Free Access
Issue
A&A
Volume 534, October 2011
Article Number A26
Number of page(s) 11
Section Planets and planetary systems
DOI https://doi.org/10.1051/0004-6361/201117091
Published online 23 September 2011

© ESO, 2011

1. Introduction

More than 500 extrasolar planets and planet candidates are currently known. Over 100 of these planets transit their central star. Ten of the transiting planets are so-called super-Earths with masses below ten Earth masses: CoRoT-7 b (Léger et al. 2009), GJ 1214 b (Charbonneau et al. 2009), Kepler-9 d (Holman et al. 2010; Torres et al. 2011), Kepler-10 b (Batalha et al. 2011), Kepler-10 c (Fressin et al. 2011), Kepler-11 b, d, e and f (Lissauer et al. 2011) , and 55 Cnc e (Winn et al. 2011; Demory et al. 2011).

The unique geometrical orientation of transiting planets offers the opportunity for the spectral characterization of their atmospheres. Transmission spectroscopy during the primary transit is more practical at ultraviolet (UV), visible and near-infrared (IR) wavelengths, because the stellar signal is stronger towards these shorter wavelengths. In contrast, emission spectroscopy during the secondary eclipse is easier in the mid-IR because the planet-star flux ratio is higher in this wavelength regime. Thus, both methods provide complementary results.

Emission and transmission observations have been performed for almost 30 extrasolar planets to date. Thermal emission of radiation has been detected (e.g., Deming et al. 2005; Harrington et al. 2006; Knutson et al. 2007; Alonso et al. 2009). Furthermore, the chemical composition of both the atmospheres and exospheres of some of these planets has been determined. Atoms (H, C, O, Na, Fe, Mg) and molecules (CO, CO2, CH4, H2O) have been detected (e.g., Charbonneau et al. 2002; Vidal-Madjar et al. 2004; Tinetti et al. 2007; Grillmair et al. 2008; Swain et al. 2009; Stevenson et al. 2010; Madhusudhan et al. 2011). Although there is some ongoing discussion as to whether some of these observations are valid (Gibson et al. 2011; Mandell et al. 2011), atmospheric characterization of exoplanets is indeed feasible with current instrumentation. For two transiting super-Earths, spectroscopic observations have already been performed (CoRoT-7 b, Guenther et al. 2011; and GJ 1214 b, Bean et al. 2010; Désert et al. 2011; Croll et al. 2011; Crossfield et al. 2011). For CoRoT-7 b, upper limits to the extension of the exosphere have been obtained. The observations of GJ 1214 b have helped to constrain the range of possible atmospheric scenarios, favoring either a cloud-free water vapor atmosphere or a cloudy hydrogen-dominated, methane-depleted atmosphere. These two planets are, however, far too hot to be considered habitable in the classical sense of life as we know it on Earth.

Nevertheless, studies of the potential atmospheric signatures of terrestrial habitable planets have been performed, to predict signal strengths and assess observation strategies (e.g., Des Marais et al. 2002; Segura et al. 2003; Ehrenreich et al. 2006; Kaltenegger & Traub 2009; Miller-Ricci et al. 2009; Deming et al. 2009; Belu et al. 2011; Rauer et al. 2011).

The extrasolar planet GL 581 d (Udry et al. 2007; Mayor et al. 2009) is the first potentially habitable super-Earth (Wordsworth et al. 2010b; von Paris et al. 2010; Hu & Ding 2011; Kaltenegger et al. 2011; Wordsworth et al. 2011). The orbital inclination of GL 581 d has been shown to lie in the range 40 − 88°, based on photometric constraints for GL 581 b (López-Morales et al. 2006) and dynamical simulations of the whole system (Mayor et al. 2009). Hence, GL 581 d is unlikely to transit its central star. However, we used GL 581 d as an analogue of similar, transiting systems that are anticipated to be found in the near future. In this study, we illustrate the possible spectroscopic characterization of potentially habitable planets based on a wide range of atmospheric model scenarios of GL 581 d from von Paris et al. (2010), following the strategy outlined below. Synthetic spectra of some specific, potentially habitable GL 581 d model atmospheres were presented by Kaltenegger et al. (2011) and Wordsworth et al. (2011). However, these authors discussed neither the potential detectability (i.e., signal-to-noise ratios) of spectral features nor the possibility of a detailed characterization of their model atmospheres.

The paper is organized as follows: Sect. 2 briefly outlines the aforementioned observing strategy. Section 3 presents the atmospheric scenarios and the models used. We present our results in Sect. 4 and our conclusions in Sect. 5.

2. Potential atmospheric characterization of terrestrial exoplanets

2.1. Observations

The observable quantity during transit measurements is the wavelength-dependent planet-to-star contrast ratio, i.e. the transit or eclipse depth. Based on the knowledge of the stellar radius, transmission spectra thus measure the apparent planetary radius as a function of wavelength. However, this depends critically on the accurate characterization of the central star, which is the main source of uncertainty for derived planetary properties (see, e.g., the case of GJ 1214 b, Carter et al. 2011). From the stellar properties such as spectrum and radius, and adopting a baseline value for the planetary radius, emission spectra during secondary eclipse can be translated into brightness temperature spectra. The brightness temperature spectra are particularly illustrative since the brightness temperature is the apparent atmospheric temperature at a given wavelength. For optically thin spectral windows, the brightness temperature corresponds to the surface temperature. In the following, we assume that the stellar properties as well as the geometric planet radius are known exactly.

