Globular clusters in the Milky Way and in other galaxies have generated a rich lore of data and theoretical interpretations for galaxy formation. Ultimately, however, we must return to the question that occupied even the earliest historical thinking about old stellar populations: how representative are the globular clusters of the stellar content of galactic halos? Are they a clearly distinguishable stellar population on their own?
For the Milky Way, we have the ability to bring in the full array of photometric, kinematic, and chemical composition properties of both kinds of stars, and the potential links of the clusters with such components as the metal-poor halo, the thick disk, the old disk, or accreted satellites, are numerous and have generated a continually growing literature. In progressively more distant galaxies, the information is inevitably more restricted until, at distances far beyond the Local Group, we can refer to the stellar halo component only by its integrated light, smoothing over all detail. For giant E galaxies, this problem has been especially unfortunate, since this is the one type of galaxy not represented in the Local Group, and the nearest rich collections of ``typical'' ellipticals are in the Virgo and Fornax clusters, some ~ 15 - 18 Mpc away.
There is, however, one giant E galaxy which is much closer: NGC 5128, the dominant member of the sparse Centaurus group at d 4 Mpc. Because of its well known inner gas and dust lane and active central regions where star formation is taking place, it has long been regarded as ``peculiar'' and was thus unduly neglected as a place where we might learn about the properties of normal ellipticals. However, much evidence accumulated in the past decade or more has revealed that major galaxies of all types undergo episodes of satellite accretion or stripping, gas infall, and even mergers with other large galaxies, and that what we see happening now in NGC 5128 is not an especially unusual event. Different theoretical channels exist for forming large elliptical galaxies (hierarchical merging of gas clouds in the early universe, or later merging of disk galaxies, to name two), and it is perhaps true that no single one will be found ``typical'' of all of them. But we now have no compelling reason to discard NGC 5128 as a basis for testing out many ideas about the way E galaxies are built, and eventually, we may find that it is just as representative of gE's as the Milky Way is of (say) large spirals.
NGC 5128 contains a healthy population of some ~ 1600 globular clusters, with a specific frequency SN = 2.6 at a normal level for ellipticals in small groups (Harris 1991). Their spatial distribution, metallicities, and radial velocities have been studied in a series of papers by G. Harris and colleagues (1992 and references cited there). The GCS has been found to display the distinctly bimodal MDF that appears in many other gE's (Forbes et al. 1997), with roughly equal numbers of clusters in each component. In this respect, NGC 5128 so far reveals nothing fundamentally new. But of paramount importance for our purposes is that NGC 5128 is close enough for us to obtain (with HST imaging) color-magnitude photometry for its old-halo red giant stars, and thus directly compare the metallicity distribution functions of the clusters and the halo stars. At present, it is the only gE galaxy which is within reach in this way. With later and more advanced imaging tools (NGST and beyond), the stellar populations in the ellipticals of Virgo and Fornax will also come within our grasp; but NGC 5128 is an important prelude to what we can expect to glean from these systems.
Direct color-magnitude photometry of the halo stars in NGC 5128 was first obtained with HST by Soria et al. (1996) and then to a deeper level by G. Harris et al. (1999 [HHP99]). This latter study produced the color-magnitude diagram shown in Figure 9, showing that the halo of NGC 5128 is dominated by an old stellar population with an extremely broad range of color across the red giant branch. Only a tiny fraction of this range can be due to age differences (see the comparison of isochrones made in HHP99); most of it must be generated by a large spread of metallicity.
