Optical images of the extragalactic sky show a variety of galaxies with various luminosities and distances, all seen in projection. During the past decade ultra-deep optical imaging using CCDs over the wavelength range 0.3-1 µm have revealed a high surface density of faint blue galaxies. At a flux level corresponding to 1 photon/pixel/minute collected in a 4-meter telescope there are about 300,000 galaxies per square degree on the sky. These galaxies have apparent magnitudes between 25 and 28 B mag. Although too faint for spectroscopic redshift determination, several tests indicate that the redshift of these galaxies extends between 0.7 and 3. Observations suggest that we may be seeing many of these galaxies at an epoch of formation of much of their stellar content. The resulting UV-bright spectrum, when redshifted to redshifts of 1-3, could produce the observed blue spectral shape.
The extragalactic background radiation in the optical part of the spectrum originates in young stellar populations in galaxies at moderate to high redshift. The subject has a long history, but only recently have CCD surveys achieved the required sensitivity to capture most of the extragalactic light of the night sky. Perhaps one of the more intriguing questions is the nature of the 25-27 B magnitude so-called "faint blue galaxies" (FBGs). While there are an increasing number of clues to the physical nature of the FBGs (colors, numbers, correlations, morphological scale lengths, and rough redshift limits), I think it is fair to say that we don't even fully understand the type mix and star formation activity of the more nearby 0.1 < z < 0.4 galaxies. For example, the possibility of a redshift-dependent luminosity function is still an open question. The situation for the FBGs, being a superposition of galaxies over a greater span of look-back time, is bound to be more complicated. Happily, the summed extragalactic background light at some wavelength may be calculated directly from the galaxy number counts. Of course this does not affect our state of ignorance regarding the faintest sources of this radiation, particularly at blue and UV wavelengths.
The intensity of the extragalactic background light (EBL) is influenced strongly by galaxy evolution (stellar lifetimes and population evolution), and to a much lesser extent by cosmology (Harrison 1964). Number-magnitude counts which rise with magnitude like dex (0.4 mag) or faster continue to add to the EBL at the faintest magnitudes, and if this slope never falls below 0.4 the surface brightness of the extragalactic sky would approach that of a typical galaxy. These considerations are related to Olbers' paradox. The finite luminous lifetime of stellar populations offers a way out of both problems.
1.1 Theoretical Estimates of the EBL
Whitrow & Yallop (1965) first gave the form of the bolometric flux in an arbitrary cosmology with coeval galaxies starting at some formation redshift. Partridge & Peebles (1967) estimated the light from primeval galaxies in a model in which star formation began at high redshift giving the EBL a 30 µm IR excess and large angular sizes. The optical EBL is probably due to evolving spirals: Wyse (1985) estimated that less than 18% of the EBL light is from ellipticals, and that the elliptical light is red. Models of mild continuous star formation in galaxies tend to predict EBL similar to that observed in the faint galaxy population. Yoshii & Takahara (1988) calculated a no-evolution EBL flux of F = 1.1 x 10-6 erg cm-2 sec-1 sr-1 at 0.36 µm wavelength. For evolving galaxies with formation redshifts ranging from 3 to 5 (and nearly independent of deceleration parameter) they estimate an EBL about two times higher than this, close to the observed value, but the shape of their EBL spectrum is very different than that observed (see below). EBL from faint galaxies is very sensitive to evolution and relatively insensitive to cosmology, particularly at wavelengths where much of the redshifted UV from early star formation appears. Differential galaxy counts in the blue part of the spectrum are relatively more sensitive to luminosity evolution. The sum total EBL, being an integral over the flux times number counts at each wave band, is even less sensitive to cosmology. Generally, small changes in the galaxy luminosity function can compensate for changes in .
