Spectrally resolved helium absorption from the extended atmosphere of a warm Neptune-mass exoplanet

See allHide authors and affiliations

Science  21 Dec 2018:
Vol. 362, Issue 6421, pp. 1384-1387
DOI: 10.1126/science.aat5879

Helium escaping from hot gas giants

Many gas giant exoplanets orbit so close to their host star that they are heated to high temperatures, causing atmospheric gases to escape. Gas giant atmospheres are mostly hydrogen and helium, which are difficult to observe. Two papers have now observed escaping helium in the near-infrared (see the Perspective by Brogi). Allart et al. observed helium in a Neptune-mass exoplanet and performed detailed simulations of its atmosphere, which put constraints on the escape rate. Nortmann et al. found that helium is escaping a Saturn-mass planet, trailing behind it in its orbit. They combined this with observations of several other exoplanets to show that atmospheres are being lost more quickly by exoplanets that are more strongly heated.

Science, this issue p. 1384, p. 1388; see also p. 1360


Stellar heating causes atmospheres of close-in exoplanets to expand and escape. These extended atmospheres are difficult to observe because their main spectral signature—neutral hydrogen at ultraviolet wavelengths—is strongly absorbed by interstellar medium. We report the detection of the near-infrared triplet of neutral helium in the transiting warm Neptune-mass exoplanet HAT-P-11b by using ground-based, high-resolution observations. The helium feature is repeatable over two independent transits, with an average absorption depth of 1.08 ± 0.05%. Interpreting absorption spectra with three-dimensional simulations of the planet’s upper atmosphere suggests that it extends beyond 5 planetary radii, with a large-scale height and a helium mass loss rate of ≲3 × 105 grams per second. A net blue-shift of the absorption might be explained by high-altitude winds flowing at 3 kilometers per second from day to night-side.

HAT-P-11b is a transiting, warm Neptune-class exoplanet (27.74 ± 3.11 Earth masses, 4.36 ± 0.06 Earth radii) that orbits its star in 4.89 days (13). Its orbit is near the edge of the evaporation desert, a region at close orbital distances characterized by a lack of observed Neptune-mass exoplanets (4, 5). The evaporation desert can be explained as the result of heating planetary atmospheres through stellar radiative flux: Planets that are insufficiently massive lose their gaseous atmospheres through its expansion and hydrodynamic escape (6, 7). The upper atmosphere of planets in mild conditions of irradiation, such as HAT-P-11b, could extend without being subjected to substantial loss and yield deep transit. The low density of HAT-P-11b and the detection of water in its atmosphere (8) suggest a hydrogen-helium–rich atmosphere clear of aerosols down to an altitude corresponding to 1 mbar atmospheric pressure.

We observed two transits of HAT-P-11b with the CARMENES (Calar Alto high-Resolution search for M dwarfs with Exoearths with Near-infrared and optical Échelle Spectrographs) (9) instrument on the Calar Alto 3.5 m telescope on 7 August 2017 (visit 1) and 12 August 2017 (visit 2). CARMENES consists of two high-resolution spectrographs covering parts of the visible (5200 to 9600 Å) and near-infrared (9600 to 17,100 Å) domains with spectral resolving powers of ~95,000 and ~80,000, respectively. We analyzed data from the near-infrared channel. The data are automatically reduced with the CARMENES Reduction and Calibration pipeline (10), which applies a bias, flat-field, and cosmic ray correction to the raw spectra. A flat-relative optimal extraction (FOX) (11) and wavelength calibration (defined in vacuum) were then applied to the spectra (12). We observed HAT-P-11 for 6 and 5.8 hours in visits 1 and 2, respectively, in 53 and 51 exposures, each of 408 s. Among these spectra, 19 and 18 were obtained during the 2.4-hour duration of the planetary transit in visits 1 and 2, respectively (13).

