Skip to main content
SearchLoginLogin or Signup

Anomaly detection techniques for unusual chemistry in exoplanet atmospheres

Presentation #200.04 in the session Exoplanet Atmospheres: Giant Planets (Oral Presentation)

Published onOct 23, 2023
Anomaly detection techniques for unusual chemistry in exoplanet atmospheres

The characterization of the chemical composition and physical parameters of the atmospheres of extra-solar system planets is the primary goal of transit transmission spectroscopy. The process of extracting this information involves sophisticated forward radiative transfer models of varying complexity, which assume a particular chemical composition, thermal structure and/or dynamics to fit to the observed spectrum. The extreme diversity among the thousands of planets that have been discovered up to date is bound to result in unique and unexpected chemical compositions that are difficult to foresee and directly implement in the existing models. We demonstrate the use of various machine learning anomaly detection techniques to identify planets with unusual chemistry. We use a large database of over 100,000 synthetic spectra, which were prepared for the 2022 ARIEL data challenge. We consider several machine learning methods for both outlier and novelty detection and demonstrate that they are successful in flagging planets with chemical compositions which deviate from the chemical inventory of the default model.

No comments here