Thesis

The James Webb Space Telescope (JWST) is opening new opportunities to study super-Earths and mini-Neptunes, the most common types of exoplanets in our Galaxy and which have no equivalent in our Solar System. These planets provide key insights into planetary formation and evolution, particularly the transition between rocky planets with thin atmospheres and planets surrounded by substantial volatile-rich envelopes.

This PhD project aims to characterise a population of super-Earths and mini-Neptunes observed with JWST, investigate the diversity of their atmospheres, and develop robust analysis methods to improve the scientific return of these observations. A major challenge is the presence of instrumental systematics, which can obscure or mimic atmospheric signals and affect the reliability of inferred planetary properties. A major challenge lies in instrumental systematics, which can compromise the robustness of inferred planetary properties.

The project will establish a homogeneous framework for reducing and analysing observations from different JWST instruments, wavelength ranges and observing modes, with a particular focus on MIRI. Machine Learning techniques, especially XGBoost, will be investigated to identify, predict and correct instrumental trends. Their performance will be compared with established reduction methods, and their impact on spectral uncertainties and atmospheric constraints will be quantitatively assessed.

Atmospheric properties, including molecular abundances, temperature structures, clouds and hazes, will be inferred using the Taurex retrieval framework and Bayesian methods. Applying a consistent methodology across a population of small exoplanets will enable meaningful comparisons between targets and help constrain the relationships between atmospheric composition, planetary properties, formation and evolution.

Finally, the project will prepare for ESA’s Ariel mission, scheduled for launch in 2031, by using the derived atmospheric properties as reference benchmarks and testing the developed analysis methods on simulated Ariel-like datasets. The results will contribute to a better understanding of small exoplanets and support the interpretation of future large-scale atmospheric surveys.