C/O RATIO AS A DIMENSION FOR CHARACTERIZING EXOPLANETARY ATMOSPHERES

At high temperatures, two distinct regimes of atmospheric chemistry can be identified based on the C/O ratio. Figure 1 shows the mixing ratios of major O and C based molecules as a function of the C/O ratio, for nominal conditions in the observable atmospheres of hot Jupiters (T = 2000 K, P = 1 bar). A sharp transition in the abundances of all the molecules, except CO, occurs at C/O = 1. For example, in going from the oxygen-rich solar abundances with C/O = 0.5 to a carbon-rich value of 1.5, the H₂O mixing ratio drops by about three orders of magnitude and the CH₄ mixing ratio increases by over three orders of magnitude. In addition, several hydrocarbons besides CH₄, such as C₂H₂, C₂H₄, and HCN, also become markedly abundant for C/O ⩾ 1, whereas CO₂ becomes negligible, as shown in Figure 1. These extreme changes represent the two distinctly observable differences between oxygen-rich (C/O < 1) and carbon-rich (C/O > 1) atmospheres at high temperatures, which make hot Jupiters ideal laboratories for constraining C/O ratios. On the other hand, at these T and P the CO abundance for C/O between 0.5 and 1.5 remains nearly unchanged in comparison. Consequently, the CO/H₂O ratio, which is nearly unity for C/O = 0.5, can attain values of up to 10–10⁴ for higher C/O ratios.

Zoom In Zoom Out Reset image size

The influence of C/O on the molecular abundances is temperature dependent, as shown in Figure 2. The mixing ratios of the major species H₂O, CO, and CH₄ in a hydrogen-dominated atmosphere are governed by the following summary reaction:

$$\begin{equation} \mathrm{CH_4 + H_2O \rightleftharpoons CO + 3 H_2.} \end{equation} \tag{ 1 }$$

Zoom In Zoom Out Reset image size

At a nominal pressure (P) of ∼1 bar, reaction (1) favors the conversion of CH₄ to CO at high T (≳1300 K), and the reverse reaction is favored at lower T. Consequently, CH₄ is the dominant carbon-bearing molecule at low T and CO is the dominant carbon-bearing molecule at high T. In an atmosphere with a solar abundance composition (C/O = 0.5) which has surplus oxygen, most of the oxygen is contained in H₂O at low T. At high T, about half the available oxygen pairs up with the carbon atoms in CO, and the remaining half is present in H₂O. Therefore, H₂O is abundant at all T for a solar abundance atmosphere. On the other hand, for C/O ⩾1, H₂O is still the dominant oxygen carrier at low T (≲ 1200 K) where CO is minimal, but at high T where CO is dominant, there is little excess oxygen available to form H₂O, contrary to the solar abundance case. The excess carbon available for the C/O >1 case is present in the form of abundant CH₄ and higher hydrocarbons such as C₂H₂, C₂H₄, and HCN.

The relative changes in the H₂O and CH₄ abundances between the oxygen-rich and carbon-rich regimes as a function of temperature are shown in Figure 3. The figure shows enhancements in CH₄ and depletion in H₂O for C/O = 1 and C/O = 2, compared to C/O = 0.5 (solar value); the absolute mixing ratios for the different C/O ratios are shown in Figure 4. For T ≲ 1200 K at P ∼ 1 bar, CO is less abundant, and CH₄ is the dominant carbon carrier and H₂O the dominant oxygen carrier, for all C/O ratios. In such atmospheres, estimating the C/O ratio requires high precision, to within a factor of two, estimation of H₂O and CH₄ abundances; currently molecular abundances for exoplanetary atmospheres are measured to only within a factor of 10 at best (Madhusudhan & Seager 2009; Madhusudhan et al. 2011b). On the other hand, for T ≳ 1200 K at P ∼ 1 bar, as is the case for the majority of transiting exoplanets known, the CH₄ enhancement and H₂O depletion increase with temperature, and can be over two (three) orders of magnitude for C/O = 1 (C/O = 2) and T ≳ 2000. In absolute terms, as shown in Figure 4, for C/O ⩾1 the CH₄ volume mixing ratios can be as high as ∼10⁻⁵ even at T ≳ 2500 K, contrary to values ≲ 10⁻⁸ predicted by a solar C/O ratio. At the same temperatures, for C/O ⩾ 1, the H₂O mixing ratios can be ≲ 10⁻⁶, while solar abundances predict ∼5 × 10⁻⁴. These CH₄ and H₂O mixing ratios for C/O ⩾1 at high T are in agreement with the observational constraints for WASP-12b reported by Madhusudhan et al. (2011b). Additionally, other hydrocarbons such as C₂H₂, C₂H₄, and HCN, which had not been considered by Madhusudhan et al. (2011b), are also expected to be present in the atmosphere of WASP-12b, as has also been suggested by previous studies (Madhusudhan et al. 2011c; Kopparapu et al. 2012). In Section 5.6, we demonstrate that model spectra which include these species are consistent with spectroscopic observations of WASP-12b.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The C/O ratio places important constraints on the likelihood of thermal inversions in exoplanetary atmospheres, as discussed in detail in Madhusudhan et al. (2011c). Here, we briefly summarize those results. Thermal inversions are caused by the presence of chemical species in the observable atmosphere which are strong absorbers of incident UV and visible light. Traditionally, TiO and VO have been proposed to cause thermal inversions in highly irradiated hot Jupiter atmospheres (Hubeny et al. 2003; Fortney et al. 2008, but cf. Spiegel et al. 2009). However, Madhusudhan et al. (2011c) show that for C/O ⩾1, the TiO and VO abundances in chemical equilibrium are lower by about two orders of magnitude compared to their abundances for a solar C/O ratio which was assumed in previous models of thermal inversions (Burrows et al. 2008; Fortney et al. 2008; Spiegel et al. 2009). Non-equilibrium effects such as gravitational settling can further deplete the abundances (Spiegel et al. 2009). Additionally, stellar activity might also affect the photochemical stability of inversion, causing compounds (Knutson et al. 2010). Consequently, if C/O ⩾1, even the most highly irradiated atmospheres cannot host thermal inversions due to TiO/VO, which might explain the lack of a thermal inversion in WASP-12b (Madhusudhan et al. 2011b, 2011c). Conversely, very highly irradiated atmospheres which do not host thermal inversions may likely be carbon-rich.

