Austrian Early Career Conference 2024

Contribution:
Poster

Authors:
M. P. K. Buchner [1]; R. Jarolim [1]; A. M. Veronig [1,2]

Affiliations:
1: University of Graz, Institute of Physics, Graz, Austria; 2: University of Graz, Kanzelhöhe Observatory for Solar and Environmental Research, Treffen am Ossiacher See, Austria

Title:
Advancing solar coronal magnetic field modeling through a non-force-free approach in physics-informed neural networks

Abstract:
While the solar magnetic field in the photosphere is commonly observed, its behavior in the upper atmospheric layers is difficult to assess observationally. However, observations support that magnetic field changes in the uppermost layer (the solar corona) are the primary source for solar eruptions. Therefore, modeling methods are used to reconstruct the coronal magnetic field from photospheric magnetic field maps. Recent methods for magnetic field extrapolations build on the force-free field assumption to reduce computational costs. While this approach can provide a realistic approximation of the solar magnetic field this assumption is not valid in the lower solar atmospheric layers. In this study, we expand on the physics-informed neural network (PINN) architecture developed by Jarolim et al. (2023). The adopted framework extrapolates photospheric vector magnetic field maps from the Helioseismic Magnetic Imager (HMI) onboard NASA's Solar Dynamics Observatory (SDO). Governed by the physical equations the training process tries to find a balance between the equations and observational data. To enhance the accuracy of the existing model and allow further investigations on the complex physical processes in solar active regions, our goal is to integrate non-force-free dynamics. Our method introduces the plasma pressure as a new quantity into the extrapolation process and experiments with force-free sub models which get pre-trained to a certain stage. This strategy is expected to yield a model that creates an improved representation of the solar magnetic field, particularly in active regions, which are the primary drivers of space weather impacts. By the refined coronal field model, we aim to contribute to a deeper understanding of solar atmospheric dynamics and its potential impacts on Earth.