Erratum to : The bias of DLAs at z ~ 2.3 : evidence for very strong stellar feedback in shallow potential wells

The paper 'The bias of DLAs at z ~ 2.3: evidence for very strong stellar feedback in shallow potential wells' was published in MNRAS 440, 2313 (2014). A numerical error in the calculation of the DLA bias affects our conclusion about the virial velocity of dark matter halo below which the neutral hydrogen fraction is sharply suppressed. A corrected version of fig. 5 is shown below (Fig. 1); all other figures in our original paper are not affected by the error, though we have changed our conclusions about which set of parameter values should be fiducial. In light of this corrected prediction of the DLA bias, our fiducial model parameters are as follows: uv,0 = 35 km s-1 αe = 3, (1) dN/dX = 0.08 c0 = 40, (2) xpeak = 0.7 xFWHM = 0.8, (3) The parameters of our revised [M/H]-Mv model are chosen by maximizing the likelihood: (αM, αv, βM, σ[M/H]) = (0.47, 0.08, -1.34, 0.3). As a result, a model in which the fraction of neutral hydrogen in DM haloes drops sharply below ~35 km s-1 can account for the observed value of the DLA bias, as well as the DLA column density abundance and velocity width distribution. This is broadly in agreement with our previous modelling of the absorption and Ly α emission properties of DLAs (Barnes & Haehnelt 2010). In our model, the DLA halo mass-metallicity relation is [M/H]mean = (0.47 ± 0.1)log (Mv/1011M⊙) - 1.34, with no significant metallicity-velocity width relation at fixed halo mass. Our main conclusion is unchanged: stellar feedback in shallow potential wells is more efficient than realized in many current numerical galaxy formation models. The required efficiency is, however, significantly less extreme than we had suggested previously. We agree now with Bird et al. (2014) that there is no significant tension between the DLA bias measurement and other observed DLA properties.