Item

Abstract

When fitting transiting exoplanet lightcurves, it is usually desirable to have ranges and/or priors for the parameters which are to be retrieved that include our degree of knowledge (or ignorance) in the routines which are being used. In Markov Chain Monte Carlo (MCMC) routines, for example, these enter as prior distributions. These can either represent our current knowledge of the distribution of such parameters (e.g., based on their observed values) or physically plausible parameters ranges to be sampled. Among the parameters that are constrained by transiting exoplanet lightcurves, there are two which are of much physical significance: the impact parameter of the orbit, $b = (a/R_*)\cos i $, and the planet-to-star radius ratio, $p = R_p/R_s$ (which defines the transit depth, $\delta = p^2$). These two are natural parameters to extract and constrain as they usually have well defined limits. A common set of "uninformative" priors used for those two parameters are uniform priors. However, this poses a sampling problem especially important for grazing orbits: given that we sample a value $p_i$ from the prior on $p$, the only physically plausible values for $b$ to be sampled given $p_i$ are those that satisfy $b < 1 + p_i$. If we simply reject the sample if the sampled value of b is greater than $1 + p_i$, then we will reject points from a significant portion of the prior area depending on its size. It is desirable, thus, to have an algorithm that efficiently samples values from the physically plausible zone in the $(b,p)$ plane. Here we present such an algorithm.