Dissipative issue of high-order shock capturing schemes with non-convex equations of state (Q1000945)

It is well known that, closed with a non-convex equation of state (EOS), the Riemann problem for the Euler equations allows for non-standard waves, such as split shocks, sonic isentropic compressions or rarefaction shocks. Loss of convexity leads in this case to non-uniqueness of entropic or Lax solutions, which can only be resolved via the Liu-Oleinik criterion (equivalent to the existence of viscous profiles for all admissible shock waves). This suggests that in order to capture the physical solution, a numerical scheme must provide an appropriate level of dissipation. A legitimate question then concerns the ability of high-order shock capturing schemes to naturally select such a solution. To investigate this question and evaluate modern as well as future high-order numerical schemes, there is therefore a crucial need for well-documented benchmarks. Here the authors propose a thermodynamically consistent non-convex EOS that can be easily introduced in Eulerian as well as in Lagrangian hydrocodes for test purposes, along with a reference solution for an initial value problem exhibiting a complex composite wave pattern (the Bizarrium test problem). Two standard Lagrangian numerical approaches, both based on a finite volume method, are then reviewed (vNR and Godunov-type schemes) and evaluated on this Riemann problem. In particular, the authors give a complete description of several state-of-the-art high-order Godunov-type schemes applicable to general EOSs. They show that this particular test problem remains quite severe when working with high-order schemes, and recommend it as a benchmark for devising new limiters and/or next-generation highly accurate schemes.