|Annu. Rev. Astron. Astrophys. 2002. 40:487-537
Copyright © 2002 by Annual Reviews. All rights reserved
Weinberg (1977) observed that “the theory of the formation of galaxies is one of the great outstanding problems of astrophysics, a problem that today seems far from solution.” Although the past two decades have seen considerable progress, Weinberg's assessment remains largely true.
Eggen, Lynden-Bell and Sandage (1962; ELS) were the first to show that it is possible to study galactic archaeology using stellar abundances and stellar dynamics; this is probably the most influential paper on the subject of galaxy formation. ELS studied the motions of high velocity stars and discovered that, as the metal abundance decreases, the orbit energies and eccentricities of the stars increased while their orbital angular momenta decreased. They inferred that the metal-poor stars reside in a halo that was created during the rapid collapse of a relatively uniform, isolated protogalactic cloud shortly after it decoupled from the universal expansion. ELS are widely viewed as advocating a smooth monolithic collapse of the protocloud with a timescale of order 108 years. But Sandage (1990) stresses that this is an over-interpretation; a smooth collapse was not one of the inferences they drew from the stellar kinematics.
In 1977, the ELS picture was challenged by Searle (see also Searle & Zinn 1978) who noted that Galactic globular clusters have a wide range of metal abundances essentially independent of radius from the Galactic Center. They suggested that this could be explained by a halo built up over an extended period from independent fragments with masses of ∼ 108 M⊙. In contrast, in the ELS picture, the halo formed in a rapid free-fall collapse. But halo field stars, as well as globular clusters, are now believed to show an age spread of 2−3 Ga (Marquez & Schuster 1994); for an alternative view, see Sandage & Cacciari (1990). The current paradigm, that the observations argue for a halo that has built up over a long period from infalling debris, has developed after many years of intense debate.
This debate parallelled the changes that were taking place in theoretical studies of cosmology (e.g., Peebles 1971, Press & Schechter 1974). The ideas of galaxy formation via hierarchical aggregation of smaller elements from the early universe fit in readily with the Searle & Zinn view of the formation of the galactic halo from small fragments. The possibility of identifying debris from these small fragments was already around in Eggen's early studies of moving groups, and this is now an active field of research in theoretical and observational stellar dynamics. It offers the possibility to reconstruct at least some properties of the protogalaxy and so to improve our basic understanding of the galaxy formation process.
We can extend this approach to other components of the Galaxy. We will argue the importance of understanding the formation of the galactic disk, because this is where most of the baryons reside. Although much of the information about the properties of the protogalactic baryons has been lost in the dissipation that led to the galactic disk, a similar dynamical probing of the early properties of the disk can illuminate the formation of the disk, at least back to the epoch of last significant dissipation. It is also clear that we do not need to restrict this probing to stellar dynamical techniques. A vast amount of fossil information is locked up in the detailed stellar distribution of chemical elements in the various components of the Galaxy, and we will discuss the opportunities that this offers.
We are coming into a new era of galactic investigation, in which one can study the fossil remnants of the early days of the Galaxy in a broader and more focussed way, not only in the halo but throughout the major luminous components of the Galaxy. This is what we mean by The New Galaxy. The goal of these studies is to reconstruct as much as possible of the early galactic history. We review what has been achieved so far, and point to the future.
1.2. Near-Field and Far-Field Cosmology
What do we mean by the reconstruction of early galactic history? We seek a detailed physical understanding of the sequence of events which led to the Milky Way. Ideally, we would want to tag (i.e., associate) components of the Galaxy to elements of the protocloud – the baryon reservoir which fueled the stars in the Galaxy.
From theory, our prevailing view of structure formation relies on a hierarchical process driven by the gravitational forces of the large-scale distribution of cold, dark matter (CDM). The CDM paradigm provides simple models of galaxy formation within a cosmological context (Peebles 1974, White & Rees 1978, Blumenthal et al. 1984). N-body and semi-analytic simulations of the growth of structures in the early universe have been successful at reproducing some of the properties of galaxies. Current models include gas pressure, metal production, radiative cooling and heating, and prescriptions for star formation.
The number density, properties and spatial distribution of dark matter halos are well understood within CDM (Sheth & Tormen 1999, Jenkins et al. 2001). However, computer codes are far from producing realistic simulations of how baryons produce observable galaxies within a complex hierarchy of dark matter. This a necessary first step towards a viable theory or a working model of galaxy formation.
In this review, our approach is anchored to observations of the Galaxy, interpreted within the broad scope of the CDM hierarchy. Many of the observables in the Galaxy relate to events which occurred long ago, at high redshift. Figure 1 shows the relationship between look-back time and redshift in the context of the ΛCDM model: The redshift range (z ≲ 6) of discrete sources in contemporary observational cosmology matches closely the known ages of the oldest components in the Galaxy. The Galaxy (near-field cosmology) provides a link to the distant universe (far-field cosmology).
Figure 1.Look-back time as a function of redshift and the size of the Universe (Lineweaver 1999) for five different world models. The approximate ages of the Galactic halo and disk are indicated by hatched regions.
