On the numerical overshoots of shock-capturing schemes

TL;DR

This paper introduces a simple metric to benchmark shock-capturing schemes focusing on overshoot errors, revealing non-monotonous behavior with CFL number and dependence on wave wavenumber, thus aiding robustness assessment.

Contribution

It proposes a new metric for quantifying shock-capturing overshoots and analyzes their behavior across different schemes, highlighting previously unknown non-monotonous and wavenumber-dependent effects.

Findings

Overshoot amplitude varies non-monotonously with CFL number

Overshoot amplitude depends significantly on reduced wavenumber

New insights into shock-capturing scheme robustness

Abstract

This note introduces a simple metric for benchmarking shock-capturing schemes. This metric is especially focused on the shock-capturing overshoots, which may undermine the robustness of numerical simulations, as well as the reliability of numerical results. The idea is to numerically solve the model linear advection equation with an initial condition of a square wave characterized with different wavenumbers. After one step of temporal evolution, the exact numerical overshoot error can be easily determined and shown as a function of the CFL number and the reduced wavenumber. With the overshoot error quantified by the present metric, a number of representative shock-capturing schemes are analyzed accordingly, and several findings including the amplitude of overshoots non-monotonously varying with the CFL number, and the amplitude of overshoots significantly depending on the reduced wavenumber of the square waves (discontinuities), are newly discovered, which are not before.