The physical properties and detectability of reionization-epoch galaxies

We present predictions drawn from cosmological hydrodynamic simulations for the physical, photometric and emission-line properties of galaxies present during the latter stages of reionization from z = 9 to 6. We find significant numbers of galaxies that have stellar masses exceeding 10(9) M-circle dot during this epoch, with metallicities exceeding one-thirtieth solar. Far from primeval 'first-star' objects, these objects exhibit a significant Balmer break, are likely to have reionized their infall regions prior to z = 9, are dominated by atomic rather than molecular cooling and are expected to be forming few if any metal-free stars. By z = 6, the space density of M-* > 10(10) M-circle dot objects is roughly equivalent to that of luminous red galaxies today. Galaxies exhibit a slowly evolving comoving autocorrelation length from z = 9 to 6, continuing a trend seen at lower redshifts in which the rapidly dropping bias counteracts rapidly increasing matter clustering. These sources can be marginally detected using current instruments, but modest increases in sensitivity or survey area would yield significantly increased samples. We compare to current observations of the z approximate to 6 rest-ultraviolet and Ly alpha line luminosity functions, and find good agreement. We also compare with the z similar to 7 object observed by Egami et al., and find that such systems are ubiquitous in our simulations. The intrinsic Ly alpha luminosity function evolves slowly from z = 9 to 6, implying that it should also be possible to detect these objects with upcoming narrow-band surveys such as Dark Ages z Ly alpha Explorer (DAzLE), if as we argue the detectability of Ly alpha does not drop significantly to higher redshifts. We make predictions for near-infrared surveys with the James Webb Space Telescope, and show that while a high density of sources will be found, Population III objects may remain elusive. We present and compare simulations with several recipes for superwind feedback, and show that while our broad conclusions are insensitive to this choice, a feedback model based on momentum-driven winds is mildly favoured in comparisons with available data.