More than 1,200 new planets confirmed using new technique for verifying Kepler data

The researchers' discovery hinges on a technique developed at Princeton that allows scientists to efficiently analyze thousands of signals Kepler has identified to determine which are most likely to be caused by planets and which are caused by non-planetary objects such as stars. This automated technique -- implemented in a publicly available custom software package called Vespa -- computes the chances that the signal is in fact caused by a planet.

The researchers used Vespa to compute the reliability values for over 7,000 signals identified in the latest Kepler catalog, and verified the 1,284 planets with 99 percent certainty. They also independently verified more than 700 additional planet signals that had already been confirmed as planets by other methods. In addition, the researchers identified 428 candidates as likely "false positives," or signals generated by something other than a planet.

Timothy Morton, lead author of the study and a Princeton associate research scholar of astrophysical sciences, developed Vespa because the vast amount of data Kepler has gathered since its 2009 launch has made the traditional method of confirming planets by direct ground-based follow-up observation untenable, he said. Follow-up observations of Kepler data had confirmed a little more than a thousand planets prior to the Princeton-NASA announcement.

"Vespa is a culmination of a change in attitude about how we deal with these large-data surveys," Morton said. "This new problem Kepler created is that we now have thousands of new planet candidates. Astronomers knew we couldn't follow up all of these in the traditional way, but there was nothing to replace it. This result now puts a number on exactly how likely it is that each detected object is a planet."

Kepler, which ended data collection for its primary mission in 2013, operated by precisely measuring the brightness of many stars simultaneously. The satellite looked for stars that exhibited subtle and regular dimming, which indicates that an orbiting planet is passing in front of, or transiting, that star.

However, some scenarios can mimic the signature of a transiting planet, such as two stars that orbit each other, and provide a false positive signal. Distinguishing between true planets and false positives is one of the central challenges for any transiting planet survey, Morton said.

Joshua Winn, an associate professor of physics at the Massachusetts Institute of Technology and former member of the Kepler team, said that efficient methods to confirm planets will become more crucial as NASA plans and launches more space telescopes, such as the Transiting Exoplanet Survey Satellite (TESS), which is expected to find tens of thousands of exoplanets.

The one-by-one confirmation process of Kepler data has been an obstacle to scientists making as many discoveries from the data as they might have, said Winn, who is the deputy science director of TESS.

"The data rate from [TESS] will be even higher than that of Kepler, making it even more important to have trustworthy algorithms to calculate the odds that a given signal really arises from a planet as opposed to an 'imposter,'" said Winn, who is familiar with the Vespa research but had no role in it.

"Morton and his colleagues have provided an automated and effective means of assessing the probability that a given signal arises from a planet," he said. "It's exactly what the exoplanet community needs to tell the real gold apart from the counterfeits."

The Vespa technique works by comparing the details of a transiting planet signal -- specifically its duration, depth and shape -- against simulated planetary and false positive signals to indicate the type of signal the candidate most likely is. At the same time, Vespa factors in the projected distribution and frequency of star types in the galaxy from which the signal originated to determine the chances that a planet with the characteristics being analyzed would exist.

Vespa was designed to supplement the internal vetting of Kepler data, Morton said. A signal from the raw Kepler data is designated a planet candidate after a series of tests within the data-processing pipeline that are intended to weed out as many imposters as possible. These procedures rule out a large portion of false positives, Morton said. In fact, more than 3,000 of the signals from the latest catalog were already identified as false positives well before Morton applied the recent Vespa analysis. Vespa is more likely to validate an internal ruling that a signal came from a small planet because of their frequency and unambiguous signal, Morton said.

"If you have something that passes all those tests, then it's likely to be a planet," Morton said. "We know small planets are common, so if Kepler sees a small-looking planet candidate and it passes the strict internal vetting, it's more likely to be a planet than a false positive because it's hard to mimic that signal with anything else."

On the other hand, if a planet candidate has the characteristics of a Jupiter-sized planet, Vespa is less likely to verify it as a planet. The signal could very well emanate from something else because of the relative rarity of Jupiter-sized planets, Morton said. A number of planet candidates are three to four times larger than Jupiter, which means that Kepler most likely detected double-star systems in which one star was passing in front of the other.

"It's easier to mimic something the size of Jupiter, and we know Jupiter-sized planets are less common," Morton said. "So the likelihood of a Jupiter-sized candidate actually being a planet that large is typically relatively low."