During a transit, the atmosphere of a planet blocks a fraction of the stellar light at a given wavelength, depending on its structure and composition. We retrieved the near-infrared transmission spectrum of the exoplanet atmosphere by calculating the ratio between the in-transit spectra and an out-of-transit master spectrum (fig. S1), representing the unocculted star. The out-of-transit master for each visit was determined by co-adding spectra taken before and after transit (13). Because of the change in radial velocity arising from the planet’s motion, the spectrum of its atmosphere experiences a spectral shift during the transit. Transmission spectra were calculated for each in-transit exposure, offset in wavelength to the planet’s rest frame, and co-added (1416) to search for absorption from the planet atmosphere. HAT-P-11b has an eccentric orbit [eccentricity (e) = 0.26] that causes the planetary radial velocity to increase from −36 km‧s−1 to −24 km‧s−1 during the transit. As a result, any absorption signatures from the planet atmosphere are expected to be blue-shifted with respect to their rest wavelengths in the stellar reference frame. This helps to distinguish between signals with stellar or planetary origins. The signal-to-noise ratio also increases because the planet absorption is offset from the equivalent stellar line, unlike planets on circular orbits (fig. S2) (13). A search for atmospheric absorption features in excess of the planetary continuum absorption signal, optical transit depth ~3400 ppm, revealed absorption in the near-infrared He i triplet (10,832.06, 10,833.22, and 10,833.31 Å in vacuum) (Figs. 1 and 2). The He i triplet originates from a transition between the 23P state and the metastable 23S state, which can be populated by recombination from the singly ionized state or by collisional excitation from the ground state (17). The triplet is spectrally and temporally resolved during the transit owing to the high spectral resolution and fast cadence of the observations. The two strongest lines of the triplet are blended, whereas the third, weakest (and bluest) line is resolved from the two others. These transitions occur in a spectral region devoid of strong water absorption lines or OH emission lines from Earth’s atmosphere (13). The spectral region is also devoid of strong stellar absorption features (fig. S1) (13).

Fig. 1 He i transmission spectra of HAT-P-11b as a function of orbital phase.

(A) Plotted in the star rest frame. (B) Plotted in the planet rest frame. Horizontal orange dotted lines correspond to the beginning and end of transit. Green lines show the three He i transitions in the planet rest frame. Excess atmospheric absorption is visible as a white signal centered on the He i transitions, following the planetary motion, and occurring only during the transit. (C and D) The equivalent simulated maps for our best-fitting atmospheric model (13).

Fig. 2 Average He i transmission spectrum of HAT-P-11b in the planetary rest frame and transit light curve.

(A) Transmission spectra in visit 1 (blue) and visit 2 (orange), showing the absorption signature centered on the He i triplet (rest wavelengths shown as dashed gray lines). The black points show the weighted average over the two visits, and the red line is our best-fitting model. Wavelengths are in the planet rest frame. (B) Light curves derived from the spectra in (A) normalized to the expected planetary continuum absorption and integrated over 10,832.84 to 10,833.59 Å. Plotting symbols are the same as in (A), and the theoretical planet light curve without helium absorption is shown in gray. The black light curve was binned in phase. The green band is the time window (−1 hour to +1 hour) used to produce the average spectrum in (A). Vertical gray dashed lines correspond to the beginning and end of the transit.

Absolute fluxes cannot be determined from ground-based high-resolution spectra because of the variability of Earth’s atmosphere and light losses at the spectrograph entrance. Although this does not prevent the detection of spectrally localized absorption features arising from the planet atmosphere, it does mean that these observations are not sensitive to any continuum occultation by the atmosphere. To highlight the excess atmospheric He i absorption from HAT-P-11b and to allow comparison with simulations, we rescaled the continuum of each individual transmission spectrum using the theoretical transit white light curve of the planet (table S1) (13). The rescaled transit light curve integrated over the spectral range 10,832.84 to 10,833.59 Å is shown in Fig. 2, including the most significant excess atmospheric absorption of 1.08 ± 0.05% (21σ, where σ is the standard deviation). The absorption signature is centered on the He i triplet in the planet reference frame, occurs during the planet transit, and is repeatable over two visits (0.82 ± 0.09% in visit 1 and 1.21 ± 0.06% in visit 2), so it arises from helium in the atmosphere of HAT-P-11b (Fig. 1 and fig. S2). The difference in absorption between the two visits could arise from variability in the size of the atmosphere or its helium density. The peak of the resolved helium absorption profile reaches ~1.2%, corresponding to an equivalent opaque radius of ~2.29 planetary radii (Rp).

We interpreted the observations of HAT-P-11b using three-dimensional (3D) simulations of its atmosphere using the EVaporating Exoplanets (EVE) code (18, 19) [a detailed description is provided in (13)]. EVE was used to generate theoretical spectra after absorption by the planet and its upper atmosphere, accounting for limb-darkening, for the partial occultation of the stellar disk during ingress/egress, and for 3D effects linked to the atmospheric structure. The upper atmosphere consists of the thermosphere, the layer heated by the stellar irradiation, and the exosphere, the above layer in which the gas becomes collisionless (13). The thermosphere is parameterized with an isotropic, hydrostatic density profile defined by the ratio between the temperature of the thermosphere and its mean atomic weight, Tth/μ. We included a constant upward velocity vth to account for the bulk expansion of the thermosphere driven by the stellar extreme ultraviolet (XUV) irradiation. This expansion can lead to substantial mass loss, so we modeled the exosphere of HAT-P-11b by releasing metastable helium atoms at the top of the thermosphere (13). The altitude of the thermopause (the thermosphere/exosphere boundary, also known as the exobase) (Rtrans) is a free parameter in the fitting, as is the mass loss rate of metastable helium Embedded Image. Monte Carlo particle simulations are used to compute the dynamics of the atmosphere under the influence of the planet and star gravity, and the stellar radiation pressure. The density profile of metastable helium in the thermosphere is scaled so that it matches the density of exospheric metaparticles at the thermopause. We assume that the low densities in the collisionless exosphere prevent the formation of additional metastable helium atoms or their de-excitation through collisions (17).

