{"id":"f6324945-3656-4730-ae0f-a548be4fe34a","arxiv_id":"2411.09690","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Only two of 51 core-collapse supernovae show late-time UV emission in Hubble data, and AT2018cow is brighter and far more compact, favoring an accreting central engine over supernova interaction.","lead":"The authors searched Hubble UV images of 51 nearby core-collapse supernovae taken two to five years after discovery and found only two possible detections, both previously known interacting supernovae. Comparing those measurements with the peculiar transient AT2018cow, they argue its steady late-time UV glow is too bright and too compact to be powered by supernova interaction, favoring an accreting central engine.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The interaction-exclusion conclusion hinges on the unverified single-blackbody interpretation of AT2018cow's late UV emission; if the UV source is not one blackbody photosphere, the compact-radius argument in §4.4 does not exclude interaction.","rationale":"The reader's weakest-assumption analysis correctly identifies the blackbody interpretation of AT2018cow's late UV emission as the load-bearing premise in §4.4. My independent reading agrees: the empirical sample comparison is carefully done and supports the statement that late-time UV detections in CCSNe are rare and preferentially associated with interacting events, but it does not by itself show that AT2018cow cannot be interacting. The decisive step is the radius argument, which depends on a single-blackbody SED with R~40 R_sun. That assumption is plausible but not independently tested in this paper; the authors themselves insert 'if it is a BB at these epochs', signalling the conditionality. A compact UV source could in principle arise in interaction, for example from a thin cooling shock or hotspot whose projected emitting area is far smaller than the overall CSM photosphere, so the radius argument needs direct SED-model testing. The manuscript's other strengths (sample construction, astrometric alignment, artificial-star upper limits, chance-alignment estimates, and consistency with two known interacting SNe) are real and independent support for the observational survey itself, but they do not remove the need to test the blackbody premise. Since the reader's CONDITIONAL verdict already reflects this unverified premise, no verdict change is needed; the condition should be discharged by the SED test above.","tokens_in":27385,"tokens_out":2967,"duration_ms":34720,"concrete_test":"Re-analyse the late-time SED of AT2018cow at the 713- and 1474-day epochs using all available HST F225W/F336W photometry together with contemporaneous optical/IR data. Fit three models: (a) a single blackbody, (b) a two-component blackbody plus power-law or line emission, and (c) an optically thin shock/nebular model. If model (b) or (c) gives a statistically comparable fit (e.g., Δχ² < 9 for two extra parameters) or if the inferred photospheric radius changes by more than an order of magnitude, the compact-radius argument in §4.4 fails and the interaction channel remains viable. This test uses only existing archival data and does not require new observations.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is that AT2018cow's late-time UV emission was not driven by interaction, based on two strands: (i) its brightness relative to a 17-SN subsample closer than AT2018cow, and (ii) the very small photospheric radius (~40 R_sun) reported in Inkenhaag et al. (2023). Strand (i) alone is weak: ATLAS17lsn is ~3 mag brighter than AT2018cow and ASASSN-17qp ~3 mag fainter, so AT2018cow is not atypical for interacting SNe; the '75 percent would be detected' argument only says a source as bright as AT2018cow would be recovered, not that interaction cannot produce it. Strand (ii) is therefore load-bearing. Section 4.4 invokes the radius with the caveat 'if it is a BB at these epochs', but the paper does not independently test that premise. If AT2018cow's late UV is a non-thermal component, a two-component SED, or emission from a compact shocked region embedded in a larger CSM interaction, then a small apparent blackbody radius does not rule out interaction. The premise is imported from earlier work and is exactly the assumption that would have to fail for the conclusion to change.