An elliptical accretion disk following the tidal disruption event AT 2020zso

Aims. The modelling of spectroscopic observations of tidal disruption events (TDEs) to date suggests that the newly formed accretion disks are mostly quasi-circular. In this work we study the transient event AT 2020zso, hosted by an active galactic nucleus (AGN; as inferred from narrow emission line diagnostics), with the aim of characterising the properties of its newly formed accretion flow. Methods. We classify AT 2020zso as a TDE based on the blackbody evolution inferred from UV/optical photometric observations and spectral line content and evolution. We identify transient, double-peaked Bowen (N III), He I, He II, and H alpha emission lines. We model medium-resolution optical spectroscopy of the He II (after careful de-blending of the N III contribution) and H alpha lines during the rise, peak, and early decline of the light curve using relativistic, elliptical accretion disk models. Results. We find that the spectral evolution before the peak can be explained by optical depth effects consistent with an outflowing, optically thick Eddington envelope. Around the peak, the envelope reaches its maximum extent (approximately 10(15) cm, or similar to 3000-6000 gravitational radii for an inferred black hole mass of 5-10 x 10(5) M-circle dot) and becomes optically thin. The H alpha and He II emission lines at and after the peak can be reproduced with a highly inclined (i  = 85 +/- 5 degrees), highly elliptical (e  = 0.97 +/- 0.01), and relatively compact (R-in = several 100 R-g and R-out = several 1000 R-g) accretion disk. Conclusions. Overall, the line profiles suggest a highly elliptical geometry for the new accretion flow, consistent with theoretical expectations of newly formed TDE disks. We quantitatively confirm, for the first time, the high inclination nature of a Bowen (and X-ray dim) TDE, consistent with the unification picture of TDEs, where the inclination largely determines the observational appearance. Rapid line profile variations rule out the binary supermassive black hole hypothesis as the origin of the eccentricity; these results thus provide a direct link between a TDE in an AGN and the eccentric accretion disk. We illustrate for the first time how optical spectroscopy can be used to constrain the black hole spin, through (the lack of) disk precession signatures (changes in inferred inclination). We constrain the disk alignment timescale to > 15 days in AT2020zso, which rules out high black hole spin values (a  < 0.8) for M-BH  similar to  10(6) M-circle dot and disk viscosity alpha  greater than or similar to  0.1.