Detecting Proxima b's atmosphere with JWST targeting CO2 at 15 micron using a high-pass spectral filtering technique
I. Snellen, J.-M. Desert, L. Waters, T. Robinson, V. Meadows, E. van Dishoeck, B. Brandl, T. Henning, J. Bouwman, F. Lahuis, M. Min, C. Lovis, C. Dominik, V. Van Eylen, D. Sing, G. Anglada-Escude, J. Birkby, M. Brogi
aa r X i v : . [ a s t r o - ph . E P ] A ug Draft version August 25, 2017
Typeset using L A TEX default style in AASTeX61
DETECTING PROXIMA B’S ATMOSPHERE WITH JWST TARGETING CO AT 15 MICRON USING AHIGH-PASS SPECTRAL FILTERING TECHNIQUE
I.A.G. Snellen, J.-M. D´esert, L.B.F.M. Waters,
3, 2
T. Robinson,
V. Meadows, E.F. van Dishoeck, B.R. Brandl,
T. Henning, J. Bouwman, F. Lahuis, M. Min, C. Lovis, C. Dominik, V. Van Eylen, D. Sing, G. Anglada-Escud´e, J.L. Birkby, and M. Brogi
13, 14 Leiden Observatory, Leiden University, Postbus 9513, 2300 RA Leiden, The Netherlands Anton Pannekoek Institute for Astronomy, University of Amsterdam, P.O. Box 94249, 1090 GE Amsterdam, the Netherlands SRON Netherlands Institute for Space Research, Sorbonnelaan 2, 3584 CA Utrecht, the Netherlands Department of Astronomy and Astrophysics, University of California, Santa Cruz, CA 95064, USA NASA Sagan Fellow Astronomy Department, University of Washington, USA Delft University of Technology, Faculty of Aerospace Engineering, Kluyverweg 1, 2629 HS Delft, The Netherlands Max-Planck-Institute for Astronomy, Koenigstuhl 17, 69117 Heidelberg, Germany Observatoire de Gen`eve, Universit´e de Gen`eve, 51 chemin des Maillettes, 1290 Versoix, Switzerland School of Physics, University of Exeter, Exeter, UK School of Physics and Astronomy, Queen Mary University of London, 327 Mile End Road, London E1 4NS, UK Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge MA 02138, USA Center for Astrophysics and Space Astronomy, University of Colorado at Boulder, Boulder, CO 80309, USA NASA Hubble Fellow
ABSTRACTExoplanet Proxima b will be an important laboratory for the search for extraterrestrial life for the decades ahead.Here we discuss the prospects of detecting carbon dioxide at 15 µ m using a spectral filtering technique with theMedium Resolution Spectrograph (MRS) mode of the Mid-Infrared Instrument (MIRI) on the James Webb SpaceTelescope (JWST). At superior conjunction, the planet is expected to show a contrast of up to 100 ppm with respectto the star. At a spectral resolving power of R=1790–2640, about 100 spectral CO features are visible within the13.2-15.8 µ m (3B) band, which can be combined to boost the planet atmospheric signal by a factor 3–4, dependingon the atmospheric temperature structure and CO abundance. If atmospheric conditions are favorable (assuming anEarth-like atmosphere), with this new application to the cross-correlation technique carbon dioxide can be detectedwithin a few days of JWST observations. However, this can only be achieved if both the instrumental spectral responseand the stellar spectrum can be determined to a relative precision of ≤ × − between adjacent spectral channels.Absolute flux calibration is not required, and the method is insensitive to the strong broadband variability of the hoststar. Precise calibration of the spectral features of the host star may only be attainable by obtaining deep observationsof the system during inferior conjunction that serve as a reference. The high-pass filter spectroscopic technique withthe MIRI MRS can be tested on warm Jupiters, Neptunes, and super-Earths with significantly higher planet/starcontrast ratios than the Proxima system. Corresponding author: Ignas [email protected] INTRODUCTIONThe discovery of the exoplanet Proxima b through long-term radial velocity monitoring (Anglada-Escud´e et al. 2016)is exciting for two reasons. First, it confirms that low-mass planets are very common around red dwarf stars, a picturethat was already emerging from both transit and radial velocity surveys (Berta et al. 2013; Dressing & Charbonneau2015). Second, the proximity of this likely temperate rocky planet at a mere 1.4 parsec from Earth makes it mostfavourable for atmospheric characterisation, making Proxima b an important laboratory for the search for extrater-restrial life for the decades ahead.Proxima b is found to orbit its host star in 11.2 days, placing it at an orbital distance of 0.0485 au. Since theluminosity of the host star is only 0.17% of that of our Sun, the level of stellar energy the planet receives is 30% lessthan the Earth, but nearly 70% more than Mars. This means that in principle it could have surface conditions thatsustain liquid water - generally