By Kurtis Archer | 2025-04-24

A team of researchers at Switzerland's University of Bern and National Center of Competence in Research PlanetS have taken a huge step in the search for extraterrestrial life.

A machine-learning algorithm with model training in artificially generated planetary system data has discovered 44 stars that have a very high probability of having a rocky planet not dissimilar to Earth in their habitable zones.

"Earth-like planets" are exoplanets that have been identified as having a similar mass to Earth and reside within their star's habitable zone. Historically, they have been found by chance while surveying astronomical data -- but now scientists are using artificial intelligence (AI) to find them.

Jeanne Davoult is an astronomer who developed the algorithm while she was at the University of Bern.

Because humans have discovered only about 6,000 exoplanets so far, AI had to use available data to create synthetic planetary systems to serve as inputs so the algorithm could be trained on identifying patterns and making predictions.

"Our model is based on an algorithm that I developed and that was trained to recognize and classify planetary systems that harbor Earth-like planets," Davoult said in a statement.

"The model identified 44 systems that are highly likely to harbor undetected Earth-like planets. A further study confirmed the theoretical possibility for these systems to host an Earth-like planet," she said.

The journal Astronomy & Astrophysics published the study online April 9.

Training the algorithm

Researchers trained and tested the algorithm with data from the so-called Bern Model of Planet Formation and Evolution, which the scientific community has used and continually developed since 2003.

The model provided 53,882 simulated planetary systems around three different types of stars: G-type stars like Earth's sun, and two types of red dwarfs -- ones with about half the mass of Earth's sun and ones with about a fifth of the mass of Earth's sun.

The algorithm began searching for patterns and correlations, one of which was the existence of an outer gas giant coinhabiting a system with an inner rocky planet -- much like Jupiter and Earth.

Conversely, an anticorrelation was found between gas giants closer to their star and the presence of rocky planets near the same orbit space.

The lack of inner rocky planets in these systems is because these "hot Jupiters" formed farther away from the star but migrated inward over time, disturbing other planetary paths along the way.

A deep correlation Davoult found in earlier research is that the properties of the innermost planet in a system are a huge marker for how likely the system is to have an Earth-like planet in its habitable zone.

The mass, radius and orbital period of the innermost planet seem to be strong indicators for whether an alternate Earth exists in the system.

Another correlation shows that around stars like Earth's sun, Earth-sized planets in the habitable zone are likelier if the innermost detectable planet's orbital period is more than 10 days, or if its radius is greater than 2.5 times the radius of Earth.

Researchers trained the algorithm with these correlations and more, on the simulated data.

"The results are impressive: the algorithm achieves precision values of up to 0.99, which means that 99% of the systems identified by the machine-learning model have at least one Earth-like planet," Davoult said.

Search for alien life

With the algorithm recognizing patterns in the simulated data, researchers then had it look at real observations of star systems in our universe, which is how astronomers discovered the 44 stars with likely Earth-like satellites.

The team further studied the 44 systems identified by the algorithm and found that 95.5% of them would remain stable with the addition of an Earth-like planet.

These findings allow astronomers to focus on specific star systems while searching for life, instead of blindly reviewing all observable systems.

The European Space Agency (ESA) is planning a mission called PLAnetary Transits and Oscillations of stars (PLATO). It expects PLATO to discover several thousand planets by tracking their passages in front of their stars.

By applying Davoult's algorithm to PLATO's findings, scientists will be able to devote their attention to the systems found to have a higher chance of supporting a planet like Earth.

"This is a significant step in the search for planets with conditions favorable to life and, ultimately, for the search for life in the Universe," said Yann Alibert, a colleague on Davoult's team and co-author of the study, in the same statement that quoted Davoult.

Future space missions

Co-author Romain Eltschinger highlighted the broader value of the research results.

"These results are important for the scientific community, and particularly for future space missions such as PLATO or future mission concepts like LIFE, which will be dedicated to the discovery and characterization of small, cold planets," Eltschinger said in the statement.

The PLATO mission will use 26 cameras to observe more than 200,000 systems to determine what exoplanets may exist within their star's habitable zone.

The mission aims to measure the sizes of the exoplanets and discover if some of them have moons and rings. PLATO will launch in 2026 and will make its observations from a halo orbit at Lagrange point L2.

The Switzerland-based Large Interferometer For Exoplanets project is designed to measure and analyze the composition of exoplanets and investigate if they contain atmospheric biomarkers -- signs of alien life.