Buried Treasure: Astronomers Find Exoplanets Hidden in Old Hubble Data


Over the past 21 years, the Hubble Space Telescope has gathered boatloads of data, with the Hubble archive center filling about 18 DVDs for every week of the telescope’s life. Now, with improved data mining techniques, an intense re-analysis of HST images from 1998 has revealed some hidden treasures: previously undetected extrasolar planets.

Scientists say this discovery helps prove a new method for planet hunting by using archived Hubble data. Also, discovering the additional exoplanets in the Hubble data helps them compare earlier orbital motion data to more recent observations.

How did astronomers detect the previously unseen exoplanets, and can the methods used be applied to other HST data sets?

This isn’t the first time hidden exoplanets have been revealed in HST data – In 2009 David Lafreniere of the University of Montreal recovered hidden exoplanet data in Hubble images of HR 8799. The HST images Lafreniere studied were taken in 1998 with the Near Infrared Camera and Multi-Object Spectrometer (NICMOS). The outermost planet orbiting HR 8799 was identified and demonstrated the power of a new data-processing technique which could tease out faint planets from the glow of their central star.

Four giant planets are now known to orbit HR 8799, the first three of which were discovered in 2007/2008 in near-infrared images taken with instruments at the W.M. Keck Observatory and the Gemini North telescope by Christian Marois of the National Research Council in Canada. In 2010 Marois and his team uncovered a fourth, innermost, planet. What makes the HR 8799 system so unique is that it is the only multi-exoplanet star system that has been directly imaged.

The new analysis by Remi Soummer of the Space Telescope Science Institute has found all three of the outer planets. Unfortunately, the fourth, innermost planet is close to HR 8799 and cannot be imaged due obscuration by the the NICMOS coronagraph that blocks the central star’s light.

When astronomers study exoplanets by directly imaging them, they study images taken several years apart – not unlike methods used to find Pluto and other dwarf planets in our solar system like Eris. Understanding the orbits in a multi-planet system is critical since massive planets can affect the orbits of their neighboring planets in the system. “From the Hubble images we can determine the shape of their orbits, which brings insight into the system stability, planet masses and eccentricities, and also the inclination of the system,” says Soummer.

Making the study difficult is the extremely long orbits of the three outer planets, which are approximately 100, 200, and 400 years, respectively. The long orbital periods require considerable time to produce enough motion for astronomers to study. In this case however, the added time span from the Hubble data helps considerably. “The archive got us 10 years of science right now,” Soummer says. “Without this data we would have had to wait another decade. It’s 10 years of science for free.”

Given its 400 year orbital period, in the past ten years, the outermost planet has barely changed position. “But if we go to the next inner planet we see a little bit of an orbit, and the third inner planet we actually see a lot of motion,” Soummer added.

When the original HST data was analyzed, the methods used to detect exoplanets such as those orbiting HR 8799 were not available. Techniques to subtract the light from a host star still left residual light that drowned out the faint exoplanets. Soummer and his team improved on the previous methods and used over four hundred images from over 10 years of NICMOS observations.

The improvements on the previous technique included increasing contrast and minimizing residual starlight. Soummer and his team also successfully removed the diffraction spikes, a phenomenon that amateur and professional telescope imaging systems suffer from. With the improved techniques, Soummer and his team were able to see two of HR 8799’s faint inner planets, which are about 1/100,000th the brightness of the host star in infra-red.

Soummer has made plans to next analyze 400 more stars in the NICMOS archive with the same technique, which demonstrates the power of the Hubble Space Telescope data archive. How many more exoplanets are uncovered is anyone’s guess.

Finding these new exoplanets proves that even after the HST is no longer functioning, Hubble’s data will live on, and scientists will rely on Hubble’s revelations for years as they continue in their quest to understand the cosmos.

Source: Hubble Space Telescope Mission Updates

12 Replies to “Buried Treasure: Astronomers Find Exoplanets Hidden in Old Hubble Data”

    1. This is why the new space telescope needs to be completed and flown. It will enable at least as much science as the Hubble telescope has.

