Prospects for Characterizing the Atmosphere of Proxima Centauri b
DDraft version November 1, 2016
Preprint typeset using L A TEX style emulateapj v. 12/16/11
PROSPECTS FOR CHARACTERIZING THE ATMOSPHERE OF PROXIMA CENTAURI b
Laura Kreidberg & Abraham Loeb Draft version November 1, 2016
ABSTRACTThe newly detected Earth-mass planet in the habitable zone of Proxima Centauri could potentiallyhost life – if it has an atmosphere that supports surface liquid water. We show that thermal phasecurve observations with the
James Webb Space Telescope ( JWST ) from 5 − µ m can be used to testthe existence of such an atmosphere. We predict the thermal variation for a bare rock versus a planetwith 35% heat redistribution to the nightside and show that a JWST phase curve measurement candistinguish between these cases at 4 σ confidence, assuming photon-limited precision. We also considerthe case of an Earth-like atmosphere, and find that the ozone 9 . µ m band could be detected withlonger integration times (a few months). We conclude that JWST observations have the potential toput the first constraints on the possibility of life around the nearest star to the Solar System.
Subject headings: planets and satellites: atmospheres — planets and satellites: individual: ProximaCentauri b INTRODUCTION
Anglada-Escud´e et al. (2016) recently announced theexciting discovery of a potentially habitable planet orbit-ing our nearest neighboring star, Proxima Centauri. Theplanet has a minimum mass of 1 . M ⊕ and an insolationequal to two thirds that of Earth, suggesting that it couldhave a rocky surface with temperatures appropriate forthe existence of liquid water.Recent transit surveys such as Kepler have shown thatEarth-like planets like these are very common – theyare found around 10 - 25% of stars (e.g. Petigura et al.2013; Dressing & Charbonneau 2013, 2015). However,none of the Earth analogs detected to date have beenfeasible targets for atmosphere characterization becausetheir host stars are too distant. At a distance of just oneparsec, Proxima b provides the first opportunity for de-tailed characterization of a habitable world beyond theSolar System. An immediate question is whether Prox-ima b has an atmosphere at all. Tidally locked planetsorbiting M-dwarfs face unique challenges to their atmo-spheric stability. The atmosphere may “collapse” if thevolatile inventory freezes out and becomes trapped onthe nightside (Joshi et al. 1997). The atmosphere is alsosubject to erosion by stellar winds, which are denser andfaster for M-dwarfs than Sun-like stars (Zendejas et al.2010). Proxima also has a high rate of flaring activitythat may further threaten the planet’s atmosphere (Dav-enport et al. 2016). To reveal Proxima b’s evolutionaryhistory and potential for hosting life, a first step is toascertain whether its atmosphere has survived. POSSIBLE APPROACHES FOR ATMOSPHERECHARACTERIZATION
Proxima b is not likely to transit its host star. Thetransit probability is only 1%, and photometric transitsearches have not revealed the planet (Kipping et al.
E-mail: [email protected] Harvard Society of Fellows, Harvard University, 78 Mt.Auburn St., Cambridge, MA 02138, USA Harvard-Smithsonian Center for Astrophysics, 60 GardenStreet, Cambridge, MA 02138, USA µ m. Next-generation ground-based extremely large telescopes (ELTs) will be capableof such measurements, but they will not be available untilthe mid-2020s.Method (ii) is a more promising approach for near fu-ture characterization of the planet. Variation in reflectedstarlight can be detected by combining high-resolutionground-based optical spectroscopy with high contrastimaging (Riaud & Schneider 2007; Snellen et al. 2015).The spectroscopy is sensitive to the Doppler shift ofstarlight reflected by the planet as it orbits, and the highcontrast imaging suppresses the stellar signal. This mea-surement can reveal the planet’s orbital inclination andthe albedo scaled by the projected area. Lovis et al.(2016) recently performed a feasibility study for apply-ing this technique to Proxima b with existing observingfacilities. They propose combining the SPHERE high-contrast imager and the ESPRESSO spectrograph on theVery Large Telescope (VLT), and find that the planet canbe detected at 5 σ confidence in 20 - 40 nights of observ-ing time (assuming an Earth-like albedo). Lower albedo(expected for an airless planet) will make the detectionmore challenging with current facilities, but it would bewithin reach of the ELTs. In addition to reflected light,this method may also be sensitive to OI auroral emission(Luger et al. 2016).In the remainder of this paper, we focus on method(iii). The basic idea is that a tidally locked planet mayhave a temperature gradient from the dayside to thenightside. As the planet orbits, the fraction of the day- a r X i v : . [ a s t r o - ph . E P ] O c t Kreidberg & Loebside that is visible varies. The thermal phase variationcan be predicted exactly for a planet with no atmosphere,assuming the inclination is known via method (ii). If anatmosphere is present, it tends to redistribute the heat tothe nightside, thus lowering the amplitude and changingthe color of the thermal phase variation. This idea hasbeen discussed before (Gaidos & Williams 2004; Seager &Deming 2009; Selsis et al. 2011; Maurin et al. 2012; Selsiset al. 2013), and and applied specifically to climate mod-els of Proxima b by Turbet et al. (2016). Here we per-form detailed signal-to-noise calculations and simulate aretrieval of the planet’s atmospheric properties based onpossible observations with the
