Cosmic ménage à trois: the origin of satellite galaxies on extreme orbits

We examine the orbits of satellite galaxies identified in a suite of N-body/gasdynamical simulations of the formation of L_* galaxies in a Lambda cold dark matter universe. The numerical resolution of the simulations allows us to track in detail the orbits of the ~ 10 brightest satellites around each primary. Most satellites follow conventional orbits; after turning around, they accrete into their host halo and settle on orbits whose apocentric radii are steadily eroded by dynamical friction. As a result, satellites associated with the primary are typically found within its virial radius, r_vir, and have velocities consistent with a Gaussian distribution with mild radial anisotropy. However, a number of outliers are also present. We find that a surprising number (about one-third) of satellites identified at z = 0 are on unorthodox orbits, with apocentres that exceed their turnaround radii. These include a number of objects with extreme velocities and apocentric radii at times exceeding ~3.5r_vir (or, e.g. >~1 Mpc when scaled to the Milky Way). This population of satellites on extreme orbits consists typically of the faint member of a satellite pair whose kinship is severed by the tidal field of the primary during first approach. Under the right circumstances, the heavier member of the pair remains bound to the primary, whilst the lighter companion is ejected on to a highly energetic orbit. Since the concurrent accretion of multiple satellite systems is a defining feature of hierarchical models of galaxy formation, a fairly robust prediction of this scenario is that at least some of these extreme objects should be present in the Local Group. We speculate that this three-body ejection mechanism may be the origin of (i) some of the newly discovered high-speed satellites around M31 (such as Andromeda XIV); (ii) some of the distant fast-receding Local Group members, such as Leo I and (iii) the oddly isolated dwarf spheroidals Cetus and Tucana in the outskirts of the Local Group. Our results suggest that care must be exercised when using the orbits of the most weakly bound satellites to place constraints on the total mass of the Local Group.