C/O ratios derived from observations can help constrain the chemical nature of inversion-causing absorbers. Besides TiO and VO, other oxygen bearing molecules are also unlikely to cause thermal inversions in C-rich atmospheres. For C/O ⩾1, almost all the oxygen is occupied by CO, which is the dominant O bearing molecule at high T, leaving little O for other oxygen-bearing molecules. As such, if thermal inversions are detected in high C/O atmospheres, they must be formed either by hydrocarbons or by other oxygen-free compounds, e.g., sulfur species (Zahnle et al. 2009). Hydrocarbon-based photochemical products are responsible for thermal inversions in giant planets in the solar system, which are at much lower temperatures (T ∼ 50–150) compared to hot Jupiters (Chamberlain 1978; Yung & Demore 1999). However, whether such absorbers can sustain in highly irradiated hot Jupiters (T ∼ 1000–3000) remains to be investigated in future studies. Thus, a simultaneous constraint on the C/O ratio and the presence of a thermal inversion from a given data set are essential to characterize the chemistry of inversion-causing absorbers in irradiated giant planet atmospheres.

The measurability of the C/O ratio in an exoplanetary atmosphere depends on the characteristic temperature of the atmosphere. As shown in Figures 2 and 3, the variations in the H₂O and CH₄ mixing ratios with the C/O ratio are significant only for T ≳ 1200 K. Above 1200 K, the differences in the H₂O and CH₄ abundances between the O-rich and C-rich regimes increase with T, reaching factors as high as 10³ (see Figure 3). Consequently, spectra of hotter atmospheres are generally more likely to carry stronger signatures of their C/O ratios and would hence be more effective in distinguishing between low (<1) and high (>1) C/O models.

The chemistry in the observable atmosphere (at P ∼ 0.01–1 bar; supplementary information of Madhusudhan et al. 2011b) is influenced by the temperature and pressure in the lower layers of the atmosphere. Vertical mixing in the atmosphere transports species from hotter lower regions of the atmosphere to upper regions, resulting in nearly uniform concentrations in the observable atmosphere (Cooper & Showman 2006; Zahnle et al. 2009; Line et al. 2010; Madhusudhan & Seager 2011; Moses et al. 2011). Therefore, the pressure level from where the species can be transported up (called the "quench" pressure, P_q) and the temperature (T_q) there are the equilibrium conditions governing the observed concentrations of the species.

While P_q and T_q can be computed for planet-specific models based on non-equilibrium chemistry (e.g., Madhusudhan & Seager 2011; Moses et al. 2011), we adopt generic values for our current purpose. For irradiated giant planet atmospheres, such as HD 189733b and HD 209458b, P_q typically lies in the 1–10 bar range (Cooper & Showman 2006; Madhusudhan & Seager 2011; Visscher & Moses 2011). We assume a nominal value of P_q = 1 bar; larger P_q requires stronger vertical mixing. As for T_q, the dayside P–T profiles of strongly irradiated atmospheres asymptote to isotherms at high optical depth, for P ∼ 1–10 bar (Burrows et al. 2008; Hansen 2008; Guillot 2010). The corresponding temperature (T_q) is a function of the visible and infrared opacities, the Bond albedo, and day–night energy redistribution (Hansen 2008; Guillot 2010; Heng et al. 2012), and can be quite high (up to a factor of ∼1.4) compared to the equilibrium temperature which is given by

$$\begin{equation} T_{\rm eq} = T_\star [(1-A_B)(1-f_r)(R_\star ^2/2a^2)]^{1/4}, \end{equation} \tag{ 2 }$$

where, T_⋆ and R_⋆ are the effective temperature and radius of the host star, and a is the orbital separation. A_B is the Bond albedo and f_r is the fraction of incident flux redistributed to the night side (Madhusudhan & Seager 2009). Consequently, based on comparisons with published profiles of several irradiated atmosphere models (Burrows et al. 2008; Fortney et al. 2008; Madhusudhan & Seager 2009), we find that a lower limit on T_q can be approximated by

$$\begin{equation} T_{\rm q} \gtrsim T_\star [R_\star ^2/2a^2]^{1/4} = T_{\rm eq}|_{f = 0}, \end{equation} \tag{ 3 }$$

where f = 0 stands for A_B = f_r = 0.

We note that our estimation of T_q is only nominal, since several factors of non-equilibrium chemistry, such as vertical eddy mixing and photochemistry, can play a role in determining the quench levels of the species and T_q can be different for different species (Moses et al. 2011).

The above estimates for P_q and T_q can be used to identify systems that are most suitable for C/O measurements. Irradiated atmospheres with T_q ≳ 1200 K at P_q = 1 bar can be expected to contain measurable signatures of their C/O ratios, as shown in Figures 2 and 3. The signatures are enhanced with temperature and are maximum for T ≳ 2000 K.

The chemistry in C-rich (C/O ⩾ 1) and O-rich (C/O < 1) atmospheres presents distinct spectroscopic signatures that are observable with current and forthcoming instruments. Atmospheres of transiting planets can be observed in transit spectra as well as in thermal spectra at occultation.⁴ A transit spectrum constrains the chemistry at the limb of the planetary atmosphere, whereas a thermal spectrum constrains the chemistry as well as the temperature structure in the dayside atmosphere of the planet (Madhusudhan & Seager 2009).

