Inferring Exoplanet Atmospheres From Spectra may be Premature

Broadband Eclipse Spectra of Exoplanets are Featureless

Authors:

Hansen et al

Abstract:

Spectral retrieval methods leverage features in emission spectra to constrain the atmospheric composition and structure of transiting exoplanets. Most of the observed emission spectra consist of broadband photometric observations at a small number of wavelengths. We compare the Bayesian Information Criterion (BIC) of blackbody fits and spectral retrieval fits for all planets with eclipse measurements in multiple thermal wavebands, typically hot Jupiters with 2-4 observations. If the published error bars are taken at face value, then eight planets are significantly better fit by a spectral model than by a blackbody. In this under-constrained regime, however, photometric uncertainties directly impact one's ability to constrain atmospheric properties. By considering the handful of planets for which eclipse measurements have been repeated and/or reanalyzed, we obtain an empirical estimate of systematic uncertainties for broadband eclipse depths obtained with the Spitzer Space Telescope: sigma_sys = 5E-4. When this systematic uncertainty is added in quadrature to published uncertainties, the Bayesian evidence for spectral features disappears: blackbodies have better BIC for all planets. Stratospheric inversions, high C/O ratios, disequilibrium chemistry, and He-dominated atmospheres have been inferred from spectral features in broadband eclipse photometry, and are therefore suspect. We conclude that statements about atmospheric composition and structure based solely on photometry are premature.