Pale Orange Dots: The Impact of Organic Haze on the Habitability and Detectability of Earthlike Exoplanets

The stellar spectra we use in our models include the Archean Sun (2.7 Ga), the modern Sun, AD Leo (M3.5V), GJ 876 (M4V), Eridani (K2V), and σ Boötis (F2V). This stellar sample spans a range of activity levels, UV fluxes, and UV spectral slopes.

Our modern solar spectrum was modeled by Chance & Kurucz (2010), and our "Archean" Sun uses a modified spectrum based on the wavelength-dependent correction accounting for solar evolution from Claire et al. (2012) for 2.7 Ga. This correction scales the absolute level of flux and accounts for the higher levels of solar activity, and therefore more UV radiation, expected from a younger Sun.

To test the impact of the UV spectrum of M dwarfs on haze generation, we compare results from two M dwarfs with different activity levels. M dwarfs can be highly active with frequent high-energy flares, although older M dwarfs may be more quiescent (West et al. 2008). For a highly active flaring star, we use a time-averaged observed spectrum of AD Leo, an M dwarf with frequent flare events (Hawley & Pettersen 1991; Hunt-Walker et al. 2012). Our spectrum of AD Leo is discussed in Segura et al. (2005). We also test haze generation using a spectrum for GJ 876, a known M4V planet host (Von Braun et al. 2014). The spectrum of GJ 876 is described in Domagal-Goldman et al. (2014) based on the spectrum reported in France et al. (2012).

In addition to the M dwarfs, we use K2V ( Eridani) and F2V (σ Boötis) spectra described in Segura et al. (2003). Eridani (3.2 pc) is a young star encircled by a dust ring (Greaves et al. 1998) and is one of the closest known exoplanet hosts (Hatzes et al. 2000). Eridani is also chromospherically active (Noyes et al. 1984). σ Boötis is an F2V star 15.5 pc away.

For all stars except the modern Sun, we scale their total integrated fluxes to the solar constant for Earth at 2.7 Ga, which was 80% less than the modern value (0.8 × 1360 W m⁻²) to compare with our Archean results. Unlike the Archean Sun, the other stars do not include a wavelength-dependent correction for stellar evolution. Figure 2 shows a comparison of the stellar spectra in the UV, visible, and near-infrared (NIR) together with the UV cross sections of several important gases. The plot of UV stellar spectra shows the actual resolution of the wavelength grid used by the photochemical model. These stellar spectra at full resolution are available for download on the VPL Spectral Database.¹¹

Zoom In Zoom Out Reset image size

A total surface pressure of 1 bar is assumed in all situations. The nominal results presented here are for pCO₂ = 0.01 bar and CH₄/CO₂ = 0.2 (Figure 3). Note that the CH₄/CO₂ ratios we refer to apply to the planetary surface because CH₄ does not follow an isoprofile in the atmospheres we simulate. CO₂, on the other hand, is well mixed. This CO₂ level is consistent with the paleosol measurements of Driese et al. (2011), and this CH₄/CO₂ ratio is sufficient to form organic haze on Archean Earth. Molecular oxygen (O₂) is set at a mixing ratio of 1 × 10⁻⁸, corresponding to a time after the origin of oxygenic photosynthesis but prior to oxygen accumulation in the atmosphere. Haze particles are treated as fractals composed of 0.05 μm sized spherical monomers, similar to the size of the monomers in Titan's hazes (Rannou et al. 1997; Tomasko et al. 2008) and the same size as the monomers used by Wolf & Toon (2010) in their study of fractal haze on Archean Earth. Haze scattering properties are derived using the optical constants of Khare et al. (1984) through the fractal mean-field approximation (Botet et al. 1997). The optical constants of Khare et al. (1984) were measured for Titan simulant hazes, but optical constants for Archean-analog haze have only been measured at one wavelength (532 nm) in a previous study (Hasenkopf et al. 2010). The haze optical constants of Khare et al. (1984) produce a reasonable match to the haze measurement of Hasenkopf et al., and an extended discussion of our choice of optical constants can be found in our previous study (Arney et al. 2016).

Zoom In Zoom Out Reset image size

We use the HITRAN 2012 linelists to generate our spectra (Rothman et al. 2013). The solar zenith angle is set at 60° for the direct-imaging spectra, which approximates the average incoming solar flux at quadrature. As in ${\mathtt{Atmos}}$, we use fractal particles with scattering, absorption, and extinction efficiencies generated with the mean-field approximation (Botet et al. 1997) for our ${\mathtt{SMART}}$ simulations, and we use the optical constants from Khare et al. (1984).

To explore the effect of the stellar UV spectrum on haze formation, we ran ${\mathtt{Atmos}}$ to obtain chemically and climatically self-consistent results for Archean-like atmospheres under the influence of different spectra from the host star. Most of these models were run with CH₄/CO₂ = 0.2. In the case of AD Leo and the K2V star, which did not form hazes with CH₄/CO₂ = 0.2, we ran additional atmospheres with higher CH₄/CO₂ ratios until a haze formed. These additional simulations have CH₄/CO₂ = 0.9 and 0.3, respectively. The F2V star did not form hazes at any CH₄/CO₂ ratios tested here (up to CH₄/CO₂ = 2).

The compositions of our planets' atmospheres are strongly influenced by their host star's spectrum despite equivalent gaseous surface boundary conditions, underscoring the importance of photochemistry in the atmospheres of exoplanets. Table 1 shows the diurnally averaged stellar UV fluxes incident on the planet for near-UV (NUV, 300–400 nm), mid-UV (MUV, 200–300 nm), far-UV (FUV, 130–200 nm), and our photochemical model's Lyα bin for each star. We also define and show "interval 1" (I1) as wavelengths between 120 and 140 nm and "interval 2" (I2) as wavelengths between 140 and 160 nm. I1 corresponds to the peak of the CH₄ UV cross section, and I2 to the peak of the CO₂ UV cross section. Note that a star's activity level (including Lyα emission and extreme UV flux) is affected by a number of parameters including the stellar age and rotation rate. For instance, younger stars tend to have higher activity levels (e.g., West et al. 2008). Stars that produce higher levels of FUV radiation than NUV and MUV tend to generate larger quantities of haze-destroying oxygen radicals because FUV can dissociate CO₂ and H₂O (Section 3.2). An exception is GJ 876, which also has a higher proportion of FUV relative to NUV and MUV; however, GJ 876 has a lower absolute level of UV flux at these wavelengths. A test scaling GJ 876's total amount of UV radiation to the total level of UV radiation produced by AD Leo diminishes GJ 876's haze production rate. Similarly, a test decreasing AD Leo's total UV flux to that of GJ 876 increases its haze production rate. Consequently both the slope of the incident radiation (ratio of FUV to NUV or MUV) and the overall intensity of that radiation appear to affect haze production.

