Annu. Rev. Astron. Astrophys. 1999. 37: 603-648 Copyright © 1999 by Annual Reviews. All rights reserved

5. SPECTRAL AND PHOTOMETRIC CHARACTERISTICS OF THE UVX

5.1. Incidence, Spectral Shape, and Line Features

Except in cases of obscuration by a major dust lane (e.g. in edge-on S0-Sb objects), far-UV radiation has been detected in all nearby early-type systems observed with adequate S/N. As noted above (Section 3), this implies the presence of sources with T_e > 8500 K. Rifatto and colleagues (1995a, 1995b) have compiled all UV observations of galaxies published before 1990 and attempted to place them on a homogeneous system, which is a challenge owing to the varied types of experiments involved and the relatively low photometric precision which is typical. Their list includes 94 galaxies of type Sb or earlier with UV detections at < 2100 Å. The list does not include later photometry or imaging from the SCAP/FOCA balloon experiments (Milliard et al 1992, Donas et al 1995, Treyer et al 1998), Atlas/FAUST (Deharveng et al 1994), Astro/UIT (Stecher et al 1997), Astro/HUT (Kruk et al 1995), or the Maoz et al (1996) HST/FOC nuclear survey. Combined, these roughly double the total number of far-UV E-Sb detections, and HST is continually enlarging this sample. It is worth emphasizing that extragalactic UV observations are largely confined to relatively nearby, bright systems (except for very distant objects where the redshift brings the restframe UV into the bands accessible from the ground).

The early IUE spectra of the nuclei of bright ellipticals and spiral bulges (Johnson 1979, Bertola et al 1980, Perola & Tarenghi 1980, Nørgaard-Nielsen & Kjærgaard 1981, Oke et al 1981, Bertola et al 1982, Deharveng et al 1982, O'Connell et al 1986) showed immediately that the strong, broad emission lines characteristic of active nuclei were absent, excluding the AGN hypothesis. Signals for 2400 Å were, however, very weak and subject to several kinds of detector noise (which generated some spurious claims of narrow coronal or chromospheric emission lines). Except in the brightest sources, it was necessary to average far-UV fluxes over bandwidths of ~ 50 Å.

Burstein et al (1988, hereafter BBBFL) produced the largest homogeneous set of good IUE spectra for early-type galaxies (32 objects). Two examples are shown in Figure 4. The flux rapidly declines shortward of 3300 Å. Strong absorption features from Mg I (2852 Å), Mg II (doublet at 2800 Å), Fe I (numerous lines), and other metallic species are easily detectable down to ~ 2500 Å, as are strong discontinuities caused by metallic blanketing at 2640 and 2900 Å. The mid-UV line spectrum closely resembles that of F-G dwarf stars (see Fanelli et al 1992), the spectral types expected for the main sequence turnoff in an old E-galaxy population. The flux reaches a minimum in the range 2000-2600 Å, where the S/N is almost always rather poor, then usually rises steeply again to shorter wavelengths. No maximum is detected in the rising component longward of the IUE cutoff at ~ 1150 Å.

Figure 4. IUE spectra of two galaxies lying at the extremes of UVX behavior. M32 has the smallest known UV upturn, while NGC 4649 has one of the strongest. Data plotted with open triangles have 20 Å binning, while the solid line has 8 Å binning. The zero point of the NGC 4649 spectrum has been shifted -3.0 mags. Neither spectrum is corrected for redshift or foreground extinction. The FUV data for M32 are too poor to judge the slope of its UVX component. The weakness of the absorption lines near 2800 Å in NGC 4549 is caused by filling by the smooth UVX component, which contributes over 70% of the light at 2700 Å in this object. Reprocessed and recalibrated IUE spectra courtesy of RC Bohlin.

At the resolution permitted by the noise, the typical far-UV IUE spectrum is a relatively smooth continuum with an equivalent temperature T_e 2000K. The spectral slope of the upturn is roughly constant, so that the far-UV rise begins at longer wavelengths in galaxies with brighter UVX components (e.g. NGC 4649 in Figure 4). The contribution of the UVX component to the mid-UV light can be appreciable, ranging up to 75% at 2700 Å for objects like NGC 4649 (BBBFL, Ponder et al 1998, Dorman et al 1999), though this drops rapidly at longer wavelengths because of the steep rise in the spectrum of the cooler main sequence turnoff stars. In objects with the smallest UVX components (e.g. M32 and NGC 4382), the spectra appear to flatten below 2000 Å, rather than rise, but the S/N is too poor to estimate a temperature (BBBFL and Figure 4).

In the great majority of cases, E galaxy spectra longward of 3200 Å are quite similar to one another; the large spectral anomalies associated with the UVX are confined to the vacuum UV. This suggests that the stars responsible for the UVX are well segregated in the color-magnitude diagram from the bulk of the population. An interesting exception is M87, where anomalies are detectable up to 4000 Å (Bertola et al 1982, McNamara & O'Connell 1989); they are spatially extended and may be related to massive star formation in M87's cooling flow (see Section 8).