In this section, we briefly illustrate the effect of the C/O ratio on the spectra of highly irradiated hydrogen-dominated planets. Models for cooler atmospheres have been reported by Kuchner & Seager (2005) and Madhusudhan et al. (2011c). For ease of illustration, we consider only four molecules (H₂O, CO, CH₄, and CO₂) in the models discussed here; detailed models including additional opacities due to HCN and C₂H₂ that are possible in C-rich atmospheres will be demonstrated in Section 5. Model transit and thermal spectra of transiting hot Jupiters with different C/O ratios and temperature profiles are shown in Figures 5 and 6. Here, we use the planetary and stellar parameters of the canonical hot Jupiter HD 189733b, and vary the temperature profile (adapted from Madhusudhan & Seager 2009) to span different high-temperature regimes, which is equivalent to changing the orbital separation.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We generate model spectra for two different temperature profiles, shown in the insets in Figure 6, with different quench temperatures. The quench temperature (T_q), discussed in Section 3.3, is the temperature of the lower atmosphere (at P ∼ 1 bar), which typically governs the chemistry in the observable atmosphere. Spectra are shown for T_q = 2000 K and T_q = 2600 K, and for C/O ratios of 0.5, 1.0, and 2.0, for each T_q. For both T_q, models with C/O = 0.5 show strong H₂O features near 1.1 μm, 1.4 μm, 1.8 μm, and 3 μm, which are clearly distinguishable from the lack of the same features for models with the C/O ⩾ 1. On the other hand, models with C/O ⩾ 1 have enhanced CH₄ abundances which cause greater absorption in the CH₄ bands near 1.2 μm, 1.3 μm, 2.3 μm, 3.3 μm, and 7.8 μm. However, the strength of the CH₄ features depends on the temperature and C/O ratio. For T_q = 2000, the methane features are only marginally detectable for a model with C/O = 1, but become the dominant spectroscopic features for C/O = 2. On the other hand, for T_q = 2600, only the strongest CH₄ bands, at 3.3 μm and 7.8 μm, are prominent for C/O ⩾ 1. The reason for the relatively weaker CH₄ features for T ∼ 2600, even for C/O ⩾ 1, is that even though the CH₄ enhancement over solar models is as high as for T ∼ 2000, the absolute CH₄ abundances are lower by a factor of ∼10, as shown in Figure 2. Finally, all the models show strong CO features at 2.3 μm and 4.8 μm, since CO is a prominent molecule in both C-rich and O-rich regimes.

The spectroscopic features discussed above can also be observed in thermal emission spectra observed at secondary eclipse. Model spectra of hot Jupiters in thermal emission are shown in Figure 6. In the absence of thermal inversions, spectra of high temperature O-rich models show strong absorption in the H₂O bands and weak absorption in CH₄ bands, and vice versa for C-rich models. It must be noted that for thermal spectra stronger absorption means lower flux and hence a weaker signal.

The hot Jupiter XO-1b (McCullough et al. 2006) has presented one of the extreme counterexamples to the TiO/VO hypothesis to date. The planet orbits a G dwarf on a 4 day orbit, and is one of the least irradiated hot Jupiters whose atmospheres have been observed to date, with an equilibrium temperature of ∼1400 K. As such, Fortney et al. (2008) predicted that the dayside atmosphere of XO-1b would not be able to host a thermal inversion since it would not be hot enough to maintain gaseous TiO or VO in the upper atmosphere. Subsequently, Machalek et al. (2008) reported observations of thermal emission from XO-1b in the four channels of Spitzer IRAC photometry (at 3.6 μm, 4.5 μm, 5.8 μm, and 8 μm). Interpreting the data using models of Burrows et al. (2008), Machalek et al. (2008) reported a thermal inversion in the dayside atmosphere of XO-1b. Given the low irradiation received at its substellar point, the inference of a thermal inversion in XO-1b presented a stark violation of the TiO/VO hypothesis.

One potential explanation to the apparent contradiction was proposed by Knutson et al. (2010) who found a correlation between inferences of thermal inversions in hot Jupiters and the chomospheric activities of their host stars. They concluded that planets orbiting chromospherically active stars did not show evidence for thermal inversions in their atmospheres, and those orbiting quiet stars showed evidence for thermal inversions. Given the low chromospheric activity of its host star, the thermal inversion in XO1b appeared to be consistent with the empirical hypothesis of Knutson et al. However, it is still unknown to date as to what molecules in the atmosphere could actually cause a thermal inversion in XO-1b that is strong enough to match the observations of Machalek et al. (2008), since gaseous TiO and VO would be clearly underabundant in such a cool atmosphere (Spiegel et al. 2009). In order to create the thermal inversion in their models, Machalek et al. had to use an artificial opacity source in visible wavelengths with a parametric opacity, wavelength range, and location in the atmosphere (Burrows et al. 2008). Whether such a source is physically plausible in XO-1b is yet to be demonstrated.

In the present work, we report a consistent explanation to the observations of Machalek et al. (2008). The observations and two model spectra are shown in Figure 9 and the corresponding chemical compositions are shown in Table 1. The four IRAC observations are shown in blue. First, we confirm the findings of Machalek et al. that a solar abundance model cannot fit the data without a thermal inversion, as shown in the green curve in Figure 9. The model fits the data at 3.6 μm, 4.5 μm, and 8 μm to within the 1σ error bars, but underpredicts the 5.8 μm flux by ∼3σ. Our non-inverted solar abundance model provides a slightly better fit to the data than that of Machalek et al. (2008), while still grossly underpredicting the flux at 5.8 μm, in agreement with Machalek et al. Furthermore, the temperature structure required for a best-fit non-inverted solar abundance model as shown in Figure 9 is nearly an isotherm, implying that the spectrum is equivalent to a blackbody, as shown by the dotted lines, making the O-rich composition irrelevant. The requirement of a thermal inversion with a solar abundance model thus rests largely on the flux observed in the 5.8 μm IRAC bandpass which has strong spectral features of H₂O. In a solar composition model, H₂O is the dominant oxygen bearing species at the temperatures of XO-1b. If there is no thermal inversion in the atmosphere, in which case temperature monotonically decreases outward, a low flux is observed in the 5.8 μm IRAC band due to strong water absorption. The contrary observation of the high 5.8 μm flux in the data implies, in the context of a solar abundance model, that it must be due H₂O in emission rather than absorption. This is possible in the presence of a thermal inversion.⁵

