Outflow Positivity Limiting for Hyperbolic Conservation Laws. Part I: Framework and Recipe

TL;DR

This paper introduces a generalized outflow positivity limiting framework for hyperbolic conservation laws that ensures positivity constraints are maintained efficiently, high-order accuracy is preserved, and numerical stability is achieved.

Contribution

It extends Zhang and Shu's positivity limiter to a broader framework that enforces positivity at boundary nodes and uses outflow limiting, improving efficiency and robustness.

Findings

Guarantees positivity-preserving time step similar to pointwise enforcement.

Enforces positivity at boundary nodes and uses wave speed capping.

Maintains high-order accuracy and conservation in numerical schemes.

Abstract

Numerical methods for hyperbolic conservation laws are needed that efficiently mimic the constraints satisfied by exact solutions, including material conservation and positivity, while also maintaining high-order accuracy and numerical stability. Discontinuous Galerkin (DG) and WENO schemes allow efficient high-order accuracy while maintaining conservation. Positivity limiters developed by Zhang and Shu ensure a minimum time step for which positivity of cell average quantities is maintained without sacrificing conservation or formal accuracy; this is achieved by linearly damping the deviation from the cell average just enough to enforce a cell positivity condition that requires positivity at boundary nodes and strategically chosen interior points. We assume that the set of positive states is convex; it follows that positivity is equivalent to scalar positivity of a collection of affine functionals. Based on this observation, we generalize the method of Zhang and Shu to a framework that we call outflow positivity limiting: First, enforce positivity at boundary nodes. If wave speed desingularization is needed, cap wave speeds at physically justified maxima by using remapped states to calculate fluxes. Second, apply linear damping again to cap the boundary average of all positivity functionals at the maximum possible (relative to the cell average) for a scalar-valued representation positive in each mesh cell. This be done by enforcing positivity of the retentional, an affine combination of the cell average and the boundary average, in the same way that Zhang and Shu would enforce positivity at a single point (and with similar computational expense). Third, limit the time step so that cell outflow is less than the initial cell content. This framework guarantees essentially the same positivity-preserving time step as is guaranteed if positivity is enforced at every point in the mesh cell.