Seven years into its mission, NASA’s TESS satellite has already identified nearly 700 exoplanets. That seems like a lot. But the telescope monitors more than 2 million stars, and most of the signals flagged as promising are not properly classified.
Buried in that backlog are real, unconfirmed planets that exist in the data. The research team discovered and verified more than 100 of them.
Search for new exoplanets through simulation
This software is called RAVEN. This stands for RAnking and Validation of ExoplaNets. The team at the University of Warwick built it to address the stubborn problem of separating real planets from a long list of impostors fooling detection tools.
Dr. Marina Lafarga-Magro, a postdoctoral researcher there, led the study. Running RAVEN on TESS data from the first four years of the mission, her team discovered more than 100 new exoplanets and flagged thousands more likely candidates.
“Using the newly developed RAVEN pipeline, we were able to validate 118 new planets and over 2,000 high-quality planet candidates, nearly 1,000 of which were completely new,” said Lafarga, lead author of the study.
How RAVEN works
Telescopes detect small dips in starlight that indicate planet candidates. There are many other things that can mimic that signal. The way a pair of stars overlap, the way faint background stars bleed through, the noise of musical instruments, etc. Sorting out the real from the fake takes time.
RAVEN takes a different path. The developers fed the machine learning model hundreds of thousands of simulation examples (synthetic planets and synthetic analogues) to learn the differences.
Once training is complete, RAVEN handles the entire chain. Find the signals, sift through them, and use math to see which ones are the strongest. The complete pipeline design was published in a separate paper published earlier this year.
“We trained a machine learning model to identify patterns in the data that could tell us the types of events we detected,” said Dr. Andreas Hajigeorgiou, who led the pipeline development at Warwick.
A new exoplanet has also been discovered
The team aimed RAVEN at the approximately 2.2 million common stars observed by TESS during the mission’s first four-year all-sky scan, Sectors 1 to 55.
They limited their search to planets with orbits shorter than 16 days. It’s short, but wide enough to capture the most interesting up-close views of the world.
Overall, 118 planets passed validation, 31 of which had never been discovered before. In addition, RAVEN flagged over 2,000 high-confidence candidates who have not yet been formally verified.
Approximately 1,000 of them are new. A small number of signals are rare single-transit candidates, signals that crossed the star only once or twice during the observation window, suggesting much longer orbits.
strange new world
Some of the most interesting discoveries belong to two unusual categories. Very short-period planets, or planets that complete a full orbit in less than 24 hours, appeared several times during the transit.
So did the inhabitants of the Neptune Desert, a range of orbits and sizes in which Neptune-sized planets are strangely absent. A 2016 paper first delineated the boundaries, but counting the small number of residents accurately has been more difficult.
The catalog also features dense multiplanetary systems with previously unknown companion stars. In these systems, the worlds are tightly packed near a single star and can be pulled together by gravity.
Count approaching planets
The larger benefits were revealed in a collaborative study led by Dr. Kaiming Cui, also from the University of Warwick. The researchers obtained a sample of clean planets and measured how often close planets appear around Sun-like stars.
Their answer is that it hosts about 9 to 10 percent of Sun-like stars. This follows previous results from NASA’s Kepler mission, but the new measurements reduce the uncertainty by up to a factor of 10.
The same clean sample also produced the first head count for the Neptune Desert. Such planets occur in only 0.08 percent of Sun-like stars.
“For the first time, we have been able to provide accurate numbers on how uninhabited this desert is,” said Choi, lead author of the population study.
The road ahead
A complete catalog of verified planets, unverified candidates, and the tools to examine them are all now publicly available. Other researchers can select targets for follow-up studies. This means we can use telescopes to determine the planet’s mass and probe its atmosphere with instruments already in use.
Neptunian desert candidates and ultrashort-period worlds will be particularly attractive to researchers studying the atmosphere because they lie on the strange end of the predictions of current planet-formation theories.
ESA’s upcoming PLATO mission will require exactly this kind of curated target list at launch. When new equipment comes online, clean samples mean science can be done more quickly.
The number of new exoplanets becomes clearer
Until this study, no one had directly measured how rare Neptune’s desert planets actually were. Now the numbers are out. Something solid.
And Kepler-era planet approach estimates, long the gold standard, have been independently confirmed and tightened by other missions, telescopes and pipelines.
For exoplanet researchers, this opens up questions about planet formation that were previously shrouded in statistical fog. Why are some orbits empty? How do tightly packed systems maintain stability?
With a cleaner planetary count to make the math stick, the next round of theory will require more than guesswork. And the same approach can wipe out years of TESS data still waiting in the archives.
This research Royal Astronomical Society Monthly Notices.
—–
Like what you read? Subscribe to our newsletter for fascinating articles, exclusive content and the latest updates.
Check out EarthSnap, a free app from Eric Ralls and Earth.com.
—–
