|Annu. Rev. Astron. Astrophys. 1999. 37:
Copyright © 1999 by . All rights reserved
Prior to the first UV observations, there was a widespread expectation that normal elliptical galaxies would be uninteresting in the FUV (as, later, would also be the case with the X-ray and far-infrared regions). The hottest identified stellar component of any consequence was the main sequence turnoff, with a temperature (Te ~ 6000 K) too cool to produce many FUV photons. Although it was recognized that the old, metal-poor populations of globular clusters sometimes contained horizontal-branch (HB) stars with Te 10000 K, these were thought to be absent in the clusters (e.g. 47 Tucanæ) with metal abundances nearest those of massive galaxies. The only hint of hot populations in E galaxies was the presence of [O II] emission lines, though these could plausibly be explained without stellar photoionization (Minkowski & Osterbrock 1959).
FUV radiation from galaxies was first detected by the University of Wisconsin UV photometer carried on the second Orbiting Astronomical Observatory (OAO-2). The experiment obtained fluxes with an entrance aperture of 10', diameter in 7 intermediate band filters extending from 4250 Å to the FUV at 1550 Å. The first announcement (Code 1969) of results for an old population was for the central bulge (r < 900 pc) of the Local Group Sb spiral M31. As expected, the energy distribution of M31 fell steeply between 3500 and 2500 Å but then, remarkably, began to rise again at shorter wavelengths. A more recent UV-optical spectrum of M31 is shown in Figure 1. Since the energy distributions of normal stars cooler than Te ~ 8500 K (spectral type A5) decline precipitously below 1800 Å owing to absorption by metallic ionization edges (e.g. Fanelli et al 1992), the detection of any far-UV flux in galaxies implies sources with higher equivalent temperatures. After a difficult calibration process, OAO-2 photometry was ultimately published for 7 E/S0 objects and the M31 bulge (Code et al 1972, Code & Welch 1982). The OAO-2 detections of two objects were confirmed, and new detections made of another 11 E galaxies, by the Astronomical Netherlands Satellite (launched in 1974) using intermediate band photometry with a 2.5' × 2.5' aperture over the range 1550-3300 Å (de Boer 1982).
Figure 1. A composite UV-optical energy distribution for the center of the Sb galaxy M31. IUE data taken with a 10" × 20" aperture is plotted below 3200 Å, while a ground-based spectrum covering the same region is plotted above. Resolution is 20 Å below 2600 Å and 12 Å above. Irregularities in the UV spectrum below 2200 Å are mainly noise. Some of the stronger absorption line features are identified ("BL" corresponds to a strong blend of Fe and other metallic lines near 2538 Å). The "UV-upturn" is the rise in the spectrum at wavelengths shorter than 2000 Å. By simple extrapolation of the far-UV continuum slope, one finds that the upturn component contributes only about 0.3% of the V light of the galaxy. Spectrum courtesy of D Calzetti.
An immediate conclusion from the UVX observations which was emphasized by Code and his colleagues was that early-type galaxies exhibited much larger scatter in the UV than was expected from their conspicuously homogeneous behavior in the optical to near-IR (4000-20000 Å) region. The UV observations implied divergent histories at some level and were among the first indications that elliptical galaxy populations were more heterogeneous than envisioned in Baade's classic definition of Population II (Baade 1944, O'Connell 1958, O'Connell 1980, Faber et al 1995). They called into question the use of E galaxies as "standard candles" in cosmological studies. They also complicated the construction of accurate K-corrections needed to transform photometry of high redshift elliptical galaxies to standard bands in the restframe (e.g. Pence 1976, Coleman et al 1980, King & Ellis 1985, Bertola et al 1982, Kinney et al 1996).
Interpretation of the unexpected OAO-2 results was initially confused by calibration uncertainties which produced anomalously steep FUV energy distributions (in normal spirals and irregulars as well as early-type systems, see Code & Welch 1982). Code (1969), Code et al (1972) suggested that the UVX component was nonthermal radiation from an active nucleus (AGN) or scattering of photons from massive hot stars by interstellar dust. The latter would have implied that most E/S0 galaxies contain an appreciable Population I component.