Zoom In Zoom Out Reset image size

We report, however, that the observations of XO-1b can be explained without the requirement of an ad hoc thermal inversion if the atmosphere of XO-1b is carbon-rich (C/O ⩾1). A model spectrum of XO-1b with a C-rich composition and a non-inverted temperature profile is shown in red in Figure 9. The model fits all the observations to within the 1σ uncertainties, on average. The key to explaining the high flux in the 5.8 μm channel, simultaneously with the other photometric points, without a thermal inversion lies in the H₂O and CH₄ abundances expected in a C-rich atmosphere (see, e.g., Madhusudhan et al. 2011b). At the temperatures of XO-1b, C/O ⩾1 causes ∼10–100 times depletion of H₂O and ∼10–100 times enhancement of CH₄ compared to a solar abundance model (with C/O = 0.5). Additionally, other hydrocarbons such as HCN and C₂H₂ are also expected to be abundant. The low H₂O abundance causes low absorption in the 5.8 μm channel, leading to a high thermal flux emitted from the deeper layers of the atmosphere. On the other hand, the low fluxes in the 3.6 μm and 8 μm channels are explained by enhanced CH₄, HCN, and C₂H₂ absorption in both channels. The 4.5 μm flux in the C-rich model is explained by absorption due to CO and HCN. Therefore, we classify XO-1b as a C2 class planet in the framework of the 2D classification scheme presented in Section 4.

Future observations in the near-infrared with existing facilities will be required to confirm the C2 classification of XO-1b. Given that our current inference is based solely on the one photometric point at 5.8 μm, new observations are required to confirm the potentially C-rich nature of XO-1b. As shown in Figure 9, the high C/O models fitting the IRAC data predict substantially higher fluxes in the 2–3 μm range compared to a solar composition model. This difference is largely due to different continuum emission between the two models; the high C/O solutions have higher temperatures in the low atmosphere (P ≳ 1 bar) which are required to fit the high 5.8 μm IRAC point. Such differences can be used to constrain the models using ground-based high precision photometry (e.g., Lopez-Morales et al. 2010; Croll et al. 2011) and spectrophotometry (Bean et al. 2010; Mandell et al. 2011; Crossfield et al. 2011), and spaceborne spectroscopy with Hubble Space Telescope (HST) in the near term, and with JWST in the long term. Note that precisions better that 10⁻⁴ would be required to distinguish between the C-rich and O-rich models.

Care must be taken, however, in the interpretation of the C/O of XO-1b using transmission spectroscopy. Given the low irradiation levels received by the planet, a transmission spectrum of XO-1b is likely to probe even cooler regions of the atmosphere. It is possible that the temperatures in such regions can be well below the 1200 K minimum required for high-confidence estimates of C/O ratios (see Section 3.1). Consequently, spectra of the limb of the atmosphere may show absorption features of H₂O even if the C/O ⩾ 1. In such cases, establishing the C/O ratio accurately would require estimating the H₂O and CH₄ abundances to precisions within factors of a few. Such precisions would require extremely precise (10⁻⁵) spectroscopic observations which may not be feasible with current instruments. At present, one could attempt to derive such molecular abundances using the HST NICMOS spectrum of XO-1b reported by Tinetti et al. (2010), whose reported constraints agree with a potential C-rich atmosphere. However, we refrain from interpreting their data in the present work since their data reduction is currently contested (Gibson et al. 2010).

The hot Jupiter CoRoT-2b (Alonso et al. 2008) presents yet another challenge for atmospheric characterization of hot Jupiters using traditional models and classification schemes. The planet orbits a late G dwarf on a 1.7 day orbit and is one of the highly irradiated hot Jupiters predicted to host a thermal inversion in its dayside atmosphere (Fortney et al. 2008). However, the host star is known to be very chromospherically active, which according to the Knutson et al. (2010) hypothesis would suggest that there should not be a thermal inversion in its dayside atmosphere. Between the two competing predictions, however, lies the uncomfortable fact that neither models with thermal inversions nor models without thermal inversions have been able to explain the observed spectrum of CoRoT-2b (Gillon et al. 2010; Deming et al. 2011). Observations of dayside thermal emission from CoRoT-2b showed an anomalously low flux in the 8 μm Spitzer IRAC channel and high flux in the 4.5 μm channel, both of which could not be explained by previous atmospheric models (Alonso et al. 2010; Gillon et al. 2010; Deming et al. 2011; Guillot & Havel 2011). As an alternative explanation, Deming et al. (2011) required either excess emission from tidal mass loss that is too high to be realistic or the presence of an unknown absorber that selectively absorbs at wavelengths below 5 μm.

In the present work, we are able to explain all the existing observations of CoRoT-2b. Figure 10 shows the observations and two model spectra fitting the data, and the corresponding chemical compositions are shown in Table 1. The data include broadband photometric detections of thermal emission from CoRoT-2b in four infrared channels: a ground-based K_s band at 2.1 μm (Alonso et al. 2010) and three Spitzer IRAC bandpasses at 3.6 μm (Deming et al. 2011), 4.5 μm, and 8 μm (Gillon et al. 2010; Deming et al. 2011). The models shown in Figure 10 include an O-rich model (C/O < 1) and a carbon-rich model (C/O > 1). The corresponding temperature profiles are shown in the inset of Figure 10. As shown in the P–T profiles, the observations rule out a thermal inversion in the dayside atmosphere of CoRoT-2b. The low flux in the 8 μm IRAC channel, thought to be anomalous by previous studies, can be naturally explained by CH₄, HCN, and C₂H₂ absorption in the carbon-rich model and by H₂O absorption in the oxygen-rich model. The same opacity sources also explain the low flux in the 3.6 μm channel.

