Action-angle Variables for Generic 1D Mechanical Systems

Luca Biasco,Luigi Chierchia
DOI: https://doi.org/10.48550/arXiv.2003.05211
2020-03-11
Dynamical Systems
Abstract:We consider a 1D mechanical system $$\bar {\mathtt H}(\mathtt P,\mathtt Q)=\mathtt P^2+\bar {\mathtt G}(\mathtt Q)$$ in action-angle variable $(\mathtt P,\mathtt Q)$ where $\bar {\mathtt G}$ is a $2\pi$-periodic analytic function with non degenerate critical points. Then, we consider a small analytic perturbation of $\bar {\mathtt H}$ of the form $${\mathtt H}^*(\mathtt P,\mathtt Q;\hat{\mathtt P}) = \mathtt P^2+\bar {\mathtt G}(\mathtt Q)+ \eta {\mathtt F} (\mathtt P,\mathtt Q;\hat{\mathtt P})=:\mathtt P^2 + {\mathtt G}^*(\mathtt P,\mathtt Q;\hat{\mathtt P})\,, \qquad \eta\ll 1\ ,$$ where the perturbed potential $ {\mathtt G}^*$ may depend on the action $\mathtt P$ and also on parameters $\hat{\mathtt P}$ ("the adiabatic actions"); indeed, this is the form of a finite dimensional mechanical system close to an exact simple resonance after averaging over fast angles and disregarding the exponentially small remainder, see [5]. Up to a finite number of separatrices and elliptic/hyperbolic points the phase space of ${\mathtt H}^*$ is divided into a finite number of open connected components foliated by invariant circles. On every connected component we perform a (Arnold-Liouville) symplectic action-angle transformation which integrates the system. We give a complete and quantitative description of the analyticity properties of such integrating transformations, estimating, in particular, how such transformations differ from the integrating transformation for $\bar {\mathtt H}$; compare Theorem 6.1 below.
What problem does this paper attempt to address?