|Annu. Rev. Astron. Astrophys. 2002. 40:487-537
Copyright © 2002 by Annual Reviews. All rights reserved
The ultimate goal of cosmology, both near and far, must be to explain how the Universe has arrived at its present state. It is plausible – although difficult to accept – that nature provides fundamental limits of knowledge, in particular, epochs where the sequence of events are scrambled. Our intuition is that any phase dominated by relaxation or dissipation probably removes more information than it retains.
But could some of the residual inhomogeneities from prehistory have escaped the dissipative process at an early stage? We may not know the answer to this question with absolute certainty for many years. In the absence of certainty, we consider what might be the likely traces of a bygone era prior to the main epoch of baryon dissipation.
5.2. Chemical Signatures
A major goal of near-field cosmology is to tag or to associate individual stars with elements of the protocloud. For many halo stars, and some outer bulge stars, this may be possible with phase space information provided by Gaia. But for much of the bulge and the disk, secular processes cause the populations to become relaxed (i.e., the integrals of motion are partially randomized). In order to have any chance of unravelling disk formation, we must explore chemical signatures in the stellar spectrum. Ideally, we would like to tag a large sample of representative stars with a precise time and a precise site of formation.
Over the past four decades, evidence has gradually accumulated (Figure 11) for a large dispersion in metal abundances [Xi/Fe] (particularly n-capture elements) in low metallicity stars relative to solar abundances (Wallerstein et al. 1963, Pagel 1965, Spite & Spite 1978, Truran 1981, Luck & Bond 1985, Clayton 1988, Gilroy et al. 1988, McWilliam et al. 1995, Norris et al. 1996, Burris et al. 2000). Elements like Sr, Ba and Eu show a 300-fold dispersion, although [α/Fe] dispersions are typically an order of magnitude smaller.
Figure 11. Mean relative abundance ratios of light s-process elements (top panel), heavy s-process elements (middle panel), and r-process elements (bottom panel) as functions of [Fe/H]. In each panel, the dotted horizontal lines represent the solar abundance ratios of these elements. The references for the data points are given in Wallerstein et al. (1997). (We acknowledge C. Sneden for this figure.)
In their celebrated paper, Burbidge et al. (1957 – B2FH) demonstrated the likely sites for the synthesis of slow (s) and rapid (r) n-capture elements. The s-process elements (e.g., Sr, Zr, Ba, Ce, La, Pb) are thought to arise from the He-burning phase of intermediate to low mass (AGB) stars (M < 10 M⊙), although at the lowest metallicities, trace amounts are likely to arise from high mass stars (Burris et al. 2000, Rauscher et al. 2001).
In contrast to the s-process elements, the r-process elements (e.g., Sm, Eu, Gd, Tb, Dy, Ho) cannot be formed during quiescent stellar evolution. While some doubts remain, the most likely site for the r-process appears to be SN II, as originally suggested by B2FH (see also Wallerstein et al. 1997). Therefore, r-process elements measured from stellar atmospheres reflect conditions in the progenitor cloud. In support of Gilroy et al. (1988), McWilliam et al. (1995) state that the very large scatter means that n-capture element abundances in ultra-metal-poor stars are products of one or very few prior nucleosynthesis events that occurred in the very early, poorly mixed galactic halo, a theme that has been developed by many authors (e.g., Audouze & Silk 1995, Shigeyama & Tsujimoto 1998, Argast et al. 2000, Tsujimoto et al. 2000).
Supernova models produce different yields as a function of progenitor mass, progenitor metallicity, mass cut (what gets ejected compared to what falls back towards the compact central object), and detonation details. The α elements are mainly produced in the hydrostatic burning phase within the pre-supernova star. Thus α yields are not dependent on the mass cut or details of the fallback/explosion mechanism which leads to a much smaller dispersion at low metallicity.
There is no known age-metallicity relation that operates over a useful dynamic range in age and/or metallicity. (This effect is only seen in a small subset of hot metal-rich stars – see “Is There an Age-Metallicity Relation?” above). Such a relation would require the metals to be well mixed over large volumes of the ISM. For the forseeable future, it seems that only a small fraction of stars can be dated directly (see “Stellar Age Dating” above).
5.3. Reconstructing Ancient Star Groups
We now conjecture that the heavy element metallicity dispersion may provide a way forward for tagging groups of stars to common sites of formation. With sufficiently detailed spectral line information, it is feasible that the chemical tagging will allow temporal sequencing of a large fraction of stars in a manner analogous to building a family tree through DNA sequencing.
