We now know that most metal-rich stars have planets. We still can’t detect small planets around most stars or distant planets at all. We only see special combinations of planetary size and distance, and everything else is outside our ability to see. We’d love to know if solar systems could have twenty planets, or four Jupiters, or a lot of wild science fiction configurations. Now, a new algorithm created by a team led by Daniel Tamayo is being set loose to try and figure out what systems are and are not stable. According to Tamayo: Separating the stable from the unstable configurations turns out to be a fascinating and brutally hard problem. 

To make sure a planetary system is stable, astronomers need to calculate the motions of multiple interacting planets over billions of years and check each possible configuration for stability – a computationally prohibitive undertaking.

Rather than brute force their models, this new software samples how a system behaves over 10,000 orbits and looks for characteristic resonances and other dynamical characteristics that may predict instabilities. They then use artificial intelligence (AI) that has been trained to know what unstable systems look like to determine if the observed system will become unstable. They estimate this speeds up the modeling process by a factor of 100,000.

Machine learning algorithms like this one aren’t always doing what we think they’re doing, so there is still a need to run brute-force models now and then to try and verify this system’s predictions. Still, this is a start and it provides guidance on what systems can broadly be ignored as not stable, and which we expect to see over and over throughout the galaxy.

More Information

Simons Foundation news release 

“Predicting the Long-Term Stability of Compact Multi-Planet Systems,” Daniel Tamayo et al., 2020 Proceedings of the National Academy of Sciences