Annu. Rev. Astron. Astrophys. 1997. 35: 389-443
Copyright © 1997 by . All rights reserved

Next Contents Previous

6. THE STAR-FORMATION HISTORY OF FIELD GALAXIES

In the past year, there has been a remarkable synthesis of the star-formation history of field galaxies (as delineated by the observations reviewed here) and the gaseous and chemical evolution of the intergalactic gas (as delineated by the studies of QSO absorption lines). Lanzetta et al (1995), Wolfe et al (1995), and more recently Storrie-Lombardi et al (1996) analyzed the evolution of the cosmic density of neutral hydrogen with redshift via various QSO absorber samples. These data locate a redshift of z = 2-3, where the bulk of the present star-formation density is seen in neutral hydrogen clouds. Likewise, the chemical evolution of metallic clouds has been studied with redshift by Pettini et al (1994). Pei & Fall (1995), Fall et al (1996) have shown, via a remarkably simple model, how these various data can be reconciled with the history of the volume-averaged star formation. Madau et al (1996), Madau (1997) have updated this analysis with the most recent estimates of the high redshift SFR discussed in Section 5 (Figure 10). The emerging picture points to a redshift range of 1-2, in which the bulk of the present-day stellar population was assembled and perhaps most of the present-day metals produced (Cowie 1988, Songaila et al 1990, Ellis 1996b).

Figure 10

Figure 10. Element and star-formation history from the analysis of Madau (1997). Data points provide a measurement or a lower limit to the universal metal ejection rate (left ordinate) and total star-formation density (right ordinate). Triangle: local Halpha survey of Gallego et al (1995). Filled dots: CFRS redshift survey Lilly et al (1996). Filled squares: Lyman limit galaxies observed in the HDF. The dashed line depicts the fiducial rate equivalent to the mass density of local metals divided by the present age of the universe (Omega = 1, H0 = 50 kms s -1 Mpc-1).

Appealing as this picture is, not least because it has much theoretical support (Baugh et al 1996, Kauffmann et al 1996, White 1996), it rests on preliminary observations and relies on the connection of disparate data sets, each using a different indicator of star formation and each selecting only some detectable subset of the overall population. At low z, is the star-formation density derived from the Halpha surveys (Gallego et al 1995) affected by fair sample problems that have been invoked to address the apparent low normalization of the local LF (Section 4.1)? At modest redshift, the rapid increase in the integrated star-formation density (Lilly et al 1996, Cowie et al 1996) relies on the conversion of blue light or [O II] line strength to the SFR and, quantitatively, via the extrapolation of the contribution to sources whose luminosities are fainter than the redshift survey limit. These uncertainties may affect the rate of increase with redshift beyond z = 0.8 (Lilly et al 1996). It will therefore be important to extend the redshift surveys fainter even at z < 1 to verify the extrapolation, as well as to use alternative diagnostics at moderate redshift such as Halpha fluxes secured via infrared spectroscopy. There are also promising new approaches to constrain the intermediate z SFR based entirely on emission line searching (Meisenheimer et al 1996). Similarly, at high z, the LF and spatial distribution of Lyman limit-selected samples needs to be explored via panoramic surveys based on the highly successful pilot studies (Steidel et al 1996c). The star formation densities derived from these data must be regarded as lower limits until the effects of dust have been properly explored.

Despite these caveats, the trend is highly encouraging. A clear gap emerges, however, in our knowledge in the redshift range between that reviewed here at z < 1 and that revealed at z > 2.3 in the HST work. This has motivated the construction of a new generation of ground-based infrared spectrographs free from OH background light (Iwamuro et al 1994, LeFevre et al 1996c, Taylor & Colless 1996, Piché et al 1997), which aim to survey this difficult region in conjunction with optical-UV imaging. Together with NICMOS on HST, which should have powerful background-limited capabilities in both deep imaging and low-resolution grism spectroscopic modes, the intermediate redshift range can be systematically explored in a way that was highly successful for the z < 1 population.

The nonzero metallicity of the highest redshift absorbing clouds (Cowie 1996) also points to an earlier era of modest star formation possibly associated with the small but convincing population of high z galaxies already known to contain old stars (Dunlop et al 1996, Stockton et al 1995). The detection of high redshift sources at far infrared and submillimeter wavelengths (Omont et al 1996) together with the successful deployment of the Infrared Space Observatory and the SCUBA submillimeter array detector (Gear & Cunningham 1995) augurs well for surveying the z > 4 universe for earlier eras of star formation. This will remain an important observational challenge even if the bulk of the star-formation activity is convincingly demonstrated to occur at z = 1-2. To physically understand the processes that lead to galaxy formation, exploration of the high z tail is surely necessary.

Tremendous progress has been made in observational cosmology. The subject of galaxy formation and evolution has moved firmly from the realm of theoretical speculation into that of systematic observation. Some lessons are, however, being learned. Perhaps the most important of these is the dangers of relying purely on morphology and star-formation diagnostics to connect what may, in fact, be very different populations observed via different techniques at high and low z. Clearly we seek more representative physical parameters to subclassify the data sets over a range of look-back times in order to test detailed hypotheses. A further hindrance is the absence of well-defined local data of the kind needed for detailed comparisons with the high z samples (Koo & Kron 1992). However, notwithstanding the formidable challenges of studying the distant universe, the combination of deeper redshift surveys and morphologies from HST has demonstrated quantitatively the presence of rapid evolution in a subset of the population to z = 1. The absence of a dominant population of star-forming galaxies at z = 3 and the small physical sizes of the faint HDF images delineate a simple picture that is consistent with hierarchical galaxy formation and knowledge of the properties of intervening gas clouds as studied in QSO absorption lines. Galactic history seems to have been remarkably recent, which can only be our good fortune, given the power of our new facilities to observe these eras in considerable detail. The observational picture is already emerging very rapidly, but much work and ingenuity will be needed to identify the physical processes that drive the evolutionary trends now revealed.


Acknowledgments

I thank my collaborators, colleagues, and visitors at Cambridge, particularly Roberto Abraham, Matthew Colless, George Efstathiou, Masataka Fukugita, Simon Lilly, and Max Pettini for their critical and helpful comments on this review. Many workers sent detailed accounts of their views on the sensitive and complex issues discussed in Section 4. Special thanks are due to Emmanuel Bertin, Ray Carlberg, Stephane Charlot, Len Cowie, Julianne Dalcanton, Simon Driver, Harry Ferguson, Luigi Guzzo, Karl Glazebrook, David Hartwick, David Koo, Huan Lin, Stacy McGaugh, Nigel Metcalfe, John Peacock, Tom Shanks, and Elena Zucca. I apologize to these and others if I have failed to represent their particular viewpoint fairly. I thank Bernard Sadoulet, Joe Silk, and Tom Broadhurst for their hospitality and support in Berkeley during the summer of 1996, when much of this review was written. Finally, I thank Allan Sandage for his numerous suggestions and encouragement.

Next Contents Previous