Direct and iterative solution of the generalized Dirichlet-Neumann map for elliptic PDEs on square domains (Q1012512)

The authors investigate the properties of the collocation coefficient matrix associated with the generalized Dirichlet-Neumann map for Laplace's equation on a square domain. The generalized Dirichlet-Neumann map has the following form [cf. \textit{S. R. Fulton, A. S. Fokas} and \textit{C. A. Xenophontos}, J. Comput. Appl. Math. 167, No.~2, 465--483 (2004; Zbl 1050.65116) and \textit{A. G. Sifalakis, A. S. Fokas, S. R. Fulton} and \textit{Y. G. Saridakis}, J. Comput. Appl. Math. 219, No.~1, 9--34 (2008; Zbl 1152.65113)] \[ \sum_{j = 1}^{n}\left|h_{j}\right|e^{i(\beta_{j} - k m_{j})}\int_{-\pi}^{\pi} e^{-i k h_{j}s}(f^{(j)} - i g^{(j)})ds = 0,\quad k\in \mathbb{C}, \] where \(g^{(j)}\) and \(f^{(j)}\) denote the derivatives of the solution in the direction making an angle \(\beta_{j}\) with the side \(S_{j}\) and in the perpendicular direction, respectively. The authors analyse and implement direct and iterative methods for determining the solution of the generalized Dirichlet-Neumann map. They analyze and implement classical successive overrelaxation-type methods and Krylov subspace methods (stabilized bi-conjugate gradient and generalized minimal residual) for both sine and Chebyshev basis functions. Let us note that the analysis is independent of the choice of basis functions and that the authors consider the same type of boundary conditions on all sides of the square domain, as well as the case where different boundary conditions are used on each side. For the numerical experiments, the authors consider the model Laplace's equation with exact solution \(\sinh(3 x)\sin(3 y)\).