The Earth as an extrasolar transiting planet: Earth's atmospheric composition and thickness revealed by Lunar eclipse observations
Alfred Vidal-Madjar, Luc Arnold, David Ehrenreich, Roger Ferlet, Alain Lecavelier des Etangs, François Bouchy, Damien Segransan, Isabelle Boisse, Guillaume Hébrard, Claire Moutou, Jean-Michel Désert, David K. Sing, Rémy Cabanac, Christian Nitschelm, Xavier Bonfils, Xavier Delfosse, Morgan Desort, Rodrigo F. Díaz, Anne Eggenberger, Thierry Forveille, Anne-Marie Lagrange, Christophe Lovis, Francesco Pepe, Christian Perrier, Frédéric Pont, Nuno C. Santos, Stéphane Udry
aa r X i v : . [ a s t r o - ph . E P ] J un Astronomy&Astrophysicsmanuscript no. preprint June 10, 2018
The Earth as an extrasolar transiting planet
Earth’s atmospheric composition and thicknessrevealed by Lunar eclipse observations ⋆ A. Vidal–Madjar , L. Arnold , D. Ehrenreich , R. Ferlet , A. Lecavelier des Etangs , F. Bouchy , , D. Segransan ,I. Boisse , G. H´ebrard , C. Moutou , J.-M. D´esert , , D. K. Sing , , R. Cabanac , C. Nitschelm , X. Bonfils ,X. Delfosse , M. Desort , R. F. Diaz , A. Eggenberger , T. Forveille , A.-M. Lagrange , C. Lovis , F. Pepe ,C. Perrier , F. Pont , N. C. Santos , , and S. Udry Institut d’Astrophysique de Paris, UMR7095 CNRS, Universit´e Pierre & Marie Curie, 98bis, boulevard Arago, 75014 Paris, France,e-mail: [email protected] Observatoire de Haute-Provence, CNRS / OAMP, 04870 Saint-Michel-l’Observatoire, France Laboratoire d’Astrophysique de Grenoble, Universit´e Joseph Fourier, CNRS (UMR 5571), BP 53, 38041 Grenoble cedex 9, France Observatoire de Gen`eve, Universit´e de Gen`eve, 51 Chemin des Maillettes, 1290 Sauverny, Switzerland Laboratoire d’Astrophysique de Marseille, Universit´e de Provence, CNRS (UMR6110), BP 8, Technopˆole Marseille ´Etoile, 13376Marseille Cedex 12, France Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, Massachusetts 02138 USA School of Physics, University of Exeter, Exeter, EX4 4QL, UK Observatoire Midi-Pyr´en´ees, TBL, 57 Ave d’Azereix, 65000 Tarbes, France Instituto de Astronom´ıa, Universidad Cat´olica del Norte, Avenida Angamos 0610, Antofagasta, Chile Centro de Astrofisica, Universidade do Porto, Rua das Estrelas, 4150-762 Porto, Portugal
Abstract
Context.
An important goal within the quest for detecting an Earth-like extrasolar planet, will be to identify atmospheric gaseousbio-signatures.
Aims.
Observations of the light transmitted through the Earth’s atmosphere, as for an extrasolar planet, will be the first important stepfor future comparisons. We have completed observations of the Earth during a lunar eclipse, a unique situation similar to that of atransiting planet. We aim at showing what species could be detected in its atmosphere at optical wavelengths, where a lot of photonsare available in the masked stellar light.
Methods.
We present observations of the 2008 August 16 Moon eclipse performed with the SOPHIE spectrograph at the Observatoirede Haute-Provence (France). Locating the spectrograph’s fibers in the penumbra of the eclipse, the Moon irradiance is then a mix ofdirect, unabsorbed Sun light and solar light that has passed through the Earth’s atmosphere. This mixture essentially reproduces whatis recorded during the transit of an extrasolar planet.
Results.
We report here the clear detection of several Earth atmospheric compounds in the transmission spectra, such as ozone,molecular oxygen, and neutral sodium as well as molecular nitrogen and oxygen through the Rayleigh signature. Moreover, wepresent a method that allows us to derive the thickness of the atmosphere versus the wavelength for penumbra eclipse observations.We quantitatively evaluate the altitude at which the atmosphere becomes transparent for important species like molecular oxygen andozone, two species thought to be tightly linked to the presence of life.
Conclusions.
The molecular detections presented here are an encouraging first attempt, necessary to better prepare for the futureof extremely-large telescopes and transiting Earth-like planets. Instruments like SOPHIE will be mandatory when characterizing theatmospheres of transiting Earth-like planets from the ground and searching for bio-marker signatures.
Key words.
Planets and planetary systems - Eclipses - Earth - Planets and satellites: atmospheres - Astrobiology - Techniques:spectroscopic - Methods: observational
1. Introduction
Soon the
CoRoT (a space program operated by the French SpaceAgency, CNES) and
Kepler (a NASA spacecraft searching forhabitable planets) missions may discover a transiting Earth-likeplanet. It will then be of prime importance to study such an ex-oplanet’s atmospheric composition, in order to define its evo- ⋆ Detailed observations as shown in Figs. 9, 10, 11 and12 are only available in electronic form at the CDS viaanonymous ftp to cdsarc.u-strasbg.fr (130.79.128.5) or via http://cdsweb.u-strasbg.fr/cgi-bin/qcat?J/A+A/ lutionary status as compared to the known Earth one. Throughthese observations the impact of galactic environmental pertur-bations as, e.g. the passing through a dense interstellar cloud(Vidal-Madjar et al. 1978), may be directly evaluated. However,the ultimate goal of these studies will be to know if life couldhave emerged elsewhere as it did once on Earth.It has been demonstrated, as for the exoplanet HD209458b,that it is possible to detect the presence of some speciesin the planetary atmosphere, like sodium, hydrogen, oxy-gen, and carbon (Charbonneau et al. 2002, Vidal-Madjar et al.2003, 2004). More recently, additional detections have been re-
Vidal–Madjar et al.: The Earth as a transiting planet ported, including HI from recombination via the Balmer jump(Ballester et al. 2007), H the main gaseous content via Rayleighscattering (Sing et al. 2008a, 2008b, Lecavelier des Etangs et al.2008b), and TiO / VO (D´esert et al. 2008). The possible signa-ture from H O in HD209458b (Barman 2007) is now stronglyquestioned by new
HST (the Hubble Space Telescope) and
Spitzer (a Space Telescope, studying the universe in infrared)observations, revealing on the contrary the presence of high-altitude haze as well as CO in the atmosphere of HD189733b(Ehrenreich et al. 2007, Lecavelier des Etangs et al. 2008a, Singet al. 2009, D´esert et al. 2009). New transiting planets are dis-covered from space by
CoRot , launched at the end of 2006, andby
Kepler launched in March 2009. More specifically, space ob-servations are reaching high enough accuracy to enable the de-tection of Earth-size transiting planets (Bord´e et al. 2003, Rouanet al. 2009, L´eger et al. 2009, Queloz et al. 2009), some of whichare possibly “ocean-planets” (L´eger et al. 2004) or even similarto the telluric planets of our Solar System, at distances compat-ible with a so-called habitable zone. Obviously, the characteri-zation of the corresponding atmospheres would be an excitingachievement.Knowing that the probability to have a transiting Earth atabout one AU from its star is of the order of 0.5%, and thatthe number of available planets rises as the cube of their dis-tance from the Sun, Ehrenreich et al. (2006) have shown thatthe number of available transiting planets within 60 pc is simi-lar to the number of possible targets available if one looks at theplanetary emissions within a sphere of about 10 pc around theSun (the current expectations of projects like
