A Hermite WENO reconstruction for fourth order temporal accurate schemes based on the GRP solver for hyperbolic conservation laws

Zhifang Du,Jiequan Li

DOI: https://doi.org/10.1016/j.jcp.2017.11.023

2017-12-31

Abstract:This paper develops a new fifth order accurate Hermite WENO (HWENO) reconstruction method for hyperbolic conservation schemes in the framework of the two-stage fourth order accurate temporal discretization in [{\em J. Li and Z. Du, A two-stage fourth order time-accurate discretization {L}ax--{W}endroff type flow solvers, {I}. {H}yperbolic conservation laws, SIAM, J. Sci. Comput., 38 (2016), pp.~A3046--A3069}]. Instead of computing the first moment of the solution additionally in the conventional HWENO or DG approach, we can directly take the {\em interface values}, which are already available in the numerical flux construction using the generalized Riemann problem (GRP) solver, to approximate the first moment. The resulting scheme is fourth order temporal accurate by only invoking the HWENO reconstruction twice so that it becomes more compact. Numerical experiments show that such compactness makes significant impact on the resolution of nonlinear waves.

Numerical Analysis

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to develop a new fifth - order - accurate Hermite - weighted essentially non - oscillatory (HWENO) reconstruction method for the numerical simulation of hyperbolic conservation laws under the framework of a two - stage fourth - order time - discretized Lax - Wendroff - type flow solver. Specifically, the paper aims to approximate the first - order moment (i.e., the gradient of the solution) by using the existing interface values of the generalized Riemann problem (GRP) solver, thereby avoiding the need for additional calculation of the first - order moment in traditional HWENO or DG methods. This makes the final scheme fourth - order accurate in time and more compact since only two HWENO reconstructions need to be invoked.

### Main problem description

1. **Improving time accuracy**: Existing high - order numerical schemes usually rely on multi - stage time iterations or complex differential equation processing, which not only increases the computational complexity but may also lead to an increase in stencil size. The method proposed in this paper aims to reduce the amount of computation while ensuring accuracy by simplifying these steps.

2. **Optimizing the reconstruction process**: Traditional HWENO and DG methods need to independently calculate the first - order moment (gradient) of the solution. This paper proposes a new method that directly uses the interface values provided by the GRP solver to approximate the gradient, thereby reducing the computational steps and improving the compactness of the scheme.

3. **Enhancing nonlinear wave resolution**: Through a more compact reconstruction method, this scheme can significantly improve the resolution of nonlinear waves, which is particularly important for applications such as shock - capturing in hyperbolic conservation laws.

### Key points of the solution

- **Using interface values**: The interface values obtained through the GRP solver can be directly used to approximate the gradient without additional calculation.
- **Reducing reconstruction steps**: Compared with the standard RK - WENO method, this method requires only half of the reconstruction steps.
- **Maintaining high - order accuracy**: Although the computational process is simplified, this method can still maintain fourth - order time accuracy and fifth - order spatial accuracy.

### Formula summary

The main formulas involved in the paper include:

- **Time - discretization formula**:
\[
\bar{u}^{n + 1/2}_j=\bar{u}^n_j-\frac{k}{2h}\left[f^*_{j + 1/2}-f^*_{j - 1/2}\right]
\]
\[
f^*_{j + 1/2}=f(u^n_{j + 1/2})+\frac{k}{4}\frac{\partial f}{\partial u}(u^n_{j + 1/2})\left(\frac{\partial u}{\partial t}\right)^n_{j + 1/2}
\]

- **Gradient approximation formula**:
\[
\Delta u^{n + 1/2}_j=\frac{1}{h}\left(\hat{u}^{n + 1/2}_{j + 1/2}-\hat{u}^{n + 1/2}_{j - 1/2}\right)
\]

These formulas ensure the efficiency and accuracy of the method while reducing the computational complexity.