Jacobian-Free Variational Method for Constructing Connecting Orbits in Nonlinear Dynamical Systems

TL;DR

This paper introduces a Jacobian-free variational method for computing connecting orbits in high-dimensional nonlinear dynamical systems, overcoming limitations of traditional shooting methods and applicable to complex PDEs like the Kuramoto-Sivashinsky equation.

Contribution

A novel matrix-free variational approach for constructing connecting orbits that is robust, scalable, and does not require Jacobian matrices, suitable for high-dimensional systems.

Findings

Successfully applied to the Kuramoto-Sivashinsky equation

No limitation on unstable manifold dimension

Avoids exponential error amplification in chaotic systems

Abstract

In a dynamical systems description of spatiotemporally chaotic PDEs including those describing turbulence, chaos is viewed as a trajectory evolving within a network of non-chaotic, dynamically unstable, time-invariant solutions embedded in the chaotic attractor of the system. While equilibria, periodic orbits and invariant tori can be constructed using existing methods, computations of heteroclinic and homoclinic connections mediating the evolution between the former invariant solutions remain challenging. We propose a robust matrix-free variational method for computing connecting orbits between equilibrium solutions of a dynamical system that can be applied to high-dimensional problems. Instead of a common shooting-based approach, we define a minimization problem in the space of smooth state space curves that connect the two equilibria with a cost function measuring the deviation of a connecting curve from an integral curve of the vector field. Minimization deforms a trial curve until, at a global minimum, a connecting orbit is obtained. The method is robust, has no limitation on the dimension of the unstable manifold at the origin equilibrium, and does not suffer from exponential error amplification associated with time-marching a chaotic system. Owing to adjoint-based minimization techniques, no Jacobian matrices need to be constructed and the memory requirement scales linearly with the size of the problem. The robustness of the method is demonstrated for the one-dimensional Kuramoto-Sivashinsky equation.