Comparison of Different Perfectly Matched Layer (PML) Methods for Time-Domain Elastic Wave Propagation Modelling

Summary Perfectly Matched Layers (PML), introduced by Berenger in 1994, have been widely utilized as absorbing boundary conditions in numerical simulations, particularly in elastic wave propagation modeling. These layers are designed to minimize artificial reflections by effectively absorbing outgoing waves at the computational domain boundaries. However, conventional PML methods come with certain limitations, which can cause spurious reflections related to low-frequency waves and evanescent waves at the PML interface, which may affect the simulation results' accuracy. One noteworthy characteristic of conventional PML is its adoption of a nonphysical splitting of the wave field components and equations. This results in some challenges in its implementation into existing numerical modeling codes. Alternative PML methods, such as Convolutional PML (CPML) and Auxiliary Differential Equation PML (ADEPML), have been proposed to address these limitations. These methods aim to improve the performance of absorbing boundary layers and enhance the accuracy of elastic wave propagation simulations. In this paper, we delve into the details of the different PML methods employed, including Split field PML (SPML), Convolutional PML (CPML), and Auxiliary Differential Equation PML (ADEPML) with the Complex frequency shifted(CFS) tensor and their performance compared with each other.