TPF-I (the NASA,Terrestrial Planet Finder via Interferometry) or
Darwin (the ESAequivalent program). However, the present di ffi culty of the tran-siting approach is the lack of photons ( i.e. high S / N) necessary toanalyze spectra of Earth-like planets when using 2-m class tele-scopes. Lecavelier des Etangs & Ehrenreich (2005) have shownthat many hot Jupiters, hot Neptunes, ocean planets, and otherlow-density planets will be observable with larger ground basedtelescopes, like the future European Extremely-Large Telescope(E-ELT).Transmission spectroscopy has already proved its po-tential to probe atmospheres of extrasolar planets. It could verywell be the first method which will give access to the atmosphereof smaller and cooler transiting planets, hopefully Earth-like.Observing in the visible and near UV o ff ers a great ad-vantage, because of the strong opacity of some specific nar-row line species in that spectral range, and because of the of-ten more intense stellar flux at these wavelengths. During atransit event, only the upper atmosphere of the planet maybe seen, thus reducing the many perturbing e ff ects potentiallypresent when observing at lower resolution, i.e. at lower alti-tudes, like the presence of clouds, continents, oceans etc. asshown in Arnold et al. 2002, Woolf et al. 2002, Arnold 2008 andreferences therein. Indeed, in the core of the spectral lines wewill sample only the upper atmosphere, where one can expectthat the mixing of di ff erent species has already taken place, thusrevealing an “average” planetary atmosphere representative ofoverall signatures.Following the pioneering work of Sagan et al. (1993) search-ing for “life on Earth” from remote sensing, our purpose isto observe the Earth as a transiting planet in order to bet-ter test the present models of Earth-like planets and thus getready to properly analyze future observations of real transit-ing Earth-like extrasolar planets. Using the Moon as a reflec-tive surface during total or partial eclipses provides the neededsituation which furthermore can be exploited from the ground.We have conducted a first attempt during the 2008 August 16 lunar eclipse with the SOPHIE high-resolution spectrograph(Perruchot et al. 2008, Bouchy et al. 2009) at the 1.93 cm tele-scope of the OHP (Observatoire de Haute Provence) observatorycovering the 3 900 to 7 000 Å spectral range at a resolving powerof R ∼
75 000. The choice of a high-resolution spectrograph al-lows us to better identify absorbing species via their detailedspectral signatures.Other observations of this eclipse have been conducted(Pall´e et al. 2009) over a wider wavelength range (0.36 - 2.4 µ m ),but at a lower spectral resolution (R <
2. Observations
The observations were conducted with the cross-dispersed,environmentally stabilized echelle spectrograph SOPHIE,dedicated to high-precision radial velocity measurements(Perruchot et al. 2008, Bouchy et al. 2009). They were securedin high-resolution mode, i.e. the spectrograph was fed by a40 µ m slit and an optical scrambler located at the output of theoptical fiber, allowing a resolving power λ/ ∆ λ =
75 000.Two optical fibers, 3 arcsec wide, separated by 1.86 arcminwere used. Both were placed on the Moon and aligned along theeast-west direction.The spectra were extracted from the detector images with theSOPHIE pipeline, which includes localization of the orders onthe 2D-images, optimal order extraction, cosmic-ray rejection,wavelength calibration, corrections of flat-field and charge trans-fer ine ffi ciency at low signal-to-noise ratio (Bouchy et al. 2009). Table 1 lists the three main spectra used in our analysis alongwith their positions over the Lunar disk as shown in Fig. 1. Onlyone observation during the eclipse was useful, taken at 20h44 UTin the penumbra, close to the Earth shadow. Figure 1 showswhere the fibers were positioned, i.e. over high-albedo Imbrium-type or pre-Imbrium-type terrains (for more information see theURL at the end of the reference list), which maximized the fluxduring the eclipse. The two other observations listed in Table 1are two among a total of 13 calibration spectra recorded after theeclipse (full Moon outside the penumbra) at similar air masses( sec ( z ) in Table 1 where z is the zenith angle) and over the twodi ff erent regions pointed to during the eclipse. Unfortunately, af-ter the eclipse the guiding camera was saturated and an accurateposition of the fibers could not be checked and adjusted visuallyas during the eclipse. Nevertheless the telescope coordinates in-dicate that the fibers were positioned over the same kind of lunarterrains. We collected only sparse observations, because clouds appearedjust after the first exposure taken at 20h44 UT. These clouds dis-appeared for a few minutes at 22h44 UT, just before the umbraleft the Moon, though these observations are more noisy and arestill possibly a ff ected by intervening clouds. idal–Madjar et al.: The Earth as a transiting planet 3 Table 1.
Fibers positions on the Moon during and after the eclipse. Times are for mid-exposure.Date; Fiber A location Fiber B location Exposure BERV Air massTime (UT) longitude; latitude longitude; latitude time (s) (km / s) sec ( z )2008–08–16; 20:44:21 E 52 . ◦ ; N 9 . ◦ E 68 . ◦ ; N 12 . ◦ ◦ ; S 51 ◦ E 50 ◦ ; S 48 ◦ ◦ ; N 20 ◦ E 64 ◦ ; N 22 ◦ Figure 1.
Fibers A and B on the Moon in penumbra. The umbrais the dark area at the left (upper figure at 20h44 UT) and at right(lower figure at 22h44 UT). The distance between the fibers is1.86 arcmin, east-west oriented.The observations completed at 20h44 UT were done withfibers A and B at about 2 and 4 arcmin from the penumbra / umbralimit respectively.Spectra were also taken later in the night after the eclipse,when the clouds moved away. In particular at 02h56 UT spectrawere taken over the same location on the Moon and at the sameair mass as those at 20h44 UT. The spectrum taken at 00h12 UT,although pointed over another region of the Moon, is also listedbecause it was taken with a shorter time gap from the 20h44 UTeclipse observation, leading to a smaller di ff erential BERV cor-rection (“Barycentric Earth Radial Velocity” correction of theEarth’s motion along its orbit to evaluate the spectral observa-tions in heliocentric coordinates as automatically done throughthe SOPHIE pipeline). The corresponding BERV shifts are alsolisted in Table 1.The calibration spectra (taken later in the night after theeclipse) are used to correct the eclipse spectra from the direct at- mospheric absorption along the line of sight between the Moonand the OHP Observatory and to extract the altitude at which theEarth atmosphere becomes transparent enough for a grazing lineof sight.
