Jury

  • Prof. Timoteo CARLETTI (UNamur), Chair
  • Prof. Anne-Sophie LIBERT (UNamur), secretary
  • Prof. Benoît FRENAY (UNamur)
  • Prof. Ugo LOCATELLI (University of Rome Tor Vergata)
  • Prof. Konstantin BATYGIN (California Institute of Technology)
  • Prof. Adrien LELEU (University of Geneva)

Abstract

The growing number of extrasolar systems detected over the past three decades has made it necessary to develop fast and reliable methods for studying the long-term stability of planetary systems. In this work, we address the complex problem of the stability of compact three-planet systems—in which resonant and chaotic behaviors are intrinsically linked—using chaos indicators and machine learning. In the first part, we design four (variational and non-variational) chaos indicators and test their performance on a synthetic dataset. In the second part, we examine the predictive power of the chaos indicators when combined with different machine learning strategies. Finally, we apply these methods to a synthetic population generated by the Bern formation model, providing results on the long-term behavior and dynamical characterization of the multiple-planet systems in this population. Our analysis highlights the effectiveness of the dynamical tools developed here for assessing the stability of near-resonant systems and paves the way for their use in future space missions.