Combined Analysis of Dayside Spectrum for HD 189733 b


An outline of the paper: Optimal Estimation Retrievals of the Atmospheric Structure and Composition of HD 189733b from Secondary Eclipse Spectroscopy by Lee et al. (2011)

The transiting hot Jupiter HD 189733 b is one of the most favourable targets we have for measuring the structure and composition of an exoplanet’s atmosphere for a number of reasons. To start with, with a radius of around 1.1 Jupiter radii, it’s a large planet, so when it transits its 0.8 solar radii K1 host star, it causes an easy-to-detect 2.5% drop in the system’s flux. Similarly, when it passes behind the star during secondary eclipse, the system’s overall brightness drops by around 0.5% in the near-infrared. Furthermore, with a K-band magnitude of 5.5, the host star is bright, providing the large number of photons that are required to perform these high precision measurements.

For these reasons, HD 189733 b has been targeted by numerous high signal-to-noise, multi-wavelength measurements in an effort to unveil the structure and composition of its atmosphere. Most of these measurements have been made at optical and near-infrared wavelengths using the various instruments of the Hubble Space Telescope and the Spitzer Space Telescope. In particular, evidence for a high altitude haze has been uncovered by transmission spectroscopy, while other claims have been made for the detection of molecules such as water and methane. These latter results, however, are still a subject of debate.

Figure 1: The combined data set, consisting of multi-wavelength secondary eclipse measurements made with different instruments on board Hubble and Spitzer at wavelengths between 1.5-24 microns. The solid line shows the predicted spectrum from the best-fit model obtained by Lee et al.

Jae-Min Lee and his co-workers at the University of Oxford have recently performed an analysis of all the secondary eclipse measurements available for HD 189733 b over the wavelength range 1.5-24 microns (Figure 1). To do this, they considered a layered atmosphere model, where each layer has a single temperature and pressure, as well as particular abundances of different chemical species. Specifically, they considered the molecules H2, He, H2O, CO2, CO and CH4, as these are believed to be the dominant species present in hot Jupiter atmospheres. By using the known opacities of these molecules under the various conditions of each atmospheric layer, it’s possible to then use radiative transfer principles to calculate how light will propagate from the lower levels of the atmosphere to the surface, so that we can predict the emergent spectrum. Lee et al. use a clever technique to vary the free parameters (namely, the pressures, temperatures and molecular abundances of each layer) so that they optimise the match between the model spectrum and the measured data (Figure 1).

Figure 2: The pressure-temperature profile obtained by Lee et al for their best-fit model, with uncertainty indicated by the shaded grey area. An adiabatic troposphere is seen to extend from 1-0.1 bar with an isothermal stratosphere above that.

Figure 2 shows the pressure-temperature profile that Lee et al obtain for their best-fit model. It shows an adiabatic troposphere extending upwards between 1-0.1 bar, where the temperature decreases with height (i.e. decreasing pressure) in the atmosphere. This is characteristic of energy transfer by convection, whereas the isothermal stratosphere extending above it from 0.1-0.001 bar indicates an efficient redistribution of heat throughout those layers between the dayside and nightside of the planet, perhaps by high-speed winds. Another nice result from this study is shown in Figure 3. The lines that are drawn on these plots indicate how much the predicted spectrum changes as you vary the abundances of the different molecules as a function of height in the atmosphere. In other words, they show which molecules the different measurements are sensitive to, and furthermore, which layers of the atmosphere are probed. As can be seen, all of the data sets provide constraints on the abundance of H2O between 0.1-0.5 bar, and some constraints can also be placed on the abundance of CO2. Although the NICMOS data is found to be sensitive to CO at ~0.2 bar, Lee et al explain that they cannot meaningfully constrain its abundance due to degeneracies with other parameters in their model, such as the layer temperatures. Meanwhile, the contribution from CH4 to the measured spectrum is found to be negligible, and as such the authors conclude that its abundance cannot be estimated using the current data.

Figure 3: The sensitivity of the model's predicted output to the different molecular abundances as a function of height in the atmosphere, represented here by pressure. Columns show the results for different data sets.

There are a few caveats to bear in mind from this study. Firstly, Lee et al took the values for the measurements and uncertainties published in the literature at face-value. This could be significant, because it has recently been suggested that uncertainties quoted from previous analyses may be underestimated. If this is true, then the uncertainties on the estimated pressure-temperature profile and molecular abundances would also increase compared to those presented by Lee et al in this analysis. Secondly, Lee et al have not included the effect of clouds or hazes in their radiative transfer calculations. This was done partly due to the computational complexity that it would introduce, as well as the number of additional free parameters that it would require in a model that is already poorly constrained by the available data. Nonetheless, as the authors note, the incorporation of clouds and hazes should be attempted for future analyses, because it is known from transmission spectroscopy measurements that such features are indeed present in this fascinating atmosphere.


About Author

I'm a PhD student at the University of Oxford. My work focuses on transiting exoplanets and, in particular, what we can learn about the atmospheres of these systems. A large part of this involves getting a better handle on the various instrumental systematics that contaminate the small signals we're trying to measure, and devising methods to remove them from the data. I'm also investigating ways of correcting for the effect of star spots on planetary transmission and emission spectroscopy measurements. My supervisor is Suzanne Aigrain.