3. Data analysis
Because we are searching for atmospheric signatures in the gath-ered spectra, we have first to recalculate the wavelengths in thegeocentric reference frame. We know that there all atmosphericsignatures will be present at identical wavelengths, while the so-lar line signatures will be slightly displaced by an amount re-lated to the spectral shift induced by the BERV. Note howeverthat at the SOPHIE resolution of about ∼ / s, these shifts ofless than 0.5km / s are likely insignificant. The possible impact ofthe BERV e ff ect on our data is assessed in Sect. 5.Our observations present two types of Earth atmospheric ab-sorption signatures : i) the “vertical” ones produced along thepath length between the Moon and the telescope and ii) the “hor-izontal” ones due to the solar light grazing the Earth atmospherebefore reaching the Moon. We are only interested in the latterand will therefore have to correct the data from the “vertical”signatures.Finally, knowing that the SOPHIE pipeline is optimized forhigh radial velocity precisions and not for flux evaluations, wecan expect some di ffi culties related to flux estimates, i.e. echelleorder corrections as well as precise zero level evaluations. The “vertical” atmospheric transmission is a function that de-pends upon the air mass AM , noted T AM ( λ ). Along the line ofsight, an initial spectrum I ( λ ) reaching the top of the atmosphereis transformed into an observed spectrum O AM ( λ ) : O AM ( λ ) = I ( λ ) × T AM ( λ ) . (1)For a plane parallel atmosphere (an approximation valid to airmasses smaller than ∼
5) the air mass AM is equal to a = sec ( z )and the atmospheric transmission function is then simply T a ( λ ),the transmission function T at the power a (Bird et al. 1982),where T is the transmission in the vertical direction (air massof 1).Indeed for an air mass a equal to the sum of two air masses a and a , one can evaluate the observed spectrum in the followingmanner : O a + a = I × T a + a = ( I × T a ) × T a , (2)showing that the transmission function T ( λ ) follows the condi-tions : T a + a ( λ ) = T a ( λ ) × T a ( λ ) (3)and T ( λ ) = . (4) Vidal–Madjar et al.: The Earth as a transiting planet
If the transmission function T ( λ ) is known at a given airmass, it can be evaluated at any other air mass. Noting that O a ( λ ) O a ( λ ) = T a ( λ ) T a ( λ ) = T a − a ( λ ) , (5)we were able to evaluate directly the transmission function atdi ff erent air masses through the complete series of spectra gath-ered during the full Moon observations. We normalized all eval-uations extracted from observational ratios to an air mass equalto 1 by calculating O a ( λ ) O a ( λ ) ! / ( a − a ) = ( T a − a ( λ )) / ( a − a ) = T ( λ ) . (6)From the 13 full Moon observations O a ( λ ), we extracted 24couples well separated in air mass each containing independentevaluations of the transmission function T ( λ ). In Fig. 2, the av-erage of all evaluations are shown, providing the reference trans-mission function T ( λ ) that we will now use in the followingdata analysis to re-evaluate all observed eclipse spectra as if theywere observed from outside the atmosphere. Figure 2.
Average atmospheric transmission function T ( λ ) asevaluated during full Moon just after the eclipse supposed to besimilar to the one during the eclipse observations. The dashedline shows a transmission model (Hayes and Latham 1975),which includes both Rayleigh di ff usion and aerosols (opticaldepth of 0.035 at 5 320 Å typical for clear OHP nights). Theexperimental evaluation was normalized in order to match themodel at our reference wavelength λ = as well as H O absorption bands are clearly seen. Note thatthe match with the model is good except for the central regionfrom 5 000 to 6 700 Å, where a clear additional absorption is de-tected. This is the Chappuis band from ozone (noted O ), notincluded in the model calculation. One can note in Fig. 2 the SOPHIE echelle order signatures asregular wiggles over ∼
80Å from 4000 to 4500Å, where they areparticularly visible. Those instrumental signatures should not beseen in an atmospheric transmission function. These will cer-tainly be a perturbation in our data analysis, but for the time
Figure 3.
At 20h44 UT, the (Fiber B / Fiber A) ratio (solid line)is shown in the NaI doublet spectral region. The Fiber A spec-trum (dashed line) with its corresponding zero level (horizontaldashed line) as extracted from the SOPHIE pipeline as well asan H O model (dotted line) are overplotted (in arbitrary scalessimply to fit within the figure frame) to show the positions of thesolar as well as H O spectral lines in the spectrum. Most of theobserved lines are due to atmospheric water vapor. They disap-pear in the fiber’s spectral ratio, while the broad spectral wingsof the two NaI solar lines are still clearly visible.being we will try to analyze the data without a modification tothe SOPHIE pipeline. We will however evaluate the impact ofthis approximation a posteriori by looking how these order sig-natures are a ff ecting the results (see Sect. 5).In Fig. 3, both the direct Fiber A spectrum observed dur-ing the eclipse as well as the Fiber B / Fiber A flux ratio areshown close to the NaI doublet solar lines. Note that the numer-ous spectral lines, mostly due to H O in the Earth atmosphere,completely disappear from the Fiber B / Fiber A ratio. This is notsurprising because both Fibers A and B are observed through thesame atmospheric (and thus air mass) layer.Although most of the weaker solar lines are also erased inthis ratio (as it can be directly seen in Fig. 2, which also showsspectral ratios, see Eq. 6), for strong and deep solar lines like e.g. the NaI lines, as shown in Fig. 3, the signature of their broadspectral wings is still visible in the Fiber B / Fiber A ratio. Thisshould not be the case if Fiber A and B have simply a di ff erenttransmission coe ffi cients, because they are both simultaneouslyobserving (almost) the same solar spectrum.The NaI lines (reaching close to the zero level, see Fig. 3)are more sensitive to any unaccounted for, even small, zero levelshift. This is not the case for the weaker solar lines.Knowing that a zero level shift is present in the extractedspectra, we search below for some plausible explanations. Over the SOPHIE CCDs, the lowest counts are not equal to zero(even where no counts should be registered), as for instance atthe shortest wavelengths, where the instrument sensitivity dropssharply. For this reason, we assumed the possibility of a zeroo ff set in the SOPHIE extracted counts. We thus subtracted avalue from the SOPHIE spectra to make the broad NaI solar lineswings disappear (as they should). This correction is compatiblewith the data, as reported in Table 2. idal–Madjar et al.: The Earth as a transiting planet 5 Table 2.
SOPHIE fluxes as extracted from the data sets, both atthe minimum of deep absorption signatures and where the flux ismaximum. Times are for mid-exposures. Flux units are arbitrarybut comparable in relative terms because they correspond to theraw data (CCD ADU) corrected for by the same pipe line.Fiber ; Min. flux Min. flux Max.time (UT) in NaI in O fluxA ; 20:44 0.0007 0.0005 0.0205B ; 20:44 0.004 0.002 0.106A ; 02:56 0.0018 0.0010 0.0514B ; 02:56 0.0016 0.0008 0.0471We tried three zero shift level scenarios :1. we assumed that stray light due to the nearby full Moon ispresent. However, subtracting various proportions of the fullmoon spectrum to the other spectra did not induce any atten-uation of the NaI solar wings and on the contrary producedmany additional solar lines;2. we supposed that a constant value has to be subtracted fromall data, proportional to the total flux gathered in each spec-trum. Again no satisfactory solution was found;3. we assumed that a unique constant value had to be used forall spectra, a correction only related to the detector and in-dependent of the total flux. Then we found that by subtract-ing 0.0003 ± absorption lines near 6 900 Å,it is clear that this correction could not be larger than the min-imum flux seen at the bottom of the absorption lines, and thushas to be somewhere in the 0 to 0.0005 range, which we foundto be the case. A posteriori, we will also see that the final resultsare all compatible with this correction.After empirically evaluating the type of correction needed,we found that another possible explanation of atmospheric originmay be given and we develop it in the next section. Indeed, following a suggestion of our referee, we recalled thatsome of the sunlight going through the Earth atmosphere is re-fracted in the lower layers of the atmosphere, producing for anobserver on the Moon within the Earth umbra, an emission ringall along the Earth limb.This ring e ff ect can contribute here in two ways. First, it canadd a constant intensity (a few percent shift of the zero level)to sunlight as seen in scattered light from the sky, to the ex-tent that the penumbral light contains not just transmitted light,but also forward-scattered light ; this should slightly