzzling because the same pair-counting approach is applied to the full sample in the same section; please clarify why pair non-independence is deemed acceptable for the sample-level estimate but not for the single-source estimate.","section":"§5.2"},{"comment":"The in-text citation 'Gravity Collaboration et al. 2023' does not match the reference list entry 'Gravity Collaboration, Amorim, A., Bourdarot, G., et al. 2023'; please standardize the citation format.","section":"Fig. 7 caption"},{"comment":"The abstract's '40%, 78% and 95%' probabilities should carry an explicit caveat that they assume independent pairs; as written, they appear as exact probabilities rather than as descriptive fractions derived from a correlated dataset.","section":"Abstract and §5.2"}],"recommendation":"major_revision","confidential_remarks":"The paper is part of the authors' ongoing program of spectral fitting of a parent sample, and the sample-level trend in §5.2 relies almost entirely on pairs from their own previously published analyses. This is not a flaw per se, but it means the trend is a single-group measurement; an independent reanalysis by another group would strengthen confidence. The editor may also wish to note that the shortest-timescale variability claim is more fragile than the abstract suggests, as it depends on a pair of intervals that touch at the boundary for one of the two models."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe monitoring data are worth having, and the compiled trend is new, but the sample-level result is the soft spot. The paper reports 15 epochs on Mrk 477, finds NH,los varying down to ~2 weeks, and adds these to 26 sources to claim a rising probability of variability with time. That trend in Fig. 6 is the main new result.\n\nThe source analysis itself is careful. Two self-consistent torus models give consistent NH,los patterns across epochs. The tied-reflection assumption is tested rather than ignored: Appendix B compares fits on archival+Swift data against the NuSTAR campaign and shows the torus parameters are consistent. The text is explicit that the single-source probability calculation is not a proper statistical statement, which shows honesty.\n\nThe soft spot is the pair-counting statistics. Mrk 477 contributes about a third of the ~350 pairs, and those pairs are strongly clustered: one anomalous epoch generates many variable pairs at the same Delta-t. The paper acknowledges this but does not test whether the trend survives excluding Mrk 477, nor does it treat the source as the sampling unit, and Fig. 6 has no error bars. That matters because the headline claim is exactly this trend. The 14-day variability detection also rests on a single pair whose 90% intervals just touch; that is suggestive, not conclusive, though the longer-timescale variability is on firmer ground.\n\nThe BLR-cloud discussion is speculative in the good sense — they explicitly say a complex cloud distribution cannot be ruled out. But the 300-500 day crossing time is derived from a visual flattening of the same trend they are trying to explain, so treat that section as interpretation.\n\nIn short: this deserves a serious referee, but I would not accept the paper as-is. The statistical construction of the timescale trend needs a robustness check — exclude Mrk 477, or bootstrap over sources, or at least show the trend with sources as the unit. That is a tractable revision, not a fatal flaw.","headline":"New monitoring data and a plausible compiled trend, but the timescale claim needs a source-level robustness test before it carries the paper.","tokens_in":25073,"tokens_out":2695,"would_cite":true,"duration_ms":28407,"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":"Fifteen X-ray observations of Mrk 477 show its obscuring gas column changes in as little as two weeks, and pooled with 26 other sources the chance of catching obscuration variability rises from about 20% below 10 days to 60-70% beyond 5…","keywords":["obscuration variability","Seyfert 2 galaxy","active galactic nuclei","X-ray spectroscopy","torus model","broad-line region clouds","column density","Mrk 477"],"falsifier":"Observe a moderately obscured (Compton-thin) active galaxy at five or more epochs spread over a decade with simultaneous soft and hard X-ray coverage, fitting each epoch with the torus reflection parameters free; if the covering factor, average column density, or photon index must vary to produce acceptable fits, the tied-parameter assumption fails and the reported variability fractions would need revision. For the BLR hypothesis, search for a fully sampled cloud eclipse lasting longer than about 300-500 days; if such an eclipse has an inferred cloud size below $10^{15}\\,{\\rm cm}$, it would contradict the paper's claim that only torus-scale material can explain long-timescale variability.","tokens_in":24089,"feed_emoji":"🔭","tokens_out":14322,"duration_ms":120532,"temperature":0.7,"pith_summary":"Mrk 477 is a nearby galaxy whose actively accreting supermassive