Lax-Wendroff/Cauchy-Kowalevski Approach

• Lax-Wendroff/Cauchy-Kowalevski approach is a high-order, single-step explicit method that replaces time derivatives with spatial derivatives to solve hyperbolic conservation laws.
• It integrates traditional approximate-flux and modified WENO-flux discretizations to maintain accuracy in smooth regions while controlling oscillations near shocks.
• The method simplifies symbolic differentiation, achieves robust convergence and computational efficiency, and is validated through rigorous numerical benchmarks.

The Lax-Wendroff/Cauchy-Kowalevski (LW/CK) approach constitutes a family of high-order, single-step, explicit time-integration strategies for hyperbolic conservation laws. This methodology is founded on replacing time derivatives in the Taylor expansion with spatial derivatives, exploiting the structure of the underlying partial differential equations (PDEs). Recent advances focus on combining weighted essentially non-oscillatory (WENO) spatial reconstructions with approximate or exact evaluations of flux derivatives, fluctuation control mechanisms near discontinuities, and algorithmic designs that prioritize computational simplicity and robustness (Baeza et al., 29 Jan 2025).

1. High-Order Taylor–Cauchy-Kowalevski Framework

The analytic core of LW/CK methods is the Taylor expansion in time applied to the conservation law

$u_t + f(u)_x = 0$

over a uniform grid, targeting an update of the form

$$u_i^{n+1} = \sum_{\ell=0}^R \frac{(\Delta t)^\ell}{\ell!} u_{i,n}^{(\ell)} + O(\Delta t^{R+1}),$$

where $u_{i,n}^{(\ell)} = \partial_t^{\ell} u(x_i, t_n)$. Via CK arguments,

$$\partial_t^{\ell} u = -\partial_x \left( \partial_t^{\ell - 1} f(u) \right),$$

enabling recursive substitution of time derivatives with spatial derivatives. This recursive CK process builds, at each step, spatial difference approximations $\widetilde u_{i,n}^{(\ell)}$ and $\widetilde f_{i,n}^{(\ell)}$, where the latter is typically constructed via finite differences in time over Taylor-polynomial-developed "pseudo-states" (Baeza et al., 29 Jan 2025).

Error-control is enforced such that the local truncation error is $O(h^{R+1}) + O(\Delta t^{R+1})$, and the algorithm can be cast in conservative form.

2. Approximate-Flux and WENO-Flux CK Discretizations

Two principal variants are employed in the CK step:

A. Traditional Approximate-Flux CK:

Each required spatial derivative is approximated by a symmetric centered difference: $$\widetilde u_{i,n}^{(\ell)} = -\frac{1}{h} \sum_{m=-q}^q a_m \widetilde f_{i+m,n}^{(\ell-1)}$$ where $a_m$ are coefficients of a $2q+1$ point stencil, yielding $2q$-order accuracy.

B. Modified WENO-Flux CK:

To suppress oscillations near discontinuities, WENO-type non-linear convex combinations of polynomial derivatives are used to reconstruct $f(u)_x$: $$f_x(x_i) \approx \sum_{k=0}^{r-1} \omega_k q_k'(x_i) = f_x(x_i) + O(h^{2r-2}),$$ where each polynomial $q_k$ interpolates the flux on a substencil, and the weights $\omega_k$ are adapted according to smoothness indicators. This approach achieves high-order in smooth regions and avoids spurious overshoot by downweighting "bad" substencils across discontinuities (Baeza et al., 29 Jan 2025).

3. Fluctuation Propagation and Control Near Discontinuities

Near shocks, the naive use of upwind-WENO reconstructs $u^{(1)}$ with $O(1/h)$ errors, rendering higher time derivatives $O(h^{-\ell})$, which balance against $\Delta t^\ell$ to produce $O(1)$ unphysical fluctuations in the Taylor update. These lead to spurious oscillations or excessive diffusion, degrading the local order and solution quality.

In the WENO-flux variant, substencils traversing a discontinuity have their weights $\omega_k$ suppressed ($O(h^{r-1})$), eliminating the blow-up in derivative approximations. Consequently, all recursive time derivatives remain uniformly $O(1)$, yielding an essentially non-oscillatory and stable scheme even in the presence of jumps. The paper states, as a key proposition: if $u$ has a jump at $x_j$, then in the WENO-flux CK, each $\widetilde u^{(\ell)} = O(1)$ uniformly in $h$, and no $O(h^{-\ell})$ spurious terms arise (Baeza et al., 29 Jan 2025).

4. Fluctuation-Control Mechanism

A targeted fluctuation-control strategy replaces only the first time-derivative approximation in the CK recursion by the WENO-flux derivative: $$\widetilde{\widetilde u}_{i,n}^{(1)} = -\sum_{k=0}^{r-1} \omega_k q_k'(x_i),$$ for use in the CK recursion, while still utilizing the upwind result in the final update to maintain upwinding.

This is formally implemented by polynomial reconstruction on r-point substencils, with smoothness indicators $I_k$ and non-linear weights $\omega_k$ computed using classical Jiang-Shu protocols. The construction guarantees $2r-1$ order in smooth regions and automatically degrades to the optimal contribution of a smooth substencil at a shock, thus eliminating the dangerous $O(h^{-1})$ fluctuations while preserving high-order accuracy away from discontinuities (Baeza et al., 29 Jan 2025).

5. Numerical Performance and Efficiency

Comprehensive numerical experimentation reveals the following:

• All Lax-Wendroff variants (including the approximate-flux, exact-flux, and fluctuation-controlled) reach the designed fifth-order convergence in both $L^1$ and $L^\infty$ norms in smooth 2D Euler benchmarks.
• In the double Mach reflection regime, the approximate-flux and fluctuation-controlled variants exhibit significantly less dissipation and sharper shock structures compared to traditional WENO-RK3.
• Efficiency assessments indicate that fifth-order Lax-Wendroff schemes are roughly $1.4\times$ faster than RK3, with the approximate-flux version being the least expensive among the high-order LW variants (Baeza et al., 29 Jan 2025).

Method	Efficiency (CPU_time RK3 / CPU_time)
WENO5-LW5	1.44
WENO5-LWA5	1.54
WENO5-LWAF5	1.44

6. Relation to Prior Methods and Generalization

The approach generalizes the high-order Lax-Wendroff time discretization originally advanced by Qiu and Shu, which requires explicit symbolic computation of flux derivatives via Faà di Bruno expansions. The present approximate procedure avoids symbolic blow-up: all time-derivatives of the flux are estimated via finite-difference stencils, vastly simplifying implementation while achieving the same formal order and conservative structure. The method is compatible with any WENO-type spatial discretization, and the local recursion can be extended to an arbitrary order in space and time without increase in stencil size (Baeza et al., 29 Jan 2025).

7. Summary and Significance

The approximate Lax-Wendroff/Cauchy-Kowalevski methodology as developed by Baeza, Mulet, and Zorío provides an effective, high-order shock-capturing framework for hyperbolic conservation laws. The hybrid of WENO-based flux derivative reconstruction and fluctuation control resolves the classical tension between compactness, accuracy, and oscillation suppression in the presence of strong gradients. This framework delivers (i) arbitrary order in space and time, (ii) robust performance across a range of smooth and discontinuous problems, and (iii) superior computational efficiency relative to multi-stage alternatives (Baeza et al., 29 Jan 2025).