The search for signs of life implies that one understands how the observed atmosphere physically and chemically works. Knowledge of the temperature and planetary radius is crucial for the general understanding of the physical and chemical processes occurring on the planet (e.g. tectonic, hydrogen loss to space). This question is far from being easy. In theory, spectroscopy can provide some detailed information on the thermal profile of a planetary atmosphere. This, however, requires a spectral resolution and a sensitivity that are well beyond the performance of a first generation spacecraft. Thus we will concentrate on the initially available observations here.

One can calculate the stellar energy of the star F_star that is received at the measured orbital distance. The surface temperature of the planet depends on its albedo and on the greenhouse warming by atmospheric compounds. However, with a low-resolution spectrum of the thermal emission, the mean effective temperature and the radius of the planet can be obtained. The surface temperature can be estimated by adjusting thermal emission to a model. The ability to associate a brightness temperature to the spectrum relies on the existence and identification of spectral windows probing the surface or the same atmospheric levels. For an Earth-like planet there are some atmospheric windows that can be used in most of the cases, especially between 8 and 11 μm as seen in Fig. 1. This window would however become opaque at high H₂O partial pressure (e.g. the inner part of the Habitable Zone [HZ] where a lot of water is vaporized) and at high CO₂ pressure (e.g. a very young Earth [Fig. 3] or the outer part of the HZ). The accuracy of the radius and temperature determination will depend on the quality of the fit (and thus on the sensitivity and resolution of the spectrum), the precision of the Sun-star distance, the cloud coverage and also the distribution of brightness temperatures over the planetary surface. For transiting planets, for which the radius is known, the measured IR flux can directly be converted into a brightness temperature that will provide information on the temperature of the atmospheric layers responsible for the emission. If the mass of non-transiting planets can be measured (by RV and/or astrometric observations), an additional estimate of the radius can be made by assuming a bulk composition of the planet, which can then be used to convert IR fluxes into temperatures.

The orbital flux variation in the IR can distinguish planets with and without an atmosphere in the detection phase [17] and [56]. Strong variation of the thermal flux with phase reveals a strong difference in temperature between the day and night hemisphere of the planet, a consequence of the absence of a dense atmosphere. In such a case, estimating the radius from the thermal emission is made difficult because most of the flux received comes from the small and hot substellar area. The ability to retrieve the radius in such a case would depend on the assumption that can be made on the orbit geometry and the rotation rate of the planet. For instance, if the rotation is likely to have been tidally synchronized, and the inclination of the system is known, then the temperature variations from the substellar point to the terminator can be modeled accurately and the radius can be derived. However, in most of the other cases, degenerate solutions will exist. When the mean brightness temperature (T_b) is stable along the orbit, the estimated radius is more reliable. The radius can be measured at different points of the orbit and thus for different values of T_b, which should allow one to estimate the error made. The thermal light curve (i.e. the integrated infrared emission measured at different position on the orbit) exhibits smaller variations due to the phase (whether the observer sees mainly the day side or the night side) and to the season. Important phase-related variations are due to a high day/night temperature contrast and imply a low greenhouse effect and the absence of a stable liquid ocean. Therefore, habitable planets can be distinguished from airless or Mars-like planets by the amplitude of the observed variations of T _b. Note that also a Venus-like exoplanet would exhibit nearly no measurable phase-related variations of its thermal emission, due to the fast rotation of its atmosphere and its strong greenhouse effect, and can only be distinguished through spectroscopy from habitable planets. The mean value of T _b estimated over an orbit can be used to estimate the albedo of the planet, A, through the balance between the incoming stellar radiation and the outgoing IR emission. In the visible ranges, the reflected flux allows us to measure the product A × R ², where R is the planetary radius (a small but reflecting planet appears as bright as a big but dark planet). The first generation of optical instruments will be very far from the angular resolution required to directly measure an exoplanet radius. Presently, such a measurement can only be performed when the planet transits in front of its parent star, by an accurate photometric technique. If the secondary eclipse of the transiting planet can be observed (when the planet passes behind the star), then the thermal emission of the planet can be measured, allowing the retrieval of mean brightness temperature T _b thanks to the knowledge of the radius from the primary transit. If a non-transiting target is observed in both visible and IR ranges, the albedo can be estimated in the visible once the radius is inferred from the IR spectrum, and compared with one derived from the thermal emission only.

