The next question in line is whether a median attenuation of ~ 1.6 mag in the UV is reasonable at z > 2, when galaxies where at most a few Gyr old and, presumably, metal- and dust-poor. Both the Cosmic Far-IR Background (CIB) detected by COBE (, ) and the FIR-bright galaxies detected by SCUBA at z 1 (, , , ) demonstrate that dust was present at high redshift. The luminosity of the CIB is about 2.5 times higher than the luminosity of the UV-optical Background (), implying a proportionally higher contribution of the redshift-integrated dust emission. However, neither the CIB nor the SCUBA galaxies are telling us how the dust content in galaxies has evolved with redshift. In the case of the SCUBA galaxies, the redshift and luminosity distribution and the AGN fraction of the sources will need to be tackled before providing such information.
The time evolution of the UV luminosity density of galaxies and of the derivative SFR density (Figure 3) can be used to constrain the metal and dust enrichment of galaxies and, therefore, the intrinsic SFR density (). The stars which produce the observed UV luminosity at each redshift produce also metals and dust with negligible delay times, at most 100-200 Myr in the case of dust (). The obscuration from dust will produce an observed UV flux lower than the true flux. Once the effects of the dust on the observed UV emission are evaluated and removed, a new SFR density is calculated. The procedure is repeated iteratively till convergence (). A number of observational contraints are used in the model: no more than ~ 10% of the baryons are in galaxies; inflows/outflows keep the z = 0 metallicity of the gas in galaxies to about solar, with a ~ 15% mean residual gas content, and the z ~ 2-3 metallicity to about 1/10-1/15 solar (); the intrinsic SFR density at z = 0 must be comparable with that measured from H surveys (); the dust emission must reproduce the observed CIB and not exceed the FIR emission of local galaxies.
These constraints are still not enough to yield a unique solution; one of the missing ingredients is the behavior of the SFR density at z > 4, where there are no data points. Different assumptions will lead to different intrinsic SFR histories. The range of solutions is bracketed by Models A and B in Figure 3. Figure 4 shows, for each of the two solutions, the evolution of the dust column density in the average galaxy and the contribution to the CIB at selected wavelengths as a function of redshift. The latter is however dependent on the assumptions about the intrisic dust emission SED, which is not well constrained.
Figure 4. The evolution of the dust column density in galaxies (top panels), expressed as optical attenuation AV in magnitudes, and of the contribution to the CIB at selected wavelengths (bottom panels) as a function of redshift, for both Model A (left panels) and Model B (right panels). Models A and B are the intrinsic SFR densities described in Figure 3 and Section 4. For both models the dust optical depth remains relatively modest at all redshifts. Yet, this is sufficient to fully account for the CIB luminosity. The contributing flux to the CIB is shown at the observer's restframe wavelengths 140 µm, 240 µm, 450 µm, and 850 µm, in arbitrary units. The two models predict different mean redshifts for the main contributors at each wavelength. In theory, this difference could be used to discriminate between the two solutions. However, some caution should be used as the spectral shape of the CIB is quite dependent on the dust emission SED adopted for the individual galaxies (in our case a single temperature blackbody combined with a 2 emissivity model).
Model B resembles the SFR density derived from the obscuration corrected Lyman-break galaxies (Figure 3). This demonstrates that attenuations of about 1.6 mag in the UV are perfectly reasonable within the framework of a simple model of stellar and dust content evolution in galaxies.