A numerical study of the convergence properties of ENO schemes (Q808652)

Essentially nonoscillatory (ENO) finite difference schemes have been developed to treat hyperbolic conservation laws, see \textit{A. Harten}, \textit{S. Osher} [SIAM J. Numer. Anal. 24, 279-309 (1987; Zbl 0627.65102)] or \textit{A. Harten}, \textit{B. Engquist}, \textit{S. Osher} and \textit{S. Chakravarthy} [J. Comput. Phys. 71, 231-303 (1987; Zbl 0652.65067)]. Essentially nonoscillatory schemes, like any nonoscillatory high-order schemes for shock calculations, are nonlinear even for linear problems. Traditional stability analysis is not applicable. Convergence for higher order essentially nonoscillatory schemes is still an open theoretical problem. For some special initial conditions the surprising observation is made that the calculations do not converge at the expected rate, and that sometimes the error in fact increases as the discretization is refined. It can be concluded that the difficulties with the convergence must be due to the nonlinear nature of the scheme, which in turn results from the use of adaptive stencils. Concepts to overcome these difficulties have been given by \textit{C.-W. Shu} [ibid. 5, No.2, 127-149 (1990; Zbl 0732.65085)].