Research ArticlesAstronomy

Direct imaging discovery of a Jovian exoplanet within a triple-star system

See allHide authors and affiliations

Science  12 Aug 2016:
Vol. 353, Issue 6300, pp. 673-678
DOI: 10.1126/science.aaf9671

Spying a planet in a triple-star system

Thousands of extrasolar planets are now known, but only a handful have been detected in direct images. Wagner et al. used sophisticated adaptive optics to discover a planet in images of the triple-star system HD 131399 and to take a spectrum of its atmosphere (see the Perspective by Oppenheimer). The planet, about four times the mass of Jupiter, orbits around one star in the system while the other two stars move farther out. This unusual arrangement is puzzling: The planet's orbit may be stable, but it is unclear how it could have formed or migrated there. The results will be used to refine theories of planet formation.

Science, this issue p. 673; see also p. 644


Direct imaging allows for the detection and characterization of exoplanets via their thermal emission. We report the discovery via imaging of a young Jovian planet in a triple-star system and characterize its atmospheric properties through near-infrared spectroscopy. The semimajor axis of the planet is closer relative to that of its hierarchical triple-star system than for any known exoplanet within a stellar binary or triple, making HD 131399 dynamically unlike any other known system. The location of HD 131399Ab on a wide orbit in a triple system demonstrates that massive planets may be found on long and possibly unstable orbits in multistar systems. HD 131399Ab is one of the lowest mass (4 ± 1 Jupiter masses) and coldest (850 ± 50 kelvin) exoplanets to have been directly imaged.

Thousands of planets around other stars have been discovered (1, 2), revealing a greater diversity than predicted by traditional planet formation models based on the solar system. Extreme examples are planets within binary and multiple-star systems, which form and evolve in variable radiation and gravitational fields. Direct imaging allows for the detection and spectroscopic characterization of long-period giant planets, thus enabling constraints to be placed on planet formation models via predictions of planet population statistics and atmospheric properties (3). However, most direct imaging surveys have traditionally excluded visual binary or multiple systems whose separations are less than a few hundred astronomical units (AUs). These exclusions are based on the assumption that such planetary systems would either be disrupted or never form, as well as the increased technical complexity of detecting a planet among the scattered light of multiple stars. As a result of this observational bias, most directly imaged exoplanets have been found around single stars.

Because multistar systems are as numerous as single stars (4), building a complete census of long-period giant planets requires investigation of both configurations. In principal, planets on wide orbits (detectable by direct imaging) might arise more frequently in multistar systems because of planet-planet or planet-star interactions (5, 6). Such interactions could even produce planets on chaotic orbits that wander between the stars (7, 8). To investigate the frequency of long-period giant planets both around single stars and in multistar systems, we are using the Very Large Telescope (VLT) and the Spectro-Polarimetric High-Contrast Exoplanet Research instrument [SPHERE (9)] to sample a population of ~100 young single and multiple A-type stars in the nearby Upper Scorpius-Centaurus-Lupus association. Here we report the discovery of the first planet detected in our ongoing survey and the widest-orbit planet within a multistar system.

Observations and discovery of HD 131399Ab

HD 131399 (also known as HIP72940) is a triple system (10) in the 16 ± 1–million–year–old Upper Centaurus-Lupus association [UCL (1113)] at a distance of 98 ± 7 pc (14) whose basic properties are given in Table 1. The system’s membership in UCL is confirmed by its parallax and kinematics (1113), and the well-constrained age of the association provides greater confidence in the young age of the system than for most directly imaged exoplanet host stars (see supplementary text for the detailed age analysis). Despite its youth, the system shows no evidence of infrared excess, and thus its primordial disk has probably been depleted to beneath detectable levels (15).

Table 1 Basic parameters of the stars and directly imaged planet in HD 131399.

The mass, effective temperature, and spectral type of the previously unresolved B and C stars (except where noted) were estimated from their K1 luminosity (1719, 35). The planet’s temperature and spectral type were determined through spectral fitting (see next section on characterization). Apparent J, H, and K magnitudes for HD 131399A were obtained from (36). Embedded Image, solar mass; N/A, not applicable.

View this table:

We observed HD 131399 on 12 June 2015, obtaining a wide range of near-infrared spectral coverage ranging from the Y band to the K band (0.95 to 2.25 μm) and diffraction-limited imaging with an 8.2-m telescope aperture. Our observations (10) resulted in the discovery of HD 131399Ab, a point source with a 10−5 contrast to HD 131399A and a projected separation of 0.84 arc sec, or 82 ± 6 AU (Fig. 1 and Table 1). After the initial discovery, we obtained follow-up observations (10) to verify whether the faint source is physically associated with the parent star (i.e., shares common proper motion) and to improve the quality of the near-infrared spectrum, enabling characterization of the planet’s atmospheric properties.

Fig. 1 Near-infrared VLT-SPHERE images of HD 131399Ab and the hierarchical triple-star system HD 131399ABC.

(A to D) The central regions that are affected by the coronagraph and residual scattered starlight are blocked by a mask (dashed circles), with the location of star A indicated by the crosshairs. (E) Composite of the point spread function (PSF)–subtracted region (dashed region) superposed on the wide-field K1 image showing the stellar components of the system, whose luminosities are adjusted to the level of the planet for clarity. In each image, the luminosity of component A (but not components B and C) has been suppressed by the use of a coronagraph. The images in panels (A) to (C) were processed with angular and spectral differential imaging to subtract the stellar PSF, whereas panel (D) and the PSF-subtracted region of (E) were processed only with angular differential imaging (10). Images in (A) to (D) share the same field orientation.

We detected HD 131399Ab with a signal-to-noise ratio in the Y (1.04 μm), J (1.25 μm), H (1.62 μm), K1 (2.11 μm), and K2 (2.25 μm) bands of 9.3, 13.2, 15.5, 23.5, and 11.9, respectively. Following astrometric calibrations (10), we measured a positional displacement to HD 131399A of Δα (right ascension) = 12 ± 8 milli–arc sec (mas) and Δδ (declination) = 6 ± 8 mas over the 11-month baseline, where the uncertainties are dominated by the calibration of the instrument orientation across the two epochs. This allows us to reject the hypothesis of a background object, which would have moved relative to HD 131399A by Δα = 27.3 ± 0.6 mas and Δδ = 28.8 ± 0.6 mas due to the relatively high proper motion of the system (14). Assuming a Keplerian orbit for the planet with a semimajor axis equivalent to its projected separation of 82 AU yields a period of ~550 years, which, for a face-on circular orbit over 11 months, is expected to produce ~9 mas of relative motion, consistent with our observations.

The bound planet hypothesis is also supported by the low probability of detecting an unbound object within UCL that happens to share a similar spectral type to HD 131399Ab (as discussed in the next section). Following the arguments in (16), the false-alarm rate of an unassociated objected with a planetlike spectrum per field of view is ~2 × 10−7. The total false-alarm probability of one such object appearing in our 33 fields of view (so far explored in our survey) is given by the binomial distribution, resulting in a probability of ~6.6 × 10−6. Although the probability of detecting a bound giant planet is not yet well established, results from the first several hundred stars surveyed suggest that this value is around a few percent—orders of magnitude higher than the probability of detecting an unbound object with a planetlike spectrum.

Characterization of HD 131399Ab

We convert the planet’s J-, H-, and K1-band aperture photometry to a mass estimate via comparison to widely used evolutionary tracks for hot-start initial conditions (1618), in which the planet retains its initial entropy of formation. Systematic interpolation between hot-start evolutionary tracks yields a mass of 4 ± 1 Jupiter masses (MJup), which places HD 131399Ab firmly in the planetary mass regime. Even in the unlikely event that the system is much older (by a few hundred million years), companion Ab would necessarily be of planetary mass (<13 MJup). The H versus H-K color of the planet is inconsistent with the cold-start scenario, in which the planet has lost some fraction of its initial entropy due to inefficient accretion (19, 20), but consistent with hot-start models including a partly cloudy atmosphere and/or super-solar metallicity (fig. S1).

