{"id":"2157ad45-4347-411c-a574-e3b06f891a29","arxiv_id":"2502.02751","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Variability-selected AGN in dwarf galaxies host lower-mass black holes and lower-luminosity point sources than spectroscopically selected AGN, with hosts showing pseudobulges and disks.","lead":"Using Hubble Space Telescope images, the authors dissect the structures of eight dwarf galaxies whose centers flicker, a sign of small black holes. They find these variability-selected black holes are lighter and dimmer than those found by traditional spectroscopic surveys, suggesting the census of black holes in small galaxies is incomplete.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The AIC model-selection criterion in §3.1 is non-standard and internally inconsistent; because point-source and bulge parameters are taken from the selected GALFIT models, the central mass and luminosity comparisons rest on an unreliable selection step.","rationale":"The reader's weakest_assumption focused on the extrapolation of the Schutte et al. (2019) and Reines & Volonteri (2015) scaling relations below their calibration range, an acknowledged astrophysical uncertainty. That is a legitimate concern about the BH-mass half of the claim, but the paper already cautions readers about it. The AIC definition in §3.1 is a more fundamental and unacknowledged problem. As written, AIC = 2k − log(χ2_ν) cannot select the models reported in Table 2: because χ2_ν is a reduced chi-square over tens of thousands of pixels, adding a component changes log(χ2_ν) by an amount far smaller than the 2k penalty, so the criterion is strongly biased toward simpler models. The presence of PSF components in the reported best fits is thus inconsistent with the stated rule, implying the actual selection was either a different, undocumented criterion or a subjective choice. Since the point-source luminosities and bulge masses are inputs to the comparison with Reines et al. (2013) and Kimbrell et al. (2021, 2023), the central claim is not securely grounded until this is resolved. The proposed re-analysis is straightforward because the imaging and GALFIT setup are described; it would settle whether the reported models, and hence the headline conclusions, survive a correct model-selection step. The reader already reached CONDITIONAL, and this concern reinforces, rather than overturns, that verdict: the paper is a valuable observational study, but it needs a corrected and reproducible model-selection analysis before the central claim can be accepted.","tokens_in":16014,"tokens_out":10453,"duration_ms":103907,"concrete_test":"Re-run the GALFIT model selection on the eight F110W images using the standard AIC = χ2 + 2k (or BIC) with the same model set and 200×200 cutouts. Compare the best models to Table 2; for any galaxy where the selected model changes (a PSF added or removed, or a bulge/disk component changed), recompute the point-source luminosity and bulge mass, and redo the comparisons in Figures 8 and 9. If the sample median luminosities and BH masses shift by more than the quoted errors, the central claim is not robust to the model-selection step.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In §3.1, AIC is defined as AIC = 2k − log(χ2_ν). For least-squares fitting the correct AIC is χ2 + 2k (up to a constant), or equivalently 2k + χ2_ν·dof. The published formula compresses the goodness-of-fit term logarithmically. On a 200×200 pixel cutout (≈4×10^4 dof), adding a PSF changes χ2_ν by at most a few percent, so −log(χ2_ν) changes by ≪ 2k; the criterion therefore almost always selects the model with the fewest components. Yet Table 2 reports PSF components in five of eight galaxies, so the results are incompatible with the stated selection rule. Either the formula is a typo and the real decision rule is undocumented, or the reported models were selected by eye; in both cases the PSF magnitudes in Table 3 and bulge masses in Table 4 are not reproducible from the method as written. The central claim — that variability-selected dwarf AGNs have lower point-source luminosities and lower BH masses than spectroscopy-selected dwarfs — is built on these measurements. This is an internal inconsistency, not merely a calibration choice, and it can be checked directly.