Machine learning inference of the interior structure of low-mass exoplanets

TL;DR

This paper demonstrates that machine learning, specifically mixture density neural networks, can efficiently infer the layered interior structure of low-mass exoplanets from observable data like mass, radius, and Love number $k_2$, reducing reliance on computationally intensive models.

Contribution

The study introduces a novel application of MDNs to infer exoplanet interior layer distributions from limited observational data, including the use of $k_2$ to improve accuracy.

Findings

MDNs accurately infer layer thickness distributions from mass and radius.

Including $k_2$ reduces degeneracy in interior structure models.

The method is adaptable to various interior structure assumptions.

Abstract

We explore the application of machine learning based on mixture density neural networks (MDNs) to the interior characterization of low-mass exoplanets up to 25 Earth masses constrained by mass, radius, and fluid Love number $k_2$. We create a dataset of 900$\:$000 synthetic planets, consisting of an iron-rich core, a silicate mantle, a high-pressure ice shell, and a gaseous H/He envelope, to train a MDN using planetary mass and radius as inputs to the network. For this layered structure, we show that the MDN is able to infer the distribution of possible thicknesses of each planetary layer from mass and radius of the planet. This approach obviates the time-consuming task of calculating such distributions with a dedicated set of forward models for each individual planet. While gas-rich planets may be characterized by compositional gradients rather than distinct layers, the method presented here can be easily extended to any interior structure model. The fluid Love number $k_2$ bears constraints on the mass distribution in the planets' interior and will be measured for an increasing number of exoplanets in the future. Adding $k_2$as an input to the MDN significantly decreases the degeneracy of the possible interior structures.