Title: On the origin of late-time X-ray flares in UV/optically-selected tidal disruption events

Abstract: We propose a model to explain the time delay between the peak of the optical and X-ray luminosity, \dt hereafter, in UV/optically-selected tidal disruption events (TDEs). The following picture explains the observed \dt in several TDEs as a consequence of the circularization and disk accretion processes as long as the sub-Eddington accretion. At the beginning of the circularization, the fallback debris is thermalized by the self-crossing shock caused by relativistic precession, providing the peak optical emission. During the circularization process, the mass fallback rate decreases with time to form a ring around the supermassive black hole (SMBH). The formation timescale corresponds to the circularization timescale of the most tightly bound debris, which is less than a year to several decades, depending mostly on the penetration factor, the circularization efficiency, and the black hole mass. The ring will subsequently evolve viscously over the viscous diffusion time. We find that it accretes onto the SMBH on a fraction of the viscous timescale, which is $2$ years for given typical parameters, leading to X-ray emission at late times. The resultant \dt\,is given by the sum of the circularization timescale and the accretion timescale and significantly decreases with increasing penetration factor to several to $\sim10$ years typically. Since the X-ray luminosity substantially decreases as the viewing angle between the normal to the disk plane and line-of-sight increases from $0^\circ$ to $90^\circ$, a low late-time X-ray luminosity can be explained by an edge-on view. We also discuss the super-Eddington accretion scenario, where \dt\,is dominated by the circularization timescale.