A new paradigm for reproducing and analyzing N-body simulations of planetary systems
Abstract
The reproducibility of experiments is one of the main principles of the scientific method. However, numerical N-body experiments, especially those of planetary systems, are currently not reproducible. In the most optimistic scenario, they can only be replicated in an approximate or statistical sense. Even if authors share their full source code and initial conditions, differences in compilers, libraries, operating systems or hardware often lead to qualitatively different results. We provide a new set of easy-to-use, open-source tools that address the above issues, allowing for exact (bit-by-bit) reproducibility of N-body experiments. In addition to generating completely reproducible integrations, we show that our framework also offers novel and innovative ways to analyse these simulations. As an example, we present a high-accuracy integration of the Solar system spanning 10 Gyr, requiring several weeks to run on a modern CPU. In our framework, we can not only easily access simulation data at predefined intervals for which we save snapshots, but at any time during the integration. We achieve this by integrating an on-demand reconstructed simulation forward in time from the nearest snapshot. This allows us to extract arbitrary quantities at any point in the saved simulation exactly (bit-by-bit), and within seconds rather than weeks. We believe that the tools we present in this paper offer a new paradigm for how N-body simulations are run, analysed and shared across the community.
- Publication:
-
Monthly Notices of the Royal Astronomical Society
- Pub Date:
- May 2017
- DOI:
- 10.1093/mnras/stx232
- arXiv:
- arXiv:1701.07423
- Bibcode:
- 2017MNRAS.467.2377R
- Keywords:
-
- methods: numerical;
- gravitation;
- planets and satellites: dynamical evolution and stability;
- Astrophysics - Earth and Planetary Astrophysics;
- Astrophysics - Instrumentation and Methods for Astrophysics
- E-Print:
- 7 pages, 4 figures, accepted for publication in MNRAS, REBOUND code available at https://github.com/hannorein/rebound , script and data files to reproduce plots in the paper available at https://github.com/hannorein/reproducibility-paper