Differentiable Modeling of Planet and Substellar Atmosphere: High-Resolution Emission, Transmission, and Reflection Spectroscopy with ExoJAX2
Hajime Kawahara, Yui Kawashima, Shotaro Tada, Hiroyuki Tako Ishikawa, Ko Hosokawa, Yui Kasagi, Takayuki Kotani, Kento Masuda, Stevanus Nuguroho, Motohide Tamura, Hibiki Yama, Daniel Kitzmann, Nicolas Minesi, Brett M. Morris

TL;DR
This paper introduces ExoJAX2, an advanced differentiable spectrum modeling framework for exoplanets and brown dwarfs, enabling high-resolution spectral analysis with improved efficiency and broader application scope.
Contribution
We developed a fast, memory-efficient, differentiable radiative transfer algorithm that handles native-resolution spectra, enhancing retrieval analysis capabilities for high-dispersion astronomical data.
Findings
Successfully retrieved the C/O ratio of GL229 B consistent with its host star.
Constrained WASP-39 b's radial velocity using molecular line structures.
Inferred Jupiter's metallicity aligning with previous estimates.
Abstract
Modeling based on differentiable programming holds great promise for astronomy, enabling advanced techniques such as gradient-based posterior sampling and optimization. This paradigm motivated us to develop ExoJAX (Kawahara et al. 2022), the first auto-differentiable spectrum model of exoplanets and brown dwarfs. ExoJAX directly calculates cross-sections as functions of temperature and pressure to minimize interpolation errors in high-dispersion spectra, although initial work focused on narrowband emission spectroscopy. Here, we introduce a fast, memory-efficient opacity algorithm and differentiable radiative transfer for emission, transmission, and reflection spectroscopy. In the era of data-rich JWST observations, retrieval analyses are often forced to bin high-resolution spectra due to computational bottlenecks. The new algorithm efficiently handles native-resolution data, preserving…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Code & Models
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