2.2. Atmospheric characterization

An observation strategy would aim to (1) establish the existence of an atmosphere; (2) determine the major atmospheric constituents along with radiative trace gases; (3) characterize surface conditions, hence assess habitability; and ultimately; (4) search for atmospheric species that would indicate the presence of life on a planet. Usually, however, the existence of an atmosphere and its composition can only be established via atmospheric and spectral modeling with subsequent comparison to the obtained data (e.g., Madhusudhan & Seager 2009; Miller-Ricci et al. 2009; Miller-Ricci & Fortney 2010).

An additional challenge is the possible false-positive or false-negative identification of so-called biomarkers (e.g., O3, CH4, N2O), which was discussed, e.g., by Selsis et al. (2002) and Schindler & Kasting (2000). Biomarkers are atmospheric species that are assumed to indicate that a biosphere is present on the planet. On Earth, N2O, for example, is believed to be almost exclusively produced by denitrifying bacteria, and O3 is a photochemical product of oxygen that itself originates mainly from photosynthetic organisms. These biomarkers can be detected spectroscopically in the Earth’s atmosphere using absorption bands (ozone at 9.6 μm, nitrous oxide at both 7.8 and 4.5 μm and methane at 7.7 and 3.3 μm).

It has to be distinguished between the false-positive or false-negative detection of biomarker species and the false-positive interpretation of detected biomarkers as a sign for life. This is due to the fact that abiotic formation of ozone is possible (e.g., Segura et al. 2007; Domagal-Goldman & Meadows 2010), even though the magnitude of the effect is debated (Selsis et al. 2002). Note, for example, that ozone has been found in the Martian atmosphere (e.g. Yung & deMore 1999). Therefore, the detection of ozone alone could still be a false-positive detection of life.

In terms of detecting biomarkers and life, the triple signature O3, CO2, and H2O is a possible way of avoiding false-positive detections of biospheres, as proposed by, e.g., Selsis et al. (2002). Sagan et al. (1993) proposed O2 (or its tracer O3) and CH4 as combined biomarkers.

3. Models and atmospheric scenarios

3.1. Atmosphere model

The spectra shown here have been calculated using atmospheric profiles from scenarios summarized in von Paris et al. (2010). Atmospheric profiles of temperature, pressure, and water were calculated using a 1D radiative-convective model. The model is based on the original model of Kasting et al. (1984). It solves the radiative transfer equation to calculate temperature profiles in the stratosphere. The stellar flux is treated in 38 spectral intervals of varying width ranging from 0.237 to 4.545 μm. The radiative transfer scheme uses a δ-2-stream method (Toon et al. 1989) to incorporate Rayleigh scattering by N2, CO2, and H2O (Vardavas & Carver 1984; von Paris et al. 2010). In addition, absorption by H2O and CO2 is taken into account in the visible and near-IR (Pavlov et al. 2000). In the troposphere, the atmosphere is assumed to be convective. Hence, temperature profiles are assumed to be adiabatic, based on Kasting (1988) and Kasting (1991). The water profile is calculated using a fixed relative humidity profile (Manabe & Wetherald 1967). More details of the model are given in von Paris et al. (2008) and von Paris et al. (2010).

3.2. Atmospheric scenarios

The model scenarios used the orbital distance of 0.22 AU and eccentricity of e = 0.38 for GL 581 d (Mayor et al. 2009). The stellar input spectrum was based on a synthetic NextGen spectrum (Hauschildt et al. 1999) in the visible and near-IR combined with measurements by the International Ultraviolet Explorer satellite in the UV, as described by von Paris et al. (2010).

Model CO2 concentrations were chosen to be consistent with CO2 concentrations on present Venus and Mars (95%), present Earth (3.55 × 10-4) as well as assumed scenarios of the early Earth (5%, e.g., Kasting 1987). Surface pressures were chosen such that model column densities were comparable to scenarios of early Earth or early Mars in the literature (e.g., Goldblatt et al. 2009; Tian et al. 2010). Three different atmospheric scenarios were considered, defined by the CO2 concentration: low (355 ppm volume mixing ratio, vmr), medium (5% vmr) and high (95% vmr) CO2 cases. The model atmospheres further contained water and molecular nitrogen as a filling gas, i.e. the model atmospheres are CO2-H2O-N2 mixtures. In each of these cases, the surface pressure was varied from 1 to 20 bar. For the 1 and 20 bar cases, respectively, spectra are presented below. For all runs, the surface albedo was kept constant at Asurf = 0.13 which is the measured surface albedo of Earth (Rossow & Schiffer 1999). In doing so, clouds were explicitly excluded in the climate calculations of von Paris et al. (2010). Table 1 summarizes the atmospheric scenarios considered in this work.

Table 1

Atmospheric scenarios for GL 581 d considered here.

