Coupling FEM with a multiple-subdomain Trefftz method

Second-order elliptic equations (35J15) Scattering theory for PDEs (35P25) Maxwell equations (35Q61) Numerical optimization and variational techniques (65K10) Finite element, Rayleigh-Ritz and Galerkin methods for boundary value problems involving PDEs (65N30) Applications to the sciences (65Z05) Diffraction, scattering (78A45) Finite element, Galerkin and related methods applied to problems in optics and electromagnetic theory (78M10) Multipole methods applied to problems in optics and electromagnetic theory (78M16)

Review text

The authors consider the second-order scalar elliptic problem \(-\nabla \cdot \lbrack M_{\epsilon }^{-1}(x)\nabla u]-\omega ^{2}\mu (x)u=f\) in \(\mathbb{R} ^{2}\), with the Sommerfeld radiation condition: \(\nabla u\cdot x-\imath k\left\Vert x\right\Vert u=0\) when \(\left\Vert x\right\Vert \rightarrow \infty \) uniformly. Here \(u:\mathbb{R}^{2}\rightarrow \mathbb{C}\) represents the longitudinal component of the magnetic field, \(M_{\epsilon }:\mathbb{R} ^{2}\rightarrow \mathbb{C}^{2,2}\) is an inhomogeneous and anisotropic permittivity (with nonzero determinant), \(\mu :\mathbb{R}^{2}\rightarrow \mathbb{C}\) is an inhomogeneous and isotropic permeability, \(\omega \in \mathbb{R}\) is the angular frequency, \(k=\omega \sqrt{\epsilon \mu }\) is the piecewise-constant wavenumber and \(f:\mathbb{R}^{2}\rightarrow \mathbb{R}\) represents the stationary current that generates the electromagnetic field (with compact support). The authors assume that \(\epsilon \) and \(\mu \) are piecewise constant with two subdomains \(\Omega _{0}\) (unbounded) and \(\Omega _{1}\) (bounded) and they introduce the Trefftz spaces \(\mathcal{T}(\Omega _{0}):=\{v\in H_\mathrm{loc}^{1}(\Omega _{0}):\nabla ^{2}v+k_{0}^{2}v=0\), \(v\) satisfies the radiation condition\(\}\)\ and \(\mathcal{T}(\Omega _{1}):=\{v\in H^{1}(\Omega _{1}):\nabla ^{2}v+k_{1}^{2}v=0\}\). The purpose of the paper is to propose numerical simulations for this problem based on spaces spanned by multipole expansions centered in points outside each \(\Omega _{i}\), \(i=0,1\), coupled with the finite element method to discretize the usual primal variational form of the problem. The authors first present the multipole approach which here involve Hankel's functions of the first kind or Bessel's function of the first kind, depending on the domain. Each multipole is characterized by its center and the degree of the function. The authors analyze the error considering two configurations of multipoles: multipole expansions up to a fixed order 1 uniformly located on a circle at the center of each subdomain and one multipole expansion of a given order placed in the origin. To introduce the coupling with FEM, the authors consider the partition \(\mathbb{R}^{2}=\Omega _{f}\cup \Omega _{m}^{0}\cup \Omega _{m}^{1}\cup \Gamma _{f0}\cup \Gamma _{f1}\cup \Gamma _{01}\), where \(\Gamma _{f0}=\partial \Omega _{f}\cap \partial \Omega _{m}^{0}\), \(\Gamma _{f1}=\partial \Omega _{f}\cap \partial \Omega _{m}^{1}\), \(\Gamma _{01}=\partial \Omega _{m}^{0}\cap \partial \Omega _{m}^{1}\) with \(\Omega _{f}\cap \Omega _{m}^{0}=\emptyset \), \(\Omega _{f}\cap \Omega _{m}^{1}=\emptyset \), \(\Omega _{m}^{0}\cap \Omega _{m}^{1}=\emptyset \). They define \(\Omega _{m}=\Omega _{m}^{0}\cup \Omega _{m}^{1}\) and the skeleton \( \Gamma =\Gamma _{f0}\cup \Gamma _{f1}\cup \Gamma _{01}\) and they consider the magnetic (\textquotedblleft Neumann\textquotedblright ) trace operator \( \gamma :H_\mathrm{loc}^{1}(\nabla ^{2},\Omega _{\square })\rightarrow \widetilde{H} ^{-1/2}(\Gamma _{\square })\) with \(\Omega _{\square }\in \{\Omega _{f},\Omega _{m}^{0},\Omega _{m}^{1}\}\), \(\Gamma _{\square }\in \{\Gamma _{f0},\Gamma _{f1}\cup \Gamma _{01}\}\), \(H_\mathrm{loc}^{1}(\nabla ^{2},\Omega _{\square })=\{v\in H_\mathrm{loc}^{1}(\Omega ):\nabla ^{2}v\in L_\mathrm{loc}^{2}(\Omega )\}\) and \(\widetilde{H}^{-1/2}(\Gamma _{\square })\) is the dual space of \( H^{1/2}(\Gamma _{\square })\). They introduce a variational formulation of the problem and they reformulate this problem as a saddle point problem for a Lagrangian functional. They finally consider a discontinuous Galerkin method. The final part of the paper presents the results of numerical simulations obtained using piecewise-linear Lagrangian finite elements. They start with the example of a disk for which they perform a detailed error analysis. They end considering the case of a disk with different values of the parameters on each half disk.

Recommendations

Cites work

Cited in

(4)

Describes a project that uses

Uses Software

This page was built for publication: Coupling FEM with a multiple-subdomain Trefftz method

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2177934)