ABIOTIC OZONE AND OXYGEN IN ATMOSPHERES SIMILAR TO PREBIOTIC EARTH

To investigate the effects of stellar energy distribution, bulk atmospheric composition, and boundary conditions on O₃ production for an abiotic terrestrial CO₂-H₂O-N₂ atmosphere, we used a photochemical model of prebiotic Earth. This code has been previously used to study abiotic Earth (Kharecha et al. 2005), as well as extrasolar terrestrial planets with biogenic gas fluxes around stars other than the Sun (Domagal-Goldman et al. 2011; Segura et al. 2003, 2005, 2007; Misra et al. 2014). Results from the photochemical code that exhibited potentially detectable biosignature gases were used as inputs to the Spectral Mapping Atmospheric Radiative Transfer (SMART) model (developed by D. Crisp), which produced simulated spectra for these model planets. SMART has been validated against observations of Earth, Mars, and Venus (Meadows & Crisp 1996, Crisp 1997; Halthore et al. 2005; Robinson et al. 2011; Robinson et al. 2014) and has been used to simulate spectra of extrasolar planets (Kiang et al. 2007; Domagal-Goldman et al. 2011; Robinson 2011).

The photochemical model computes the chemical equilibrium for 38 chemical species (Table 1) involved in 162 reactions, including but not limited to reactions (R1–R6). The atmosphere is divided into 100 layers with a fixed vertical extent of 1 km each. The long-lived chemical species are O, O₂, O₃, H₂O, H, OH, HO₂, H₂O₂, H₂, CO, CO₂, HCO, H₂CO, CH₄, CH₃, C₂H₆, NO, NO₂, HNO, SO, SO₂, and H₂SO₄; N₂ is also included with a constant mixing ratio. The input stellar flux for the model includes the interval from 130 nm to 855 nm, as well as the Lyα flux at 121 nm. Detailed model descriptions can be found in Kharecha et al. (2005), Segura et al. (2007), and Guzmán-Marmomejo et al. (2013).

No photosynthetic source of oxygen is included in our photochemical model; the boundary conditions instead reflect a planet devoid of life (Table 1; Section 2.2). Only H, CO, O, and CO₂ were allowed to cross the top layer of the atmosphere; all other species had a "0 flux" boundary condition at the top of the uppermost layer of the atmosphere. H was allowed to flow up and out through the top of the atmosphere, in accordance with diffusion-limited escape of H to space. CO₂ flow past the top layer of the model was converted to an equal influx of CO and O, consistent with CO₂ photolysis above the top of our model grid. At the lower boundary, CH₄ was assigned a surface flux from abiotic production in the oceans resulting from serpentinization, and H₂ was given both a volcanic outgassing rate and a fixed deposition velocity into the oceans. The CO₂ surface mixing ratio and H₂ outgassing rate varied between different simulations to cover a variety of planetary conditions (Figure 2). For all other species, we assumed reaction with the surface or dissolution into oceans at rates proportional to the solubility of each species (Kharecha et al. 2005). For soluble species, we allowed the dissolution rates to go down from their dissolution-limited rates in order to maintain ocean redox balance, as explained in the next section.

Zoom In Zoom Out Reset image size

Redox balance between oxidants and reductants in the surficial environment should be the primary driver of the atmospheric chemistry of a potentially habitable but lifeless world. Any redox imbalance in the surface environment of a planet will lead to changes in reaction rates or removal of species that would drive the system back toward a dynamic, redox-balanced equilibrium. For example, if a planetary surface had an excess of oxidants, it would lead to greater rates of reactions that destroy these oxidants, or burial of minerals that would remove these oxidants from the surface. Through either pathway, their concentrations would be decreased, thereby driving the system back toward redox balance. As a result, any such imbalance would be short-lived and less likely to be observed. Conversely, a redox-balanced simulation represents a point of "dynamic equilibrium" that would be the most likely state for a planet with the given outgassing rates, bulk composition, and physical properties. Note that this dynamic equilibrium is not the same thing as chemical equilibrium due to the energy input from stars, but instead represents a steady-state atmospheric composition.

