Giant exoplanets and brown dwarfs: exploring the atmospheric retrieval method via direct imaging spectroscopy

Abstract

The retrieval method, also known as the inverse method, has become a fundamental analysis technique for modelling and understanding exoplanetary atmospheres. In their simplest form, retrieval approaches aim to obtain the best fit solution via fitting to observed spectra using an atmospheric model defined with varying degrees of flexibility and complexity. The critical chemistry and physics driving parameters sample the parameter space, guided by Bayesian statistics, with the aim of attaining a best fit. This analysis returns estimates for an object’s mass, radius, surface gravity, temperature-pressure structure and cloud properties, as well as confirming and constraining the presence and abundance of a variety of molecular species.

TauREx3 (Tau Retrieval of Exoplanets) is a Bayesian retrieval suite developed for application to spectroscopic observations of exoplanet atmospheres. In the past, retrieval techniques, including TauREx3, have mainly been applied to transit spectroscopy. Therefore, the application of retrieval analysis to directly imaged exoplanets and brown dwarfs is still greatly unexplored and novel territory. As such, we have adapted TauREx3 for analysis of near-infrared spectrophotometry from a variety of directly imaged gas giant exoplanets and brown dwarfs, including a significant expansion of the forward model’s temperature-pressure structures and cloud capabilities. The objects analysed as part of this work span the L and T spectral and temperature regimes.

We first validate TauREx3 using high-quality data of brown dwarf GJ 570 D, robustly comparing our results to those of other retrieval studies. We then explore the atmosphere of the cool, directly-imaged exoplanet 51 Eri b. This work showed evidence for the presence of ammonia in its atmosphere, as well as the ability to fit the spectra without including cloud modelling in the retrieval.

We then conduct a thorough study of L-type, low surface gravity exoplanets, free-floating objects and brown dwarfs. Our sample included VHS 1256 b, PSO 138, HR 8799cde and Beta Pic b. We employ a variety of cloud modelling approaches, condensate species, cloud particle size distributions as well as probing the inclusion of fractional (patchy) cloud coverage. In summary, these retrievals did not display a clear preference for a particular cloud modelling approach, likely due to the data quality inhibiting the ability of the retrieval to differentiate between the cloud characteristics we probed.

Finally, our retrieval framework was then tested using simulated James Webb Space Telescope (JWST) observations of VHS 1256 b and Ross 458 c. These retrievals resulted in extremely precise, but not always accurate, parameter constraints. This work demonstrated the need for causation when using retrieval analysis as we enter the new high-quality data era of JWST.