The model calculations of von Paris et al. (2010) suggested that four scenarios with high CO2 partial pressures (high-CO2 5, 10 and 20 bar, as well as medium-CO2 20 bar) were habitable, with surface temperatures above 273 K, i.e. above the freezing point of water. These results were in broad agreement with the results of Wordsworth et al. (2010b), Hu & Ding (2011), and Kaltenegger et al. (2011).

3.3. Computation of spectra and signal-to-noise ratios

thumbnail Fig. 1

Spectra of the 20 bar high-CO2 case. Left: secondary eclipse spectrum. Inlet shows the corresponding temperature-pressure profile. Right: transit depth spectrum.

Open with DEXTER

The line-by-line code MIRART-SQuIRRL (Schreier & Böttger 2003) was used to calculate high-resolution synthetic planetary emission, brightness temperature, and transmission spectra. Note that the collision-induced absorption (CIA) of CO2 is not included here, although the CIA is thought to be important in some spectral regions around 7 μm and longwards of 20 μm (e.g., Wordsworth et al. 2010a).

Furthermore, the output of the stellar radiative transfer code of the climate model (see above) was used to produce reflection spectra IR(1)where IS is the stellar spectrum, AS the spectral albedo from the stellar radiative transfer code, Rs the stellar radius, and d the orbital distance. This approach of using the climate model output to construct spectra is based on Kitzmann et al. (2011). A similar method was used by Wordsworth et al. (2011) to calculate synthetic broadband emission spectra of GL 581 d model atmospheres.

These spectra were then used to calculate contrast spectra (i.e., emission + reflection) as well as spectra of the effective tangent height of the atmosphere. Our calculations closely follow those of Rauer et al. (2011), where more details can be found. The spectra were calculated on an equidistant spectral grid (in wavenumber), hence the spectral resolution R varies between R ≈ 2 − 10 × 103, depending on wavelength.

For a detection of a spectroscopic feature, the relevant quantity is the signal-to-noise ratio (SNR). In contrast to the SNR calculations by, e.g., Kaltenegger & Traub (2009) or Rauer et al. (2011), we take into account not only the stellar photon noise, but also the thermal emission of the telescope, the zodiacal emission, and the dark noise. The spectrum of GL 581 used in the SNR calculations is the synthetic spectrum described above. We base our telescope parameters on the James Webb Space Telescope (JWST), assuming a 6.5 m aperture and a detection efficiency of 0.15 (Kaltenegger & Traub 2009). More details of the noise estimates can be found in Appendix A. The fictitious transit duration of GL 581 d, hence the assumed integration time, is calculated to be 4.15 h. The value of SNR are calculated for a spectral resolution of R = 10, which is a reasonable value in the context of exoplanet characterization.

thumbnail Fig. 2

Same as Fig. 1, but for the 1 bar low-CO2 case.

Open with DEXTER

4. Results and discussion

4.1. Presence of an atmosphere

The left part of Fig. 1 shows the contrast spectrum of the high-CO2 20 bar case. It is clearly seen that the reflection component dominates the near-IR up to about 2 μm and in the window 2.7 μm. For longer wavelengths, the emission of the planet is the main component. The broad water and CO2 absorption bands are clearly seen in the spectrum. Hence, the existence of an atmosphere, as well as the presence of water and CO2, can be inferred. Interestingly, the planet-star contrast is rather low, even though GL 581 is an M-type star and GL 581 d a super-Earth. The contrast reaches about 4 × 10-5 in the mid-IR, which is about an order of magnitude higher than the contrast between Earth and the Sun. However, it is about 100 times lower than corresponding values for hot Jupiters. The right part of Fig. 1 shows the synthetic transit depth spectrum for the 20 bar high-CO2 case. It can be seen that because of the large amounts of water and CO2, the planet appears larger than its geometric radius at all IR wavelengths. Hence, the presence of an atmosphere could also be clearly inferred from transmission spectra because of the wavelength-dependent apparent radius.

Figure 2 also shows contrast and transit spectra, but for the 1 bar low-CO2 case. The spectra are relatively flat, except for the strong CO2 fundamental bands. In these bands, the presence of an atmosphere could be inferred, as for the 20 bar high-CO2 case. Since the atmosphere is very dry (partial pressure of water less than 10-5 bar), the water bands are difficult to discern in the emission spectrum. In contrast to the high-CO2 20 bar case, the secondary eclipse spectrum of the low-CO2 1 bar case is dominated by the reflection component up to 4.5 μm, i.e. the limit of the stellar radiative transfer code of the atmospheric model.

4.2. Atmospheric characterization

After securely detecting the atmosphere, the next step involves its characterization (i.e., composition, surface pressure) and, from there, assessment of the surface conditions, hence potential habitability.

4.2.1. Surface temperature

Figure 3 shows the brightness temperature spectra of the 1 and 20 bar scenarios of the low-CO2 (left) and the high-CO2 case (right). Note that brightness temperature spectra are based only on the emission spectrum because the brightness temperatures calculated from the contrast due to the reflection spectrum would yield values up to about 700 K (near 1 μm). Therefore, a characterization of the atmospheric temperatures of terrestrial planets is not possible in the near-IR up to about 4 − 5 μm.