Most stars are born within rich clusters of many hundreds to many thousands of stars (Clarke et al. 2000, Carpenter 2000). McKee & Tan (2002) propose that high-mass stars form in the cores of strongly self-gravitating and turbulent gas clouds. The low mass stars form within the cloud outside the core, presumably at about the same time or shortly after the high-mass stars have formed. The precise sequence of events which give rise to a high-mass star is a topic of great interest and heated debate in contemporary astrophysics (e.g., Stahler et al. 2000).
A necessary condition for chemical tagging is that the progenitor cloud is uniformly mixed in key chemical elements before the first stars are formed. Another possibility is that a few high-mass stars form shortly after the cloud assembles, and enrich the cloud fairly uniformly. Both scenarios would help to ensure that long-lived stars have identical abundances in certain key elements before the onset of low-mass star formation.
For either statement to be true, an important requirement is that open clusters of any age have essentially zero dispersion in some key metals with respect to Fe. There has been very little work on heavy element abundances in open clusters. The target clusters must have reliable astrometry so as to minimize pollution from stars not associated with the cluster (Quillen 2002).
If our requirement is found not to be true, then either the progenitor clouds are not well mixed or high-mass stars are formed after most low-mass stars. A more fundamental consequence is that a direct unravelling of the disk into its constituent star groups would be impossible; in other words, the epoch of dissipation cannot be unravelled after all.
Consider the (extraordinary) possibility that we could put many coeval star groups back together over the entire age of the Galaxy. This would provide an accurate age for the star groups either through the color-magnitude diagram, or through association with those stars within each group that have [n-capture/Fe] ≫ 0 and can therefore be radioactively dated. This would provide key information on the chemical evolution history for each of the main components of the Galaxy.
But what about the formation site? The kinematic signatures will identify which component of the Galaxy the reconstructed star group belongs to, but not specifically where in the Galactic component (e.g., radius) the star group came into existence. For stars in the thin disk and bulge, the stellar kinematics will have been much affected by the bar and spiral waves; it will no longer be possible to estimate their birthplace from their kinematics. Our expectation is that the derived family tree will severely restrict the possible scenarios involved in the dissipation process. In this respect, a sufficiently detailed model may be able to locate each star group within the simulated time sequence.
In addition to open clusters, we have already argued that the thick disk is an extremely important fossil of the processes behind disk formation. The thick disk is thought to be a snap-frozen relic of the early disk, heated vertically by the infall of an intermediate mass satellite. Chemical tagging of stars that make up the thick disk would provide clues on the formation of the first star clusters in the early disk.
5.4. Chemical Abundance Space
An intriguing prospect is that reconstructed star clusters can be placed into an evolutionary sequence, i.e., a family tree, based on their chemical signatures. Let us suppose that a star cluster has accurate chemical abundances determined for a large number n of elements (including isotopes). This gives it a unique location in an n-dimensional space compared to m other star clusters within that space. We write the chemical abundance space as (Fe/H, X1/Fe, X2/Fe, ...) where X1, X2 ... are the independent chemical elements that define the space (i.e., elements whose abundances are not rigidly coupled to other elements).
The size of n is unlikely to exceed about 50 for the foreseeable future. Hill et al. (2002) present exquisite data for the metal-poor star CS 31082-001, where abundance estimates are obtained for a total of 44 elements, almost half the entire periodic table (see also Cayrel et al. 2001). In Figure 12, we show what is now possible for another metal-poor star, CS 22892-052 (Sneden et al. 2001a). The α elements and r-process elements, and maybe a few canonical s-process elements at low [Fe/H], provide information on the cloud abundances prior to star formation, although combinations of these are likely to be coupled (Heger & Woosley 2001, Sneden et al. 2001a). There are 24 r-process elements that have been clearly identified in stellar spectra (Wallerstein et al. 1997).
Figure 12. CS 22892-052 n-capture abundances (points) taken from Sneden et al. (2000) and scaled solar system abundances (solid and dashed lines) taken from Burris et al. (2000). Many of the heavy elements conform to the solar system r-process abundance pattern, although some elements show the hallmark of the s-process. This figure was originally presented in Sneden et al. (2001a).
