NEXONN: Advancing the NEXOTRANS Exoplanet Atmospheric Retrieval Framework with Transformer-Based Neural Posterior Estimation for JWST Transmission Spectroscopy of WASP-39b
Year: 2026
Type: preprint
Abstract: The James Webb Space Telescope is delivering unprecedented constraints on exoplanet atmospheres, but computationally intensive Bayesian retrievals challenge the transition from individual planets to population-level studies of planetary formation and evolution. We introduce NEXONN, a transformer-conditioned normalizing-flow framework within NEXOTRANS that enables rapid atmospheric posterior estimation with uncertainty quantification. Incorporating PandExo-based JWST noise, we benchmark NEXONN against PyMultiNest using simulated spectra and JWST observations of WASP-39 b (0.6-12 microns). For simulated MIRI/LRS data, eight posteriors achieve a mean overlap of 80.32%. Retrievals of combined NIRISS, NIRSpec, and MIRI observations yield consistent atmospheric properties. For the full dataset, NEXONN infers C/O = 0.82 (+0.01/-0.02) and [M/H] = 1.13 (+0.04/-0.04), compared with PyMultiNest values of 0.80 (+0.03/-0.01) and 1.19 (+0.05/-0.03). Divergence metrics, Wasserstein distances, posterior overlaps, and coverage tests assess posterior fidelity and uncertainty calibration, revealing parameter-dependent differences. Once trained, NEXONN performs posterior inference within seconds, enabling scalable population studies of exoplanet atmospheric composition, formation, and evolution.
Source: arXiv (Cornell University)
Language: English
Subfield: Astronomy and Astrophysics
Field: Physics and Astronomy
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7236731578