The theoretical spectra were oversampled in time and wavelength compared with the CARMENES observations. We therefore convolved the output with the instrumental response, resampled over the CARMENES wavelength scale, and averaged within the time windows of each observed exposure. The theoretical and observed time-series spectra were compared over visits 1 and 2 together (103 exposures), limiting the fitting to the spectral range 10,826 to 10,834 Å (139 pixels, defined in the star rest frame) to avoid contamination by Earth atmosphere. We calculated a grid of simulations as a function of the four free parameters in the model, using χ2 as the merit function (fig. S4).

The exploration of the model parameter space reveals a broad χ2 minimum (χ2 ~ 6130 for 14,178 data points). The best-fitting thermopause altitudes extend between 5 Rp and the Roche Lobe (the limit of the gravitational influence of the planet, at 6.5 Rp). The Tth/μ values indicate high temperatures and/or low mean atomic weight (Tth/μ ≥ 24,000 K‧amu−1), suggesting a large fraction of ionized gas and free electrons. The width of the absorption signature is dominated by thermal broadening, but the upward expansion of the thermosphere could play a role with vth up to 10 km’s−1, which is in the range of values predicted for HAT-P-11b (20). Comparison between our best-fitting signature from a radially expanding thermosphere and the observed absorption profile reveals that it is symmetrical but blue-shifted (Fig. 3). Including zonal winds flowing from day- to night-side in our best-fitting models provides a better match to the observed absorption profile (χ2 = 6121) for velocities of ~3 km’s−1 (fig. S4). 2D hydrodynamical simulations (21) have shown that such winds can form at high altitudes in the extended atmospheres of giant planets.

Fig. 3 Contribution of zonal winds to the He i absorption signature from HAT-P-11b.

The black points are the observed average over the two nights, as shown in Fig. 2A. The blue curve corresponds to a radially expanding thermosphere, and the red curve (shown in Fig. 2A) is blue-shifted by the additional contribution of zonal winds flowing from day- to night-side at 3 km‧s−1.

Our results suggest a negligible contribution from the exosphere, with Embedded Image below 3 × 105 g‧s−1. This is consistent with the spectral symmetry of the observed absorption profile near the He i triplet and the symmetry of the time series absorption around the transit center (Fig. 2). These properties demonstrate that absorption from HAT-P-11b arises mostly from spherical layers of gas likely to be still gravitationally bound to the planet. The absence of post-transit absorption, or a strong absorption signal blueward of the helium transitions, rule out an extended tail of helium trailing the planet. This is unlike the elongated hydrogen exosphere detected around GJ 436 b (22, 23), a warm Neptune with similar density to that of HAT-P-11b. Our best-fitting models yield densities of metastable helium ~10 cm−3 at altitudes between 5 and 6.5 Rp, within the range of values simulated for GJ 436 b at similar altitudes (17).

Our best-fitting simulation in Fig. 4 and fig. S5 illustrates how the radiative environment from the host star (spectral type K4) influences the exosphere of HAT-P-11b. Helium atoms in the shadow of the planet keep moving on their original escape trajectory (determined by the orbital velocity of the planet when they escaped), until they radiatively de-excite with a lifetime of 131 min (24). Outside of the planet shadow, helium atoms are blown away faster than this lifetime by the strong stellar radiation pressure. It is much stronger at the helium triplet wavelength (~10,833 Å) than for the hydrogen Lyman-α wavelength (1215.7 Å) because of the brighter near-infrared continuum. Radiation pressure on metastable helium atoms escaping HAT-P-11b is higher than the gravitational pull of the star by a factor of ~90, whereas it reaches a maximum of ~5 for the hydrogen exospheres of planets around G- and K-type stars (18). However, the low photoionization threshold of metastable helium atoms implies that their lifetime is only 2.4 min at the orbital distance of HAT-P-11b, which explains why we do not observe an extended comet-like tail trailing the planet. There are, therefore, extended upper atmospheres around both warm Neptunes HAT-P-11b and GJ 436 b. Although they have similar mass and radius, the different spectral types and XUV emission of their host stars (K- and M-type, respectively) are expected to produce different structures for their upper atmospheres. The presence of helium at high altitudes around HAT-P-11b nonetheless suggests that large amounts of hydrogen could be escaping into its exosphere.