The one thing that can change the redox balance of an Earth-mass planet is hydrogen escape to space, which can irreversibly change the net redox inventory of the surface and near-surface environment (see Catling 2014 for a review of this concept, and on redox controls in Earth's history). (For smaller planets, escape of other species, such as C and O, could also affect redox balance.) However, on timescales less than the hundreds of millions of years required for the escape process to affect the bulk redox of the surface and upper mantle, the surface and atmospheric redox state will be buffered to match the redox state of the mantle. In other words, escape can slowly change the redox state of the planet, but at any given point in the planet's evolution, the redox states of the atmosphere, the ocean, and the atmosphere–ocean system should be balanced. The simulations presented here can be considered representations of a planet in such a steady state, with models that have smaller contributions of reduced species from the subsurface simulating planets that are further along in their oxidation via atmospheric H escape.

Redox balance in this model can be tracked in units of "H₂ reducing power," measured by the amount of H₂ required to convert the redox-neutral gases (defined in our calculations as H₂O, CO₂, N₂, and SO₂ for H and O, C, N, and S, respectively) into the species in question. For example, atmospheric CH₄ can be produced by the representative reaction CO₂ + 4H₂ → CH₄ + 2H₂O; therefore, the addition of one CH₄ molecule to an atmosphere is equivalent to the addition of four molecules of H₂. We can use this scheme to tabulate the sources and sinks of H₂ for each species in the atmosphere (Table 1), in the oceans, and across the ocean–atmosphere boundary.

In past work (e.g., Segura et al. 2007), the focus was on the redox balance in the atmosphere, between the volcanic outgassing of reduced species, escape of hydrogen (a reduced species) from the top of the atmosphere, the flux of methane (a reduced species) from the ocean into the atmosphere, and any redox imbalance in the deposition of gases to the surface and oceans. This can be expressed in an equation showing the balance of sources and sinks of reductants:

$$\begin{equation} F_{{\rm volc}} \; + \;F_{{\rm CH_{4}}} \; + \;F_{{\rm ox\; dep}} = F_{{\rm esc}} \; + \;F_{{\rm red\; dep}}, \end{equation} \tag{ E1 }$$

where F_volc is the flux of reduced volcanic species, $F_{\rm CH_4}$ is the flux of methane from the oceans, F_{ox dep} is the deposition flux of oxidized species (removal of oxidants is equivalent to addition of reductants), F_esc is the rate of escape of reductants out the top of the atmosphere, and F_{red dep} is the rate of deposition of reductants. All of these are tracked in units of "H₂ reducing power" as explained in the preceding paragraph.

However, a more careful redox balance also must include reactions that occur in the oceans into which the atmospheric species are being deposited. We assume that reactions with land surfaces get propagated to the oceans through rivers, so all species deposited at the surface end up in this ocean budget. With this assumption, the redox balance is between the flux of reduced species from the atmosphere, the flux of oxidized species from the atmosphere, the flux of methane to the atmosphere, and any redox imbalance at the ocean floor, including deposition of oxidized species such as iron oxides, deposition of reduced species such as organic carbon, and the flux of reduced species such as Fe²⁺ into ocean water from the subsurface. The net reaction for ocean redox balance is therefore (again measured in units of "H₂ reducing power")

$$\begin{equation} F_{{\rm red\; dep}} \; + \;F_{{\rm floor}} = F_{{\rm ox\; dep}} \; + \;F_{{\rm CH_{4}}}, \end{equation} \tag{ E2 }$$

where F_floor is the redox imbalance at the ocean floor and F_{red dep}, F_{ox dep}, and F_CH4 represent the same values as they do in Equation (E1).

F_floor is effectively a free model parameter, for which a value must be chosen. We used values from 3 × 10⁹ to 3 × 10¹¹ molecules cm⁻² s⁻¹, using estimates for the iron oxide deposition rate on early Earth (5 × 10¹¹ molecules cm⁻² from Holland 1984) as an upper limit, with lower values simulating more oxidized planets with a higher propensity for accumulation of abiotic O₂ and O₃. These lower values of F_floor may be appropriate for planets with an oxidized bulk composition (compared to Earth), or for water-rich planets with high enough seafloor pressures that an ice VII layer forms at the ocean floor, potentially decreasing the flux of reduced mantle materials (Léger et al. 2004).

