As discussed in detail by others at this conference, the absolute brightness of the extragalactic sky at visible and infrared wavelengths carries important information on the history of star formation and galaxy evolution throughout the age of the Universe. As first emphasized by Tinsley (1973), the diffuse background at the shorter UV wavelengths is also an important element of this modern day version of "Olbers' paradox" - but with several key differences with respect to the situation in the adjacent spectral regions.
For one, the observed integrated far-UV spectre of the different classes of galaxies are not particularly well modeled or understood at present. This is especially true of the UV emission from ellipticals and the bulges of spirals, which is believed reflect complex stages of late stellar evolution (cf. Burnstein et al. 1988; King et al. 1992 and references therein). However, the global galaxy luminosity function at UV wavelengths is almost entirely dominated by the emission from massive O and B stars contained in the star-forming regions of spirals and irregulars. Consequently, in the UV late type galaxies most likely account for close to 90% of the local galaxy luminosity density and therefore dominate the integrated galaxy background (cf. Milliard et al. 1992 for a recent discussion).
Another important difference with respect to the case at visible and longer wavelengths is that the integrated UV light of galaxies is accumulated over a relatively modest cosmological pathlength. Late type galaxies are generally surrounded by large halos of neutral hydrogen which permit little or no radiation to escape below the Lyman limit at 912 Å. At an observed wavelength of limit at 1500 Å, the Lyman limit is reached at z = 0.64, corresponding to a look-back time of limit at t limit at 3h-1 Gyr. This modest look-back time, combined with the short life times of main sequence OB stars, leads to the integrated UV light of galaxies being primarily a measure of the level of on-going star formation in the relatively local Universe rather than a measure of its total accumulated stellar content, as is the case at longer visible and infrared wavelengths.
Three rather different observational and theoretical approaches have so far been employed in assessing the possible contribution of galaxies to the extragalactic UV background. Although all three methods have considerable uncertainties associated with them, they nevertheless give answers that agree to better than a factor of 3-4. The integrated light of galaxies is almost certainly that of the three sources of diffuse extragalactic UV background light considered in this review that is most likely to generate an observationally significant flux.
3.1 Integration of Theoretical Galaxy Evolution Models
Following the pioneering work of Tinsley (1972), several groups have in recent years developed very elaborate and physically self-consistent models for galaxy evolution based on stellar formation and evolution theory (e.g. Guiderdoni & Rocca-Volmerange 1991; Bruzual & Charlot 1993). Although the primary motivation for these models is to explain faint galaxy counts and colors observed in the visible, as emphasized by Tinsley (1973), the predicted integrated background spectrum provides an important observational constraint on the models.
The general expression for the background at received wavelength limit at 0 due to the integrated light of galaxies is
where (z) = 0 / (1 + z) and 0(, z) is the total (co-moving) volume emissivity due to all classes of galaxies at all wavelengths < 0 and all redshifts z 0. The function 0(, z) encapsulates the luminosity and number evolution of all galaxies in the Universe at all wavelengths at all epochs. Needless to say, this function is not very well determined at present.
One complication is that the predicted UV fluxes for models of late type star-forming galaxies are less forgiving than at visible wavelengths in the sense that the emergent UV flux is very sensitive to not only the details of the assumed star formation history (see Fig. 4 of Bruzual & Charlot 1993 for a nice illustration), but also to the amount, detailed geometry and properties of any absorbing dust present in the galaxy (e.g. Bruzual, Magris, & Calvet 1988). Given these complications (the latter of which is usually ignored), and the fact that the faint optical counts still present a considerable challenge to the models (Koo & Kron 1992), too much faith cannot presently be placed in their extrapolation to far-UV wavelengths.
A recent discussion with emphasis on the ultraviolet has been given by Martin, Hurwitz & Bowyer (1991), who show that the current evolution models predict a rather wide range of spectral shapes for the background. Nonetheless, the predicted far-UV intensities generally span the I 40 - 240 photons s-1 cm-2 sr-1 Å-1 range - suggesting that the integrated UV light of galaxies should be within reach observationally.
