I use simulations of cosmological reionization to quantify the effect of photoionization on the gas fraction in low mass objects, in particular the characteristic mass scale below which the gas fraction is reduced compared to the universal value. I show that this characteristic scale can be up to an order of magnitude lower than the linear theory Jeans mass, and that even if one defines the Jeans mass at a higher overdensity, it does not track the evolution of this characteristic suppression mass. Instead, the filtering mass, which corresponds directly to the scale over which baryonic perturbations are smoothed in linear perturbation theory, provides a remarkably good fit to the characteristic mass scale. Thus, it appears that the effect of reionization on structure formation in both the linear and nonlinear regimes is described by a single characteristic scale, the filtering scale of baryonic perturbations. In contrast to the Jeans mass, the filtering mass depends on the full thermal history of the gas instead of the instantaneous value of the sound speed, so it accounts for the finite time required for pressure to influence the gas distribution in the expanding universe. In addition to the characteristic suppression mass, I study the full shape of the probability distribution to find an object with a given gas mass among all the objects with the same total mass, and I show that the numerical results can be described by a simple fitting formula that again depends only on the filtering mass. This simple description of the probability distribution may be useful for semi-analytical modeling of structure formation in the early universe.

We develop an efficient method to study the effects of reionization history on the temperature-density relation of the intergalactic medium in the low density limit (overdensity δ < ∼ 5). It is applied to the study of photo-reionization models in which the amplitude, spectrum and onset epoch of the ionizing flux, as well as the cosmology, are systematically varied. We find that the mean temperature-density relation at z = 2 − 4 is well approximated by a powerlaw equation of state for uniform reionization models. We derive analytical expressions for its evolution and exhibit its asymptotic behavior: it is found that for sufficiently early reionization, imprints of reionization history prior to z ∼ 10 on the temperature-density relation are washed out. In this limit the temperature at cosmic mean density is proportional toWhile the amplitude of the radiation flux at the ionizing frequency of H i is found to have a negligible effect on the temperature-density relation as long as the universe reionizes before z ∼ 5, the spectrum can change the overall temperature by about 20%, through variations in the abundances of helium species. However the slope of the mean equation of state is found to lie within a narrow range for all reionization models we study, where reionization takes place before z ∼ 5. We discuss the implications of these findings for the observational properties of the Lyα forest. In particular, uncertainties in the temperature of the intergalactic medium, due to the uncertain reionization history of our universe, introduces a 30% scaling in the amplitude of the column density distribution while the the slope of the distribution is only affected by about 5%. Finally, we discuss how a fluctuating ionizing field affects the above results. We argue that under certain conditions, the loss of memory c 0000 RAS 2 Hui and Gnedin of reionization history implies that at late times, the temperature-density relation of a gas in a fluctuating ionizing background can be approximated by one that results from a uniform radiation field, provided the universe reionizes sufficiently early.

Stellar feedback plays a key role in galaxy formation by regulating star formation, driving interstellar turbulence and generating galactic scale outflows. Although modern simulations of galaxy formation can resolve scales of ∼ 10 − 100 pc, star formation and feedback operate on smaller, "subgrid" scales. Great care should therefore be taken in order to properly account for the effect of feedback on global galaxy evolution. We investigate the momentum and energy budget of feedback during different stages of stellar evolution, and study its impact on the interstellar medium using simulations of local star forming regions and galactic disks at the resolution affordable in modern cosmological zoom-in simulations. In particular, we present a novel subgrid model for the momentum injection due to radiation pressure and stellar winds from massive stars during early, pre-supernova evolutionary stages of young star clusters. This model is local and straightforward to implement in existing hydro codes without the need for radiative transfer. Early injection of momentum acts to clear out dense gas in star forming regions, hence limiting star formation. The reduced gas density mitigates radiative losses of thermal feedback energy from subsequent supernova explosions, leading to an increased overall efficiency of stellar feedback. The detailed impact of stellar feedback depends sensitively on the implementation and choice of parameters. Somewhat encouragingly, we find that implementations in which feedback is efficient lead to approximate self-regulation of global star formation efficiency. We compare simulation results using our feedback implementation to other phenomenological feedback methods, where thermal feedback energy is allowed to dissipate over time scales longer than the formal gas cooling time. We find that simulations with maximal momentum injection suppress star formation to a similar degree as is found in simulations adopting adiabatic thermal feedback. However, different feedback schemes are found to produce significant differences in the density and thermodynamic structure of the interstellar medium, and are hence expected to have a qualitatively different impact on galaxy evolution.