Efficient, uninformative sampling of limb-darkening coefficients for a three-parameter law

TL;DR

This paper introduces an efficient method for sampling limb-darkening coefficients of a three-parameter law, enabling more accurate modeling of stellar limb-darkening effects in astronomical observations.

Contribution

It presents a novel, fast sampling technique for three-parameter limb-darkening coefficients using a cone transformation, improving over previous two-parameter methods.

Findings

Uniform samples are physically valid in 97.3% of cases.

The method covers 94.4% of the allowed parameter space.

Provides open-source code for sampling and validation.

Abstract

Stellar limb-darkening impacts a wide range of astronomical measurements. The accuracy to which it is modelled limits the accuracy in any covariant parameters of interest, such as the radius of a transiting planet. With the ever growing availability of precise observations and the importance of robust estimates of astrophysical parameters, an emerging trend has been to freely fit the limb-darkening coefficients (LDCs) describing a limb-darkening law of choice, in order to propagate our ignorance of the true intensity profile. In practice, this approach has been limited to two-parameter limb-darkening laws, such as the quadratic law, due to the relative ease of sampling the physically allowed range of LDCs. Here, we provide a highly efficient method for sampling LDCs describing a more accurate three-parameter non-linear law. We first derive analytic criteria which can quickly test if a set of LDCs are physical, although naive sampling with these criteria leads to an acceptance rate less than 1%. We then show that the loci of allowed LDCs can be transformed into a cone-like volume, from which we are able to draw uniform samples. We show that samples drawn uniformly from the conal region are physically valid in 97.3% of realizations and encompass 94.4% of the volume of allowed parameter space. We provide Python and Fortran code (LDC3) to sample from this region (and perform the reverse calculation) at https://github.com/davidkipping/LDC3, which also includes a subroutine to efficiently test whether a sample is physically valid or not.