The UDF05 Follow-up of the HUDF: II. Constraints on Reionization from z-dropout Galaxies

[Abridged] We detect three (plus one less certain) z-dropout sources in two separate fields of our UDF05 HST NICMOS images. These z~7 Lyman-Break Galaxy (LBG) candidates allow us to constrain the Luminosity Function (LF) of the star forming galaxy population at those epochs. By assuming a change in only M* and adopting a linear evolution in redshift, anchored to the measured values at z~6, the best fit evolution coefficient is found to be 0.43+-0.19 mag per unit redshift (0.36+-0.18, if including all four candidates), which provides a value of M*(z=7.2)=-19.7+-0.3. This implies a steady evolution for the LBG LF out to z~7, at the same rate that is observed throughout the z~3 to 6 period. This puts a strong constraint on the star-formation histories of z~6 galaxies, whose ensemble star-formation rate density must be lower by a factor 2 at ~170 Myr before the epoch at which they are observed. In particular, a large fraction of stars in the z~6 LBG population must form at redshifts well above z~7. Extrapolating this steady evolution of the LF out to higher redshifts, we estimate that galaxies would be able to reionize the universe by z~6, provided that the faint-end slope of the z>7 LF steepens to alpha~-1.9, and that faint galaxies, with luminosities below the current detection limits, contribute a substantial fraction of the required ionizing photons. This scenario gives however an integrated optical depth to electron scattering that is ~2sigma below the WMAP-5 measurement. Therefore, altogether, our results indicate that, should galaxies be the primary contributors to reionization, either the currently detected evolution of the galaxy population slows down at z>7, or the LF evolution must be compensated by a decrease in metallicity and a corresponding increase in ionization efficiency at these early epochs.