The SDSS-DR12 large-scale cross-correlation of Damped Lyman Alpha Systems with the Lyman Alpha Forest

We present a measurement of the DLA mean bias from the cross-correlation of DLA and the Ly$\alpha$ forest, updating earlier results of Font-Ribera et al. 2012 with the final BOSS 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 $b_{\rm DLA} = 1.99\pm 0.11$; a more conservative analysis, which removes DLA in the Ly$\beta$ forest and uses only the cross-correlation at $r> 10{\rm h^{-1}\,Mpc}$, yields $b_{\rm DLA} = 2.00\pm 0.19$. This assumes the cosmological model from \cite{Planck2015} and the Ly$\alpha$ 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 $\sim 4\cdot10^{11} {\rm M_{\odot}}$ if all DLA were hosted in halos of a similar mass. In a realistic model where host halos over a broad mass range have a DLA cross section $\Sigma(M_h) \propto M_h^{\alpha}$ down to $M_h > M_{\rm min} =10^{8.5} {\rm M_{\odot}}$, we find that $\alpha > 1$ is required to have $b_{\rm DLA}> 1.7$, implying a steeper relation or higher value of $M_{\rm min}$ than is generally predicted in numerical simulations of galaxy formation.