An efficient method to reconstruct invariant manifolds of saddle points

In vector field analysis, saddle points have two different types of invariant manifolds, namely stable ones and unstable ones. The invariant manifolds represent separatrices that partition the domain of trajectories into invariant regions of different dynamics. In this work, we analyze the basins of attraction of two different stable nodes by reconstructing the separatrices of a saddle point. To this purpose we present a computational algorithm that detects the points lying on the manifold, considering the plane generated by the two stable eigenvectors of the saddle point. Finally we reconstruct the surface by using the moving least-squares approximant method.