The photochemical model automatically ensures that Equation (E1) is satisfied, for a given set of boundary conditions (Kharecha et al. 2005). To ensure that Equation (E2) was additionally satisfied for all our simulations, we created a script that changes the model's boundary conditions between simulations and repeatedly runs the photochemical code until Equation (E2) is satisfied with the given value of F_floor.

To achieve this, we used a "piston velocity" treatment of the diffusion of gases into the ocean (Kharecha et al. 2005). This treatment traces its origins to the biogeochemical work of Broecker et al. (1982) that describes and explains the behavior of elements in Earth's oceans. Under this treatment, it is assumed that the rate at which gases diffuse into the ocean can be approximated with a conceptual model of a piston pushing gases through a thin layer into the ocean. The velocity of that piston for a given gas, X, is given by v_p(X) = K_diff(X)/z_film, where v_p(X) is the piston velocity of gas X, K_diff(X) is the thermal diffusivity of gas X, and z_film is an empirically determined surface layer thickness (40 μm) that is the same for all gases. This approach is used to approximate multiple, complex phenomena, including wave breaking, bubble formation, and molecular diffusion. It yields an equation for the rate at which gas X dissolves into the ocean:

$$\begin{equation} \phi ({\rm X}) = v_p ({\rm X}) \cdot (\alpha ({\rm X}) \cdot pX - [X]_{{\rm aq}})\cdot C, \end{equation} \tag{ E3 }$$

where ϕ(X) is the rate at which gas X flows into the ocean, v_p(X) is the piston velocity for gas X, α(X) is the solubility of gas X in water, pX is the partial pressure of gas X at the bottom of the atmosphere, [X]_aq is the concentration of gas X at the top of the ocean, and C is a conversion factor (6.02 × 10²⁰ molecules cm⁻³ mol⁻¹ L).

At the start of each set of runs, we assumed that the ocean was empty of all soluble species ([X]_aq = 0), which gives the maximum rate at which any species can diffuse into the oceans:

$$\begin{equation} \phi _{\max } ({\rm X}) = v_p ({\rm X}) \cdot \alpha ({\rm X}) \cdot pX \cdot C. \end{equation} \tag{ E4 }$$

If Equation (E2) was not satisfied, we ran the photochemical model again, assuming that the ocean would start to "fill up" with the soluble species contributing the most to the model ocean's redox imbalance: if the ocean was too oxidized, we lowered the deposition velocity of the gas contributing the most to the ocean's oxidizing power; conversely, if the ocean was too reduced, we lowered the deposition velocity of the gas contributing the most to the ocean's reducing power. We set a minimum diffusion velocity of 0 for all soluble species; once a species reached this minimum deposition rate, the next-greatest contributor to the redox balance was altered. This approach is consistent with [X]_aq increasing from its minimum value of [X]_aq = 0 up to its maximum value of [X]_aq = α(X)·pX. Higher concentrations of X require production of X in the oceans, and (except CH₄, which is discussed below) this case is reserved for systems with biological production of X.

Highly reactive species (H, OH, O, HCO, CH₃, HNO) were excluded from this process, as these species would rapidly react and never become saturated in the ocean, so we assumed that the ocean is always empty of these species. Two additional species (SO₂ and H₂SO₄) are highly soluble, so we assumed that the ocean would not fill with them either. That left nine species whose deposition velocities were allowed to change between model iterations: O₂, H₂O₂, O₃, H₂, CO, H₂CO, NO, NO₂, and SO. If changes to the deposition rates of these species affected the net redox state of the ocean by less than 1%, we instead altered the CH₄ flux from the ocean to the atmosphere, within bounds of 0 and 6.8 × 10⁸ molecules cm⁻² s⁻¹, the estimate for Earth's global prebiotic CH₄ flux from serpentinization (Guzmán-Marmolejo et al. 2013). We also ran simulations where all 17 species were allowed to vary as if the ocean could fill up with them. Doing this affected the quantitative predictions of O₂ and O₃ concentrations, but the effect of these changes was not large enough to qualitatively impact the absorption features that could be remotely observable.