dilute thepenumbral spectrum and introduce a zero level correction of type (2). Second, the light reflected from the lunar surface itself willcontain the ring e ff ect, again a few percent additive backgroundspectrum, a correction of type (2.) or (3.).The ring relative intensity and spectral signature were al-ready observed during Lunar eclipses through the study of thelunar light reflected from the umbral regions, only lit by theEarth ring. Danjon (1936) very early understood that the umbralluminosity was related to the transparency of the Earth lower at-mosphere and after observing several lunar eclipses, he proposeda related five level scale for the lunar umbral luminosity.Observation revealed (as in e.g. Hernitschek et al. 2008) thata 5 magnitudes drop exists at the penumbra-umbra (PU) tran-sition, showing that the ring contribution to the penumbral fluxis on the order of 1 %. Furthermore, the ring contribution de-creases when moving away from the PU transition deeper withinthe umbra, as the expected signature of forward-scattered light.The ring contribution within the penumbral region is indeed pro-portional to the length of the Earth’s limb ( L E ) just in front of thesolar disk appearing on top of the Earth limb.Thus, when moving in the penumbral domain away fromthe PU transition, the direct solar contribution increases likethe surface of the solar disk above the Earth limb ( ∼ L E , seeFig. 4), while the ring contribution increases as the length of theEarth’s limb covered by the solar disk ( ∼ L E ). This means thatfor our observations, the relative ring e ff ect contribution shouldbe stronger in Fiber A than in Fiber B according to the geome-try of our 20h44 UT observation, for which Fiber A, presentinga lower signal, is closer to the PU transition than Fiber B. Thering e ff ect contribution in the reflected Lunar light seems thus tobetter correspond to our empirical correction scenario (3.).This is confirmed in the Gedzelman & Vollmer (2008) sim-ulation of irradiance of lunar eclipses. They furthermore showthat the ring contribution at the PU transition should present analmost flat spectral signature from 4000 to 7000Å (see their Fig.1), i.e. over the whole SOPHIE spectral range. This was directlyobserved during the 2008 August 16 lunar eclipse by Pall´e etal (2009, see top of their Fig. S3 of the supplementary section),who detected only weak spectral signatures deeper in the umbraas also predicted by Gedzelman & Vollmer (2008).In summary the ring contribution should be on the order of1 % of the signal in Fiber A, and be at a relatively lower levelin Fiber B with no wavelength dependence. This is almost ex-actly what we have found from our empirical study (3.), whichpresents the same 0.0003 zero level shift in both Fibers, i.e. in-deed on the order of 1 % in Fiber A and relatively less in Fiber B(see Table 2). Additional discussions in Sect. 5 will show thatthis correction is on the same order near the NaI doublet lines at5 900 Å and near the molecular O lines at 6 900 Å and thus isindeed weakly variable with wavelength.The ring e ff ect could also explain why the ratios (EclipseA) / (Full Moon A), (or B) could also be a ff ected, ratios that weare going to use and describe in the following sections.Both these instrumental and ring e ff ects are probably presentin our study and need to be corrected for empirically. In futurestudies we will also try to observe umbral spectra in order tohave a more quantitative evaluation of the ring e ff ect and thusinclude its contribution in our analysis.
4. Quantitative analysis
We present in this section the approach developed to extract aquantitative information from the observed spectra related to thealtitude at which the Earth atmosphere becomes transparent as a
Vidal–Madjar et al.: The Earth as a transiting planet function of wavelength, a result very similar to the one obtainedfor a transiting planet.
These unabsorbed spectra are simply recomputed by dividing theobserved spectra O a ( λ ) by the transmission function T ( λ ) at thepower of the precise air mass of the corresponding observation : I ( λ ) = O a ( λ ) / T a ( λ ) . (7)All these calculated corrected spectra are those an observershould have obtained from above the Earth atmosphere. We willnote these spectra E ( λ ) and F ( λ ), corresponding to the “eclipse”and “full Moon” corrected spectra respectively. Figure 4.
Geometry of a crescent of Sun, S , above the Earth’sreference limb (O S and R S are the center and radius of the Sun,O E and R E those of the Earth). The scene (not to scale) is ob-served from the Moon at the location where are positioned fibersA or B within the penumbra. From the observational knowledgeof S A or S B as extracted from Eq. 11 (see text), the entire de-fined geometry of the observations allows a direct evaluation ofthe corresponding arc lengths, L A or L B .From the deep penumbra, where the spectrograph’s fibers arelocated during the observation, a crescent of Sun is seen abovethe Earth’s limb. The Moon irradiance is thus a mix of directSun light (i.e. una ff ected by the Earth’s atmosphere), and solarlight that passed through the atmosphere, mostly along an arc L where the solar disk intersects the Earth’s limb (Fig. 4). This mixof direct and absorbed solar light essentially reproduces what isrecorded during a transit. Let us note S ⊙ as the surface of the fullsolar disk, S the fraction of solar surface visible above Earth’sgeometrical limb during the eclipse, and h is an equivalent heightover which the atmosphere can be considered opaque, at each Figure 5. L = f ( S ) function computed here for the geometryof the eclipse: S is the surface of the solar crescent above theEarth’s reference limb - as observed from the Moon at the loca-tion of either fiber A or B within the penumbra. From there, thediameters of the Earth and the Sun are 6 768.0 and 1 890.6 arcsecrespectively. This geometry implies the relation shown betweenthe length of the Earth’s limb covered by the Sun L , and thesurface of the solar crescent, normalized to the total solar disksurface S ⊙ as seen from the Moon. Since S A / S ⊙ or S B / S ⊙ areknown from observations, the corresponding arcs lengths L A or L B are also known, and thus h ( λ ) evaluated.wavelength, from the point of view of the geometrical cross sec-tion.The total column density along a grazing line of sight pass-ing through the terminator at di ff erent altitudes could be esti-mated following the Fortney et al. (2005) formulation: the opti-cal depth, τ , in a line of sight grazing the Earth limb at an altitude h is given by τ ( λ, h ) ≈ σ ( λ ) n ( h ) p π R E H , (8)where R E is the Earth radius, H the atmosphere scale height,and n ( h ) = n ( h = exp ( − h / H ) the volume density at the altitude h of the main absorbent with a cross section σ ( λ ). The scale heightis given by the relation H = kT /µ g , where k is the Boltzmannconstant, T the temperature, µ the mean mass of atmosphericmolecules times the mass of the proton, and g the gravity at theEarth radius, g = M E G / R E with M E equal to the Earth mass andG the gravitational constant.For any wavelength, the line of sight becomes opaque atan e ff ective altitude h for an e ff ective optical thickness of τ e f f = ff ective altitude h as a function of wavelength to model calculations of e ff ectivealtitudes.The e ff ective altitude of the atmosphere at a wavelength λ is calculated by solving the equation τ ( λ, h ) = τ e f f . Using thequantities defined above, the e ff ective altitude h is given by Eq.9. h ( λ ) = H ln (cid:16) ( n ( h = σ ( λ ) /τ e f f ) p π R E / kT µ g (cid:17) . (9)In principle, a simple relation links the two E ( λ ) and F ( λ )corrected spectra due to the geometry of the problem E ( λ ) = F ( λ ) × S − L × h ( λ ) S ⊙ . (10) idal–Madjar et al.: The Earth as a transiting planet 7 h ( λ ) ,theeffective heightof the Earthatmosphere. Figure 6.
Absorbing atmosphere thickness versus wavelength,evaluated according to Eq. 14 for the 20h44 UT eclipse obser-vation associated to the 02h56 UT full Moon ones. The refer-ence altitude equal to 0 km (dashed line) has been chosen for λ = h ( λ ) = λ . Then, S is more precisely the fraction of solarsurface visible above the Earth’s limb corresponding to our ref-erence altitude h ( λ ). It follows that Eq. 10 reduces to E ( λ ) = F ( λ ) × SS ⊙ . (11)The S / S ⊙ ratio can be geometrically calculated for a givenobservation time, but may also be directly estimated from the E ( λ ) and F ( λ ) values of the flux measured during and outsidethe eclipse at λ . We decided to take the reference wavelengthover a 20 Å range centered at λ = h ( λ ) into “real” altitudes, above sea level. Indeedall our altitude evaluations h ( λ ) relative to h ( λ ), could be posi-tive or negative depending on more or less absorption at λ thanat λ .The ratio of Eqs. 10 and 11 gives for Fiber A E A ( λ ) E A ( λ ) = F A ( λ ) F A ( λ ) × [1 − L A S A × h ( λ )] , (12)and for Fiber B E B ( λ ) E B ( λ ) = F B ( λ ) F B ( λ ) × [1 − L B S B × h ( λ )] , (13) Figure 7.
