Robustness of Cosmological Simulations I: Large Scale Structure

Abstract

The gravitationally-driven evolution of cold dark matter dominates the formation of structure in the Universe over a wide range of length scales. While the longest scales can be treated by perturbation theory, a fully quantitative understanding of nonlinear effects requires the application of large-scale particle simulation methods. Additionally, precision predictions for next-generation observations, such as weak gravitational lensing, can only be obtained from numerical simulations. In this paper, we compare results from several N-body codes using test problems and a diverse set of diagnostics, focusing on a medium resolution regime appropriate for studying many observationally relevant aspects of structure formation. Our conclusions are that -- despite the use of different algorithms and error-control methodologies -- overall, the codes yield consistent results. The agreement over a wide range of scales for the cosmological tests is test-dependent. In the best cases, it is at the 5% level or better, however, for other cases it can be significantly larger than 10%. These include the halo mass function at low masses and the mass power spectrum at small scales. While there exist explanations for most of the discrepancies, our results point to the need for significant improvement in N-body errors and their understanding to match the precision of near-future observations. The simulation results, including halo catalogs, and initial conditions used, are publicly available.