A square entropy stable flux limiter for $P_NP_M$ schemes

Abstract

We study some theoretical aspects of $P_NP_M$ schemes, which are a novel class of high order accurate reconstruction based discontinuous Galerkin (DG) schemes for hyperbolic conservation laws. The PNPM schemes store and evolve the discrete solution $u_h$ under the form of piecewise polynomials of degree $N$, while piecewise polynomials $w_h$ of degree $M \geq N$ are used for the computation of the volume and boundary fluxes. The piecewise polynomials $w_h$ are obtained from $u_h$ via a suitable reconstruction or recovery operator. The $P_NP_M$ approach contains high order finite volume methods ($N=0$) as well as classical DG schemes ($N=M$) as special cases of a more general framework. Furthermore, for $N \neq M$ and $N>0$, it leads to a new intermediate class of methods, which can be denoted either as Hermite finite volume or as reconstructed DG methods. We show analytically why $P_NP_M$ methods for $N \neq M$ are, in general, not $L^2$-diminishing. To this end, we extend the well-known cell entropy inequality and the following $L^2$ stability result of Jiang and Shu for DG methods (i.e. for $N=M$) to the general $P_NP_M$ case and identify which part in the reconstruction step may cause the instability. With this insight we design a flux limiter that enforces a cell square entropy inequality and thus an $L^2$ stability condition for $P_NP_M$ schemes for scalar conservation laws in one space dimension. Furthermore, in this paper we prove existence and uniqueness of the solution of the $P_NP_M$ reconstruction operator.