Same as Fig. 6, but here the h ( λ ) variations are binnedover 200 pixels ( ∼ ff erencewas not visible. Here we can note that both possible perturba-tions related to either the BERV shift as well as the air masscorrection are indeed negligible (see text). On the contrary somebroadband variations could be seen, probably linked to albedovariations of the di ff erent lunar regions where the fibers werepointed (see Table 1). In the blue region (4 000 to 5 000 Å), thefringes are due to an imperfect correction of the SOPHIE ordersand not to classical CCD interferential fringes (see text).in which L A and L B are the lengths of the arc of limb at the h ( λ ) level as seen from the locations of Fiber A and Fiber Brespectively, as projected on the Moon, and similarly S A and S B are the corresponding surfaces of the Sun above these Earth’slimbs.The di ff erence between the ratios for Fibers A and B, (Eq. 12– Eq. 13), is E A ( λ ) E A ( λ ) × F A ( λ ) F A ( λ ) − E B ( λ ) E B ( λ ) × F B ( λ ) F B ( λ ) ! == L B S B − L A S A ! × h ( λ ) . (14)Therefore, the measurements (left side of Eq. 14) are propor-tional to h .However, to fully solve the problem, one needs to evaluatethe factor of proportionality in front of h . We know the rela-tive angular diameters of both the Earth (above the Earth’s ref-erence limb) and the Sun, as seen from the Moon at the time ofthe observations, which unambiguously defines (geometrically)how far above that selected Earth’s limb reference the solar discemerges. From a simple geometrical calculation (see Fig. 4), onecan evaluate the corresponding arc lengths L A and L B , which aredirectly derived from the relation L = f ( S ) that we have calcu-lated for the time of the observations, as shown in Fig. 5. Finally,the direct measurements of S A and S B are extracted from the ob-servations through Eq. 11.For example, from our best observations made on August 16at 20h44 UT during the eclipse, compared to the full Moon ob-servations made on August 17 at 02h56 UT, we find the follow-ing corrected flux ratios in the 4 520 - 4 540 Å reference spectralband : Vidal–Madjar et al.: The Earth as a transiting planet E A ( λ ) / F A ( λ ) = S A / S ⊙ as seenfrom the Moon where Fiber A is located; this leads to L A =
569 arcsec; E B ( λ ) / F B ( λ ) = S B / S ⊙ as seenfrom the Moon where Fiber B is located; this leads to L B = h ( λ ),either in arcsec seen from the Moon or in km above (or below)the Earth’s reference limb at h ( λ ).The extracted values of h ( λ ) are shown in Fig. 6, obtainedwhen using the 20h44 UT eclipse observations compared to the02h56 UT full Moon ones at about the same air mass and overthe same regions on the Moon. The extracted altitudes are rela-tive to the reference altitude h ( λ ), here equal to 0. As expected,the evaluations can be either positive (higher altitudes) or nega-tive (lower altitudes).To check the validity of several of our assumptions we calcu-lated the same h ( λ ) function extracted again from the 20h44 UTeclipse observations compared to both full Moon observations,the one at 02h56 UT as well as the 00h12 UT one made at dif-ferent air masses, BERV and over distant regions on the Moon(see Table 1). This allows us to check that our evaluations are notsensitive to air masses or BERV variations (see Fig. 7), becauseboth extracted h ( λ ) functions are very similar in all their details.Although nearly identical (non discernable at full spectral reso-lution as in Fig. 6), the result indicates however that some broadband variation is present. It is not caused by an improper airmass correction because the produced e ff ect cancels in two spec-tral regions: the first one at 4 530 Å (by definition, since this isour reference wavelength) and again around 6 300 Å. That theobserved relative variation of the transmission remains small inthe blue and more important in the red is indeed in contradictionwith an air mass correction, which has to be stronger in the blue(see Fig. 2). We thus conclude that the observed relative varia-tion is more probably due to lunar albedo changes over di ff erentlunar regions.This broadband e ff ect could however be the cause of addi-tional systematic errors in the altitude evaluations. A simple visual analysis of Figs. 6 and 7 gives some idea of theerrors estimated for the h ( λ ) evaluations. First in Fig. 6, the er-ror due to the photon noise in the observations is ∼ ± ∼ ± ∼ ± ff erent regions observed that an additional systematic errordue to lunar albedo variations could be on the order of ∼ ± h ( λ )function we cannot produce absolute altitudes to better than ap-proximately ∼ ±
5. Discussion
The information extracted from the presented spectra is h ( λ ),thus we are exactly in the situation of an extrasolar planetarytransit. In e ff ect, a spectrum recorded during a transit is not thespectrum of the absorbing species, but only the spectral sig-natures (in terms of altitude) of the highest absorbing specieswithin the observed atmosphere. For instance, in the central re-gion of the spectrum (Fig. 6) the main feature is the Chappuisband of ozone, corresponding to altitudes known to be in the30 km range, other detectable species are only those able to ef-ficiently absorb above. As we will see, oxygen and sodium arealso detected with the present observations, but not H O.In order to properly interpret our observations, we have touse model calculations to estimate the altitude at which the Earthatmosphere becomes transparent, as a function of wavelengthand for a grazing line of sight exactly as in a transit situation.Standard Earth atmospheric model calculations are from e .g.Ehrenreich et al. (2006) or Kaltenegger & Traub (2009). Fromthese model calculations, one can evaluate the e ff ective Earth ra-dius at each wavelength, and from it, the e ff ective height of theEarth atmosphere as a function of wavelength. From Fig. 3 ofKaltenegger & Traub (2009), the e ff ective altitude correspond-ing to the tip of the Chappuis band absorption (near 6 000 Å), ison the order of 30 km, revealing that the atmosphere becomesnearly opaque below this altitude at this wavelength.In our model calculation we used an unidimensional sin-gle scattering transmission model based on the model de-scribed by Ehrenreich et al. (2006) to obtain the theoreti-cal transmission spectrum of the terrestrial atmosphere. Thepressure, temperature, and mixing ratio profiles of the consid-ered atmospheric components are shown in Fig. 8 and comefrom the US Standard Atmosphere 1976 spring-fall pressure-temperature profile (COSEA 1976; Cox 2000). For simplicity,we chose to only include molecular nitrogen (N ), molecularoxygen (O ), and ozone (O ) in the model. Molecular nitro-gen contributes to the transmission spectrum via the Rayleigh-scattering of light. The Rayleigh-scattering cross section of N depends on the N refractive index, which is calculated accord-ing to Sneep & Ubachs (2005). Within the SOPHIE spectralrange, molecular oxygen contributes to the transmission spec-trum through i) photoabsorptions by the forbidden Σ + g – Σ − g tran-sition bands B (1–0) at 6 880 Å and γ (2–0) at 6 280 Å, andii) through Rayleigh scattering. The line parameters for the O bands are extracted from the HITRAN 2008 molecular spectro-scopic database (Rothman et al. 2009), scaled to the atmospherictemperature and pressure profiles used, and convolved with aGaussian profile with full-width at half-maximum matching thespectrograph resolution. The Rayleigh-scattering cross sectionsof O are calculated following Sneep & Ubachs (2005) using re-fractive indices from Bates (1984). We retrieved the UV / visiblephotoabsorption cross section of O at 293 K and 1 bar from theGEISA 1997 data base (Jacquinet-Husson et al. 1999).We recall here that we are not doing a fit of our data but justcompute altitude signatures related to a “standard” Earth atmo-spheric model. The variable e ff ective altitude h (as a function of wavelength)is shown on Fig. 9 for both model calculations and as extractedfrom the observations, the overall contribution of O , O and idal–Madjar et al.: The Earth as a transiting planet 9 Figure 8. “Standard” atmosphere used in the to model calculation (see text) :
Upper left the pressure profile, upper right thetemperature profile, lower left the N (blue), O (green) and O (pink) mixing ratio as a function of altitude (km) and lower right the N (blue) Rayleigh, O (green) Rayleigh and molecular bands and O (pink) cross sections as a function of wavelength (in nm).Rayleigh scattering due to both N and O , along with theRayleigh contribution alone are also shown.In the selected reference domain from 4 520 to 4 540 Å, themodel calculation predicts an altitude equal to 23.8 km. Becauseour observational evaluation of h is relative to the altitude inthe reference domain, we shift our relative h ( λ ) evaluations by + . ∼ i.e. ∼ ± . The acute need for high spectral resolution is highlighted by thedetection of other species, which are producing signatures at al-titudes above the ozone level.
