Uncovering edge plasma dynamics via deep learning from partial observations

One of the most intensely studied aspects of magnetic confinement fusion is edge plasma behaviour, which is critical to reactor performance and operation. Drift-reduced Braginskii two-fluid theory has for decades been widely used to model edge plasmas with varying success. This Letter demonstrates that physics-informed neural networks can accurately learn turbulent field dynamics consistent with the two-fluid theory from just partial observations of a synthetic plasma's electron density and temperature for plasma diagnosis and model validation in challenging thermonuclear environments.

磁约束聚变（magnetic confinement fusion）领域中研究最为深入的方向之一便是边缘等离子体行为（edge plasma behaviour），其对反应堆的运行与性能至关重要。数十年来，减漂移布拉金斯基双流体理论（Drift-reduced Braginskii two-fluid theory）被广泛用于模拟边缘等离子体，但其应用效果参差不齐。本研究快报证实，仅通过合成等离子体（synthetic plasma）的电子密度与温度的部分观测数据，物理信息神经网络（physics-informed neural networks）便可精准学习到与该双流体理论相符的湍流场动力学，可为严苛热核环境下的等离子体诊断（plasma diagnosis）与模型验证（model validation）提供支撑。