Although it is theoretically possible for higher CH₄ fluxes to originate from a planet with a different composition, accurately calculating them requires a separate geochemical model that currently does not exist. We repeated the process of running the model and iterating on the boundary conditions until the net ocean–atmosphere system obtained redox balance. The criterion we used for this "global" redox balance was that there was not a net reduced flux to the oceans, and that any net oxidant flux to the oceans was smaller than that simulation's proscribed limit on F_floor. This is equivalent to assuming that the mantle provides the ocean with some nonzero, finite flux of reductants that can serve as a sink for oxidants from the atmosphere.

To study the effects of the bulk atmospheric composition, we varied the CO₂ surface mixing ratio from 3 × 10⁻⁴ (300 ppm, close to the modern Earth's value) to 5 × 10⁻¹ (50% of the atmosphere) and varied the H₂ outgassing rate from 2 × 10⁶ to 2 × 10¹² molecules cm⁻² s⁻¹, bracketing the modern value of 2 × 10⁹ molecules cm⁻² s⁻¹ by three orders of magnitude. We parameterized F_floor, the flux rate of reduced species (such as Fe²⁺) into the oceans, over two orders of magnitude, from 3 × 10⁹ to 3 × 10¹¹ molecules Fe²⁺ cm⁻² s⁻¹, with the top end of that range similar to the estimated rate of Fe oxide deposition on Earth prior to the rise of oxygen (5 × 10¹¹ molecules Fe cm⁻² s⁻¹).

To study the effects of the stellar energy distribution, we simulated planets with an incoming solar spectrum that would be consistent with a planet in the habitable zones of Sigma Boötis (an F2V-type star), the Sun, ε Eridani (K2V-type star), AD Leonis (M3.5V-type star) and GJ 876 (M4V-type star). The characteristics of stars used for the present simulations are presented in Table 2. The F and K stars have been used by several authors to study biosignatures in atmospheres similar to present Earth, with high O₂ and low CO₂ (e.g., Segura et al. 2005; Grenfell et al 2007; Rugheimer et al. 2013). For the M dwarfs we used spectra from AD Leonis (AD Leo; Table 2) and GJ 876. AD Leonis is a star that has been extensively used in studies of planetary atmospheres (e.g., Domagal-Goldman et al. 2011; Segura et al. 2010; Rauer et al. 2011; von Paris et al. 2013), and GJ 876 is one of the M stars that have UV flux measurements (France et al. 2013). The AD Leonis spectrum is the same as the one used in Segura et al. (2005, 2010). To compile the GJ 876 stellar input spectrum, we use the GJ 876 spectrum reported by France et al. (2012) from 130 to 320 nm combined with a NextGen model (v5) star with T_eff = 3200 K (http://hobbes.hs.uni-hamburg.de/~yeti/NG-spec.html) from 320 to 855 nm.

Reviews on the potential habitability of planets around M dwarfs (e.g., Scalo et al. 2007, Tarter et al. 2007) made clear that many properties of these stars are not fully known and thus their effects on planetary habitability have not been constrained. Particularly, chromospheric activity and its resultant UV emission are not fully understood and pose a complex problem for stellar atmospheric models (e.g., Walkowicz et al. 2008), and Segura et al. (2005) demonstrated that the slope of the UV emission from M dwarfs was relevant to the atmospheric chemistry of biosignatures as it affects destruction rates and therefore atmospheric lifetimes of potential biosignatures gases. M dwarfs are classified as active based on their Hα strong chromospheric emission, while those with Hα in absorption were "quiescent," although their UV spectra were unknown and therefore it was not possible to assess whether their chromospheres were inactive. UV spectra from very active red dwarfs are characteristically constant in the wavelength range from 100 to 300 nm. Observations from the Hubble Space Telescope (Walkowicz et al. 2008; France et al. 2012, 2013) showed that the UV emission from stars classified as inactive based on their Hα emission was similar to that of the active M dwarfs; therefore, the characteristic flat UV spectra of active M dwarfs when quiescent (no flaring) may be representative of nonactive M dwarfs as well. In particular, GJ 876 shows Hα in absorption, but it presents a measurable emission in X-rays, a characteristic of chromospherically active stars (France et al. 2012, and references therein). Observations with the Hubble Space Telescope showed that the NUV emission (200–320 nm) of GJ 876 was actually similar to that of AD Leonis (Walkowicz et al. 2008). For the relatively blue stars (F dwarfs), their high amounts of UV radiation are caused by blackbody radiation extending into the MUV, as opposed to chromospheric activity at FUV wavelengths (Figure 1).