Hills (1971) pointed out that the steep rise of the M31 UV spectrum to higher photon energies was incompatible with known nonthermal sources but was closely matched by the Rayleigh-Jeans tail of a high temperature thermal source. Based on comparison with a small sample of UV-bright stars in the globular cluster M3, he proposed that the UV upturn is produced by highly evolved, hot, low-mass stars such as the central stars of planetary nebulae, now known as post-asymptotic giant branch (PAGB) stars, or their hot white dwarf descendents. He did not require that these be members of a strong Population II (old, metal-poor) component but pointed out that their prominence would probably depend on metal abundance.
Tinsley (1972a) argued that the UV light arose instead from young, massive, main sequence stars and showed that a spectral synthesis model for an old galaxy with an exponentially declining star formation rate and an e-folding time of 2 Gyr could fit the OAO-2 flux for M31 observed at 1700 Å. This would imply that the UVX was related to a normal, if temporally extended, star formation process in early-type systems. Fuel for the star formation might be primordial gas consumed gradually over a Hubble time, mass loss from red giants, or material accreted from outside galaxies (Gallagher 1972, Tinsley 1972b, O'Connell 1980, Gunn et al 1981). Residual star formation histories of the type suggested by Tinsley would drastically change the predicted properties of E galaxies viewed at moderate look-back times, whereas the low-mass star interpretation would have less serious implications for spectral evolution.
It was implicit in these early studies that the hot components that dominated the far-UV light could be virtually undetectable at visible wavelengths - i.e. that the UV was providing entirely independent information about galaxies. Ignoring any contribution from the cool components to the UV light, the maximal fractional contribution of a hot component to the integrated V-band light of a galaxy will be pmax ~ 100.4 , where = (1500-V)hot - (1500-V)obs. A color for a typical E galaxy is (1500-V)obs ~ +3 while a component with an appropriate far-UV spectral slope (B0 equivalent) has (1500-V)hot ~ -4.5, implying that pmax ~ 0.001. This is about 50 times smaller than could be directly detected in the V-band using spectral synthesis techniques.
The early workers on the UVX realized that the best tests of the alternative interpretations were (a) UV spatial structure and (b) UV spectral features observed at higher resolution. An active nucleus would be a concentrated point source, whereas a population of low-mass stars would presumably have a smooth distribution similar to that found in the optical bands for bulges and E galaxies. Young, massive-star populations would likely have a clumpy structure, similar to the OB associations found in spiral arms, and they might well be concentrated to disks. In nearer galaxies individual massive OB stars could be isolated. Spectroscopically, a UV-bright AGN would be easy to identify on the basis of broad, high-excitation emission lines. Active massive star-forming regions characteristically exhibit strong UV resonance lines of Si IV, C IV, and other species, often with P-Cygni profiles (e.g. Kinney et al 1993), whereas the spectra of hot, low-mass stars are relatively weak-lined in the 1200-2000 Å region.
Because of limited UV observing opportunities, it would not be possible to apply these tests in a definitive way until over a decade after the discovery of the UVX. Only short-duration sounding rocket or balloon experiments were available until 1978. The most productive observing facility for the study of the UVX in the period 1978-1990 was IUE (Kondo 1987). IUE's handicaps of small effective collecting area, small entrance aperture, and limited dynamic range were beautifully compensated by its record 18 year lifetime and a capability for very long integration times, and it produced an invaluable set of UVX spectra. The fact that its point spread function was smaller than its 10" × 20" entrance aperture also meant that spatial structure could be studied to a radius of 10". After 1990, HST and the two Astro missions provided new capabilities to study the UVX.
In the next two sections, we describe the basic phenomenology of far-UV sources in bright early-type galaxies, as determined by IUE and other instruments, and how this bears on the now accepted interpretation of these as low-mass stars in old stellar populations.