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper uses HST/WFC3 F275W snapshot images of 51 nearby (z<0.065) core-collapse supernovae obtained 2–5 years after discovery to search for late-time UV emission. The authors perform Gaia/Pan-STARRS-based astrometric recalibration, DOLPHOT PSF photometry, artificial-star completeness experiments, and chance-alignment estimates. They find two point-source detections, ASASSN-17qp and ATLAS17lsn, both previously identified as interacting SNe, and place upper limits on the remaining 49 objects. Comparing the absolute UV magnitudes with AT2018cow, the authors argue that AT2018cow is not atypical among interacting SNe, but that it is brighter than most upper limits in a 17-SN subsample closer than AT2018cow. Combined with a reported late-time photospheric radius of about 40 solar radii (Inkenhaag et al. 2023), they conclude that AT2018cow's late-time UV emission was not driven by interaction and may instead trace the inner engine, possibly a long-lived accretion disk.","tokens_in":27574,"tokens_out":6934,"duration_ms":67607,"significance":"If the conclusion holds, the paper provides an important empirical constraint on the nature of AT2018cow and LFBOTs more generally, favoring engine-powered, non-interaction channels. The observational work is careful: the astrometric alignment, artificial-star limits, and synthetic photometry from archival spectra of SN2010jl and SN1993J are all thoughtfully executed, and the resulting sample of late-time UV upper limits for CCSNe is a genuinely useful resource. The two detections are consistent with independent classifications of both objects as interacting SNe, which strengthens confidence in the methodology. The main risk is not in the measurements but in the interpretive step that excludes interaction for AT2018cow, which depends on an unverified single-blackbody assumption.","major_comments":[{"comment":"The interaction-exclusion conclusion rests on the premise, imported from Inkenhaag et al. (2023), that AT2018cow's late-time UV emission is a single blackbody with photospheric radius about 40 solar radii. The text itself qualifies this with 'if it is a BB at these epochs', but no test of the single-blackbody assumption is presented here. If the late-time UV is non-thermal, a two-component SED, or emission from a compact shocked region embedded in a larger CSM interaction, a small apparent blackbody radius would not exclude an interaction-powered origin. Because this premise is load-bearing for the central conclusion, the manuscript should either demonstrate the blackbody/multi-component interpretation with the available photometry or spectra, or soften the conclusion to explicitly state that it applies only if the emission is a single blackbody photosphere.","section":"Section 4.4"},{"comment":"The numerical claim supporting the brightness argument is internally inconsistent. The paper states that only 5 out of 17 CCSNe in the closer subsample have an absolute-magnitude limit brighter than AT2018cow and the remaining 12 have a fainter limit, but then concludes that 'SNe as bright as AT2018cow would have been detected in 75 percent of the images'. With one image per SN, 5/17 is approximately 29 percent, not 75 percent; if the intended statement is that 12 of 17 limits are bright enough, the fraction is approximately 71 percent, which still does not equal 75 percent and contradicts the preceding sentence. Since the 75 percent figure is repeated in the abstract and conclusions, this needs to be corrected and the statistical argument re-stated precisely.","section":"Section 4.4"},{"comment":"Even after correcting the percentage, the brightness comparison alone has limited discriminating power for the interaction hypothesis. The detected interacting SN ATLAS17lsn is about 3 mag brighter than AT2018cow and ASASSN-17qp about 3 mag fainter, a spread that brackets AT2018cow; the Dessart et al. (2023) model comparison is explicitly acknowledged by the authors as not allowing firm conclusions. Thus the argument against interaction reduces essentially to the radius premise raised in the first major comment. The paper should state this dependence explicitly rather than presenting brightness and radius as two independent strands of evidence.","section":"Section 4.4"}],"minor_comments":[{"comment":"The abstract says 'for two CCSNe we detect a point source', but for ASASSN-17qp the source centroid is at 3.2 sigma outside the 3 sigma uncertainty region and is only associated after an additional relative-astrometry step; consider describing it as a 'possible' detection in the abstract.","section":"Abstract and Section 3"},{"comment":"The caption and footnotes should explicitly state that the 'Brightness' column entries with '>' denote 95 percent completeness upper limits from the artificial-star experiment of Section 2.6, and that 'sigma_total' is the quadratic sum in Eq. (1).","section":"Table 1"},{"comment":"There is a typo: 'SN1993C' in the discussion of the Dessart et al. (2023) model should read 'SN1993J'.","section":"Section 4.3"},{"comment":"The figure caption