Convergence analysis of a discontinuous finite element formulation based on second-order derivatives (Q2384245)

The discretization of the elliptic equation \(-\Delta u+u=f\) is based on piecewise integrals of the form \[ \int (-\Delta u+u)(-\Delta v+v)\,dx \] in the spirit of the discontinuous Galerkin method. For piecewise polynomial approximations of degree \(p \geq 2\), optimal uniform \(h\) and \(p\) convergence rates are obtained in the broken \(H^{1}\) and \(H^{2}\) norms. Convergence in \(L^{2}\) is optimal for \(p \geq 3\), if the computational mesh is strictly rectangular. If the mesh consists of skewed elements, then optimal convergence is only obtained if the corner angles satisfy a given regularity condition. For \(p = 2\), only suboptimal \(h\) convergence rates in \(L^{2}\) are obtained and for linear polynomial approximations the method does not converge.