An Improved Jacobi-Davidson Method for the Computation of Selected Eigenmodes in Waveguide Cross Sections

Numerical simulation of open structures requires the truncation of the computational domain and the definition of appropriate boundary conditions. However, any artificial boundary at finite distances causes additional undesired modes in the eigenanalysis of waveguides which must be excluded from the computed spectrum. We propose an extension of the Jacobi-Davidson method by introducing a special weighting function that enables a reliable suppression of undesired eigenmodes already during the solution process. Moreover, the number of iteration steps as well as the computation time can be drastically reduced by this process. We show the improvement of efficiency by means of an eigenmode computation in a photonic crystal fiber discretized by the finite integration technique.