For all stars, we placed the model planets in an orbit that would allow the surfaces of these planets to absorb approximately the same amount of energy as Earth absorbs from the Sun. This includes a correction factor for the broadband albedo of a planet caused by the interaction of a planet's albedo spectrum with the stellar energy distribution. This was done after the process adopted by Segura et al. (2005).

In this study, we limit ourselves to stellar spectra that have been observed or modeled based on other observations. While it is possible that extremely active stars could contribute significant energy to CO₂ photolysis and O₂/O₃ production (K. Zahnle 2013, private communication), such stars would also deliver significant energy for O₂/O₃ destruction in the form of energetic protons. Past modeling of both these effects shows that, if anything, the net impact of flares would be to diminish, not enhance, O₃ concentrations (Segura et al. 2010) and should only be important for very active stars. Another possibility for enhanced O₂ and O₃ production beyond the parameter space explored here would be planets around O, B, and A stars that have significant UV radiation even when the stars are quiescent. However, owing to the relatively short lifetimes of habitable zones around these objects, they are not typically included in target lists for future exoplanet characterization missions (Turnbull & Tarter 2003).

At least three other groups (Selsis et al. 2002; Hu et al. 2012; Tian et al. 2014) have previously reported accumulation of detectable O₂ and O₃ in planetary atmospheres using altitude-dependent photochemical models. However, the first of these studies (Selsis et al. 2002) did not balance redox in the atmosphere, which can lead to spurious accumulation of atmospheric O₂ and O₃ by neglecting sinks for these species via reaction with reductants at the planet's rocky surface or in its oceans (Segura et al. 2007).

The other two studies (Hu et al. 2012; Tian et al. 2014) produced high O₂ and O₃ while maintaining redox balance, but did so with boundary conditions that we contend are not as rigorous as those employed here. Specifically, the deposition velocities of H₂, CO, and O₂ adopted by these two groups may have enhanced O₂ and O₃ concentrations, either by impacting the bulk redox budget or by impacting the distribution of oxidized and reduced species in the atmosphere. This allows them to satisfy Equations (E1) and (E2) above, but in a biased fashion that can favor accumulation of certain oxidized species over others. (For that matter, it also can favor accumulation of certain reduced species over others). Hu et al. (2012) proscribe H₂, CO, and O₂ deposition velocities to be 0, 10⁻⁸, and 0 cm s⁻¹, respectively; they cite Kharecha et al. (2005) for the value for CO, who base the CO deposition velocity on the chemistry of CO in water, and assume that the oceans are saturated with respect to H₂ and O₂. This effectively eliminates one of the most important sinks for O₂ in an abiotic world: reactions with surface rocks and dissolved species in ocean water.

In Tian et al. (2014), the deposition velocities proscribed for H₂, CO, and O₂ were respectively 0, 10⁻⁶, and 10⁻⁶ cm s⁻¹. The rationale for this is that H₂ would not be consumed by biology in a prebiotic ocean (and therefore the oceans would be saturated in H₂) and that CO and O₂ would be consumed by redox reactions in the oceans. Both this approach and that of Hu et al. (2012) are based on assumed CO-consuming reactions in the ocean, but the soluble species that destroy CO should also act to destroy other reduced species. Not accounting for this could overestimate the rate at which CO flows into the oceans, underestimate atmospheric CO concentrations, and underestimate the rate at which atmospheric CO reacts with O₂ and O₃. A more comprehensive and less selective approach is to manage the deposition velocities of H₂, CO, and O₂—and all other soluble species—based on lowering ocean concentrations of the species contributing the most to ocean redox imbalance. This can be done in a manner that is consistent with the fundamental assumptions in the deposition velocity model, as described above (Section 2.2).