says 'the black line represent' and should be 'represents'; it would also help to state explicitly whether the arrows are 95 percent upper limits or some other limiting magnitude.","section":"Figure 2"},{"comment":"The artificial-star experiment uses one PSF model constructed from SN2017ffq for all images; the paper states that subtraction tests on one random image showed no residuals, but a short statement on how the PSF model was checked across the full range of positions and dither patterns would strengthen the reproducibility of the limits.","section":"Section 2.6"}],"recommendation":"major_revision","confidential_remarks":"The underlying observational product—late-time UV detections and upper limits for 51 CCSNe—is solid and useful. The main risk is the interpretive step in Section 4.4, where the exclusion of interaction for AT2018cow depends on a single-blackbody assumption imported from earlier work. The 75 percent versus 5/17 numerical inconsistency must be resolved, and the conclusion should be calibrated to the strength of the evidence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Candidly, the useful part of this paper is the census, and it is genuinely new. Fifty-one CCSNe with HST F275W observations at 2–5 years, a careful astrometric re-calibration, DOLPHOT photometry for the two detections, and artificial-star upper limits for the rest. Two detections in 51, both interacting SNe, with chance-alignment probabilities quoted. That is a clean empirical result: late-time UV emission in CCSNe is rare, and when it appears it points to interaction. The null result for the other 49, with published limits, is a reference dataset people will want.\n\nThe AT2018cow comparison is where I get more cautious. The paper shows AT2018cow is not unusually bright for an interacting SN — ATLAS17lsn is ~3 mag brighter, ASASSN-17qp ~3 mag fainter, so that part is honest and does not take a position. The stronger claim, that the UV is not driven by interaction and that we are seeing the inner engine, rests on the small photospheric radius (~40 R_sun) from Inkenhaag et al. 2023. Section 4.4 says 'if it is a BB at these epochs' — and that is exactly the premise that goes untested. If the UV is a non-thermal component, a two-component SED, or a compact shocked region embedded in a larger CSM interaction, the radius argument collapses. The paper flags the caveat but does not check it. The closer-sample brightness argument (17 SNe, 75 percent would have been detected) is weaker: upper limits are not detections, and the two interacting detections bracket AT2018cow in brightness, so it does not by itself exclude interaction. The radius argument is doing the load-bearing work.\n\nIs that fatal? Not for the census, which stands on its own. And the paper is honest about the conditional nature of the conclusion. But the headline claim about AT2018cow should be read as 'under a single-blackbody interpretation, interaction is disfavoured', not as a settled exclusion. A referee should push on whether the blackbody assumption holds at these epochs, and whether the two detected interacting SNe tell us anything about the spread in photospheric radii among interacting CCSNe.\n\nI would send this to a serious referee. The observational analysis is careful, the data product is useful, and the AT2018cow discussion, while conditional, is worth having in the literature. With a bit more caution in the conclusions about what the radius argument assumes, it would be a solid A&A paper. I'd cite it for the census.","headline":"A careful new census of late-time UV emission in CCSNe, with a conclusion about AT2018cow that holds only if you accept the single-blackbody radius argument.","tokens_in":28217,"tokens_out":2790,"would_cite":true,"duration_ms":27251,"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":"This paper concludes that the late-time ultraviolet emission of the peculiar transient AT2018cow is not powered by supernova ejecta interacting with circumstellar material, and is more likely radiation from the inner engine of the…","keywords":["core-collapse supernovae","AT2018cow","late-time ultraviolet emission","luminous fast blue optical transients","circumstellar medium interaction","tidal disruption events","Hubble Space Telescope","accretion disk"],"falsifier":"A direct test would be a late-time ultraviolet spectrum of AT2018cow: narrow or intermediate-width emission lines from shocked circumstellar gas would contradict the no-interaction conclusion, as would resolving an emission region larger than about 40 solar radii in size.","tokens_in":27134,"feed_emoji":"🔭","tokens_out":10655,"duration_ms":84109,"temperature":0.7,"pith_summary":"Core-collapse supernovae normally fade out of