Boundaries, Weyl groups, and superrigidity (Q422144)

Let \(\Gamma\subset G\) be an irreducible lattice in a simple Lie group with \(\mathrm{rk}_{\mathbb R}\left(G\right)\geq 2\). Margulis' celebrated superrigidity theorem decribes the representations of \(\Gamma\) to simple, center-free Lie groups \(H\) with \(\mathrm{rk}_{\mathbb R}\left(H\right)\geq 2\): if \(\rho:\Gamma\rightarrow H\) has unbounded Zariski-dense image, then \(\rho\) is the restriction of a representation \(\rho:G\rightarrow H\). \textit{G.\ Margulis} [Invent. Math. 76, 93--120 (1984; Zbl 0551.20028)] actually proved this result in a more general setting and in the meantime there have been various generalizations to nonlinear actions of \(\Gamma\). For example, \textit{É.\ Ghys} [Invent.\ Math.\ 137, No.\ 1, 199--231 (1999; Zbl 0995.57006)] showed that (under the same assumptions on \(\Gamma\)) every continuous action of \(\Gamma\) on the circle factors over the action of a finite group. \textit{M.\ Burger} and \textit{S.\ Mozes} [J. Am. Math. Soc. 9, No. 1, 57--93 (1996; Zbl 0847.22004)] proved that an action of \(\Gamma\) on a proper CAT(-1)-space \(Y\) which is non-elementary and does not leave invariant a proper subset of \(Y\) must extend to a homomorphism \(G\rightarrow \mathrm{Isom}\left(Y\right)\). The paper under review suggest a very interesting unified approach to such superrigidity results by looking at `generalized Weyl groups' \(W_{G,B}\), which the authors define as the groups of \(G\)-equivariant automorphisms of \(B\times B\) for some \(G\)-boundary \(B\). The notion of \(G\)-boundary was introduced by \textit{M.\ Burger} and \textit{N.\ Monod} in [Geom. Funct. Anal. 12, No.\ 2, 219--280 (2002; Zbl 1006.22010)] and it means by definition that \(B\) is a measurable \(G\)-space such that the action of \(G\) is amenable and the projection \(B\times B\rightarrow B\) is ergodic with polish coefficients. The definition of \(W_{\Gamma,B}\) is independent of all choices, in particular if \(\Gamma\) is a lattice in \(G\), then \(W_{\Gamma,B}\) depends not on \(\Gamma\) but only on \(G\). One knows that \(B\) is trivial if and only if \(G\) is amenable. If \(B\) is nontrivial, then the permutation of the two factors of \(B\times B\) yields a nontrivial element \(w_{\mathrm{flip}}\in W_{\Gamma,B}\). For hyperbolic groups \(\Gamma\) one actually obtains \(W_{\Gamma,B}\simeq {\mathbb Z}/2{\mathbb Z}\), but for lattices in higher rank noncompact simple algebraic groups there are always additional elements in \(W_{\Gamma,B}\). Indeed in this case one may choose \(B=G/P\) and one recovers the classical Weyl group \(W_{\Gamma,B}\cong \mathrm{Weyl}_G\), e.g.\ \(W_{\Gamma,B}\cong S_n\) for \(G=\mathrm{SL}\left(n\right)\). The paper under review considers representations \(\rho:\Gamma\rightarrow H\) of a countable group \(\Gamma\) to \(H\), where \(H\) is supposed to be the locally compact group of \(k\)-points (for a local field \(k\)) of a connected adjoint \(k\)-simple \(k\)-algebraic group. For such a group the diagonal action of \(H\) on \(H/P\times H/P\) has finitely many orbits, indexed by \(\mathrm{Weyl}_H\) with a unique full-dimensional orbit, corresponding to \(w_{\mathrm{long}}\in \mathrm{Weyl}_H\). This orbit contains \(\left(P,w_{\mathrm{long}}P\right)\) and it is the image of an embedding of \(H/{\mathcal{Z}}_H\left(A\right)\). The authors then have the following general result. If \(\rho\) is unbounded and Zariski-dense, then a) there exists a unique measurable \(\Gamma\)-equivariant map \(\phi:B\rightarrow H/P\), b) \(\phi\times \phi:B\times B\rightarrow H/P\times H/P\) factors through the embedding \(H/{\mathcal{Z}}_H\left(A\right)\rightarrow H/P\times H/P\), c) there exists a homomorphism \(\pi:W_{\Gamma,B}\rightarrow \mathrm{Weyl}_H\) with \(\pi\left(w_{\mathrm{flip}}\right)=w_{\mathrm{long}}\) and \(\left(\phi\times\phi\right)\circ w=\pi\left(w\right)\circ \left(\phi\times\phi\right)\). Moreover, if \(\Gamma\) is a lattice in a simple algebraic group \(G\), then \(\pi:W_{\Gamma,B}=\mathrm{Weyl}_G\rightarrow \mathrm{Weyl}_H\) is injective and maps \(w_{\mathrm{long}}^{\left(G\right)}\) to \(w_{\mathrm{long}}^{\left(H\right)}\). If \(\Gamma\) is an \(\widetilde{A}_2\)-group, then \(\pi:S_3\rightarrow \mathrm{Weyl}_H\) is injective and \(\pi\left(\left(1,3\right)\right)=w_{\mathrm{long}}^H\). These results already suffice to derive some results of the type ''certain \(\Gamma\) admits no unbounded Zariski dense homomorphism to certain \(H\)''. For example this is the case if \(\mathrm{rk}_k\left(H\right)=1\) and \(\Gamma\) is either a lattice in a simple \(l\)-algebraic group \(G\) with \(\mathrm{rk}_l\left(G\right)\geq 2\), or \(\Gamma\) is an exotic \(\widetilde{A}_2\)-group. Moreover the existence of an injective homomorphism \(\pi:\mathrm{Weyl}_G\rightarrow \mathrm{Weyl}_H\) with \(\pi\left(w_{\mathrm{long}}^{\left(G\right)}\right)=w_{\mathrm{long}}^{\left(H\right)}\) can be excluded in many other cases, such as \(G=\mathrm{PGL}_4\left(l\right)\) and \(H=\mathrm{PGL}_3\left(k\right)\). Finally, if \(\Gamma\) is again either a lattice in a simple algebraic group \(G\) or an \(\widetilde{A}_2\)-group, then \(\pi\) is an isomorphism of Coxeter groups and using some results on Tits buildings one can derive Margulis' superrigidity and a new result about nonlinearity of exotic \(\widetilde{A}_2\)-groups. This is a research announcement, complete proofs will appear elsewhere. The authors mention in the introduction that they also have cocycle versions (in the sense of Zimmer cocycle rigidity) for their results, which cover more new ground.