Owen [39] suggested searching for O₂ as a tracer of life. Oxygen in high abundance is a promising bio-indicator. Oxygenic photosynthesis, which by-product is molecular oxygen extracted from water, allows terrestrial plants and photosynthetic bacteria (cyanobacteria) to use abundant H₂O, instead of having to rely on scarce supplies of electron donor to reduce CO₂, like H₂ and H₂S. With oxygenic photosynthesis, the production of the biomass becomes limited only by nutriments and no longer by energy (light in this case) nor by the abundance of electron donors. Oxygenic photosynthesis at a planetary scale results in the storage of large amounts of radiative energy in chemical energy, in the form of organic matter. For this reason, oxygenic photosynthesis had a tremendous impact on biogeochemical cycles on Earth and eventually resulted in the global transformation of Earth environment. Less than 1 ppm of atmospheric O₂ comes from abiotic processes [67]. Cyanobacteria and plants are responsible for this production by using the solar photons to extract hydrogen from water and using it to produce organic molecules from CO₂. This metabolism is called oxygenic photosynthesis. The reverse reaction, using O₂ to oxidize the organics produced by photosynthesis, can occur abiotically when organics are exposed to free oxygen, or biotical by eukaryotes breathing O₂ and consuming organics. Because of this balance, the net release of O₂ in the atmosphere is due to the burial of organics in sediments. Each reduced carbon buried lets a free O₂ molecule into the atmosphere. This net release rate is also balanced by weathering of fossilized carbon when exposed to the surface. The oxidation of reduced volcanic gasses such as H₂, H₂S also accounts for a significant fraction of the oxygen losses. The atmospheric oxygen is recycled through respiration and photosynthesis in less than 10,000 yrs. In the case of a total extinction of the Earth biosphere, the atmospheric O₂ would disappear in a few million years.

Reduced gases and oxygen have to be produced concurrently to be detectable in the atmosphere, as they react rapidly with each other. Thus, the chemical imbalance traced by the simultaneous signature of O₂ and/or O₃ and of a reduced gas like CH₄ can be considered as a signature of biological activity [33]. The spectrum of the Earth has exhibited a strong infrared signature of ozone for more than 2 billion years, and a strong visible signature of O₂ for an undetermined period of time between 2 and 0.8 billion years (depending on the required depth of the band for detection and also the actual evolution of the O₂ level) [25]. This difference is due to the fact that a saturated ozone band appears already at very low levels of O₂ (10⁻⁴ ppm) while the oxygen line remains unsaturated at values below one PAL [52]. In addition, the stratospheric warming decreases with the abundance of ozone, making the O₃ band deeper for an ozone layer less dense than in the present atmosphere. The depth of the saturated O₃ band is determined by the temperature difference between the surface-clouds continuum and the ozone layer. Note again that the non-detection of O₂ or O₃ on an exoplanet cannot be interpreted as the absence of life.

