NASA’s data hides a plethora of potential planets, waiting for scientists to take a closer look at the wispy star.
In a new study, researchers reanalyzed the first year of observations by NASA’s Transiting Exoplanet Survey Satellite (TESS). They discovered 11,554 planet candidates. Of these, 10,091 have never been flagged, 1,052 were already known as TESS candidates, and 411 appeared only once when they crossed in front of the Star.
fresh truck
TESS looks for planets by observing stars dimming.
When a planet passes in front of its star, a small portion of the light is blocked. Astronomers call this a transit. TESS uses transit methodology to keep an eye on these dips.
But stars can be finicky and flicker for a variety of reasons. Another star could obscure the companion star, or nearby objects could contaminate the signal. The probe itself can leave patterns in the data. Turning dips in starlight into planet candidates is no easy task. Converting a candidate to a confirmed planet is even more difficult. Most TESS searches have favored bright stars that are easy to track and observe.
“Rather than just looking at bright stars as before, we expanded our search for planets to include fainter stars,” said Joshua Ross, a graduate research fellow at Princeton University and lead author of the study. IFL Science.
This decision revealed a huge trove of data that had been overlooked. The research team examined more than 83 million starlight records from TESS’s first year. Among them were many stars so faint that it had not been possible to examine them in detail. But they didn’t do it by hand.
planet spotting
The researchers built a semi-automated pipeline that uses machine learning to classify data. Their main tool was a random forest classifier. This is a system that lets many decision trees vote on whether the signal looks like a planet, an eclipsing binary system, or noise.
“This just gives us a much larger stellar base from which to explore these planets,” Ross said. “We developed a semi-automated pipeline that incorporates machine learning to comb through large amounts of data to find planets, and we discovered about 10,000 new planet candidates.”
Because faint stars contain different types of noise, the team trained separate models for bright and faint stars. The software learned from known TESS candidates, known eclipsing binaries, and artificial planet signals added to the real light curve. After the automatic disconnection, the researchers manually inspected approximately 50,000 potential traffic signals. All published candidates passed human review.
The new list is a map of likely targets. Some candidates may turn out to be false. TESS pixels cover a large portion of the sky, allowing light from nearby stars to mix together. The researchers also removed a number of potential eclipsing binaries and known contaminants, but cautioned that continued follow-up research is essential.
The scale of it is truly amazing. Humankind has confirmed more than 6,000 exoplanets in about 30 years. This study added over 10,000 new possibilities from one year of TESS data.
Most of the candidates seem like a big world. The study classifies 97.7% as gas giants, including a small number of Neptune-, sub-Neptune-, and super-Earth-like planets. It also includes 66 ultra-short-term candidates, or worlds that could orbit the star in less than a day.
1 confirmed case
To test the pipeline, the researchers followed up on one target, TIC 183374187. They wanted to see if this was really a planet.
Using the Planet Finder spectrometer on the 6.5-meter Magellan Clay Telescope in Chile, they measured the star’s wobble and confirmed that the candidate was TIC 183374187 b, a hot Jupiter. The planet orbits every 5.059 days, has a mass approximately 0.56 times that of Jupiter, and a radius approximately 1.25 times that of Jupiter.
Its host star is old and metal-poor, and likely belongs to the Milky Way’s thick disk, an ancient population of stars. Such a hot Jupiter baffles astronomers because giant planets are thought to form far from stars where there is enough cold gas and ice to build them. However, these worlds end up roasting in narrow orbits that last only a few days. Finding planets around old stars like this provides astronomers with another system to test how giant planets form, persist, and migrate inward over billions of years.
2nd year
The catalog is still in its early stages, as the first year’s search focused mainly on TESS’s southern sky observations. Ross and his colleagues have already begun work on TESS’s second year of observations, with the spacecraft observing another half of the sky and later revisiting some stars.
Some planets orbit too slowly to be clearly visible within the 27-day TESS observation window. A single tilt can suggest a planet, but repeated tilts are required to measure its orbit. By linking observations taken several months apart, the team can recover candidates over longer periods of time and reject signals that appear only once due to noise or contamination.
They’re also trying to change search itself. The next version of the pipeline is designed to examine multiple observation windows across multiple viewing windows, rather than processing each short TESS fixation almost independently. This could improve the signal of dark planets and discover small worlds that were unremarkable on the first pass.
The work is now moving from discovery to triage. Astronomers will need to decide which candidates are worthy of telescope time, which candidates are likely to be frauds, and which can answer larger questions about how planets form around different types of stars.
This study Astrophysical Journal Appendix Series.