Using the integral field spectrograph (21) on SPHERE, we obtained a 0.95- to 1.65-μm spectrum. This spectrum allows the characterization of water and methane absorption bands within 1.4 to 1.6 μm, although the signal-to-noise ratio in the individual spectral channels at shorter wavelengths is too poor to be useful in spectral analysis. In the K-band, where the contrast with the star is more favorable, the dual-band images also probe the 2.2-μm methane absorption. Like the exoplanet 51 Eridani b (16) and other field (nonexoplanet) T-type brown dwarfs, the near-infrared spectrum of HD 131399Ab (Fig. 2) displays prominent methane and water absorption bands. The data are in good agreement with models of exoplanetary atmospheres (18), allowing us to estimate the atmospheric properties of effective temperature (Teff) and surface gravity (g). Systematic exploration of interpolated atmospheric models indicates Teff = 850 ± 50 K and Embedded Image (centimeters per square second), where the uncertainty in surface gravity is mostly dominated by systematic uncertainties within the models (namely in the cloud properties) and not by the model-data fit. Comparison to standard classifications of field brown dwarfs (Fig. 2B) indicates a spectral type of T2 to T4.

Fig. 2 Near-infrared spectrum of HD 131399Ab.

(A) HD 131399Ab spectrum (black) alongside the best-fit model atmosphere in red (18), with Teff = 850 K and log(g) = 3.8 cm/s2, showing water and methane absorption in the atmosphere with the approximate absorption regions indicated by the gray dashed lines. The spectrum of the T-type exoplanet 51 Eri b (16) is shown in blue, scaled by 50% to roughly match the luminosity of HD 131399Ab. Fλ, specific flux; Z/H, metallicity. (B) Near-infrared spectrum of HD 131399Ab and spectra of standard field brown dwarfs (39, 40), with each 1.4- to 2.4-μm spectrum normalized independently in λFλ units (equivalent to power per unit area). The objects’ labels correspond to the object designations from the Two Micron All Sky Survey (J2000 hours and minutes of right ascension) and the spectral type. Vertical error bars indicate 2σ photometric uncertainties horizontal bars denote photometric bandpass.

The transition between L and T spectral types (Teff ~ 2100 to 1300 K and Teff ~ 1300 to 600 K, respectively) is marked by the appearance of H- and K-band methane absorption in the atmospheres of the cooler T dwarfs. In J versus J-H color-magnitude space (Fig. 3), this appears as bluer color (more negative J-H color) compared with the hotter L dwarfs. At the threshold of the L-T transition, the photosphere becomes brighter in the J band as silicate clouds transition from above to below the photosphere (22). The fact that cloudy directly imaged exoplanets (such as HR8799bcde, β Pic b, or 2M1207b) appear at the bottom of the L-dwarf sequence points to cloud layers in these low-gravity objects that are thicker than in their higher-gravity brown dwarf counterparts (23, 24). In contrast, HD 131399Ab and the two other directly imaged T-type exoplanets follow the T-dwarf sequence, which we interpret as evidence for a similarity between the mostly or fully cloud-free atmospheres of these exoplanets and cool field brown dwarfs. HD 131399Ab is the closest directly imaged exoplanet to the L-T transition, which is consistent with the partly cloudy atmosphere suggested by the H versus H-K hot-start model predictions (fig. S1).

Fig. 3 J-H color-magnitude diagram of brown dwarfs and directly imaged giant exoplanets.

HD 131399Ab falls among the methane-dominated T dwarfs near the L-T transition. The L- and T-dwarf data (with parallax-calibrated absolute magnitudes) were obtained from (41), whereas the directly imaged exoplanet data are from (16, 4247). Vertical and horizontal error bars indicate 2σ photometric uncertainties.

Orbital characterization of HD 131399

HD 131399Ab is the widest known exoplanet that orbits within a triple system (Figs. 4 and 5). Because the presence of a second and third star can greatly limit the phase space where planetary orbits are stable, observing a system in this configuration is thought to be unlikely (7, 25). In our ongoing survey, we have imaged 18 single A-type stars and 15 binary or triple-star systems with separations similar to HD 131399A-BC. Although the sample size is small, it is surprising to us that the first planet detected in our survey is in a triple system.

