First sunspot image taken by Daniel K. Inouye Solar Telescope. (Credit: NSO/AURA/NSF)
A collaboration between researchers in ICS, the UH Institute for Astronomy, and the National Solar Observatory has published another article using artificial intelligence (AI) to reveal the structure of the Sun’s atmosphere. The team has been training large neural networks on magnetohydrodynamic simulations in order to predict 3D magnetic and velocity fields of the Sun’s photosphere from spectropolarimetric observations taken by the Daniel K. Inouye Solar Telescope on Haleakalā.
Their latest paper, published in the Astrophysical Journal on Dec. 11 2025, uses physics-informed neural networks to resolve details of the Sun’s atmosphere in a physically-consistent manner. Peter Sadowski, an ICS Associate Professor involved in the project, says this work highlights the growing importance of AI in the physical sciences. “Combining AI with traditional computational science techniques is very powerful because physics simulations can provide large training datasets for neural network models.” ICS postdoc Yannik Glaser and PhD student Linnea Wolniewicz were also involved. The project is supported by the NSF and the National AI Research Resource Pilot Program.
The work was also highlighted in a UH news article this month: