Solar flares (SF) are energetic solar events that can influence space weather and disrupt space-based technologies. Many data-driven forecasting models rely on parameterized, hand-engineered features derived from active regions (ARs). However, such parameterization compresses the spatial structure of ARs and may discard information relevant to flare initiation. In this work, we investigate the spatial distributions of several AR properties for flare forecasting using deep convolutional neural networks. To standardize input dimensions across ARs, we propose a resizing technique called Pad-crop. Our results show that models trained on different physical properties achieve comparable forecasting performance, indicating that the morphology of ARs in-its-own may encode meaningful flare signatures. Gradient-weighted Class Activation Mapping (Grad-CAM) analysis further indicates that the models focus on physically meaningful zones of the ARs, suggesting that relevant features are being captured. In addition, an analysis of false-positive predictions suggests that intermediate C-class flares during the prediction window may contribute to model confusion. These findings highlight the importance of the spatial structure in AR properties and motivate further exploration of morphology-based approaches for SF forecasting. The source code associated with this work will be made publicly available.
Three-dimensional (3D) magnetic reconnection at null points is a fundamental process in the solar corona, underlying flares, jets, and the rapid restructuring of coronal magnetic fields. We present an analytical study of the ideal current-driven kink instability in the two main 3D models of magnetic reconnection, namely the torsional spine and torsional fan magnetic field geometries. For sufficiently large azimuthal field amplitudes, the total twist exceeds the Kruskal–Shafranov threshold of 2π . The linear growth rates are derived from the ideal magnetohydrodynamic energy principle for line-tied cylindrical flux tubes, in the static limit of the equilibrium flows. We show that both the torsional spine and torsional fan become unstable to the m=1 kink mode, with Alfvénic growth rates. The instability deforms the null-point geometry strongly and creates current sheets where subsequent magnetic reconnection can change the magnetic topology substantially. These results imply that twisted null-point configurations in the solar corona can evolve rapidly into kink-unstable states, providing a natural trigger for impulsive energy release and topological restructuring in flares, jets, and coronal heating events.
Flare ribbon morphology is often used as an observational proxy for magnetic reconnection and the underlying coronal magnetic topology. However, complex ribbon shapes do not necessarily imply a specific large-scale topology, particularly when different atmospheric diagnostics sample different parts of the same event. We examine this problem using an apparently confined C2.2 flare in NOAA active region (AR) 13078 on 2022 August 16, observed with the Solar Dynamics Observatory/Atmospheric Imaging Assembly (SDO/AIA), Solar Dynamics Observatory/Helioseismic and Magnetic Imager (SDO/HMI), New Vacuum Solar Telescope (NVST) H α imaging, and a non-force-free Field (NFFF) extrapolation. The flare shows a strongly wavelength-dependent ribbon morphology. In lower-atmospheric diagnostics, especially AIA 1600 Å and NVST H α , it appears mainly as a compact two-ribbon flare rooted around a polarity inversion line (PIL). In contrast, AIA 304 Å reveals a C-shaped ribbon with an extended tail linking the compact flaring core to nearby pores and plage. The extended AIA 304 Å morphology appears to develop through the progressive brightening of connected, low-lying magnetic loops, rather than representing a separate ribbon system. The pre-flare phase shows localized ultraviolet (UV)/extreme-ultraviolet (EUV) brightenings together with loops in hot AIA channels. Hot AIA channels reveal a sigmoid-like structure associated with the same low-lying magnetic system that hosts the active-region filament and nearby arch filament system. During the flare, the H α filamentary material becomes less visible in places, most likely due to heating, enhanced background emission, and opacity changes, rather than complete disappearance or eruption. The extrapolated field contains a coronal null point at higher altitude; however, its fan and spine footpoints are not co-spatial with the observed flare ribbons or with the extended tail brightening. This suggests that the large-scale null-point topology is unlikely to be the primary cause of the C-shaped ribbon morphology. Nevertheless, the extrapolation does reproduce the low-lying magnetic connectivity associated with the active-region filament, the arch filament system, and loops linking the positive-polarity sunspot to surrounding negative-polarity pores and plage. Together, the observations and extrapolated magnetic field topology support a scenario in which the extended tail of the C-shaped ribbon results from the progressive formation or successive heating of the loops connecting the sunspot and plage, rather than from reconnection at the higher coronal null.
We present an analytical study of the time-dependent magnetic helicity creation in the separator magnetic reconnection model. The background field connects two 3D null points via a separator, and a localized flux ring perturbation with a linearly growing amplitude drives the reconnection. Using a gauge-invariant formulation, we derive a closed-form expression for the helicity creation rate. The rate scales linearly with the product of the background and perturbation field strengths, the geometric dimensions of the flux ring, and the square of its displacement from the separator axis. For typical coronal parameters, the model yields an analytical scaling law for the helicity source; a realistic reconnection diffusion region creates helicity at a negligible rate compared to the active-region helicity budget. The handedness of the created twist is determined by the sign of the perturbation amplitude, providing a clear observational signature. A comparison with a simple linear null point shows that the separator geometry is essential for a net helicity creation. These results provide the first analytical description of time-dependent helicity variation during separator reconnection and offer testable predictions for spectropolarimetric observations.
The general formula for the acoustic wave energy flux derived for sound generation in non-isothermal turbulent fluids by Routh and Musielak (2025, Fluids 10, 156) is applied, for the first time quantitatively, to the superadiabatic layer of the solar convection zone, using thermodynamic profiles from a standard solar model and an eddy-time correlation function calibrated against three-dimensional (3D) solar convection simulations. Three main results are obtained. First, an exact, closure-independent relation is derived for vertically propagating waves, F_m/F_q = 9 Ω _0^4/(ω ^2-Ω _0^2)^2 , where F_m and F_q are the monopole and quadrupole contributions to the acoustic wave energy flux, ω is the angular wave frequency, and Ω _0 is the acoustic cutoff frequency of the stratified background medium. This relation is valid for any turbulence spectrum and temporal correlation function; it shows that the temperature-gradient monopole source dominates over the classical Lighthill quadrupole only in a narrow band just above Ω _0 . Second, the dipole source term F_d produces a net negative correction to the emitted flux ( -8% to -11% ): the temperature gradient partially suppresses, rather than enhances, the emission through this term. Third, the non-isothermal corrections are strongly band-selective: the frequency-integrated total flux changes by only a few percent, but the flux in the chromospherically relevant 0 – 20 mHz band is enhanced by ≈ 15% (with the monopole alone contributing +26% ), precisely the part of the spectrum available for chromospheric heating. The classical quadrupole mechanism remains dominant throughout the source layer, the total fluxes ( 5.5 – 6.6× 10^7 erg cm−2 s−1) agree with the classical benchmarks, and the sensitivity of these results to the assumed turbulence closure is quantified, providing a framework for future calculations based on turbulence statistics extracted from numerical convection simulations.