Fig. 4 Schematic illustration of the components of the HD 131399 hierarchical triple-star system and comparison to the solar system.

(A) The dashed ellipse (top left) shows a preliminary orbit for the planet; the dashed curve at the bottom of this panel indicates our best-fit orbit of the BC pair. The orbit shown for the planet has orbital elements that are consistent with the data, although the astrometric uncertainties permit a substantial range of orbits (see the supplementary text for parameter ranges). (B) The image in (A) is reproduced with the orbits of the solar system planets overlaid. The underlying image is a composite image of the actual PSFs superposed on a dark-sky background. The image is composed of SPHERE J-, H-, and K-band PSFs for components A and Ab (colored as blue, green, and red, respectively) and the monochromatic K-band PSF of components B and C. For clarity, the luminosity of the planet is enhanced by a factor of 105, and because only K-band photometry exists for components B and C, their colors have been adjusted to be representative of typical G and K stars.

Fig. 5 Ratio of semimajor axes of planets that orbit one star of a multiple system (satellite, or S-type, planets) to the semimajor axes of their host systems.

The gray solid line at one-third times the binary separation represents the approximate critical radius of tidal truncation and orbital stability in the coplanar case (25). Although the critical radius varies somewhat for different parameters of stellar mass ratio, eccentricity, and inclination, HD 131399Ab is much closer to the critical radius than any other known exoplanet. For systems in which either the planet or stars lack precise orbital solutions, their projected separations are plotted instead (denoted by triangular plot points instead of circles). This includes HD 131399Ab, although from the results of the preliminary orbit fit, the semimajor axes of this system are indeed similar to the projected separations. See table S4 for the list of included objects and their associated references. MJ, Jupiter mass.

We used astrometric observations dating back to 1897 (table S2) (26) to fit the orbit to a grid of models for a binary system using the center of mass of the BC system. Our neglect of the BC orbit is motivated by the system’s hierarchical nature and the fact that most previous data could not resolve the pair. Our best-fit model (fig. S2 and table S3) consists of a semimajor axis of Embedded Image AU, eccentricity of Embedded Image, and inclination of Embedded Image with respect to the plane of the sky, where the Embedded Image subscript denotes the values for the BC orbit around HD 131399A. Using only the newer, more reliable data permits a wider range of Embedded Image AU, Embedded Image, and Embedded Image. The available astrometry for the planet does not permit a robust orbital solution, though we performed a preliminary orbit fit to obtain the plausible parameter ranges of Embedded Image AU, Embedded Image, and Embedded Image, with no single solution being strongly preferred.

The orbital configuration of HD 131399 results in a more dynamically extreme configuration than for any known exoplanet within a binary or multiple system (Fig. 5 and table S4), with the ratio of semimajor axes Embedded Image. Values of q < 0.23 require higher planetary eccentricities (ep > 0.3) to maintain the ≥82 AU observational constraint on the planet’s projected separation. The most dynamically similar planets to HD 131399Ab are γ Cephei Ab (27), discovered via radial velocity measurements; HD 41004Ab (28); and HD 142Ac (29), for which q ~ 0.1. Perhaps the most similar well-studied example is the transiting system Kepler-444, which hosts five sub–Earth-sized planets within 0.1 AU from the primary Kepler-444A (30). The latter stellar system is likewise a hierarchical triple, with a tight M-dwarf binary at 66 AU from the planet-hosting primary star. Though similar to these other systems, HD 131399 stands out due to the proximity of the planet’s orbit to that of the other stars in the system.