The left part of Fig. 3 shows that for the low-CO2 scenarios, the surface temperature could be inferred from the brightness temperature spectra because the atmosphere is transparent except in the CO2 fundamental bands. In contrast, as can be seen in the right part of Fig. 3, the difference between the brightness temperature and the surface temperature is always non-zero in the high-CO2 20 bar case. This means that the emission spectrum does not allow to determine the surface temperature, hence to assess potential habitability. The reason for this is that the atmosphere is optically thick for thermal radiation due to the large amounts of CO2 and water in the atmosphere (von Paris et al. 2010). In the 1 bar high-CO2 case, some spectral windows would still allow for the determination of the surface temperature mostly because the atmosphere is much drier than in the high-CO2 20 bar case.

thumbnail Fig. 3

Brightness temperature spectra (based on planetary emission only) of the 1 and 20 bar scenarios. Left: low-CO2 case. Right: high-CO2 case. Surface temperatures are indicated by horizontal dashed lines.

Open with DEXTER

thumbnail Fig. 4

Comparison of secondary eclipse spectra to study the effect of varying surface pressure. Scenarios are as indicated. Inlets show the corresponding temperature-pressure profiles.

Open with DEXTER

thumbnail Fig. 5

As Fig. 4, but for transmission spectra instead of secondary eclipse spectra.

Open with DEXTER

4.2.2. Surface pressure

In Fig. 4, the 1 and 20 bar runs with high and low CO2 concentrations are compared to each other to illustrate how surface pressures could be inferred from secondary eclipse spectra. In the low-CO2 case (left), the main effect can be seen in the 15 μm band, which is considerably broader for the 20 bar run than for the 1 bar run. This is simply because the line center becomes optically thick at pressures of about 100 mbar, whereas the line wings are transparent up to pressures of the order of 5 − 10 bar. Also, the 2.7 and 4.3 μm bands are much deeper in the 20 bar scenario compared to the 1 bar scenario. In the high-CO2 case (right), the pressure effect in the 15 μm CO2 band is much less pronounced. Line wings are already saturated at pressures well below 1 bar. Owing to the large difference in surface temperature (see inlet), some spectral regions (e.g., windows near 8 − 9 μm or around 11 μm) differ considerably. Furthermore, in the near-IR, the reflection spectra show large differences, owing to the strong decrease in the spectral albedo with surface pressure. This illustrates the use of near-IR secondary eclipse measurements to determine atmospheric characteristics besides temperature.

Figure 5 shows the effect of changing surface pressure on the transmission spectrum for the low and the high-CO2 1 and 20 bar cases. The spectra show significant differences. For example, in the 9.5 μm CO2 band, tangent heights differ by about a factor of 2 to 3. However, in terms of absolute height, this amounts to at most 5 − 10 km. Note that the CO2 bands at 7 − 10 μm are visible even in the spectra of the high-pressure low-CO2 scenarios, in contrast to the emission spectra where these bands did not appear (see Fig. 4).

These results imply that for the cases studied, transmission spectra are in general more sensitive to surface pressure than emission spectra, for two main reasons. Firstly, the tangent height depends to first order on the atmospheric scale height, H ~  T/ma (T is the surface temperature and ma the mean molecular weight of the atmosphere). Thus, for higher surface pressures, and correspondingly higher surface temperatures, scale heights are larger. Secondly, for higher surface pressures, atmospheres extend further out into space. For example, the 20 bar high-CO2 case has its model lid (corresponding to a pressure of 6.6 × 10-5 bar) at 20 km, whereas the 1 bar case has its model lid at 13 km altitude, which corresponds roughly to a three scale heights-difference.

4.2.3. Surface albedo

The reflectivity of the surface, called the surface albedo, is an important parameter for helping distinguish between different types of surfaces, such as oceans or ice. Reflection spectra could in principle enable us to constrain the value of the surface albedo.

However, for optically thick atmospheres, as discussed above, the reflection spectrum cannot be used to determine the surface albedo in the high-CO2 20 bar case. Without the influence of the atmosphere, the reflection contrast would be 7.1 × 10-9 (with a fixed surface albedo of Asurf = 0.13, see Sect. 3.2). As can be seen from Fig. 1, the reflection contrast is far lower than this value, indicating that even cloud-free atmospheres could inhibit surface characterization via reflection spectra. The same is true for the high-CO2 1 bar and the low-CO2 20 bar cases (not shown). Only in the low-CO2 1 bar case would the reflection spectrum allow us to infer the surface albedo (see Fig. 2).

thumbnail Fig. 6

Comparison of secondary eclipse spectra to study the effect of varying CO2 concentration. Scenarios are as indicated. Inlets show the corresponding temperature-pressure profiles.