When the different boundary conditions are taken into account, we are able to replicate prior work, with the exception of the Tian et al. (2014) results. When we adopt the boundary conditions of Hu et al. (2012), we replicate their qualitative results of relatively high O₂ and O₃. We also replicate those in Segura et al. (2007) for cases with similar stellar fluxes, H₂ outgassing rates, and surface CO₂ mixing ratios. However, we were not able to replicate the results of Tian et al. (2014) using photochemical models (Domagal-Goldman et al. 2011; Guzmán-Marmomejo et al. 2013; Kurzweil et al. 2013) that share a common heritage and a similar chemistry network with their model, even when using the boundary conditions given in Tian et al. (2014) and the same source for the stellar flux (France et al. 2012). In general, using similar methodologies, we predict at least an order of magnitude lower O₂ column densities than Tian et al. (2014) for all our models. If we adopt our boundary condition method (Section 2.2), we find orders of magnitude lower O₂ concentrations. For example, for a 5% CO₂ atmosphere with the same H₂ outgassing rates as Tian et al. (2014), we predict an O₂ column depth of 8 × 10¹⁸ molecules cm⁻², compared to the Tian et al. (2014) value of 8 × 10²² molecules cm⁻².

We used our radiative transfer model (SMART) to simulate the reflection and emission spectra of the six planets shown in Tables 3 and 4. Figures 3–5 show spectra in the UV, visible near-IR (NIR), and mid-IR, respectively. For comparison, the top panels of each of these figures show a spectrum of modern Earth, generated by the well-validated Virtual Planetary Laboratory 3D spectral Earth model (Robinson et al. 2010, 2011, 2014).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For all the model planets, the simulated reflectance spectra show the presence of the O₃ Hartley bands, which are seen in absorption from 0.2 to 0.3 μm (Figure 3). O₃ absorption is detectable at the 3σ level for abiotic planets around σ Boötis (F2V), the Sun (G2V), ε Eridani (K2V), AD Leonis (M3.5V), and GJ 876 (M4V) for missions that can achieve a signal-to-noise ratio (S/N) of 3, 16, 64, 8, and 38, respectively, in the UV (Table 4). However, these O₃ signals for many of these model planets will be much weaker than the O₃ feature caused by a biosphere. It will therefore be distinguishable if the bottom of the absorption feature in the albedo spectrum can be measured. A stronger O₃ absorption feature is present in the simulated albedo spectrum of the model planet around σ Boötis (an F2V star), making the detection of the presence of O₃ relatively easy for any mission that can detect radiation in the UV wavelength range, such as the Occulting Ozone Observatory (O3). The shape of the feature is also very similar to the one in Earth's modern spectrum, meaning that a relatively high spectral resolution is required to differentiate between the O₃ levels for the abiotic model planet and inhabited Earth. Our simulation of the abiotic planets around σ Boötis also exhibits an O₃ feature at 9.6 μm (Figure 5), which overlaps strong absorption by the doubly hot band of CO₂ at 9.4 μm. The ozone absorption is more cleanly seen when the absorption from CO₂ is removed from the spectrum (the gray spectrum in panel 2 of Figure 5), as could be done by modeling and removal during analysis of an observed spectrum. Detecting this feature, once the carbon dioxide has been modeled and removed, requires an S/N > 15 (Table 4). The abiotically generated ozone predicted for the σ Boötis planet would not likely be detectable by the James Webb Space Telescope (JWST). However, if this instrument is used to characterize transit spectra of any potentially habitable super-Earth planets (Deming et al. 2009; Kaltenegger & Traub 2009), the cautions above should be taken into consideration.

CH₄ features are potentially detectable on the model planet around GJ 876, given measurements with an S/N above 17 at 1.7 μm and complete removal of the overlapping CO₂ absorption feature that dominates at this wavelength, or an S/N above 25 at 3.3 μm, with independent limits on CO₂ and CO concentrations. However, no O₃ or O₂ feature can be detected for the same planet for an S/N < 30. The reason for the higher concentrations of CH₄ on GJ 876 (compared to models of planets around other stars) is the lower O₂ concentrations and a correspondingly lower sink for CH₄. The first two of these features could be detectable by a flagship-class mission with a UV–NIR wavelength range that can also observe the O₃ feature at 0.25 μm. This combination could constitute a false positive for life, without a strategy (which could include NIR observations) to discriminate its abiotic origin.

