[Paper Review] Splitting and composition methods in the numerical integration of differential equations

Sergio Blanes, Fernando Casas+1 more|ArXiv.org|Dec 1, 2008

Numerical methods for differential equationsMathematics82 references81 citations

TL;DR

This paper provides a comprehensive survey of splitting and composition methods for numerical integration of ordinary differential equations (ODEs), focusing on their ability to preserve geometric structures like symplecticity and conservation laws. The methods decompose complex vector fields into simpler, integrable components, enabling high-order, structure-preserving integrators that exhibit superior long-term stability and error behavior compared to standard schemes.

ABSTRACT

We provide a comprehensive survey of splitting and composition methods for the numerical integration of ordinary differential equations (ODEs). Splitting methods constitute an appropriate choice when the vector field associated with the ODE can be decomposed into several pieces and each of them is integrable. This class of integrators are explicit, simple to implement and preserve structural properties of the system. In consequence, they are specially useful in geometric numerical integration. In addition, the numerical solution obtained by splitting schemes can be seen as the exact solution to a perturbed system of ODEs possessing the same geometric properties as the original system. This backward error interpretation has direct implications for the qualitative behavior of the numerical solution as well as for the error propagation along time. Closely connected with splitting integrators are composition methods. We analyze the order conditions required by a method to achieve a given order and summarize the different families of schemes one can find in the literature. Finally, we illustrate the main features of splitting and composition methods on several numerical examples arising from applications.

Motivation & Objective

• To provide a unified and comprehensive overview of splitting and composition methods for numerical integration of ODEs.
• To clarify the theoretical foundations of these methods, particularly their backward error analysis and geometric structure preservation.
• To analyze the order conditions and construction techniques for high-order schemes using composition techniques.
• To illustrate the practical performance and advantages of these methods through numerical examples from diverse applications.
• To highlight open challenges such as stability, coefficient optimization, and extension to stochastic and PDE systems.

Proposed method

• Decompose the vector field of an ODE into m simpler components, each of which is exactly integrable.
• Construct numerical integrators by composing the exact flows of these components in a specific sequence with weighted time steps.
• Use composition techniques to achieve arbitrary order by satisfying specific order conditions derived from Lie algebraic expansions.
• Apply the method of modified equations to interpret the numerical solution as the exact solution of a perturbed system with similar geometric properties.
• Optimize free parameters in the composition by minimizing the norm of the leading error term in the modified equation.
• Analyze linear stability using the harmonic oscillator as a model problem, identifying stability thresholds for different schemes.

Experimental results

Research questions

• RQ1 How can splitting methods be systematically constructed to achieve arbitrary order while preserving geometric structures?
• RQ2 What are the necessary and sufficient order conditions for composition methods to achieve a given order of accuracy?
• RQ3 How does the backward error analysis of splitting schemes explain their long-time stability and favorable error propagation?
• RQ4 What are the limitations of high-order splitting methods, particularly regarding stability and the presence of negative or complex coefficients?
• RQ5 In what ways can splitting methods be adapted for problems with variable time steps or ill-posed dynamics?

Key findings

• Splitting methods preserve key geometric properties such as symplecticity, volume preservation, time-symmetry, and conservation of first integrals, leading to superior long-term behavior.
• High-order splitting schemes can be constructed via composition of lower-order flows, with order conditions derived from Lie algebraic expansions and commutator identities.
• The presence of negative coefficients in schemes of order higher than two is unavoidable, though complex coefficients with positive real parts can be used as an alternative.
• Linear stability analysis shows that schemes like leapfrog have a stability threshold of |hλ| ≤ 2, and high-order methods often suffer from small stability regions, limiting practical use.
• Optimizing coefficients to minimize the norm of the leading error term improves efficiency, but higher-order error terms may dominate performance, indicating a need for full asymptotic error estimation.
• Splitting methods have been successfully extended to stochastic differential equations and PDEs, preserving intrinsic conservation laws and improving long-time accuracy.