Abstract. We study a Hamiltonian describing a pendulum coupled with several anisochronous oscillators, devising an asymptotic expansion for the splitting (matrix) associated with a homoclinic point. This expansion consists of contributions that are manifestly exponentially small in the limit of vanishing hyperbolicity, by a shift-of-contour argument. Hence, we infer a similar upper bound on the splitting itself.
Main Concepts and Results
1.1.Background and history. The study of "separatrix splitting" dates back to Poincaré's classic Les Méthodes Nouvelles de la Mécanique Céleste [Poi93].Starting with Kolmogorov's 1954 note [Kol54], it was proved in a series of papers over a period of twenty years that quasiperiodic motions (invariant tori) are typical for nearly integrable Hamiltonians [Mos62,Arn63,Mos66a,Mos66b,Mos67], and that motions which become quasiperiodic asymptotically in time (stable/unstable manifolds) are stable under small perturbations [Mos67,Gra74]. Arnold [Arn64] described a mechanism how a chain of such "whiskered" tori could provide a way of escape for special trajectories, resulting in instability in the system. (A trajectory would typically lie on a torus and therefore stay eternally within a bounded region in phase space.) The latter is often called Arnold mechanism and the general idea of instability goes by the name Arnold diffusion. It is conjectured in [AA68] that Arnold diffusion due to Arnold mechanism is present quite generically, for instance in the three body problem.Arnold mechanism is based on Poincaré's concept of biasymptotic solutions, discussed in the last chapter of [Poi93], that are formed at intersections of whiskers of tori. Following such intersections, a trajectory can "diffuse" in a finite time from a neighbourhood of one torus to a neighbourhood of another, and so on.Chirikov's report [Chi79] is a very nice physical account on Arnold diffusion, while Lochak's compendium [Loc99] discusses more recent developments in a readable fashion and is a good point to start learning about diffusion. Gelfreich's introduction [Gel01] to splitting of separatrices is excellent, and we recommend it to anyone intending to study the topic. From there one should advance to [GL01], which covers more topics with more details. The extensive memoir [LMS03] by Lochak, Marco, and Sauzin is written from the geometric point of view. It has a historical flavor, making it interesting and accessible to virtually anyone.For the separatrix splitting of the periodically forced pendulum, see [HMS88,SMH91,DS92, EKS93]. In the case of the standard map, an asymptotic expression of the splitting has been obtained in [Gel99] of a pendulum coupled to d rotators, with φ ∈ S 1 := R/2πZ and I ∈ R the coordinate and momentum of the pendulum, and ψ ∈ T d := (S 1 ) d and A ∈ R d the angles and actions of the rotators, respectively. The perturbation f is assumed to be real-valued and real-analytic in its arguments, and λ is a (small) real number, whereas g > 0. This Hamiltonian is sometimes called the generalized Ar...