Absorption features from other biosignature gases in these atmospheres would be much more difficult to detect (Table 4). The next strongest O₂ feature occurs at 0.76 μm, and for the relatively low abiotically produced O₂ abundances, this gas is only detectable with measurements that have an S/N above 380. This is an extremely difficult measurement to make and therefore is likely not a concern for proposed exoplanet characterization missions. We did not include any O₂ dimer features, as the spectral model used here predates the improvement needed to study them (Misra et al. 2014).

By observing the atmospheric context more completely, discrimination between biological and abiological sources for the O₃ is possible. Ideally, this would involve observations of the stellar spectrum, broad-wavelength planetary spectra sufficient to constrain other gas concentrations, and detailed atmospheric models similar to the ones used in this study. Knowledge of the UV stellar energy distribution is needed to provide inputs to photochemical models that could calculate the fluxes needed to maintain the concentrations of various gases derived from the spectrum. These models could then attempt to constrain a number of spectral features that are particular to an atmosphere with abiotic O₃.

First, the quantification of the abundance of O₃ could indicate atmospheric (and abiotic) O₃ production, as the O₃ column density in all our simulations of abiotic planets is at least an order of magnitude less than Earth's modern O₃ column density (Table 3). For most of the abiotic O₃ concentrations modeled here, quantification would require measurement of the depth of the 0.25 μm O₃ absorption feature, which is not saturated, and exhibits at least a small amount of transmission through the planet's atmosphere. The lone exception is the abiotic planet simulated around σ Boötis, which, like modern Earth, exhibits a spectrum that is completely opaque from 0.23 to 0.31 um. This case would be much more difficult to discriminate as abiotically generated, as the saturated ozone band would give only a lower limit on the ozone abundance in the atmosphere. It would not be possible to tell whether the actual abundance was much higher and therefore potentially biologically generated.

Second, a mission could discriminate between biotic and abiotic sources of O₃ by measuring the O₃/O₂ ratio of the atmosphere. Despite detectable O₃ concentrations, in all cases there was a lack of an observable O₂ absorption feature at 0.76 μm. The lack of this O₂ feature would indicate relatively low O₂ and high O₃/O₂ compared to modern Earth, suggesting an abiotic O₃ source. A mission with a very broad wavelength range could potentially detect a CO feature at 2.3 μm or 4.6 μm. This would indicate significant CO₂ photolysis rates as a potential abiotic source of atomic O and atmospherically derived O₂ and O₃. If the amount of CO were quantified, along with that of O₃, it might also indicate a significant but unutilized energy source for life. Similarly, CO₂ features in the infrared around 10 μm would indicate a relatively high abundance of CO₂ compared to modern Earth and hence a potential large source of O through CO₂ photolysis. These features appear in our models of high-CO₂ abiotic planets (Figure 5) but are absent in the spectrum of modern Earth (Figure 5, top panel) because Earth's pre-anthropogenic CO₂ abundance is not high enough to create these features. (The absorption feature at 9.6 μm in Earth's spectrum is caused by O₃, not by CO₂.) This strategy for discriminating false and true positives is similar to the one suggested by Selsis et al. (2002). In addition to the search for molecular O₂ features, a search for O₂ dimer features beyond 1 μm could also inform planet characterization (Misra et al. 2014). Although we did not explicitly model this feature, the O₂ concentrations in our model simulations of abiotic planets are orders of magnitude lower than those required to create such features on planets with 1 bar of total pressure.

