f81b8aae02a265af.tex
1: \begin{abstract}
2: We use high resolution Eulerian hydrodynamics simulations to study
3: kinematic properties of the low ionization species in damped
4: Ly$\alpha$ systems at redshift $z=3$. Our adaptive mesh refinement
5: simulations include most key ingredients relevant for modeling neutral
6: gas in high-column density absorbers: hydrodynamics, gravitational
7: collapse, continuum radiative transfer above the hydrogen Lyman limit
8: and gas chemistry, but no star formation. We model high-resolution
9: Keck spectra with unsaturated low ion transitions in two Si II lines
10: ($1526$ and $1808\AA$), and compare simulated line profiles to the
11: data from the SDSS DLA survey.
12: 
13: We find that with increasing grid resolution the models show a trend
14: in convergence towards the observed distribution of HI column
15: densities. While in our highest resolution model we recover the
16: cumulative number of DLAs per unit absorption distance, none of our
17: models predicts DLA velocity widths as high as indicated by the data,
18: suggesting that feedback from star formation might be important. At
19: $z=3$ a non-negligible fraction of DLAs with column densities below
20: $10^{21}\cm^{-2}$ is caused by filamentary structures in more massive
21: halo environments. Lower column density absorbers with $N_{\rm HI} <
22: 10^{21.4}\cm^{-2}$ are sensitive to changes in the UV background
23: resulting in a $10\%$ reduction of the cumulative number of DLAs for
24: twice the quasar background relative to the fiducial value, and nearly
25: a $40\%$ reduction for four times the quasar background. We find that
26: the mass cut-off below which a large fraction of dwarf galaxies cannot
27: retain gas after reionization is $\sim 7\times 10^7\msun$, lower than
28: the previous estimates. Finally, we show that models with
29: self-shielding commonly used in the literature produce significantly
30: lower DLA velocity widths than the full radiative transfer runs which
31: essentially render these self-shielded models obsolete.
32: \end{abstract}
33: