The SDSS-DR12 large-scale cross-correlation of damped Lyman alpha systems with the Lyman alpha forest

We present a measurement of the damped Ly alpha absorber (DLA) mean bias from the cross-correlation of DLAs and the Ly alpha forest, updating earlier results of Font-Ribera et al. (2012) with the final Baryon Oscillations Spectroscopic Survey data release and an improved method to address continuum fitting corrections. Our cross-correlation is well fitted by linear theory with the standard Lambda CDM model, with a DLA bias of b(DLA) = 1.99 +/- 0.11; a more conservative analysis, which removes DLA in the Ly beta forest and uses only the cross-correlation at r > 10 h(-1) Mpc, yields b(DLA) = 2.00 +/- 0.19. This assumes the cosmological model from Planck Collaboration (2016) and the Ly a forest bias factors of Bautista et al. (2017) and includes only statistical errors obtained from bootstrap analysis. The main systematic errors arise from possible impurities and selection effects in the DLA catalogue and from uncertainties in the determination of the Ly alpha forest bias factors and a correction for effects of high column density absorbers. We find no dependence of the DLA bias on column density or redshift. The measured bias value corresponds to a host halo mass similar to 4 x 10(11) h(-1)M(circle dot) if all DLAs were hosted in haloes of a similar mass. In a realistic model where host haloes over a broad mass range have a DLA cross-section Sigma(M-h) proportional to M-h(alpha) down to M-h > M-min = 10(8.5) h(-1)M(circle dot), we find that alpha > 1 is required to have b(DLA) > 1.7, implying a steeper relation or higher value of M-min than is generally predicted in numerical simulations of galaxy formation.