The best predictor of whether a planet forms haze in our photochemical scheme appears to be the absolute level of flux in the I2 bin. Here, the F2V star and AD Leo produce the most and second most flux, respectively, and these stars are least and second least efficient at haze formation as we discuss below. The K2V star has the third largest amount of flux in the I2 bin, and while it is able to form a haze, it requires a higher CH₄/CO₂ ratio to do so than the stars with the lower I2 levels. GJ 876 has the least flux in I2, and as we will show below, it is also the star around which hazes form most easily. This suggests that oxygen species produced by CO₂ photolysis are the primary haze destroyers.

A well-characterized UV spectrum of the host star and good constraints on the planet's orbit will be important considerations for predicting incident UV radiation on a planet, generating hazes using photochemical models, and placing general constraints on the photochemistry and climate of a planet.

Table 2 presents a comparison of the total integrated column densities of key gases in the atmospheres of our simulated environments, including some of the hydrocarbons involved in haze formation and oxygen radicals involved in destroying hydrocarbons. The values presented in this table are divided by the nominal total integrated column density for Archean Earth (with CH₄/CO₂ = 0.2) at 2.7 Ga, and the diversity of gas abundances for each star clearly illustrates how photochemistry impacts these atmospheres. In this table, C₄H₂ and C₅H₄ are direct precursors to haze particles according to our simplified haze formation scheme as discussed above (Pavlov et al. 2001a). Ethane (C₂H₆), also shown, forms from photochemical reactions involving CH₄ and may be important for warming organic-rich atmospheres (Haqq-Misra et al. 2008).

In addition to the gas profiles, ${\mathtt{Atmos}}$ was also used to calculate the temperature profiles for the simulations shown in Table 2. The results of these climate calculations are provided in Table 3. The diversity of surface temperatures in this table is due to the climatic effects of different haze thicknesses, different greenhouse gas abundances, and the spectral energy distribution of the host star. These effects are discussed in detail in the sections below.

To illustrate these atmospheres' gas, haze, and temperature profiles, these quantities are shown for the planets with CH₄/CO₂ = 0.2 in Figure 4. The profiles for the nominal Archean environment are shown with green lines for comparison.

Zoom In Zoom Out Reset image size

The photochemical production of oxygen-bearing gases has an important impact on the ability of each atmosphere simulated here to form haze. The generation of O from CO₂ photolysis (and photolysis of other O-bearing species such as NO₂ and H₂O) leads to the creation of oxidized species (including O itself) that can react with organics on the reaction pathway to haze formation.

Table 4 shows a comparison of the principal photolysis reactions involving CO₂, H₂O, and NO₂ in each atmosphere that produce hydrocarbon-consuming oxygen species. An asterisk marks the fastest reaction in each atmosphere. CO₂ photolysis is the most efficient source of O and O(¹D) in every atmosphere. The reaction rates scale with the amount of UV flux able to dissociate a given species. For instance, compare the reaction rates in Table 4 with the UV fluxes and cross sections in Figure 2: stars with elevated fluxes at the wavelengths overlapping the UV cross sections of these O-producing species produce higher amounts of oxygen species through photolysis. In general, the more oxygen an atmosphere produces from photolysis of species like H₂O, CO₂, and NO₂, the thinner the hazes. Thus, to predict whether a star is likely to have a planet with organic haze in the habitable zone, we will require measurements of the shape of its UV spectrum, especially for wavelengths controlling CO₂ photolysis, which is the dominant predictor of haze destruction.

Figure 5 shows the hydrocarbon chemical reaction network with the fastest reactant rates that lead to haze production or haze sinks for each star. "HCAER" in this figure stands for the hydrocarbon species that most efficiently condenses out as haze particles, C₄H₂. Species outlined by hexagons represent major sinks of haze-forming gases, and essentially a truncation of the haze-forming reaction network. The dominant overall pathway to haze formation for the haze-forming planets examined here is

$$\begin{eqnarray}&&{\mathrm{CH}}_{4}+h\nu \to {{\mathrm{CH}}_{2}}^{3}+{\rm{H}}+{\rm{H}}\end{eqnarray} \tag{ R1 }$$

$$\begin{eqnarray}&&{{\mathrm{CH}}_{2}}^{3}+{\rm{H}}\to \mathrm{CH}+{{\rm{H}}}_{2}\end{eqnarray} \tag{ R2 }$$

$$\begin{eqnarray}&&\mathrm{CH}+{\mathrm{CH}}_{4}\to {{\rm{C}}}_{2}{{\rm{H}}}_{4}+{\rm{H}}\end{eqnarray} \tag{ R3 }$$

$$\begin{eqnarray}&&{{\rm{C}}}_{2}{{\rm{H}}}_{4}+h\nu \to {{\rm{C}}}_{2}{{\rm{H}}}_{2}+{\rm{H}}+{\rm{H}}\end{eqnarray} \tag{ R4 }$$

$$\begin{eqnarray}&&{{\rm{C}}}_{2}{{\rm{H}}}_{2}+h\nu \to {{\rm{C}}}_{2}{\rm{H}}+{\rm{H}}\end{eqnarray} \tag{ R5 }$$

$$\begin{eqnarray}&&{{\rm{C}}}_{2}{\rm{H}}+{{\rm{C}}}_{2}{{\rm{H}}}_{2}\to {{\rm{C}}}_{4}{{\rm{H}}}_{2}+{\rm{H}}.\end{eqnarray} \tag{ R6 }$$

