Regularity and rigidity of asymptotically hyperbolic manifolds (Q436151)

Let \({\bar X}^{n+1}\) be a manifold with boundary \(M^n\), and suppose a metric \(g\) is given on \(X^{n+1}:={\bar X}^{n+1}\backslash M^n\), making it a complete Riemannian manifold. For instance, consider the Poincaré ball model: the hyperbolic metric can be conformally transformed into the Euclidean one by simply rescaling, and the latter obviously extends to the boundary. In the general situation, let \(x\) be a defining function of the boundary \(M^n\) of \({\bar X}^{n+1}\), that is, \(x\restriction_{X^{n+1}}>0\), \(x\restriction_{M^n}=0\), \(\text{d}x\neq 0\) at every point of \(M^n\). If \(x^2g\) extends to a metric \(\bar g\) on \({\bar X}^{n+1}\), satisfying the Hölder condition \(C^{k,\mu}\), then the Riemannian manifold \(X^{n+1}\) is said to be conformally compact of regularity \(C^{k,\mu}\). Under some mild conditions, \(X^{n+1}\) asymptotically looks like the hyperbolic space, and that is why \(X^{n+1}\) it is called asymptotically hyperbolic. The interest in the subject is not solely due to having some roots in hyperbolic geometry. Relevant applications to general relativity can hardly be surprising (see, e.g., [\textit{E. Witten}, Adv. Theor. Math. Phys. 2, No. 2, 253--291 (1998; Zbl 0914.53048)]). Moreover, since the conformal class of \(\bar{g}\restriction_{M^n}\) depends on \(g\) and not on the choice of the defining function \(x\), the above construction turns out to be a valuable tool in conformal geometry, too. Further features may be appreciated by making a comparison with the earlier subject of asymptotically locally Euclidean manifolds. The problem is to find sufficient conditions for a complete Riemannian manifold to be conformally compact of reasonable regularity, that are ``intrinsic'', i.e., only involving \(\left(X^{n+1},g\right)\). The main natural requirement is the following asymptotically hyperbolic condition of order \(a\) on the curvature \[ \left\|\text{Rm}-K\right\|\leq C\exp\left(-a\rho\right), \] where \(\text{Rm}\) is the Riemann tensors of \(g\), \(K_{ijkl}=-\left(g_{ik}g_{jl}-g_{il}g_{jk}\right)\) (recall that \(\text{Rm}=K\) for a hyperbolic space, in the sign convention they use), \(\rho\) is the distance function from a fixed point, and \(C\) is a positive constant. A more technical condition is the existence of an essential subset (a notion not to be confused with other meanings in different contexts). This condition was used to address the same problem in the paper [\textit{E. Bahaud} and \textit{R. Gicquaud}, J. Geom. Anal. 21, No. 4, 1085--1118 (2011; Zbl 1259.53038)], which may be considered as a starting point for the work under review. At the beginning, as in the aforementioned article, the authors exploit the analysis of ODEs and keep an additional hypothesis of \(b\)-th order decay on the covariant derivatives of \(\text{Rm}\). Then, by using tools from elliptic PDE theory, they introduce suitable harmonic coordinates near infinity. As a first outcome, they obtain \(C^{2,\mu}\) regularity for some \(\mu\in (0,1)\) when \(b>2\). Next, noticeably, the additional condition on the covariant derivatives is dropped by means of Ricci flows techniques and some estimates due to W. X. Shi. This way, only assuming asymptotic hyperbolicity of order \(a\) and the existence of an essential subset, they prove \(C^{0,\frac23 a}\) regularity for \(0<a<1\), \(C^{1,\mu}\) regularity for some \(\mu\in (0,1)\) when \(2-\frac1{n+1}<a<2\), and \(C^{1,\mu}\) regularity for all \(\mu\in (0,1)\) when \(a\geq 2\). The result in the case \(a>2\) is credited to be somewhat optimal in [\textit{R. Gicquaud}, ``Conformal compactification of asymptotically locally hyperbolic metrics. II: weakly ALH metrics'', Preprint, \url{arXiv:1109.5096}, p. 4]. In the decade before publication of the article under review, several authors proved that, under one or another weak additional assumption, \(X^{n+1}\) is forced to be nothing but a hyperbolic space. In other words, they stated rigidity theorems in the subject. Since most of them require some regularity at infinity, the results discussed above allow one to replace that requirement with an intrinsic condition. More specifically, the rigidity result the title refers to performs such an improvement on a theorem from [\textit{S. Dutta} and \textit{M. Javaheri}, Adv. Math. 224, No. 2, 525--538 (2010; Zbl 1210.53040)]. Namely, the authors prove that (still in the presence of an essential subset) when \(n\geq 4\), \(a>2\), and, in addition, if one has \(\text{Ric}\geq -ng\) and \(X^{n+1}\) is simply connected at infinity (the only ``less intrinsic'' condition left), then \(\left(X^{n+1},g\right)\) is a standard hyperbolic space; moreover, the same is true, if instead of \(n\geq 4\), one has \(n=3\) and \(\int_{X^{n+1}}\left\| \text{Rm}-K\right\|\text{d}\mu_g<\infty\).