Discovery and spectroscopy of the young jovian planet 51 Eri b with the Gemini Planet Imager

See allHide authors and affiliations

Science  02 Oct 2015:
Vol. 350, Issue 6256, pp. 64-67
DOI: 10.1126/science.aac5891

An exoplanet extracted from the bright

Direct imaging of Jupiter-like exoplanets around young stars provides a glimpse into how our solar system formed. The brightness of young stars requires the use of next-generation devices such as the Gemini Planet Imager (GPI). Using the GPI, Macintosh et al. discovered a Jupiter-like planet orbiting a young star, 51 Eridani (see the Perspective by Mawet). The planet, 51 Eri b, has a methane signature and is probably the smallest exoplanet that has been directly imaged. These findings open the door to understanding solar system origins and herald the dawn of a new era in next-generation planetary imaging.

Science, this issue p. 64; see also p. 39


Directly detecting thermal emission from young extrasolar planets allows measurement of their atmospheric compositions and luminosities, which are influenced by their formation mechanisms. Using the Gemini Planet Imager, we discovered a planet orbiting the ~20-million-year-old star 51 Eridani at a projected separation of 13 astronomical units. Near-infrared observations show a spectrum with strong methane and water-vapor absorption. Modeling of the spectra and photometry yields a luminosity (normalized by the luminosity of the Sun) of 1.6 to 4.0 × 10−6 and an effective temperature of 600 to 750 kelvin. For this age and luminosity, “hot-start” formation models indicate a mass twice that of Jupiter. This planet also has a sufficiently low luminosity to be consistent with the “cold-start” core-accretion process that may have formed Jupiter.

Several young, self-luminous extrasolar planets have been directly imaged at infrared (IR) wavelengths (18). The planets directly imaged to date are massive [(estimated at 5 to 13 Jupiter masses (MJ)] and positioned at large separations [9 to 650 astronomical units (AU)] from their host star, compared with planets in our solar system. Photometry and spectroscopy can be used to probe the atmospheres of these young jovian planets, providing clues about their formation. Several unexpected results have emerged. The near-IR colors of these planets are mostly red, indicating cloudy atmospheres similar to those of brown dwarfs of spectral type L. Methane absorption features are prominent in the near-IR spectra of T dwarfs [effective temperature (Teff) < 1100 K], as well as in the giant planets of our solar system, but such features are weak or absent in the directly imaged exoplanets (4, 911). Most young planets appear to be methane-free, even at temperatures where equivalent brown dwarfs show evidence of methane, suggesting nonequilibrium chemistry and persistent clouds that are probably age- and mass-dependent (1, 1215).

In spite of uncertainties about their atmospheric properties, the luminosities of these planets are well constrained. Luminosity is a function of age, mass, and initial conditions (16, 17) and hence can provide insights into a planet’s formation. Rapid formation (e.g., through global disk instabilities acting on a dynamical time scale) yields high-entropy planets that are bright at young ages (“hot start”). Alternatively, two-stage formation—in which the development of a dense solid core is followed by gas accretion through a shock, as is likely in the case of Jupiter—can produce a range of states, including lower-entropy planets that are cooler and slightly smaller in radius (“cold start”). The young directly imaged planets are almost all too bright for the cold-start model to apply, except for specific accretion shock properties; however, their formation is also difficult to explain by global instability, which should operate preferentially at higher masses and at large semimajor axis separations (18, 19). In addition, these planets are close to the limit of sensitivity for first-generation large-telescope adaptive optics (AO) systems. The goal of the latest generation of surveys, which use dedicated high-contrast AO coronagraphs (2023) such as the Gemini Planet Imager (GPI) and its counterparts, is to expand the sample of directly imaged planets to include closer separations, lower masses, and lower temperatures, a crucial empirical step toward investigating the above modes of formation.

The Gemini Planet Imager Exoplanet Survey (GPIES) is targeting 600 young nearby stars with the GPI instrument. The star 51 Eridani (51 Eri) was chosen as an early target for the survey because of its youth and proximity. Its stellar properties are given in Table 1. The star exhibits weak mid- and far-IR excess emission, indicating low-mass inner (5.5 AU) and outer (82 AU) dust belts (24, 25). It also has two distant (~2000 AU) stellar companions, which constitute the 6-AU–separation M-dwarf binary system GJ 3305 (26). 51 Eri and GJ 3305 were classified in 2001 as members of the β Pictoris moving group (27), and subsequent measurements support this identification (28). The estimated age of the β Pictoris moving group ranges from 12 to 23 million years (My) (27, 2932). Giving strong weight to the group’s lithium-depletion boundary age, we adopted an age of 20 ± 6 My for all four components of the 51 Eri system (28).

