FIGURE 1. A comparison of the observations with simulated water absorption.

In these simulated transmission spectra, the water mixing ratio profile is assumed to be constant² and equal to 5  10^-4. The observations are indicated with black triangles and error bars at 1, and the coloured rhombi and stars indicate the different models integrated over the IRAC bands. To match the observations, the planetary radius at 10 bar corresponds to a transit depth of 2.28%. Blue trace, colder terminator¹⁸ temperature–pressure profile; orange trace, warmer terminator¹⁸; black traces, constant temperature at 500 K (solid line) and 2,000 K (dotted line). Warmer temperatures increase the atmospheric scale-height (that is, the vertical distance over which the pressure decreases by a factor of e), hence the atmosphere is optically thick at higher altitudes. This explains the differences among the three classes of spectra at wavelengths shorter than 3.5 m and longer than 4.5 m, where the water opacities are far less temperature-dependent. The opposite is true for the water opacities in the 3.5–4.5 m wavelength range, which might be orders of magnitude smaller at 500 K rather than at 2,000 K, so for colder temperature profiles the weaker water lines are optically thick at 10 bar or deeper. An increase/decrease of the mixing ratio by a factor of 10 with respect to the standard case considered will cause, as a main effect, an increase/decrease of 0.03–0.04% in the total absorption due to water. As a secondary effect, the absorption gradient between 3.6 and 5.8 m gets steeper for lower water mixing ratios, but this trend is marginal compared to the role played by temperature. CO, if present in sufficient abundance to be detected², would show its distinctive signature in the 4.5–4.9 m spectral range (see ref. 11 for details). This is a spectral region than can be observed with IRAC (channel centred at 4.5 m).