Before we embark on a detailed overview of the relevant data, we give a descriptive working picture of the sequence of events involved in galaxy formation. For continuity, the relevant references are given in the main body of the review where these issues are discussed in more detail.
1.3. A Working Model of Galaxy Formation
Shortly after the Big Bang, cold dark matter began to drive baryons towards local density enhancements. The first stars formed after the collapse of the first primordial molecular clouds; these stars produced the epoch of reionization. The earliest recognizable protocloud may have begun to assemble at about this time.
Within the context of CDM, the dark halo of the Galaxy assembled first, although it is likely that its growth continues to the present time. In some galaxies, the first episodes of gas accretion established the stellar bulge, the central black hole, the first halo stars and the globular clusters. In the Galaxy and similar systems, the small stellar bulge may have formed later from stars in the inner disk.
The early stages of the Galaxy's evolution were marked by violent gas dynamics and accretion events, leading to the high internal densities of the first globular clusters, and perhaps to the well-known black hole mass−stellar bulge dispersion relation. The stellar bulge and massive black hole may have grown up together during this active time. We associate this era with the Golden Age, the phase before z ∼ 1 when star formation activity and accretion disk activity were at their peak.
At that time, there was a strong metal gradient from the bulge to the outer halo. The metal enrichment was rapid in the core of the Galaxy such that, by z ∼ 1, the mean metallicities were as high as [Fe/H] ∼ -1 or even higher. In these terms, we can understand why the inner stellar bulge that we observe today is both old and moderately metal rich. The first halo stars ([Fe/H] ≈ -5 to -2.5) formed over a more extended volume and presumably date back to the earliest phase of the protocloud. The first globular clusters formed over a similar volume from violent gas interactions ([Fe/H] ≈ -2.5 to -1.5). We believe now that many of the halo stars and globulars are remnants of early satellite galaxies which experienced independent chemical evolution before being accreted by the Galaxy.
The spread in [Fe/H], and the relative distribution of the chemical elements, is a major diagnostic of the evolution of each galactic component. If the initial mass function is constant, the mean abundances of the different components give a rough indication of the number of SN II enrichments which preceded their formation, although we note that as time passes, an increasing fraction of Fe is produced by SN Ia events. For a given parcel of gas in a closed system, only a few SN II events are required to reach [Fe/H] ≈ -3, 30 to 100 events to get to [Fe/H] ≈ -1.5, and maybe a thousand events to reach solar metallicities. We wish to stress that [Fe/H] is not a clock: Rather it is a measure of supernova occurrences and the depth of the different potential wells that a given parcel of gas has explored.
During the latter stages of the Golden Age, most of the baryons began to settle to a disk for the first time. Two key observations emphasize what we consider to be the mystery of the main epoch of baryon dissipation. First, there are no stars with [Fe/H] < -2.2 that rotate with the disk. Second, despite all the activity associated with the Golden Age, at least 80% of the baryons appear to have settled gradually to the disk over many Ga; this fraction could be as high as 95% if the bulge formed after the disk.
About 10% of the baryons reside in a thick disk which has [Fe/H] ≈ -2.2 to -0.5, compared to the younger thin disk with [Fe/H] ≈ -0.5 to +0.3. It is striking how the globular clusters and the thick disk have similar abundance ranges, although the detailed abundance distributions are different. There is also a similarity in age: Globular clusters show an age range of 12 to 14 Ga, and the thick disk appears to be at least 12 Ga old. Both the thick disk and globulars apparently date back to the epoch of baryon dissipation during z ∼ 1-5.
Figure 2 summarizes our present understanding of the complex age-metallicity distribution for the various components of the Galaxy.
Figure 2. The age-metallicity relation of the Galaxy for the different components (see text): TDO – thin disk open clusters; TDG – thick disk globulars; B – bulge; YHG – young halo globulars; OHG – old halo globulars. The blue corresponds to thin disk field stars, the green to thick disk field stars, and the black shows the distribution of halo field stars extending down to [Fe/H] = -5.
It is a mystery that the thick disk and the globulars should have formed so early and over such a large volume from material that was already enriched to [Fe/H] ∼ -2. Could powerful winds from the central starburst in the evolving core have distributed metals throughout the inner protocloud at about that time?
Finally, we emphasize again that 90% of the disk baryons have settled quiescently to the thin disk since z ∼ 1.
1.4. Timescales and Fossils
The oldest stars in our Galaxy are of an age similar to the look-back time of the most distant galaxies in the Hubble Deep Field (Figure 1). For the galaxies, the cosmological redshift measured from galaxy spectra presently takes us to within 5% of the origin of cosmic time. For the stars, their upper atmospheres provide fossil evidence of the available metals at the time of formation. The old Galactic stars and the distant galaxies provide a record of conditions at early times in cosmic history, and both harbor clues to the sequence of events that led to the formation of galaxies like the Milky Way.
The key timescale provided by far-field cosmology is the look-back time with the prospect of seeing galaxies at an earlier stage in their evolution. However, this does not imply that these high-redshift objects are unevolved. We know that the stellar cores of galaxies at the highest redshifts (z ∼ 5) observed to date exhibit solar metallicities, and therefore appear to have undergone many cycles of star formation (Hamann & Ferland 1999). Much of the light we detect from the early universe probably arises from the chemically and dynamically evolved cores of galaxies.
