Decode the atmospheric secrets of distant worlds with BeAR - a GPU-accelerated Bayesian retrieval engine that transforms exoplanet spectra into detailed atmospheric compositions and structures.

Every photon of light from a distant exoplanet or brown dwarf carries encrypted information about its atmosphere - temperature layers, chemical composition, cloud structures, and more. But deciphering this spectral code requires sophisticated statistical detective work, comparing observations against millions of possible atmospheric models to find the best matches.

BeAR (Bern Atmospheric Retrieval) tackles this computational challenge head-on using Bayesian nested sampling powered by NVIDIA CUDA acceleration. Built on the foundation of HELIOS-r2, this open-source toolkit generates posterior probability distributions for atmospheric parameters while calculating Bayesian evidence to rank competing models. The GPU acceleration isn’t just a nice-to-have - it delivers 10-100x speedups over CPU-only approaches, making previously intractable retrievals feasible. The new Python interface (pyBeAR) democratizes access further, allowing researchers to perform retrievals or generate synthetic spectra directly from Python workflows.

Real scientific discoveries are already flowing from BeAR, including detailed brown dwarf atmospheric characterizations and exoplanet secondary eclipse analyses. As next-generation telescopes like JWST flood us with high-quality exoplanet spectra, tools like BeAR become essential infrastructure for unlocking the atmospheric diversity of worlds beyond our solar system.

⭐ Stars: 11
💻 Language: C++
🔗 Repository: NewStrangeWorlds/BeAR