Differentiable Stellar Atmospheres with Physics-Informed Neural Networks

Jiadong Li*, Mingjie Jian*, Yuan-Sen Ting*, Gregory M. Green · 2025

Read Differentiable Stellar Atmospheres with Physics-Informed Neural Networks by Jiadong Li*, Mingjie Jian*, Yuan-Sen Ting*, Gregory M. Green (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-06357.txt.

Description

Star atmospheres are usually computed by heavy physics codes that cannot be tuned by gradient methods. Jiadong Li and co-authors replace that bottleneck in this 2025 paper with a neural network taught physics rules. Their model, Kurucz-a1, mimics classic one-dimensional atmosphere calculations while honoring the balance of pressure and gravity inside a star.

Trained with that physical law as a constraint, the network matched and even beat the classic code on balance and on the sun's observed spectrum. Because the network is differentiable, it can plug into modern fitting pipelines. Astronomers could then tune shared physical constants across many stars at once. The authors frame it as a step toward data-driven stellar physics at survey scale.

Catalog record

CREATORJiadong Li*, Mingjie Jian*, Yuan-Sen Ting*, Gregory M. Green
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE15 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-06357

Browse the full library · About the Tower of Babel