Non-virtually abelian anisotropic linear groups are not boundedly generated (Q2073759)

This is an important paper which introduces new methods for dealing with problems involving the bounded generation of groups. A group \(G\) is said to be \textit{boundedly generated} or having property (BG) if there exists a finite subset \(X\) of \(G\) and a positive integer \(m\) such that each element \(g\) of \(G\) can written in the form \[g=x_1^{n_1}\cdots x_m^{n_m},\] where \(x_i \in X\) and \( n_i \in \mathbb{Z}\). By definition every group having (BG) is finitely generated. The simplest examples are finitely generated virtually nilpotent groups. The finitely generated groups which do not have (BG) include all non-cyclic free groups of finite rank. The first examples of \textit{non-virtually solvable} groups having (BG) are due to \textit{D. Carter} and \textit{G. Keller} [Am. J. Math. 105, 673--687 (1983; Zbl 0525.20029)]. They show that \(\mathrm{SL}_n(\mathcal{O})\), where \(\mathcal{O}\) is a ring of algebraic integers and \(n \geq 3\), is boundedly generated and the bounding number \(m\) is determined by \(n\) and the discriminant of \(\mathcal{O}\). In addition they show that \(X\) can be chosen to consist entirely of elementary (and hence unipotent) matrices. This result (which does not hold in general for \(n=2\)) has subsequently been extended to all Chevalley groups of rank \(>1\) and most quasi-split groups. Although bounded generation is a purely combinatorial property of groups it has surprisingly many important consequences and applications. For example it is known that bounded generation for \(S\)-arithmetic subgroups of absolutely almost simple algebraic groups over a number field \(K\), where \(S\) is a finite set of valuations of \(K\) containing all the non-Archimedean valuations, ensures that such groups have the \textit{congruence subgroup property} (subject to some natural assumptions). More precisely this means that for these groups the \textit{congruence kernel,} as originally defined by Serre, is \textit{finite}. The property (BG) plays a crucial role in the proof of the Margulis-Zimmer conjecture for commensurated subgroups of higher rank \(S\)-arithmetic subgroups of Chevalley groups as well as the estimation of Kazhdan constants. In addition the natural extension of property (BG) to profinite groups, \((\mathrm{BG})_{\mathrm{pr}}\), plays a significant role. In particular the pro-\(p\) groups having \((\mathrm{BG})_{\mathrm{pr}}\) are precisely the \(p\)-adic analytic groups. A linear group \(\Gamma \subset \mathrm{GL}(K)\), where \(K\) is a field of characteristic zero, is said to be \textit{anisotropic} if \(\Gamma\) consists entirely of semi-simple elements. To date all the \(S\)-arithmetic subgroups of absolutely almost simple algebraic groups over a number field \(K\) known to be boundedly generated are not anisotropic. One of the principal aims of this paper is to find examples of boundedly generated linear groups which are generated by semi-simple elements. The main results follow from the following. Theorem A. Let \(\Gamma \subset \mathrm{GL}_n(K)\) be a linear group over a field \(K\) of characteristic zero which is not virtually solvable. Then in every possible presentation (BG) for \(\Gamma\) at least two of the elements \(x_i\) are not semi-simple. After reducing to the case where \(K\) is a number field the proof follows from a technical result involving a finite set of matrices in \(\mathrm{GL}_n(\overline{\mathbb{Q}})\) whose eigenvalues satisfy a condition called \textit{multiplicatively independence}. The proof which is long and intricate makes use of properties of \textit{generic elements} in Zariski-dense subgroups as well as Laurent's theorem from Diophantine geometry. This has a number of important consequences. Corollary B. An anisotropic linear group \(\Gamma \subset \mathrm{GL}_n(K)\) over a field of characteristic zero has (BG) if and only if it is finitely generated and virtually abelian. Theorem C. Let \(G\) be an algebraic group over a number field \(K\) and let \(S\) be a finite set of valuations of \(K\) containing all the archimedean ones. Then the infinite \(S\)-arithmetic subgroups of \(G\), where \(G\) is absolutely almost simple and \(K\)-anisotropic, are not boundedly generated. It is clear that by definition if \(\Gamma\) is a discrete group having property (BG) the its profinite completion \(\widehat{\Gamma}\) has property \((\mathrm{BG})_{\mathrm{pr}}\). The converse does not hold. Corollary D. There exist residually finite finitely generated groups \(\Gamma\) not having property \(\mathrm{(BG)}\) for which \(\widehat{\Gamma}\) has property \((\mathrm{BG})_{\mathrm{pr}}\). Diophantine geometry together with the existence of generic elements can provide an apparently novel way to determine whether (or not) a group is boundedly generated. In this paper this approach has extended existing results. For example a standard method for proving a group does \textit{not} have (BG) reduces to showing that its second bounded cohomology is an infinite dimensional real vector space. Theorem C proves the same result for some groups whose second bounded cohomology vanishes. It is hoped, for example, that these methods can provide a verifiable sufficient condition for some free amalgamated products to have bounded generation.