A Physics‐Informed Neural Networks Approach for Accurate Prediction of Solitary Waves in Magnetized Plasma

ABSTRACT In the present work, we propose a new paradigm for the simulation of solitary waves in plasma with the help of Physics‐Informed Neural Networks (PINNs). PINNs is a type of neural network architecture aimed at solving a complicated problem described by the physical laws. For this purpose, we investigated the nonlinear ion acoustic solitary structure formation in a two‐component magnetized plasma. The evolution of the energetic electrons and those trapped in the plasma potential well is modeled by the so‐called Cairns–Gurevich distribution. A KdV‐like equation describing the nonlinear behavior of the ion acoustic wave (IAW) was found analytically by using the standard reductive perturbation technique and the appropriate independent variables. Besides, we address how deep learning can be integrated with plasma physics by focusing on the Korteweg‐de Vries (KdV) equation that describes the propagation of ion‐acoustic solitary waves. We demonstrate that PINNs can manage solitary waves without significantly deviating from physical principles. Results demonstrate enhanced computational efficiency as compared to conventional numerical approaches.