Machine learning: investigating the relationship between biogenic atmospheres, exoplanets and planetary systems

Abstract

The research proposal was to study planetary systems with exoplanets capable of sustaining biogenic atmospheres and understand their role in this context. For this task, the Earth similarity index (ESI) coupled to a multilevel modeling perspective was considered, as it is an index for classifying exoplanets by similarity to Earth. The characterization of these systems is important because astronomical missions whose objective is to identify biological markers that reveal the presence of life will be able to choose priority targets and eliminate less promising targets. 72 extrasolar systems were studied using unsupervised and supervised Machine Learning techniques with the aim of identifying the formation of clusters and investigating the multilevel relationship between planets and planetary systems. The work demonstrates that a wide variety of types of exoplanets can probably harbor atmospheres capable of being studied remotely, although these results do not consider the real internal constitutions of the objects studied, as they are unknown, preventing a historical reconstruction of the evolution process of these planets. The multilevel approach demonstrates that approximately 54% of the variation in the ESI value is due to the effect of the conditions of the planetary system where the exoplanet under study is located.