Accretion shocks and cold filaments in galaxy formation

A generic expectation for gas accreted by high-mass haloes is that it is shock-heated to the virial temperature of the halo. In low-mass haloes, or at high redshift, however, the gas cooling rate is sufficiently rapid that an accretion shock is unlikely to form close to the virial radius. Instead, the accretion shock will form at smaller radii, perhaps close to the central galaxy. Semi-analytic models have always made a clear distinction between the regimes in which accretion is limited by the cooling time of hot gas and by the dynamical time-scale of the halo, using simple estimates of the mass-scale at which the transition from rapid to slow cooling occurs. In this work, we revisit this issue using the latest understanding and calibration of accretion shock formation. Starting from the well-established Galform code, we investigate the effect of accounting for the presence or otherwise of an accretion shock close to the virial radius using the shock stability model of Birnboim & Dekel. As expected, when we modify the code so that there is no effective feedback from galaxy formation, we find that so-called ‘cold-mode’ accretion is the dominant channel for feeding gas into the galaxies at high redshifts, such that 90 per cent of baryons in galaxies (averaged over all galaxies) arrive via this channel. The mass-scale at which the rapid to slow cooling transition occurs is significantly affected at high redshifts and accretion rates become dominated by cold-mode accretion. However, the impact of this change in the cooling channel on galaxies properties is mitigated by compensating effects in the star formation and feedback cycle. When effective feedback, which reheats gas from galaxies to the virial temperature but which allows no gas to escape from a halo, is included in the model, we find that the ‘cold mode’ is even less apparent because of the presence of gas ejected from the galaxy's disc, although it can still contribute almost 50 per cent of the net inflow rate when averaged over all galaxies. Thus, the inclusion of the latest calibration of accretion shock physics makes little difference to basic results from earlier semi-analytic models, which used a simpler treatment. We conclude that this ‘cold-mode’ physics is already adequately accounted for in semi-analytic models and that feedback represents a much larger uncertainty than any of these effects.