An \(hp\)-adaptive discontinuous Galerkin method for plasmonic waveguides (Q491570)

The author presents a numerical method for the analysis of a rectangular metallic plasmonic waveguide embedded in an infinite homogeneous dielectric. This simple device is thought as a representative example for real world problems in plasmonics, a branch of nanophotonics. The mathematical model of the metallic waveguide of infinite length consisting of lossy inhomogeneous isotropic medium and surrounded by dielectric material, developed in [\textit{P. Berini}, ``Plasmon-polariton waves guided by thin lossy metal films of finite width: bound modes of symmetric structures'', Phys. Rev. B 61, No. 15, 10484--10503 (2000; \url{doi:10.1103/physrevb.61.10484})], is described with uncoupled Maxwell's equations that results in time-harmonic vectorial wave equations assuming that the dielectric permittivity is a scalar function in \(x\) and \(y\) and the permeability is equal \(1\). Considering the propagation of TE and TH waves in the direction of the waveguide yields a couple of uncoupled scalar equations, that means, eigenvalue problems have to be solved. The unbounded domain is truncated resulting in a rectangular domain with homogeneous Dirichlet boundary conditions. The author vindicates the eligibility of the truncation by results given in his earlier paper about photonic crystal applications [\textit{S. Giani}, Computing 95, No.~5, 395--414 (2013; Zbl 1311.78015)] which show that the modes decrease exponentially away from the waveguide. A discontinuous Galerkin method is chosen to get discrete eigenvalue problems with the advantage that discontinuities across elements are allowed. This property facilitates the \(hp\)-version of the discontinuous Galerkin method. The locally refined meshes in form of quadrilaterals and the variable polynomial approximation orders are based on a posteriori error estimation using an auxiliary problem which is already analyzed in the author's recent paper [\textit{S. Giani} et al., Appl. Numer. Math. 62, No. 2, 121--140 (2012; Zbl 1236.65141)]. The method is validated by numerical computations with a focus on the convergence of the adaptive algorithm demonstrated by a short program. Especially, the gain in efficiency is outlined using the \(hp\)-adaptivity rather than only an \(h\)-refinement. A problem with strong singularities in the gradient of the eigenfunctions is chosen in order to show the robustness of the algorithm. The results are interpreted as modes on the surface of the metallic waveguide. The program packages AptoFEM (finite elements), ARPACK (eigenvalue problem), and MUMPS (linear systems) are used to create the numerical results. Data about the dimension of the problems and computing times are not given.