AT 2018cow at ~5 years: additional evidence for a tidal disruption origin

TL;DR

This study presents a new epoch of HST UV observations of AT 2018cow at approximately 1900 days post-explosion to test the tidal disruption event (TDE) hypothesis. Photometric analysis across four filters shows a UV plateau and slow fading consistent with accretion-disk evolution expected from a TDE around an intermediate-mass black hole with $M_{ m BH}$ near $10^{3}\,M_\odot$, corroborated by a recovered $\log(M_{ m BH})=3.3\pm0.4$ when including the latest data. Comparisons with SN–CSM interaction models indicate faster fading for such scenarios, strengthening the TDE interpretation while acknowledging that deeper late-time data (∼2800–3000 days) are needed to conclusively exclude CSM interaction. Overall, the results extend the case for a TDE origin for AT 2018cow and illustrate how late-time UV observations can discriminate between disk-dominated TDEs and interacting SNe.

Abstract

The Luminous Fast Blue Optical Transient (LFBOT) AT 2018cow is the prototype of its class with an extensive set of multi-wavelength observations. Despite a rich data set there is, still, no consensus about the physical nature and origin of this event. AT 2018cow remained UV bright 2-4 years after the explosion, which points at an additional energy injection source, most likely from an accretion disk. We present additional late time UV data obtained with the Hubble Space Telescope, to show there is no significant fading in the optical since the last epoch and only marginal fading in the UV. The new UV data points match the predictions of previously published accretion disk models, where the disk is assumed to form from the tidal disruption of a low mass star by an intermediate mass black hole. This consistency provides evidence that AT 2018cow could indeed be a tidal disruption event. The marginal decay is in contrast with the predictions of light curves produced by interacting supernovae. The difference between expectations for disc emission and interacting supernovae will further increase with time, making data at even later times a route to robustly rule out interaction between supernova ejecta and circumstellar material.

Figures (3)

• Figure 1: Lightcurves in four HST filters from Inkenhaag2023, with the data present in this work added (data after t=1750 d is the new data), to show the rate of decay in all four filters slowed down significantly at later times, resembling the lightcurves of TDEs. The lightcurves in the different filters are offset for clarity as indicated in the legend, and different colours correspond to the different filters (F225W in blue, F336W in red, F555W in yellow and F814W in green). We note that the 1$\sigma$ error bars on the magnitudes are small, the horizontal bars through the markers are the endcaps of these error bars. The absolute magnitudes plotted in blue and red (UV filters) are intrinsic to the source, while the magnitudes plotted in yellow and green data (optical filters) likely originate in an underlying extended source Inkenhaag2023.
• Figure 2: Top panel: Late time (>660 days) UV lightcurve of AT 2018cow in the F225W (blue) and F336W (red) filters, assuming all emission is coming from AT 2018cow. The red and blue lines are the disk modelling from Inkenhaag2023 in these same filters, which were fit to the first two epochs of data shown only and extrapolated out to later times here (no further fitting has been performed). The third epoch at 1900 days is consistent with the predictions from this model, under the assumption that all UV light come from AT 2018cow (see Section \ref{['sec:discussion']}). The purple and pink stars are synthetic photometric measurements in the F275W filter from two interacting supernovae, SN 2010jl and SN 1993J, respectively, as presented in Inkenhaag2025. The dashed lines are straight line fits to the magnitudes as a function of time, extrapolated to show the decay rate of in the lightcurve to later times. The black line is the model for an interacting SN from Dessart2023, extrapolated after 1000 days. Bottom panel: Zoom-in on the lightcurve of AT 2018cow, where the circular data points represent the UV magnitudes assuming all the UV light is emitted by the transient, the square markers represent the UV magnitudes when a constant underlying source with the average SEDs of the compact SF regions from Inkenhaag2023 is subtracted, and the triangular data points represent the UV magnitudes when a constant underlying source with the bluest SED that has F815W-F555W colours redder than AT 2018cow is subtracted (see the main text for details). Note that the circular markers for the F336W filter are behind the square and triangular markers for the last two epochs. We have also plotted the average disk model for each of the cluster assumptions, which include the data of the third epoch, in the same style as in the top panel.
• Figure 3: Late time (>660 days) UV lightcurves of AT 2018cow assuming no underlying cluster (top panel), an average cluster obtained from the compact SF regions in the host galaxy (middle panel) or a blue cluster (bottom pabel), see the main text for details. F225W magnitudes are plotted in blue and F226W magnitudes are plotted in red. The lines in each panel are 100 disk models for each filter, from which the mass of the BH has been deduced, with the colours corresponding to the filter. These disk models are the models from Inkenhaag2023, but repeated to include the third epoch in the fit. We find BH masses in each case which are consistent with the BH mass found in Inkenhaag2023.