thought as a prerequisite for the emergence and evolution of biological activity (e.g.Kasting et al. 1993; Kopparapu et al. 2013). Although other habitats can be envisaged outside the so called ’habitablezone’ , such as under the icy surface of Jupiter’s moon Europa (e.g Reynolds et al. 1983; Kargel et al. 2000), it israther unlikely that signs of biological activity under such conditions could be detected in extrasolar planet systems(Lovelock 1965; Segura et al. 2005).It is highly debatable whether Earth-mass planets in the habitable zones of red dwarf stars, such as Proxima b, couldsustain or have ever sustained life. First, it is expected that the pre-main sequence of red dwarf stars lasts up to a billionyears during which the stellar luminosity is significantly higher than during the main-sequence lifetime of the star.This means that the planet will have had a significantly hotter climate early on, during which it may have lost mostor maybe all of its potential water content (Ramirez & Kaltenegger 2014; Luger & Barnes 2015). Second, Proxima, asa large fraction of red dwarf stars, is a flare star that actively bombards the planet atmosphere with highly energeticphotons and particles (Khodachenko et al. 2007; Lammer et al. 2007), possibly causing a large fraction of the planetatmosphere to be lost. Thirdly, planets in the habitable zone of red dwarf stars are expected to be tidally locked, andmay be synchronously rotating – always faced with the same side to the host star (e.g. Ribas et al. 2016; Barnes et al.2016). It is not clear whether a habitable climate can be sustained with such a configuration (e.g. Kite et al. 2011).However, several theoretical endeavours, also in the wake of the Proxima b discovery, show that despite these possibledrawbacks the planet may well host an atmosphere with liquid water on its surface (Tarter et al. 2007; Ribas et al.2016; Turbet et al. 2016).Several studies have investigated the potential detectability of Earth-like atmospheres of planets orbiting late M-dwarfs using high-dispersion spectroscopy. Snellen et al. (2013) calculated whether the transmission signature ofmolecular oxygen of a twin earth-planet in front of a mid-M dwarf could be observed using high-dispersion spectroscopy(see also Rodler & L´opez-Morales 2014) and showed that it would require a few dozen transits with the EuropeanExtremely Large Telescope (E-ELT) to reach a detection. However, it is very unlikely that Proxima b is transiting(Kipping et al. 2017). A more promising avenue is to combine high-dispersion spectroscopy with high-contrast imaging(HDS+HCI Sparks & Ford 2002; Riaud & Schneider 2007; Snellen et al. 2014, 2015; Kawahara et al. 2014; Luger et al.2017). Snellen et al. (2015) simulated observations with the E-ELT using optical HDS+HCI of a then still hypotheticalEarth-like planet around Proxima, showing that detection of such planet would be possible within one night. Recently,Lovis et al. (2016) argued that if the new ESPRESSO high-dispersion spectrograph at the ESO Very Large Telescope(VLT) can be coupled with the high-contrast imager SPHERE, and the latter has a major upgrade in adaptive opticsand coronagraphic capabilities, a detection of Proxima b is within reach.Since the next-generation of extremely large ground-based telescopes is at least 5 to 10 years away, the JamesWebb Space Telescope (JWST) could be a more immediate option to detect an atmospheric signature of Proximab. Unfortunately, simple diffraction arguments tell us that the JWST is not large enough to spatially separate theplanet from its host star - with a maximum elongation of 37 mas ( ∼ λ/D at 1 µ m). Several studies (Greene et al.2016; de Wit et al. 2016) show that atmospheric characterisation of super-Earths transiting small M-dwarf stars iswithin range of the JWST. However, the probability that Proxima b transits its host star is only 1.3%. Instead,Kreidberg & Loeb (2016) discuss the possibility of detecting the thermal phase curve with the JWST Mid-InfraredInstrument MIRI, using its slitless 5 - 12 µ m Low Resolution (LRS) mode . Since the planet is expected to be tidallylocked, the night-to-dayside temperature gradient will result in a variation in apparent thermal flux as function oforbital phase. Depending on the orbital inclination, and on whether the planet has an atmosphere (which affects theredistribution of absorbed stellar energy around the planet), variations of up to ∼
35 ppm are expected in the LRSwavelength regime. They show that in the ideal case of photon-limited precision, one can indeed detect the phase
Figure 1.
