Efficient enforcement of far-field boundary conditions in the Transformed Field Expansions method

The Method of Transformed Field Expansions (TFE) has been demonstrated to be a robust and highly accurate numerical scheme for simulating solutions of boundary value and free boundary problems from the sciences and engineering. As a Boundary Perturbation Method it builds highly accurate solutions based upon exact solutions in a simple, canonical, geometry and corrects these via Taylor series to fit the actual geometry at hand. The TFE method has significantly enhanced stability properties when compared with other Boundary Perturbation approaches, however, this comes at the cost of requiring a full volumetric discretization as opposed the surface formulation that other methods can realize. In this paper we outline two techniques for ameliorating this shortcoming, first by employing a Legendre Spectral Element Method to implement efficient, graded meshes, and second by utilizing an Artificial Boundary with a Transparent Boundary Condition placed quite close to the boundary of the domain. In this contribution we focus on the specific problem of simulating the Dirichlet–Neumann operator associated to Laplace’s equation on a periodic cell (which arises in the water wave problem). While the details of our results are specific to this problem, the general conclusions are valid for the wider class of problems to which the TFE method can be applied. For each technique we discuss implementation details and display numerical results which support the conclusion that each of these techniques can greatly reduce the computational cost of using the TFE method.

► The TFE method is a stabilized Boundary Perturbation approach, however, it uses a volumetric (not surface) discretization. ► We discuss two methods for fixing this shortcoming. ► First, a Legendre Spectral Element Method to implement graded meshes. ► Second, we use an Artificial Boundary with a Transparent Boundary Condition placed close to the boundary of the domain. ► We discuss implementation details and display numerical results which show that each greatly reduces the computational cost.