Topologically crossing heteroclinic connections to invariant tori. (Q1406503)

Let \(M\) be a symplectic manifold of dimension \(2n_c+ 2n_h\), with \(n_c\), \(n_h> 0\), and let \(f_\mu: M\to M\) be a family of symplectic diffeomorphisms that depends smoothly on \(\mu\) with \(|\mu- \mu_0|< a_0\) for some \(a_0> 0\). We assume that each \(f_\mu\) has a partially hyperbolic fixed point \(p_\mu\) such that the derivative of \(f_\mu\) at \(p_\mu\) has \(n_h\) eigenvalues \(\lambda\) with \(|\lambda|< 1\), \(n_h\) eigenvalues \(\lambda\) with \(|\lambda|> 1\) and \(2n_c\) eigenvalues \(\lambda\) with \(|\lambda|= 1\). We also assume that \(f_{\mu_0}\) satisfies a twist condition on the center manifold \(W^c(p_{\mu_0})\) of \(p_{\mu_0}\). By the KAM theorem, there exists a family of tori \(\{T_{\mu,\alpha} \}_{\alpha \in{\mathcal I}}\) invariant under \(f_\mu\), indexed by a Cantor set \({\mathcal I}\). We assume also that the unstable manifold \(W^u (T_{\mu,\alpha})\) of the torus \(T_{\mu,\alpha}\) has a topologcial crossing intersection with the stable manifold \(W^s(T_{\mu, \beta})\) of the torus \(T_{\mu, \beta}\). The tori \(\{T_{\mu, \alpha} \}\) are called transition tori for \(f_\mu\). Under these assumptions, the authors prove the following shadowing lemma: For any bi-infinite collection \(\{T_{\mu,\alpha_i}\}_{i\in \mathbb{Z}}\) of transition tori and any collection \(\{\varepsilon_i\}_{i \in\mathbb{Z}}\) of positive numbers \(\varepsilon_i\), there exists a sequence \(\{n_i\}_{i\in\mathbb{Z}}\) of natural numbers and an orbit \(\{z_i\}_{i\in \mathbb{Z}}\) of \(f_\mu\) such that \(z_{i+1}= f^{n_i}_\mu (z_i)\) and \(d(z_i, T_{\mu,\alpha_i})< \varepsilon_i\) for all \(i\in \mathbb{Z}\). The authors also prove the following theorem of the existence of symbolic dynamics: For any finite collection of transition tori \(\{T_{\mu,\alpha_i}\}_{i= 1,2,\dots, d}\), any \(\varepsilon> 0\) and any positive integer \(N\) there exist \(n>N\) and an invariant set \(S_\mu\) relative to \(f^n_\mu\) such that (i) \(S_\mu\) is contained in an \(\varepsilon\)-neighborhood of \(\bigcup T_{\mu,\alpha_i}\), (ii) there exists a surjective continuous map \(\rho_\mu: S_\mu\to \Sigma_d\) with \(\rho_\mu\circ f^n_\mu= \sigma\circ \rho_\mu\), where \((\Sigma_d,\sigma)\) denotes a full shift over \(d\) symbols. For the proofs, the authors recall the notion of correctly aligned windows. Most part of the paper is devoted to the construction of windows, on which the proofs relies.