Zoom In Zoom Out Reset image size

In this pathway, CH needs to react with CH₄ to form C₂H₄, but in atmospheres that generate large quantities of oxygen species, CH can instead readily react with O to form CO via CH + O $\to$ CO + H. Once C₂H₄ forms, it can exit the haze-formation path if it reacts with O to form HCO, or it can continue on the haze-formation path if it is photolyzed to form C₂H₂. Then, C₂H₂ can be photolyzed to produce C₂H, but if C₂H₂ instead reacts with O in an oxygen-rich atmosphere, it can form ${{\mathrm{CH}}_{2}}^{3}$; although ${{\mathrm{CH}}_{2}}^{3}$ is involved in haze formation, going from C₂H₂ to ${{\mathrm{CH}}_{2}}^{3}$ does not advance further in the haze formation scheme toward higher-order hydrocarbons. If the C₂H produced from C₂H₂ reacts with O or O₂, it will result in HCO, which is not useful for haze formation.

Once the reaction network has produced the gases needed to form C₄H₂ (HCAER), it condenses out as haze particles, via C₂H₂ + C₂H $\to$ C₄H₂ + H. Alternatively, C₂H₂ can react with CH to form C₃H₂, which begins a second chain of polymerizations that can lead to C₅H₄, which also condenses out as haze particles (HCAER2). However, this is less efficient than the process forming C₄H₂ and is not shown in Figure 5 or in the reaction network outlined above.

Table 5 shows the ratios of the total integrated reaction rates that remove hydrocarbons from the haze-formation chain (via reaction with oxygen radicals) versus the reactions that step toward the formation of haze particles. The reactions in the table represent key stages of the dominant haze-formation process outlined above. There is a clear difference between the planets that form haze and those that do not. The haze-poor planets generally favor reactions with oxygen radicals over reactions that lead to haze formation. As discussed previously, the primary driver of this difference in behavior is the amount of flux in the 140–160 nm region, where CO₂ is most efficiently photolyzed, because oxygen sourced from CO₂ photolysis is the primary source of haze-precursor destruction, although the other oxygen-bearing gases contribute as well.

An alternative, but less efficient, way of forming C₂H₄ involves the formation of ethane. This haze-formation pathway requires production of the methyl radical (CH₃). Although oxygen radicals can frustrate haze formation later on in the reaction network, they are initially helpful in forming CH₃. For every star but AD Leo, the most efficient vectors toward forming CH₃ are

$$\begin{eqnarray}&&{\mathrm{CH}}_{4}+{\rm{O}}\to {\mathrm{CH}}_{3}+\mathrm{OH}\end{eqnarray} \tag{ R7 }$$

$$\begin{eqnarray}&&{\mathrm{CH}}_{4}+\mathrm{OH}\to {\mathrm{CH}}_{3}+{{\rm{H}}}_{2}{\rm{O}}.\end{eqnarray} \tag{ R8 }$$

AD Leo, meanwhile, most efficiently forms CH₃ from CH₄ photolysis due to its high Lyα output overlapping with the peak of the CH₄ UV cross section, but this is only a factor of 1.15 times faster than CH₄ + OH.

Once CH₃ forms, an exit from the haze-formation network occurs if it reacts with O to form formaldehyde, CH₂O, which ultimately ends up as CO₂:

$$\begin{eqnarray}&&{\mathrm{CH}}_{3}+{\rm{O}}\to {\mathrm{CH}}_{2}{\rm{O}}+{\rm{H}}.\end{eqnarray} \tag{ R9 }$$

The above oxidation of CH₃ competes with the formation of ethane (C₂H₆) via reaction of CH₃ with another CH₃, or more commonly with CH₃CO (produced from CH₃ + CO):

$$\begin{eqnarray}&&{\mathrm{CH}}_{3}+{\mathrm{CH}}_{3}\to {{\rm{C}}}_{2}{{\rm{H}}}_{6}\end{eqnarray} \tag{ R10 }$$

$$\begin{eqnarray}&&{\mathrm{CH}}_{3}\mathrm{CO}+{\mathrm{CH}}_{3}\to {{\rm{C}}}_{2}{{\rm{H}}}_{6}+\mathrm{CO}.\end{eqnarray} \tag{ R11 }$$

A number of reactions can then occur with C₂H₆ that are relevant to haze formation. Ethane can react with oxygen radicals to form C₂H₅. Then, either C₂H₅ can react with H or O₂ to re-form CH₃, or it can react with CH₃ to advance toward C₂H₄. Alternatively, C₂H₆ can be photolyzed to form C₂H₄ directly (and, less efficiently, C₂H₂). Once C₂H₄ forms, haze formation advances toward C₄H₂ (HCAER) through the same steps outlined above in the dominant haze formation network after C₄H₂ is formed.

Below, we present an analysis of haze formation and its climatic consequence for each host star type compared to our nominal Archean results to explore the differing atmospheric compositions and temperatures of these worlds.

We present a detailed discussion of haze formation for Archean Earth orbiting the Sun at 2.7 billion years ago in Arney et al. (2016). Haze formation for pCO₂ = 0.01 begins to noticeably impact the Earth's spectrum at CH₄/CO₂ = 0.18. The temperature of the planet's surface drops from ∼284 K when no haze is present to 272 K when a haze is in place at CH₄/CO₂ = 0.2. Although this surface temperature is below the freezing point of water, 3D climate studies have suggested planets like early Earth with global average temperatures down to 250 K can still maintain stable open ocean waters near the equator (Charnay et al. 2013), so this low temperature can still be considered "habitable" because the planet could still support liquid water at the surface.

3.3.1. Hazes with the Modern Solar Constant

3.3.2. M Dwarfs

3.3.3. K2V Dwarf

3.3.4. F2V Dwarf

Reflectance, thermal emission, and transit transmission spectra for the Archean-analog planets are presented in Figure 6 for all stellar spectral types studied here. All of these planets have CH₄/CO₂ = 0.2 except the spectrum labeled "K2V—haze," which has CH₄/CO₂ = 0.3, and "AD Leo—haze," which has CH₄/CO₂ = 0.9, the ratios required to form haze for these planets. At CH₄/CO₂ = 0.2, planets around AD Leo, the K2V star, and the F2V star do not have spectrally apparent hazes in their atmospheres, but the Archean Sun, modern Sun, and GJ 876 planets do. As noted before, the F2V star does not generate organic hazes even at CH₄/CO₂ ratios greater than unity.

