The AI model could properly identify turbulent horizontal motion under the surface using only temperature and vertical motion data collected from the solar photosphere’s surface. This may aid our understanding of solar convection and the processes that cause explosions and jets to emerge from the Sun.
“Using the spatial distributions of temperature and vertical velocity, we created a unique convolutional neural network to estimate the spatial distribution of horizontal velocity,” stated a team of researchers led by astronomer Ryohtaroh Ishikawa of the National Astronomical Observatory of Japan.
“This allowed for the accurate detection of both spatially distributed and concentrated characteristics. [..] When compared to those described in prior studies, our network performed better on practically all spatial scales.”
The solar photosphere is the portion of the Sun’s atmosphere that we call its surface. It’s the solar atmosphere’s lowest layer, and it’s where solar activity like sunspots, solar flares, and coronal mass ejections begin.
The photosphere’s surface is not homogeneous if you look closely. It’s coated with densely packed portions that are lighter in the middle and darker near the edges. Granules are the tops of convection cells in the solar plasma, and they’re termed granules. As it flows outwards and cools, hot plasma rises in the centre and then falls back down at the edges.
We can estimate their temperature and velocity via the Doppler effect when we observe these cells, but horizontal motion cannot be measured immediately. Smaller scale flows in these cells, on the other hand, can interact with solar magnetic fields and cause other solar phenomena to occur. Turbulence is also expected to contribute to the heating of the solar corona, thus scientists are interested in learning more about how plasma behaves in the photosphere.
Ishikawa and his colleagues created numerical simulations of plasma turbulence and trained their neural networks using three separate types of simulation data. They discovered that the AI could accurately depict horizontal flows in simulations that would be undetected on the real Sun based just on temperature and vertical flow data.
This means we can feed it solar data and expect it to return answers that are compatible with what is happening on our intriguing, dangerous star.
The neural network, on the other hand, requires some fine-tuning. While the AI was able to detect large-scale flows, it struggled to pick out tiny details. Because the accuracy of small-scale turbulence is critical for particular calculations, the researchers believe that fixing this issue should be the next step in their software development.
By comparing the results of the three convection models, we observed that the rapid decrease in coherence spectrum occurred on the scales that were lower than the energy injection scales, which were characterised by the peaks of the power spectra of the vertical velocities. “This implies that the network was not appropriately trained to reproduce the velocity fields on small scales generated by turbulent cascades,” they wrote in their paper.
“Future research could look into these issues.”
Closer to home, the researchers are refining their programme to better comprehend turbulence in fusion plasmas, which is another important future application.
The findings were reported in the journal Astronomy & Astrophysics.