Galactic winds and the Lyα forest

We study the effect of galactic outflows on the statistical properties of the Lyα forest and its correlation with galaxies. The winds are modelled as fully ionised spherical bubbles centered around the haloes in an N-body simulation of a ΛCDM model. The observed flux probability distribution and flux power spectrum limit the volume filling factor of bubbles to be less than 10%. We have compared the mean flux as a function of distance from haloes with the Adelberger et al. (ASSP) measurement. For a model of bubbles of constant size surrounding the most massive haloes, bubble radii of > ∼ 1.5 hMpc are necessary to match the high transmissivity at separations ≤ 0.5 hMpc but the increase of the transmissivity at small scales is more gradual than observed. The cosmic variance error due to the finite number of galaxies in the sample increases rapidly with decreasing separation. At separations ≤ 0.5 hMpc our estimate of the cosmic variance error is ∆F̄ ∼ 0.3, 30% higher than that of ASSP. The difficulty in matching the rise in the transmissivity at separations smaller than the size of the fully ionised bubbles surrounding the haloes is caused by residual absorption of neutral hydrogen lying physically outside the bubbles but having a redshift position similar to the haloes. The flux level is thus sensitive to the amplitude of the coherent velocity shear near halos and to a smaller extent to the amplitude ot thermal motions. We find that the velocity shear increases with halo mass in the simulation. A model where LBGs are starbursts in small mass haloes matches the observations with smaller bubble radii than a model where massive haloes host the LBGs. If we account for the uncertainty in the redshift position of haloes, a starburst model with a bubble radius of 1 hMpc and a volume filling factor of 2% is consistent with the ASSP measurements at the 1-1.5σ level. If this model is correct the sharp rise of the transmissivity at separations ≤ 0.5 hMpc in the ASSP sample is due to cosmic variance and is expected to become more moderate for a larger sample.