gabrielfougeron
commented
9 years ago Current investigations tend to show that : (cf Gronchi, G. (2002). Theoretical and computational aspects of collision singularities in the N-body problem.)
- Non-collision singularities are arbitrarily close to collision singularities in some sense.
- Binary singularities are ALWAYS collision singularities.
- Binary collision regularization is an active area of research, and might prove feasable (in our code).
It might thus be interesting to restict ourselves to the case of binary collisions.
This issue concerns the singularity of the system of ODEs for long time integration. These events can be categorized into two groups :
Currently, the collision model is the following : Two planets collide iff they are closer than a given threshold. The collision happends right now, in particular, it can only happen at a given time step.
Idea : Predict the time of impact and only then decide if the bodies should collide. More generally, one can criticize the physical grounds of the naive model, which is not satisfying enough. A better model should be designed, giving more space for physical interpretation.