3.2 UV Background Fluctuation Measurements
An almost heroic observational attempt to measure the galaxy contribution to the UV background has been undertaken by Martin & Bowyer (1989), who by means of a sounding rocket experiment searched for the small-scale fluctuations in the far-UV background expected from the integrated light of galaxies. This technique, which was first pioneered by Schectman (1973; 1974) at visible wavelengths, consists of fitting the measured power spectrum of the UV background to that expected from galaxies calculated on the basis of observed visible light correlation functions and assumed models for galaxy spectral evolution. According to Martin & Bowyer (1989) the observed radial power spectrum of the UV background on angular scales of 6-12' is consistent with the P -1.2 power law signature anticipated from galaxies (where is the inverse angular scale). However, as mentioned in Section 2.3, since the observed background fluctuations in the 1.5 x 3.0° area sampled by Martin & Bowyer (1989) were of very low amplitude ( I /I 5%) compared to the average background of < I > 220 photons s-1 cm-2 sr-1 Å-1, the amplitude of the power spectrum is also small: P( = 200 rad-1) = 3 ± 1 10-4(photons s-1 cm-2 sr-1 Å-1)2 rad-2. Consequently, Martin & Bowyer (1989) concluded that the integrated light of galaxies can at most contribute 20% of the total (galactic and extragalactic) background observed, corresponding to an intensity of I 40±13 photons s-1 cm-2 sr-1 Å-1.
Although subsequent analysis has questioned the validity of the fluctuation results (due to contamination from UV starlight scattered off the IRAS cirrus; cf. Sasseen et al. 1993), it is nonetheless remarkable that this intensity estimate is comparable to those obtained from the theoretical galaxy evolution models and the direct UV galaxy counts.
3.3 Extrapolation of Ultraviolet Galaxy Counts
The most convincing and direct demonstration that galaxies must provide a significant contribution to the extragalactic UV background has recently been given by Armand, Milliard & Deharveng (1993). These authors base their analysis on observed UV ( 2000 Å) galaxy counts obtained with a balloon borne UV telescope (Milliard et al. 1992). The balloon counts are complete down to a UV magnitude of m 18.5, which by itself yields an integrated galaxy background of I 30 photons s-1 cm-2 sr-1 Å-1. Armand et al. extrapolate this resolved portion of the background to fainter magnitudes by use of the Guiderdoni & Rocca-Volmerange (1991) galaxy evolution models. Because of the various observational and theoretical uncertainties, the total background can only be predicted with certainty to lie in the range I 40-130 photons s-1 cm-2 sr-1 Å-1. Nonetheless, this flux is in good agreement with those obtained through the other less direct approaches above, and clearly demonstrate that galaxies must be a significant source of extragalactic UV background radiation.
3.4 The Integrated UV Light of Quasars and AGNs
As opposed to the situation at higher X-ray energies, quasars and active galactic nuclei play only a marginal role in the case of the background in the UV. This conclusion follows implicitly from a point alluded to in Section 2.2, namely that current models for quasar and AGN evolution have difficulties explaining the level of meta-galactic ionizing background deduced from the proximity effect displayed by the Lyman forest absorption lines, and that this intensity is to begin with very faint compared to a nominal extragalactic UV background of I 100 photons s-1 cm-2 sr-1 Å-1.
Estimates of the quasar contribution to the UV background obtained through integration of models for quasar evolution are sensitive to the assumptions made concerning the average far- and extreme-UV quasar spectrum, the quasar turn-on epoch, and the intervening absorption. As an illustration, Figure 5 show the anticipated background spectra calculated for the three evolution models adopted by Bajtlik et al. (1988), and assuming an average F -0.5 quasar continuum spectrum. If the intervening Lyman continuum absorption is ignored, the resulting integrated background spectrum is a power law with the same spectral index as that assumed for the quasars. Including the effects of absorption lowers the predicted fluxes further by a factor ~ 3, and decreases the sensitivity to evolution effects by quenching the background contributions from higher redshifts emitted in the Lyman continuum. The intensity of the integrated quasar flux is predicted to be I 10 photons s-1 cm-2 sr-1 Å-1 throughout the far-UV. Very similar results are obtained for other quasar models (Martin & Bowyer 1989; Martin et al. 1991 and references therein). It follows that quasars probably at most contribute a few percent of the nominal extragalactic UV background flux.
Figure 5. The integrated UV background due to quasars calculated for the three evolution models adopted by Bajtlik et al. (1988) and used in Figure 2. An F -0.5 quasar continuum spectrum was assumed. The dashed curves show the unabsorbed backgrounds and the full curves the same spectra when intervening Lyman continuum absorption is taken into account.
As a parenthetical aside, the reason that quasars are a dominant background source in the X-rays, but not in the adjacent UV, can be traced to the fact that the nominal extragalactic far-UV background intensity of I 100 photons s-1 cm-2 sr-1 Å-1 is actually considerably brighter in terms of energy per octave in frequency than the extragalactic background flux in the X-rays.