Planetary system

Stability (learning theory)

Physics

Term (time)

Astronomy

Machine learning

Satellite

Exoplanet

Computer science

Solar System

Orbital elements

Computational physics

Planet

Orbital mechanics

Astrobiology

Proceedings of the National Academy of Sciences

Significance Observations of planets beyond our solar system (exoplanets) yield uncertain orbital parameters. Particularly in compact multiplanet systems, a significant fraction of observationally inferred orbital configurations can lead to planetary collisions on timescales that are short compared with the age of the system. Rejection of these unphysical solutions can thus sharpen our view of exoplanetary orbital architectures. Long-term stability determination is currently performed through direct orbital integrations. However, this approach is computationally prohibitive for application to the full exoplanet sample. By speeding up this process by up to five orders of magnitude, we enable precise exoplanet characterization of compact multiplanet systems and our ability to examine the stability properties of the multiplanet exoplanet sample as a whole.

Predicting the long-term stability of compact multiplanet systems