The size of m is likely to be exceedingly large for the thin disk where most of the baryons reside. For a rough estimate, we take the age of the disk to be 10 Ga. If there is a unique SN II enrichment event every 100 years, we expect of order 108 formation sites. Typically, a SN II event sweeps up a constant mass of 5 × 104 M⊙ (Ryan et al. 1996, Shigeyama & Tsujimoto 1998). Simple chemical evolution models indicate that this must be of the right order to explain the metallicity dispersion at low [Fe/H] (Argast et al. 2000). Roughly speaking, there have been 103 generations of clouds since the disk formed, with about 105 clouds in each star-forming generation, such that cloud formation and dispersal cycle on a 107 year timescale (Elmegreen et al. 2000).
Whereas the total number of star clusters over the lifetime of the thin disk is very large, the size of m for the stellar halo (Harding et al. 2001), and maybe the thick disk (Kroupa 2002), is likely to be significantly smaller. Our primary interest is the oldest star clusters. Reconstructing star clusters within the thick disk is a particularly interesting prospect since the disk is likely to have formed within 1−1.5 Ga of the main epoch of baryon dissipation (Prochaska et al. 2000).
The task of establishing up to 108 unique chemical signatures may appear to be a hopeless proposition with current technology. But it is worth noting that more than 60 of the chemical elements (Z > 30) arise from n-capture processes. Let us suppose that half of these are detectable for a given star. We would only need to be able to measure two distinct abundances for each of these elements in order to achieve 109 independent cells in -space. If many of the element abundances are found to be rigidly coupled, of course the parameter space would be much smaller.
It may not be necessary to measure as many as 30 elements if some can be found which are highly decoupled and exhibit large relative dispersions from star to star. Burris et al. (2000) demonstrate one such element pair, i.e., [Ba/Fe] and [Sr/Fe]. It is difficult at this stage to suggest which elements are most suited to chemical tagging. In part, this depends on the precise details and mechanism of formation of the n-capture elements at low [Fe/H].
The element abundances [Xi/Fe] show three main peaks at Z ≈ 26, Z ≈ 52, and Z ≈ 78; the last two peaks are evident in Figure 12. There have been suggestions that the r-process gives rise to random abundance patterns (e.g., Goriely & Arnould 1996), although this is not supported by new observations of a few metal-poor stars. Heavy r-process elements around the second peak compared to the Sun appear to show a universal abundance pattern (Sneden et al. 2000, Cayrel et al. 2001, Hill et al. 2002). However, Hill et al. find that the third peak and actinide elements (Z ≥ 90) are decoupled from elements in the second peak. We suspect that there may be a substantial number of suitable elements (∼ 10) which could define a sufficiently large parameter space.
Our ability to detect structure in -space depends on how precisely we can measure abundance differences between stars. It may be possible to construct a large database of differential abundances from echelle spectra, with a precision of 0.05 dex or better; differential abundances are preferred here to reduce the effects of systematic error.
5.5. Chemical Trajectories
Our simple picture assumes that a cloud forms with a unique chemical signature, or that shortly after the cloud collapses, one or two massive SN IIs enrich the cloud with unique yields which add to the existing chemical signature. The low-mass population forms with this unique chemical signature. If the star-formation efficiency is high (≳ 30%), the star group stays bound although the remaining gas is blown away. If the star-formation efficiency is low (≲ 10%), the star cluster disperses along with the gas. In a closed box model, the dispersed gas reforms a cloud at a later stage.
In the closed box model, each successive generation of supernovae produces stellar populations with progressive enrichments. These will lie along a trajectory in -space which can be identified in principle using minimum spanning tree methods (Sedgewick 1992). The overall distribution of the trajectories will be affected by fundamental processes like the star formation efficiency, the star formation timescale, the mixing efficiency, the mixing timescale, and the satellite galaxy infall rate.
As we approach solar levels of metallicity in [Fe/H], the vast number of trajectories will converge. By [Fe/H] ≈ -2.5, AGB stars will have substantially raised the s-process element abundances; by [Fe/H] ≈ -1, Type Ia supernovae will have raised the Fe-group abundances. Star clusters that appear to originate at the same location in this -space may simply reflect a common formation site, i.e., the resolution limit we can expect to achieve in configuration space. The ability to identify common formation sites rests on accurate differential abundance analyses (Edvardsson et al. 1993, Prochaska et al. 2000).
Even with a well-established family tree based on chemical trajectories in the chemical -space, this information may not give a clear indication of the original location within the protocloud or Galactic component. This will come in the future from realistic baryon dissipation models. Forward evolution of any proposed model must be able to produce the chemical tree.
