Monthly Notices of the Royal Astronomical Society (February 21, 2013) 429 (2): 1088-1101
Beyond convergence studies and comparison of different codes, there are essentially no controls on the accuracy in the non-linear regime of cosmological N body simulations, even in the dissipationless limit. We propose and explore here a simple test which has not been previously employed: when cosmological codes are used to simulate an isolated overdensity, they should reproduce, in physical coordinates, those obtained in open boundary conditions without expansion. In particular, the desired collisionless nature of the simulations can be probed by testing for stability in physical coordinates of virialized equilibria. We investigate and illustrate the test using a suite of simulations in an Einstein de Sitter cosmology from initial conditions which rapidly settle to virial equilibrium. We find that the criterion of stable clustering allows one to determine, for given particle number N in the “halo” and force smoothing, a maximum red-shift range over which the collisionless limit may be represented with desired accuracy. We also compare our results to the so-called Layzer Irvine test, showing that it provides a weaker, but very useful, tool to constrain the choice of numerical parameters. Finally we outline in some detail how these methods could be employed to test the choice of the numerical parameters used in a cosmological simulation.