Mapping the Sun's upper photosphere with artificial neural networks
arXiv:2101.11445 · doi:10.1051/0004-6361/202140424
Abstract
We have developed an inversion procedure designed for high-resolution solar spectro-polarimeters, such as Hinode/SP or DKIST/ViSP. The procedure is based on artificial neural networks trained with profiles generated from random atmospheric stratifications for a high generalization capability. When applied to Hinode data we find a hot fine-scale network structure whose morphology changes with height. In the middle layers this network resembles what is observed in G-band filtergrams but it is not identical. Surprisingly, the temperature enhancements in the middle and upper photosphere have a reversed pattern. Hot pixels in the middle photosphere, possibly associated to small-scale magnetic elements, appear cool at the log(tau_500)=-3 and -4 level, and viceversa. Finally, we find hot arcs on the limb side of magnetic pores, which we interpret as the first direct observational evidence of the "hot wall" effect in temperature.
Submitted to Astronomy and Astrophysics. Comments are welcome
References in corpus (10)
- The NumPy array: a structure for efficient numerical computation
- An Open Source, Massively Parallel Code for Non-LTE Synthesis and Inversion of Spectral Lines and Zeeman-induced Stokes Profiles
- Inversion of the radiative transfer equation for polarized light
- Recovering Thermodynamics from Spectral Profiles observed by IRIS: A Machine and Deep Learning Approach
- Stokes Inversion based on Convolutional Neural Networks
- Machine learning initialization to accelerate Stokes profile inversions
- Inferring Vector Magnetic Fields from Stokes Profiles of GST/NIRIS Using a Convolutional Neural Network
- The Solar Oxygen Abundance from an Empirical Three-Dimensional Model
- Improved detection of farside solar active regions using deep learning
- A Non-Linear Magnetic Field Calibration Method for Filter-Based Magnetographs by Multilayer Perceptron