Zoom In Zoom Out Reset image size

Existing observations of CoRoT-2b are less constraining of its C/O ratio, implying that it could be classified as O1 or C2 according to the 2D classification scheme proposed in Section 4, although the C2 class is marginally favored. We find that all the observations can be explained to within the 1σ uncertainties on average by a wide range of models with carbon-rich (C/O > 1) and oxygen-rich (C/O < 1) compositions. The Spitzer IRAC 5.8 μm flux that implied a substantial depletion of H₂O in XO-1b is unavailable for CoRoT-2b, leading to a poorer constraint on the H₂O abundance in the latter. Thus, it is expected that both H₂O-rich and H₂O-poor solutions are possible for CoRoT-2b, implying low C/O (<1) and high C/O (>1) ratios, respectively. This degeneracy occurs because H₂O, CH₄, and HCN, all have absorption features in the 3.6 and 8 μm IRAC channels. A 5.8 μm observation could have broken the degeneracy with an independent constraint on the H₂O abundance; the IRAC 5.8 μm bandpass does not contain significant CH₄ features. However, since the C-rich model provides a marginally better fit to the data, especially at 8 μm, we tentatively classify CoRoT-2b as a C2 class planet in the framework of our 2D classification scheme proposed in Section 4.

New observations of CoRoT-2b would be required to conclusively constrain its C/O ratio. As shown in Figure 10, the two spectral models differ substantially in their absorption features. In particular, observations using the HST Wide Field Camera 3 (WFC3) instrument in the 1.1–1.7 μm bandpass would be able to conclusively determine the H₂O abundance based on the strong H₂O feature in the O-rich model around 1.4 μm. In addition, ground-based observations with higher precision than that currently available in the K_s band would be able to better constrain the lower atmosphere temperature profile of the atmosphere, which in turn could differentiate between the different models. Another alternative would be to use transmission spectroscopy and photometry from ground and space to constrain the atmospheric composition of the day–night terminator of the planet.

WASP-14b is a transiting hot Jupiter orbiting an F5 main-sequence star at an orbital separation of 0.036 AU (Joshi et al. 2009). With an incident irradiation of 2.5 × 10⁹ erg cm⁻² s⁻¹, leading to a zero-albedo equilibrium temperature of ∼2200 K, WASP-14b belongs to the class of highly irradiated hot Jupiters that are expected to host strong thermal inversions based on the TiO/VO hypothesis of Fortney et al. (2008). However, using three channels of Spitzer IRAC photometry, at 3.6 μm, 4.5 μm, and 5.8 μm, Blecic et al. (2012) reported the lack of a thermal inversion in the dayside atmosphere of WASP-14b. The absence of a thermal inversion in WASP-14b was evident primarily from a low brightness temperature of ∼1600 K in the 8 μm channel compared to temperatures above 2200 K observed in the lower wavelength channels at 3.6 μm and 4.5 μm, indicating temperature decreasing outward in the atmosphere.

While the lack of a thermal inversion in WASP-14b may contradict the prediction of Fortney et al. (2008), it is also possible that condensation and gravitational settling of TiO/VO (Spiegel et al. 2009) might preclude the formation of a thermal inversion. It is unclear if photospheric activity of the host star may have caused the disappearance of TiO and VO. Knutson et al. (2010) find that host star WASP-14 is not very active implying that the planet should have had a thermal inversion. But Knutson et al. (2010) also highlight the difficulty in accurately establishing the Ca ii H and K indices for F stars, such as WASP-14b. Alternately, as shown in Madhusudhan et al. (2011c) and in the present work, it may also be possible that the atmosphere of WASP-14b is carbon-rich (i.e., C/O ⩾ 1) in which case TiO and VO are naturally low in abundance.

We confirm the conclusions of Blecic et al. (2012) that the limited data available cannot conclusively constrain the C/O ratio of the dayside atmosphere of WASP-14b. Two model spectra fitting the observations are shown in Figure 11, and Table 1 shows the model compositions. In the O-rich model, the dominant absorption in the 8 μm and 3.6 μm IRAC channels is caused by H₂O, whereas that in the C-rich model is caused by HCN, CH₄, and C₂H₂. Both models contain absorption features due to CO in the 4.5 μm bandpass. The temperature profiles for the two models are shown in the inset in Figure 11. Since the data rule out a thermal inversion in the atmosphere regardless of the C/O ratio, both models have similar lower atmospheric temperatures.

Zoom In Zoom Out Reset image size

Given that both C-rich and O-rich models explain the data almost equally well, WASP-14b could be classified either as an O1 or a C2 system. Note that despite the high irradiation, WASP-14b cannot be classified as an O2 system because of the lack of a strong thermal inversion and hence falls in the gray area between the O1 and O2 classes shown in Figure 7. However, since an oxygen-rich model (in green) provides a marginally better fit to the 8 μm IRAC point than the carbon-rich model (in red), one may tentatively classify WASP-14b as an upper-O1 class planet, as shown in Figure 8.

New observations are needed to establish the C/O ratio of WASP-14b. The degeneracy between the two models arises from the lack of a spectral measurement in the IRAC 5.8 μm channel. In principle, a 5.8 μm IRAC data point, if available, could have easily removed the degeneracy. Given that the 5.8 μm IRAC channel is no longer available on Spitzer, an alternative way to constrain the water abundance with existing facilities would be to observe the H₂O bands in the HST WFC3 bandpass between 1.1 and 1.7 μm.