Zoom In Zoom Out Reset image size

Haze absorbs strongly at blue and UV wavelengths, causing the reflectance spectra (top panel of Figure 6) of the hazy worlds to have lower albedos at these wavelengths. When the haze is thick enough to affect the spectrum, it creates a large absorption feature at these short wavelengths. Thus, rather than the Rayleigh scattering-induced increase in reflectivity at short wavelengths seen for the haze-free planets, hazy worlds produce their peak spectral brightness at visible wavelengths. The UV–blue haze absorption feature can be seen for the Archean and modern Sun, GJ 876, and hazy K2V planets, although the sparse haze around the AD Leo planet with CH₄/CO₂ = 0.9 is thin enough to be almost spectrally indistinguishable from a clear-sky world (Figure 6).

There are large spectral differences for the planet orbiting the modern Sun compared to the hazy planets around stars emitting Archean-like levels of radiation, and these are primarily due to the effects of atmospheric temperature on particle coagulation timescales and therefore the size of haze particles (Arney et al. 2016). The modern Sun's planet has the peak of its reflectance spectrum at λ ∼ 0.7 μm, compared to λ ∼ 0.5 μm for the Archean, GJ 876, and K2V planets, and this is due to the larger particles in the modern Sun planet's atmosphere: the maximum radius of the Archean, GJ 876, and K2V haze particles plotted here is ∼0.5 μm versus ∼0.79 μm for the modern Sun's planet. Absorption and scattering efficiencies, Q_abs and Q_scat, are both larger for bigger fractal particles, and Q_scat also trends toward a flatter wavelength dependence as particle size grows (Figure 1). Increased absorption (higher Q_abs) deepens the short-wavelength absorption feature produced by a thick haze of larger particles. Meanwhile, the larger scattering efficiency (higher Q_scat) at longer wavelengths for larger particles increases the brightness of the planet at these wavelengths, pushing the peak of the reflectance spectrum redward. This demonstrates the need to simulate particles in coupled photochemical-climate models to capture the effects of atmospheric temperature on particle size and the resulting impacts on the planetary spectrum.

The impact of haze on the temperature structures of the atmospheres simulated here can also be seen in the thermal radiation spectra (middle panel of Figure 6). Hazes absorb UV photons and warm the stratosphere similarly to ozone on modern-day Earth. Signatures of warm stratospheres (thermal inversions) in the hazy atmospheres can be seen in the thermal emission spectra as CH₄ and CO₂ in emission (rather than absorption) near 8 and 15 μm for the Archean Sun, modern Sun, and hazy K2V spectra. As discussed in Section 3.3.2, the haze around the GJ 876 planet, however, does not produce a strong thermal inversion because its star emits less UV radiation (see also its temperature profile in Figure 4). On the other hand, both of the M dwarf planets have higher surface temperatures for the reasons discussed in Section 3.3.2, and this is apparent from the larger amounts of thermal radiation emitted by these worlds in the atmospheric window between roughly 9 and 11 μm. Ethane, a strong greenhouse gas, can be seen near 12 μm in all spectra as a photochemical consequence of the large quantities of methane in these atmospheres compared to modern-day Earth.

Hazes also strongly impact transit transmission spectra. Our transit spectra (bottom panel of Figure 6) include the effect of atmospheric refraction (Misra et al. 2014a, 2014b), and this makes it impossible to probe below 15–20 km in altitude for all model planets, including those with haze-free atmospheres. For atmospheres with haze, the minimum altitude that transit observations can probe is set by the altitude where the haze becomes optically thick. Note that because all of these transit spectra sense altitudes in the stratosphere, water vapor cannot be detected. Stratospheres on traditionally habitable planets are dry; a wet stratosphere would indicate a planet undergoing a runaway greenhouse effect. The transit spectra of the hazy worlds exhibit a scattering slope in the visible and NIR due to a combination of haze scattering and Rayleigh scattering. The thick haze shown around the modern Sun in particular produces a relatively featureless, sloped spectrum in which absorption features from gases are obscured at visible and NIR wavelengths shorter than ∼2 μm. At longer IR wavelengths where the haze is relatively transparent, its impact on the transit transmission spectra is diminished, and absorption features, particularly for CO₂ and CH₄, become apparent even for the spectrum of the modern Sun.

The transit spectra are sensitive to hazes that are barely detectable in reflected light, because of the longer path length taken by light in transit observations. The haze around the AD Leo planet with CH₄/CO₂ = 0.9 makes it scarcely distinguishable from a planet without haze in reflected light. However, the AD Leo haze is more apparent in the transit transmission spectrum compared to the haze-free planets.

As first discussed by Wakeford and Sing (2015) and by Arney et al. (2016), there is an absorption feature from the haze itself near 6 μm (caused primarily by C=C and C=N stretching) that may allow remote identification of hydrocarbon hazes on exoplanets. This feature produces the increase in effective tangent height in the hazy transit transmission spectra at this wavelength. Ethane and CH₄ absorption overlaps with the haze's 6 μm absorption feature, but the haze feature can be distinguished by higher opacity centered around 6.3 μm. We focus on this feature in Figure 7, which compares the AD Leo planet with a sparse haze to the Modern Sun planet with a thick haze. The AD Leo planet has more CH₄ and C₂H₆ than the modern Sun's planet. There is a peak in the haze extinction coefficient near 6.3 μm, which causes an increase in absorption for the modern Sun planet. The AD Leo planet's spectrum in this region is controlled by the behavior of the CH₄ and C₂H₆ absorption cross sections because its haze is very thin.

Zoom In Zoom Out Reset image size

In addition to the 6 μm feature, there is a much weaker haze absorption feature near 3 μm that is most easily seen as a small bump in the modern Sun spectrum. The weakness of the 3 μm haze feature makes it unlikely to be detectable. Both the 6 μm and 3 μm features can be seen as peaks in the haze Q_abs curve in Figure 1. These peaks appear to be general features of organic haze and are not specific to our use of the optical constants of Khare et al. (1984) (see Figure 14 in Arney et al. 2016 for a comparison of haze optical constants in the literature).