N₂O is produced in abundance by life but only in negligible amounts by abiotic processes. Nearly all of Earth's N₂O is produced by the activities of anaerobic denitrifying bacteria. N₂O would be hard to detect in the Earth's atmosphere with low resolution, as its abundance is low at the surface (0.3 ppmv) and falls off rapidly in the stratosphere. Spectral features of N₂O would become more apparent in atmospheres with more N₂O and/or less H₂O vapor. Segura et al. [52] have calculated the level of N₂O for different O₂ levels and found that, although N₂O is a reduced species compared to N₂, its levels decreases with O₂. This is due to the fact that a decrease in O₂ produces an increase of H₂O photolysis resulting in the production of more hydroxyl radicals (OH) responsible for the destruction of N₂O. The detection of H₂O and CO₂, not as biosignatures themselves, are important in the search for signs of life because they are raw materials for life and thus necessary for planetary habitability. The methane found in the present atmosphere of the Earth has a biological origin, except for a small fraction produced abiotically in hydrothermal systems, where hydrogen is released by the oxidation of Fe by H₂O, and reacts with CO₂. Depending on the degree of oxidation of a planet's crust and upper mantle, such non-biological mechanisms can also produce large amounts of CH₄ under certain circumstances. Therefore, the detection of methane alone cannot be considered a sign of life, while its detection in an oxygen-rich atmosphere would be difficult to explain in the absence of a biosphere. Note that methane may have been detected on Mars [38], while the atmosphere of Mars contains 0.1% of O₂ and some ozone. In this case, the amounts involved are extremely low and the origin of the Martian O₂ and O₃ is known to be photochemical reactions initiated by the photolysis of CO₂ and water vapor. If confirmed, the presence of methane could be explained by subsurface geochemical processes, assuming that reducing conditions exist on Mars below the highly oxidized surface. The case of NH₃ is similar to that of CH₄. They are both released into Earth's atmosphere by the biosphere with similar rates but the atmospheric level of NH₃ is orders of magnitude lower due to its very short lifetime under UV irradiation. The detection of NH₃ in the atmosphere of a habitable planet would thus be extremely interesting, especially if found with oxidized species. There are other molecules that could, under some circumstances, act as excellent biomarkers, e.g., the manufactured chloro-fluorocarbons (CCl₂F₂ and CCl₃F) in our current atmosphere in the thermal infrared waveband, but their abundances are too low to be spectroscopically observed at low resolution.

We need to address the abiotic sources of biomarkers, so that we can identify when it might constitute a “false positive” for life. CH₄ is an abundant constituent of the cold planetary atmospheres in the outer solar system. On Earth, it is produced abiotically in hydrothermal systems where H₂ (produced from the oxidation of Fe by water) reacts with CO₂ in a certain range of pressures and temperatures. In the absence of atmospheric oxygen, abiotic methane could build up to detectable levels. Therefore, the sole detection of CH₄ cannot be attributed unambiguously to life.

O₂ also has abiotic sources: the first one is the photolysis of CO₂, followed by recombination of O atoms to form O₂ (O + O + M → O₂ + M); a second one is the photolysis of H₂O followed by escape of hydrogen to space. The first source is a steady state maintained by the stellar UV radiation, but with a constant elemental composition of the atmosphere while the second one is a net source of oxygen. In order to reach detectable levels of O₂ (in the reflected spectrum), the photolysis of CO₂ has to occur in the absence of outgassing of reduced species and in the absence of liquid water, because of the wet deposition of oxidized species. Normally, the detection of the water vapor bands simultaneously with the O₂ band can rule out this abiotic mechanism [52], although one should be careful, as the vapor pressure of H₂O over a high-albedo icy surface might be high enough to produce detectable H₂O bands. In the infrared, this process cannot produce a detectable O₃ feature [57]. The loss of hydrogen to space can result in massive oxygen leftovers: more than 200 bars of oxygen could build up after the loss of the hydrogen contained in the Earth Ocean. However, the case of Venus tells us that such oxygen leftover has a limited lifetime in the atmosphere (because of the oxidation of the crust and the loss of oxygen to space): we do not find O₂ in the Venusian atmosphere despite the massive loss of water probably experienced in the early history of the planet. Also, such evaporation-induced build up of O₂ should occur only closer to a certain distance from the Star and affect small planets with low gravity more dramatically. For small planets (< 0.5 M_Earth) close to inner edge of the HZ (< 0.93 AU from the present Sun), there is a risk of abiotic oxygen detection, but this risk becomes negligible for big planets further away from their star. The fact that, on the Earth, oxygen and indirectly ozone are by-products of the biological activity does not mean that life is the only process able to enrich an atmosphere with these compounds. The question of the abiotic synthesis of biomarkers is crucial, but only very few studies have been dedicated to it [32], [44], [54], [57] and [62].