Future confirmation of the C/O ratio of WASP-14b would lead to one of two interesting conclusions. If the C/O is found to be less than unity, based on HST observations, it will be a stringent constraint on the irradiation evidence for gravitational settling of TiO/VO according to the Spiegel et al. (2009), and will place a lower-limit on the irradiation level that defines the gray region between the O1 and O2 classes in Figures 7 and 8. Alternately, if the C/O is found to be ⩾1, WASP-14b will join WASP-12b in the new class of carbon-rich planets (CRPs; Madhusudhan et al. 2011c).

The transiting hot Jupiter WASP-19b orbits a G8 main-sequence star at a separation of 0.017 AU (Hebb et al. 2010). At an irradiation level of 4 × 10⁹ erg cm⁻² s⁻¹, it is one of the very highly irradiated hot Jupiters with a zero-albedo equilibrium temperature of 2400 K. Consequently, it is also expected to host a strong thermal inversion due to TiO/VO if the atmosphere is in chemical equilibrium and the composition is oxygen-rich as predicted by Fortney et al. (2008). However, using six channels of broadband photometry obtained using Spitzer and ground-based facilities, Anderson et al. (2012) report the non-detection of a thermal inversion in the dayside atmosphere of WASP-19b. Their inference was based on observations of thermal emission in four channels of Spitzer IRAC photometry at 3.6 μm, 4.5 μm, 5.8 μm, and 8 μm, and two ground-based photometric measurements at 1.6 μm (Anderson et al. 2010) and 2.1 μm (Gibson et al. 2010). The inference of a non-inverted profile is evident from the higher brightness temperatures observed in the 1.6–3.6 μm bands compared to the longer wavelength bands, suggesting a temperature decreasing outward in the atmosphere.

Even though the lack of a noticeable thermal inversion in WASP-19b defies the predictions of the TiO/VO hypothesis of Fortney et al. (2008), alternate explanations exist. It might be unlikely that gravitational settling (Spiegel et al. 2009) alone can deplete all the TiO and VO in this highly irradiated atmosphere. However, Anderson et al. (2012) note that the planet orbits a chromospherically active star, in which case, according to the hypothesis of Knutson et al. (2010), the high UV flux incident on the planetary atmosphere could potentially dissociate the inversion-causing molecules. An alternative explanation, as suggested by Madhusudhan et al. (2011c), could be that the atmosphere is carbon-rich in which case TiO and VO would be naturally underabundant and hence the lack of a thermal inversion. Therefore, WASP-19b could belong to either class O1/O2 or C2, depending on the C/O ratio.

Current observations of WASP-19b are inconclusive about its C/O ratio but nominally favor a C/O ⩾ 1. The data and two model spectra of WASP-19b are shown in Figure 12 (see Table 1 for the chemical compositions of the models). As demonstrated by Anderson et al. (2012), the majority of available data are fit equally well by O-rich as well as C-rich models. The large error bar on the flux in the 5.8 μm Spitzer IRAC channel makes it difficult to constrain the H₂O abundance which is one of the major discriminator between the O-rich and C-rich models. And, as in the case of WASP-14b, the remaining three Spitzer data at 3.6 μm, 4.5 μm, and 5.8 μm are also fit equally well by both models, although the 3.6 μm flux is better explained by the C-rich model. The opacity sources in the different IRAC channels are the same as those discussed for WASP-14b. The photometric data at 1.6 μm and 2.1 μm do not particularly discriminate between the two models because both models have similar lower atmospheric temperatures since neither of them hosts a thermal inversion.

Zoom In Zoom Out Reset image size

However, an important discrimination is provided by the z'-band observation (Burton et al. 2012) centered at 0.9 μm, in which TiO is a prominent opacity source for the O-rich model. Even though TiO and VO may not be abundant enough in the upper atmosphere, due to either gravitational settling or chromospheric activity, in an O-rich atmosphere it should still be present in the lower atmosphere around P ∼ 1 bar where the temperatures are hottest and where the incident irradiation does not reach with significant intensity. Consequently, if C/O < 1 and there is no thermal inversion, highly irradiated atmospheres like WASP-19b should display noticeable TiO and VO absorption between 0.7 and ∼1.1μm, which is observable as decreased flux in the z', z, and J bands, as can be seen in Figure 12. The flux observed in the z' band is better fit by the C-rich model which does not have significant TiO/VO absorption. Consequently, we tentatively classify WASP-19b as a C2 class planet in Figure 8 in view of the marginally better fit provided by the C-rich model at 0.9 μm and 3.6 μm.

New observations are required to conclusively determine the C/O ratio of WASP-19b and to classify it as O2 or C2. As shown in Figure 12, new ground-based data in the z and J bands will be able to discriminate between the different flux continua at 1 μm and 1.2 μm predicted by the O-rich and C-rich models. Additionally, observations with HST WFC3 would be able to provide a robust discrimination between the two models in the H₂O bands between 1.1 μm and 1.7 μm.

The transiting hot Jupiter WASP-33b, orbiting an A5 main-sequence star at an orbital separation of 0.03 AU, is one of the most highly irradiated hot Jupiters known (Collier-Cameron et al. 2010; Herrero et al. 2011). At an incident irradiation exceeding 10¹⁰ erg cm⁻² s⁻¹, and an equilibrium temperature exceeding 3300 K, it the hottest planet in our sample and is comparable in temperature to low-mass stars. Consequently, WASP-33b is the most suitable hot Jupiter to study thermal and chemical characteristics of irradiated giant planets. If the atmosphere of WASP-33b is oxygen-rich, TiO and VO should be abundant in the atmosphere, despite any gravitational settling. Consequently, an oxygen-rich WASP-33b would offer the most unambiguous opportunity to detect a thermal inversion in a hot Jupiter atmosphere and place it in the O2 class discussed in Section 4. On the other hand, the lack of a thermal inversion in WASP-33b would be a strong indication of a carbon-rich atmosphere, i.e., a C2-class planet.

