N-body realizations of cuspy dark matter haloes

Abstract

We describe an algorithm for generating equilibrium initial conditions for numerical experiments with dark matter haloes. Our haloes are modelled using a general form for the mass density p{r), making it possible to represent most of the popular density profiles in the literature. The finite mass 7-models and the cuspy density profiles found in recent high-resolution cosmological TV-body simulations having a density power-law fall-off at large distances proportional to are included as special cases. The algorithm calculates the phase-space distribution function of each model assuming spherical symmetry and either an isotropic velocity dispersion tensor or an anisotropic velocity dispersion tensor of the type proposed by Osipkov and Merritt. The particle velocities are assigned according to the exact velocity distribution, making this method ideal for experiments requiring a high degree of stability. Numerical tests confirm that the resulting models are highly stable. This approach is motivated by the instabilities that arise when a local Maxwellian velocity distribution is adopted. For example, after approximating the velocity distribution by a Gaussian we show that a Hernquist halo with an initial r(^-1) density cusp immediately develops a constant density core. Moreover, after a single crossing time the orbital anisotropy has evolved over the entire system. Previous studies that use this approximation to construct halo or galaxy models could be compromised by this behaviour. Using the derived distribution functions we show the exact 1-d velocity distributions and we compare them with the Gaussian velocity distributions with the same second moment for different distances from the halo centre. We show that instabilities arise because a Gaussian velocity distribution is a very poor approximation to the true velocity distribution of particles. We also perform a series of numerical simulations evolving several dark matter halo models in isolation, with the intention of checking the stability of the initialization procedure in both configuration and velocity space. A subset of the models are evolved under the assumption that the velocity distribution at any given point is a Gaussian and the time evolution of the density profiles and velocity structure is monitored. Finally, a number of applications are discussed, including issues of relaxation in dark matter haloes as well as mergers of haloes in scattering experiments. [brace not closed]