Schematic representation of the high-pass spectral technique. The left panel shows a toy-model planet spectrum (blue)with its low-resolution spectrum overlaid (black). The right panel shows the high-pass spectrally filtered signal consisting of thedifference between the high- and low resolution spectrum. Only the high-frequency components of the spectrum are preserved,making the high-pass filtered spectrum significantly easier to calibrate at a cost of information on the planet continuum fluxlevel. variation over a planet orbit (11.2 days - 268 hrs). However, a particular concern is the intrinsic variability of ProximaCen - which is known to be a flare star. A month-long observation program of the MOST satellite (Davenport et al.2016) detected on average two strong optical flares a day. Extrapolating their result to lower energies and mid-infraredwavelength implies that the star exhibits ∼
50 flares a day at levels >
500 ppm - an order of magnitude stronger thanthe expected phase variation amplitude of the planet. Hence it will be challenging to discern the planet signal withsuch observations.In this paper we discuss a new application to the cross-correlation technique used to probe exoplanet atmospheres(e.g Snellen et al. 2010), targeting the 15 µ m CO planet signal with the Medium Resolution Spectrograph (MRS)mode of MIRI. Importantly, it is unaffected by broadband flux variations of the host star. Furthermore, a detectionof CO would constitute conclusive evidence that Proxima b contains an atmosphere, and provide constraints to itstemperature structure. If Proxima b is indeed a terrestrial planet that formed at or near its current semi-majoraxis, it would have been subjected to the super-luminous phase of the star for up to 160 My (Barnes et al. 2016). Inthis time it may have undergone ocean loss and a runaway greenhouse, or had all but the heaviest molecules in itsatmosphere stripped early on, with the possibility of longer-term replenishment by volcanic outgassing over its 5 Gyrhistory (Lammer et al. 2007; Ribas et al. 2016; Meadows et al. 2016). All of these evolutionary processes would haveincreased the likelihood that the planetary atmosphere currently contains CO .In Section 2 we describe the details ofthe method, of the MRS mode of MIRI, and present simulations, including atmospheric modelling. The results arepresented and discussed in Section 3. A HIGH-PASS SPECTRAL FILTERING TECHNIQUEThe key to any exoplanet atmospheric observation is the ability to separate planet signatures from the overwhelmingflux of the host star. The cross-correlation technique used to probe exoplanet atmospheres does not work in the case ofProxima b with the JWST, since the spectral resolution is not sufficient to use the change in the radial component of theplanet orbital velocity to filter out the planet light (e.g. Snellen et al. 2010), neither is the angular resolution sufficientto separate the planet from the star (in which case the HDS+HCI technique could be used). In this case we proposeto use a new version of these techniques in which a spectral feature is targeted in the integrated planet+star spectrumthat can be differentiated from the stellar spectrum and attributed to the planet. Generally, this would not workbecause it constitutes an absolute spectrophotometric measurement requiring instrumental calibration and knowledgeof the stellar spectrum to a level significantly better than the planet signal. However, if the instrument spectralresolution is sufficiently high, the requirements on calibration and stellar spectrum knowledge can be significantlyrelaxed by probing the high-pass spectrally filtered signature instead of the absolute spectrophotometric signal. In thiscase the planet signal is the difference in flux density relative to a low-resolution mean. E.g. in the case of a molecularband composed of a series of distinct absorption lines, the high-pass filtered spectral signature consists of series ofpeaks and valleys in the spectrum (Fig. 1). This has the advantage that the planet signal is spread out over manypixels and consists of positive and negative high-frequency components. In particular low-frequency components inthe spectrum are notoriously difficult to calibrate due to required accuracies in the spectrum of the calibration source,and in stray-light and background corrections. However, these are not important in this case.We argue that the spectral filtering technique will work well for the MRS mode of MIRI targeting CO at 15 µ m. Inthe case of Proxima b, its variations in the radial component of its orbital velocity of up to 50 km sec − correspondsto a shift of ± MRS mode of MIRI