Table 1 Properties of 51 Eridani and 51 Eridani b.
View this table:

We observed 51 Eri in the H band (1.65 μm) in December 2014, as the 44th target in the GPIES campaign. GPI observations produce spectroscopic cubes with a spectral-resolving power of 45 over the entire field of view. A companion planet designated 51 Eri b, was apparent after subtraction of the point spread function (PSF). The planet is located at a projected separation of 13 AU, and its spectra exhibit distinctive strong methane and water-vapor absorption (Figs. 1 and 2). We observed 51 Eri again in January 2015 to broaden the wavelength coverage, using GPI (J band, 1.25 μm) and the W. M. Keck Observatory’s Near Infrared Camera 2 (NIRC2; Lp band, 3.8 μm). The observed spectra are highly similar to those of a field brown dwarf of spectral type T4.5 to T6 (Fig. 2). The J-band spectrum confirmed methane absorption at this wavelength, and the extremely red H-Lp color is similar to that of other cool, low-mass objects (Fig. 3). The signal-to-noise ratio at J-band wavelengths is inferior to that at H-band wavelengths, and extraction introduces additional systematic effects. The J-band detection is reliable (>6σ), but the fluxes in individual spectral channels are less certain. However, the methane feature was robustly detected at both bands (28).

Fig. 1 Images of 51 Eri and 51 Eri b (indicated by the arrow) after PSF subtraction.

(A) H-band GPI image from December 2014. (B) J-band GPI image from January 2015. (C) Lp-band NIRC2 image from January 2015.

Fig. 2 J- and H-band spectra for 51 Eri b from GPI data, after PSF subtraction.

Strong methane absorption, similar to that of Jupiter, is apparent. (Top) Spectra for the hotter young planetary object 2M 1207 b (purple) and a high–mass-field T6 brown dwarf from the SpeX library (orange) (43) are overplotted. (Bottom) Observed J and H spectra and Lp photometry with two model fits overlaid: a young, low-mass, partly cloudy object (TB-700K, green) and a higher-mass cloud-free object (SM-750K, pink). The main source of error in the extracted spectra is residual speckle artifacts, so errors in neighboring spectral channels are strongly correlated; error estimation is discussed in (28). λFλ, flux.

Fig. 3 Color-magnitude diagram of brown dwarfs (gray and black) and planetary-mass objects (colors).

51 Eri b is indicated with a red star, distinct from most other planets in the methane-dominated T-dwarf region of the diagram. The Lp photometry for field brown dwarfs is taken from (45, 46) or converted from the Wide-field Infrared Survey Explorer W1 band (47) using an Lp-versus-W1 linear fit. Parallaxes are available for all objects plotted (46). MLp, Lp-band absolute magnitude.

Demonstrating common proper motion (33), or showing that the probability of a foreground or background contaminant is extremely low, establishes the nature of directly imaged planets. The interval between the December 2014 and January 2015 observations is too brief, given our astrometric accuracy (28), to show that 51 Eri b and 51 Eri share proper motion and parallax. However, nondetection of 51 Eri b in archival data from 2003 (28) excludes a stationary background source and requires proper motion within ~0.1 arc sec/year of 51 Eri. The strong methane absorption that is evident for 51 Eri b is found only in T-type or later brown dwarfs. We determined the probability of finding a T dwarf in our field by merging the observed T-dwarf luminosity functions (27, 28) and adopting the spectral types and absolute magnitudes for T dwarfs (34), from which we calculated a false alarm rate of 1.72 × 10−7 methane objects (i.e., types T0 to T8.5) per GPI field (>5σ). The proper motion constraint eliminates a further 66% of likely background T-dwarf proper motions. The total false alarm probability after observing 44 targets is the probability of a T-spectrum object appearing in 44 Bernoulli trials, given by the binomial distribution, which yields a final probability of 2.4 × 10−6. Although the occurrence rate of planetary companions is not known with precision, the detection of planetary objects at similar physical separations to 51 Eri b, such as β Pic b and HR8799 e, indicates that the rate is >10−3 per star. Hence, with the high-quality spectrum available to us, it is much more likely that 51 Eri b is a bound planetary companion than a chance alignment.

We used planetary atmosphere and evolution models to estimate the properties of 51 Eri b. We first fitted the observed J- and H-band spectra using standard cloud-free equilibrium-chemistry models, with radii constrained based on mass as given by evolutionary tracks, similar to those in (35). This constrained fit gives an effective temperature of 750 K, with a radius [0.76 Jupiter radius (RJ)] and surface gravity similar to those of an old (10 billion years), high-mass brown dwarf. A similar, though less extreme, result (small radii and hence high masses and old ages) is associated with several model fits to the HR8799 observations of the HR8799 planets (13, 15, 16), even though high masses are excluded by dynamical stability considerations (36). This model was not constrained to fit the Lp-band observation but does so within 1.6σ.