the ultraviolet within a few weeks, so a supernova that still glows in the UV years later needs an extra power source: either the ejecta colliding with circumstellar material, or a direct view down to the central engine. The authors use 51 nearby core-collapse supernovae observed with the Hubble Space Telescope in the ultraviolet 2-5 years after explosion, and find only two with a point source at the supernova position, both already known as interacting supernovae. Against this census, AT2018cow's late-time UV brightness is not atypical for an interacting supernova, but it is brighter than the upper limits for most supernovae closer than itself. The key additional fact is that AT2018cow's emission sits on a blackbody photosphere of only about 40 solar radii, orders of magnitude smaller than an interacting supernova's photosphere, which would be embedded in the circumstellar material. The paper concludes that AT2018cow's late-time UV emission is not driven by interaction, and that we are probably seeing the inner region of the explosion, possibly a long-lived accretion disk, as expected in tidal disruption models.","feed_headline":"AT2018cow's late UV light points to an engine, not a supernova","feed_subtitle":"A 51-supernova UV census finds only two detections, both interacting; AT2018cow is brighter and far more compact.","key_machinery":"The load-bearing object is a comparative census: 51 nearby core-collapse supernovae observed in one ultraviolet filter (F275W) between 2 and 5 years after discovery, with positions tied to Gaia/Pan-STARRS astrometry, point-source detections confirmed by PSF photometry, and magnitude upper limits from an artificial star experiment. The decisive comparison is absolute UV magnitude versus time since discovery, with AT2018cow's late-time light curve and a reference interacting-supernova UV model overlaid; this is what shows that AT2018cow is bracketed by the two detections yet brighter than most nearby upper limits. The second mechanism is the photospheric-radius argument: AT2018cow's emission is a blackbody with radius about 40 solar radii (measured in the authors' earlier work), and any interacting core-collapse supernova that is a blackbody at these epochs would have its photosphere embedded in circumstellar material at a much larger radius, ruling out interaction as the driver.","core_discovery":"Out of 51 nearby core-collapse supernovae (z<0.065) imaged with HST/WFC3 F275W within 2-5 years of discovery, the authors find a likely point source at the supernova position in only two: ASASSN-17qp and ATLAS17lsn, both of which had already been identified as interacting supernovae. In absolute UV magnitude, AT2018cow sits between these two detections, so a late-time UV detection by itself is not evidence against a supernova nature. But when the sample is restricted to supernovae closer than AT2018cow, the transient is brighter than the upper limits on most of them, and its blackbody photospheric radius of roughly 40 solar radii is orders of magnitude smaller than the photosphere an interacting core-collapse supernova would have in its circumstellar medium. The paper's conclusion is that AT2018cow's late-time UV emission was not driven by interaction; instead, we are likely seeing the inner region of the explosion, perhaps a long-lived accretion disk, a scenario naturally expected in tidal disruption models and less straightforward in supernova scenarios.","pith_inferences":["If the engine interpretation is correct, other luminous fast blue optical transients should show similar late-time ultraviolet persistence; a targeted UV survey of LFBOTs at 2-5 years would test this prediction.","The radius argument relies on the emission being a single blackbody; if future observations reveal a non-thermal component (for example from shocks or a jet), the case against interaction would need revisiting.","The same snapshot-and-artificial-star approach could be extended to larger transient samples with upcoming wide-field ultraviolet surveys to map what fraction of core-collapse supernovae show late-time UV excess and to separate interacting from engine-powered events.","A distinctive prediction of the accretion-disk scenario is a slowly declining, possibly variable UV source; interaction-powered emission, by contrast, would be expected to steepen as the shock decelerates through the circumstellar medium."],"forward_implications":["Late-time ultraviolet emission between 2 and 5 years after explosion is rare among core-collapse supernovae: only 2 of 51 show a likely point source, and both are known interacting events, so such detections are a practical marker for ongoing ejecta-CSM interaction.","AT2018cow's late-time UV brightness is bracketed