James Webb Space Tele-scope ( JWST , scheduled for launch in 2018). METHODS
Toy Climate Model
To predict the thermal emission from Proxima b, werequire a model of its climate. We use a simple modelthat makes the following assumptions: if no atmosphereis present, all the incident stellar flux will be absorbedand reradiated on the planet’s dayside as a blackbody.We assume the emissivity of the surface is unity; for morediscussion of this point, see §
5. On the other hand, ifthere is an atmosphere, it can advect heat to the night-side.We also assume the planet is tidally locked, and checkthis assumption by calculating the tidal locking timescalefrom Gladman et al. (1996). For a tidal Q of 100 andan initial rotation rate of one cycle per day, the lockingtimescale is t lock ∼ yr, which is short compared to theage of the system. However, we note that if the planethas a non-zero eccentricity, it is possible that the lock-ing timescale if significantly longer (Coleman et al. 2016;Ribas et al. 2016). Continued radial velocity monitoringof the system will be important for precise constraints onthe eccentricity. We also note that moons could poten-tially delay tidal locking of the planet to the star; how-ever, moons are unlikely to be present around Proximab due to instability of their orbits on gigayear timescales(Sasaki & Barnes 2014).Based on these assumptions, we use an analyticclimate model of the form: σT = (cid:26) S × (1 − A ) × F/ , π/ < | θ | < πS × (1 − A ) × [ F/ − F ) cos z ] , | θ | ≤ π/ S = 0 . S ⊕ = 890 W / m is the Proxima b ir-radiance, A is the Bond albedo, 0 < F < . z is the zenith angle, θ is the longitude, and σ isthe Stefan-Boltzmann constant. We neglect internal heatflux because it contributes a negligible fraction of the to-tal energy budget (assuming an Earth-like heat flux oforder 100 mW / m Davies & Davies 2010).The above expression behaves well for the limitingcases F = 0 and F = 0 .
5. For the zero redistributioncase, the dayside incident flux is proportional to the co-sine of the zenith angle, and the nightside recieves zeroflux, in agreement for the expected climate of a planetwith no atmosphere. For the other limiting case F = 0 . Angular separation from substellar point (deg) T e m pe r a t u r e ( K ) F = 0.0F = 0.01F = 0.05F = 0.25F = 0.5
Fig. 1.—
Temperature maps for a range of values for the heatredistribution parameter F . This calculation assumes an insolation S = 890 W / m and an albedo of 0.1. where half the incident flux is redistributed to the night-side, the planet is isothermal. In Figure 1, we show ex-ample temperature maps for a range of values for F . JWST /MIRI signal-to-noise calculation
To calculate the feasibility of detecting Proxima b’sthermal emission with
JWST we used the beta versionof the
JWST
Exposure Time Calculator (ETC, avail-able at jwst.etc.stsci.edu ) to estimate signal-to-noise(SNR) for MIRI observations of Proxima Cen. We con-sidered observations with the LRS spectrograph usingthe slitless mode optimized for exoplanet observations(Kendrew et al. 2015) as well as photometric imagingobservations at λ > µ m. We did not consider MRSspectroscopy, because MRS is an integral field spectro-graph, and slit losses are a major concern for precisionexoplanet atmosphere characterization (Beichman et al.2014).For our model spectrum, we assumed a 3000 K black-body normalized to the K-band magnitude of ProximaCentauri. Line blanketing in the optical and near-IR cancause the spectrum to depart from a blackbody, but thiseffect is weaker at the wavelengths sensed by MIRI wherefewer lines are present. To check whether line blanketingaffects the normalization of the spectrum, we compared aPHOENIX stellar atmosphere model to a blackbody andfound that at K-band they agree to within 10 percent(Husser et al. 2013).From the input stellar spectrum, the ETC producesthe expected count rate per resolution element in pho-toelectrons per second. For slitless LRS spectroscopy,which has a resolution of 100 at 7 . µ m, the count rateranges from 3 . × e / s / resolution element near 5 . µ mto 4 × at 10 µ m. For the filters, central wave-lengths of λ = [12 . , , . , . , .
