N-body, parallel simulation using a Barnes-Hut algorithm: performance versus accuracy

The Barnes-Hut method facilitates prioritizing two-body interactions in an N-body system according to their likely significance in calculating the system's dynamics. In particular, it allows a consistent segregation of two-body interactions into those that should be treated by direct calculation versus those that can be aggregated in subsets and then treated by mean-field approximations. In this paper we describe the principles of the Barnes-Hut method, its use in parallelized N-body simulations, and the performance/accuracy trade-offs it presents. We present the latter in the context of results from simulation cases: N-bodies interacting via a gravitational potential, and N-bodies interacting via a Lennard- Jones potential. These should be available in the near future to operate as part of the ``Bootable Cluster CD'' parallel computation environment of the National Computational Science Institute of the Shodor Educational Foundation.