by the two interacting-supernova detections, so a persistent UV source alone does not argue against a supernova nature.","Among core-collapse supernovae closer than AT2018cow, 12 of 17 would have been detectable if they were as bright as AT2018cow, making the transient unusually bright for a supernova at that distance.","Given the roughly 40-solar-radius photosphere, an interacting-supernova explanation would require a far larger photosphere in circumstellar material, so the ultraviolet emission is more plausibly powered by the inner engine, such as a long-lived accretion disk.","If this conclusion holds, AT2018cow's late-time behavior favors tidal-disruption-like models with a persistent accretion disk over standard or interacting supernova models."],"supporting_citations":[{"why":"Supplies AT2018cow's late-time UV light curve and the roughly 40-solar-radii photospheric radius that rules out a CSM-embedded photosphere.","marker":"Inkenhaag et al. 2023"},{"why":"First reported the late-time (>2 yr) UV bright source at AT2018cow's position, the phenomenon the paper seeks to interpret.","marker":"Sun et al. 2022"},{"why":"Provides the interacting-supernova UV light-curve model that the paper overplots to interpret detections and non-detections.","marker":"Dessart et al. 2023"},{"why":"Identifies ASASSN-17qp as a long-term interacting emitter, establishing that one of the two UV detections is an interacting supernova.","marker":"Gutiérrez et al. 2020"},{"why":"Identifies ATLAS17lsn as an interacting supernova with massive circumstellar shells, supporting the second detection being interaction-powered.","marker":"Smith & Andrews 2020"},{"why":"Reports more than 1700 days of optical-to-infrared emission of ATLAS17lsn attributed to CSM interaction, strengthening the association of its UV detection with an interacting supernova.","marker":"Moran et al. 2023"},{"why":"Supplies the artificial star experiment method used to set the upper limits on the 49 non-detected supernovae.","marker":"Eappachen et al. 2022"},{"why":"Provides late-time UV data of SN2010jl used as a comparison interacting supernova in the brightness plots.","marker":"Fransson et al. 2014"}],"fun_headline_variants":["AT2018cow UV not from interaction, hints at central engine","Late UV in AT2018cow points to central engine, not supernova interaction","51 supernovae: only two UV detections, AT2018cow stands out as compact","AT2018cow's persistent UV: engine-driven, not supernova collision","Late-time UV survey rules out interaction for AT2018cow"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusion that interaction cannot explain AT2018cow assumes that its late-time ultraviolet emission is a single blackbody photosphere of about 40 solar radii, and that any interacting core-collapse supernova would have a photosphere embedded in its circumstellar material that is much larger than this.","fun_headline_variants_meta":{"raw":{"variants":["AT2018cow UV not from interaction, hints at central engine","Late UV in AT2018cow points to central engine, not supernova interaction","51 supernovae: only two UV detections, AT2018cow stands out as compact","AT2018cow's persistent UV: engine-driven, not supernova collision","Late-time UV survey rules out interaction for AT2018cow"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001128,"raw_usage":{"total_tokens":4808,"prompt_tokens":1184,"completion_tokens":3624,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":800,"completion_tokens_details":{"reasoning_tokens":3518}},"tokens_in":800,"tokens_out":3624,"duration_ms":23370,"temperature":1.0,"reasoning_tokens":3518,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T20:23:10.113257+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct test would be a late-time ultraviolet spectrum of AT2018cow: narrow or intermediate-width emission lines from shocked circumstellar gas would contradict the no-interaction conclusion, as would resolving an emission region larger than about 40 solar radii in size.","supporting_citations":[{"cited_title":"R., Crowther, P","cited_arxiv_id":null,"evidence_quote":"First reported the late-time (>2 yr) UV bright source at AT2018cow's position, the phenomenon the paper seeks to interpret."},{"cited_title":"& Andrews, J","cited_arxiv_id":null,"evidence_quote":"Identifies ATLAS17lsn as an interacting supernova with massive circumstellar shells, supporting the second detection being interaction-powered."},{"cited_title":"2023, A&A, 669, A51","cited_arxiv_id":null,"evidence_quote":"Reports more than 1700 days of optical-to-infrared emission of ATLAS17lsn attributed to CSM interaction, strengthening the association of its UV detection with an interacting supernova."}],"review_version":1}