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes HST/WFC3 imaging of eight dwarf galaxies that host AGN candidates selected via optical photometric variability in Baldassare et al. (2020). The authors use GALFIT to decompose the galaxies into Sérsic bulges, disks, nuclear components, and point sources, finding that regular hosts are best fit by pseudobulges rather than classical bulges. They estimate black hole masses from bulge-mass and total-stellar-mass scaling relations, obtaining a range of ~10^3.6–10^6.6 M_sun, and compare the point-source luminosities and BH masses with samples selected via optical spectroscopy (Reines et al. 2013; Kimbrell et al. 2021, 2023). The central claim is that variability selection can uncover lower-mass BHs and lower-luminosity AGN than optical spectroscopic selection.","tokens_in":16244,"tokens_out":3324,"duration_ms":33021,"significance":"If the central claim holds, this is an important step for understanding the occupation fraction of intermediate-mass black holes and the biases of different AGN selection techniques in dwarf galaxies. The paper presents a rare, homogeneous HST dataset and makes a direct comparison of variability-selected versus spectroscopically selected AGN hosts. The authors are appropriately cautious about the extrapolation of BH scaling relations, which is a strength. However, the reliability of the point-source measurements and the resulting mass/luminosity comparisons depends on a GALFIT model-selection step whose stated criterion appears internally inconsistent; this must be resolved before the central claim can be accepted.","major_comments":[{"comment":"The AIC definition as AIC = 2k − log(χ2_ν) is not the standard Akaike Information Criterion for least-squares fitting, which is (up to a constant) χ2 + 2k, or equivalently 2k + χ2_ν · dof. For the 200×200 pixel cutout used here, dof is ~4×10^4, so adding a PSF changes χ2_ν by at most a few percent; the log term in Eq. (3) would then vary by much less than the 2k penalty, and the criterion would almost always select the model with the fewest components. Yet Table 2 reports PSF components in five of the eight galaxies. This is an internal inconsistency: either Eq. (3) is a typo and the actual decision rule is undocumented, or the models were selected by eye rather than by the stated AIC. Because the point-source magnitudes and bulge parameters from these selected models feed the luminosity and mass comparisons that constitute the main claim, the selection step must be made reproducible and correct.","section":null},{"comment":"The text states that fits were obtained for all galaxies except NSA 156688, which is edge-on, and that 'four of them are fit with point sources.' Both statements conflict with Table 2, which lists a F110W fit for NSA 156688 (with a PSF magnitude of 22.50) and shows PSF entries for five galaxies (67333, 51928, 124554, 104881, 156688). Please clarify the sample of galaxies that enter the morphological and point-source analyses, and ensure the text and table agree.","section":null},{"comment":"The black hole masses are estimated by extrapolating scaling relations calibrated on more massive galaxies down to total stellar masses of ~10^7 M_sun and bulge masses that yield log M_BH ≈ 3.6–6.6. The authors state this extrapolation should be treated with caution, but the central comparison with the Reines et al. (2013) sample (Fig. 9) relies entirely on these extrapolated values. The paper should quantify how sensitive the claimed lower-mass result is to plausible changes in the low-mass slope or normalization of the relations, for example by recomputing the comparison using the upper/lower scatter bounds on log M_BH or by using an alternative relation. As written, the conclusion that variability selection 'finds lower mass black holes' is not robust to the acknowledged possibility that the relations change at low mass.","section":null},{"comment":"The evidence that the bright point sources are AGN rather than nuclear star clusters is presented as internally contradictory. Figure 5 shows that the point-source luminosities are not significantly brighter than the predictions for an NSC, while Figure 7 (and the text) states that the inferred stellar masses of the point sources are 'over massive' given the galaxy stellar mass, leading to the conclusion that the emission is not entirely stellar. These two statements are in tension: if the sources are underluminous for an NSC, they should not appear overmassive unless the mass-to-light ratio is anomalous. The paper needs either a model that reconciles these findings (e.g., a composite NSC+AGN scenario with quantitative predictions) or a clearer explanation of why the color-based mass estimate is preferred over the luminosity-based comparison. Without a firm classification of the point sources as AGN, the luminosity comparison in Fig. 8 is not necessarily a comparison of AGN luminosities.","section":null}],"minor_comments":[{"comment":"The entries in Table 4 for log M_BH are labeled 'bulge' and 'total' but the text says Eq. (5) is used only for the four galaxies with regular morphology and Eq. (6) for irregular galaxies; Table 4 lists values for both columns for several galaxies, which is confusing. Please clarify which column was used for the final BH mass estimates