Hazes are strong absorbers at UV wavelengths (Figure 6) and so could potentially act as a UV shield for planetary surfaces. In particular, fractal organic hazes could have provided a UV shield for the anoxic Archean atmosphere (Wolf & Toon 2010; Arney et al. 2016), especially for DNA-damaging UVC radiation ($\lambda \lt 0.280$ μm). Since the Archean likely lacked an O₂/O₃ shield, another shielding agent would have assisted the development of land-based life.

Table 6 summarizes the UV flux at the surface (W m⁻²) for UVA (λ = 0.315–0.400 μm), UVB (λ = 0.280–0.315 μm), and UVC (λ < 0.280 μm) radiation for each of our planets. For comparison, we also include the surface UV fluxes our model calculates for the actual modern-day Earth atmosphere. Note that the "Archean Sun" results presented here are not the same as the results presented for UV shielding in our earlier work (Arney et al. 2016). The haze for the "Archean Sun" here refers to simulations with CH₄/CO₂ = 0.2 for pCO₂ = 0.01, and this haze is slightly thinner than the one discussed in the context of UV shielding in Arney et al. (2016), which referred to CH₄/CO₂ = 0.21 for pCO₂ ∼ 0.02. The hazy "modern Sun" planet has less UVA and UVB at the surface than the actual modern-day Earth, illustrating how the broadband UV absorption by organic haze cuts down UVA and UVB far better than gases in the actual modern-day atmosphere. The haze-free UV fluxes we quote here are comparable to the fluxes for similar stars found by Rugheimer et al. (2015) in a study of the UV surface environment of Earthlike planets orbiting various stellar types.

The surface UVC fluxes of the "Modern Sun—haze" planet and the "GJ 876—haze" planet are higher than we currently experience on Earth but should be easily tolerated by Chloroflexus aurantiacus, an anoxygenic phototroph that has been studied as an analog for Archean photosynthetic organisms (Pierson et al. 1992). Chloroflexus aurantiacus was shown in Pierson et al. (1992) to exhibit moderate growth under UVC fluxes comparable to or lower than the fluxes calculated here for every star in Table 6 except the F2V star, the modern Sun with no haze, and the Archean Sun with no haze. Of course, life can also take refuge from UV radiation under other types of chemical or physical UV shields (e.g., within a liquid water column) so even these higher UV fluxes do not necessarily prohibit life (Cockell 1998). Still, UV shielding is an important consideration for planetary habitability, so despite their cooling effects, UV-blocking hazes like the ones studied here may actually enhance planetary habitability.

Our analysis does not consider M dwarf flaring events, which can increase the UV irradiance by orders of magnitude (Segura et al. 2010). Since we have shown that stars with very high UV flux—particularly high FUV fluxes— do not form hazes as readily or at all compared to stars with lower FUV fluxes, frequent flaring events are expected to have a deleterious effect on a haze layer, although we have not examined the effects of time-dependent flares here.

Organic haze's strong absorption features provide an indirect way to sense atmospheres rich in CH₄ even if the CH₄ absorption features themselves are not distinguishable. Because attempts to characterize exoplanets have been frustrated by the presence of atmospheric aerosols (e.g., Kreidberg et al. 2014), haze is typically considered to obscure planetary characteristics. However, for the organic hazes presented here, gaseous absorption features can still be seen for λ > 0.5 μm in reflected light and for λ > 1 μm in transit transmission even in the hazy spectra.

Although they may obscure aspects of the planetary environment, organic hazes have the potential to unveil interesting ongoing planetary processes. The presence of an organic haze implies an active source of methane, particularly in high CO₂ atmospheres like Archean Earth, which requires a CH₄/CO₂ level > 0.1—and therefore a substantial CH₄ flux of the order of ∼1 × 10¹¹ molecules cm⁻² s⁻¹ before haze formation occurs. This methane flux is comparable to the rate of methane production by biology on modern Earth (Kharecha et al. 2005), so hazes in atmospheres with Archean-like CO₂ levels could signal possible biological activity.

Because methane can be produced by a variety of biological and nonbiological means, there is no reason to expect organic-rich planets in the habitable zone to be rare. We should therefore be prepared for the detection of hazy habitable planets orbiting G, K, and M dwarfs. In this section, we discuss the detectability of organic haze around M dwarfs with JWST, and around G and K dwarfs with a future large 10 m direct-imaging telescope.

3.6.1. Simulated JWST Observations

M dwarf planet hosts will be important targets for transit transmission observations by the JWST because the ratio of the planet's size relative to the star's size is largest for M dwarfs. Thus, their transit transmission signals are larger than for equal-radius planets orbiting stars of higher mass. Planets in the habitable zone also orbit closer to M dwarf stars, so their transits occur more frequently than for planets orbiting stars of higher mass.

Figure 8 shows the results of our simulated observations over 65 hr of integration time (10 transits) per instrument for a planet orbiting GJ 876. The pink line shows the simulated spectrum, and the orange points with error bars denote the simulated JWST observations. The gray line shows the planet without haze, which is included for comparison. The error bars are calculated assuming photon-limited noise, which is the same assumption made in Schwieterman et al. (2016). The large error bars at wavelengths longer than 8 μm are due to spectral noise caused by the dim stellar blackbody at these wavelengths.

Zoom In Zoom Out Reset image size

We calculate the continuum level for the JWST observations by fitting a polynomial to continuum regions. To determine the detectability of spectral features, we determine the continuum level around absorption features, then subtract that continuum. We calculate the signal-to-noise ratio (S/N) of absorption features in our simulations by binning across the features and comparing the signal to the noise level. The noise level is computed from the error on the binned absorption features and the error on the continuum estimate added in quadrature.