Open with DEXTER

4.2.4. Atmospheric composition

Figure 6 shows secondary eclipse spectra at equal surface pressures (20 bar), but for different CO2 concentrations (high, medium, low). It is rather difficult to distinguish between the medium and high-CO2 scenarios, except for some narrow spectral windows. When comparing Fig. 4 with Fig. 6, it is therefore difficult to decide whether the shape of the spectrum is actually due to a pressure difference at high CO2 concentration or a concentration difference at high surface pressures. It is, however, possible to distinguish between high or low CO2 concentrations by the presence of many weak CO2 bands in the high-CO2 case, e.g. at around 7, 9 and 10 μm. These bands do not appear in the spectrum unless CO2 concentrations exceed several percent. Additionally, the presence of water can be inferred from the strong 6.3 μm fundamental band and the signatures of the rotation band in the mid- to far-IR. Thus, dry and moist atmospheres could be distinguished.

To show the effect of varying the CO2 concentration on transmission spectra, Fig. 7 compares the 20 bar scenarios for low, medium and high CO2 concentrations. Interestingly, the medium-CO2 case shows a more pronounced 15 μm band of CO2 despite the fact that the high-CO2 scenario is significantly warmer at the surface (by about  ~40 K). This is due to the lower mean atmospheric molecular mass (hence, higher scale height) which is  ~29 g mol-1 in the medium-CO2 case and  ~43 g mol-1 in the high-CO2 case. The weak bands around 7 and 10 μm do not differ by much. Nevertheless, in the water rotation bands (longwards of 20 μm) and the strong CO2 fundamentals, rather large differences in the spectrum can be seen. This indicates that transmission spectroscopy could be used to distinguish between the different CO2 concentrations.

thumbnail Fig. 7

Same as Fig. 6, but for transmission spectra instead of secondary eclipse spectra.

Open with DEXTER

The same is true when comparing the low and high-CO2 scenarios. Here, however, the differences in the spectrum are visible in the weak bands of CO2 (e.g., around 10 μm) rather than in the strong fundamentals (4.3 and 15 μm).

thumbnail Fig. 8

Possibility of a false-positive biomarker detection for different scenarios, as indicated. Modern Earth spectrum taken from Grenfell et al. (2011). Planck curves of corresponding surface temperatures as dashed lines. Smoothed (red) and binned spectra (blue) are for a spectral resolution of 10.

Open with DEXTER

4.2.5. Surface conditions

In summary, the characterization of surface conditions on GL 581 d using secondary eclipse spectroscopy would be rather difficult, even if the planet were transiting. It would be possible to approximately constrain CO2 and water concentrations in the atmosphere. However, constraining surface pressures or surface temperatures, hence assess habitability, is complicated by degeneracies, as stated above. Overall, our results imply that it is easier to characterize the atmospheric scenarios of GL 581 d with transmission spectroscopy than emission spectroscopy, especially when considering the habitable scenarios with massive CO2 atmospheres. However, surface pressures cannot be inferred directly with transmission spectra because effective tangent heights are always of about of a few kilometres. Nevertheless, when combining several spectral bands and using both transmission and emission spectra, it may be possible to constrain the surface pressure, as well as both the CO2 concentrations and the presence of water. Thus, in principle, surface conditions could be assessed by means of atmospheric modeling.

4.3. Biomarkers

Although the model scenarios used in this work only consider H2O-CO2-N2 atmospheres, we can compare the computed spectra with spectral signatures of modern Earth to discuss the potential for false-positive or false-negative identifications of biomarkers.

4.3.1. False-positive detections of biomarkers

As an illustration, Fig. 8 compares high-resolution and binned (with R = 10) emission spectra of GL 581 d with those of a modern Earth. The modern Earth spectrum (taken from Grenfell et al. 2011) is shown in the upper left, the GL 581 d scenarios from this work are (clockwise) high-CO2 20 bar, high-CO2 1 bar, and the medium-CO2 20 bar case. The spectra are centered around 9.6 μm, which is the position of the main absorption band of ozone. This absorption feature is close to an absorption band of CO2 at around 9.5 μm, which is clearly seen in all GL 581 d spectra in Fig. 8.

In addition, we indicate the position of the 7.7 μm band of methane (overlapping the 7.8 μm band of nitrous oxide), which is again close to absorption bands of CO2 at 7.5 and 8 μm. The methane band is hardly discernible in the Earth spectrum, at the considered resolution of R = 10. However, in the GL 581 d scenarios shown in Fig. 8, absorption is seen, which in this case is due to CO2. These CO2 bands at biomarker positions (7.5, 8, 9.5 μm) are also present in the transmission spectra (e.g., Fig. 7), implying that transmission spectra might also suffer from the possibility of false-positive biomarker detections.

thumbnail Fig. 9

Effect of O3 isoprofiles on emission spectra. Constant O3 concentrations as indicated. Left: high-CO2 20 bar, right: Medium-CO2 20 bar. Horizontal lines represent binned spectra for a spectral resolution of 10.

Open with DEXTER

At this low spectral resolution, the spectral features around 9.6 μm, caused by CO2 in the GL 581 d cases or ozone in the Earth case, are similar. Thus, the CO2 bands could be mistaken for actual biomarker signals, at the expected low spectral resolution used for exoplanet characterization. This is the case for the emission spectra of medium and high-CO2 scenarios, whereas in the low-CO2 case (not shown in this figure), the respective CO2 bands are too weak to produce false detections. In the transmission spectra (e.g., Fig. 7), these bands also appeared for high-pressure, low-CO2 scenarios, which implies that false-positives are possible in transmission spectra even for these scenarios.

