Zimmer's conjecture: subexponential growth, measure rigidity, and strong property (T) (Q2092825)

Let \(\Gamma\) be a lattice in a non-compact, simple Lie group \(G\) of rank \(n\ge 2\), such as \(G=\mathrm{SL}(n+1,{\mathbb R})\). Margulis' superrigidity theorem says that linear representations of \(\Gamma\) are either restrictions of a Lie group representation or have finite image. Its nonlinear version is Zimmer's conjecture which says that such lattices can not act on low-dimensional manifolds except for actions factoring over a finite group action. For example, for a lattice in \(\mathrm{SL}(n+1,{\mathbb R})\) actions on closed manifolds of dimension smaller than \(n\) should factor over a finite group action. (The bound is sharp because \(\mathrm{SL}(n+1,{\mathbb R})\) and its lattices act on the \(n\)-dimensional projective space.) There had been results for group actions on the circle (\(n=3\)) by Witte, Ghys, Burger-Monod, and for group actions on surfaces (\(n=4\)) by Farb-Shalen, Franks-Handel. A survey of the state of the conjecture around 2008 can be found in a survey by \textit{D. Fisher} [in: Geometry, rigidity, and group actions. Selected papers based on the presentations at the conference in honor of the 60th birthday of Robert J. Zimmer, Chicago, IL, USA, September 2007. Chicago, IL: University of Chicago Press. 72--157 (2011; Zbl 1264.22012)]. The paper under review proves Zimmer's conjecture for \(C^2\)-actions of cocompact lattices in \(\mathrm{SL}(n+1,{\mathbb R})\). The result was afterwards improved to \(C^1\)-actions by \textit{A. Brown} et al. [Compos. Math. 158, No. 3, 529--549 (2022; Zbl 07532079)]. The paper appeared as a preprint in August 2016. The proof of Zimmer's conjecture for the non-cocompact lattice \(\mathrm{SL}(n+1,{\mathbb Z})\), which follows a similar approach but is technically harder, appeared as a preprint by the same authors in October 2017 and is published in [Invent. Math. 221, No. 3, 1001--1060 (2020; Zbl 1482.22023)]. An overview of the proof is given in a talk in Seminaire Bourbaki by \textit{S. Cantat} [Astérisque 414, 1--48, Exp. No. 1136 (2019; Zbl 1485.22021)] and in several other published surveys by different authors. The actual result of the paper under review is that a \(C^2\)-action of \(\Gamma\) on a manifold \(M\) of dimension smaller \(n\) must leave some Riemannian metric invariant, which implies that the action factors over some compact group and hence (in the presence of unipotent elements) over some finite group. The idea behind the proof is to provide a converse to the fact that an action preserving a Riemannian metric needs to have vanishing Lyapunov exponents \(L(\gamma,x)=\liminf_{n\to\infty}\frac{1}{n}\int_X\log\Vert D_x\gamma^n\Vert d\mu(x)\). The authors start with an invariant measure, which exists by work of Zimmer, and prove subexponential growth of derivatives of group elements, which with Lafforgue's strong property T implies the existence of an invariant Riemannian metric. So, the key step is that for the actions in question the derivatives of group elements must have subexponential growth with respect to word length, that is, for any \(\epsilon>0\) exists \(C_\epsilon>0\) with \(\Vert D_x\gamma\Vert\le C_\epsilon e^{\epsilon l(\gamma)}\) for all \(x\in M,\gamma\in\Gamma\), where \(l(\gamma)\) is the word length of \(\gamma\) with respect to some fixed generating system. Among other things this uses Ratner's measure rigidity theorem and the work of Brown-Rodriguez-Wang, applied to the induced \(G\)-action on an auxiliary ``suspension space'', which is an \(M\)-bundle over \(\Gamma\backslash G\) and is a tool to go from a lattice action to a Lie group action. For \(A\subset \mathrm{SL}(n+1,{\mathbb R})\) the subgroup of positive diagonal matrices, the failure of uniform subexponential growth of derivatives would imply the existence of an \(a\in A\) and an \(a\)-invariant probability measure \(\mu\) on the suspension space with a positive Lyapunov exponent for the fiberwise derivative cocycle. The new idea is to construct from \(\mu\) a \(G\)-invariant measure \(\mu^\prime\) on the suspension space such that the fiberwise derivative cocycle continues to have positive Lyapunov exponent for some \(s^\prime\in A\). This yields a contradiction with Zimmer's cocycle rigidity theorem as there are no nontrivial linear representations in dimension smaller \(n\). This contradiction proves uniform subexponential growth of derivatives. For the final step, the authors then need Lafforgue's strong property T, which for cocompact lattices in \(\mathrm{SL}(n+1,{\mathbb R})\) has been proved by \textit{T. de Laat} and \textit{M. de la Salle} [Proc. Lond. Math. Soc. (3) 111, No. 4, 936--966 (2015; Zbl 1328.22003)]. The authors use strong property T and uniform subexponential growth of derivatives to produce an invariant Riemannian metric for the action. Namely, strong property T is used to construct a sequence of averaging operators on the space of sections of the bundle of symmetric \(2\)-tensors on \(M\), which applied to an arbitrary Riemannian metric in the limit yields an invariant Riemannian metric. As the authors say in the introduction, the proof has many surprising features, including the use of hyperbolic dynamics to obtain an essentially elliptic result, and the use of homogeneous dynamics to obtain results on nonlinear actions. A detailed sketch of the proof is given at the end of the introduction.