black hole is seen through a veil of gas, and this paper asks whether that veil changes over time. It compiles 15 X-ray observations taken over 13 years with four telescopes and fits them simultaneously with self-consistent torus models, finding that the line-of-sight gas column oscillates between about $1.5$ and $7\\times10^{23}\\,{\\rm cm}^{-2}$ and can change on timescales as short as two weeks. The paper also pools these data with 26 previously studied sources, roughly 350 observation pairs in total, and finds that the chance of seeing a change in obscuration grows from about 20% for observations less than 10 days apart to 60-70% for separations beyond five years. If this is right, single snapshot comparisons routinely underestimate how often obscured active galaxies are changing, and a small number of epochs is usually enough to catch variability. The authors argue the timescale trend points to torus-scale material as the main source of long-term obscuration changes, with broad-line-region clouds confined to the shortest separations.","feed_headline":"Obscuring gas around a black hole shifts in as little as 2 weeks","feed_subtitle":"Fifteen X-ray snapshots show Mrk 477's gas veil changes in ~14 days; variability odds rise to 60-70% after 5 years.","key_machinery":"The argument runs on simultaneous X-ray spectral fitting with two self-consistent torus models: borus02, a homogeneous-reflector model, and UXCLUMPY, a clumpy-cloud model. All 15 spectra are fit together with the torus reflection parameters (covering factor, average torus column density, cloud-distribution width, inclination) and the photon index tied across epochs, while the line-of-sight column density $N_{\\rm H,los}$ and a cross-normalization constant are free in each epoch; this design is what separates obscuration changes from intrinsic luminosity changes. The variability claim is then carried by a pair-counting statistic: across the 105 epoch pairs of Mrk 477 and about 350 pairs in the full 27-source sample, a pair counts as variable when its two $N_{\\rm H,los}$ values are inconsistent at 90% confidence, and the fraction of such pairs is computed as a function of the time separation $\\Delta t$.","core_discovery":"The paper's central claim is that Mrk 477, a Seyfert 2 active nucleus viewed through obscuring gas, shows frequent line-of-sight column density variability, with $N_{\\rm H,los}$ moving between roughly $1.5\\times10^{23}$ and $7\\times10^{23}\\,{\\rm cm}^{-2}$ and changes appearing on timescales down to about two weeks. Of the 105 pairs of observations, 42 are inconsistent at 90% confidence with a single column density, so the probability that two randomly drawn epochs disagree is about 40%; the chance of catching at least one variable pair rises to 78% with three observations and 95% with four. Combined with 26 other sources analyzed with the same method, the variable-pair fraction increases from about 20% at $\\Delta t<10$ days to roughly 60-70% at $\\Delta t>5$ years, while the average column change in variable pairs stays near $30{-}40\\times10^{22}\\,{\\rm cm}^{-2}$ except for the shortest separations, where it drops to about $5\\times10^{22}\\,{\\rm cm}^{-2}$. The paper further claims that this timescale dependence is difficult to explain if all obscuration variability comes from broad-line-region clouds, because above a cloud's crossing time the variability probability should become flat; matching the observed rise would require such clouds two to three orders of magnitude larger than previous measurements, so the data favor variability arising from torus-scale structure, with broad-line-region clouds possibly dominating only the $\\Delta t<10$ day regime.","pith_inferences":["Beyond the paper: the rising variable-pair fraction with time implies that two-epoch 'changing-look' classifications separated by years will overstate how quickly or dramatically a source is transitioning, because the same source would likely show a different column if sampled again.","Beyond the paper: if the $\\Delta t<10$ day regime is indeed dominated by broad-line-region clouds, rapid-cadence monitoring of Compton-thin Seyfert 2 galaxies should find frequent small-amplitude eclipses with $\\Delta N_{\\rm H,los}\\sim5\\times10^{22}\\,{\\rm cm}^{-2}$, a testable prediction the paper does not itself make.","Beyond the paper: applying the same pair-count statistic to the growing X-ray archive of other obscured active galaxies would test whether the >5-year variable-pair fraction continues to rise, and would show how much Mrk 477's dense monitoring skews the current sample.","Beyond the paper: the tied-reflection assumption could be stress-tested on sources with stronger reflection than Mrk 477, where torus geometry changes would leave clearer spectral signatures; if reflection variability turns out to be common, the derived $N_{\\rm H,los}$ histories of all 27 sources would need to be re-fit."],"forward_implications":["With only two observations of a genuinely variable