CH₄ and CO₂ can be detected across the NIRISS and NIRSpec bands. The CH₄ feature near 1.7 μm has S/N = 3.0, and CO₂ at 2 μm has S/N = 2.8. Detections of features at shorter wavelength are $\lt 1\sigma$. However, the broader and stronger CH₄ feature near 2.3 μm is detectable at S/N = 6.4. For the NIRSpec absorption features, the opacity of the haze is negligible, and we measure S/N = 6.4 for the CH₄ feature near 3.3 μm and S/N = 5.9 for the CO₂ feature near 4.3 μm. In principle, therefore, it will be possible to measure CH₄ and CO₂ abundances and the CH₄/CO₂ ratio in transit transmission for the atmospheric altitudes probed by transit observations.

The haze is more transparent at the long-wavelength end of the NIRISS bandpass and in NIRSpec. The abundances of CH₄ and CO₂ inferred from longer wavelengths where the haze is less opaque will be larger than the gas abundances inferred from the absorption features at shorter NIRISS wavelengths that are truncated by the haze. Based on the size of the error bars in the continuum regions around absorption bands, the absorption features in the hazy spectrum for λ < 2.5 μm are about 2–10σ shallower than they would be in a spectrum without haze. The inconsistency between the retrieved gas abundances at longer wavelengths and abundances retrieved at shorter wavelengths would suggest the presence of a haze whose opacity increases toward shorter wavelengths.

Detection of spectral features in MIRI is in principle more challenging than in NIRISS and NIRSPec because of the decline of the stellar blackbody at these longer wavelengths. However, if we define the MIRI continuum from between the first point in the MIRI bandpass and the points between 8 and 10 μm, we can measure S/N = 5.1 for the set of absorption features between 6 and 8 μm, which includes the haze absorption feature. Higher S/N would be needed to distinguish the haze from other absorbers in this region, but the presence of haze can be inferred separately from the NIRISS and NIRSpec observations of the CH₄/CO₂ ratio and the depths of the NIRISS absorption features compared to the expected haze-free level.

The instrument models above do not include any contributions from systematic noise. Systematic noise sources come, in large part, from instrumentation and detectors, and will not be fully characterized until after launch. They will also tend to decrease in time as instrument and detector models are improved and new observing techniques are developed. The simulated observations presented here indicate that achieving a combined noise (random plus systematic) at the level of several ppm will be essential for characterizing hazy exo-Earths. Such precision may be possible if the systematic noise sources are characterized to a level well below the random noise. However, if the JWST systematic noise represents a floor at the >10 ppm level, as proposed by Greene et al. (2016), then characterizing the hazy exo-Earths presented here becomes extremely difficult, and the error bars on our simulated JWST measurements will become much larger.

3.6.2. Simulated Direct-imaging Observations

Unlike for JWST observations, M dwarf planet hosts are generally poor targets for direct-imaging surveys because it will likely not be possible to angularly separate their habitable planets from their host stars except for the closest M dwarfs (e.g., Proxima Centauri b, Anglada-Escudé et al. 2016). The inner working angle (IWA), which defines the smallest angular separation between a planet and its host star at which the planet can be detected, scales with Nλ/D where D is the telescope diameter and N is a small-valued constant of order 10⁰. F dwarf habitable planets, which will naturally orbit farther from their stars than planets orbiting cooler hosts (Kopparapu et al. 2013), are most likely to be observable outside the IWA for the star types we simulate here, but we have shown that Archean-like worlds orbiting F dwarfs are less likely to have organic hazes. Note that F dwarfs are less numerous than stars of lower mass, so the distance to any one of them is likely to be larger than for G, K, and M dwarfs, and so the angular separation between planet and star may still pose a problem. G and K dwarf planets, on the other hand, may have organic haze, and such stars will be important targets for future direct-imaging missions (Stark et al. 2014).

We tested what a hazy Archean Earth-analog orbiting the modern Sun, the Archean Sun, and the K2V dwarf would look like to a future 10 m LUVIOR-type space telescope (Postman et al. 2010; Bolcar et al. 2015; Dalcanton et al. 2015) using the instrument noise model for a coronagraph described in Robinson et al. (2016). The star–planet systems are assumed to be located at a distance of 10 pc. The results of these simulations are presented in Figure 9. A spectrum with the haze removed (gray line) is presented alongside the hazy spectrum (pink line) for comparison in all three cases. The "observed" spectra are simulated assuming 200 hr (roughly one week) of integration time per coronagraphic bandpass (which may not span the entire wavelength region of interest) for a planet at quadrature. If the planets were at a distance of 3 pc instead of 10 pc, the integration time needed to achieve the same S/N decreases by about an order of magnitude; we use 10 pc here to be conservative and to be able to consider the challenges of detecting distant planets. The spectral resolution (R = λ/${\rm{\Delta }}\lambda$) is 70, and the telescope wavelength range is 0.4–3 μm. We chose an outer working angle of 20λ/D, and an IWA of 3λ/D. For a 10 m mirror, the IWA limits the longest wavelengths that can be observed in all three cases: the planets orbiting the G2V star cut off near 1.5 μm, and the planet orbiting the K2V star cuts off near 1 μm. Visible and NIR wavelength ranges are observed by separate detectors as described in Robinson et al. (2016). It is assumed that the telescope system will be cooled to a sufficiently low temperature to minimize detector thermal noise (T < 80 K) that would otherwise contribute to spectral noise in the NIR. Thermal noise should not contribute appreciably to wavelengths <1.6 μm, so our assumption of a cold telescope should not strongly impact the results shown here.

Zoom In Zoom Out Reset image size

A throughput of 5% is assumed from the work of Robinson et al. (2016), although LUVOIR may have a higher throughput closer to 20% (A. Roberge, personal communication). For this reason, the simulations presented here may be considered conservative.

To test the detectability of spectral features, we followed a similar procedure to our JWST analysis. We fit a polynomial to the continuum regions near features of interest, then removed the continuum using the procedure described above. We then binned across the absorption features to determine their S/N compared to the continuum level.

