On the convergence of a local third order shock capturing method for hyperbolic conservation laws (Q521918)

For systems of first-order hyperbolic equations, the authors start from the class of conservative schemes as given by \textit{C.-W. Shu} and \textit{S. Osher} [J. Comput. Phys. 83, No. 1, 32--78 (1989; Zbl 0674.65061)]. Considering mainly the 1D case, they consider a reconstruction procedure of the flux from an integral and two derivatives per cell given by a second degree polynomial. This method includes the use of a harmonic mean of the derivatives and is called piecewise polynomial harmonic method. It is local total variation bounded and monotone, third order accurate in smooth regions of the solution and does not introduce artificial local extremes. They also formulate an algorithm which is based on a global Lax-Friedrichs decomposition. This procedure in many dimensions is applied coordinatewise and for nonlinear systems using the local characteristic fields. For the Euler equations of gas dynamics they use Roe averages. In case of convergence of their scheme, having consistency and conservativity, they are sure that the limit is a weak solution. The main contribution of the authors is the proof of Lipschitz continuity of the numerical flux. Then, imagining a strongly stable Runge-Kutta time discretization, they get total variation stability on finite time intervals and convergence of the scheme. In their numerical tests they compare to third-order essentially non-oscillatory method on 1D Riemann problems for the Euler equations (where they also hint on a fifth-order generalization), but report on 2D problems with shocks, too. They conclude that their method gives competitive results using less CPU time due to the more localized and cheaper reconstruction process.