Two new planets discovered thanks to artificial intelligence

The planets discovered this time belonged to Kepler's extended mission, called K2.

To find them, the team, led by a University of Austin student, Anne Dattilo, created an algorithm that examines the data collected by Kepler to detect signals missed by traditional planetary search methods. In the long run, the process should help astronomers find many missing planets hidden in Kepler's data. The discoveries have been accepted for publication in an upcoming issue of The Astronomical Journal.

Other members of the team include NASA Sagan member UT Austin Andrew Vanderburg and Google engineer Christopher Shallue. In 2017, Vanderburg and Shallue used for the first time artificial intelligence to discover a planet around a Kepler star, a planet known to house seven planets. This discovery made this solar system the only one to have as many planets as ours.

Dattilo explained that this project required a new algorithm, the data taken during Kepler's extended K2 mission being very different from those collected during the initial mission of the spacecraft.

"K2 data is harder to exploit because the spacecraft is moving all the time," said Vanderburg. This change occurred after a mechanical failure. While the mission planners were finding a workaround, the spacecraft had remained with a flicker that the AI ​​had to take into account.

The Kepler and K2 missions have already discovered thousands of planets around other stars, and an equal number of candidates are waiting to be confirmed. So, why do astronomers need to use artificial intelligence to search more in Kepler archives?

"The artificial intelligence will help us to search the dataset uniformly," Vanderburg said. "Even if all the stars were surrounded by a planet the size of the Earth, we will not find them all when we look at Kepler – it's just because some data is too noisy, or the planets are sometimes not aligned properly, we need to correct the ones we missed, we know there are many planets we do not see for these reasons.

"If we want to know how many planets there are in total, we need to know how many planets we have found, but we also need to know how many planets we have missed." That's where it comes in, "he said. he explained.

The two planets that Dattilo's team found "are very typical of the planets found in K2," she said. "They are very close to their host, they have short orbital periods and they are hot, they are slightly larger than the Earth."

One of the two planets calls K2-293b and orbits a star at 1,300 light-years from the constellation Aquarius. The other, K2-294b, gravitates around a star 1,230 light-years away, also in Aquarius.

Once the team used their algorithm to find these planets, she then studied the host stars using ground-based telescopes to confirm that the planets are real. These observations were made with the 1.5 meter telescope of the Whipple Observatory of the Smithsonian Institution in Arizona and with the Gillett Telescope of the Gemini Observatory in Hawaii.

The future of the concept of AI to find planets hidden in data sets looks promising. The current algorithm can be used to probe the entire K2 dataset, said Dattilo – about 300,000 stars. She also believes that this method is applicable to the research mission of Planet TESS, successor to Kepler, launched in April 2018. Kepler's mission ended later that year.

Dattilo plans to continue using the AI ​​for planet hunting when it will enter high school in the fall.

Explore further:
The first Kepler Space Telescope exoplanet confirmed, ten years after its launch