The Medium Resolution Spectrograph (MRS) mode of the Mid-Infrared Instrument (MIRI Rieke et al. 2015;Wright et al. 2015) on JWST utilises an integral field spectrograph (IFS) which has four image slices producingdispersed images of the sky on two 1024x1024 infrared detector arrays, which provide R = 1300-3600 integral fieldspectroscopy over a λ = 5 − µ m wavelength range (Wells et al. 2015; Labiano et al. 2016). The spectral windowis divided in four channels covered by four integral field units: (1) 4.96-7.77 µ m, (2) 7.71-11.90 µ m, (3) 11.90-18.35 µ m, and (4) 18.35-28.30 µ m.Two grating and dichroic wheels select the wavelength coverage within these four channels simultaneously, dividingeach channel into three spectral sub-bands indicated by A , B , and C respectively. To obtain a complete spectrum overthe whole MIRI band one has to combine exposures in the three spectral settings, A, B, and C. Since we are primarilyinterested in the 15 µ m CO feature, only one setting will be sufficient: 3 B covering the 13.2 − µ m range. Notethat the same setting will deliver the 1B (5.6 – 6.7 µ m), 2B (8.6 – 10.2 µ m) & 4B (20.4 – 24.7 µ m) wavelength rangesfor free. Of these, 2B is particularly interesting since it contains the ozone absorption feature. This is briefly discussedin Section 3.5.The IFS of Channel 3 consists of 16 slices (width = 0.39 ′′ ), each containing 26 pixels (0.24 ′′ ) providing a field ofview of ∼ ′′ × ′′ . The spectrum is dispersed over 1024 pixels (1 pix = 2.53 nm = 52 km sec − ) at a spectral resolvingpower of R ∼ − ).2.2. Modelling Proxima b and its atmosphere
Exoplanet Proxima b is found to orbit its host star in 11.186 +0 . − . days (Anglada-Escud´e et al. 2016). The amplitudeof its radial velocity variations corresponds to a minimum mass of 1.27 +0 . − . M Earth . If the mean density of the planetis the same as that of the Earth, and its orbit is nearly edge-on, it will have a radius of ∼ Earth . Proxima Cenhas an estimated mass of 0.123 ± Sun , implying an orbital semi-major axis of 0.0485 +0 . − . AU, correspondingto a maximum angular separation of 37.5 mas. Proxima Cen has an effective temperature of T Eff = 3042 ±
117 K,radius of 0.141 ± Sun , and bolometric luminosity of L=0.0017 L
Sun (Doyle & Butler 1990; S´egransan et al. 2003;Demory et al. 2009).Due to the close vicinity of the planet to its host star, it is generally assumed that Proxima b is tidally locked, meaningthat the same dayside hemisphere is eternally facing the star. The planet effective dayside temperature will stronglydepend on its Bond albedo and global circulation patterns. If due to atmospheric circulation the absorbed stellarenergy is homogeneously distributed over the planet, and the Bond albedo is similar to that of Earth (A B =0.306), thedayside equilibrium temperature of Proxima b is 235 K. If there is effectively no circulation and the absorbed stellarenergy is instantaneously reradiated, its observed dayside temperature could be as high as 300 K (and even up to 320K for a moon-like albedo) . In the other extreme case in which the planet has an albedo such as Venus (A B =0.9) witha very effective atmospheric circulation, its dayside effective temperature could be as low as 145 K. For the calculationsbelow we assume a continuum brightness temperature of 280 K at 15 µ m and a near-transiting orbital inclination,corresponding to a planet/star contrast ratio of 6 × − at superior conjunction.Simulated high-resolution emission spectra of Proxima b were generated by the Spectral Mapping AtmosphericRadiative Transfer (SMART) model assuming it to have an atmosphere such as Earth, using opacities from theLine-By-Line ABsorption Coefficient (LBLABC) tool (both developed by D. Crisp; see Meadows & Crisp 1996). TheHITRAN 2012 line database (Rothman et al. 2013) was used as input to LBLABC, which generates opacities at ultra-fine resolution (resolving each line with >
10 resolution elements within the half-width) on a grid of pressures and
Figure 2.
