Deep Learning based approach for investigating Electromagnetic Wave Propagation in Plasmas

2D simulations of EM wave-plasma interaction leading to complex scattering patterns are computationally expensive and challenging. The scattering pattern is primarily determined by the plasma density profile and the incident wave frequency. We propose a novel U-Net based deep learning (DL) approach, with a series of encoder and decoder units

with skip connections, for studying microwave-plasma interaction. The scattering of a plane EM wave with fixed frequency (1 GHz ) and amplitude incident on a plasma medium with different Gaussian density profiles (leading to absorption, reflection and transmission) have been considered in this work. Training data has been generated using conventional FDTD based method. We evaluate the results obtained from the network, using various metrics like SSIM index, average percent error and mean square error, with the data obtained from well-established FDTD based technique. The trained DL model is able to reproduce the scattering pattern of the EM wave from different plasma profiles with an error margin of less than 2%. Ablation studies along with network visualization using feature maps has also been performed. The DL technique proposed in this work is significantly fast as compared to the existing computational techniques and can be used as an alternative computational approach for investigating EM-plasma interaction.