One possible way of avoiding such false-positive detections in the cases presented here (i.e., CO2-rich atmospheres) is to exploit the dual nature of the CO2 bands around 7 and 10 μm. If spectral observations are performed, e.g., at 9.5 and 10.5 μm, and both filters detect a deep absorption, then the spectral signatures are most likely due to CO2. Hence, a possible spectral characterization with respect to biomarkers should be done in all main IR CO2 bands (2−15 μm) in order to avoid false-positive detections.

4.3.2. False-negative detection of biomarkers

To investigate the possibility of false-negative detections of ozone, we calculated model spectra of the high-CO2 20 bar case and of the medium-CO2 20 bar case, where in addition we introduced artificial profiles of ozone in the line-by-line spectral calculations. These ozone profiles have not been previously implemented in the 1D climate calculations. However, they are not expected to strongly influence the temperature structure of the atmosphere (e.g., Kaltenegger et al. 2011) because of the lack of UV radiation emitted by GL 581 in the wavelength regime where ozone heating occurs.

In a first attempt, we used isoprofiles of ozone in the calculation of the spectra. Chosen concentrations were 10-8, 10-7, 10-6, and 10-5, respectively. This choice of concentrations broadly covers the range of ozone concentrations found in the present Earth atmosphere (10-8 near the surface, about to almost 10-5 in the mid-stratosphere).

The results of this sensitivity analysis are shown in Fig. 9 for the high-CO2 and medium-CO2 20 bar cases. A detection of ozone at low concentrations of 10-8 or 10-7 would clearly not be possible at low spectral resolution. For the high-CO2 20 bar case, even a concentration of 10-6 would be very challenging to detect. In both cases shown here, however, it would be possible to infer ozone levels of 10-5 since the calculated intensity drops by about a factor of 3 − 8 at the center of the ozone fundamental band.

In reality, ozone profiles will most likely not be in the form of isoprofiles. The production of ozone proceeds through the three-body reaction (2)where M is, on Earth, typically nitrogen. The production of the atomic oxygen needed in Eq. (2) requires photolysis of molecular oxygen, carbon dioxide, or water. Hence, there is a trade-off between the UV radiation necessary for photolysis and the density required to permit the three-body production reaction, which is why photochemical models generally predict a distinct maximum of atmospheric ozone concentrations around the 1 − 10 millibar pressure range.

The actual location of this ozone maximum depends on the stellar UV radiation field, the ozone and the oxygen content of the atmosphere (Selsis et al. 2002; Segura et al. 2003; Segura et al. 2007; Domagal-Goldman & Meadows 2010; Grenfell et al. 2011). Therefore, we inserted artificial ozone profiles CO3 as a function of pressure p based on a Gaussian profile (3)where Cmax is the maximum ozone concentration reached at pressure pm. Figure 10 shows the approximation of an Earth ozone profile with Eq. (3). The agreement near the maximum of the ozone layer is rather good.

thumbnail Fig. 10

Modern Earth mean ozone profile from Grenfell et al. (2011) and an approximation (in red) given by eq. 3 with Cmax = 9 ppm and pm = 6 mbar.

Open with DEXTER

Table 2

SNR values for secondary eclipse spectroscopy. .

Table 3

SNR values for transmission spectroscopy.

Figure 11 shows the effect of varying Cmax on emission spectra at constant pm of 10 mbar. We show spectra for the high-CO2 and medium-CO2 20 bar cases. In Fig. 12, the effect of varying pm on the spectrum at constant Cmax of 10 ppm is illustrated for the high-CO2 20 bar case.

Note that tropospheric concentration maxima, i.e. at pressures higher than 100 mbar, are usually not expected from atmospheric chemistry modeling. Maxima are instead found in the lower to upper stratosphere. Hence, the scenarios for pm of 100 or 1000 mbar shown in Fig. 12 are likely to overestimate the ozone column.

From Fig. 11, it is evident that it is very difficult to infer the presence of ozone even at high concentrations of about 10 ppm. Only for the profile with Cmax = 100 ppm would it be possible to detect an additional absorption besides the CO2 band.

Figure 12 shows that for a value of pm = 100 mbar or higher, it would be possible to detect ozone in the spectrum. However, as stated above, such a tropospheric maximum of ozone is not very probable.

These sensitivity studies illustrate that an inferred absence of ozone could be possible for the atmospheric scenarios considered in this work, even if ozone were present in large concentrations of 10-6 − 10-5. Hence, the possibility of false-negatives of ozone also needs to be taken into account when investigating CO2-rich, potentially habitable atmospheric scenarios.

thumbnail Fig. 11

Effect of O3 profiles on emission spectra. Parameters for Eq. (3): Cmax as indicated, pm = 10 mbar. Left: high-CO2 20 bar, right: Medium-CO2 20 bar. Horizontal lines represent binned spectra for a spectral resolution of 10.

