COSMOLOGICAL EVOLUTION OF ATOMIC GAS AND IMPLICATIONS FOR 21 cm H I ABSORPTION

Galaxy disks are shown to contain a significant population of atomic clouds of 100 pc linear size which are self-opaque in the 21 cm transition. These objects have H I column densities as high as 10(23) cm(-2) and contribute to a global opacity correction factor of 1.34 +/- 0.05 that applies to the integrated 21 cm emission to obtain a total H I mass estimate. High-resolution, opacity-corrected images of the nearest external galaxies have been used to form a robust redshift zero distribution function of H I, f (N-H (I), X, z = 0), the probability of encountering a specific H I column density along random lines of sight per unit comoving distance. This is contrasted with previously published determinations of f (N-H (I), X) at z = 1 and 3. A systematic decline of moderate column density (18 < log(N-H (I)) < 21) H I is observed with decreasing redshift that corresponds to a decline in surface area of such gas by a factor of five since z = 3. The number of equivalent Damped Lyman Alpha absorbers (log(N-H (I)) > 20.3) has also declined systematically over this redshift interval by a similar amount, while the cosmological mass density in such systems has declined by only a factor of two to its current, opacity-corrected value of Omega(H) (DLA)(I) (z = 0) = 5.4 +/- 0.9 x 10(-4). We utilize the tight but strongly nonlinear dependence of 21 cm absorption opacity on column density at z = 0 to transform our high-resolution H I images into ones of 21 cm absorption opacity. These images are used to calculate distribution and pathlength functions of integrated 21 cm opacity. We suggest that that this z = 0 calibration may also apply at higher redshift. In this case, the incidence of deep 21 cm absorption systems is predicted to show very little evolution with redshift, while that of faint absorbers should decline by a factor of 5 between z = 3 and the present. We explicitly consider the effects of H I absorption against background sources that are extended relative to the 100 pc intervening absorber size scale. Extended background sources result in dramatically altered distribution and pathlength functions which are insensitive to the predicted redshift evolution. Future surveys of 21 cm absorption will require very high angular resolution, of about 15 mas, for their unambiguous interpretation.