|Annu. Rev. Astron. Astrophys. 1999. 37:
Copyright © 1999 by . All rights reserved
7.2. Inferences About Chemical Abundances and Ages
The evolutionary model sets agree that the observed far-UV fluxes can be produced by old (5-15 Gyr), metal-rich (Z ~ 1-3 Z) populations given favorable, but not unreasonable, assumptions about Y / Z, R, and MENV. They also agree that changes in Z alone cannot reproduce the Faber correlation, unless it is assumed that Y / Z 2.5 or that R increases with Z (as foreseen by GR). Models based on more conservative assumptions about mass loss require that populations with significant UVX components be older than about 10 Gyr.
As noted in Section 6.3 and illustrated in Figure 10, the metal abundance Z, if decoupled from Y and mass loss, has only a secondary influence on the spectrum of the hot components at wavelengths above 1200 Å. There is a somewhat greater effect on the temperature distribution of light in the 900-1200 Å range (Brown et al 1997). Y can have a significant effect on UV flux production, but its influence can be masked by changes in mass loss. Furthermore, the UVX absorption line spectrum is apparently subject to atmospheric diffusion effects, as discovered by Brown et al (1997). Therefore, far-UV spectra in the 1200-2000 Å range cannot easily be used to infer abundances of the hot star populations of E galaxies.
It is also premature to try to use the UVX to age-date elliptical galaxies. The far-UV appears promising for this purpose because it is the most rapidly evolving part of a single-burst galaxy spectrum. A number of studies have noted that the "turn-on" of the UVX (which occurs at the age when MTO drops to the point that the assumed RGB mass loss is able to fill the smaller envelope channels) marks an obvious spectral transition which might be used to age-date E galaxies at moderate lookback times (e.g. GR, Magris & Bruzual 1993, Chiosi 1996, Bressan et al 1996, Chiosi et al 1997, Yi et al 1999). The amplitude of the transition in UV colors is large. Unfortunately, its timing in the models is very sensitive to assumptions about mass loss and helium abundance, making results strongly model-dependent. This can be seen in the cases presented by Yi et al (1997b, 1999).
An interesting related example is the interpretation of the UVX as the product of a metal-poor subpopulation of extremely old (18-20 Gyr) stars, presumably the earliest generation to form in massive galaxies (Lee 1994, Park & Lee 1997). The models employed by Lee and Park adopt atypically small values for both total RGB mass loss (a fixed 0.22 M) and MENV (0.02 M). Larger values for these parameters would significantly reduce the inferred ages. The models also do not fit the UVX spectra very well, having the flatter energy distributions for 1500-3500 Å characteristic of metal poor systems (Park & Lee 1997), even when mixed metallicities and larger effective mass loss are included (Yi et al 1999).
It is clear in general that assumptions about R and MENV largely determine the outcome of the far-UV evolutionary synthesis models developed to date and any conclusions about t, Y, or Z that may emerge from them. This is necessarily so, given the circumstance that changes in only a few 0.01 M in MENV can radically affect the UV output of stars. Age and mass loss can be traded for one another, and without a more deterministic theory of mass loss, derived ages are not reliable.
These remarks apply to the far-UV, hot-star spectra of old populations. The situation is quite different in the mid-UV region (2400-3200 Å) where the light becomes dominated by cool stars (Te ~ 5500-7500 K) near the main sequence turnoff. Turnoff light is directly sensitive to both age and abundance, whereas red giant light, which is important longward of 4500 Å, is insensitive to age. The principal obstacle to exploitation of the mid-UV as a population diagnostic is the lack of good "libraries" of mid-UV stellar energy distributions. Empirical datasets (e.g. Fanelli et al 1992) tend to be limited to solar abundance, whereas theoretical ones (e.g. Kurucz 1991) have serious shortcomings due to difficulties in treating the overwhelming UV line blanketing. HST is beginning to fill the gap, if slowly, with high quality, medium-resolution spectra (e.g. Heap et al 1998). Another technical difficulty with the mid-UV is that contamination by the long-wavelength tail of the UVX energy distribution must be removed. Although this contributes over 50% of the 2700 Å light in many cases (BBBFL, Ponder et al 1998), correction for the UVX component is straightforward because it has a smooth and well-determined shape (Dorman et al 1999).
Preliminary synthesis models of the mid-UV using theoretical stellar spectra confirm the expectation that it will be an excellent age/abundance diagnostic. DOR's experiments with fitting broad-band UV colors of E galaxies showed that 2500-V yields much more information on age and composition than does 1500-V. They estimate that (2500-V) / log Z ~ 2.7 for old populations, a much higher sensitivity than for most optical-IR indices (e.g. Worthey 1994). The mid-UV continuum is especially useful in placing limits on the contribution of metal-poor populations to galaxy light. The available mid-UV models show, for instance, that the metal-poor fraction in E galaxies is much smaller than predicted by simple "closed box" nucleosynthetic models (e.g. Tantalo et al 1996, Worthey et al 1996). Empirically, mid-UV spectral features also strongly distinguish the populations of globular clusters and E galaxy cores from one another (e.g. Ponder et al 1998).
One of the most important applications of mid-UV stellar population analysis will be to the spectra of high redshift galaxies. Age-dating of distant objects that are passively evolving, such as LBDS 53W091, with z = 1.55, can constrain the earliest epoch of star formation, and hence cosmology (e.g. Spinrad et al 1997).