Planet model spectra (see Sect. 2.2) assuming a standard Earth atmospheric model for temperatures and gasmixing ratios, with on the right the assumed T/p profile. The upper panel shows the case where the stratospheric temperatureswere artificially made isothermal and equal to the tropopause temperature (isothermal stratosphere model), the middle panelshows a model for clear sky conditions (clear atmosphere model), and the lower panel shows a spectrum for a case with opaquehigh-altitude cirrus clouds (optically thick cirrus model). temperatures that spans a range relevant to Earth’s atmosphere. Following this, SMART – which has been extensivelyvalidated against moderate- to high-resolution observations of Earth (Robinson et al. 2011) – was used to simulatespectra at 5 × − cm − resolution (corresponding to R > at the simulated wavelengths).Our spectral simulations used a standard Earth atmospheric model for temperatures and gas mixing ratios(McClatchey et al. 1972), the spectra of which are shown in Figure 2. To bound certain extremes in thermal emission,model runs were performed for both clear sky conditions (the ‘clear atmosphere model’) and for an opaque high-altitudecirrus cloud (located at 0.2 bar, near the tropopause Muinonen et al. 1989), called the ‘optically-thick cirrus model’.Also, to explore a situation with large thermal contrast between the surface and stratosphere, a case where Earth’sstratospheric temperatures were artificially made isothermal and equal to the tropopause temperature (210 K) wassimulated (‘isothermal stratosphere model’). 2.3. Simulated observations
First, an estimate of the expected signal-to-noise (S/N) for Proxima Cen with MIRI is obtained from the betaversion of the JWST exposure time calculator . The 12 µ m and 22 µ m flux densities of Proxima have been deter-mined by the NASA Wide-field Infrared Survey Explorer (WISE) to be 924 mJy (m W3 = 3.838 ± W4 =3.688 ± µ m flux densities of 816, 713, and 630 mJy respectively. These fluxes are fed to the exposure time calculatorfor channel 3B of the MIRI MRS mode. A detector setup of 5 groups and fast readout gives an integration time of http://jwst.etc.stsci.edu χ interval is used to determine the statisticaluncertainties of a possible CO detection. These simulations were repeated for the three different models, and for arange star/planet contrasts corresponding to different orbital phases or different effective dayside temperatures. RESULTS AND DISCUSSION3.1.
Detectability
Our simulations show that the 15 µ m CO high-pass filtered signal of the Earth-mass planet can be detected within alimited amount of observing time. The MRS mode of MIRI at the JWST will in 24 hours integration time (excludingoverheads) deliver a R=1790–2640 spectrum of Proxima Cen between 13.2 and 15.8 µ m at a S/N of ∼ σ contrast limit of ∼ × − . While the high-frequency features in the filtered planet spectrum are typically at a 1–3 × − , there are about 100 within the targetedwavelength range - combining to a detection at a ∼ σ level. It means that while the continuum planet/star contrast isat a level of 6 × − ( ∼ σ per wavelength step), the combined spectrally filtered signal over the 3B band is about afactor 3–4 higher. The top-right panel of Fig. 3 shows the statistical confidence intervals for 5 ×
24 hrs of observationsif no CO signal is present, while the bottom-right panel shows the same, but then with the clear-atmosphere CO model spectrum for a face-on planet injected, indicating it can be detected at nearly 4 σ within this exposure time.Hence, while the individual CO features are not visible in the simulated spectrum, their combined signal can be clearlydetected. Results are very similar for the Isothermal Stratosphere model and the Optically-Thick Cirrus model.3.2. Important prerequisites
The stellar spectrum and its variability
We have made several assumptions that are vital for the high-pass spectral filtering technique to succeed in detectingCO in Proxima b. First, it is assumed that the high-frequency components of the spectrum of the host star itselfare perfectly known. Although low-resolution (R=600) mid-infrared spectra of M-dwarfs taken with the Spitzer SpaceTelescope (Mainzer et al. 2007) seem featureless, Phoenix model spectra (Allard et al. 2012) show that the 13.2-15.8 µ m wavelength region of an M5V dwarf star harbours thousands of H O lines, collectively resulting in wavelength-to-wavelength variations of ∼
2% in the MIRI MRS spectrum. It means that these features need to be calibrated to betterthan a relative precision of 1% for them not to interfere with the planet CO signal and not to act as an extra noisesource. The star is also expected to have numerous but much weaker lines from hot CO in this MRS band, resultingin wavelength-to-wavelength fluctuations of a few times 10 − - hence which need to be calibrated to a relative precisionof ∼ ≤ − . This means that a deep stellar spectrum may need to be obtained near orat inferior conjunction when the contribution from planet emission from the system is the smallest, which would limitthe sensitivity of this method for low orbital inclinations and would take as much time as the observations at superiorconjunction.Since the data at superior and inferior conjunction must be taken at least ∼ ≤ − on this time scale. We used archival data of Proxima from the UVES spectrograph https://phoenix.ens-lyon.fr/Grids/BT-Settl/CIFIST2011 2015/SPECTRA/ Figure 3.
