Evolution of the HI Content

 
Figure: Cosmological density of neutral gas, incidence of CIV absorption, and comoving density of luminous QSOs as a function of redshift from QSO absorption-line statistics. Top panel. Mean cosmological density of neutral gas, , normalized to the critical density (Storrie-Lombardi et al 1996; Rao et al 1995 (z=0); Storrie-Lombardi & Wolfe priv. comm.) Middle panel. Number of CIV metal-line absorption systems per unit redshift, n(z) (Steidel 1990); z=0.3 point from Bahcall et al 1993). Filled points from Steidel indicate rest frame equivalent widths Å; open points are for Å. Hatched areas indicate the range () for unevolving cross sections since z=1.5, beyond which redshift CIV can be measured with ground-based telescopes. Bottom panel. Comoving density of optically selected QSOs: filled squares from Schmidt et al 1994; open circles from Hewitt et al 1993). km sMpc,

A measurement of the total HI content contained in nearby galaxies and HI clouds is an important constraint in the bigger picture of galaxy evolution on cosmological time scales. Although the neutral gas in large galaxies at present is often considered to be a minor component that is used as tracer for kinematics or as a dwindling source of fuel for star formation, there is now strong evidence from the studies of QSO absorption lines that the present HI content of the universe is but a fraction of what it was at redshift . This finding comes from the statistics of the ``damped Lyman-'' class of absorption line that is identified with dynamically cold layers akin to the disks of familiar nearby spiral galaxies (Wolfe et al 1986, Lanzetta et al 1995, Storrie-Lombardi et al 1996). Figure 1 summarizes the presently available measurements of the neutral gas content as a function of z (Storrie-Lombardi et al 1996), along side a plot of the incidence of CIV absorption lines, which provide an indication of the cross section presented by clouds of ionized, metal-rich gas (Steidel 1990, Bahcall et al 1993), and the comoving density of luminous optically selected QSOs (Schmidt et al 1994, Hewett et al 1993). Taken literally, the evidence points to an epoch around when the neutral gas mass density, , hit a maximum, at roughly the same time that the CIV cross section for strong absorption lines ( Å) began a sharp increase. This is also a time when luminous QSOs were most abundant. These indicators testify that we are seeing substantial redistribution of gas, as witnessed by the formation of ionized metal-rich galaxy halos and the efficient fueling of active galactic nuclei. The surge in neutral gas content indicates that protogalactic gravitational potentials were deep enough that gas was confined to sufficiently high density that it was at least momentarily immune to ionization by star bursts and ionizing background radiation. This may be the epoch that disk galaxies formed as secondary infall of gas occurred into galactic potentials formed in the first round of galactic bulge formation. The disk formation would be accompanied by halo enrichment, either by in situ star formation or by metal pollution of the extended halo region by winds from the new star forming regions of the disk. An alternative view is that this is also likely to be an epoch when small protogalactic lumps are merging vigorously, and star burst within the lumps would be effective at ejecting metals into an extended region that would at later times constitute the ``halo'' region of the merger product.

Figure 1 summarizes only the neutral atomic and ionized gas components. A complete balance requires an accounting for all the universe's baryons, as gas is exchanged between neutral, ionized, and molecular phases, as well as the path of stellar evolution leading to the current state where far more baryons are contained in stars than in neutral gas. Although the present HI content is only % of the mass in stars, there was a period at when the HI content apparently was of order half of the present stellar mass. Estimates of the mass content in ionized halos suggest that they probably contain about ten times the mass contained in the damped Lyman- absorbers at their peak (Petitjean et al 1993). This interpretation of the CIV data relies on theoretical modeling, with large uncertainties due to ionization level and carbon abundance. It is striking that the recent HST observations (Bahcall et al 1993) are consistent with no evolution in the absorption cross section presented by high column density CIV systems from (Steidel 1990) to , implying that large quantities of ionized gas may still be present, either in the form of extended halos or in intergalactic clouds whose mass could far exceed the visible stellar mass in galaxies. It has been argued that the CIV systems, together with the more diffuse clouds of the Lyman- forest, may contain a substantial fraction of the baryon content of the Universe (Rauch & Haehnelt 1995, Haehnelt et al 1996). The hydrogen neutral fraction of these ionized clouds would provide column densities of HI well below the regime ordinarily probed by 21cm line observations. At present, the molecular gas mass content of galaxies appears to be roughly equal to the neutral atomic mass (cf. Kennicutt et al 1994).

 
Figure 2: Star formation rate (SFR) and density of neutral gas as a function of time compared with . Bottom panel from references in Figure 1; from Lanzetta et al 1995. Rising dashed curve is KTC model for increasing stellar mass; dot-dash is KTC model for declining cold gas content, adjusted as described in text to indicate only the atomic fraction. The cross hatched band indicates the range of models proposed by Pei and Fall for the true , corrected for selection effects caused by dusty damped Lyman- absorbers. Top panel Relative star formation rate for models by Lanzetta et al (1995) (dotted), KTC (dashed), and Pei and Fall (1995) (solid: I = model with infall; C,O = ``closed box'' and outflow models).

Surveys of nearby volumes in the 21 cm line are important in anchoring the z=0 point of Figure 1. A complete inventory of neutral gas in the nearby Universe, of the sort that is being provided by unbiased 21cm line surveys, will have tight error bars and thus will carry large statistical weight in models that describe the evolution of . Note that the statistical errors of the z>1.6 measurements of are so large that these high z points are consistent with no evolution at all. In principle, the low z regime (0 < z < 1.6) of the diagram can be measured using QSO absorption line methods in much the same way as the high z points. However, the observations are difficult since the Lyman- line is not redshifted into the optical window until , so the low z absorption line work must be done from space observatories such as IUE and HST (Lanzetta et al 1995, Rao et al 1995).

When the data points are plotted as a function of time, as in Figure 2, it is clear that the single low z QSO absorption line point applies to a time span well longer than half the age of the Universe. Several selection effects make it difficult to obtain a reliable sampling of damped Lyman- absorbers at low z. Cosmological factors, as well as the apparent shrinking of damped Lyman- absorption cross section, , with increasing age of the Universe (Lanzetta et al 1995), act to make them very rare at recent times: , where k=2 for and k=3 for . There are also selection effects that are likely to influence the QSO absorption line measurements and that affect the low z measurement most strongly: (1) The presence of dust in disk galaxies is expected to become increasingly important as they evolve and may act to selectively attenuate the light from QSOs behind them, causing these lines of sight to be under represented in QSO samples (Fall & Pei 1989, Pei & Fall 1995, Webster et al 1995) although this view is contested (Boyle & Di Matteo 1995). (2) Gravitational lensing may act to selectively amplify background objects into QSO samples, but also may bend the light path so as to dodge the high HI column densities of the disk (Bartelmann & Loeb 1996, Smette et al 1995a,1996).

Figure 2 includes curves to indicate trends in stellar evolution relative to the decline in with time. Recent analyses of the depletion of the neutral gas content of galaxies over cosmological times due to star formation have been presented by Lanzetta et al (1995) and Pei and Fall (1996). A related study by Kennicutt et al (1994, KTC) addresses the prolonging of the current star formation rate in disks due to delayed gas return as stellar populations age. An example of the KTC models is presented in Figure 2 to illustrate both the rise of stellar mass with time and decline of neutral gas content with time for a disk system without added inflow or allowing mass to escape. For this display, the KTC model has been scaled so that the final stellar mass and the final HI mass are consistent with the observations at ; the relative proportion of HI to H has been adjusted for this display to vary linearly with time from at high z to 2/3 at the present time. The slope of the KTC model at the time marked ``Now'' in Figure 2 can be compared with the steeper slope drawn to indicate the rate at which the current star formation rate would consume the present atomic hydrogen content, exhausting the supply in only 1.5 Gyr if there were no additional reservoirs of molecular gas or contribution from delayed stellar return.

The models of Lanzetta et al require a balance between the decline of and a rise in stellar mass; along with assumptions that stellar return occurs instantaneously as each generation of stars is formed, this constrains the mean star formation rate history of the Universe. In this picture, the star formation rate as a function of time has a surge at early epochs (see top panel in Figure 2) when little metal enrichment has occurred. A consequence is that their models produce uncomfortably large numbers of stars at with low metal abundances. Pei and Fall suggest that this problem can be solved with a family of dusty models, which imply that the damped Lyman- statistics drastically underestimate the neutral gas content at all redshifts. In Pei and Falls' models, the peak of star formation rate is delayed to an epoch at 1 to 2 when the ISM has acquired higher metal abundances. The KTC models that are tuned to describe the ecology of large galaxy disks have a more nearly uniform, gently declining star formation rate. The current generation of radio telescopes is not sensitive enough to resolve this controversy by simply looking back to measure the HI density at 1/2. On the other hand, choosing complete samples of radio selected high z quasars for background probes would remove possible selection effects by dust.

At z=0, recent 21 cm line measurements are indicating that the bulk of the atomic hydrogen content of the nearby universe is bound into galaxies with optical counterparts (Zwaan et al 1996, Schneider 1996, Szomoru et al 1994, Henning 1995, Briggs 1990). Furthermore the normalization of the HI mass function seems to be well understood (cf. Zwaan et al 1996, Rao & Briggs 1993), although there is still concern over the normalization of even the optically determined luminosity function (cf Ellis et al 1996, Glazebrook et al 1995). Clearly the determination of the integral HI content of the local universe is a measurement that the Parkes Multibeam Survey will clarify since it will be complete, unbiased by extinction and optical surface brightness, and will have well understood sensitivity limits.

© Copyright Astronomical Society of Australia 1997