A new numerical method for scalar eigenvalue problems in heterogeneous, dispersive, sign-changing materials
Martin Halla, Thorsten Hohage, Florian Oberender
TL;DR
This work tackles time-harmonic scalar transmission problems with sign-changing coefficients arising from dispersive materials. It introduces a finite element discretization based on a $T$-coercive, weakly coercive reformulation obtained via a global reflection operator, enabling stable discretization with standard FE spaces for both source and dispersive eigenvalue problems. The authors provide explicit constructions and bounds for the reflection operators, along with a carefully designed quadrature strategy to handle the nonlocal transformation, and validate the approach through 2D and 3D numerical experiments, including a dispersive eigenvalue problem solved by a contour integral method. The method offers a tractable, convergent alternative to optimization-based approaches, with practical relevance for simulating surface plasmons and other sign-changing coefficient phenomena in heterogeneous, dispersive media.
Abstract
We consider time-harmonic scalar transmission problems between dielectric and dispersive materials with generalized Lorentz frequency laws. For certain frequency ranges such equations involve a sign-change in their principle part. Due to the resulting loss of coercivity properties, the numerical simulation of such problems is demanding. Furthermore, the related eigenvalue problems are nonlinear and give rise to additional challenges. We present a new finite element method for both of these types of problems, which is based on a weakly coercive reformulation of the PDE. The new scheme can handle $C^{1,1}$-interfaces consisting piecewise of elementary geometries. Neglecting quadrature errors, the method allows for a straightforward convergence analysis. In our implementation we apply a simple, but nonstandard quadrature rule to achieve negligible quadrature errors. We present computational experiments in 2D and 3D for both source and eigenvalue problems which confirm the stability and convergence of the new scheme.