From model spectrum to simulated MRS MIRI observations assuming the clear-atmosphere model. The top-leftpanel show the model spectrum convolved to the resolution of the MRS 3B channel of MIRI, normalised relative to the averagestar flux. The middle-left panel shows its associated high-pass spectral signal and the lower-left panel with noise added asif Proxima was observed for 5 ×
24 hrs. The top-right panel shows the statistical 1, 2, and 3 σ confidence intervals when noCO signal is present, pointing to a 3 σ upper limit of 50 ppm for the planet/star contrast. The bottom-right panel shows thesame, but with the simulated CO signal present - detected at ∼ σ . The y-axis indicates the mean planet/star contrastof the (non-spectrally filtered) template spectrum between 13.2 – 13.5 µ m. Note that these simulations assume that both theinstrumental spectral response and the stellar spectrum at a wavelength-to-wavelength scale have been determined, which islikely to require extra deep observations at inferior conjunction that serve as a reference (see Section 3.2). (R=100,000) at the Very Large Telescope separated by 4 days (October 10 & 14, 2009) to assess the optical variabilityof the star. For each of the two nights, a few dozen spectra were combined and the 868 – 878 nm wavelengthrange extracted which is dominated by hundreds of TiO lines but is free of telluric lines. The averaged spectrumwas subsequently convolved with a Gaussian to mimic the resolution of MIRI and subsequently binned to match itswavelength steps (in ∆ λ/λ ). After dividing out a linear trend with wavelength, the standard deviation of the ratio ofthe resulting spectra of the two nights is 4 × − . Since these data are possibly limited by flat fielding uncertainties,and variability in the mid-infrared is expected to be lower, this result is encouraging.3.2.2. Instrument calibration
Another prerequisite is that the spectral responses of the MRS pixels of MIRI can be adequately calibrated. Neitherabsolute flux calibration nor the low-frequency spectral response are important, but the sensitivity of one wavelengthrelative to the next is crucial - e.g. the spectral pixel-to-pixel calibration of the flat field. Potentially challenging isfringing, a common characteristic of infrared spectrometers. It is caused by interference at plane-parallel surfaces inthe light-path of the instrument. Experiences with data from ISO and Spitzer show that it can be removed down to thenoise level (e.g. Lahuis & van Dishoeck 2000). Wells et al. (2015) have characterised the fringing of the MIRI detectorsin the laboratory and identify three fringe components with scale lengths (in wave number) of 2.8, 0.37 and 10-100cm − , originating from the detector substrate, dichroic, and fringe beating respectively. The planet CO features alsoshow a regular pattern, but with a characteristic scale length of ∼ − , which fortunately is significantly differentfrom these fringe components. A potentially unwelcome source of error may be fringing in combination with dithering.Small residuals left over after defringing, combined from different dither positions, may be challenging to calibrate.In the signal-to-noise calculations presented above we assumed that the instrument calibration is perfect. For it notto add an extra source of noise, the wavelength-to-wavelength precision of the flat fielding and fringe removal must be ≤ − . If this level can be reached for individual IFS pixel, a tailored dithering strategy will subsequently push thecalibration noise to below 10% of the noise budget for a 24h observation. A single observation will have the star lightmostly distributed over 2 slices × √ × × ≈ Planet spectrum
In addition, temporal stellar atmospheric disturbances can modify the chemical composition of a planet atmosphere,meaning that flares could temporarily change the CO abundance of Proxima b. Such effects has been investigated byVenot et al. (2016), who find that although the abundances of some chemical species can be significantly altered deepin to the atmosphere ( ∼ is expected to only be affected very high in the atmosphere at < × − mbar- suggesting that the 15 µ m CO will hardly be affected.In principle, the observations could also be sensitive to other planets in the system. Since such planet would likely bein a significantly wider orbit and be colder, the expected signal would be smaller. The CO signal of such hypotheticalplanet could be distinguished from that of Proxima b since its superior conjunctions would occur on different epochs.3.3. Phase variations
A detection of CO will provide us with both the strength of the planet signal and its radial velocity. These can beused to constrain the orbital inclination of Proxima b. An example of such observation is shown in Fig. 4, showingin the left panel the expected variation in contrast and radial velocity and their uncertainties for 9 ×
24 hrs exposuresfor each three measurements at orbital phase φ =0.25, 0.5, and 0.75 - assuming an orbital inclination of near 90 o .These represent significantly longer exposures than that presented in Fig. 3. The right panel shows the same but foran inclination of i = 30 o , resulting in a smaller variation in contrast and radial velocity as function of phase. Theobservations at inferior conjunction may need to serve as a reference for the stellar spectrum (see Sect. 3.2.1). Inboth cases it is assumed that all thermal flux originated from the dayside hemisphere of the planet with an effectivetemperature of 280 K.Since the strength of the signal at φ =0.25 and 0.75 can be up to a factor two lower than that at φ = 0 .