Fig. 4 The best-fitting EVE simulation of the HAT-P-11b helium absorption time series.

The system is shown during transit ingress; egress and mid-transit are shown in fig. S5. Distances are defined with respect to the star center. (A) View of the exosphere from above the planet. Metastable helium atoms are colored as a function of their radial velocity in the stellar rest frame (color bar). The dashed circle is the projected transition between the exosphere and thermosphere regimes. All particles in this projected view are outside of the thermosphere. The eccentric orbit of HAT-P-11b (black disk, plotted above the exosphere for the sake of clarity) is shown as a green curve. The tail is due to particles in the planet shadow being protected from photo-ionization. (B and C) View along the LOS (line of sight) toward Earth, showing the (B) thermospheric and (C) exospheric regimes separately. (B) The thermosphere is colored as a function of the column density of metastable helium. (C) Particles in the exosphere are colored as in (A). (B) shows the grids discretizing the stellar disk, the thermosphere, and the planetary disk.

Theoretical models (17, 25) have predicted that the metastable He i triplet can be used to trace atmospheric evaporation. Because the 10,833-Å He i triplet is not absorbed by the interstellar medium, it allows probing planetary systems farther from Earth than the H i Lyman-α at 1215 Å (26). Although early searches were unsuccessful because of instrumental limitations (27), an unresolved detection of metastable helium on the inflated gas giant WASP-107b has been achieved with the Hubble Space Telescope (HST) (28). Because of the low spectral resolution of HST, the helium triplet in WASP-107b was covered with just 1 pixel. We have calculated that observations of HAT-P-11b helium atmosphere with the James Webb Space Telescope (JWST) would measure the triplet with a high sensitivity but over just 2 pixels (13). High-resolution spectrographs have the ability to spectrally resolve the He i transitions, aiding in the separation of planetary from stellar signals. As shown by our results (Fig. 2) (29), resolved observations of the He i triplet provide additional constraints on the extended atmospheres of exoplanets, from their thermosphere to exosphere.

Supplementary Materials

Materials and Methods

Tables S1 and S2

Figs. S1 to S6

References (3145)

References and Notes

  1. Material and Methods are available as supplementary materials.
Acknowledgments: We acknowledge the Geneva exoplanet atmosphere group for fruitful discussions and the support of X. Dumusque for the DACE platform. Funding: This work was based on observations collected at the Centro Astronómico Hispano Alemán (CAHA), operated jointly by the Max-Planck Institut für Astronomie and the Instituto de Astrofisica de Andalucia (CSIC) under OPTICON program 2017B/026, “Multi-wavelength observations of the warm Neptune HAT-P-11b: A journey across the atmosphere.” This project has received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement 730890. This material reflects only the authors views, and the Commission is not liable for any use that may be made of the information contained therein. This work has been carried out in the frame of the National Centre for Competence in Research “PlanetS” supported by the Swiss National Science Foundation (SNSF). R.A., V.B., D.E., C.L., L.P., F.P., and A.W. acknowledge the financial support of the SNSF. A.W. acknowledges the financial support of the SNSF through the grant P2GEP2_178191. A.L.d.E. thanks the CNES for financial support. This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (project FOUR ACES; grant agreement 724427). This project has also received funding from the European Research Council (ERC) under the European Union’s Seventh Framework Programme (Fp7/2007-2013)/ERC grant agreement 337592. Author contributions: R.A. coordinated the study, performed the data reduction, and analyzed the results. V.B. developed the EVE code, based on previous code written by V.B. and A.L.d.E.; R.A. and V.B. performed the EVE simulations and wrote the paper. J.J.S. performed the HST and JWST simulations. R.A., V.B., C.L., D.E., L.P., A.W., and F.P. wrote the OPTICON proposal. All authors participated in the discussion and interpretation of the results, and commented on the manuscript. Competing interests: The authors declared no competing interests. Data and materials availability: The CARMENES raw and reduced data can be obtained from the Calar Alto Observatory archive at under program number 051. The data and simulation outputs used to produce each figure are available at (30). The EVE code is described in (13, 18, 28) and available at

Stay Connected to Science

Navigate This Article