Radiative transfer simulations of cosmic reionization I: methodology and initial results

We present a new hybrid code for large volume, high resolution simulations of cosmic reionization, which utilizes a N-body algorithm for dark matter, physically motivated prescriptions for baryons and star formation, and an adaptive ray tracing algorithm for radiative transfer of ionizing photons. Two test simulations each with 3 billion particles and 400 million rays in a 50 Mpc/h box have been run to give initial results. Halos are resolved down to virial temperatures of 10^4 K for the redshift range of interest in order to robustly model star formation and clumping factors. This is essential to correctly account for ionization and recombination processes. We find that the halos and sources are strongly biased with respect to the underlying dark matter, re-enforcing the requirement of large simulation boxes to minimize cosmic variance and to obtain a qualitatively correct picture of reionization. We model the stellar initial mass function (IMF), by following the spatially dependent gas metallicity evolution, and distinguish between the first generation, Population III (PopIII) stars and the second generation, Population II (PopII) stars. The PopIII stars with a top-heavy IMF produce an order of magnitude more ionizing photons at high redshifts z>10, resulting in a more extended reionization. In our simulations, complete overlap of HII regions occurrs at z~6.5 and the computed mass and volume weighted residual HI fractions at 5<z<6.5 are both in good agreement with high redshift quasar absorption measurements from SDSS. The values for the Thomson optical depth are consistent within 1-sigma of the current best-fit value from third-year WMAP.