Finally, the lack of a CH₄ absorption feature could indicate an abiotic source of O₂ or O₃. Looking at the entire suite of our simulations (Figure 6), there is a clear trend: decreasing O₂ and O₃ concentrations are both correlated with increasing CH₄ concentrations. This trend also holds for inhabited planets, but at O₂ and CH₄ concentrations that are orders of magnitude greater. Because of this trend, we expect that any atmospheres with properties beyond the parameter space explored here would exhibit undetectable amounts of CH₄ if they had higher O₂ or O₃ concentrations than the atmospheres simulated in this study. This stands in contrast to modern Earth, which has orders of magnitude more O₂, O₃, and CH₄ and features from these gases that are correspondingly easy to detect compared to the features modeled here for abiotic planets. Although some of our abiotic simulations (e.g., the simulation of a planet around GJ 876 shown in Figures 4 and 5) had detectable features from both O₃ and CH₄, placing upper limits on the column depths of these two gases would indicate potential abiotic sources, following the logic in earlier parts of this section. Such quantification—or at least the placement of low upper limits on these gases—could identify potential sources of abiotic O₂ and O₃ production.

Zoom In Zoom Out Reset image size

Taken together, these features would hint at an abiotic source of the O₃ and would cast doubt on conclusions of a biological O₃ source. This indicates a need for broad-wavelength searches for life on exoplanets: to discriminate an abiotic or biological source, a UV-visible mission that attempts to observe an O₃ feature at 0.3 μm should at the least attempt to quantify the O₃ and detect O₂ at 0.76 μm. This means having a spectral resolution of at least R = 60 and S/N > 10 (although trades between R and S/N are possible; see Stapelfeldt et al. 2014). Preferably, it should extend into the IR so it would also be capable of observing CH₄ and CO absorption features. Table 4 shows that missions in the IR that observe the O₃ and CH₄ features in low resolution do not detect both for an abiotic planet; this would allow such a mission to discriminate between biosignatures and a false positive for life. Finally, although the simulations presented here do not contain significant SO₂, accumulation of that gas could also lead to a high inventory of atmospheric O atoms, or give clues to the amount of volcanic outgassing on the planet (Kaltenegger & Sasselov 2010).

The implications of a potential false positive for life for planned or proposed exoplanet characterization missions will depend on the capability of the mission to observe the wavelengths of the features discussed above. For example, the abiotic O₃ absorption features are not within JWST's accessible wavelength range. JWST could observe the O₃ feature at 9.6 μm, but in all our simulations in which that feature was detectable, CH₄ concentrations were low and likely undetectable at 3.3 μm. Therefore, JWST can avoid this problem if both CH₄ and O₃ concentrations are constrained, and conclusions are not drawn on an O₃ absorption feature alone. Flagship-class space-based telescopes designed to characterize extrasolar planets with UV-visible NIR photons (e.g., Cash 2006; Soummer et al. 2009) would have the aperture size and wavelength range needed to observe the abiotic Hartley O₃ and CH₄ features. Taken on their own, the simultaneous presence of CH₄ and O₃ in these atmospheres could be misinterpreted as evidence for disequilibrium brought about by biological activity. However, if the observations are further scrutinized, discrimination between biological and abiological O₃ sources should be possible. The large telescope size would enable the placement of tight lower limits on O₂ concentrations, casting doubt on the presence of biological O₂ production. Additionally, the observation of multiple CO absorption features would enable the identification of a potential abiotic O₃ source: photolysis of a large CO₂ reservoir. Finally, combining this information with observations of the star's energy distribution via photochemical models (such as those used here) would provide confirmation that the O₃ likely originated from photolysis of potentially abiotic gases.

These abiotic ozone sources would prove more difficult to missions that would attempt to use band filters to specifically target the UV absorption feature from O₃, such as the proposed Occulting Ozone Observatory (Pravdo et al. 2010; Savransky et al. 2010). These proposals call for an initial focus on this relatively easy-to-detect feature. However, this focus comes at the price of less information about the atmospheric and stellar context for the observed features. In this case, observations of the star's energy distribution would be critical. By using these data as inputs to photochemical models, the potential for an abiotic O₃ source could be identified. However, the biological and abiological sources for the O₃ feature could not be discriminated, only highlighted as possibilities. If bands targeting the smaller absorption feature from O₂ are also included, this could aid in O₃ source attribution. However, because this feature is smaller, longer observations would be required to detect it. Thus, the ideal role of this mission is likely as a "first step" in generating a prioritization list of extrasolar planets for detailed follow-up observations.