Abstract
The large dynamic range in some astrophysical N-body problems led to the use of adaptive multi-time-steps; however, the search for optimal strategies is still challenging. We numerically quantify the performance of the hierarchical Hamiltonian Splitting (HHS) integrator for collisionless simulations using a direct summation code. We compare HHS with the constant global time-step leapfrog integrator, and with the adaptive one (AKDK). We find that HHS is approximately reversible, whereas AKDK is not. Therefore, it is possible to find a combination of parameters such that the energy drift is considerably milder for HHS, resulting in a better performance. We conclude that HHS is an attractive alternative to AKDK, and it is certainly advantageous for direct summation and P3M codes. Also, we find advantages with GADGET4 (Tree/FMM) HHS implementation that are worth exploring further.
