Optimal Localization of Eigenfunctions in an Inhomogeneous Medium
band gap phenomenonDirichlet eigenvalue problemeigenfunctionelectromagnetic wavesfinite difference schememinimization methodmode localizationmultiple eigenvaluenumerical experimentsoptical devicesoptimal designspectral problemsteepest descent method
Laplace operator, Helmholtz equation (reduced wave equation), Poisson equation (35J05) Variational methods for second-order elliptic equations (35J20) General topics in linear spectral theory for PDEs (35P05) Estimates of eigenvalues in context of PDEs (35P15) Numerical methods based on necessary conditions (49M05) Numerical optimization and variational techniques (65K10) Physical optics (78A10) Optimization problems in optics and electromagnetic theory (78M50) Methods of reduced gradient type (90C52)
The authors study a problem arising in the design of optical devices that exploit the photonic band gap phenomenon. This phenomenon refers to the existence of a certain frequency band in which waves having frequency in that band cannot propagate in the medium. A band gap can be created in a medium with periodic structure. If a band gap exists, it is possible to create a standing wave with frequency in the gap by introducing pattered defects. To model that defect, a localized perturbation of the real dielectric property is introduced in the equation that governs the transverse electric-mode for electromagnetic waves in two dimensions. The authors address the problem of finding the perturbation so that the defect mode produced has desired properties. To simplify, they pose the problem on a bounded domain and deal with the Dirichlet eigenvalue problem \(\triangle u+\lambda\varepsilon u=0\). The admissible class of coefficient \(\varepsilon\) is the subset \(A\) of \(L^\infty\) functions upper and lower bounded away from zero. The goal is to find \(\varepsilon\) such that the eigenfunction \(u(\varepsilon)\) is the ``most localized one, in the sense that a weighted \(L^2\)-norm of \(\sqrt{\varepsilon}u\), named \(J(\varepsilon,u)\), reaches its minimum. Since the problem of minimizing \(J(\varepsilon, u(\varepsilon))\) over \(A\) is not well posed, authors propose two constrained problems. The ``global one lies in considering only eigenfunctions associated to the first \(N\) eigenvalues. A difficulty arises when the solution occurs at a multiple eigenvalue. The ``local one lies in restricting the admissible \(\varepsilon\) to a subset in a way that multiple eigenvalues are excluded. The disadvantage is that due to the nonconvexity of this subset, any derivative-based minimization method would be forced to search locally near the value of \(\varepsilon\) taken as basis to define this subset. In the last part of the article, the authors introduce a computational method which can be adapted for solving both problems and that can be viewed as a steepest descent approach with trajectory continuation. To illustrate some numerical experiments about the algorithm, eigenvalue problem is discretized by a simple 5-point finite difference scheme on a uniform square grid, using Matlab.
- A numerical approach for defect modes localization in an inhomogeneous medium
- Optimization of quasi-normal eigenvalues for 1-D wave equations in inhomogeneous media; description of optimal structures
- Numerical studies of the optimization of the first eigenvalue of the heat diffusion in inhomogeneous media
- Maximal and minimal norm of Laplacian eigenfunctions in a given subdomain
- Maximizing Band Gaps in Two-Dimensional Photonic Crystals
- Reduced basis approximation and a posteriori error estimates for parametrized elliptic eigenvalue problems
- Optimization of scattering resonances
- Optimization of quasi-normal eigenvalues for 1-D wave equations in inhomogeneous media; description of optimal structures
- A numerical approach for defect modes localization in an inhomogeneous medium
- A general framework for localization of classical waves. I: Inhomogeneous media and defect eigenmodes
- Optimization of spectral functions of Dirichlet-Laplacian eigenvalues
- Data-driven reduced order modeling for parametric PDE eigenvalue problems using Gaussian process regression
- Review of computational approaches to optimization problems in inhomogeneous rods and plates
- Maximization of the quality factor of an optical resonator
This page was built for publication: Optimal Localization of Eigenfunctions in an Inhomogeneous Medium
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4830559)
