[Paper Review] A class of generalized additive Runge-Kutta methods

Adrian Sandu, Michael Günther|arXiv (Cornell University)|Oct 21, 2013

Numerical methods for differential equations11 references3 citations

TL;DR

This paper introduces Generalized Additive Runge-Kutta (GARK) methods, a novel class of numerical integrators that extend traditional additive Runge-Kutta schemes by allowing different stage values to be used as arguments for different components of the right-hand side. The framework enables higher-order accuracy, improved stability, and enhanced flexibility for solving systems with multiple scales or physical processes, particularly through implicit-explicit and implicit-implicit formulations with decoupled stability properties.

ABSTRACT

This work generalizes the additively partitioned Runge-Kutta methods by allowing for different stage values as arguments of different components of the right hand side. An order conditions theory is developed for the new family of generalized additive methods, and stability and monotonicity investigations are carried out. The paper discusses the construction and properties of implicit-explicit and implicit-implicit,methods in the new framework. The new family, named GARK, introduces additional flexibility when compared to traditional partitioned Runge-Kutta methods, and therefore offers additional opportunities for the development of flexible solvers for systems with multiple scales, or driven by multiple physical processes.

Motivation & Objective

• To develop a generalized framework for additive Runge-Kutta methods that allows different stage values as inputs to different components of the right-hand side.
• To extend order conditions theory, stability, and monotonicity analyses to this new class of methods.
• To enable the construction of implicit-explicit and implicit-implicit GARK schemes with improved stability and flexibility for multi-scale or multi-physics problems.
• To provide a foundation for designing high-order, stable, and efficient solvers for systems with multiple stiff and non-stiff components.

Proposed method

• The GARK method generalizes traditional ARK by introducing separate stage values $ Y_i^{\{q\}} $ for each component $ f^{\{q\}} $, allowing different arguments in the right-hand side evaluations.
• Order conditions are derived using NB-series theory, extending the classical order conditions framework to the generalized setting.
• Stability is analyzed via algebraic stability and absolute monotonicity, with new definitions adapted for the GARK structure.
• The method supports both implicit-explicit (IMEX) and fully implicit (implicit-implicit) formulations, with coupling coefficients $ A^{\{q,m\}} $ enabling flexible stage value dependencies.
• A generalized Butcher tableau is introduced to represent the method, with separate coefficient matrices for each component interaction.
• The framework allows for constructing methods with decoupled stability properties, such as $ P^{\{i,j\}} = 0 $ for $ i \neq j $, enhancing stability control.

Experimental results

Research questions

• RQ1 Can a generalized additive Runge-Kutta framework be developed that allows different stage values as arguments for different components of the right-hand side?
• RQ2 How can order conditions be derived for this new class of methods using NB-series theory?
• RQ3 What are the conditions for algebraic stability and absolute monotonicity in the generalized GARK framework?
• RQ4 Can implicit-explicit and implicit-implicit GARK schemes be constructed with improved stability and order accuracy?
• RQ5 Is it possible to design GARK methods that are both high-order and stability-decoupled for multi-physics systems?

Key findings

• The paper establishes a complete order conditions theory for GARK methods using NB-series, enabling the construction of high-order schemes.
• A new definition of algebraic stability is introduced and applied to GARK schemes, ensuring stability under general conditions.
• The method supports the construction of stability-decoupled schemes where $ P^{\{i,j\}} = 0 $ for $ i \neq j $, improving control over stability behavior.
• A second-order, algebraically stable GARK method is constructed with $ P^{\{1,2\}} = P^{\{2,1\}} = 0 $, demonstrating practical applicability.
• The framework allows for the development of transposed-classical IMEX schemes and extends the concept of stiff accuracy to the generalized setting.
• Theoretical analysis confirms monotonicity under step size restrictions, and the method supports multirate and symplectic extensions for future work.