Discontinuous Galerkin methods for first-order hyperbolic problems

We consider discontinuous Galerkin (DG) finite element approximations of a model scalar linear hyperbolic equation. We show that in order to ensure continuous stabilization of the method it suffices to add a jump-penalty-term to the discretized equation. In particular, the method does not require upwinding in the usual sense. For a specific value of the penalty parameter we recover the classical discontinuous Galerkin method with upwind numerical flux function. More generally, using discontinuous piecewise polynomials of degree k, the familiar optimal O(h^{k+1/2}) error estimate is proved for any value of the penalty parameter. As precisely the same jump-term is used for the purposes of stabilizing DG approximations of advection-diffusion operators, the discretization proposed here can simplify the construction of discontinuous Galerkin finite element approximations of advection-diffusion problems. Moreover, the use of the jump-stabilization makes the analysis simpler and more elegant.