Partial hyperbolicity and homoclinic tangencies (Q2018706)

Let \(M\) be a compact manifold. It is interesting and significant to describe the dynamics of a large class of diffeomorphisms of \(M\). This goal was achieved in a satisfactory way about 40 years ago for the hyperbolic diffeomorphisms. However, we are far from well understanding the dynamics beyond hyperbolicity and one would like to extend the previous properties to weaker forms of hyperbolicity. We say that a diffeomorphism \(f : M \longrightarrow M\) exhibits a \textit{homoclinic tangency} if there is a hyperbolic periodic orbit whose invariant manifolds \(W^s(O)\) and \(W^u(O)\) have a non-transverse intersection. \(f\) is \textit{partially hyperbolic} if there exists a filtration \[ \emptyset=U_0\subset U_1\subset\cdots\subset U_s=M \] such that the maximal invariant set \(\Lambda_i\) in each level \(U_i\setminus U_{i-1}\) is partially hyperbolic. That is, the tangent bundle of \(\Lambda_i\) splits into invariant linear subbundles: \[ T_{\Lambda_i}M=E^s\oplus E^c_1\oplus\cdots\oplus E^c_k\oplus E^u, \] such that each central bundle \(E^c_j\) is one-dimensional and there exists a forward iterate \(f^N\) which satisfies: (1) \(\|f^Nv\|\leq 1/2\) for each unitary \(v\in E^s\). (2) \(\|f^{-N}v\|\leq 1/2\) for each unitary \(v\in E^u\). (3) \(\|Df^N_xu\|\leq (1/2)\|Df^N_xv\|\) for each \(x\in \Lambda_i\), each \(j=0, \dots, k\) and any unitary vectors \(u\in E^s\oplus\cdots \oplus E^j,v\in E_{j+1}\oplus\cdots \oplus E^u\). Let Diff\(^1(M)\) be the set of \(C^1\) diffeomorphisms and \(\overline{HT}\) be the set of diffeomorphisms exhibiting a homoclinic tangency. The main result of this paper is: Any diffeomorphism \(f\in\mathrm{Diff}^1(M)\) can be approximated in Diff\(^1(M)\) by diffeomorphisms which exhibit a homoclinic tangency or by partially hyperbolic diffeomorphisms. This theorem is proved by showing that a partially hyperbolic diffeomorphism is a residual set in \(f\in\mathrm{Diff}^1(M)\setminus \overline{HT}\). In the reviewer's opinion, this paper gives a global picture of \(C^1\) dynamics which help us well to understand the dynamics behavior of diffeomorphisms. The technique the authors apply to obtain the main result is complicated and exquisite.