A Mildly Relativistic Outflow Launched Two Years after Disruption in the Tidal Disruption Event AT2018hyz

We present late-time radio/millimeter (as well as optical/UV and X-ray) detections of the tidal disruption event (TDE) AT2018hyz, spanning $970 - 1300$ d after optical discovery. In conjunction with earlier deeper limits, including at $\approx 700$ d, our observations reveal rapidly rising emission at $0.8-240$ GHz, steeper than $F_\nu\propto t^5$ relative to the time of optical discovery. Such a steep rise cannot be explained in any reasonable scenario of an outflow launched at the time of disruption (e.g., off-axis jet, sudden increase in the ambient density), and instead points to a delayed launch. Our multi-frequency data allow us to directly determine the radius and energy of the radio-emitting outflow, showing that it was launched $\approx 750$ d after optical discovery. The outflow velocity is mildly relativistic, with $\beta\approx 0.25$ and $\approx 0.6$ for a spherical and a $10^\circ$ jet geometry, respectively, and the minimum kinetic energy is $E_K\approx 5.8\times 10^{49}$ and $\approx 6.3\times 10^{49}$ erg, respectively. This is the first definitive evidence for the production of a delayed mildly-relativistic outflow in a TDE; a comparison to the recently-published radio light curve of ASASSN-15oi suggests that the final re-brightening observed in that event (at a single frequency and time) may be due to a similar outflow with a comparable velocity and energy. Finally, we note that the energy and velocity of the delayed outflow in AT2018hyz are intermediate between those of past non-relativistic TDEs (e.g., ASASSN-14li, AT2019dsg) and the relativistic TDE Sw\,J1644+57. We suggest that such delayed outflows may be common in TDEs.