2025ApJ...987..120W

We develop a novel approach to performing precision inference on tidally perturbed dwarf galaxies. We use a Bayesian inference framework of implicit likelihood inference, previously applied mainly in the field of cosmology, based on forward simulation, data compression, and likelihood emulation with neural density estimators. We consider the case of NGC 205, a satellite of M31. NGC 205 exhibits an S-shape in the mean line-of-sight velocity along its semimajor spatial axis, suggestive of tidal perturbation. We demonstrate that this velocity profile can be qualitatively reproduced even if NGC 205 was in a spherically symmetric and isotropic state before its most recent pericenter passage. We apply our inference method to mock data and show that the precise shape of a perturbed satellite's sky-projected internal velocity field, mapped across its entire face, can be highly informative of both its orbit and total mass density profile, even in the absence of proper motion information. For NGC 205 specifically, our method is hampered because the available data only covers a line along its semimajor axis. This shortcoming could be addressed with another round of observations.