Flat rank of automorphism groups of buildings (Q2463777)

We let \(G\) denote a totally disconnected locally compact group. The authors' eventual aim is a classification of these groups and they are here concerned with the flat rank as an invariant for \(G\). Topologising Kac-Moody groups provides the authors with examples of such groups. \textit{V. Kac} [Math. USSR, Izv. 2, 1271--1311 (1968); translation from Izv. Akad. Nauk SSSR, Ser. Mat. 32, 1323--1367 (1968; Zbl 0222.17007)] had been concerned with graded Lie algebras, \textit{R. V. Moody} [J. Algebra 10, 221--230 (1968; Zbl 0191.03005)] concerned with semi-simple Lie algebras. It was \textit{R. V. Moody} and \textit{K. L.Teo} [J. Algebra 21, 178--190 (1972; Zbl 0232.20089)], referring to \textit{C. Chevalley} [Tôhoku Math. J., II. Ser. 7, 14--66 (1955; Zbl 0066.01503)], who introduced groups associated with these algebras for fields of characteristic \(p>0\). The authors avoid a definition of a Kac-Moody group but their primary interest is in groups over fixed finite fields and having generalised Cartan matrices; an example [see \textit{B. Rémy}, Geom. Funct. Anal. 14, No. 4, 810--852 (2004; Zbl 1056.22015)] is the linear Kac-Moody group \(\text{SL}_n(F_q[t,t^{-1,1}])\) over Laurent polynomials, for a field \(F_q\) of characteristic \(p\) with \(q\) elements. The authors restrict their investigation to Kac-Moody groups \(G\) which act as automorphism groups of twin buildings with sufficiently transitive actions. A Coxeter complex is made up of a Weyl group \(W\), classically a finite rank Coxeter group, the related Coxeter matrices and a set \(S\) of involutions. A (Weyl) chamber system \(\mathcal{C}\) is a finite simplicial complex in which all the cells are homeomorphs of simplices. The chambers are the maximal simplices and are called adjacent if they have a face of co-dimension 1 in common. The Weyl group is a Coxeter group of finite rank generated by the reflections across the walls of the chambers or by the adjacency relations. Automorphisms of \((W,S)\) act as permutations of the chamber system. A building [\textit{N. Bourbaki}, Groupes et algèbres de Lie, Chapitre IV (Paris 1968; Zbl 0186.33001)], denoted \(\mathcal{X}\), is a cell complex, not necessarily of finite rank, made up of unions of simplicial complices. It can be divided into apartments such that any two chambers are contained in a common apartment and two apartments containing a given chamber are isomorphic. A Coxeter complex has a single apartment. The Coxeter complices and Coxeter groups are directly connected in that the automorphism system of a Coxeter complex is the Coxeter group itself; the relation between buildings and groups is much more complicated. Instead of Coxeter systems one may use more abstractly defined buildings, viz. the system of BN-pairs of [\textit{J. Tits}, Lect. Notes Math. 386. Berlin-Heidelberg-New York: Springer-Verlag (1974; Zbl 0295.20047)]. Given a group acting simplicially on a building, transitively on pairs of chambers and on the apartments, \(B\) and \(N\) are subgroups associated with the chambers and the apartments respectively; \(N/{N \cap B}\) can be identified as a Weyl group. It is indeed a (possibly infinite) Coxeter group with \(S\) uniquely determined by the BN-pair. Different BN-pairs can be associated with the same building. The authors assume further criteria to ensure that the buildings used in the article have finite \(S\). When a Kac-Moody group \(\Lambda\) over a finite field acts on twin buildings \(\mathcal{X}_{\pm}\) (denoted positive and negative) the topological Kac-Moody group \(\overline{\Lambda}\) can be determined by a sufficiently transitive action of the (minimal) Kac-Moody group as automorphisms on an associated building; it is defined to be the closure of \(\Lambda/K\) in the (permutation) topology of the automorphism group, where \(K\) denotes the common kernel for the positive and negative buildings. The topological Kac-Moody group based on \(\text{SL}_{n}(F_{q}[t, t^{-1}])\) is dealt with by Rémy (loc. cit.) and denoted there as \(\text{PSL}_{n}(F_{q}(t))\). The underlying building is of affine type \(\widetilde{A}_{n-1}\). Kac-Moody groups have no explicit arithmetic and algebraic structure. However the Weyl group ensures that the higher rank Kac-Moody groups over large enough finite fields are like Lie groups in having a lattice structure, viz., a discrete subgroup with a finite covolume (with respect to the G-invariant measure) [\textit{A. Borel} and \textit{Harish-Chandra}, Ann. Math. (2) 75, 485--535 (1962; Zbl 0107.14804)]. The lattices need not be co-compact. There is a division of types of Coxeter systems corresponding to surfaces of constant closure in a manifold, viz., spherical, Euclidean and hyperbolic geometrical