A fully automated optics system that corrects atmospheric blurring of celestial objects has imaged 715 star systems thought to harbour planets, completing each observation in less time than it takes to read this article.
At the time of writing, observations from the Kepler Space Mission have yielded more than 975 confirmed exoplanet detections from 4,234 candidates1. These candidates are identified by small, periodic drops in the brightness of the star, indicating that a planet might be transiting in front of it2. This is perhaps the conceptually simplest method of finding exoplanets, and it remains the only approach that can find planets with the proper orbit and radius to potentially support life. However, follow-up observations of high spatial resolution are needed to confirm and characterize each candidate system detected by the Kepler mission. Writing in The Astrophysical Journal, Law et al.3 describe how they have used a robotic adaptive optics system4,5 to follow up 715 of the Kepler candidate star systems in just 36 hours of observing time.
In planetary-transit observations, the size of the planet can be inferred from the relative dip in star brightness measured during the transit. Only a small fraction of exoplanets will transit their star when viewed from Earth, but a statistical analysis6 of the Kepler candidates detected in observations of more than 100,000 stars indicates that exoplanets may be relatively common. This includes Earth-sized planets with orbits that would permit liquid water to exist on the planets' surfaces6. Because Kepler was designed to continually monitor many thousands of stars, its images lack the spatial resolution needed to characterize individual star systems in further detail. This means that various false-positive detections — for example, those associated with the partial eclipse of a star in a binary system by its companion star — cannot be ruled out, and that possible binary host stars (or stars in the foreground or background by coincidence) cannot necessarily be identified.
The presence or absence of a stellar companion to a 'primary host star' is important information for understanding the formation and development of planetary systems. It also affects the estimation of the planet's size from the transit: the relationship between star brightness and planetary radius is more complex if there is a stellar companion, leading to incorrect results if the existence of the companion star is unknown. For these and other reasons, follow-up observations with high angular resolution are needed to fully understand each Kepler candidate.
High-resolution follow-up images could be collected by a space-based observatory such as the Hubble Space Telescope, but observing thousands of candidate systems would monopolize this limited resource. Obtaining such images from the ground is made difficult by the blurring ('seeing') introduced by atmospheric turbulence, and by the resulting inhomogeneity in the density and refractive index of the air. In the past two decades, ground-based observatories have begun using a technology known as adaptive optics to measure and correct this blurring in real time7. Many of these systems now use lasers to create artificial 'guide stars' on the sky to measure the blurring, and then correct it for science targets that are themselves too faint to be used for such measurement — as is the case for many of the Kepler candidates. Adaptive optics surveys of the candidate systems began in 2011–12 (refs 8,9), but these initial studies were limited to fewer than 100 targets because of the time taken to set up and initiate each observation, typically at least 15–20 minutes per target for most current adaptive optics systems.
The robotic adaptive optics system (Robo-AO) used by Law and colleagues supersedes these constraints. The system has been designed4,5 for highly efficient, automated high-resolution observing on 1- to 3-metre-class telescopes, and has been mounted on the 60-inch (1.5-metre) telescope at the Palomar Observatory in California (Fig. 1). The atmospheric blurring at Palomar Observatory is typically about 0.65 arcseconds. Robo-AO sharpens star images to about 0.12–0.15 arcseconds in diameter4,5 — not far from the 0.09-arcsecond value that is theoretically possible with a 1.5-metre telescope in space. This performance has enabled Law et al. to resolve 53 of the 715 Kepler candidates observed by Robo-AO so far into multiple stars. Forty-three of these 53 are new discoveries, including one that is a probable false positive for a candidate exoplanet.
Of course, automated observing at a rate of 200–250 targets per night, as Law and co-workers have done, creates a substantial data cleaning and analysis task. To detect and characterize companion stars that are significantly fainter than their primaries, the authors have developed a fully automated data-processing pipeline. More than 800 short (115-millisecond) exposures are collected of each target, and these must first be calibrated, centred and averaged into a single image. The light from the primary star is then subtracted from the image, and an automated target-detection algorithm is applied. Companion stars that are as faint as one one-hundredth to four one-thousandths of the primary can be detected in median to good atmospheric conditions; this is not faint enough to find exoplanets, but more than sufficient to identify many companion stars. Once the brightness of a companion star has been measured, its mass and diameter can be determined from standard stellar models and from the characteristics of the primary star.
Law et al. have provided updated estimates for the planetary radii of each of the Kepler candidates with a fainter companion star. Five small planet candidates have been confirmed to be less than twice the diameter of Earth, but a larger number of other candidates could be significantly bigger than this threshold if they are found to be orbiting the fainter companion star instead of the primary (this would require future observations of their transits with Robo-AO or some other high-resolution system). The team also suggests that several of the stars with multiple Kepler candidate planets are likely to be coincident multiples — two separate planetary systems orbiting both stars of the binary pair. In addition, the Robo-AO observations so far yield plausible (98% confidence) evidence3 that giant planets with orbital periods of less than 15 days are two to three times more likely than longer-period planets to be found in binary-star systems. This suggests that companion stars have a role in creating close-in giant planets and stabilizing their orbits. The researchers expect to observe every Kepler candidate using Robo-AO by the end of 2014 to confirm this conclusion, and aim to develop a more comprehensive statistical sample of planetary systems associated with binary stars.
More generally, these results are a convincing indication that laser-guide-star adaptive optics is now ready for highly efficient, quantitatively precise, high-resolution astronomical observations. Current and future adaptive optics systems on much larger telescopes than the Palomar 60-inch telescope — such as Keck, Gemini and the European Southern Observatory's Very Large Telescope — can produce even sharper images of even fainter objects, but much work will be needed to match the degree of automation and efficiency already demonstrated by Robo-AO.