Uncovering turbulent plasma dynamics via deep learning from partial observations.

Area of Science:

• Plasma Physics
• Magnetic Confinement Fusion
• Turbulence Modeling

Background:

• Edge plasma turbulence is crucial for magnetic confinement fusion reactor performance.
• Drift-reduced Braginskii two-fluid theory is a standard but limited model for boundary plasmas.
• Accurate modeling of edge turbulence is essential for both theoretical and experimental advancements.

Purpose of the Study:

• To develop a novel framework for understanding and modeling edge plasma turbulence.
• To improve the accuracy of turbulence simulations beyond conventional equilibrium models.
• To enable learning turbulent fields from partial observational data.

Main Methods:

• Implemented a physics-informed deep learning framework.
• Constrained the deep learning model using partial differential equations derived from two-fluid theory.
• Trained the model using partial observations of electron pressure.

Main Results:

• The deep learning framework accurately learned turbulent fields consistent with two-fluid theory.
• This approach successfully utilized partial electron pressure observations, overcoming limitations of conventional models.
• Demonstrated the capability to reconstruct complex turbulent dynamics from sparse data.

Conclusions:

• Physics-informed deep learning offers a powerful new paradigm for studying edge plasma turbulence.
• This technique enhances the validation of magnetized plasma turbulence theories.
• It paves the way for advanced plasma diagnostics and improved fusion reactor design.