We find that, for all three stars, the haze absorption feature at UV–blue wavelengths is easily detected against the expected extrapolated continuum level. The haze absorption feature of the modern Sun has an extremely robust detection of S/N = 70, the Archean Sun haze has S/N = 12, and the K2V haze has S/N = 27. Our extrapolated continuum only extends the polynomial fit to longer wavelengths and does not include the expected Rayleigh scattering, so these detections are underestimates compared to a model that includes Rayleigh scattering. An absorption feature from overlapping CH₄ and H₂O can be seen near 0.72 μm for all three planets. It can be detected at S/N = 3.1 for the modern Sun, S/N = 2.6 for the Archean Sun, and S/N = 3.3 for the K star.

Features can be seen in the spectra of the modern and Archean Sun near 1.15 and 1.4 μm that are caused by overlapping H₂O and CH₄ absorption bands. The 1.15 and 1.4 μm features are detected at S/N = 16 and S/N = 11 for the modern Sun, and at S/N = 11 and S/N = 9 for the Archean Sun, respectively. At 10 pc, these features are cut off by the K star IWA and cannot be observed. Because of the ∼1 μm IWA cutoff for a planet of a K star at 10 pc, the haze absorption band is the strongest feature that can be seen, providing indirect evidence of methane.

The large error bars around 1 μm are caused by falloff in the quantum efficiencies of both CCD and InGaAs detectors assumed by the simulator, making it difficult to detect spectral features here. In the NIR, considering the frequent overlap between CH₄ and H₂O bands, good sensitivity to the clean, methane-free water bands near 0.82 μm and especially the stronger band near 0.94 μm (Figure 6) is crucial in order to detect and quantify the abundance of water even for the G star spectra. Because water could not be cleanly detected in these spectra with the assumed detector technology, retrievals of gas abundances would likely exhibit degeneracies in the retrieved amounts of H₂O and CH₄.

Note that organic haze itself could be an indirect sign of water on Earthlike planets. Because sources of methane on an Earthlike planet are likely to involve water (either through biological production or through serpentinization, the dominant abiotic source of methane on Earth, and one that requires water), the indirect detection of methane on a terrestrial planet could also be argued to suggest the presence of water, particularly in an atmosphere with CO₂ that necessitates a vigorous CH₄ flux to produce haze.

CO₂ is important to detect for this reason and others. For instance, measuring CO₂ abundance can constrain the redox state of the atmosphere, and its presence in a planetary atmosphere can help to determine whether a planet is in fact terrestrial in the absence of other data for mass or planetary radius. Unfortunately, CO₂ cannot be detected in reflected light for any planets shown here. Although there is a CO₂ feature near 1.6 μm, it is not detectable at the spectral resolution and noise level we simulate. The strongest CO₂ band shortward of 5 μm is near 4.3 μm. However, this band would be difficult to measure in direct imaging. We assume a cold telescope in the simulations here, but thermal radiation from a telescope that is not cryogenically cooled would be very significant for wavelengths longer than about 1.8 μm. In addition, 4.3 μm may not be accessible with the telescope's IWA (as in the examples shown here). Assuming IWA = 3λ/D, a telescope would have to be about 27 m in diameter to reach 4.3 μm for a planet that is 1 au from its star at a distance of 10 pc. On the other hand, a 10 m mirror would be sufficient to reach 4.3 μm for a target at 3 pc.

Longer wavelengths such as 4.3 μm may be more easily observable with LUVOIR for targets whose geometry allows for transit transmission observations. It may also be possible to observe exoplanets such as the ones simulated here with the next generation of ground-based observatories with larger mirror diameters than LUVOIR using adaptive optics and advanced coronography. These observatories will have to contend with spectral contamination from Earth's atmosphere, but Snellen et al. (2013) has suggested that the Doppler shift of the planet could be used to disentangle its spectrum from Earth's atmosphere.

Our results suggest that G dwarfs, K dwarfs, and some M dwarfs are more likely to generate hydrocarbon hazes in Earthlike atmospheres than F dwarfs and stars with frequent flare events such as AD Leo. To generate hazes, stars need sufficient UV flux to drive the relevant photochemistry through reactions such as CH₄ + hν (λ < 150 nm) $\to$ CH₃ + H, but too much FUV flux generates oxygen radicals through reactions such as CO₂ + hν(λ < 200 nm) $\to$ CO + O and CO₂ + hν (λ < 200 nm) $\to$ CO + O(¹D) that halt the haze formation process by oxidizing hydrocarbon photochemical products.

However, our model assumes a mechanism proposed for the formation of Titan's haze (Allen et al. 1980; Yung et al. 1984) such that haze formation occurs through formation of acetylene (C₂H₂) and its further polymerization to higher order hydrocarbons. In reality, this scheme is likely overly simplistic. For example, measurements of Titan's hazes by the Cassini spacecraft have discovered nitrile chains and nitrogen-bearing polycyclic aromatic hydrocarbons that suggest nitrogen-bearing compounds may be important to haze formation on Titan (Waite et al. 2007; López-Puertas et al. 2013).

Titan's atmosphere is extremely reducing, but Archean Earth's atmosphere was probably less so, containing non-negligible amounts of CO₂ (Kasting 1993; Driese et al. 2011). Interestingly, laboratory experiments have suggested that the presence of oxygen may not be as harmful to haze production as our haze-formation scheme suggests here. For example, Trainer et al. (2006) showed that haze formation in a CH₄/N₂ mixture containing CO₂ was more efficient than in a CH₄/N₂-only mixture because the oxygen atoms produced by CO₂ photolysis were incorporated into the haze molecules. Furthermore, DeWitt et al. (2009) showed the existence of carbonyl and carboxyl groups in aerosol analogs with C/O = 0.1. Hörst & Tolbert (2014) showed that CO, also, can benefit aerosol formation and be a source of oxygen incorporation into aerosol molecules. Recently, Hicks et al. (2016) showed that oxygen from CO₂ incorporated into haze molecules can comprise 10% of the mass of Archean haze particles.

These complexities suggest that an updated study incorporating these mechanisms into our photochemical model will be necessary to determine their impact on haze formation for the planets simulated here. We may find that haze formation is enhanced relative to our findings for planets orbiting stars with efficient oxygen production, and the hazes that form in more oxygen-rich atmospheres may differ in composition and spectral properties from those in more oxygen-poor atmospheres. Updates to our photochemical model including incorporation of laboratory studies we are involved with will allow us to examine these issues are part of our ongoing work and future work.

