New software framework for studying exoplanet atmospheres
The Bern Atmospheric Retrieval code uses graphics processing units to analyze high and low resolution data.
Researchers developed BeAR, the Bern Atmospheric Retrieval code. This framework is the successor of Helios-r. It performs joint retrievals of exoplanet atmospheres.
The code uses a high-resolution cross-correlation framework for emission and transmission spectra. It combines log-likelihoods from high- and low-resolution data. The forward model and likelihood run entirely on a graphics processing unit (GPU).
A GPU is a graphics processing unit. A forward model is a model atmosphere. A retrieval is a process that uses data to find model parameters.
BeAR tested the hot Jupiter WASP-77Ab using Gemini South/IGRINS and JWST/NIRSpec dayside spectra. It reproduced published abundances, C/O ratios, and metallicities to within sigma. It also matched the thermal structure and line contrast to within sigma.
The joint retrieval confirmed a sub-solar metallicity and a near-solar C/O ratio. One forward model takes about ms on an NVIDIA GH200. This allows millions of models per minute on a GPU cluster.
Injection--recovery tests show that the posteriors underestimate the uncertainties of abundance ratios by about a factor of two.
Key facts
- BeAR performs joint retrievals by adding log-likelihoods from high- and low-resolution data.
- A full IGRINS retrieval with free parameters takes to minutes on one NVIDIA GeForce RTX 5080.
- The framework uses a GPU-accelerated system for emission and transmission spectra.