Surveys for z > 3 damped Ly alpha absorption systems: The evolution of neutral gas

We have completed spectroscopic observations using LRIS on the Keck 1 teles cope of 30 very high redshift quasars, 11 selected for the presence of damp ed Ly alpha absorption systems and 19 with redshifts, z > 3.5 not previousl y surveyed for absorption systems. We have surveyed an additional 10 QSOs w ith the Lick 120" and the Anglo-Australian Telescope. We have combined thes e with previous data, resulting in a statistical sample of 646 QSOs and 85 damped Ly alpha absorbers with column densities N-Ht greater than or equal to 2 x 10(20) atoms cm(-2) covering the redshift range 0.008 less than or e qual to z less than or equal to 4.694. Four main features of how the neutra l gas in the universe evolves with redshift are evident from these data. 1. For the first time, we determine a statistically significant steepening in the column density distribution function at redshifts z > 4.0 (greater t han 99.7% confidence). The steepening of the distribution function is due t o both fewer very high column density absorbers (N-HI greater than or equal to 10(21) atoms cm(-2)) and more lower column density systems (N-HI = 2-4 x 10(20) atoms cm(-2)). 2. The frequency of very high column density absorbers (N-HI greater than o r equal to 10(21) atoms cm(-2)) reaches a peak in the redshift range 1.5 < z < 4, when the universe is 10%-30% of its present age. Although the sample size is still small, the peak epoch appears to be 3.0 less than or equal t o z less than or equal to 3.5. The highest column density absorbers disappe ar rapidly toward higher redshifts in the range z = 3.5 --> 4.7 and lower r edshifts z = 3.0 --> 0. None with column densities log N-HI greater than or equal to 21 have yet been detected at z > 4, although we have increased th e redshift path surveyed by approximate to 60%. 3. With our current data set, the comoving mass density of neutral gas, Ome ga (g), appears to peak at 3.0 < z < 3.5, but the uncertainties are still t oo large to determine the precise shape of Omega (g). The statistics are co nsistent with a constant value of Omega (g), for 2 < z < 4. There is still tentative evidence for a drop-off at,- > 4, as indicated by earlier data se ts. If we define R-g* drop Omega (g)/Omega (*), where R-g is the ratio of t he peak value of Omega (g) to Omega (*), the mass density in galaxies in th e local universe, we find values of R-g* = 0.25-0.5 at,- - 3, depending on the cosmology. For an Omega = 1 universe with a zero cosmological constant, R-g* = 0.25-0.5. For an Omega = 1 universe with a positive cosmological co nstant (Omega (Lambda) = 0.7, Omega (M) = 0.3), we find R-g* approximate to 0.25. For a universe with Omega (Lambda) = 0 and Omega (M) = 0.3, we find R-g* approximate to 0.3. 4. Omega (g) decreases with redshift for the interval z = 3.5 --> 0.008 for our data set, but we briefly discuss new results from Rao & Turnshek for 3 < 1.5 that suggest that <Omega>(g) (z < 1.5) may be higher than previously determined. To make the data in our statistical sample more readily available for compa rison with scenarios from. various cosmological models, we provide tables t hat include all 646 QSOs from our new survey and previously published surve ys. They list the minimum and maximum redshift defining the redshift path a long each line of sight, the QSO emission redshift, and when an absorber is detected, the absorption redshift and measured H I column density.