Near-field cosmology provides a dynamical timescale, τD ∼ (Gρ)−1/2, where ρ is the mean density of the medium. The dynamical timescale at a radial distance of 100 kpc is of order several Ga, so the mixing times are very long. Therefore, on larger scales, we can expect to find dynamical and chemical traces of past events, even where small dynamical systems have long since merged with the Galaxy.
We note that the CDM hierarchy reflects a wide range of dynamical timescales, such that different parts of the hierarchy may reveal galaxies in different stages of evolution. In this sense, the hierarchy relates the large-scale density to the morphology and evolution of its individual galaxies; this is the so-called morphology-density relation (Dressler 1980, Hermit et al. 1996, Norberg et al. 2001). Over a large enough ensemble of galaxies, taken from different regions of the hierarchy, we expect different light-weighted age distributions because one part of the hierarchy is more evolved than another. In other words, the evolution of small-scale structure (individual galaxies) must at some level relate to the environment on scales of 10 Mpc or more.
The near field also provides important evolutionary timescales for individual stars and groups of stars (see “Stellar Age Dating” below). Individual stars can be dated with astero-seismology (Christensen-Dalsgaard 1986, Gough 2001) and nucleo-cosmochronology (Fowler & Hoyle 1960, Cowan et al. 1997). Strictly speaking, nucleo-cosmochronology dates the elements rather than the stars. Coeval groups of stars can be aged from the main-sequence turn-off or from the He-burning stars in older populations (Chaboyer 1998). Furthermore, the faint end cut-off of the white dwarf luminosity function provides an important age constraint for older populations (Oswalt et al. 1996). Presently, the aging methods are model dependent.
1.5. Goals of Near-Field Cosmology
We believe that the major goal of near-field cosmology is to tag individual stars with elements of the protocloud. Some integrals of motion are likely to be preserved while others are scrambled by dissipation and violent relaxation. We suspect that complete tagging is impossible. However, some stars today may have some integrals of motion that relate to the protocloud at the epoch of last dissipation (see “Zero Order Signatures – Information Preserved Since Dark Matter Virialized” below).
As we review, different parts of the Galaxy have experienced dissipation and phase mixing to varying degrees. The disk, in contrast to the stellar halo, is a highly dissipated structure. The bulge may be only partly dissipated. To what extent can we unravel the events that produced the Galaxy as we see it today? Could some of the residual inhomogeneities from prehistory have escaped the dissipative process at an early stage?
Far-field cosmology currently takes us back to the epoch of last scattering as seen in the microwave background. Cosmologists would like to think that some vestige of information has survived from earlier times (compare Peebles et al. 2000). In the same spirit, we can hope that fossils remain from the epoch of last dissipation, i.e., the main epoch of baryon dissipation that occurred as the disk was being assembled.
To make a comprehensive inventory of surviving inhomogeneities would require a vast catalog of stellar properties that is presently out of reach (Bland-Hawthorn 2002). The Gaia space astrometry mission (Perryman et al. 2001), set to launch at the end of the decade, will acquire detailed phase space coordinates for about one billion stars, within a sphere of diameter 20 kpc (the Gaiasphere). In “The Gaiasphere and The Limits of Knowledge” below, we look forward to a time when all stars within the Gaiasphere have complete chemical abundance measurements (including all heavy metals). Even with such a vast increase in information, there may exist fundamental – but unproven – limits to unravelling the observed complexity.
The huge increase in data rates from ground-based and space-based observatories has led to an explosion of information. Much of this information from the near field is often dismissed as weather or unimportant detail. But in fact fundamental clues are already beginning to emerge. In what is now a famous discovery, a large photometric and kinematic survey of bulge stars revealed the presence of the disrupting Sgr dwarf galaxy (Ibata et al. 1994), now seen over a large region of sky and in a variety of populations (see “Structures in Phase Space” below). Perhaps the most important example arises from the chemical signatures seen in echelle spectroscopy of bulge, thick disk and halo stars. In “Epilogue: Challenges for the Future,” we envisage a time when the analysis of thousands of spectral lines for a vast number of stars will reveal crucial insights into the sequence of events early in the formation of the Galaxy.
In this review, we discuss fossil signatures in the Galaxy. A key aspect of fossil studies is a reliable time sequence. In “Stellar Age Dating,” we discuss methods for age-dating individual stars and coeval groups of stars. In “Structure of the Galaxy,” we describe the main components of the Galaxy. In “Signatures of Galaxy Formation,” we divide the fossil signatures of galaxy formation into three parts: zero order signatures that preserve information since dark matter virialized; first order signatures that preserve information since the main epoch of baryon dissipation; and second order signatures that arise from major processes involved in subsequent evolution. In “The Gaiasphere and The Limits of Knowledge,” we look forward to a time when it is possible to measure ages, phase space coordinates and chemical properties for a vast number of stars in the Galaxy. Even then, what are the prospects for unravelling the sequence of events that gave rise to the Milky Way? We conclude with some experimental challenges for the future.