squishyplanet

Welcome to the documentation for squishyplanet! This lightweight package is designed to produce realistic lightcurves and phase curves of non-spherical exoplanets.

Most of the time, assuming that an exoplanet is a perfect sphere is a great approximation. However, in cases where we both a) expect the planet to be slightly deformed (either through gravitational interaction with its host star or through its own rapid rotation) and b) have high-precision data, fitting for its triaxial shape can provide additional constraints on the planet’s interior properties and evolution. squishyplanet can generate models of these triaxial planets, then leaves the choice of inference framework up to you.

In the limiting case where the planet is forced to be spherical, squishyplanet is designed to be as accurate as jaxoplanet and starry (see Compare with jaxoplanet notebook and Create a phase curve). Since squishyplanet uses its own implementation of the polynomial limb darkening model presented in Agol, Luger, and Foreman-Mackey 2020, it can handle complex limb darkening profiles even while accounting for the planet’s non-circular, potentially time-varying, projected shape.

Though it is generally slower than jaxoplanet given its increased reliance on numerical solutions and more complex underlying model, squishyplanet is also built on JAX and can be just-in-time compiled for speed. Users can expect reasonably-sized transit-only calculations to take ~10s of ms. However, phase curve calculations, which rely on Monte Carlo integrations at each timestep, are much slower than transit calculations. So, users should expect phase curve evaluations to be much slower, on the order of 100s of ms per evaluation. Be aware of this when selecting an inference framework if trying to fit actual data.

Also note that although JAX can technically compute the gradient of the likelihood function with respect to the model parameters, as of now, these calculations take significantly longer than the forward model evaluations. This is a performance issue that may be revisited in the future, but for now, gradient-dependent techniques like HMC are likely impractical with squishyplanet.

We recommend that potential users start with the geometry visualizations to get a sense of the coordinate system and how the planet is defined. Those interested in contributing to the code, or who find issues/want some clarification, should check out the contributing page and open an issue or pull request. Happy squishing!

Attribution

[insert JOSS citation/bibtex here someday]