À propos d'équations génériques (Q4032640)

It is proved that a stable soluble group which satisfies \(x^ n=1\) generically has finite exponent dividing some power of \(n\). This partially answers a question by Poizat, who observed that this property holds for algebraic groups, even without the solubility condition. By a Theorem of Baudisch such a group is nilpotent-by-finite; here an easier proof of that fact is given. If just one generic element has exponent 3, then Poizat has shown that all \(G^ 0\)-conjugacy classes are Abelian. Now it is proved that a stable group with nilpotent conjugacy classes is itself nilpotent (a corresponding result holds for soluble instead of nilpotent), hence a group with generic element of order 3 is again nilpotent-by-finite. (Meanwhile this result has been extended to the case of arbitrary prime exponent, but only for soluble groups [the author, ``Commutator conditions and splitting automorphisms for stable groups'', Arch. Math. Logic 32, 223-228 (1993)].) A second part deals with centralisers of involutions in stable periodic groups. It is shown that if the generic type is foreign to the centraliser of some involution \(i\), then \(i\) inverts an Abelian subgroup of finite index. This generalises the previously known case when \(C(i)\) is finite. A similar result holds for 2-elements.