Scientists have announced a groundbreaking discovery lurking beyond our solar system that rewrites humanity's understanding of the cosmos. A new study led by Princeton University in New Jersey identified more than 10,000 new possible planets trillions of miles from Earth. That included at least 11 worlds described by scientists as 'super-Earths.'
Using artificial intelligence to comb through NASA's old Transiting Exoplanet Survey Satellite (TESS) data from more than 83 million stars, the researchers uncovered thousands of hidden worlds that had gone undetected for years. This discovery suggests the formation of planets may be vastly more common than scientists previously realized, including around faint, overlooked stars once thought too difficult to study.
If even a fraction of the new candidates are confirmed, it would dramatically expand the known population of worlds beyond our solar system and strengthen the idea that planets may outnumber stars in the Milky Way. Currently, the NASA Exoplanet Science Institute has only confirmed the existence of 6,286 exoplanets throughout the galaxy. An exoplanet refers to any world outside of our solar system, from frozen balls to huge gas giants like Jupiter. To this point, fewer than 900 of these planets have been confirmed using the data on distant stars coming from TESS.
In their new paper, researchers said: 'Our findings more than double the number of known TESS exoplanet candidates.' The researchers hunted for 'transiting' planets, ones that crossed directly in front of their star as seen from Earth's perspective, causing a small, regular dip in the star's brightness. Previous TESS searches focused on brighter stars, which were easier to see. The goal of the new study, published in The Astrophysical Journal Supplement Series, was to scan much fainter stars, up to 10,000 times dimmer than humans can see with the naked eye.
The researchers started with raw data from TESS's first full year of observations, which took pictures of huge patches of sky every 30 minutes between 2018 and 2019. They then used the project's specially processed 'light curves,' which were graphs of how each star's brightness changed over time, to examine 83 million stars. To achieve this massive undertaking, the Princeton team trained two common types of AI models, called Random Forest classifiers, to sort through the data on these millions of star systems. These models learned to tell apart real planet transits from fake signals like eclipsing binary stars or other 'background noise' created in the void of space.
The AI helped flag promising candidates from 83 million light curves, which humans then double-checked. Ultimately, 11,554 candidates made the list, of which 10,091 were completely new worlds that had never been seen around these faint stars and were spotted transiting in front of their home suns multiple times. Another 411 potential planets were seen transiting only once in front of their home stars. Of those worlds, just 11 have been declared Earth-like, with scientists saying these 'super-Earths' may be one to two times larger than our planet, with a rocky surface.
However, the study did not mention if liquid water was discovered on any of the new planets, which would be a key factor in the search for alien life. It also did not determine if any of the planets sit in the 'habitable zone' of these solar systems, which scientists believe is a major requirement for life as we know it to exist. The so-called habitable zone is also known as the Goldilocks zone, where a planet is just the right distance from its home star, making it not too hot and not too cold for liquid water to form on the surface.
Results revealed that the vast majority of the new planets appear to be gas giants, worlds composed of dense gases such as hydrogen and helium, featuring a thick atmosphere and small rocky or metallic core. In our solar system, gas giants include Jupiter and Saturn. 'This detection demonstrates our pipeline's ability to identify real, previously undiscovered, transiting planets. Overall, this work shows that large-scale, machine learning–assisted transit searches of TESS [Full Frame Images] can significantly expand the census of transiting planet candidates, in particular around faint stars,' the team said in a statement.
