The Magnetic Structure And Evolution Of The Filament-Sigmoid And Prominence-Cavity Systems

The Sun feeds the earth,people look at the Sun.Observing with telescopes,we find that the Sun is not everlasting and unchanged.Some transient phenomena are detected and called solar activities.Long time observation shows that the solar activities have cyclical characters,the most obvious one is the 11-year solar cycle.Solar eruptions are the most dramatic solar activities,which is a process of energy release.It will cause disastrous space weather events and affect human activities such as space explorations and ground communications,if the solar eruptions are facing the Earth.Solar flares,filaments eruptions and coronal mass ejections are considered as different phenomena of the same physical process,i.e.,Magnetic Reconnection.Although many observed phenomena can be explained by the existing models;some critical physical mechanisms are unclear and debated,such as,how does magnetic field energy accumulate before the eruptions,what are the triggers,and how does the energy release during the eruption?In order to have a better understanding about the solar eruptions,we study the structure and evolution of the recently observed filament-sigmoid and prominence-cavity systems by coronal magnetic field reconstruction and MHD simulation.In the end,taking advan-tage of the recently developed neural networks,we explore the possibility of building mappings between different observed solar images and further to improve the image quality of solar observations.In summary,the thesis contains two parts:explaining the physical processes of solar activities,and exploring new accessory methods of solar observations.The following shows the results of these two aspects.The magnetic flux rope(MFR) is a bundle magnetic field lines surrounding the central axis and the basic plasmon structure in the universe.Solar eruptions usually contain a MFR which plays a major role during energy storage and release.The hot channels are observed after the launch of SDO.Some researchers deduce that the hot channels are characteristics of MFR.Due to the lack of measurement of the coronal magnetic field,the magnetic structure of the hot channels is unclear.We investigate the formation and magnetic topology of four flare/CME events with filament-sigmoid systems,and reconstruct the coronal magnetic field matching the observed hot channels well based on the observed magnetic field on photosphere.And then we explore the relationship between the modeled MFR and the observed hot channels and filaments.Observations reveal that the sigmoidal hot channels are located above the filaments,and they appear in pairs prior to eruption.The formation of hot channels usually takes several to dozens of hours during which two J-shape sheared arcades gradually evolve into sigmoidal hot channels,then they keep stable for tens of minutes or hours and erupt.While the low-lying filaments show no significant change.We construct a series of magnetic field models with the flux rope insert method and find that the best-fit pre-flare models contain magnetic flux ropes with hyperbolic flux tubes(HFTs).The field lines above the HFT correspond to the high-lying hot channel,while those below the HFT surround the underlying filaments.In particular,the continuous and long field lines representing the flux rope located above the HFT match the observed hot channels well in three events.While for SOL2014-04-18 event,the flux bundle that mimics the observed hot channel is located above the flux rope.The flux rope axis lies in a height range of 19.8 Mm and 46 Mm above the photosphere for the four events,among which the flux rope axis in SOL2012-07-12 event has a maximum height,which probably explains why it is often considered as a double-decker structure.The formation of the hot channels is simulated by models with increasing axial flux,which suggests that the high-lying hot channel may be formed by magnetic reconnections increasing axial flux.Further,the magnetic reconnections in an MHD simulation are studied.The prominence-cavity system and the filament-sigmoid system correspond to the same magnetic structure:MFR.The difference is that prominence-cavity systems are usu-ally located in quiet regions seen off the solar limb,while filament-sigmoid systems often lie in active regions observed on the solar disk.This prominence-cavity system includes a hot cavity surrounding a prominence with prominence horns and a central hot core above the prominence.The evolution of the system from quasi-equilibrium to eruption can be divided into four phases:quasi-static phase,slow rise phase,fast rise phase,and propagation phase.The emerged MFR stays quasi-static during which mag-netic reconnection occurs at the overlying high-Q(squashing factor)apex region,which gradually evolves into a hyperbolic flux tube(HFT).Then the system enters a slow rise phase,during which the overlying magnetic reconnection between the MFR and the overlying arcades at the apex HFT removes the confinement of the overlying magnetic field,thus engines the slow rise of the flux rope.Once the MFR reaches the regime of torus instability,the fast rise of the MFR begins.Another HFT quickly forms at the dip region under the MFR,followed by the explosive flare reconnection.The system enters the propagation phase with the acceleration decreasing to about 0 km/s2,once its apex reaches the height of about one solar radius above the photosphere.The simulation reproduces the main processes of one group of prominence eruptions,especially those occurring on the quiet Sun.Finally,we explore the mappings between input(A)and target(B)datasets built by the conditional Generative Adversarial Networks(cGANs)and show the upper and lower limits of the mappings.The mappings with the input and target datasets being the same are defined as AA-mappings,others are defined as AB-mappings which include random mappings trained by the input of datasets with random-pixel images.The paper starts with AB-mappings from solar images taken in 6563 A by the Global Oscillation Network Group(GONG)on the ground to those observed in eight different wavelengths by SDO/AIA in space.The fake AIA images generated by the eight models match the corresponding real AIA images well in large scale structures such as active regions and prominences.But the small scale coronal loops and filament threads are difficult to reconstruct.Further,the AA-mappings and random mappings give the upper and lower limits of mapping results measured by 4 evaluation functions.We find that through the AA-mappings the enhancement of the signal-to-noise ratio(SNR)is greater than that of 3-frame-stacked real observations.The denoising effect still needs to be tested by experiments from other disciplines and fields.Based on this exploration,we carry out two experiments which are obtaining space-based SDO/MHI Solar magnetograms from ground-based Ha observations by deep learning and denoising SDO/MHI Solar magnetograms by self2self.And positive results have been achieved.In summary,the magnetic flux rope is the core structure of the filament-sigmoid and prominence-cavity systems.We reconstruct the magnetic field of the observed hot channels based on the observed photosphere magnetic field for the first time.Magnetic flux ropes with hyperbolic flux tubes are constructed and consistent with the observed filament-sigmoid systems.The field lines above the HFT correspond to the high-lying hot channel,while those below the HFT surround the underlying filaments.We find that the hot channels are formed by the increment of the axial flux driven by the magnetic reconnections of sheared arcades above the filaments.We study the structure and evo-lution of the MFR in an MHD simulation of the prominence-cavity system.Four phases of filament eruptions are reproduced.We find that the magnetic reconnections between the MFR and the overlying magnetic field and flare reconnections below the MFR play major roles at the slow rise and fast rise phases,respectively.The above-mentioned works give us a better understanding about the magnetic structure and eruption mech-anism of the solar eruptions.In the end,we point out the limitations as well as the lower and upper limits of mappings between different solar images built by the condi-tional Generative Adversarial Networks.Further we explore noise reduction by neural network deep learning.