Abstract

Poster - Splinter eScience

Thursday, 15 September 2022, 16:15   (SFG 1040 / virtual eScience)

Extracting information on exoplanets from transit spectroscopy utilizing deep learning

C Kilby, T Carroll, K Poppenhaeger
Leibniz-Institut für Astrophysik Potsdam (AIP)

Exoplanets cause distortions of the stellar spectral line profiles during transits. The shape of these distortions encodes information about the exoplanet properties. However, stellar magnetic activity can also cause line distortions which can be difficult to tease apart. In this master thesis, we train a neural net to make predictions about the particular sky-projected position of the exoplanet over the stellar disk. As a training set, we simulate stellar line profiles distorted by the Rossiter-McLaughlin effect during an exoplanetary transit. Our preliminary results show that we can extract the sky-projected spin-orbit misalignment angle and impact parameter of the system through analysis of just a few line profiles. Even with a low vsini, there is sufficient information within the line profile distortions to determine the spin-orbit misalignment angle of a given system. In future simulations, we will introduce noise and more stellar characteristics, such as differential rotation, into the simulated data to see if the model is able to learn to a similar degree of accuracy.