Strongly dense free subgroups of semisimple algebraic groups. (Q1932665)

The authors define and prove the existence of so-called strongly dense free subgroups in semisimple algebraic groups. These results are used to answer questions on expansion properties of Cayley graphs over finite simple groups of Lie type in a later paper. In fact, the results are of much more general interest with many prospective applications, some of which are carried out in the paper. As can be expected of these authors, the description of the results detailed in the introduction is extremely lucid. The so-called Hausdorff-Banach-Tarski paradox entails finding free groups of rigid rotations acting on spheres. For a fixed-point free action by a free group on the sphere \(S^n\) to exist, topological considerations (Lefschetz numbers, in particular) show that \(n\) has to be odd. In a paper published in volume 29 of L'Enseignement Mathématique, \textit{P. Deligne} and \textit{D. Sullivan} proved that for \(n>1\), there is a free non-Abelian group of rigid rotations which acts freely on \(S^{2n-1}\) [Enseign. Math., II. Sér. 29, 145-150 (1983; Zbl 0521.57035)]. In the paper following it in the same issue [ibid. 29, 151-164 (1983; Zbl 0533.22009)], \textit{A. Borel} proved -- among other things -- a corresponding (necessarily weaker) result for even-dimensional spheres. Borel shows the existence of a free non-Abelian subgroup \(\Gamma\) in any compact, connected, semisimple Lie group \(U\) such that for any proper closed subgroup \(V\) of \(U\), nontrivial elements of \(\Gamma\) (acting by left translations on \(U/V\)) fix \(\chi(U/V)\) points. The existence of free subgroups in real Lie groups or, more generally, in algebraic groups over various fields has diverse, interesting implications. The existence of free subgroups in semisimple groups is itself an immediate consequence of the (deep) theorem known as the ``Tits alternative'' which shows that a finitely generated group of matrices over a field must either be virtually solvable or must contain a free non-Abelian group. The interest is to produce free subgroups with additional properties as this has consequences in seemingly unrelated questions. The authors define an abstract subgroup \(\Gamma\) of an algebraic group \(G\) to be strongly dense if any two noncommuting elements of \(\Gamma\) generate a Zariski-dense subgroup. The main theorem of this paper asserts: Let \(G\) be a connected, semisimple algebraic group over an uncountable algebraically closed field \(k\). If \(G\) is not the group of type \(C_2\) when \(k\) has characteristic \(3\), there exists a free group \(\Gamma\) of rank \(2\) in \(G(k)\) which is strongly dense. As the authors observe, the statement also implies the existence of strongly dense free subgroups of any finite rank \(>1\) and of countably infinite rank since non-Abelian subgroups of strongly dense subgroups are strongly dense. The theorem is also valid for the exceptional case of type \(C_2\) over \(k\) of characteristic \(3\) but this is proved in a later paper (the reference to this is missing). Indeed, many of the arguments used in the proof of the main theorem are valid in this exceptional case, too, but there are some additional difficulties. The authors also point out that an argument using the Baire category theorem implies that a generic pair of elements in \(G(k)\times G(k)\) generates a strongly dense free subgroup of \(G\). Further, the theorem is valid for more general fields like those which have infinite transcendence degrees over their prime fields; there are weaker results valid for algebraically closed fields which are not algebraic closures of finite fields. Corollaries are drawn on probabilistic generation in finite simple groups and on simultaneous paradoxical decompositions which generalize Borel's results. The following two interesting problems are raised by the authors: Problem 1: J. Tits showed that every finitely generated (Zariski-) dense subgroup \(\Gamma\) of \(G\) contains a dense free subgroup. A challenging problem is to determine whether or not \(\Gamma\) always contains a strongly dense free subgroup. Problem 2: Borel showed (in the paper quoted above) that word maps \(w\colon G\times G\to G\) are dominant if \(G\) is a semisimple algebraic group over \(k\) and \(w\) is a nontrivial word in the free group \(F_2\). Can one characterize the set of pairs \((w_1,w_2)\) of words in the free group \(F_2\) such that the double word map \(G\times G\to G\times G\); \((a,b)\mapsto(w_1(a,b),w_2(a,b))\) is dominant?