1] have count ratesof [1 . × , . × , . × , . × ]. These valuesare in good agreement with signal-to-noise predictionsfrom Cowan et al. (2015).For LRS spectroscopy, the Proxima spectrum saturatesthe detector over the range 5 − µ m, even in the short-est exposure time (0.15 sec). However, there is a possi-bility of implementing alternate readout modes (NikoleLewis, priv. comm.) that could decrease the number ofsaturated pixels and also improve the duty cycle of the Time (days) P l ane t/ s t a r f l u x ( pp m ) Time (days) P l ane t/ s t a r f l u x ( pp m ) Fig. 2.—
Thermal phase curves for a bare rock (left) and a planet with 35% heat redistribution. The models both assume an inclinationof 60 degrees and an albedo of 0.1.
10 15 20 25
Wavelength (microns) D a y - n i gh t c on t r a s t ( pp m ) Fig. 3.—
The phase variation spectrum for Proxima b. The mod-els (blue and red curves) correspond to the difference between themeasured star+planet spectrum at phase 0.5 and at phase 0.0 forthe case of a rock (no heat redistribution, blue), and a planet withan atmosphere that advects 35% of the heat to the nightside (red).The data points are simulated MIRS/LRS measurements from 5 -12 µ m and MIRI imaging measurements > µ m. The uncertain-ties for the simulated data are based on the photon noise for thedifference between two phase curves bins, where the spectrum ineach bin is co-added over an integration time of 24 hours. observations.In either case, our science goals are not strongly af-fected by saturating the bluest pixels ( < µ m) becausethe signal is small (a few ppm) at those wavelengths.The photometric observations do not saturate as quicklyand can therefore achieve 80% or higher duty cycle. Forour final SNR calculations, we assumed a 50% duty cyclefor LRS and an 80% duty cycle for the photometry. Wealso assumed that the noise is photon-limited; i.e., uncer-tainty due to background and flat-fielding are negligible. RESULTS
Predicted thermal phase variation
We used the toy climate model to predict the IR phasevariation of Proxima b over the course of its orbit. Weconsidered two scenarios: a “rock” case, with zero as-sumed heat redistribution, and an “atmosphere” case,with moderate redistribution ( F = 0 . µ m.For both scenarios, we assumed an inclination of 60 ◦ (the median value for an isotropic distribution of incli-nations) and a Bond albedo of 0.1 (a typical value forrocky bodies; Usui et al. 2013). Physically realistic atmo-spheres would likely have a higher albedo (e.g., Earth’salbedo is 0.3). However, assuming the same albedo is amore conservative choice because it means the planet’sproperties are harder to distinguish. We used the valuesreported in Anglada-Escud´e et al. (2016) for the planet’sphysical and orbital parameters. We assumed a planetradius of 1 . R ⊕ , based on predictions from the terrestrialplanet mass-radius relation (Chen & Kipping 2016). See § − µ m for the two cases.We calculate the phase curve directly from the tem-perature map using the SPIDERMAN software package forPython (in development on GitHub at https://github.com/tomlouden/SPIDERMAN
Louden & Kreidberg) . Thepeak-to-trough phase variation at 10 µ m for the rock caseis 35 ppm. The amplitude is sensitive to wavelength,varying by over an order of magnitude between 5 and10 µ m. This strong wavelength dependence results fromthe ratio of blackbody intensities for the planet versus thestar. The planet’s emission peaks near 10 µ m, whereasthe star peaks in the optical and decreases steeply withwavelength. The atmosphere case ( F = 0.35) shows asimilar wavelength dependence in the phase curve am-plitude; however, the overall amplitude is scaled downby a factor of two compared to the rock. Simulated Spectrum and Retrieval of AtmosphericProperties
Using the climate model and
JWST noise estimatesdescribed in §
3, we simulated a measurement of the ther- Kreidberg & Loeb r p = . +0 . − . . . . . A A = . +0 . − . . . . . r p . . . . F . . . . A . . . . F F = . +0 . − . Fig. 4.—
Pairs plot for MCMC fit to the MIRI/LRS thermal vari-ation spectrum showing the posterior distributions for the planet-to-star radius ratio r p , the Bond albedo A , and the fractional heatredistribution F . mal phase variation for Proxima b. Following Selsis et al.