The spectral region containing the NaI doublet is shown inFig. 10. To repeat the discussion presented in Sect. 3, concerningthe zero flux level correction, the result without any correctionapplied is also shown. Obviously, the atmospheric NaI doubletdoes not show up without that correction, and furthermore thebroad wings of the two NaI solar lines are still present while ourextraction process should have entirely washed out any solar linesignature, as it is indeed almost perfectly the case in Figs. 6 or 9.This correction introduces an additional systematic error,which produces an altitude shift of about 1.5 km per 0.0001 step
Figure 9.
Observed binned variations (solid line) of the e ff ective altitude h compared to model calculations (also binned). TheRayleigh-alone model calculation is shown (dashed line) along with the complete atmospheric model calculation (dotted line),which includes N , O and O (see text).change in the zero level (see Fig. 10). This is small compared tothe systematic error due to the instrumental order correction (onthe order of ± Oand O produce strong absorption signatures in the Earth at-mosphere and, as shown in Fig. 2, they have relatively similarsignatures near 5 900 Å (in the NaI lines vicinity) for H O oraround 6 300 Å for O . We will see in the sections below thatO is indeed easily detected in the grazing (transit like) observa-tions of the atmosphere, while we see here that none of the nu-merous H O lines present in that spectral range (see Fig. 3) aredetected. The reason is very simple: H O is hidden in the tran-sit like observations because it is at lower altitudes than ozone,while O , a major constituent of the atmosphere, is still presentin su ffi cient quantities at higher altitudes, and indeed shows upabove the ozone layer. This is exactly what models do predict(see e.g. Kaltenegger & Traub 2009) and thus reveals the qualityand consistency of our analysis, from which only species thoughtto show up are indeed detected.The zero level correction induces additional very large errorson the estimated altitude of the NaI spikes: the NaI 5 890 Å linepeaks up from 35 to 100 km, while the 5 896 Å one from 35 to65 km altitudes. NaI is known to be present in the Earth atmosphere within alayer at about 92 km, presenting an average thickness of about11 km (Moussaoui et al. 2010). If it was completely opaque inthe two NaI lines, we should have found the same level in bothlines at lower altitudes than the 92 km, depending on the ratiobetween the atmospheric line width ( ∼ ∼ i.e. ∼
50 km.This could correspond to our 0.0002 zero level correction,leading to both line levels at about 35 km altitude, a little lowwhen compared to our evaluation.However, we can also directly evaluate the line’s opacityin the 92 km altitude layer by using the observed average ver-tical sodium column density of the layer to be on the orderof 4.10 m − (Moussaoui et al. 2010). Using the average alti-tude, thickness, and “vertical” column density of the layer, wecan translate these numbers into an average “horizontal” col-umn density at 92 km of altitude and found it to be on the or-der of 3.10 cm − , leading to an optical thickness in the coreof the line on the order of 3 in the strongest NaI line and 1.5in the weaker one. Both lines are thus opaque, but not stronglyopaque. The known variability of the NaI layer could be the rea-son for this slightly less opacity, which renders it probable that idal–Madjar et al.: The Earth as a transiting planet 11 Figure 10.
NaI doublet spectral region. The shifted h ( λ ) variations (solid lines) are shown along with the level of the model ozoneprediction (dotted line). Upper plot.
The zero level correction as detailed in Sect. 3 is not applied. The two NaI narrow atmosphericsignatures do not show up while through our complete data analysis method the broad wings of the sodium solar lines are obviouslystill present, contrary to expectation (the solar broad lines wings should not appear in any information uniquely related to the Earthatmosphere).
Lower plot.
The zero level correction by subtracting a 0.0003 value is now applied (see text). The NaI solar wingshave disappeared, while the narrow NaI atmospheric signatures clearly show up. To show the impact of the zero level correction,the two evaluations corresponding to the 0.0002 and 0.0004 zero corrections are overplotted (thin dash-dotted lines). This gives adirect idea of the induced error bars on our evaluation.a marginally opaque layer was observed during the eclipse, thenleading to two di ff erent altitudes for the observed doublet spikescorresponding to the also plausible values as the ones observedwith our 0.0003 zero corrections, which are at 55 and 45 kmrespectively. From this brief quantitative discussion, we note that the0.0004 zero correction is certainly a limit because, in that case,the altitude of the strongest line is at more than 100 km, i.e. toohigh, even above the altitude of the layer itself.The NaI atmospheric detection is certainly compatible withwhat is known about the Earth layer, both qualitatively and quan- titatively. This shows that our high-resolution approach couldgive us access to narrow line signatures, and thus gives us confi-dence in searching for other species. We detail this search in thefollowing sections. Two molecular oxygen absorption bands have their signaturesin the observed spectral range. We used for them the same zerolevel correction as for the NaI lines.The strongest one, the O forbidden Σ + g – Σ − g transition bandB (1–0) at 6 880 Å, is shown in Fig. 11. All molecular bandpeaks are clearly present, undoubtedly signing the O detection.To evaluate the impact of the zero level adjustment, the threeevaluations of h ( λ ) made with the three values, 0.0003 ± ∼± ∼±
10 km in the peaks,showing that their absolute estimates from these observationsare quite imprecise.However, the average peak altitude and its relative heightvariations are certainly more meaningful. In particular, the av-erage height of the 0.0004 correction produces signatures closerto the model predictions, suggesting that the zero level correc-tion may be slightly wavelength-dependant or that its acceptablerange is reduced to 0.00035 ± The other oxygen molecular band in the observed spectral rangeis the γ (2–0) at 6 280 Å. It is shown in Fig. 12, for which thesame zero level corrections have again been used.The narrow spike positions of the band are clearly detected.This band is weaker however and emerges only by ∼ demonstrates the quality and precision of our approach,which allows us to detect atmospheric signatures just beyondthe photon noise. Indeed the photon noise induces in that regionof high solar flux and high instrument sensitivity only a ± molecular band we have checked theperfect match of the peak’s spectral positions with the predictedones while it seems that again their heights drop faster thanthe peak’s altitude variations evaluated in the model calculation.This appears to confirm that the “standard” atmospheric modelselected for that comparison is probably hotter than the observedatmosphere, particularly in the sampled ∼
30 km altitude range.If this were true it may also explain the slight mismatch foundbetween the predicted and the observed O Chappuis band (see Fig. 9), in which the model is higher by ∼ Figure 12. γ (2–0) 6 280 Å O molecular band. Upper plot.
Same background correction as in Fig. 10 (see text). The solidline represents the altitude information as extracted from theSOPHIE data set.
Lower plot.