While they efficiently absorb the visible light, photosynthetic plants have developed strong infrared reflection (possibly as a defense against overheating and chlorophyll degradation), resulting in a steep change in reflectivity around 700 nm, called the red-edge. The primary molecules that absorb the energy and convert it to drive photosynthesis (H₂O and CO₂ into sugars and O₂) are chlorophyll A (0.450 μm) and B (0.680 μm). The exact wavelength and strength of the spectroscopic “vegetation red edge” (VRE) depends on the plant species and environment. On Earth around 440 million years ago [42] and [48], an extensive land plant cover developed, generating the red chlorophyll edge in the reflection spectrum between 700 and 750 nm. Averaged over a spatially unresolved hemisphere of Earth, the additional reflectivity of this spectral feature is typically only a few percent [25] and [36]. Several groups [1], [9], [37], [63] and [68] have measured the integrated Earth spectrum via the technique of Earthshine, using sunlight reflected from the non-illuminated, or “dark”, side of the moon. Earthshine measurements have shown that detection of the Earth's VRE is feasible if the resolution is high and the cloud coverage is known, but is made difficult owing to its broad, essentially featureless spectrum and cloud coverage [36]. Our knowledge of the reflectivity of different surface components on Earth – such as deserts, ocean and ice – helps in assigning the VRE of the Earthshine spectrum to terrestrial vegetation. The Earth's hemispherical integrated vegetation red-edge signature is weak, but planets with different rotation rates, obliquities, land-ocean fraction, and continental arrangement may have lower cloud-cover and higher vegetated fraction [50]. Knowing that other pigments exist on Earth and that some minerals can exhibit a similar spectral shape around 750 nm [49], the detection of the red-edge of the chlorophyll on exoplanets, despite its interest, will not be unambiguous. Assuming that similar photosynthesis would evolve on a planet around other stellar types, possible different types of spectral signature have been modeled that could be a guide to interpreting other spectral signatures. These signatures will be difficult to verify through remote observations as being of biological origin.

Another topic that has been proposed to discover continents and seas on an exoplanet is the daily variation of the surface albedo in the visible [14] and [49]. On a cloud-free Earth, the diurnal flux variation in the visible caused by different surface features rotating in and out of view could be high, assuming hemispheric inhomogeneity. When the planet is only partially illuminated, a more concentrated signal from surface features could be detected as they rotate in and out of view on a cloudless planet. Earth has an average of 60% cloud coverage that prevents easy identification of the features without knowing the cloud distribution. Clouds are an important component of exoplanet spectra because their reflection is high and relatively flat with wavelength. Clouds hide the atmospheric molecular species below them, weakening the spectral lines in both the thermal infrared and visible [25] and [36]. In the thermal infrared, clouds emit at temperatures that are generally colder than the surface, while in the visible the clouds themselves have different spectrally-dependent albedo that further influence the overall shape of the spectrum. Clouds reduce the relative depths, full widths, and equivalent widths of spectral features. If one could record the planet's signal with a very high time resolution (a fraction of the rotation period of the planet) and SNR, one could determine the overall contribution of clouds to the signal [11] and [40]. During each of these individual measurements, one has to collect enough photons for a high individual SNR per measurement to be able to correlate the measurements to the surface features, which precludes this method for first generation missions that will observe a minimum of several hours to achieve a SNR of 5 to 10. For the Earth [11] and [40], these measurements show a correlation to the Earth's surface features because the individual measurements are time resolved as well as having an individual high SNR, making it a very interesting concept for future generations of missions.

On the Earth, photosynthetic organisms are responsible for the production of nearly all of the oxygen in the atmosphere. However, in many regions of the Earth, and particularly where surface conditions are extreme, for example, in hot and cold deserts, photosynthetic organisms can be driven into and under substrates where light is still sufficient for photosynthesis. These communities exhibit no distinguishable detectable surface spectral signature. The same is true of the assemblages of photosynthetic organisms at more than a few meters depth in bodies of water. These communities are widespread and dominate local photosynthetic productivity. Fig. 2 shows known cryptic photosynthetic communities and the calculated disk-averaged spectra of such hypothetical cryptic photosynthesis worlds. Such worlds are Earth-analogs that would not exhibit a distinguishable biological surface feature like the VRE in the disc-averaged spectrum but still be inhabited, and remotely detectable [10].