Figure 9. Color-magnitude diagram in (I, V-I) for a halo field in NGC 5128, at a projected distance of 21 kpc south of the center of the galaxy (adapted from Harris et al. 1999). The limiting magnitude I 27 reaches to within about one magnitude above the expected level of the horizontal branch. The population of stars is dominated entirely by an old red-giant branch, with negligible contribution from any young or intermediate-age stars. Fiducial lines for different metallicities ([Fe/H] = -2.2, -1.3, -0.75, -0.25, +0.1 from left to right) are superimposed on the data to indicate the very large abundance range in the NGC 5128 halo. |
Interpolating within the fiducial lines, whose locations are a very nonlinear function of [Fe/H], we obtain the MDF shown in Figure 10. About two-thirds of the stars belong to a ``metal-rich'' component (MRC) starting at Z 0.25 Z and trailing off to above-solar metallicity. The remaining one-third falls into the ``metal-poor'' component (MPC) starting at very low abundance ([Fe/H] < -2) and merging at its upper end with the onset of the MRC.
Figure 10. Metallicity distribution function for the red giant stars in the halo of NGC 5128, deduced from the color-magnitude data in the previous figure over the magnitude range 25 < I < 26 (adapted from Harris et al. 1999). The MDF is shown as the number of stars per unit heavy-element abundance Z / Z. The dashed lines show the simple one-zone chemical enrichment model which has a characteristic exponential decay shape. The halo was enriched in two ``bursts'': the metal-poor component with a yield y 0.002 (left-hand curve), and the metal-richer component with y 0.006. |
HHP99 show that these two components can be matched within the context of a classic one-zone chemical enrichment model (e.g., Pagel & Patchett 1975) in which an amount of gas G with initial abundance Z0 is gradually converted to a mass of stars S, with each generation of stars ejecting back some enriched material to the interstellar medium. If the star formation is left to run to completion, then G -> 0 asymptotically, while the heavy-element abundance Z continually increases. Assuming complete mixing and instantaneous recycling, we find by solving the gains-and-losses equation for Z as a function of the remaining gas mass G that the relative number of stars at metallicity Z is described by a curve with a characteristic exponential-decay shape,
Here y is the ``yield rate'', the ratio of ejected mass in
heavy elements to the mass locked up in dead stellar remnants
for any given generation of stars.
This simplest of enrichment models, of course, has no direct
timescale information in it, and does not take into account any
of the interesting and important
details in the enrichment process such as
the different contributions of Type I and II supernovae.
Nevertheless, its characteristic shape stands out remarkably well
(Figure 10): the MPC started at primordial composition
Z0 0 and
was enriched at a rate y ~ 0.002, while the MRC began
at Z0 0.25
Z and was
enriched with a clearly
higher yield y 0.006.
The notably different yield
ratios required by the data might, perhaps, represent different
mixtures of Type I and II supernova or different IMFs in
the two components. However, a more likely possibility may
be connected with the fact that the MPC makes up only a
minority of the stars (at most one-third of the halo, and perhaps
much less if the MRC comprises a higher proportion of
the inner halo; that is, if the halo has a metallicity gradient).
Within the context of an in situ formation picture, this
would mean that the first, earliest burst started from near-primordial
composition and used up only a fraction of the gas, leaving it
pre-enriched to Z ~ 0.25
Z and waiting
for the second, more major burst to start. The effective yield in the
first burst would have been lowered if there had been substantial
gas loss from the MPC star-forming
clouds, as was noted long ago by
Hartwick (1976) as a
basic explanation for the low metallicity of the Milky Way halo stars.
An situ synthesis of the formation of the NGC 5128 halo
might then go something like this
(HHP99;
Forbes et al. 1997):
Round One: The protogalaxy consists of
many pre-galactic fragments (SGMCs) of primordial composition,
distributed throughout a still-lumpy larger potential well
(Pudritz 1998Z).
Star formation begins within these small clouds and consumes
a fraction of the gas supply; but the ensuing stellar winds and
supernovae interrupt the process, heating
most of the (unused, but now partially enriched)
gas and driving it out of these small local potential wells
(Dekel & Silk 1986).
The residue of this
epoch is seen in the MPC stars, and in the metal-poor globular clusters.
Round Two: The gas filling the halo has now
cooled and re-collapsed inward,
and star formation begins again. This time, almost
all the gas is used up, since it lies deeper within the fully
formed potential well of the galaxy and does not escape.