The O band spectral structureis shown to demonstrate the clear coincidence in position of therepeated spikes signing the O detection. Other observations of the 2008 August 16 lunar eclipse werecompleted through inter-calibrated optical and near-infraredground-based observations at the William Herschel and NordicOptical Telescopes by Pall´e et al. (2009), to provide continuouswavelength coverage from 0.36 to 2.40 µ m. Their approach wasvery similar to ours, except that their spectral resolution is some-what lower ( ∼ ∼
75 000). They also focused onumbral, penumbral, and out-of-the-eclipse observations withoutdetailing (as we did) the precise location within the penumbrafrom where the observations were made. This is important be-cause it leads toward only qualitative detections instead of quan-titative ones.Furthermore, the Pall´e et al. (2009) transmission spectrumis “calculated by computing the ratio of the umbra / penumbra idal–Madjar et al.: The Earth as a transiting planet 13 Figure 11. O forbidden Σ + g – Σ − g transition band B (1–0) at 6 880 Å. The same background correction as in Fig. 10 has been applied(see text). Upper plot.
The solid line represents the altitude information extracted from the
SOPHIE data, while the dotted line isthe model calculation, in which O absorption is included from 6 856 Å upwards. The narrow atmospheric signatures due to the O absorption are clearly seen. The sharp peaks perfectly match the model with respect to spectral positions, while their heights couldbe quite di ff erent. Lower plot.
The thick solid line again represents the evaluated altitude for the 0.0003 zero level correction, whilethe thin dash-dotted lines show the extreme possible variations of the estimated levels due to the 0.0002 (lower evaluations in thepeaks and higher ones in the continuum) and 0.0004 (higher evaluations in the peaks and lower ones in the continuum) zero levelcorrections.regions taken at the same averaged air mass in order to mini-mize the local atmospheres telluric line variations”. This meansthat the observed transmission spectrum they evaluated is notthe transit-like spectrum of the Earth. Indeed the umbra spectraresulting uniquely from rays deflected by the lower and denselayers of the atmosphere are unobservable during a real transit of an extrasolar planet since the corresponding stellar light de-flected through the lower layers of the atmosphere moves awayfrom the observer line of sight. What Pall´e et al. (2009) havereported are identifications of di ff erent atmospheric specie sig-natures only present in the lower parts of the atmosphere as seen within the umbra, but not the real transit signatures, which areonly observable within the penumbra, as we did. Figure 13.
The h ( λ ) variation as extracted from the region of theCaII H line. Because the S / N is quite poor in that region thephoton noise in this part of the spectrum is more than 10 timeshigher than in the other spectral regions. Large signatures similarto the NaI ones should be visible however, but according to thelow CaII content of the Earth atmosphere, they are simply notobservable, as it is the case here.In particular, Pall´e et al. (2009) have not detected the NaI sig-natures. They are produced at very high altitudes only ( ∼
92 km)and should be indeed unobservable in the lunar umbra. Theymention the detection of CaII signatures however, while it is alsoknown that in the Earth atmosphere CaII ions are only presentat high altitudes (Granier et al. 1989). CaII is furthermore about120 times fainter than NaI, thus the CaII detection (and not theNaI one) is quite a surprise. Because we have in our hands muchlower S / N SOPHIE observations at our disposal, we tried how-ever to look for possible CaII signatures. As shown in Fig. 13,nothing shows up. We conclude that the Pall´e et al. (2009) CaIIdetection is probably an artifact.To enforce that argument we note that Pall´e et al. (2009)clearly detected H O signatures (a low altitude species)at ∼ and Rayleigh signatures(see Fig. 9). This is why the di ff erent H O signatures we haveseen in our observation of the “vertical” Earth atmosphere (seeFig. 2) do not show up in any of our “horizontal” transit-like sig-natures within the Earth atmosphere, because they are simply toolow in altitude as predicted from the model calculations (see e.g.
Fig. 10 where several H O lines should have shown up aroundthe NaI doublet, if present).The Pall´e et al. (2009) study in a larger spectral domain thanours reveals the great variety of detectable species, while ours re-veals how precisely the atmospheric content could be analyzed. and O in anextrasolar planetatmosphereduring transits The purpose of this work is also to show and evaluate the fea-sibility of the detection of O and O in the atmosphere of anEarth-like extrasolar planet during transits. We will not repeat the arguments of Lecavelier des Etangs& Ehrenreich (2005) here but only recall the extreme di ffi cultyof these observations, which demand 10 − − − accuracy tobe obtained over the observed stellar flux before, during, andafter the planetary transit. This is a real challenge even merelyin terms of photon noise, and asks for very large telescopes inthe 10 to 30 m diameter range. This is only possible from theground in a relatively short-term perspective, in particular in theframe of the ELTs (Extremely Large Telescopes) concepts thatare planned to operate during the coming decade.The
ELTs are however observing through the Earth atmo-sphere, and it is di ffi cult to argue that such high precision accu-racies will be achieved.Our observations show that although the O Chappuis bandseems to be easier to detect according to the broad spectral rangeover which it extends, it is still possible to think that uncontrolledsystematics due to atmospheric fluctuations will perturb the de-tection. This is the obvious di ffi culty of broadband detections. Figure 14.
Correlation function extracted from the O forbidden Σ + g – Σ − g transition band B (1–0) at 6 880 Å by using a maskcentered on each of the spike positions over a ± observation via high-resolutionspectroscopy signatures is very promising because it can beshown that the addition of all spectral lines over only the O forbidden Σ + g – Σ − g transition band B (1–0) at 6 880 Å is equiv-alent to a detection over a ∼ detection among the feasible ones from the ground with ELT telescopes at least in terms of photon fluxes. But the use of highspectral resolution and stable spectrographs, similar to the onepresently used for radial velocity searches of extrasolar planets,has proved to be possible in a quite di ffi cult case, which is toobserve the Earth’s atmosphere as a transiting planet through theEarth atmosphere itself, i.e. all searched for and perturbing spec-tral signatures are exactly at the same wavelengths. For transit-ing extrasolar planets the situation will be more favorable be-cause a simple shift of ∼
10 to 30 km / s or more will be enoughto clearly separate them as shown through the correlation func-tion extracted with an O like mask (see Fig. 14). This shouldbe a relatively common observational situation. Note the sec-ondary peaks of the correlation function which are due to the rel- idal–Madjar et al.: The Earth as a transiting planet 15 atively regular separation between the successive O line peaks.Because these are however well separated, this should not pre-vent O detection over this molecular band.Our observation shows that with high spectral resolution,systematics are much easier to control because extremely nearbyspectral regions behave similarly to each other in terms of anysystematic e ff ects. This may be one of the very few possible ap-proaches to reach this goal, which is detecting O , one of thestrongest signatures of life bearing atmospheres.Observing O and O species in extrasolar Earth like plane-tary atmospheres, both accessible from the ground in the samespectral range, shows that these searches will be attainable for ELT observatories equipped with high spectral resolution andhigh stability spectrographs.
6. Conclusion
We have shown that Lunar eclipse spectral observations are ableto reveal the atmospheric content of the Earth’s atmosphere, inparticular, broadband signatures of O and Rayleigh scatteringas well as narrowband features of NaI and O observed at highspectral resolution. As these observations mimic future studiesof transiting extrasolar planets, we are confident that quantitativeinformation about extrasolar atmospheres will be within reach.Furthermore, both Rayleigh scattering and O are broadband andextend across the visible part of the spectrum, just where solar-like stars have their maximum flux, making it easier to obtainthe high signal-to-noise values necessary for a positive exoplanetdetection.More studies of Lunar eclipses should be completed in orderto better quantify the present detections and more precisely showthe feasibility of future extrasolar Earth-like planets studies.These studies, with extremely large ground-based telescopes,should allow at least the detections of ozone and O . In any case,these observations will be extremely di ffi cult because the re-quired accuracy is in the 10 − to 10 − range. This is why observ-ing in the visible range, where orders of magnitude more photonsare available, could ultimately be one of the most promising ap-proaches. Acknowledgements.