1.2 Upper Limits to the Diffuse EBL
It is useful to derive the diffuse EBL for comparison with other observations of diffuse background light at various wavelengths. Upper limits to the EBL have been set by all-sky photometry (Roach & Smith 1968; Mattila 1976; Dube et al. 1977, 1979). Progress in this field was reviewed by Toller (1983). The optical background is dominated by foreground emission (atmosphere, zodiacal light, galactic cirrus, aurora) which makes any absolute EBL measurement difficult; this makes optical EBL more elusive than its counterparts at some other wavelengths (e.g., sub-mm, radio, X-ray) where direct absolute measurements of the diffuse background emission are possible. As a result, discussion of the optical EBL is necessarily focussed on the contribution to the EBL from discrete sources to about 1 arcminute diameter. Total diffuse EBL chopping experiments have suffered from systematic errors larger than the EBL itself. For example, in chopping between a Lynds dark nebula and nearby blank sky the Lynds nebulae were unfortunately brighter than neighboring blank sky because of the back scattered Galactic starlight from nebular dust. One would also need to know where the "dark" nebula is in relation to the edge of the galaxy. These studies are very important since they are the only ones that attempt a direct measurement of the diffuse optical EBL on angular scales larger than one arcminute. A series of observations currently being undertaken by R. Bernstein, B. Madore, & W. Freedman using the Carnegie telescopes in a clever modern variation of Dube's experiment may yield a more accurate determination of the diffuse optical EBL.
A convenient but unfortunate unit often used to express the light of the night sky is S10(), which is the equivalent bolometric surface brightness of a 10th mag star of specified spectral type measured at wavelength , if its light is uniformly spread over a square degree of sky. One gets a feeling for the possibilities for systematic error in all-sky EBL photometry by reviewing the relative intensity of the competing sources: in units of S10,V, airglow = 40 (atomic lines) plus 50 (bands and continuum), zodiacal light away from the zodiac = 100, stars fainter than 6th V mag = 30 (galactic poles) or 95 (mean sky) or 320 (mean in galactic plane), diffuse galactic light and cirrus = 20, summing to a mean sky zenith background of 290 S10,V at low geomagnetic latitudes with no moon. By comparison, the detected EBL from discrete galaxies less than 1 arcmin in size is 0.5 S10,V, a mere 1.7 x 10-3 of the other diffuse optical backgrounds. CCDs can, however, be calibrated to a reproducible accuracy of 10-5 so that accurate subtraction of diffuse backgrounds on angular scales larger than the CCD is possible in principle. Since measurements are always made at some wavelength, and the spectrum of the EBL is not known a-priori, a more physical unit would be F = F in W m-2 sr-1, which gives equal weight to the energy contribution at each wavelength.
All sky photometry has yielded upper limits around 4 S10(V), where 1 S10(V) = 1.2 x 10-9 erg cm-2 sec-1 sr-1 Å-1 = 5.3 x 10-9 W m-2 sr-1, at 4400 Å. Dube et al. (1979) observed an EBL at 5115 Åof 1 ± 1.2 S10 (5115 Å) corresponding to F < 10-5 erg cm-2 sec-1 sr-1 at 5100 Å, about three times higher than the EBL due to the discrete faint blue galaxies (FBGs). If the objects dominating the EBL appear smaller than 5 arcsec, as the data imply, then the direct CCD imaging sensitivity is over 100 times the sensitivity of the integrated sky chopping techniques. The diffuse EBL on scales of 10-30 arcsec at K-band is currently better constrained: assuming Gaussian statistics Boughn et al. (1986) obtained an upper limit for the total diffuse K-band EBL of 7 x 10-20 erg cm-2 Hz-1 sr-1 = 9 x 10-6 erg cm-2 s-1 sr-1 at 2.2 µm for fluctuations on scales of > 10 arcsec, only slightly above the summed flux from discrete objects in K-band (see below). Gunn (1965) & Shectman (1974) considered statistically the fluctuations in the EBL due to galaxy clustering. This is directly measurable in deep wide-field CCD imaging data, but involves assuming a distribution function. This may soon be possible, since some ultra-deep CCD fields have been reimaged many times, permitting extraction of the fluctuation distribution function separately for the detector + sky noise and the diffuse EBL.