Solving Maxwell equations in a closed cavity, and the question of 'spurious modes'

It is pointed out that for a closed cavity should not be solved by the 'vectorial' analog of the classical finite-element method (that is, with node-based finite elements). It is suggested that there are good reasons to use edge-elements instead. Such elements (whose degrees of freedom correspond to the edges of the mesh, not its nodes) not only guarantee the right kind of continuity for fields like h and e, but also prevent the appearance of unwanted, unphysical divergent solutions in the problem of eigenmodes, the so-called spurious modes.<
>