2006ApJ...648..299S

We use our very deep U_nGRI catalog of z∼4, 3, and 2 UV-selected star-forming galaxies to study the cosmological evolution of the rest-frame 1700 Å luminosity density. The ability to reliably constrain the contribution of faint galaxies is critical here, and our data do so by reaching deep into the galaxy population, to M^*_LBG+2 at z∼4 and deeper still at lower redshifts (M^*_LBG=-21.0 and L^*_LBG is the corresponding luminosity). We find that the luminosity density at z≳2 is dominated by the hitherto poorly studied galaxies fainter than L^*_LBG, and, indeed, the bulk of the UV light at these epochs comes from galaxies in the rather narrow luminosity range L=(0.1-1)L^*_LBG. Overall, there is a gradual rise in total luminosity density starting at ≳4 (we find twice as much UV light at z∼3 as at z∼4), followed by a shallow peak or plateau within z∼3-1, finally followed by the well-known plunge to z∼0. Within this total picture, luminosity density in sub-L^*_LBG galaxies at z≳2 evolves more rapidly than that in more luminous objects; this trend is reversed at lower redshifts, z≲1–a reversal that is reminiscent of galaxy downsizing. We find that within the context of commonly used models there seemingly are not enough faint or bright LBGs to maintain ionization of intergalactic gas even as recently as z∼4, and the problem becomes worse at higher redshifts: apparently the universe must be easier to reionize than some recent studies have assumed. Nevertheless, sub-L^*_LBG galaxies do dominate the total UV luminosity density at z≳2, and this dominance highlights the need for follow-up studies that will teach us more about these very numerous but thus far largely unexplored systems.