We next fitted a model to the J-H spectra and Lp photometry using a linear combination of cloudy and cloud-free surfaces and nonequilibrium chemistry, and we allowed the planet’s radius to vary independently of the radii given by evolutionary tracks. Models of this type generally produce reasonable fits to other directly imaged planets (1113, 15, 37, 38). This model produced a slightly lower effective temperature. The spectral shape and colors only weakly constrain gravity but favor lower masses, and the radius (~1 RJ) is consistent with evolutionary tracks, given the age of the system. Table 2 summarizes the results of the modeling. With the spectral and atmospheric uncertainties, a wide range of other models (including those with temperatures as high as ~1000 K) are also broadly consistent with the observations. The low temperature is supported by the evidence of strong methane absorption that is not observed for other planets of similar age.

Table 2 Modeling results for 51 Eri b.
View this table:

The value of log(L/L), –5.4 to –5.8 (where L/L is the planet’s luminosity normalized by that of the Sun), is similar in all models, regardless of temperature or clouds. Combined with the age, the luminosity can be used to estimate the mass of the planet. For a hot-start model, this corresponds to a mass of ~2 (t/20 My)0.65 [(L/2 × 10−6L☉;)0.54 MJ, the lowest-mass self-luminous planet directly imaged to date (t, age of the planet). 51 Eri b, unlike other young (<100 million-year-old) planetary-mass companions, has a low enough luminosity to be consistent with cold-start core-accretion scenarios. In cold-start evolution, luminosity at an age of 20 My is nearly independent of mass, so the mass of 51 Eri b would be between 2 and 12 MJ.

51 Eri b and the GJ 3305 binary system form a hierarchical triple configuration (28), but the companion pair is far enough away that the planet is expected to be dynamically stable in its current orbit (26). Moreover, the young age of the system suggests that although long-term dynamical effects, such as secular Lidov-Kozai oscillations, might have altered the planet's eccentricity and inclination, it is unlikely that they have had time to produce the extreme eccentricities required for tidal friction to alter the planet's semimajor axis (39). The formation of a ~2-MJ planet at an orbital distance of ~15 AU around a Sun-like star can be explained by modest extensions to the core-accretion theory. Early versions of the theory found that accretion of the core at larger orbital distances is in danger of taking too long, failing to capture the natal gas before it dissipates (40). 51 Eri b is close enough to the star that this may be less of a problem, and the addition of migration (41) or pebbles that experience gas drag (42) also helps overcome this time-scale difficulty.

The transition from L-type to T-type planets appears to occur over a narrow range of temperatures, between ~1000 K (HR8799 b and PSO J318.5-22) (42) and 700 K (51 Eri b). Direct determination of an object’s mass, either through spectral surface gravity indicators or reflex astrometry of the primary star, could determine whether it formed through hot- or cold-start processes. 51 Eri b provides an opportunity to study in detail a planet that is still influenced by the initial conditions of its formation. With a methane-dominated spectrum, low luminosity, and a potentially low-entropy start, 51 Eri b is a bridge between wider-orbit, hotter, more massive planets and planets at Jupiter-like scales.

Supplementary Materials

Materials and Methods

Figs. S1 to S3

Tables S1 to S3

References (4890)


  1. Material and methods are available as supplementary materials on Science Online.
  2. This research has benefitted from the SpeX Prism Spectral Libraries, maintained by A. Burgasser at
  3. ACKNOWLEDGMENTS: This work is based on observations obtained at the Gemini Observatory, which is operated by the Association of Universities for Research in Astronomy under a cooperative agreement with NSF on behalf of the Gemini partnership, whose membership includes: NSF (United States); the National Research Council (Canada); the Comisión Nacional de Investigación Científica y Tecnológica (Chile); the Australian Research Council (Australia); the Ministério da Ciência, Tecnologia e Inovação (Brazil); and Ministerio de Ciencia, Tecnología e Innovación Productiva (Argentina). The research was supported by grants from NSF, including AST-1411868 (B.M., K.F., J.P., and A.R.), AST-0909188 and AST-1313718 (J.R.G., P.K., R.D.R., and J.W.), AST-1413718 (M.P.F. and G.D.), and AST-1405505 (T.B.). Support was also provided by grants from NASA, including NNX14AJ80G (B.M., F.M., E.N., and M.P.), NNH15AZ591 (D.S. and M.M.), NNX15AD95G (J.R.G. and P.K.), NNX11AD21G (J.R.G. and P.K.), and NNH11ZDA001N (S.M. and R.P.). J.R., R.D., and D.L. acknowledge support from the Fonds de Recherche du Quebec. Support is acknowledged from NSF fellowships DGE-123825 (A.Z.G.), DGE-1311230 (K.W.-D.), DGE-1232825 (S.G.W.), and DGE-1144087 (L.W.H.). Portions of this work were performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under contract DE-AC52-07NA27344. GPI data are archived at the Gemini Science Archive:
View Abstract

Stay Connected to Science

Navigate This Article