However, the -space will provide a vast amount of information on chemical evolution history. It should be possible to detect the evolution of the cluster mass function with cosmic time (Kroupa 2002), the epoch of a starburst phase and/or associated mass ejection of metals to the halo (Renzini 2001), and/or satellite infall (Noguchi 1998).
As we go back in time to the formation of the disk, we approach the chemical state laid down by population III stars. The lack of stars below [Fe/H] ≈ -5 suggests that the protocloud was initially enriched by the first generation of stars (Argast et al. 2000). However, the apparent absence of any remnants of population III remains a puzzle: Its stars may have had a top-heavy initial mass function, or have dispersed into the intra-group medium of the Local Group. If one could unravel the abundances of heavy elements at the time of disk formation, this would greatly improve the precision of nucleo-cosmochronology (see “Stellar Age Dating” above).
5.6. Candidates for Chemical Tagging
Chemical tagging will not be possible for all stars. In hot stars, our ability to measure abundances is reduced by the stellar rotation and lack of transitions for many ions in the optical. The ideal candidates are the evolved FGK stars that are numerous and intrinsically bright. These can be observed at echelle resolutions (R > 30,000) over the full Gaiasphere. Moreover, giants have deep, low-density atmospheres that produce strong low-ionization absorption lines compared to higher gravity atmospheres. Even in the presence of significant line blending, with sufficient signal, it should be possible to derive abundance information by comparing the fine structure information with accurate stellar synthesis models. Detailed abundances of large numbers of F and G subgiants would be particularly useful, if it becomes possible to make such studies, because direct relative ages can be derived for these stars from their observed luminosities.
It is not clear at what [Fe/H] the r-process elements become swamped by the ubiquitous Fe-group and s-process elements. At a resolution of R ∼ 105, many r-process elements can be seen in the solar spectrum, although the signal-to-noise ratio of about 1000 is needed, and even then the spectral lines are often badly blended (Kurucz 1991, 1995). Travaglio et al. (1999) suggest that the s-process does not become significant until [Fe/H] ≈ -1 because of the need for pre-existing seed nuclei (Spite & Spite 1978, Truran 1981), although Pagel & Tautvaisiene (1997) argue for some s-process production at [Fe/H] ∼ -2.5. Prochaska et al. (2000) detected Ba, Y and Eu in a snapshot survey of thick disk G dwarfs in the solar neighborhood with -1.1 ≲ [Fe/H] ≲ -0.5. This survey only managed to detect a few transitions in each element, although their spectral coverage was redward of 440 nm with SNR ≈ 100 per pixel at R ≃ 50,000. Longer exposures with R ∼ 105 and spectral coverage down to 300 nm would have detected more heavy elements.
In our view, observations of nucleosynthetic signatures of metal-poor stars provide a cornerstone of near-field cosmology. Success in this arena requires major progress across a wide front, including better atomic parameters (Truran et al. 2001), improved supernova models, better stellar synthesis codes and more realistic galaxy formation models. There are no stellar evolutionary models that lead to a self-consistent detonation and deflagration in a core-collapse supernova event or, for that matter, detonation in a thermonuclear explosive event. Realistic chemical production at the onset of the supernova stage requires a proper accounting of a large number of isotope networks (400−2500) that cannot be adequately simulated yet. Modern computers have only recently conquered relatively simple α networks involving 13 isotopes. The inexorable march of computer power will greatly assist here.
There is also a key experimental front both in terms of laboratory simulations of nucleosynthesis, and the need for major developments in astronomical instrumentation (see “Epilogue: Challenges for the Future”). Many authors (e.g., Sneden et al. 2001b) have stressed the importance of greatly improving the accuracy of transition probabilities and reaction rates for both heavy and light ion interactions. This will be possible with the new generation of high-intensity accelerators and radioactive-beam instruments (Käppeler et al. 1998, see Manuel 2000).
Progress on all fronts will require iteration between the different strands. Already, relative r-process and α element abundances for metal-poor stars have begun to constrain the yields for different stellar masses and associated mass cuts of progenitor supernovae (Mathews et al. 1992, Travaglio et al. 1998, Ishimaru & Wanajo 2000).
It is an intriguing thought that one day we may be able to identify hundreds or thousands of stars throughout the Gaiasphere that were born within the same cloud as the Sun.