On the evaluation of the eigenvalues of the finite differences Laplacian over a hexagon (Q1100865)

The paper is devoted to the construction of a fast discrete method for solving the partial eigenvalue problem \(-\Delta u=\lambda u\) in a regular hexagon D under homogeneous Dirichlet boundary conditions on \(\partial D\). First the authors reduce the discrete eigenvalue problem based on a 7-point finite difference approximation to -\(\Delta\) over an equilateral triangular grid into three different subproblems connected with discrete eigenvalue problems for the Laplacian over an equilateral triangle. These subproblems are then solved by well-known iterative methods such as the inverse power method with orthogonalization or with shifting and others. The key to an efficient algorithm is now a fast direct method for solving the systems of linear equations arising at each iteration step of the inverse powers iteration for the sub-problems. The complexity of the direct method is of the order \(O(h^{-2} \log h^{-1})\) if one omits some costs before starting the iterations, where h denotes the usual discretization parameter. The fast direct method proposed in the paper can be used for solving the Helmholtz equation in an equilateral triangle and, by means of capacitance matrix techniques, in more general regions as well.