The authors thank the sta ff of Haute-Provence Observatoryfor their contribution to the success of the SOPHIE project and their supportat the 1.93-m telescope. We thank the “Programme National de Plan´etologie”(PNP) of CNRS / INSU, the Swiss National Science Foundation, and the FrenchNational Research Agency (ANR-08-JCJC-0102-01 and ANR-NT05-4-44463)for their continuous support of our planet-search programs.We also thank W.A. Traub, our referee, for mentioning the ring e ff ect thatpossibly explains the zero shift correction and E. Pall´e for very constructive dis-cussions before the observing campaigns of the 2008 August 16 Moon eclipse.DE acknowledges financial support from the Centre National d’EtudesSpatiales (CNES) and NCS the support from the European ResearchCouncil / European Community under the FP7 through a Starting Grant, as wellas from the Fundac¸˜ao para a Ciˆencia e a Tecnologia (FCT), Portugal, througha Ciˆencia 2007 contract funded by FCT / MCTES (Portugal) and POPH / FSE(EC), and in the form of grants reference PTDC / CTE-AST / / / CTE-AST / / References
Arnold, L., Gillet, S., Lardi`ere, O., Riaud, P., & Schneider, J. 2002, A&A 392 ,231Arnold, L. 2008, Space Science Reviews, 135 , 323Ballester, G.E., Sing, D.K., & Herbert, F. 2007, Nature 445, 511Barman, T. 2007, ApJ 661, L191Bates, D.R. 1984, P&SS, 32, 785Bird, R.E., Hulstrom, R.L., Kliman, A.W., & Eldering, H.G. 1982, AppliedOptics 21, 1430Borchi, F., & Pommereau, J.-P. 2007, Atmos. Chem. Phys. 7, 2671 Bord´e, P., Rouan, D., & L´eger, A. 2003, A&A, 405, 1137Bouchy, F., H´ebrard, G., Udry, S., et al. 2009, A&A, 505, 853Charbonneau, D., Brown, T.M., Noyes, R.W., & Gilliland, R.L. 2002, ApJ 568,377COESA (Committee on Extension to the Standard Atmosphere), U.S. StandardAtmosphere 1976 (Washington DC: Government Printing O ffi ce)Cox, A.N. 2000, Allen’s Astrophysical Quantities (4th ed.; New York: AIP)Danjon, A. 1936, Annales de l’Observatoire de Strasbourg 3, 139D´esert, J.-M., Vidal-Madjar, A., Lecavelier Des Etangs, A., Sing, D., Ehrenreich,D., H´ebrard, G. & Ferlet, R. 2008, A&A 492, 585D´esert, J.-M., Lecavelier des Etangs, A., H´ebrard, G., Sing, D.K., Ehrenreich,D., Ferlet, R., & Vidal-Madjar, A. 2009, ApJ 699, 478Ehrenreich, D., Tinetti, G., Lecavelier Des Etangs, A., Vidal-Madjar, A., &Selsis, F. 2006, A&A 448, 379Ehrenreich, D., H´ebrard, G., Lecavelier des Etangs, A., Sing, D.K., D´esert, J.-M.,Bouchy, F., Ferlet, R., & Vidal-Madjar, A. 2007, ApJ 668, L179Fortney, J. J. and Marley, M. S. and Lodders, K. and Saumon, D. and Freedman,R. 2007, ApJ, 627, L69Gedzelman, S. D., & Vollmer, M. 2008, Applied Optics 47, 149Granier, C., Jegou, J.P., & M´egie, G. 1989, Journal of Geophys. Res. 94, 9917Hayes, D.S., & Latham, D.W. 1975, ApJ 197, 593Hernitschek, N., Schmidt, E. & Vollmer, M. 2008, Applied Optics 47, 62Jacquinet-Husson, N., Ari´e, E., Ballard, J., et al. 1999, JQSRT 62, 205Kaltenegger, L., & Traub, W.A. 2009, ApJ 698, 519Lecavelier des Etangs, A., & Ehrenreich, D. 2005, 39 th ESLAB Symposium onTrends in Space Science and Cosmic Vision 2020, Edt. F. Favata, J. Sanz-Forcada, A. Gim´enez, and B. Battrick. ESA SP-588, 73Lecavelier Des Etangs, A., Pont, F., Vidal-Madjar, A., & Sing, D. 2008a, A&A481, L83Lecavelier Des Etangs, A., Vidal-Madjar, A., D´esert, J.-M., & Sing, D. 2008b,A&A 485, L865L´eger, A.,Selsis, F., Sotin, C., et al. 2004, Icarus 169, 499L´eger, A., Rouan, D., Schneider, J., et al. 2009, A&A 506, 287Millier, F., Vidal-Madjar, A., Guidon, J., & Roble, R.G. 1979, Geophys. Res.Let. 6, 863Moussaoui, N., Clemesha, B.R., Holzl¨ohner, R., Moussaoui, Simonich, D.M.,Bonaccini Calia, D., Hackenberg, W., & Batista, P.P. 2010, A&A 511, A31Pall´e, E., Osorio, M.R.Z., Barrena, R., Monta˜n´es-Rodrguez, P., & Martn, E.L.2009, Nature 459(7248), 814Perruchot, S., Kohler, D., Bouchy, F., et al. 2008, in Ground-based and AirborneInstrumentation for Astronomy II. Edited by McLean, Ian S. & Casali, MarkM., Proceedings of the SPIE, Volume 7014, pp. 70140J-12Queloz, D., Bouchy, F., Moutou, C., et al. 2009, A&A 506, 303Rothman, L.S., Gordon, I.E., Barbe, A., et al. 2009, JQSRT 110, 533Rouan et al. 2009, CoRoT International Symposium I, Paris, 2-5 feb. 2009Sagan, C., Thompson, W. R., Carlson, R., Gurnett, D., & Hord, C. 1993, Nature,365, 715Sing, D.K., Vidal-Madjar, A., D´esert, J.-M., Lecavelier des Etangs, A., &Ballester, G. 2008a, ApJ 686 , 658Sing, D.K., Vidal-Madjar, A., Lecavelier des Etangs, A., D´esert, J.-M., Ballester,G.E., & Ehrenreich, D. 2008b, ApJ 686 , 667Sing, D.K., D´esert, J.-M., Lecavelier Des Etangs, A., Ballester, G.E., Vidal-Madjar, A., Parmentier, V., H´ebrard, G., & Henry, G.W. 2009, A&A 505, 891Sneep, M., & Ubachs, W. 2005, JQSRT 92, 293Thomason, L.W., & Taha, G. 2003, Geophys. Res. Let. 30, 1631Vidal-Madjar, A., Laurent, C., Bruston, P., & Audouze, J. 1978, ApJ 223, 589Vidal-Madjar, A., Lecavelier des Etangs, A., D´esert, J.-M., Ballester, G.E.,Ferlet, R., H´ebrard, G., & Mayor, M. 2003, Nature 422, 143Vidal-Madjar, A., D´esert, J.-M., Lecavelier des Etangs, A., H´ebrard, G.,Ballester, G.E., Ehrenreich, D., Ferlet, R., McConnell, J.C., Mayor, M., &Parkinson, C.D. 2004, ApJL 604, L69Woolf N.J., Smith P.S., Traub W.A., & Jucks, K.W. 2002, ApJ 574, 430http: // / resources / mapcatalog / usgs / I703 //