Rates of mixing for the Weil-Petersson geodesic flow: exponential mixing in exceptional moduli spaces (Q529612)

The authors prove that the geodesic flow of a negatively curved surface \(S\) of genus \(g\) with \(n\) cusps of a prescribed type is exponentially mixing. As a consequence they show that the Weyl-Petersson geodesic flow on \(T^{1}\mathfrak{M}_{(g,n)}\) mixes exponentially fast when \((g,n) = (1,1)\) or \((0,4)\). It was previously shown by many of the same authors that the WP geodesic flow on \(T^{1}\mathfrak{M}_{(g,n)}\) is mixing for all \((g,n)\) but fails to be exponentially mixing for \((g,n) \notin \{ (1,1), (0,4)\}\). Consider the surface of revolution \(X\) obtained by revolving the curve \(y = x^{r}\), \(r > 2\) around the \(x\)-axis. This surface has a single cusp singularity at the point \(p_{0} = (0,0,0)\). If \(p \in S = X - \{p_{0} \}\), then the Gaussian curvature at \(p\) is given by \(K(p) = - \frac{r(r-1)}{\delta(p)^{2}} + O(\delta(p)^{-1})\), where \(\delta(p)\) is the Riemannian distance between \(p\) and \(p_{0}\). Next, consider surfaces \(X\) with finitely many punctures \(\{p_{1}, \dots , p_{k} \}\) such that the metric in \(S = X - \{p_{1}, \dots , p_{k} \}\) is \(C^{5}\) and negative and lifts to a geodesically convex metric on the universal cover \(\tilde{S}\). Assume also that the Gaussian curvature in \(S\) is given by \(K(p) = \sum_{i=1}^{k} - \frac{r_{i}(r_{i} - 1)}{\delta_{i}(p)^{2}} + O(\delta_{i}(p)^{-1})\), where \(r_{i} > 2\) and \(\delta_{i}(p)\) is the Riemannian distance between \(p\) and \(p_{i}\) for all \(i\). The authors impose the further condition that \(||\nabla^{j}K(p)|| = \sum_{i=1}^{k} O(\delta_{i}(p)^{-2 - j})\) for \(j = 1,2,3\) and all \(p \in S\). For punctured surfaces \(S\) satisfying the conditions above, the authors show that the geodesic flow on \(T^{1}S\) is exponentially mixing. To obtain the exponential mixing the method used is to construct a surface \(\Sigma\) transverse to the geodesic flow \(\{\varphi_{t} \}\) and a return map \(F : \Sigma \rightarrow \Sigma\) that is \(C^{1 + \alpha}\) for some \(\alpha \in (0,1)\). One then applies work of \textit{V. Araújo} and \textit{I. Melbourne} [Ann. Henri Poincaré 17, No. 11, 2975--3004 (2016; Zbl 1367.37033)]. However, a priori, the flow \(\{\varphi_{t} \}\) is incomplete and the stable and unstable bundles \(W^{s}\) and \(W^{u}\) may not be \(C^{1 + \alpha}\), a necessary condition for constructing a \(C^{1+ \alpha}\) return map \(F : \Sigma \rightarrow \Sigma\). This technical difficulty is overcome by changing the Riemannian metric on \(T^{1}S\) and performing a time change on the geodesic flow \(\{\varphi_{t} \}\) to obtain a volume-preserving Anosov flow \(\{\psi_{t} \}\) on a complete Riemannian manifold of finite volume. The stable and unstable foliations for \(\{\psi_{t} \}\) are \(C^{1 + \alpha}\) by the theory of Anosov flows, and it follows that the stable and unstable bundles for \(\{\varphi_{t} \}\) are also \(C^{1 + \alpha}\). One concludes that \(\{\varphi_{t} \}\) possesses stable and unstable foliations \(W^{s}\) and \(W^{u}\) that are locally uniformly \(C^{1 + \alpha}\). From this one is able to construct a \(C^{1+\alpha}\) return map \(F : \Sigma \rightarrow \Sigma\).