Observations of the dayside atmosphere of WASP-33b nominally allow both possibilities, O2 as well as C2, thus necessitating new observations to break the degeneracy. The available data comprise of ground-based photometry in the z' band (Smith et al. 2011) and the K_s band (Deming et al. 2012), and Spitzer IRAC photometry in the 3.6 μm and 4.5 μm channels (Deming et al. 2012). Deming et al. (2012) conclude that all the existing observations are consistent with either a carbon-rich atmosphere without a thermal inversion (i.e., a C2-type atmosphere) or an oxygen-rich atmosphere with a thermal inversion (i.e., an O2-type atmosphere). While our results in the present work generally agree with the assessment of Deming et al., we note that a C2-type atmosphere for WASP-33b explains the data marginally better than one of O2 type. The data and two model spectra of WASP-33b are shown in Figure 13. Table 1 shows the chemical compositions of the models.

Zoom In Zoom Out Reset image size

The data in the z band and in the two Spitzer bands, at 3.6 μm and 4.5 μm, can be explained almost equally well by both models (C2 and O2), but for very different reasons. The high flux observed in the z' band is naturally explained by the lower atmosphere of the non-inversion model. And since that model is C-rich, no significant TiO or VO absorption is expected and hence the high predicted flux in the z' band. On the other hand, the O2 model which has a thermal inversion has a cooler lower atmosphere than the non-inverted model. However, because this model is O-rich, substantial TiO and VO is present and is seen as an emission feature, as opposed to absorption, due to the thermal inversion. Consequently, the O2 model predicts a higher brightness temperature in the z' band than its lower atmospheric temperature and hence explains the high observed z'-band flux. Similar reasoning explains the fits of both models to the Spitzer data at 3.6 μm and 4.5 μm. At 3.6 μm the low observed flux is explained by the inversion model based on the lower temperature of its lower atmosphere and lack of any significant opacity source at 3.6 μm in its oxygen-rich atmosphere for T ⩾ 2500 K. The same model explains the higher 4.5 μm flux due to CO in emission. On the other hand, the non-inverted C-rich model explains the 3.6 μm point due to strong absorption by HCN, CH₄ and C₂H₂ in the 3.6 μm channel and absorption due to CO and HCN in the 4.5 μm channel.

The K_s-band flux, however, favors a C-rich atmosphere without a thermal inversion. The O2 model predicts over 2σ lower planet–star flux contrast than that observed, whereas a C2 model fits the data point rather precisely, within the 1σ errors. The K_s band is devoid of strong molecular absorption and hence probes thermal emission from the deepest layer of a planetary atmosphere (P ∼ 1 bar) until continuum opacity makes the atmosphere optically thick. Consequently, the difference between the K_s-band fluxes predicted by the two models directly reflects the difference between the isothermal temperatures of the deep atmosphere in the two models. More generally, since temperature profiles with thermal inversions have cooler lower atmospheres compared to those without thermal inversions, for the same irradiation, thermal emission in opacity windows such as the J, H, and K bands tends to be typically lower for models with thermal inversion.

WASP-33b is one of the most ideal targets for follow-up observations with existing instruments. The very high planet–star flux contrast implies that thermal emission from the planet can easily be detected with routine observations in the near-IR as has already been demonstrated by Smith et al. and Deming et al. New ground-based observations in the J band (1.2 μm) and H band (1.6 μm), and HST observations with the WFC3 instrument should be able to conclusively distinguish between the two possibilities of an O2 and a C2 atmosphere. As shown in Figure 13, J- and H-band observations should be able to place tight constraints on the temperature in the lower atmosphere and hence the presence of a thermal inversion, given the large difference in flux contrast between the two models in those bands. Furthermore, an oxygen-rich WASP-33b must show strong emission features of H₂O in the HST WFC3 bandpass due to a thermal inversion in accordance with the TiO/VO hypothesis of Fortney et al. (2008). On the other hand, weak features in the water bands would indicate a carbon-rich atmosphere in WASP-33b.

The atmosphere of the transiting hot Jupiter WASP-12b (Hebb et al. 2009) has generated substantial interest in the recent past. Being one of the most highly irradiated hot Jupiters, with an equilibrium temperature of ∼2600 K, initial models assuming solar abundance (i.e., O-rich) compositions predicted strong thermal inversions to be present in the dayside atmospheres of planets like WASP-12b (Fortney et al. 2008). Even considering depletion due to gravitational settling, TiO was expected to be abundant in the dayside atmosphere of WASP-12b and, hence, cause a thermal inversion (Spiegel et al. 2009). Furthermore, an O-rich atmosphere would imply the presence of abundant H₂O and negligible CH₄ (Burrows & Sharp 1999). However, observations of the dayside atmosphere of WASP-12b suggested a stark contrast to previous theoretical predictions which assumed solar abundances. Using multi-band photometric observations over a wide wavelength range (1–10 μm) obtained with Spitzer (Campo et al. 2011) and from ground (Croll et al. 2011), Madhusudhan et al. (2011b) reported the detection of a carbon-rich atmosphere (C/O ⩾ 1), implied by an underabundance of H₂O and overabundance of CH₄, and the lack of a strong thermal inversion in the dayside atmosphere of WASP-12b. Consequently, it is apparent that WASP-12b cannot be characterized solely on the basis of incident irradiation, and assuming a solar C/O ratio, as per previous classification schemes (e.g., Fortney et al. 2008). However, the C/O ⩾ 1 and the lack of a strong thermal inversion in WASP-12b are both consistent with the properties of a C2-class planet according to our 2D classification scheme suggested in Section 4.