The heavy-element abundance drives on upward to Solar metallicity.
This second stage, producing the MRC stars and clusters,
is in fact what forms the bulk of the galaxy that we see today.
The time interval between these two major phases is a matter of
conjecture, but we can speculate that round Two
must have happened at least 1 - 2 Gy after Round One
(that is, one or two free-fall times, enough for the ejected gas
to cool, dissipate, and collapse back inward).
Round Three: The galaxy continues to build by
later and more minor accretion of material, representing
the ongoing phase of accumulation that we are witnessing now.
Occasional smaller satellites are captured, bringing their
gas into the central ~ 5 kpc region that we see as the
current site of star formation. The effects on the stellar
content of the outer halo, however, appear to be minor.
This scenario can be checked against the MDF for
the globular clusters (Figure 11,
adapted from HHP99).
The metal-rich clusters
have a mean metallicity and MDF dispersion that are indistinguishable
from the metal-richer stellar component - a strong consistency
test that the two types of objects formed together in the major
star-forming phase that produced most of the galaxy. Two interesting
differences between the clusters and the field stars are, however,
that (a) the clusters in the metal-poor ``mode''
make up about 60% of the halo GCS,
whereas the MPC stars make up at most 1/3 of the stars; and
(b) the MDF for the metal-poor clusters is sharply
peaked at [Fe/H] = -1.2, and does not have the smooth ramp-up
displayed by the MPC stars.
An intriguing point connected with the comparison in
Figure 11
is that the specific frequencies of the MPC and MRC
must be quite different: the MPC contains one-third or less of
the stars, but about two-thirds of the clusters. If we take these
proportions literally, then we find (see
HHP99)
SN(MPC)
4.3, rather like the Fornax and Virgo ellipticals, while
SN(MRC) 1.5
- a low value more like disk galaxies.
This discrepancy would be consistent with the idea that globular
clusters, as the densest units within the star-forming gas clouds,
formed earliest; but that in ``Round One'', subsequent star formation
was interrupted before it could run to completion, passing much of the
gas on to the second round.
Figure 11. Metallicity distribution
function (number of objects per [Fe/H] bin) for the halo stars
in NGC 5128 (upper panel) and for the halo globular clusters
(lower panel), with data from
HHP99 and
G. Harris et
al. (1992).
We must also ask whether or not other distinctly different models
can produce the combination of features we observe. One alternate
picture is the accretion model of
Côté et
al. (1998).
The basis of their picture is that the MRC is regarded
as the original E galaxy (formed in a single burst), while the
MPC is accumulated later by accreted smaller galaxies, which have
lower metallicity. The metal-poor halo and globular clusters
are thus envisaged to
build up over time. Large numbers of small galaxies are needed in
this scheme: in the case of NGC 5128, the MPC stars make up as much
as 1/3 of the halo, which would require the accretion of
200
dwarf galaxies. For comparison, the present-day Centaurus group
contains about 20 dwarfs, mostly irregulars (all of which have low
SN; it is the dE galaxies which have the higher
SN that we need
for the MPC halo). These factors present difficulties for the
halo accretion picture which may not be easy to overcome (see
additional discussion in the next section).
An alternate and well known approach is the merger scenario
outlined by
Schweizer (1987) and
Ashman & Zepf (1992),
whereby a gE galaxy
forms by the amalgamation of large disk galaxies (such as, indeed,
we see happening today in cases like the Antennae).
Here, the MRC is postulated to form during the merger, as long as the
incoming progenitor galaxies are gas-rich. In the case of NGC 5128,
we see at least 2/3 of the halo is metal-rich; thus, the merging
``galaxies'' would have had to be almost entirely gaseous in the
first place. The distinction between this approach and a
Searle-Zinn-like in situ model, in which a large galaxy builds
up by hierarchical clustering of gas clouds,
is therefore considerably blurred. We will return to these
model comparisons in our later case studies.