We use a small suite (~300) of N-body simulations (10) to demonstrate that stable orbital configurations that are consistent with the astrometric constraints exist for all four bodies. This holds even for some of the more extreme configurations (i.e., smaller A-BC semimajor axis and higher eccentricity). The current astrometry also permits unstable orbits for the planet. Given the young age of the system, the planet might be on an unstable orbit, perhaps due to planet-planet or planet-star scattering, and could yet be ejected to become a free-floating planetary-mass object. This is not the most likely scenario, as the time scale for the planet to suffer an ejection or collision is only a few million years (25). In all cases, the orbit of HD 131399Ab is non-Keplerian, as the planet’s orbital parameters (a, e, and i) undergo complex evolution due to the influence of the BC pair (fig. S3).

Formation of HD 131399Ab and the origin of its long-period orbit

Given its location in a triple system, a broad set of formation pathways is possible for HD 131399Ab. Because planet formation is inhibited in the outer disk regions due to the strong perturbations from the binary (31, 32), it is unlikely that HD 131399Ab formed in isolation on its present long-period orbit around HD 131399A and is now on a stable orbit around HD 131399A. We speculate that the planet may have arrived at its present orbit through one of three possible scenarios. Scenario (i): The planet formed on a short orbit around star A and subsequently underwent a planet-planet scattering event that ejected it to its current long-period orbit (33). This scenario requires the presence of a massive planet on a shorter-period orbit. Such a planet could have evaded detection if it were beneath our sensitivity limits (see supplementary online text for details). As a consequence we would also expect the Ab orbit to be rather eccentric. Scenario (ii): HD 131399Ab formed as a circumbinary planet around components B and C and underwent a scattering event via interactions with another planet or with the binary itself (6). This scenario would also be most consistent with an eccentric Ab orbit. Scenario (iii): The planet formed around either component before the A-BC system arrived in its present configuration. The stellar orbits could have evolved subsequently due to interactions with the natal disks or secular effects (34). This scenario does not require the presence of a second close-in massive planet, though the resulting outer planetary orbit may be indistinguishable. Thus, it is possible that the planet is no longer orbiting the star around which it formed. These scenarios are also consistent with HD 131399Ab obtaining an orbit around all three components, although the short lifetime of such an orbit makes this configuration unlikely.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S4

Tables S1 to S4

References (4881)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
Acknowledgments: This work is based on observations performed with VLT/SPHERE under program IDs 095.C-0389A [principal investigator (PI): D.A.] and 296.C-5036A (PI: K.W.). K.W. is supported by the NSF Graduate Research Fellowship Program under grant 2015209499. The results reported herein benefited from collaborations and/or information exchange within NASA’s Nexus for Exoplanet System Science (NExSS) research coordination network sponsored by NASA’s Science Mission Directorate. This research has benefitted from the SpeX Prism Spectral Libraries and the Washington Double Star Catalog maintained by the U.S. Naval Observatory at All atmospheric models used in this study can be found online at, and all of the raw data products for HD 131399Ab and associated calibrations may be obtained from the ESO archive at SPHERE is an instrument designed and built by a consortium consisting of IPAG (Grenoble, France), Max-Planck-Institut für Astronomie (MPIA) (Heidelberg, Germany), Laboratoire d’Astrophysique de Marseille (Marseille, France), Laboratoire d’Etudes Spatiales et d’Instrumentation en Astrophysique (Paris, France), Laboratoire Lagrange (Nice, France), Istituto Nazionale di Astrofisica (INAF) Osservatorio di Padova (Italy), Observatoire de Genève (Switzerland), ETH Zurich (Switzerland), Nederlandse Onderzoekschool voor de Astronomie (NOVA) (Netherlands), Office National d’Etudes et de Recherches Aérospatiales (France), and the Netherlands Institute for Radio Astronomy (ASTRON) (Netherlands) in collaboration with ESO. SPHERE was funded by ESO, with additional contributions from CNRS (France), MPIA (Germany), INAF (Italy), FINES (Switzerland), and NOVA (Netherlands). SPHERE also received funding from the European Commission Sixth and Seventh Framework Programmes as part of the Optical Infrared Coordination Network for Astronomy (OPTICON) under grants RII3-Ct-2004-001566 for FP6 (2004–2008), 226604 for FP7 (2009–2012), and 312430 for FP7 (2013–2016).

Stay Connected to Science

Navigate This Article