Predicting exoplanet phase curves from climate parameters using neural networks

Abstract:

The James Webb Space Telescope (JWST) will give us new opportunities to characterize terrestrial exoplanets. Thermal phase curves observed by JWST as a planet orbits its host star will allow us to probe the planet’s atmospheric dynamics. Determining a planet’s atmospheric properties from its phase curve requires running a Global Climate Model (GCM) many times. The problem is that GCMs are so numerically expensive that it will be difficult to do this in practice. We present a novel approach in which we train a neural network with GCM output to predict thermal phase curves from six planetary parameters: mass, density, surface pressure, optical thickness, and the mean molecular weight of the chemical species in its atmosphere. The goal is to provide this algorithm to astronomers so that they can quickly understand what their observed phase curves imply about a planet’s atmosphere.