Speaker
Description
Solar coronal magnetic fields store the magnetic energy that drives solar eruptions, such as flares and coronal mass ejections, which significantly impact space weather. Nonlinear force-free fields (NLFFFs) are commonly used to model the 3D coronal magnetic fields. We develop a physics-informed neural operator (PINO) model that learns the solution operator mapping 2D photospheric vector magnetic fields to 3D NLFFFs. The model is trained using both physics losses derived from the NLFFF partial differential equations and data losses from target NLFFF solutions. We first validate our approach on an analytical NLFFF model. Subsequently, we train and evaluate the model using 2,327 numerically computed NLFFF samples from 211 active regions in the Institute for Space-Earth Environmental Research (ISEE) database. Our results show that the trained PINO model can reconstruct NLFFFs in under one second on a single consumer-grade GPU, enabling real-time reconstruction of 3D coronal magnetic fields. For 30 selected active regions, the AI-generated NLFFFs exhibit qualitative and quantitative similarity to the target NLFFFs. Additionally, the magnetic energy evolution of the AI-generated NLFFFs for active region AR 11158 appears similar to both the target NLFFFs and those obtained from existing methods. Our model could be integrated into physics-based space weather forecasting frameworks, such as the flare forecasting method proposed by Kusano et al. (2020).
| Do you plan to attend in-person or online? | In-person |
|---|