5, calibrationof the instrumental response and the stellar spectrum will be even more important. If the orbital inclination is low, theplanet will never be seen entirely face-on, reducing the maximum signal at superior conjunction. However, it wouldalso mean that the mass of the planet is higher, meaning that the planet radius could be larger than assumed above(in particular if such more massive Proxima b is volatile rich), possibly counteracting the reduction in expected planetsurface brightness. 3.4.
Atmospheric characterisation
We performed our MRS MIRI simulations for three different atmospheric model spectra, 1) for a planet withan isothermal stratosphere, 2) for a planet with an inversion layer and clear atmosphere, and 3) with inversion layercombined with optically thick cirrus. Independent of which model we use, the increase in S/N over that expected for onewavelength step are very similar for all models at a factor 3–4. We also experimented by using the Earth transmissionspectrum instead, which is similar to the isothermal stratosphere spectrum, but with significant differential signal atthe heart of the CO band. This provides a S/N increase of a factor of ∼ τ = 1 surface. In the center of the strongest CO lines, where the opacityis greatest, we probe the atmosphere at the highest altitudes. Therefore, in the case of a strong thermal inversion(the clear atmosphere and the optically-thick cirrus models) the atmosphere will be warmer at such low pressure -resulting in emission lines in stead of absorption lines when the atmosphere is cooler at higher altitudes. This impliesthat a detection will also constrain the temperature structure of the upper atmosphere, giving additional insights inhigh-altitude atmospheric processes. It will not just merely be a detection of the planet atmosphere, which can becompared with theoretical models. E.g. Segura et al. (2005) argue that Earth-like planets orbiting M-dwarfs are likelyto have relatively cool, bordering on isothermal stratospheres – even with O present.Several features from other molecules are present in the 13.2 - 15.8 µ m wavelength range, such as C H at 13.7 µ mand HCN at 14 µ m, which may be included in the atmospheric spectral template if needed (and if expected to bepresent in the planet atmosphere). 3.5. Prospects of detecting ozone
When CO is targeted in the MRS 3B band, the 1B, 2B, and 4B bands are observed simultaneously. Interestingly, the2B band, ranging from 8.6 to 10.3 µ m covers the 9.6 µ m ozone band. While CO in the atmosphere of Proxima b maybe likely, if the planet did undergo ocean loss early it its history to generate a massive O atmosphere (Luger & Barnes2015), it is also possible that O , photochemically-produced from the O , is also present in higher abundances thanseen on Earth (Meadows et al. 2016). O is of course also of high interest as it can be used as a proxy for the O biosignature from a photosynthetic biosphere. Detection of O with JWST would therefore provide an intriguing firsthint that life might be present on an extrasolar planet, although O production via abiotic O from ocean loss wouldfirst have to be ruled out.The expected S/N per wavelength step in a 24 hour observation is 2 × , about a factor of two higher than in the3B band because of the high stellar flux. On the other hand, the expected continuum planet/star contrast is a factor ∼ µ m. Unfortunately, the individual lines within the 9.6 µ m band are moretightly packed than the lines in the 15 µ m CO band, i.e. the ozone band is not fully resolved at the MRS resolutionof R=2800 in this wavelength range. This means that if ozone is present in the atmosphere of Proxima b, it’s spectraldifferential signal will be about a factor 3–4 smaller than that of CO . We estimate that in the best case one couldexpect a 2 σ result in 20 days of JWST observing. We also note that the prerequisite for spectral calibration is alsomore stringent by a factor 2 compared to the CO case. Hence, although observations of ozone come for free when theCO band is targeted, it is unlikely this could result in a firm detection and is probably beyond the limit of what theJWST can achieve. 3.6. Strategy for proof of concept and other prospects
We envisage two ways to show proof of concept for the high-pass spectral filtering technique with the MRS mode ofMIRI. First, the method can be used on exoplanet targets with significantly higher planet/star contrasts. For example,a T=1000 K hot Jupiter orbiting a solar type star will have a 15 µ m contrast of 10 − , a factor 20 higher than Proximab. It means that for a host star whose 15 µ m flux is 40 times (4 magnitudes) fainter than Proxima b, CO will stillbe detected 10x faster. Also, one could aim for cool Neptunes or super-Earths orbiting nearby M-dwarfs, such asGliese 687b. If the spectrum of this particular planet exhibits a CO absorption feature it can be detected a factor 4faster than in the case of Proxima b. From a theoretical point of view it will be important to identify those planetsthat are expected to have CO in their atmospheres and select those with the most favourable stellar magnitudes andplanet/star contrasts.Ultimately, one should target Proxima itself, gradually increasing the integration time and validating at each stepthat the expected S/N limits are being reached. Detecting CO in the atmosphere of Proxima b will be a major stepforward in our quest for potential habitats and signs of extraterrestrial life.The detection of specific spectral features expected in a planet atmosphere becomes orders of magnitude morepowerful if the planet can also be angularly separated from the planet (e.g. Snellen et al. 2015). In such case thestellar spectrum can be effectively removed from all pixels in the IFU (since it is identical everywhere), after which theresidual spectra can be searched for the planet features. Hoeijmakers et al. (2017; in prep.) show that this technique isvery effective for the SINFONI and OSIRIS IFU spectrographs, located at the VLT and Keck Telescopes respectively,which have spectral resolving powers similar to that of MIRI (and the NIRSPEC IFU). If we could point the JWST0 !" & !" & ()*+,"-*.!*/!&0+
12 ’’ 31212 ’ ’2 31
Figure 4.
Examples of phase variation observations of the 15 µ m CO feature of Proxima b. The left shows the expectedvariation for an edge-on orbit, with the light blue regions indicating the 1 σ confidence interval for a 9 ×
24 hr observation ata given orbital phase (hence nearly 650 hrs of observations in total - significantly more than the simulations presented in Fig.3). The three dark-blue regions indicate the 1 σ confidence intervals for particular observations at an orbital phase of φ =0.5(superior conjunction), 0.25 and 0.75 - showing the variation in radial velocity and contrast. The right panel shows the samebut for an inclination of i = 30 o , resulting in a smaller variation in contrast and radial velocity as function of phase. In bothcases it is assumed that all thermal flux originated from the dayside hemisphere of the planet with an effective temperature of280 K. directly at α Centauri A using the MIRI MRS, only 1–1.5 ′′ away the starlight is at the level of that of Proxima implyingthat an Earth-size planet in the habitable zone of α Cen A could possibly be detected in 24 hrs. Unfortunately, α Centauri A will saturate the MIRI detectors within a small fraction of a second, irreversibly damaging the instrument.Possible ways to mitigate this issue need to be investigated.ACKNOWLEDGMENTSSnellen acknowledges funding from the European Research Council (ERC) under the European Union’s Horizon 2020research and innovation programme under grant agreement No 694513, and from research program VICI 639.043.107,which is financed by The Netherlands Organisation for Scientific Research (NWO). D´esert acknowledge funding fromthe European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme(grant agreement nr 679633; Exo-Atmos). Robinson and Birkby gratefully acknowledge support from the NationalAeronautics and Space Administration (NASA) through the Sagan Fellowship Program executed by the NASA Ex-oplanet Science Institute. Support for this work was provided in part by NASA through Hubble Fellowship grantHST-HF2-51336 awarded by the Space Telescope Science Institute, which is operated by the Association of Universi-ties for Research in Astronomy, Inc., for NASA, under contract NAS5-26555. Meadows and Robinson are members ofthe NASA Astrobiology Institute’s Virtual Planetary Laboratory Lead Team, supported by NASA under CooperativeAgreement No. NNA13AA93A. REFERENCES
Allard, F., Homeier, D., & Freytag, B. 2012, PhilosophicalTransactions of the Royal Society of London Series A,370, 2765Anglada-Escud´e, G., Amado, P. J., Barnes, J., et al. 2016,Nature, 536, 437 Barnes, R., Deitrick, R., Luger, R., et al. 2016, ArXive-prints, arXiv:1608.06919Berta, Z. K., Irwin, J., & Charbonneau, D. 2013, ApJ, 775,91Davenport, J. R. A., Kipping, D. M., Sasselov, D.,Matthews, J. M., & Cameron, C. 2016, ApJL, 829, L311