[Paper Review] Spatial Manifestations of Order Reduction in Runge-Kutta Methods for Initial Boundary Value Problems

Rodolfo R. Rosales, Benjamin Seibold+2 more|arXiv (Cornell University)|Dec 4, 2017

Numerical methods for differential equations46 references4 citations

TL;DR

This paper identifies and analyzes spatial boundary layers that cause order reduction in high-order Runge-Kutta methods for time-dependent initial-boundary-value problems (IBVPs). By applying singular perturbation theory and modal analysis, it shows that conventional boundary conditions induce spurious boundary layers that degrade global accuracy; it introduces weak stage order (WSO) and modified boundary conditions as remedies, restoring full order for diagonally implicit schemes.

ABSTRACT

This paper studies the spatial manifestations of order reduction that occur when time-stepping initial-boundary-value problems (IBVPs) with high-order Runge-Kutta methods. For such IBVPs, geometric structures arise that do not have an analog in ODE IVPs: boundary layers appear, induced by a mismatch between the approximation error in the interior and at the boundaries. To understand those boundary layers, an analysis of the modes of the numerical scheme is conducted, which explains under which circumstances boundary layers persist over many time steps. Based on this, two remedies to order reduction are studied: first, a new condition on the Butcher tableau, called weak stage order, that is compatible with diagonally implicit Runge-Kutta schemes; and second, the impact of modified boundary conditions on the boundary layer theory is analyzed.

Motivation & Objective

• To understand the spatial manifestation of order reduction in Runge-Kutta methods when applied to time-dependent IBVPs.
• To explain why conventional boundary conditions—while enforcing exact values at stages—induce spurious boundary layers that reduce global accuracy.
• To develop and analyze remedies, including weak stage order (WSO) and modified boundary conditions, that restore full convergence order.
• To establish that order reduction in IBVPs arises solely from time discretization, even when spatial resolution is extremely fine.
• To provide a theoretical framework based on singular perturbation and modal analysis that explains persistent boundary layer effects over many time steps.

Proposed method

• Conducts a modal analysis of the numerical scheme to decompose the spatial error into regular and boundary layer components.
• Applies singular perturbation theory to the semi-discrete system (spatially continuous, time-discretized) to model the error as a boundary layer problem.
• Derives asymptotic expansions of the numerical solution in powers of the time step $\Delta t$, identifying the leading-order error terms.
• Introduces the concept of weak stage order (WSO) as a new condition on the Butcher tableau that ensures order preservation for diagonally implicit Runge-Kutta schemes.
• Analyzes the impact of modified boundary conditions on the boundary layer structure and its persistence over time.
• Uses spectral decomposition of the operator $M = \sum_{i=1}^s \lambda_i^j \vec{r}_i \vec{\ell}_i^T$ to relate stage order conditions to error behavior.

Experimental results

Research questions

• RQ1 Why do high-order Runge-Kutta methods suffer from order reduction in IBVPs despite formal high-order accuracy?
• RQ2 What geometric structures—specifically boundary layers—arise in the spatial error due to time discretization, and why do they persist over time?
• RQ3 How does the mismatch between interior and boundary error approximations lead to a singularly perturbed problem in the global error?
• RQ4 Can weak stage order (WSO) be defined for diagonally implicit Runge-Kutta schemes to prevent order reduction?
• RQ5 To what extent can modified boundary conditions mitigate or eliminate the spatial boundary layer effects that cause order reduction?

Key findings

• Boundary layers form in the spatial error due to a mismatch between interior and boundary approximations, even when the solution is highly accurate at the boundary.
• Conventional boundary conditions that enforce exact values at each stage induce spurious boundary layers that degrade the global convergence order.
• The leading-order error term in the spatial approximation is $O(\Delta t^p)$ only if weak stage order (WSO) conditions are satisfied, otherwise it degrades to $O(\Delta t^{p-1})$.
• Weak stage order (WSO) is a new, compatible condition for diagonally implicit Runge-Kutta schemes that ensures full order of accuracy by aligning stage order with the structure of the boundary layer problem.
• Modified boundary conditions can suppress or eliminate boundary layer effects, thereby restoring the formal order of convergence.
• The analysis shows that order reduction in IBVPs is a time-discretization phenomenon independent of spatial resolution, as long as the spatial discretization converges to the semi-discrete limit.