On superstable CSA-groups (Q925848)

A group is CSA (conjugately separated abelian) if all maximal abelian subgroups are conjugately separated, or malnormal (i.e.\ \(H\cap H^g=\{1\}\) for all \(g\notin H\)). Equivalently, the centralizer of every non-trivial element is abelian and self-normalizing. Existentially closed CSA-groups had been proposed by Poizat as candidates for a so-called bad group, i.e.\ a group of finite Morley rank whose proper connected definable subgroups are all nilpotent; this was refuted by the author and \textit{E. Jaligot} [J. Algebra 280, No.\ 2, 772--796 (2004; Zbl 1080.20026)]. This paper continues the study of CSA-groups under stability assumptions. In particular, it is shown that a non-abelian superstable CSA-group has an infinite simple definable subgroup whose proper definable subgroups are all abelian. This easily implies a result of \textit{Y. Mustafin} and \textit{B. Poizat} [J. Algebra 297, No.\ 1, 155--167 (2006; Zbl 1095.03021)], namely that a superstable model of the universal theory of non-abelian free groups is abelian, and also that a superstable torsion-free hyperbolic group is cyclic. The existence of a non-abelian CSA-group of finite Morley rank is equivalent to the existence of a simple bad group whose proper definable subgroups are abelian; the question is raised whether in fact every simple bad group must contain a simple bad CSA-group. Finally, the author shows that an existentially closed involution-free CSA-group is not superstable, either.