per object.","section":null},{"comment":"The redshifts in Table 1 include NSA 67333 at z=0.0020 with a distance of 11.2 Mpc, which is inconsistent with a Hubble-flow distance at that redshift; the distance appears to come from a peculiar-velocity model. Please state the distance source or note that distances are not purely from redshifts.","section":null},{"comment":"The sentence 'The two galaxies that are not fit with point sources are star forming and have no clear photometric center' does not match Table 2, where three galaxies (124477, 88260, 57867) lack PSF entries. Please reconcile the count.","section":null},{"comment":"The text uses 'varaiations' and 'over massive'—these should be corrected to 'variations' and 'overmassive'.","section":null},{"comment":"The definition of α_ox in Eq. (7) is missing a minus sign if it is intended to follow the standard convention α_ox = 0.3838 log(f_2keV / f_2500) = -0.3838 log(f_2500 / f_2keV). Please check the sign convention and the direction of the offset in Fig. 4.","section":null},{"comment":"The appendix note for NSA 51928 says 'The PSF is off nuclear,' but the main text states all point sources are 'central.' Please clarify the spatial offset and its implications for the AGN interpretation.","section":null}],"recommendation":"major_revision","confidential_remarks":"The paper addresses a timely question and the HST dataset is valuable. The main concern is not the scientific goal but the reproducibility of the model-selection step: the AIC formula appears to contradict the reported results, and the internal counts of PSF fits and successful fits are inconsistent. These issues are fixable but require reanalysis or careful documentation. The BH mass extrapolation is a separate, acknowledged weakness; although the authors are candid, the central claim depends on it and should be stress-tested quantitatively before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is the first morphological study of variability-selected dwarf AGN hosts, and the HST data plus the comparison to spectroscopically selected samples make it worth reading. The authors decompose eight galaxies with GALFIT, estimate black hole masses from scaling relations, and compare point-source luminosities and masses to BPT- and broad-line-selected dwarfs. They are also appropriately careful about the point-source/star cluster ambiguity and about extrapolating the scaling relations below their calibration range.\n\nThe main problem is in the model selection. Equation 3 defines AIC = 2k − log(χ2_ν). For the 200×200 pixel cutout used here, that is effectively thousands of degrees of freedom, and the log term changes by a small amount when a PSF is added. The criterion would almost always select the model with the fewest components. Table 2, however, reports PSFs in five of eight galaxies and multiple components in several. So either the formula is a typo and the actual decision rule is undocumented, or the models were chosen by eye. In both cases the PSF magnitudes and bulge masses that drive the central mass/luminosity comparison are not reproducible from the method as written. This is fixable, but it has to be fixed before the central claim can be trusted.\n\nThere are two smaller issues. The text and abstract say black hole masses range from 10^3.7 to 10^6.6 M_sun, but the maximum in Table 4 is 10^5.73. That mismatch needs to be reconciled. And the conclusions say the authors 'demonstrate' lower masses and luminosities; with eight galaxies and 0.55–0.68 dex scatter in the scaling relations, 'suggest' is more honest. The scaling-relation extrapolation itself is acknowledged and flagged, which I credit.\n\nWhat is genuinely new is the selection channel. Variability selection finds AGN that BPT diagnostics miss, and the result that these point sources are fainter than spectroscopically selected AGN but brighter than inactive dwarfs is a useful data point for IMBH occupation fractions. The morphology result—pseudobulges rather than classical bulges—extends earlier work to this sample.\n\nThis paper deserves a serious referee. The question matters, the data are real, and the flaws are identifiable and fixable. I would send it out with a specific request: document the actual model-selection procedure, correct Eq. 3, fix the mass-range inconsistency, and soften the conclusions to match the evidence. The audience is the dwarf AGN and IMBH community.","headline":"First morphological study of variability-selected dwarf AGN hosts with useful HST data, but the model-selection criterion as written contradicts the reported fits, and the central mass/luminosity comparison rests on that step.","tokens_in":16832,"tokens_out":5145,"would_cite":true,"duration_ms":46285,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Photometric variability