Since haze can cool a planetary climate, there may be an inner edge of the "hazy habitable zone" (HHZ) closer to the star than the traditional boundaries of the habitable zone (Kasting 1993; Kopparapu et al. 2013) for planets with organic-rich atmospheres. However, the results presented here indicate that the inner edge of the HHZ will not be relevant to certain types of stars such as F and M dwarfs. As we have seen, F2V planets with atmospheres containing CO₂ and H₂O do not generate this haze even at high CH₄/CO₂ ratios because of the buildup of haze-destroying oxygen-containing species. Some M dwarf planets are able to generate haze for the types of atmospheres considered here, but its cooling effects would be small because the M dwarf spectral output is in a wavelength range where these hazes are relatively transparent.

The outer edge of the habitable zone (OHZ) may also be affected by organic hazes. The OHZ is traditionally defined as the distance where CO₂ greenhouse warming is balanced by Rayleigh scattering from additional CO₂. In principle, the warming potential for organic-rich planets at the OHZ would be limited by the formation of haze and its attendant antigreenhouse effect. This process could define the "maximum greenhouse effect" for organic-rich worlds orbiting G and K stars. However, all of this is subject to the caveat that habitable planets near the OHZ may not be able to generate organic hazes in the first place. If the maximum CO₂ greenhouse limit allows for several bars of CO₂, then implausibly large CH₄ fluxes may be required to achieve a high enough CH₄/CO₂ ratio to create a haze in such atmospheres.

The haze's broadband UV and blue wavelength absorption feature is prominent and distinctive in reflected light, but to ensure accurate interpretation of this feature, it is important to explore similar UV absorbers that might mimic this feature in a planet's spectrum. We compare this haze's spectral signature with other short-wavelength absorbers in Figure 10, which plots hazy Archean Earth alongside modern Earth with clouds, Venus, Mars, modern Earth with a Mars-like surface, and Earth with a ZnS haze. The ZnS spectrum is not intended to be physically realistic and is provided simply to show the absorptive effects of ZnS particles, which have strong UV absorption similar to organic haze. We also show hazy Archean Earth with water clouds constructed using a weighted average of 50% haze-only, 25% haze and cirrus cloud, and 25% haze and stratocumulus cloud (Robinson et al. 2011, and see also our discussion of water clouds in hazy Archean spectra in Arney et al. 2016).

Zoom In Zoom Out Reset image size

In the modern Earth atmosphere, the ozone Chappuis band is a broad feature centered near 0.5–0.7 μm, but its absorption does not continue farther into the UV the way haze does, so the Rayleigh scattering slope becomes prominent for λ < 0.5 μm, distinguishing this spectrum from a hazy one. In addition, hydrocarbon hazes are unlikely to be present in atmospheres with spectrally apparent ozone (Section 4.4).

Mars is red because iron oxide absorbs strongly at blue wavelengths. A spectrum of the Earth with the wavelength-dependent surface albedo of Mars shows what Earth could look like with a surface rich in iron oxide. On an Earthlike planet with a 1 bar atmosphere, the blue-absorbing iron oxide feature is unlikely to be mistaken for hydrocarbon haze because of increased reflectivity for λ < 0.5 μm due to Rayleigh scattering. It is also important to note that at low spectral resolution, iron oxide could also be mistaken for ozone absorption. However, Rayleigh scattering is non-apparent on Mars itself, so the strong iron oxide absorption feature could mimic haze. Mars's iron oxide feature could be distinguished from haze by the absence of CH₄ features in the spectrum.

Earth with a thick haze of ZnS particles is the closest mimic we found to our cloud-free spectrum of Archean Earth with a hydrocarbon haze. However, ZnS is not a realistic aerosol candidate for Earthlike atmospheres because it condenses at temperatures close to 1000 K (Morley et al. 2012). Charnay et al. (2015) and Morley et al. (2015) show what a ZnS haze would look like in a more realistic atmosphere for GJ 1214b. Therefore, constraints on surface temperature or semimajor axis could eliminate this as a potential source of UV absorption.

Venus's broad UV absorption caused by its unknown UV absorber (Markiewicz et al. 2014) and SO₂ can also mimic the UV absorption of organic haze. However, the Venus spectrum lacks CH₄ features and strong water features that would indicate habitability. Venus's oxidizing atmosphere is very different from that of Archean Earth.

Organic haze's blue and UV wavelength absorption feature together with observations of methane bands would strongly imply the existence of haze in an atmosphere. The UV absorbers we compared to here can be distinguished from organic haze through the appearance of the Rayleigh scattering slope or the lack of CH₄ features, or else (in the case of ZnS) they are extremely unlikely for an Earthlike atmosphere.

Although the F2V planet produces a significant number of oxygen radicals compared to our other stars, the absolute level of oxygen in its atmosphere is not large enough to be detectable. The column densities of O₂ and O₃ in the F2V atmosphere are $9.0\times {10}^{18}$ molecules cm⁻² and $2.61\times {10}^{14}$ molecules cm⁻², respectively. The column density of methane is $4.27\,\times {10}^{22}$ molecules cm⁻². As discussed in Domagal-Goldman et al. (2014), it is difficult to accumulate abiotic oxygen at detectable levels in atmospheres rich in organics because reactions with reduced gases are major oxygen sinks. Domagal-Goldman et al. (2014) show that at lower CH₄ column densities than the ones we simulate here (e.g., ∼10¹⁶ molecules cm⁻²), O₂ and O₃ can reach column densities of 10¹⁸–10²¹ molecules cm⁻² and 10¹⁶–10¹⁸ molecules cm⁻², respectively, and O₃ can produce significant spectral signatures. Another study, by Harman et al. (2015), showed that O₂ and O₃ from CO₂ photolysis can produce spectral signatures in atmospheres with low CH₄ mixing ratios different from those simulated here. However, the model of Harman et al. (2015) produces similar O₃ and O₂ column depths when using similar assumptions to ours for the CH₄ and CO₂ mixing ratios, and broadly reproduces the trends seen here between different stellar types (C. Harman 2016, personal communication).