Study of late-time ultraviolet emission in core collapse supernovae and its implications for the peculiar transient AT2018cow

Context. Over time, core-collapse supernova (CCSN) spectra become redder due to dust formation and cooling of the SN ejecta. An ultraviolet (UV) detection of a CCSN at late times will thus indicate an additional physical process, such as an interaction between the SN ejecta and the circumstellar material, or viewing down to the central engine of the explosion. Both of these models have been proposed to explain the peculiar transient AT2018cow, a luminous fast blue optical transient detected in the UV two to four years after the event, with only marginal fading over this time period. Aims. To identify whether the late-time UV detection of AT2018cow could indicate that it is a CCSN, we investigate whether CCSNe are detectable in the UV between two and five years after the explosion. We determine how common late-time UV emission in CCSNe is and compare those CCSNe detected in the UV to the peculiar transient AT2018cow. Methods. We used a sample of 51 nearby (z < 0.065) CCSNe observed with the Hubble Space Telescope within two to five years of discovery. We measured their brightness or determined an upper limit on the emission through an artificial star experiment in cases of no detection. Results. For two CCSNe, we detected a point source within the uncertainty region of the SN position. Both have a low chance alignment probability with bright objects within their host galaxies. Therefore, they are likely to be related to their SNe, which are both known to be interacting SNe. Conclusions. Comparing the absolute UV magnitude of AT2018cow at late times to the absolute UV magnitudes of the two potential SN detections, there is no evidence that a late-time UV detection of AT2018cow is atypical for interacting SNe. However, when limiting the sample to CCSNe closer than AT2018cow, we see that it is brighter than the upper limits on most CCSN non-detections. Combined with a very small late time photospheric radius of AT2018cow, this leads us to conclude that the late-time UV detection of AT2018cow was not driven by interaction. Instead, it suggests that we are possibly viewing the inner region of the explosion that is perhaps due to the long-lived presence of an accretion disc. Such properties are naturally expected in tidal disruption models and are less straightforward (though not impossible) in SN scenarios.