Simulating Cosmic Reionization

TL;DR

This paper presents the largest and most detailed cosmological simulations of cosmic reionization, resolving dwarf galaxy halos and modeling radiative feedback to understand the Epoch of Reionization.

Contribution

It introduces highly scalable N-body and radiative transfer codes capable of simulating large volumes with detailed galaxy resolution for reionization studies.

Findings

Simulations resolve halos >10^8 Solar masses in volumes up to (163 Mpc)^3.

Efficient parallelization on thousands of cores enables large-scale cosmological modeling.

Initial scientific results shed light on the role of dwarf galaxies in cosmic reionization.

Abstract

The Cosmic Dark Ages and the Epoch of Reionization constitute a crucial missing link in our understanding of the evolution of the intergalactic medium and the formation and evolution of galaxies. Due to the complex nature of this global process it is best studied through large-scale numerical simulations. This presents considerable computational challenges. The dominant contributors of ionizing radiation were dwarf galaxies. These tiny galaxies must be resolved in very large cosmological volumes in order to derive their clustering properties and the corresponding observational signatures correctly, which makes this one of the most challenging problems of numerical cosmology. We have recently performed the largest and most detailed simulations of the formation of early cosmological large-scale structures and their radiative feedback leading to cosmic reionization. This was achieved by running extremely large (up to 29 billion-particle) N-body simulations of the formation of the Cosmic Web, with enough particles and sufficient force resolution to resolve all the galactic halos with total masses larger than 10^8 Solar masses in computational volumes of up to (163 Mpc)^3. These results were then post-processed by propagating the ionizing radiation from all sources by using fast and accurate ray-tracing radiative transfer method. Both of our codes are parallelized using a combination of MPI and OpenMP and to this date have been run efficiently on up to 2048 cores (N-body) and up to 10000 cores (radiative transfer) on the newly-deployed Sun Constellation Linux Cluster at the Texas Advanced Computing Center. In this paper we describe our codes, parallelization strategies, scaling and some preliminary scientific results. (abridged)