in dwarf galaxies exposes active galactic nuclei with estimated black hole masses of roughly $10^{3.6}$ to $10^{6.6}$ solar masses, lower on average than those found by optical spectroscopy.","keywords":["dwarf galaxies","active galactic nuclei","intermediate-mass black holes","photometric variability","galaxy morphology","Hubble Space Telescope","black hole scaling relations","nuclear star clusters"],"falsifier":"Dynamical mass measurements of these nuclei, for example from spatially resolved molecular gas kinematics, would settle the claim: if any nucleus comes out above roughly $10^6$ solar masses, the conclusion that variability selection finds lower-mass black holes than spectroscopy would fail. A complementary check is deep optical spectroscopy looking for broad hydrogen-$\\alpha$ or high-ionization lines; their absence would put the AGN interpretation in doubt.","tokens_in":15789,"feed_emoji":"🕳️","tokens_out":12419,"duration_ms":107381,"temperature":0.7,"pith_summary":"Most dwarf-galaxy black hole searches use optical spectroscopy, a method that can miss active galactic nuclei (AGN) whose emission lines are diluted by star formation or absent because the galaxy has little gas. This paper instead studies eight dwarf galaxies whose AGN were found through photometric variability, meaning their brightness changes over time, using Hubble Space Telescope images in three bands. Light-profile fits yield estimated black hole masses of roughly $10^{3.6}$ to $10^{6.6}$ solar masses, lower on average than dwarf AGN found by spectroscopy, and the nuclear point sources are brighter than ordinary star clusters yet dimmer than spectroscopically selected AGN. The authors conclude that variability selection can uncover a population of lower-mass, lower-luminosity AGN in dwarf galaxies that spectroscopic surveys systematically miss.","feed_headline":"Variability finds smaller black holes in dwarf galaxies","feed_subtitle":"Hubble images of eight galaxies suggest black holes near 10^4 solar masses, below spectroscopically found AGN.","key_machinery":"The argument is carried by photometric variability as an AGN selector that does not depend on emission-line ratios, combined with two-dimensional light-profile decomposition of the Hubble images into disk, bulge, unresolved point-source, and nuclear components. Black hole masses are not measured directly but inferred from two scaling relations: the bulge-stellar-mass to black-hole-mass relation calibrated down to dwarf galaxies, and the total-stellar-mass to black-hole-mass relation for local AGNs. The nature of each point source is tested by comparing its near-infrared, optical, and ultraviolet luminosities and colors with stellar population models and with nuclear star cluster scaling relations, while X-ray observations of the same galaxies provide an independent check.","core_discovery":"The paper's central discovery is a population offset: active dwarf galaxies identified by optical variability host estimated black hole masses of $10^{3.6}$ to $10^{6.6}$ solar masses, with nuclear point sources about forty times dimmer than dwarf AGN selected by the standard emission-line BPT diagram (an emission-line test for AGN activity) and about ten times brighter than point sources in non-active dwarfs. The host galaxies are morphologically heterogeneous: three of eight are irregular, and the regularly structured ones are best fit by pseudo-bulges rather than classical bulges. Five of eight central point sources have colors inconsistent with pure star clusters, and three galaxies have X-ray detections consistent with AGN activity, supporting the interpretation that these are genuine active nuclei. On the paper's reading, these are low-luminosity AGN that spectroscopy misses, either because star formation dilutes their line emission or because the galaxies are too gas-poor to show standard AGN line ratios.","pith_inferences":["If the population offset holds up, the local occupation fraction of black holes in dwarf galaxies may be higher than spectroscopic surveys imply, because variability catches nuclei that leave no AGN signature in line ratios.","A decisive extension would be a dynamical mass measurement for one of these nuclei, for example from spatially resolved molecular gas kinematics; a mass above roughly $10^6$ solar masses would overturn the lower-mass conclusion.","The five point sources with red, cluster-inconsistent colors could be followed up with mid-infrared spectroscopy or deeper X-ray observations to confirm the AGN interpretation independently of the scaling relations.","Applying the same imaging analysis to larger variability-selected samples, such as those expected from wide-field time-domain surveys, could map how bulge and disk structure correlate with black hole mass across the $10^3$ to $10^6$ solar mass range."],"forward_implications":["Spectroscopic AGN searches in dwarf galaxies are incomplete at the low-luminosity end; variability monitoring can recover a population they miss.","Black hole occupation fractions inferred from spectroscopy alone may underestimate how many low-mass galaxies harbor intermediate-mass black holes.","The prevalence of pseudo-bulges among the regularly structured hosts supports the idea that low-mass black holes grow through secular, disk-driven processes rather than major mergers.","The intermediate brightness and cluster-inconsistent colors of the point sources imply that these nuclei contain both AGN and stellar cluster light, so separating the two requires multi-band, high-resolution follow-up.","Wide-field time-domain surveys can build much larger samples of variability-selected dwarf AGN and test how host morphology correlates with black hole mass."],"supporting_citations":[{"why":"Parent photometric variability survey that originally identified the eight AGN candidates; supplies the sample.","marker":"Baldassare et al. (2020)"},{"why":"Calibrates the bulge mass–black hole mass relation down to dwarf galaxies; used for the four regularly structured objects.","marker":"Schutte et al. (2019)"},{"why":"Provides the total stellar mass–black hole mass relation for local AGNs; used for the irregular galaxies.","marker":"Reines & Volonteri (2015)"},{"why":"Spectroscopically selected dwarf AGN sample whose broad hydrogen-alpha black hole masses are the direct comparison for the lower-mass claim.","marker":"Reines et al. (2013)"},{"why":"BPT-selected dwarf AGN hosts whose morphologies and point-source luminosities are compared with the variability-selected sample.","marker":"Kimbrell et al. (2021)"},{"why":"Non-active dwarf galaxies providing the point-source luminosity baseline that lies below the variability-selected nuclei.","marker":"Kimbrell et al. (2023)"},{"why":"Chandra X-ray observations of the same galaxies; detections in three objects support the AGN interpretation of the point sources.","marker":"Messick et al. (2023)"},{"why":"Nuclear star cluster scaling relations and mass-to-light ratios used to test whether the point sources are consistent with stellar clusters.","marker":"Georgiev et al. (2016)"},{"why":"Provides the light-profile modeling code used for all galaxy decompositions.","marker":"Peng et al. (2010)"}],"fun_headline_variants":["Variability reveals dwarf black holes spectroscopy misses","Flickering dwarfs host black holes too small for spectroscopy","Variable AGN in dwarfs host low-mass black holes","Dwarf galaxies that flicker host lighter black holes","Variability spots faint AGN in dwarf galaxies"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The black hole masses are not measured; they are estimates from scaling relations calibrated on more massive galaxies and extrapolated down to about $10^4$ solar masses, and if those relations change shape at low masses the claimed contrast with spectroscopically selected AGN collapses.","fun_headline_variants_meta":{"raw":{"variants":["Variability reveals dwarf black holes spectroscopy misses","Flickering dwarfs host black holes too small for spectroscopy","Variable AGN in dwarfs host low-mass black holes","Dwarf galaxies that flicker host lighter black holes","Variability spots faint AGN in dwarf galaxies"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000862,"raw_usage":{"total_tokens":3754,"prompt_tokens":975,"completion_tokens":2779,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":591,"completion_tokens_details":{"reasoning_tokens":2703}},"tokens_in":591,"tokens_out":2779,"duration_ms":19360,"temperature":1.0,"reasoning_tokens":2703,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T11:14:56.284767+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Dynamical mass measurements of these nuclei, for example from spatially resolved molecular gas kinematics, would settle the claim: if any nucleus comes out above roughly $10^6$ solar masses, the conclusion that variability selection finds lower-mass black holes than spectroscopy would fail. A complementary check is deep optical spectroscopy looking for broad hydrogen-$\\alpha$ or high-ionization lines; their absence would put the AGN interpretation in doubt.","supporting_citations":[{"cited_title":"J., Reines, A","cited_arxiv_id":null,"evidence_quote":"Non-active dwarf galaxies providing the point-source luminosity baseline that lies below the variability-selected nuclei."},{"cited_title":"2023, The Astrophysical Journal, 953, 18, doi: 10.3847/1538-4357/acdc90","cited_arxiv_id":null,"evidence_quote":"Chandra X-ray observations of the same galaxies; detections in three objects support the AGN interpretation of the point sources."}],"review_version":1}