Skip to content
Home » Research


Solar Active Region Magnetic Field

AR11158Solar active regions host sunspots and strong magnetic fields. The excess magnetic energy powers flares and coronal mass ejections (CMEs). Solar eruption is the drivers of near-Earth space weather events.

  • We used HMI vector magnetic field maps to track the evolution of AR 11158 over 5 days (Sun et al. 2012a). A time sequence of force-free field (NLFFF) models revealed the gradual increase of magnetic energy prior to an eruptive X2 flare, and a sudden decrease post-flare.
  • We used NLFFF models to probe the magnetic topology of active regions. Our work on coronal null points explained the geometry of a highly inclined jet (Sun et al. 2012b), and a circular-ribbon flare with extended late-phase extreme ultraviolet emission (Sun et al. 2013).
  • Active regions can possess “non-neutralized” electric currents in the photosphere. We surveyed 30 regions of Cycle 24: half of them were active with flares and CMEs; half were relatively quiet (Avallone & Sun 2020). We showed that the “non-neutrality” was significantly higher for the flare-active group than the inactive one, as predicted by theory. We then used an analytical model to argue that a simple projection effect could account for the observed features (Sun & Cheung 2021). The poloidal currents, when projected vertically onto the surface, can dominate the observed signal when the flux rope is partially emerged, as in reality.

    Eruptiveness of Solar and Stellar Flares

    eruptionWhat determines whether a flare will be accompanied by a CME? On the Sun, the flare-CME association rate approaches 1 for GOES X-class flares. For cool stars, however, very few CME candidates have been reported, in contrast to the frequent detection of “super flares”.

  • AR 12192 hosted the largest sunspot group since 1990. It is the most flare-productive site of Cycle 24, but surprisingly generated no CME. Our study (Sun et al. 2015b) proposed the eruptiveness is limited by some relative measure of magnetic non-potentiality over the restriction of background field. The region serves as a solar analog for the “missing stellar CME conundrum”.
  • Using an analytical model with the observed stellar parameters , we showed that typical coronal fields on active cool stars are indeed more stable against the “torus instability” compared to the Sun (Sun, Török, & DeRosa 2021). Their larger spots, stronger dipolar fields, and a more closed magnetic topology will provide tighter confinement on the ejecta, which naturally reduces the CME occurrence.

    Deep Learning for DKIST

    DLThe 4-meter DKIST will provide revolutionary diagnostic capability of the solar atmosphere with its high cadence, high resolution, excellent polarization sensitivity, and multi-line coverage. The process to infer physical parameters from the polarized spectra (Stokes profiles) is known as “inversion”. Traditional inversion algorithms, which optimize a 1D atmospheric model iteratively, are slow and unable to keep up with the large data flow of DKIST (~20 TB per day). In contrast, deep learning methods are well suited for inverting large volumes of high-dimensional data, and can be orders of magnitude faster. Our NSF/AAG project will will develop deep learning inversion algorithms tailored for DKIST.

    We will first perform realistic MHD simulations of the solar photosphere based on published work, and forward synthesize a large library of DKIST-like Stokes profiles. Using the library as the input and the known MHD ground-truth as the target, we will then train, validate, and benchmark a set of convolutional neural networks that can rapidly perform the inversion. Finally, as DKIST data become available, we will apply adversarial domain adaption techniques to reduce the systematic differences between the simulated and real data. Our method will exploit the spatiotemporal structures of the data that have so far been ignored, and directly infer important physical quantities, such as the Poynting flux, that are difficult to estimate with traditional methods.

    Polar Fields and Solar Cycle

    AIAThe Sun undergoes 11-year activity cycles during which the global-scale field reverses polarity. The poloidal component manifests as unipolar regions at high latitudes, i.e., polar fields. Their strength is a good indicator of the magnitude of the upcoming cycle.

  • We showed (Sun et al. 2015a) that the two poles evolved out of sync during Cycle 24. The evolution was determined by the asymmetric active region emergence, and was further modulated by time-varying meridional flows.
  • Polar field is difficult to measure due to the inclined viewing angle from the ecliptic. We proposed a spatiotemporal interpolation method (Sun et al. 2011; Sun 2018) to better estimate its strength, which proved to improve space weather modeling.
  • Polar fields originate from small flux concentrations a few arcsecond across in the photosphere. Our analysis (Sun et al. 2021) revealed that the orientations of the field vector crucially depend on the data resolution. We plan to use DKIST data to study them at high resolution and to better estimate their contribution to the heliospheric magnetic flux.

    Lorentz Force and the Momentum Processes During Solar Eruptions

    AIAActive region photospheric fields evolve rapidly during major eruptions. Theoretical arguments have linked these changes (“magnetic imprints”) to the Lorentz force that drives plasma motion. This has profound implication on the less discussed momentum processes during solar eruptions.

  • We have created high-cadence vector magnetograms from HMI observations (Sun et al. 2017). The data revealed highly structured spatiotemporal patterns of the magnetic imprints.
  • “Bald patch” magnetic topology refers to the location where U-shaped field lines turn tangent to the photosphere. Simulations suggest that they may be destroyed by reconnection during an eruption. We showed (Lee et al. 2021), for the first time, that a segment of bald patch in AR 12673 indeed rapidly “disintegrated” during an eruptive X9 flare.