systems. When \(W\) is a finite Weyl group the building is said to be of spherical type since the Coxeter complex is a topological sphere. The chambers simplicially intersecting the sphere to form a triangulation. Work on groups over finite fields of characteristic \(p > 0\) leads to a new class, affine buildings, where each apartment comes from a tiling of Euclidean space; there will be doubly infinite sequences of chambers (graphically represented by circuits). Affine buildings will be spherical `at \(\infty\)' but the underlying dimension drops. The affine Weyl groups are extensions of finite Weyl groups; they have maximal Abelian subgroups such that the quotients are finite Weyl groups. The Coxeter complex for an affine Weyl group is the Euclidean plane tiled by equilateral triangles. A locally finite infinite (simplicial) tree without end-points is an affine building. If the tree has less than three edges at each vertex the tree is a Coxeter complex of algebraic \(\mathbb Z\)-rank 1 since the Coxeter generators are unrelated; the apartments are copies of \(\mathbb R\). For spherical buildings the chambers are simplices and for affine buildings the chambers are products of simplices. The bilinear forms for linear representations of hyperbolic Coxeter groups are negative at each vertex; removing any vertex results in positivity of the form. Following a suggestion of M. Gromov, the group together with word length may be looked upon as a metric space. Hyperbolic groups act `geometrically' on hyperbolic buildings. A metric space is called Gromov-hyperbolic [\textit{M. Gromov}, Soft and hard symplectic geometry. ICM Series. AMS, Providence, RI (1987; Zbl 0664.53016)] if, for some \(\delta \geq 0\), all triangles are `thin' in the sense that for any geodesic triangle each side of the triangle is contained in a \(\delta\)-neighbourhood of the other two sides. A classical hyperbolic space is Gromov-hyperbolic. Let \(B(G)\) denote the set of compact open subgroups of \(G\); for the existence see [\textit{D. van Dantzig}, Compos. Math. 3, 403--426 (1936; Zbl 0015.10202)]. These form a basis of neighbourhoods of the identity in \(G\). To investigate bicontinuous automorphisms on \(G\) the authors restrict themselves to automorphisms to \(B(G)\). Given an automorphism \(\alpha\) the authors impose a discrete metric on \(B(G)\) as a distance between \(U\) and \(\alpha(U)\) in such a way that \(G\) acts as isometries. The displacement function for an automorphism \(\alpha\) is defined as the minimum distance between \(U\) and \(\alpha(U)\) for \(U \in B(G)\). For an automorphism \(\alpha\), a compact open subgroup of \(G\) is called tidy if it minimises the displacement function, so every automorphism of \(G\) has a tidy compact open subgroup. A subgroup of \(\Aut(G)\) is called flat if all its elements have a common tidy \(U \in B(G)\); a subgroup of \(G\) is called flat if the group of inner automorphisms induced by the subgroup is flat, so every \(U \in B(G)\) is flat. The authors show that a subgroup of \(G\) which is a finite extension of a finitely generated Abelian group is flat. Elaborating on \textit{G. A. Willis}' [Math. Ann. 300, No. 2, 341--363 (1994; Zbl 0811.22004)] description of tidyness the authors call the integer \(s_{G}= \min \{[\alpha(V): \alpha(V) \cap V]\mid V \in B(G)\}\) the scale of \(\alpha\). Given a flat subgroup \(H\) of \(\Aut(G)\) and defining \( H_{1} \) to be the subgroup \(\{\alpha \in H: s_{G}(\alpha)= 1= s_{G}(\alpha^{-1})\}\), then \(H / H_{1}\) is a maximal free Abelian subgroup. The flat rank of \(H\) is defined as \(\mathbb Z\)-rank of \(H / H_{1}\). The flat-rank of G is defined to be the supremum of flat-ranks of the flat subgroups. One of the main theorems is on a lower bound for the flat rank. The minimal Kac-Moody group \(\Lambda\) is assumed to be locally finite (i.e., it has a finite number of edges at each vertex). The algebraic rank of a group \(H\), denoted alg-rk(\(H\)), is the maximum rank of the free abelian subgroups of \(H\). By lifting a free abelian subgroup \(A\) of \(W\) to a flat subgroup of \(\overline{\Lambda}\) with flat rank equal to \(\text{rank}_{\mathbb Z}(A)\), the authors show that \(\text{alg-rk}(W)\leq\text{flat-rk}(\overline{\Lambda})\). The proof is valid also for the geometric completion of \(\Lambda\). For the corresponding theorem for an upper bound instead of the topological Kac-Moody group the authors consider a geometric completion which we denote as \(\widetilde{\Lambda}\). A contractible metric space with unique geodesics between points and such that the distances between points are convex functions is called a CAT(0) space (i.e., comparison of Aleksandrov-Toponogov where the 0 indicates that the curvature of the geodesics is non-positive [see the