We examine the abundance , clustering and metallicity of Damped Lyman- \alpha Absorbers ( DLAs ) in a suite of hydrodynamic cosmological simulations using the moving mesh code AREPO .
We incorporate models of supernova and AGN feedback , as well as molecular hydrogen formation .
We compare our simulations to the column density distribution function at z = 3 , the total DLA abundance at z = 2 - 4 , the measured DLA bias at z = 2.3 and the DLA metallicity distribution at z = 2 - 4 .
Our preferred models produce populations of DLAs in good agreement with most of these observations .
The exception is the DLA abundance at z < 3 , which we show requires stronger feedback in 10 ^ { 11 - 12 } h ^ { -1 } M _ { \odot } mass halos .
While the DLA population probes a wide range of halo masses , we find the cross-section is dominated by halos of mass 10 ^ { 10 } -10 ^ { 11 } h ^ { -1 } M _ { \odot } and virial velocities 50 - 100 km s ^ { -1 } .
The simulated DLA population has a linear theory bias of 1.7 , whereas the observations require 2.17 \pm 0.2 .
We show that non-linear growth increases the bias in our simulations to 2.3 at k = 1 h \mathrm { Mpc } ^ { -1 } , the smallest scale observed .
The scale-dependence of the bias is , however , very different in the simulations compared against the observations .
We show that , of the observations we consider , the DLA abundance and column density function provide the strongest constraints on the feedback model .