Abstract
Abstract. The efficient determination of left eigenvectors in the method of lines (MoL) is described in this paper. The electromagnetic fields are expanded into eigenmodes and the eigenmodes are determined from an explicit matrix eigenvector problem. To study complicated structures with a moderate numerical effort, the analysis is done with a reduced set of these eigenmodes. The enforcements of the continuity of the transverse electric and magnetic fields at interfaces leads to expressions with rectangular matrices. Now left eigenvectors can be considered as inverse of these rectangular matrices. Until now, the left eigenvectors were determined from a second explicit eigenvalue problem. Here, it is shown how they can be determined with simple matrix products from previously determined right eigenvectors. This is done by utilizing the relation between the transverse electric and magnetic fields. The derived formulas hold for structures with Dirichlet, Neumann or periodic boundary conditions and the materials may be lossy. Open structures are modeled with perfectly matched layers (PML). To verify the expressions, various devices that contain such PMLs and lossy metals were studied. In all cases, error measures show that the algorithm derived in this paper works very well.
Highlights
Complicated waveguide structures can be analyzed analytically only in exceptional cases
Usually the analytic expressions exist of infinite series, which must be truncated in practical applications
Numerical results are shown where we demonstrate that the developed expressions work in case of perfectly matched layers and for lossy materials
Summary
Complicated waveguide structures can be analyzed analytically only in exceptional cases. There exist various ways in computing the eigenmodes and it is not always easy to determine all important ones with analytic expressions in case of complex structures. We use the Method of Lines (MoL) where the eigenmodes are determined after discretization with finite differences from an explicit eigenvalue problem. When a reduced set of eigenmodes is used one has to determine inverses of rectangular (not square) matrices. For this reason, (Schneider, 1999) proposed the use of a pseudo inverse (Strang, 1986). The given methods work quite well, but the numerical cost for determining the pseudo inverse or the left eigenvectors is quite high.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have