A posteriori error estimates for numerical solutions to hyperbolic conservation laws (Q2036630): Difference between revisions

The paper at hand studies a posteriori criteria for the convergence of numerical schemes for systems of hyperbolic conservation laws in one spatial dimension. The results can be understood in the wider context of convergene of approximation schemes (not necessarily numerical methods but also, e.g., vanishing viscosity) for systems of hyperbolic conservation laws in one spatial dimension. Previous convergence results relied on compensated compactness and, thus, do not provide information on uniqueness of limits or convergence rates. The paper at hand aims at overcoming these limitations by using the $L^1$ stability theory, developed mainly by the first author, to compare approximate solutions to the entropy solution semi-group. There is, however, a fundamental obstruction to proving a priori results in this setting, since the counterexample from [\textit{A. Bressan} et al., Commun. Pure Appl. Math. 59, No. 11, 1604--1638 (2006; Zbl 1122.35074)] shows that in certain cases standard schemes such as Godunov's method are unstable in total variation. The authors consider a large class of numerical schemes satisfying the Lipschitz continuity of the approximate solution (in a suitable sense) and consistency with the weak forms of the equation and of the entropy inequality. They show that, if post-processed versions of (actually computed) numerical solutions to such schemes satisfy two criteria (that can be verified in an a posteriori sense), namely that the total variation (in space) of the approximate solution remains small at all times, outside some small strips in space-time the oscisllation of numerical solutions remains small then the $L^1$ error is small.