Time stepping N-body simulations
read the original abstract
Leapfrog integration has been the method of choice in N-body simulations owing to its low computational cost for a symplectic integrator with second order accuracy. We introduce a new leapfrog integrator that allows for variable timesteps for each particle in large N-body simulations. Tests with single particles in fixed potentials show that it behaves as a symplectic integrator. We then examine the results of both standard leapfrog and our temporally adaptive leapfrog on full N-body integrations of clusters and large scale structure establishing accuracy criteria for both methods. The adaptive method shows significant speed-ups over single step integrations---but the integrator no longer appears to be symplectic or, in the case of large scale structure simulations, accurate. This loss of accuracy appears to be caused by the way that the timestep is chosen, not by the integrator itself. We present a related integration technique that does retain sufficient accuracy. Although it is not symplectic, it is apparently better than previous implementations and is our current integrator of choice for large astrophysical simulations. We also note that the standard leapfrog difference equations used in cosmological N-body integrations in comoving coordinates are not symplectic. We derive an implementation of leapfrog that is in comoving canonical coordinates to correct for this deficiency.
This paper has not been read by Pith yet.
Forward citations
Cited by 1 Pith paper
-
Perturbation-theory informed integrators for cosmological simulations
LPT-matched integrators for cosmological simulations outperform FastPM with O(1-100) timesteps while convergence is limited to order 3/2 post-shell-crossing due to acceleration field irregularity.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.