Ultraviolet radiative feedback during the advanced stages of reionization

The ionizing ultraviolet background (UVB) during reionization can suppress the gas content of low-mass galaxies, even those capable of efficient atomic cooling (i.e. with virial temperatures Tvir ≳ 104 K). This negative radiative feedback mechanism can thus reduce the star formation efficiencies of these haloes, which may delay the completion of reionization. In this work, we explore the importance of UV radiative feedback on T vir ≳ 104 K haloes during the middle and late stages of reionization. We do not try to self-consistently model reionization; instead, we explore a large parameter space in an attempt to draw general, robust conclusions. We use a tiered approach. Using one-dimensional hydrodynamical simulations, we model the ability of gas to collapse on to haloes of various masses under UVBs of various intensities. We then generate realistic, parametrized maps of the inhomogeneous UVB, using large-scale semi-numeric simulations. By combining these results, we find that under all reasonably conservative scenarios, UV feedback on atomically cooled haloes is not strong enough to notably delay the bulk of reionization. Such a delay is only likely if ionizing efficiencies of z ≳ 10 sources are much higher (over two orders of magnitude) than z ∼ 6 data seem to imply. Towards the end of reionization, star formation can be quenched only in haloes in a narrow mass range close to the atomic-cooling threshold. This result depends only weakly on the intensity of the UVB: quenching star formation in haloes just twice as massive requires an order of magnitude increase in source ionizing efficiencies. This implies that the natural time-scale for the bulk of reionization is the growth of the global collapsed fraction contained in Tvir ≳ 104 K haloes. Thus, the likely reionization scenario would involve a small H ii filling factor 'tail' extending to high redshifts, governed by more complicated feedback on Tvir ≲ 104 K objects, followed by a period of relatively rapid evolution in the H ii filling factor. Furthermore, our results underscore the importance of extended dynamic ranges when modelling reionization. Simulations must be capable of resolving haloes with mass ≳108 M⊙, even when modelling the late stages of reionization, while at the same time being large enough to capture H ii regions several tens of Mpc in size. © 2008 The Authors