Detection of solar internal flows with numerical simulation and machine learning

Abstract The solar interior is filled with turbulent thermal convection, which plays a key role in energy and momentum transport and generation of the magnetic field. Turbulent flows in the solar interior cannot be optically detected due to its significant optical depth. Currently, helioseismology is the only way to detect the internal dynamics of the Sun. However, long-duration data with a high cadence is required and only a temporal average can be inferred. To address these issues effectively, in this study, we develop a novel method to infer solar internal flows using a combination of radiation magnetohydrodynamic numerical simulations and machine/deep learning. With the application of our new method, we can evaluate the large-scale flow at 10 Mm depth from the solar surface with three snapshots separated by an hour. We also apply the method to observational data. Our method is highly consistent with the helioseismology, although the amount of input data is significantly reduced.