    2. This is why the new space telescope needs to be completed and flown. It will enable at least as much science as the Hubble telescope has.

  1. Sounds awesome, hopefully the innermost one isn’t fake, direct imaging suffered a little blow lately.

    1. Are you referring to Fomalhaut b? Because one lead author has made an efficient (it seems to me) riposte in what looks like a grudge match:

      “”You have one scientist trying to create a controversy out of nothing,””.

      “For the latest image, Kalas had to resort to another Hubble instrument, the Space Telescope Imaging Spectrograph. He says the switch to a different detector may explain the planet’s slight deviation from its expected position. He has been allocated time to take another picture with that same instrument next summer.”

      “In an e-mail to Schneider today, Kalas said that, to be scrupulously fair, Schneider should also mention that there are doubts associated with 1RXJ1609, a planet imaged directly at infrared wavelengths that Jayawardhana co-discovered and announced in 2008, a few months before the unveiling of Fomalhaut b.”

      Someone taking a swing in a grudge match isn’t actually delivering a blow as much as overblown concern. We will have to wait and see if this is something real IMO.

      1. Although the situation isn’t completely clear cut, 1RXJ1609B is likely just a low-mass brown dwarf or some sub-deuterium-burning object made during the early stages of the host star’s formation, unlike bona fide planets formed after circumstellar material collapsed into a disk. The emerging consensus is that HR 8799bcde, beta Pic b, and Fomalhaut b are the bona fide planets detected by direct imaging (though perhaps the “directness” is up for debate with Fomalhaut b). Other recently discovered objects (1RXJ1609B, 2M 1207B, etc.) have a much hazier status and probably have more in common with low-mass brown dwarf companions than things like Jupiter.

    2. (from someone in the field)
      – Extremely unlikely to be fake. Their signal-to-noise for HR 8799d (SNR ~ 5) is just good enough given the statistics of pixel intensities in speckle-dominated regions where you just have noise (e.g. in such regions you’re likely to have, by chance, several pixels with SNR ~ 3). Furthermore, as you can see from their Figure 1 (http://arxiv.org/abs/1110.1382), they detect HR 8799d separately in the two different rolls (e.g. where the detector angle of orientation differs). This is important since each roll position samples different regions of the HST/NICMOS detector: were their detection just a HST/NICMOS artifact, they shouldn’t see a detection in both rolls.

      A few other lines of evidence (there are more)… the centroid position for HR 8799d doesn’t jump around alot depending on how strict/loose they are with their data reduction technique, certainly not any more in absolute terms (e.g. in fractions of a pixel) than HR 8799c or d and is very consistent between the two rolls (their Fig. 4). This is ever the more impressive since the d planet sits right next to the diffraction spikes in one of the rolls. Finally, they go through a detailed “false alarm probability” estimate by comparing the pairs of pixels separated by ~30 degrees (the difference in angle between the two rolls) and find in only three cases (out of the 96 they studied) is the signal-to-noise greater than 3 and the “position” of the speckles consistent between the two rolls. Thus, they estimate a false alarm probability of ~3%.

      – As far as direct imaging “suffering a blow as of late”, that’s not actually true All that happened is one astronomer in the field simply declaring (at a meeting) that Fomalhaut b isn’t a planet and this expression of skepticism got picked up as “news” because this person is very media-savvy. It would be one thing if he had presented a reanalysis of the Fomalhaut data or analysis of his own data disputing Fomalhaut’s status as a planet. This didn’t happen, and the only substantive arguments given are rather feeble, in my opinion.

      1. Scientist := knows how to maximize signal-to-noise ratio.

        Media-savvy scientist := also knows how to maximize noise-to-signal ratio.

  2. Sounds awesome, hopefully the innermost one isn’t fake, direct imaging suffered a little blow lately.

Comments are closed.