Computation of invariant tori by orthogonal collocation (Q1971371): Difference between revisions

This paper is concerned with the computation of invariant tori for systems of ordinary differential equations. Assuming that the differential system is written in the form \( \theta'= f(\theta,r)\), \(r'= g(\theta,r)\) where \(\theta\in S^p\) are the angular variables and \( r \in \mathbb R^q\) are the ``radial'' variables. The existence of an invariant manifold \( r = r(\theta)\) implies that this function satisfies a first order partial differential equation (PDE) together with the periodicity boundary conditions. In this context the authors propose a Newton iterative approach to compute the invariant tori, so that at each step an Hermite cubic orthogonal collocation at a given mesh is carried out. By using a linear differential equation as model problem the authors prove for a uniform mesh that the proposed approach is stable and fourth order accurate in the norm \( \|\cdot \|_\infty\). It is remarked that these properties also hold for non uniform meshes with respect to the maximal step size. The paper ends presenting the numerical results for two problems: In the first one they start with a linear PDE with a known solution to check the order of accuracy of the proposed method for the model problem. In the second one they consider a systems of two non linear coupled oscillators such that without coupling each oscillator possess an invariant torus, and explore the behaviour of the method for different values of the coupling parameter.