Floer homology on the time-energy extended phase space (Q722384)

Let $(M,\omega)$ be an exact symplectic manifold sympletically convex at infinity, $H$ a smooth autonomous Hamiltonian on $M$ having a compact regular level $\Sigma=H^{-1}(0)$, which is of restricted contact type with respect to a global primitiv $\lambda$ of $\omega$. Then, introducing the Rabinowitz action functional \[ A_H(x,\tau)=\int_\mathbb{T}x^\ast\lambda-\tau\int_\mathbb{T}H(x(t))dt, \quad \mathbb{T}=\mathbb{R}/\mathbb{Z}, \] the Rabinowitz-Floer homology $\{RFH_k(\Sigma,M)\}_{k\in\mathbb{Z}}$ is defined. This paper shows the Rabinowitz Floer homology can be seen as a standard Floer homology for fixed perodic Hamiltonian orbits on the extended phases space $\tilde{M}=M\times T^\ast\mathbb{R}$, endowed with the symplectic form $\tilde(\omega)=\omega\times d\tau\wedge d\sigma$, where $H$ is lifted as $\tilde{H}(x,\sigma.\tau)=\tau H(x)$. The authors say that this result might produce a non-trivial product in Rabinowitz-Floer homology. Precisely, this statement is shown assuming the first Chern class of $TM$ vanishes on tori, $H$ is constant and positive outside of $M_0=M\setminus\iota(\sigma_\infty\times(0,+\infty)$, the flow $\phi^t_R$ of the Reeb vector field $R$ of $(\Sigma,\lambda|_\Sigma)$ is Morse-Bott, with special almost complex structure whose properties are described in \S1.2 (cf. \S1.3). The authors lift the Rabinowitz action functional to $\tilde{M}$ as \[ \mathbb{A}_{\tilde{h}}(\tilde{x})=\int_\mathbb{T}\tilde{x}^\ast\tilde{\lambda}-\int_\mathbb{T}\tilde{H}(\tilde{x}(t))dt. \] Here a loop of compatible almost complex structures $\tilde{J}_t(x,\tau,\sigma)=J_t(x,\tau)\times\hat{J}_1$ is chosen. Where $\hat{J}_1$ is the standard complex structure on $T^\ast\mathbb{R}\cong\mathbb{C}$, and $J\in\{J_t\}$ is a family of almost complex structures on $M$ compatible with $-\omega$. Then the Floer negative gradient equation of $\mathbb{A}_{\tilde{H}}$ for $\tilde{u}=(u,\zeta,\eta): \mathbb{R}\times\mathbb{T}\to \tilde{M}$ takes the form \begin{align*} \partial_s u+J_t(u,\eta)(\partial_t u-\eta X_H(u))&=0. \\ \partial_s\zeta-\partial_t\eta&=0, \\ \partial_s\eta+\partial_t\zeta-H(u)&=0. \end{align*} Here $X_H$ means the corresponding Hamiltonian vector field; $\iota_{X_H}\omega=-dH$. $\tilde{H}$ is not coercive; unbounded form below and from above. But in \S3, the authors show uniform bounds for the energy of solutions of the Floer equations associated to $\tilde{H}$ imply that these solutions take values in a compact subset of $\tilde{M}$, modulo transforms of the variable $\sigma\in\mathbb{R}$. Adopting these, the following Theorem is derived (cf. \S2.2): \textbf{Theorem 1}. The Floer complex $(F_\ast(\tilde{H},f),\partial)$ over $\mathbb{Z}_2$ produced by the critical points of an auxiliary Morse function $f$ on the manifold of 1-periodic orbits of $X_{\tilde{H}}$ and by the zero-dimensional space of finite-energy solutions of the above Floer equation modulo the $\mathbb{R}$-action is well defined. Periodic orbits of $\tilde{H}$ come in non-compact families, due to the presence of the $\mathbb{R}$-action on the time-energy extended phase space. Its Floer equation is also invariant with respect to this action. This involves revisiting the proof of transversality in Hamiltonian Floer theory, in order to check that we can actually work with an $\mathbb{R}$-invariant almost complex structure $J$. This is done in \S4, the last section, together with computation of the Fredholm index of the operators which arise when linearising the Floer equation. Since $\tilde{H}$ is autonomous, after modding out the $\mathbb{R}$-action, the periodic orbits of $X_{\tilde{H}}$ come in continuous families. This fact is dealt with by the standard way of counting Floer trajectories with cascades [\textit{U. Frauenfelder}, Int. Math. Res. Not. 2004, No. 42, 2179--2269 (2004; Zbl 1088.53058); \textit{A. Banyaga} and \textit{D. E. Hurtubise}, Algebr. Geom. Topol. 13, No. 1, 237--275 (2013; Zbl 1261.57029)]. By this fact, and results in \S4, the following Theorem is obtained (\S2.3). \textbf{Theorem 2}. There is a chain complex isomorphism \[ \Phi: RF_\ast(H,f)\to F_\ast(H,f). \] Before showing that the Floer homology of $\tilde{H}$ is isomorphic to the Rabinowitz Floer homology of $(\Sigma, M)$, Floer homology on the extended phase space $\tilde{M}$ is explained in \S1.