Numerical Methods for Approximating Invariant Manifolds of Delayed Systems

Abstract

In this paper we develop new methods for computing k-dimensional invariant manifolds of delayed systems for $k\geq2$. Our current implementation is built for $k=2$ only, but the numerical and algorithmic challenges encountered in this case will be also present for any $k>1$. For small delays, we consider methods for approximating delay differential equations (DDEs) with ordinary differential equations (ODEs). Once these approximations are made, any existing method for computing invariant manifolds of ODEs can then be used directly. We derive bounds on errors incurred by the most natural of these approximations. For large delays, we extend to DDEs the method originally introduced by Krauskopf and Osinga [Chaos, 9 (1999), pp. 768–774] for invariant manifolds of ODEs. We test the convergence of the resulting algorithms numerically and further illustrate our approach by computing two-dimensional unstable manifolds of equilibria in the context of phase-conjugate feedback lasers.