(2011), we define the phase variation as the differencebetween the star + planet spectrum at phase 0.5 andphase 0.0. We show the results in Figure 3. We plot sim-ulated data for LRS as well as all the photometric filters.However, note that each data set (LRS + each filter in-dividually) requires a complete phase curve observationto obtain. Full phase coverage is required because thedetectors are expected to have percent-level sensitivityvariations over time, which make it impossible to stitchtogether segments of the phase curve observed at differ-ent epochs.We wish to know how robustly we can determine theheat redistribution on the planet based on these mea-surements. The key parameters that the spectrum de-pends on are the orbital inclination, the planet-to-starradius ratio, the albedo, and the heat redistribution. Weassume the inclination is known exactly and that the ra-dius ratio is known to a precision of 10% (for discussionof this point, see section § A and the heat redistribution F .To assess how tightly we can constrain the albedo andheat redistribution, we ran an MCMC fit to the sim-ulated LRS spectrum from Figure 3, assuming a fixedinclination, a Gaussian prior on the planet radius withstandard deviation 10% of the best fit radius value, andalbedo A and redistribution F were free parameters. Weused the emcee package to perform the fit (Foreman-Mackey et al. 2013). Figure 4 shows the resulting poste-rior distribution of the fit parameters. We measure theheat redistribution to be F = 0 . +0 . − . , which is incon-sistent with the moderate redistribution atmosphere caseat 4.5 σ confidence. The albedo is also well constrained,to A = 0 . +0 . − . . These results demonstrate that a singleMIRI/LRS phase curve observation is a powerful diag-nostic of the presence of an atmosphere on Proxima b. The 10 µ m ozone feature We also explored the feasibility of detecting an ozoneabsorption feature from the planet at 10 µ m. This isa prominent feature of Earth’s IR emission spectrum,and noteworthy as a potential biosignature (Segura et al.2005; Lin et al. 2014), though it can also arise fromarise abiotically from evaporated oceans (e.g. Ribas et al.2016). For this case, we assumed Earth-like atmosphericproperties: a Bond albedo A = 0 . HubbleSpace Telescope .The continuum normalized spectrum of the star +planet is shown in Figure 5. The ozone feature doesnot vary with planet orbital phase, but it is in princi-ple detectable from a very high signal-to-noise combinedspectrum , because M-dwarfs are too hot to have abun-dant ozone in their photospheres. However, the predictedfeature amplitude is small – less than one part per mil-lion over a narrow band. For a qualitative illustration ofhow much observing time is required to detect the fea-ture, we plotted a simulated spectrum co-added from 60days total integration. We note that this model spec-trum is the most challenging case to detect. If there isa temperature contrast between the day and nightside,the ozone feature depth on the dayside would be a factorof several larger than for the isothermal scenario, and inaddition, the periodicity of the signal would help distin-guish it from variation in the stellar continuum due tochanging star spot coverage. In any case, such an obser-vation would be tremendously exciting if successful, butwe discuss several important caveats regarding feasibilityin the next section. ASSUMPTIONS & CAVEATS
In our analysis, we make several important assump-tions about the planetary system and the data obtainablewith
JWST , which we outline below:1.
The star is a perfect blackbody.
In reality, the stel-lar spectrum has a forest of atomic and molecu-lar absorption lines (mainly due to water). Modelinfrared spectra for mid-M dwarfs depart from ablackbody at the 1% level at the wavelength andresolution of MIRI/LRS (Veyette et al. 2016), andabsorption features will change in amplitude as starspots with varying water content rotate in and outof view. Proxima’s star spot properties not known,but assuming 1% variability (appropriate for a 1%covering fraction and 300 Kelvin temperature dif-ference), the stellar spectrum will vary at the 100-ppm level. Correcting for this effect is particularlyimportant for the detection of the 9.8 µ m ozone fea-ture, which is only one ppm in amplitude. Even forthe thermal phase variation, which is periodic andlarger in amplitude, the changing starspot coverage C o n t i nuu m - n o r m a li z e d f l u x - ( × ) Fig. 5.—
Continuum normalized star + planet spectrum (blueline). The absorption feature centered at 9.8 µ m corresonds toan ozone band. The feature at 8 µ m is due to methane, whichis also unexpected in the stellar spectrum (assuming equilibriumchemistry; Heng et al. 2016). The simulated data assumed photon-limited precision from 60 days of co-added observations. could pose a significant challenge. The star’s rota-tion period is 83 days (Anglada-Escud´e et al. 2016),but individual spots can rotate out of view ontimescales comparable to the planet’s orbit. There-fore, robust detection of the planet signal will re-quire improved stellar models and water lines listsin the infrared (Fortney et al. 2016), as well as de-tailed characterization of the stellar spot coverage.Before undertaking an intensive JWST observingcampaign, it will be important to assess whetherthese improvements are feasible at the level of pre-cision required.2.