Open with DEXTER

4.4. Detectability

The main spectral bands investigated for detectability are the 2.0, 2.7, 4.3, 7.7, 9.5, and 15 μm CO2 absorption bands as well as the 6.3 μm H2O absorption band in transmission. For secondary eclipse spectroscopy, only the 9.5 and 15 μm CO2 and the 20 μm H2O bands were considered. In addition, two filters outside broad absorption bands are calculated, namely at 8.5 and at 11.5 μm. These two filters offer the possibility to characterize surface temperatures in optically thin atmospheres.

Furthermore, six representative scenarios are considered, namely the 1 and 20 bar runs of the low, medium, and high-CO2 cases.

Table 2 summarizes the SNR values for secondary eclipse spectroscopy. It is clearly seen that, even though GL 581 is a very close star (6.27 pc, Butler et al. 2006), obtainable SNR values are very small, only reaching values of 0.3 − 0.77. The reason for the low SNR values in the mid- to far IR is the effect of the zodiacal background chosen here (see Appendix A), which reduces SNR values by about a factor of five at 20 μm and a factor of two at 15 μm compared to the photon-limited case. A similar result was found by, e.g., Belu et al. (2011).

Table 3 summarizes the SNR values for transmission spectroscopy which are mostly below unity. The exceptions are those for the 2.0, 2.7, and 4.3 μm near-IR fundamentals of CO2. In these bands, characterization of the GL 581 d scenarios, as outlined above, could be feasible. A clear distinction between, e.g., high and low-CO2 cases is still very difficult, given that the SNR values are only marginally larger than unity. One way to increase the SNR is to co-add the transit data because the SNR scales with the square root of the number of transits. Assuming a five-year mission lifetime of JWST (3 − 4 transits observable per year) would lead to SNRs of about a factor of four higher than the values shown in Tables 2 and 3.

5. Conclusions

We have presented synthetic emission and transmission spectra for a wide range of atmospheric scenarios of the potentially habitable super-Earth GL 581 d. These spectra were used as an example of possible future spectroscopic investigations of candidate habitable planets.

Water and carbon dioxide could be clearly seen in the calculated spectra because of prominent absorption bands, indicating the presence of an atmosphere.

The determination of surface temperatures has been possible for model atmospheres with either low surface pressures or low CO2 content. The potentially habitable, CO2-rich scenarios did not allow for the characterization of surface temperatures. Thus, their potential habitability could not be assessed directly from the spectra, in agreement with a model spectrum from Kaltenegger et al. (2011). The additional determination of atmospheric conditions was complicated by degeneracies between the surface pressure and the CO2 concentration. However, when combining observations in several spectral bands and using both transmission and emission spectra, inferring approximative CO2 concentrations and surface pressures would be possible.

thumbnail Fig. 12

Effect of maximum of O3 profiles on emission spectra for the high-CO2 20 bar case. Parameters for Eq. (3): Cmax = 10 ppm, pm in mbar as indicated. Horizontal lines represent binned spectra for a spectral resolution of 10.

Open with DEXTER

With currently planned telescope designs such as JWST, the typical SNR values for emission and transmission spectroscopy are, however, rather low, implying that the detection of an atmosphere of a “GL 581 d-like” transiting planet is challenging. We have been able to calculate reasonable, single-transit SNR only for three near-IR CO2 fundamentals at 2.0, 2.,7 and 4.3 μm (transmission spectroscopy). In the mid- to far-IR, thermal and zodiacal background noises inhibit SNRs above unity.

Our results indicate that the search for biomarkers in CO2-rich or high-pressure atmospheres would suffer from the possibility of false-positive detections, in agreement with previous studies (e.g., Schindler & Kasting 2000; or Selsis et al. 2002). This is due to absorption bands of CO2, which occur close to the main biomarker absorption bands. However, if the main CO2 IR bands were to be observed simultaneously, such false-positive detections could possibly be avoided. An investigation of this possibility will be the subject of future modeling including Earth-like atmospheric composition. Our results also imply that CO2 absorption bands could mask the spectroscopic features of ozone, hence produce false-negative detections. This was shown to be possible even if ozone were present in rather large concentrations of up to 10-5.

Acknowledgments

This research has been supported by the Helmholtz Association through the research alliance “Planetary Evolution and Life”. Philip von Paris and Pascal Hedelt acknowledge support from the European Research Council (Starting Grant 209622: E3ARTHs). Insightful discussions with A.B.C. Patzer and F. Selsis are gratefully acknowledged. We thank the anonymous referee for his/her constructive remarks, which helped to improve the paper.

References

Appendix A: Noise contributions

The total SNR Sp of a planetary feature is calculated with (A.1)where Fp is the planetary signal measured during secondary eclipse (emitted and reflected spectrum) or primary eclipse (additional transit depth due to atmospheric absorption) and σtotal represents the total noise of the measurement. In addition to the stellar noise σS, the main contributions to the noise budget come from the zodiacal emission of the solar system (denoted σzodi), the thermal emission of the mirror and, in the case of the JWST, the sun shade (denoted σthermal). Both components need to be taken into account in the SNR calculations since their effect on detectability could be significant, especially in the mid-IR, longwards of 10 μm (e.g., Deming et al. 2009; Belu et al. 2011). In addition, the dark noise σd is considered here. Hence, we have (A.2)The thermal noise was obtained from (A.3)where ϵ is the emissivity, Bλ(TJWST) the blackbody emission of the telescope and sun shade, tt the integration time (here assumed to be the transit duration, i.e. 4.15 h), npx the number of pixels used for the integration on the detector, λ the considered wavelength and R the spectral resolution, AE the emitting area, qe = 0.15 the quantum efficiency of the telescope (e.g., Kaltenegger & Traub 2009; Rauer et al. 2011), and SA the solid angle per pixel.

