Numerical resolution limits on subhalo abundance matching

Abstract

Subhalo abundance matching (SHAM) inserts galaxies into dark matter only simulations of the growth of cosmic structure in a way that requires minimal assumptions about galaxy formation. A galaxy is placed at the potential minimum of each distinct self-bound subhalo with a luminosity which is a monotonically increasing function of the maximum mass (or circular velocity) attained over the subhalo's earlier history. Galaxy and subhalo properties are linked by matching model and observed luminosity functions. Simulated structures can then be compared in detail with observation, for example, through galaxy correlation statistics, group catalogues, or galaxy-galaxy lensing. Robust astrophysical conclusions can be drawn from such a comparison only on scales which are unaffected by the numerical limitations of the simulation. Here we compare results for the Millennium Simulation (MS) with subhalos defined using the SUBFIND to those obtained applying identical analysis to the much higher resolution Millennium-II. Correlation statistics on scales between 200 kpc and 2 Mpc converge to within 20% only for subhalos with masses at infall corresponding to at least 1000 simulation particles in the MS. Numerically converged results can be obtained to much lower infall particle number, if galaxies are followed even after their associated subhalos have been tidally disrupted, as in most recent semi-analytic galaxy formation simulations. This allows robust comparison between simulation and observation over a wider dynamic range in mass than for a SHAM analysis which ignores such "orphan" galaxies.