One crucial factor in interpreting planetary spectra is the point in the evolution of the atmosphere when its biomarkers and its habitability become detectable. Spectra of the Earth exploring temperature sensitivity (hot house and cold scenario) and different singled out stages of its evolution [41] and [47], as well as exploring the evolution of the expected spectra of Earth [25], produce a variety of spectral fingerprints for our own planet. These spectra will be used as part of a big grid to characterize any exoplanets found and influence the design requirements for a spectrometer to detect habitable planets.

The spectrum of the Earth has not been static throughout the past 4.5 Ga. This is due to the variations in the molecular abundances, the temperature structure, and the surface morphology over time. At about 2.3 Ga oxygen and ozone became abundant, affecting the atmospheric absorption component of the spectrum. At about 0.44 Ga, an extensive land plant cover followed, generating the red chlorophyll edge in the reflection spectrum. The composition of the surface (especially in the visible), the atmospheric composition, and temperature-pressure profile can all have a significant influence on the detectabilty of a signal. Fig. 3 shows theoretical visible and mid-infrared spectra of the Earth at six epochs during its geological evolution [25] and [28]. The epochs are chosen to represent major developmental stages of the Earth, and life on Earth. If an extrasolar planet is found with a corresponding spectrum, we can use the stages of evolution of our planet to characterizing it, in terms of habitability and the degree to which it shows signs of life. Furthermore, we can learn about the evolution of our own planet's atmosphere and possible the emergence of life by observing exoplanets in different stages of their evolution. The Earth's atmosphere has experienced a dramatic evolution over 4.5 billion years, and other planets may exhibit similar or greater evolution, and at different rates. It shows epochs that reflect significant changes in the chemical composition of the atmosphere. The oxygen and ozone absorption features could have been used to indicate the presence of biological activity on Earth anytime during the past 50% of the age of the solar system. Different signatures in the atmosphere are clearly visible over Earth's evolution and observable with low resolution.

The range of characteristics of planets is likely to exceed by far our experience with the planets and satellites in our own Solar System. For instance, models of planets more massive than our Earth – rocky SuperEarths – need to consider the changing atmosphere structure, as well as the interior structure of the planet [51] and [65]. Also, Earth-like planets orbiting stars of different spectral type might evolve differently [19], [23], [52], [53] and [55]. Modeling these influences will help to optimize the design of the proposed instruments used to search for Earth-like planets. The spectral resolution required for optimal detection of habitability and biosignatures has to be able to detect those features on our own planet for the dataset we have over its evolution.

Using a numerical code that simulates the photochemistry of a wide range of planetary atmospheres, several groups have simulated a replica of our planet orbiting different types of stars: an F-type star (more massive and hotter than the Sun) and a K-type star (smaller and cooler than the Sun) (see Fig. 4). The models assume same background composition of the atmosphere as well as the strength of biogenic sources.

A planet orbiting a K star has a thin O₃ layer, compared to Earth's one, but still exhibits a deep O₃ absorption: indeed, the low UV flux is absorbed at lower altitudes than on Earth which results in a less efficient warming (because of the higher heat capacity of the dense atmospheric layers). Therefore, the ozone layer is much colder than the surface and this temperature contrast produces a strong feature in the thermal emission. The process works the other way around in the case of an F-type host star. Here, the ozone layer is denser and warmer than the terrestrial one, exhibiting temperatures about as high as the surface temperature. Thus, the resulting low temperature contrast produces only a weak and barely detectable feature in the infrared spectrum.

This comparison shows that planets orbiting G (solar) and K-type stars may be better candidates for the search for the O₃ signature than planets orbiting F-type stars. This result is promising since G and K-type stars are much more numerous than F-type target stars [26], the latter being rare and affected by a short lifetime (less than 1 Gyr).