" Eigenvalue Computations by the Discontinous Galerkin Method "
Abstract:
We analyse the spectral correctness of
Discontinuous Galerkin (DG) methods on two different examples: the Laplace
eigenproblem, as a prototype of an operator with a compact inverse, and
the Maxwell eigenproblem, as a prototype of an operator with a non-compact
inverse.
We show that, for the first example, a huge class of DG methods enjoys
optimal spectral properties, whereas, for the second example, some
conditions should be imposed on the DG method and on the conformity
of the mesh in order to ensure spectral correctness.
Numerical tests showing the sharpeness of our theory are presented:
examples of optimal spectral approximation are discussed and we also
comment upon DG methods and meshes which generate
spourious eigenvalues.