New data confirm a high C/O ratio and a weak thermal inversion in the dayside atmosphere of WASP-12b. Using new observations in the Spitzer IRAC channels at 3.6 μm and 4.5 μm, in addition to existing data, Cowan et al. (2012) confirmed that their best-fitting models required a high C/O ratio and the lack of a strong thermal inversion in the dayside atmosphere of WASP-12b (see Section 5.2.1 of Cowan et al.). It is to be noted, however, that since Cowan et al. used forward models (Burrows et al. 2008) for their interpretation, as opposed to a retrieval algorithm of Madhusudhan et al., a statistical estimate of the C/O ratio was not possible in the Cowan et al. study, despite their confirmation of a C/O ⩾1 implied by their high CO/H₂O requirement. New studies have also reported ground-based photometric (Zhao et al. 2012) and spectroscopic (Crossfield et al. 2012) observations of thermal emission from WASP-12b in the K band (around 2.1 μm), and confirmed previous measurements in the same bandpass by Croll et al. (2011). More recently, Swain et al. (2012) reported observations of WASP-12b in the near-infrared (1–1.7 μm) using the HST WFC3. Based on their thermal emission spectra, Swain et al. suggested that even though models with C/O ⩾ 1 best explained their data, they could not provide conclusive constraints on the C/O ratio. However, Swain et al. stress that explaining their data required a substantial paucity of H₂O (≲ 10⁻⁶). It is well known that such low levels of H₂O is possible if either C/O ⩾ 1 or if the overall metallicity of the planet is low (e.g., see Section 3; also see Seager et al. 2005; Madhusudhan et al. 2011c; Moses et al. 2012). In the latter case, both C/H and O/H would have to be underabundant in WASP-12b by about two orders of magnitude below the stellar values, a scenario which would not be able to explain the high CO required to explain the Spitzer data in Madhusudhan et al. (2011b) and Cowan et al. (2012), which Swain et al. did not consider including in their model fits. Furthermore, a hundred times lower metallicity (particularly C and O abundances) in a giant planetary atmosphere relative to its host star would be challenging to explain using the standard core accretion model for giant planet formation (see, e.g., Owen et al. 1999; Atreya & Wong 2005).

In the present work, we find that all existing observations of thermal emission from WASP-12b are consistent with a high C/O ratio and a weak thermal inversion in its dayside atmosphere, confirming the findings of Madhusudhan et al. (2011b). The data include four channels of ground-based near-infrared photometry in the z' (Lopez-Morales et al. 2010), J, H, and K bands (Croll et al. 2011), four-channels of Spitzer IRAC photometry (Campo et al. 2011; for 3.6 μm and 4.5 μm we use the new values from Cowan et al. 2012), and a near-infrared spectrum in the HST WFC3 bandpass (Swain et al. 2012). We have also updated our spectral models, over those used in Madhusudhan et al. (2011b), by including molecular opacities due to C₂H₂ and HCN, which could be important sources of opacity in carbon-rich atmospheres (see Section 3). Figure 14 shows all the data along with two spectral models, and the chemical compositions of the models are shown in Table 1. We find that a carbon-rich model (with C/O ⩾ 1) provides the best fit to all the data whereas a solar abundance model is inconsistent with the data, in agreement with the interpretation of Madhusudhan et al. (2011b). The primary difference between the fits in Madhusudhan et al. (2011b) and the present work lies in the interpretation of the absorption in the 8 μm IRAC channel. While Madhusudhan et al. (2011b) suggested that the low flux at 8 μm was solely due to CH₄ absorption, we find that the low flux could be due to a combination of CH₄ and HCN opacity. The uncertainties in high temperature opacities of hydrocarbons (Tinetti et al. 2010) make it difficult to judge the exact relative contributions of CH₄ and HCN, but it is possible that in C-rich atmospheres HCN may dominate the 8 μm opacity, as has also been suggested by Kopparapu et al. (2012) and Moses et al. (2012).

Zoom In Zoom Out Reset image size

We classify WASP-12b as a C2 class planet, based on all observations of its dayside atmosphere. The HST WFC3 spectrum reported by Swain et al. (2012) provides a particularly unambiguous constraint on the C/O ratio as shown in the inset of Figure 14. The WFC3 data by themselves alone conclusively rule out H₂O absorption (as also reported by Swain et al. 2012) which is a prominent feature around 1.4 μm, and hence rule out an O-rich atmosphere. Furthermore, the small absorption feature in the data at 1.57 μm appears to be consistent with the small C₂H₂ feature in the C-rich model, although the detection is only marginal. The data also rule out a strong thermal inversion in the atmosphere, which is apparent from the lack of any strong emission features in the spectrum, as shown in Figure 14.

Additional constraints on the atmospheric composition of WASP-12b are expected in future work. In this work, we have not considered observations of transmission spectra of WASP-12b which probe the atmospheric composition at the day–night terminator of the planet. Cowan et al. (2012) reported transmission photometry using Spitzer at 3.6 μm and 4.5 μm. They observed unusually high absorption in the 3.6 μm channel compared to the 4.5 μm channel, which they were unable to explain using any composition in their atmospheric models, in which the dominant sources of opacity were only H₂O and CO, and, probably, CH₄. Swain et al. (2012) reported transmission spectroscopy of WASP-12b using HST WFC3. In order to explain their data, Swain et al. required extremely low levels of H₂O (≲ 5 × 10⁻⁵) and a possible presence of TiH and CrH in the atmosphere. Such a composition is possible if either the atmospheric metallicity is at least two orders of magnitude lower than the stellar metallicity or if the metallicity is normal but the C/O ratio is ⩾1 as has been inferred on the day side by Madhusudhan et al. (2011b) and in the present work. Future observations of WASP-12b in transit in the near-infrared (between 0.7 μm and 1.1 μm) could also place constraints on the TiO/VO abundances in its atmosphere. In an upcoming effort, we plan to reanalyze all existing transmission data of WASP-12b and fit them with our models which include a broader set of opacities, than previous studies, including HCN and C₂H₂, to obtain statistical constraints on the atmospheric C/O ratio at the day–night terminator of WASP-12b.