obscured active galaxy, the chance of seeing the variability is about 40%, so archival two-epoch comparisons systematically underreport obscuration variability.","Monitoring campaigns that obtain three or four epochs have a 78-95% chance of catching at least one variable pair, making modest campaigns a cost-effective way to identify variable sources.","Across the 27-source sample, the variable-pair fraction grows from about 20% at separations under 10 days to 60-70% beyond 5 years, implying that long-term obscuration variability is common among Compton-thin Seyfert 2 nuclei.","The average column change in variable pairs is roughly constant at $30{-}40\\times10^{22}\\,{\\rm cm}^{-2}$ across timescales, except below 10 days where it drops to about $5\\times10^{22}\\,{\\rm cm}^{-2}$, suggesting a distinct short-timescale population of low-column clouds.","The timescale dependence favors torus-scale material as the dominant source of obscuration variability, with broad-line-region clouds contributing mainly at the shortest separations."],"supporting_citations":[{"why":"Supplies the borus02 homogeneous-torus model used to fit reflected emission and derive torus geometry.","marker":"Baloković et al. 2018"},{"why":"Supplies the UXCLUMPY clumpy-torus model whose cloud distribution was calibrated using observed column density distributions and eclipsing-event frequencies.","marker":"Buchner et al. 2019"},{"why":"Provides the largest previous sample of fully observed eclipsing events and the reference for the trend of increasing eclipse probability at longer timescales.","marker":"Markowitz et al. 2014"},{"why":"Early evidence that X-ray obscuration changes support an inhomogeneous torus, the physical picture this work quantifies.","marker":"Risaliti et al. 2002"},{"why":"Defines the parent sample of Compton-thin obscured AGN with NuSTAR and Swift-BAT detections from which Mrk 477 is drawn, and first flagged Mrk 477 as variable.","marker":"Zhao et al. 2021"},{"why":"Established the same simultaneous torus-model fitting methodology and analyzed earlier sources in the 27-source sample.","marker":"Torres-Albà et al. 2023"},{"why":"Applied the same methodology to additional parent-sample sources, providing the comparison data used in the pair-count statistics.","marker":"Pizzetti et al. 2024"},{"why":"Measured short BLR cloud eclipses with small column changes, the reference for cloud sizes and short-timescale variability amplitudes.","marker":"Maiolino et al. 2010"},{"why":"Measured broad-line-region sizes used to test whether BLR clouds could be large enough to explain variability at longer timescales.","marker":"Gravity Collaboration et al. 2023"}],"fun_headline_variants":["Mrk 477's gas veil shifts in just 2 weeks","Seyfert 2's obscuring gas changes in ~14 days","Gas column varies in 2 weeks around Mrk 477","Swift and NuSTAR track rapid gas changes in Mrk 477"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the reflecting structure around the black hole and the intrinsic spectral shape stay completely unchanged across the 13 years, so every spectral difference between epochs must be assigned to the gas directly in front of the source; if the reflector itself evolves, some of that evolution would be misread as line-of-sight obscuration variability.","fun_headline_variants_meta":{"raw":{"variants":["Mrk 477's gas veil shifts in just 2 weeks","Seyfert 2's obscuring gas changes in ~14 days","Gas column varies in 2 weeks around Mrk 477","Swift and NuSTAR track rapid gas changes in Mrk 477"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000251,"raw_usage":{"total_tokens":1677,"prompt_tokens":1186,"completion_tokens":491,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":802,"completion_tokens_details":{"reasoning_tokens":415}},"tokens_in":802,"tokens_out":491,"duration_ms":5816,"temperature":1.0,"reasoning_tokens":415,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T20:34:05.996689+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Observe a moderately obscured (Compton-thin) active galaxy at five or more epochs spread over a decade with simultaneous soft and hard X-ray coverage, fitting each epoch with the torus reflection parameters free; if the covering factor, average column density, or photon index must vary to produce acceptable fits, the tied-parameter assumption fails and the reported variability fractions would need revision. For the BLR hypothesis, search for a fully sampled cloud eclipse lasting longer than about 300-500 days; if such an eclipse has an inferred cloud size below $10^{15}\\,{\\rm cm}$, it would contradict the paper's claim that only torus-scale material can explain long-timescale variability.","supporting_citations":[{"cited_title":"Hydrogen Column Density Variability in a Sample of Local Compton-Thin AGN II","cited_arxiv_id":"2403.06919","evidence_quote":"Applied the same methodology to additional parent-sample sources, providing the comparison data used in the pair-count statistics."}],"review_version":1}