As originally emphasized by Tinsley (1972a), the characteristic FUV slope of UVX galaxies is consistent with star-forming models in which massive O and early B stars dominate the light. These require star formation to have occurred within the last 10-20 Myr. Tinsley did not discuss the far-UV spectral shape of her models, but later studies (e.g. Wu et al 1980, Gunn et al 1981, Nesci & Perola 1985, Rocca-Volmerange & Guiderdoni 1987, Bica & Alloin 1988, Burstein et al 1988, Ferguson et al 1991, Bruzual & Charlot 1993) would show that the young star and old, evolved star models could produce almost indistinguishable FUV energy distributions at low spectral resolution (see Figure 5). It would be necessary to consider other information, especially spectral features, to resolve the ambiguity.

Figure 5. The Astro/HUT spectrum of the gE galaxy NGC 1399 in the Fornax Cluster (see Figure 2). The histogram is the observed flux in 10 Å bins. The solid line shows the best-fitting Kurucz (1991) solar abundance model atmosphere, which has Te = 24000 K. The dashed line is a model from Rocca-Volmerange & Guiderdoni (1988) for an old galaxy with continuing star formation. This contains hotter starlight than is present in the galaxy. The inset shows the observed spectrum near the C IV 1550 Å doublet compared with continuous star forming models. From Ferguson et al (1991).

Because of the limited signal-to-noise of individual spectra, IUE studies of possible far-UV absorption lines have been based on summed spectra for either the same or several objects. A serious complication is a systematic background of "fixed pattern" and camera artifact features in long-exposure IUE spectra (Crenshaw et al 1990). Welch (1982) analyzed 12 exposures of the center of M31 and detected weak absorption features at 1260 Å (Si II + S II), 1302 Å (O I + Si II + Si III), and 1335 Å (C II). The features were confirmed by BBBFL in M31, but they were weak or absent in a summed spectrum for 3 bright UVX E galaxies (BBBFL). These lines are characteristic of normal early-B stars (e.g. Fanelli et al 1992) but are considerably stronger in the stars than in any of the UVX sources. The strong Si IV (1400 Å) or C IV (1550 Å) absorption features associated with massive O stars were absent in both M31 and the summed E spectrum.

Welch (1982) and BBBFL argued that the weakness of the massive OB-star spectral features was inconsistent with recent star formation as the source of the UVX. BBBFL pointed out additional evidence in the form of continuum shapes. Although it is possible to produce a star-forming model whose spectrum matches the typical steep UVX far-UV spectral slope (see Figure 5), in fact many systems with younger populations (e.g. NGC 205 or 5102) have rather flat energy distributions in the 1200-3000 Å region. This is the signature of an aging starburst or a young starburst containing local extinction (Kinney et al 1993). Furthermore, the predicted young-star contamination of the optical band if the UVX originates in massive stars is significantly larger than limits from careful spectral synthesis studies - e.g. the 2% maximum at 4000 Å set by Rose (1985) using high-resolution spectra of 12 E galaxies. The uniformity of the UV-upturn slope and the absence of warm-star effects at wavelengths longer than 2000 Å are therefore additional evidence against the involvement of massive stars in the UVX phenomenon.

The best far-UV spectra of UVX galaxies, covering the range 900-1800 Å, were obtained by Astro/HUT with a photon-counting detector and calibration superior to IUE's. HUT spectra of 8 early-type systems confirmed the weakness of the massive OB-star spectral features (Ferguson et al 1991, Davidsen & Ferguson 1992, Brown et al 1995, Brown et al 1997). For example, they placed an upper limit on the strength of C IV 1550 Å of 3 Å equivalent width in NGC 1399, which formally excluded spectral synthesis models for continuous star formation with normal metal abundances (see Figure 5). Equally important, by extending spectral coverage below 1150 Å, HUT was able to detect turnovers in the UVX spectra that place firm upper limits on their effective temperatures of T_e 25000 K (equivalent to a B0 V star). This excludes continuing star formation models with a normal initial mass function (IMF). Only contrived models (e.g. invoking a truncated IMF or an unprecedented synchronization of star formation in different galaxies) can reconcile young populations with these results. The narrow T_e range, however, also appears to require an unusual degree of "fine-tuning" in an old population interpretation.

The HUT spectra also provide excellent limits on the amount of internal interstellar extinction. Because the slope of the far-UV spectral rise is nearly at the maximum encountered among hot stars (Dean & Bruhweiler 1985, Fanelli et al 1992), there is little room for interstellar reddening. In the HUT data, there is no evidence for extinction significantly in excess of the expected Galactic foreground. Since dust is normally associated with star-forming regions, this is yet further evidence of their unimportance in UVX galaxies.

The spectroscopic evidence therefore strongly corroborates the structural evidence from the last section that massive stars are not responsible for the UV-upturn.

Recently, Bica et al (1996) have produced composite spectra for groups of E galaxies from the IUE archives. They find evidence for broad absorption features at 1400 and 1600 Å in most UVX sources, which they identify with the Ly- satellite lines in intermediate-temperature (DA5) white dwarf stars. If true, this would be remarkable since only a very unusual population would have the requisite concentration of white dwarfs. The features are not present in the better quality HUT spectra discussed above, but Bica et al suggest they are confined only to the nuclei and have been diluted by the larger HUT entrance aperture. This can be checked with HST/STIS.