We assumed a telescope and sun shade temperature TJWST of 45 K with an emissivity of ϵ = 0.15 (Belu et al. 2011) and a total emitting surface of sun shade and mirror combined of AE = 240 m2 (Nella et al. 2002). The number of pixels npx was calculated to be (A.4)where nspatial and nspectral the number of pixels in spatial and spectral (per μm) direction, respectively. We assumed nspectral = 32.33 pixel μm-1, mimicking the MIRI instrument in the 5 − 11 μm spectral range (Belu et al. 2011). The number of pixels in the spatial direction is obtained from (A.5)with dt the telescope diameter (i.e., 6.5 m in the case of JWST) and ps = 0.11′′ per pixel the instrumental pixel scale. The number calculated from Eq. (A.5) is then rounded up to the closest multiple of 4. The solid angle per pixel is calculated with (1 sq. deg  sr) (A.6)The zodiacal emission IZ (given in W m-2 μm-1 sr-1) used in the noise calculations is taken from one example measurement in the ecliptic plane presented by Kelsall et al. (1998, their Fig. 9). Note that this choice is a rather pessimistic assumption because some potential targets will probably be located towards higher ecliptic latitudes, hence the corresponding zodiacal noise would be somewhat smaller. We calculate the zodiacal noise via (A.7)where Atel is the telescope area, assuming a circular aperture of 6.5 m.

The dark noise contribution to the noise budget is calculated from (A.8)where dc = 0.03 e pixel-1 s-1 is the dark current (MIRI, see e.g. Deming et al. 2009; or Belu et al. 2011).

All Tables

Table 1

Atmospheric scenarios for GL 581 d considered here.

Table 2

SNR values for secondary eclipse spectroscopy. .

Table 3

SNR values for transmission spectroscopy.

All Figures

thumbnail Fig. 1

Spectra of the 20 bar high-CO2 case. Left: secondary eclipse spectrum. Inlet shows the corresponding temperature-pressure profile. Right: transit depth spectrum.

Open with DEXTER
In the text
thumbnail Fig. 2

Same as Fig. 1, but for the 1 bar low-CO2 case.

Open with DEXTER
In the text
thumbnail Fig. 3

Brightness temperature spectra (based on planetary emission only) of the 1 and 20 bar scenarios. Left: low-CO2 case. Right: high-CO2 case. Surface temperatures are indicated by horizontal dashed lines.

Open with DEXTER
In the text
thumbnail Fig. 4

Comparison of secondary eclipse spectra to study the effect of varying surface pressure. Scenarios are as indicated. Inlets show the corresponding temperature-pressure profiles.

Open with DEXTER
In the text
thumbnail Fig. 5

As Fig. 4, but for transmission spectra instead of secondary eclipse spectra.

Open with DEXTER
In the text
thumbnail Fig. 6

Comparison of secondary eclipse spectra to study the effect of varying CO2 concentration. Scenarios are as indicated. Inlets show the corresponding temperature-pressure profiles.

Open with DEXTER
In the text
thumbnail Fig. 7

Same as Fig. 6, but for transmission spectra instead of secondary eclipse spectra.

Open with DEXTER
In the text
thumbnail Fig. 8

Possibility of a false-positive biomarker detection for different scenarios, as indicated. Modern Earth spectrum taken from Grenfell et al. (2011). Planck curves of corresponding surface temperatures as dashed lines. Smoothed (red) and binned spectra (blue) are for a spectral resolution of 10.

Open with DEXTER
In the text
thumbnail Fig. 9

Effect of O3 isoprofiles on emission spectra. Constant O3 concentrations as indicated. Left: high-CO2 20 bar, right: Medium-CO2 20 bar. Horizontal lines represent binned spectra for a spectral resolution of 10.

Open with DEXTER
In the text
thumbnail Fig. 10

Modern Earth mean ozone profile from Grenfell et al. (2011) and an approximation (in red) given by eq. 3 with Cmax = 9 ppm and pm = 6 mbar.

Open with DEXTER
In the text
thumbnail Fig. 11

Effect of O3 profiles on emission spectra. Parameters for Eq. (3): Cmax as indicated, pm = 10 mbar. Left: high-CO2 20 bar, right: Medium-CO2 20 bar. Horizontal lines represent binned spectra for a spectral resolution of 10.

Open with DEXTER
In the text
thumbnail Fig. 12

Effect of maximum of O3 profiles on emission spectra for the high-CO2 20 bar case. Parameters for Eq. (3): Cmax = 10 ppm, pm in mbar as indicated. Horizontal lines represent binned spectra for a spectral resolution of 10.

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.