strangely titled book by \textit{M. W. Davis}, Buildings are CAT(0), Geometry and cohomology in group theory. Cambridge: Cambridge University Press. Lond. Math. Soc. Lect. Note Ser. 252, 108--123 (1998; Zbl 0978.51005)]). It is assumed that there is a \(W\)-valued (Coxeter) `distance' between chambers; the length of words comprising the distance function provides a metric for the building. They assume also that \(G\) acts strongly transitively, or even the weaker \(\delta\)-2 transitively, on the building. A gallery is a finite sequence of chambers such that consecutive chambers have a common face. The shortest galleries between pairs of chambers are the geodesics. The CAT(0)-realisation of \(W\), denoted \(|W|_0\), is the quotient of \(W\) by equivalence classes (called `residues'). The general idea for equivalence is so that, for any \(T \subset S\), to any chamber \(c\) in \(\mathcal{C}\) there would be a unique chamber \(c'\) in the Coxeter \(T\)-subsystem at a minimal distance from \(c\). The authors assume throughout that the chamber systems have `finite thickness'; thickness involves a number of residues but not chambers directly. The reviewer does not understand how the authors manage to assume that the geometric completion does not depend on the CAT(0) realisation. The geometrical rank of a complete CAT(0) space \(X\), denoted \(\text{rk}(X)\), is defined as the maximal dimension of isometrically embedded Euclidean subspaces in \((G,\delta)\). There is a problem for determining flat ranks in that a Gromov-hyperbolic CAT(0) building has no flats. Under the assumption of a co-compact action of the group, the authors consider quasi-isometries and quasi-flats. Quasi-flats take over as hyperbolicity is invariant under quasi-isometries. A (\(\lambda,k)\)-quasi isometry \(f\) is of the form \(\lambda ^{-1} \operatorname{dist} f(x),f(y)-k\leq\operatorname{dist}f(x), f(y)\leq \lambda\operatorname{dist} f(x),f(y) +k\). Heuristically it has the effect of looking, at a long distance, like an isometry. A quasi-flat is a subspace which is quasi-isometric to a Euclidean space. The authors then use a theorem of \textit{B. Kleiner} [Math. Z. 231, No. 3, 409--456 (1999; Zbl 0940.53024)] for Gromov-hyperbolic buildings, showing that since quasi-isometries are invariant and that a quasi-flat is, dependent on the parameters \(\lambda\) and \(k\), arbitrarily close to a flat. The authors prove that the algebraic rank for the Weyl group of the building is less than or equal to the geometric rank of \(|W|_0\). Incorporating the CAT(0) realisation of \(G\) one has \(\text{alg-rk}(W)\leq \text{flat-rk}(\widetilde{G}) \leq\text{flat-rk}(W)\). This is a global measurement of the flatness. The authors prove that the CAT(0)-realisation of a building has the same rank as any of its apartments. The reviewer is not convinced that the authors have proved that the inequality holds also for \(\overline{\Lambda}\) as implied in their abstract. The final theorem shows the existence, for each \(n \geq 1\), of totally disconnected locally compact groups \(G_n\) with flat rank dimension \(n\) which are simple, compactly generated and not linear. For linear examples one may use linear groups like the \(\text{PSL}_n(F_q(t))\) for \(n \geq 3\). For nonlinear models the authors use hyperbolic extensions of affine diagrams. They fix a field of cardinality at least five and let \(D_1 = W_1\) be the hyperbolic Coxeter diagram consisting of a circuit with 5 vertices all of whose edges are labeled \(\infty\). The vertices are unrelated algebraically, so the maximal rank of an Abelian subgroup of \(W_{1}\) will be 1 and \(\text{alg-rk}(W_1) = \text{rk}(|W_1|_0) = 1\). A Gromov-hyperbolic diagram can be isometrically embedded in a metric tree. To get a nonlinear group the authors find a particular type of affine building [see Rémy (loc. cit.)] and the diagram \(\widetilde{A}_1\) with edges labeled \(\infty\) is linked by edges labeled \(\infty\) to each vertex of \(D_1\); these are free products of the respective groups. The proof of nonlinearity is involved and involves constructing the right type of building to get a suitable kernel for the linear group. If \(n > 1\) the same procedure is used to get a sufficiently large Abelian subgroup of the Weyl group. The translation subgroup corresponding to the \(\tilde{A}_{n}\)-subdiagram is an Abelian subgroup of rank \(n\) in \(W_{n}\). The proof would involve also arrangements of the coefficients of the generalised Cartan matrices to yield the \(W_{n}\). The geometric completions provide suitable locally compact groups for the theorem. Similar nonlinearities arise in quantum gravity, see \textit{S. Mizoguchi, K. Mohri} and \textit{Y. Yamada} [Classical Quantum Gravity 23, No. 9, 3181--3193 (2006; Zbl 1152.83417)].