Introduce physical boundaries
Closed this issue · 4 comments
Priors are fixed parameter ranges
Some combinations of these are unphysical
Currently, no check is made and the model is always calculated
We can introduce sensible boundaries of moons e.g. outside of Hill sphere (like TLS: not search for things that can't exist)
Then, Pandora can return a model for those unphysical situations so that their LogLikelihood is very bad (e.g., np.ones for the moon)
Assuming small planetary eccentricity:
r_hill = params.a_planet * (params.M_planet / (3 * M_star)) ** (1/3)
Assuming density of planet and moon is similar:
roche_limit = 1.25992 * params.r_planet ** (1/3)
r_hill Needs additional parameter: M_star. Estimate from R_star (which we need anyways) or let provide by user?
For unphysical moons: Should Pandora in fitting mode
- Return a model as if no moon would be present (planet-only), or
- Return a model containing nothing (which has very bad log-likelihood)?
Calculate stellar mass from a_planet and per_planet, and then use for Hill radius calculation?
Added hill_sphere_threshold which the user can set to e.g. 1.0 or 1.2.
Roche: Requires assumptions on density. What if gas planet and iron moon?