ULTRAVIOLET-RADIATION FROM EVOLVED STELLAR POPULATIONS .2. THE ULTRAVIOLET UPTURN PHENOMENON IN ELLIPTIC GALAXIES

We discuss the far-ultraviolet upturn phenomenon (UVX) observed in elliptical galaxies and spiral galaxy bulges. Our premise is the the UV radiation from these systems emanates primarily from extreme horizontal branch (EHB) stars and their progeny. Such objects have zero-age horizontal branch envelope masses M(env)(0) less than or similar to 0.05 M(circle dot) . Local examples of EHB stars exist in some globular clusters and in the Galactic disk field and serve both as a guide and constraint. We derive the broad-band UV colors 1500-V and 2500-V for globular clusters and elliptical galaxies from the available satellite data and investigate color-color and color-line strength correlations. There are several important distinctions between clusters and galaxies. They do not occupy a single Mg-2-color sequence. Clusters can be bluer than any galaxy in 15-V and 25-V,implying larger hot star populations, but galaxies are significantly bluer than clusters in 15-25 at a given 15-V. We attribute this primarily to the effect of metal abundance on the mid-UV (2500 Angstrom) light. These redder colors of the galaxies also imply that the UVX in galaxies is not produced by metal-poor subpopulations similar to the clusters. We develop a simple spectral synthesis formulation for all phases of single star evolution from the zero-age main sequence (ZAMS) to the white dwarf cooling track that requires only one or two parameters for each choice of age and abundance. We provide the ingredients necessary for constructing models with arbitrary horizontal branch (HE) morphologies in the age range 2 < t < 20 Gyr and for six metallicities in the range -2.26 < [Fe/H] < 0.58; we also consider the effect of enhanced Y in metal-rich models. UV properties of the models are predicted using the Kur;ucz (1991) atmospheres. The maximum lifetime UV output is produced by EHB stars with M0(env) similar to 0.02 M circle dot and can be up to 30 times higher than for post-asymptotic giant branch (P-AGB) stars. The ultraviolet output of old populations is governed primarily by the distribution of M(env)(0), P(M(env)(0)), on the ZAHB. The UV output is not very sensitive to [Fe/H] or to Y, but it can change very rapidly with M(env)(0). Thus it is extremely sensitive to the precise nature of giant-branch mass loss. Because this process is not well understood physically, we choose to leave mass loss as an implicit free parameter. Our models use simple descriptions of P(M(env)(0),) to bracket the colors produced from any real distribution of stars. Our models accurately predict the range of UV colors observed for the globular clusters, given known constraints on their age, abundances, and HE morphologies. Clusters with ''blue HB'' morphologies do not require the hotter EHB stars to explain their UV colors, although a small EHB population is consistent with our models. The largest known population of these stars in a cluster, as a fraction of the total HE, is similar to 20% in omega Cen. For [Fe/H] greater than or similar to - 0.5, however, blue HE stars will be rare. As a consequence, we find that models with [Fe/H] greater than or similar to 0 that do not contain EHB stars cannot reproduce the colors of most of the galaxies. However, only small EHB fractions are required: less than or similar to 5% for the bulk of the E galaxies and similar to 20% for those with the strongest UVX. These results are independent of the assumed [Fe/H]. The EHB fraction required for most galaxies is comparable to the fraction of hot subdwarfs in the Galactic disk. Most of these are EHB stars, and their existence considerably strengthens the case for EHB populations as the source of elliptical galaxy UV light. The models also predict that the fraction of the far-UV light from P-AGB stars, which are spatially resolvable in nearby galaxies, is similar to 70% and similar to 20% for moderate UVX and strong UVX systems, respectively. We find that 25 - V, but not 15 - V, is sensitive to the age and abundance, though these cannot always be cleanly distinguished. The galaxy colors place limits of [[Fe/H]] > -0.5 and less than 15% on the contribution of globular cluster-type populations to the V light. Galaxy colors are consistent with solar-abundance models with ages in the range 6-14 Gyr. However, the 25-V colors of the galaxies other than the strong UVX systems are too blue to be consistent with [Fe/H] > 0.2 for any age. This may be additional evidence that [Mg/Fe] > 0 in elliptical galaxies. UV colors for M32 are consistent with the solar abundance, intermediate-age (4-6 Gyr) population inferred from optical/IR observations. We discuss several implications of the observations and the models, including the question of light metal versus iron peak enhancements in galaxies, whether the UV color-Mg-2 correlation is continuous or discrete, effects of helium abundance on the UVX, and the key question of whether red giant branch mass loss can be large enough to produce the necessary EHB population in the strong UVX galaxies.