摘要
We apply a semi-analytic galaxy formationmodel to two high-resolution cosmological N-body simulations to investigate analogues of the Milky Way system. We select these according to observed properties of the Milky Way rather than by halo mass as in most previous work. For disc-dominated central galaxies with stellarmass (5-7)×10 <sup>10</sup> M <sub>⊙</sub> , the median host halo mass is 1.4 × 10 <sup>12</sup> M <sub>⊙</sub> , with 32-68 per cent percentile spread in the range [0.86, 3.1] × 10 <sup>12</sup> M <sub>⊙</sub> , consistent with dynamical measurements of the Milky Way halo mass. For any given halo mass, the probability of hosting a MilkyWay system is low, with a maximum of ~20 per cent in haloes of mass ~10 <sup>12</sup> M <sub>⊙</sub> . The model reproduces the V-band luminosity function and radial profile of the bright (M <sub>V</sub> < -9) Milky Way satellites (r < 280 kpc). Galaxy formation in low-mass haloes is found to be highly stochastic, resulting in an extremely large scatter in the relation between M <sub>V</sub> (or stellar mass) for satellites and the depth of the subhalo potential well in which they live, as measured by the maximum of the rotation curve, V <sub>max</sub> . Following Sawala et al., we account for baryonic effects on the growth of dark halo potentials by rescaling V <sub>max</sub> . This adjustment alleviates the 'too big to fail' problem - we find that, in 35 per cent of Milky Way-like systems, three or fewer of the top 12 satellites (in order of V <sub>max</sub> ) have V <sub>max</sub> > 30 km s <sup>-1</sup> . Our model predicts that around half of the dark matter subhaloes with V <sub>max</sub> > 20 km s <sup>-1</sup> host satellites fainter than M <sub>V</sub> = -9 and so may be missing from existing surveys.