The precision of the measurements is photon-limited.
Past observations with space-based tele-scopes have been successful at reaching the photonlimit (Kreidberg et al. 2014; Ingalls et al. 2016).These results are encouraging; however, they havenot approached ppm-level precision, and MIRI hasa different type of detector (arsenic-doped silicon).Testing the precision of the MIRI detectors earlyin the mission will be key for guiding potential ob-servations of Proxima Centauri.3.
The inclination and planet-to-star radius ratio canbe determined.
These quantities are necessary forinterpreting the thermal phase variation. We as-sume the inclination will be measured with thecombination of high-contrast imaging and high-resolution spectroscopy (method (ii) of § ◦ (Brogi et al. 2012). The planet-to-star radius ratiocan also be estimated precisely: the mass-radiusrelation for terrestrial bodies is tight (with scatterless than five percent; Dressing et al. 2015; Chen& Kipping 2016). Therefore, the dominant sourcesof uncertainty for the planet-to-star radius ratioare the stellar radius, which is known to about 5percent from interferometric observations (Demoryet al. 2009) and the planet minimum mass, whichis already known to 10 percent (Anglada-Escud´eet al. 2016).4. Heated rock radiates with a blackbody spectrum.
Forthis to be the case, the emissivity of the rock mustbe unity. Rocky material tends to have high emis-sivity in the IR (near 0.9), but the exact value de-pends on wavelength and the composition of therock, and can drop as low as 0.5 (Karr 2013). De-tailed modeling of the impact of emissivity on thepredicted thermal phase variation is beyond thescope of this paper, but it should be consideredin future study of Proxima b. CONCLUSION
In this paper, we outlined an observational test of theexistence of an atmosphere on Proxima b. By combiningintensive observing programs from the ground and space,it is possible to precisely measure the fraction of incidentflux that is redistributed to the nightside of the planet.In the case of no redistribution, one could infer theplanet does not have an atmosphere and is unlikely tohost life. By contrast, if we do find evidence for sig-nificant energy transport, this would indicate that anatmosphere or ocean are present on the planet to helptransport the energy. In that case, Proxima b would bea much more intriguing candidate for habitability. Eitherway, these observations will provide a major advance inour understanding of terrestrial worlds beyond the SolarSystem.We thank Nikole Lewis, Natasha Batalha, and KlausPontoppidan for tips about
JWST
SNR calculations. Weare also grateful to Jayne Birkby and Mercedes L´opez-Morales for insightful discussions about what observa-tions are possible for Proxima Cen b. We appreciatethoughtful comments on the manuscript from Ed Turnerand Tom Greene. We thank Colin Goldblatt for gra-ciously sharing his knowledge of rocks. We also thankSarah Rugheimer and Mark Veyette for making theirmodels publicly available. Finally, we thank the refereefor his or her comments, which improved the quality ofthe discussion.
REFERENCESAnglada-Escud´e, G., Amado, P. J., Barnes, J., et al. 2016,Nature, 536, 437Beichman, C., Benneke, B., Knutson, H., et al. 2014, PASP, 126,1134Brogi, M., Snellen, I. A. G., de Kok, R. J., et al. 2012, Nature,486, 502Chen, J., & Kipping, D. M. 2016, ArXiv e-prints,arXiv:1603.08614 Coleman, G. A. L., Nelson, R. P., Paardekooper, S.-J., et al.2016, ArXiv e-prints, arXiv:1608.06908Cowan, N. B., Greene, T., Angerhausen, D., et al. 2015, PASP,127, 311Davenport, J. R. A., Kipping, D. M., Sasselov, D., Matthews,J. M., & Cameron, C. 2016, ApJ, 829, L31Davies, J. H., & Davies, D. R. 2010, Solid Earth, 1, 5