Superstable groups (Q1080416)

This paper is an important contribution to the theory of stable groups. Powerful use is made of Lascar's U-rank, enabling new results as well as generalizations of the finite rank theory to be proved for arbitrary superstable groups. From now on let G denote a (saturated) superstable group (possibly with additional structure). A fundamental observation is that if \(p\in S_ 1(G)\) then the U ranks of p and of Stab(p) are not too different, more precisely \(U(p)+\omega^{\beta}>U(G)\) iff \(U(Stab((p))+\omega^{\beta}>U(G)\). It follows that if \(U(G)=\omega^{\alpha_ 1}k_ 1+...+\omega^{\alpha_ n}k_ n\) and \(U(p)=\omega^{\alpha_ 1}k_ 1+...+\omega^{\alpha_ r}k_ r\) (r\(\leq n)\) then \(U(Stab(p))=U(p)\). This shows that G has normal definable subgroups of each rank \(\omega^{\alpha_ 1}k_ 1+...+\omega^{\alpha_ r}k_ r\) (r\(\leq n)\). So if G is simple then U(G) is a monomial. (A similar argument shows that if G is a field then U(G) is a monomial.) Next, Zilber's ''indecomposability theorem'' (itself a generalization of a fact about algebraic groups) is generalized. Some notation: (\(\infty)\)- definable \(X\subset G\) is \(\alpha\)-indecomposable if for every definable subgroup \(H\subset G:\) \(| X/H| =1\) or \(U(X/H)\geq \omega^{\alpha}\). If X is a subgroup then \(\alpha\)-indecomposability is also called \(\alpha\)-connectedness. Theorem. Let \(U(G)=\omega^{\alpha}k+\beta\) \((\beta <\omega^{\alpha})\). Let \(X_ i(i\in I)\) be a family of (\(\infty)\)-definable \(\alpha\)-indecomposable subsets of G, each containing the identity. Then the subgroup H of G generated by the \(X_ i\) is (\(\infty)\)-definable and \(\alpha\)-connected. A consequence of this theorem is that if G is \(\alpha\)-connected and \(g\in G\) then the normal subgroup of G generated by g is definable (and \(\alpha\)-connected). Another result is that if \(U(G)=\omega^{\alpha}k+\beta\) \((\beta <\omega^{\alpha})\), then G has a definable abelian subgroup H with \(U(H)\geq \omega^{\alpha}\). This is used to give a simple proof of a result of Cherlin that every superstable division ring is commutative. Having read this paper, one is led to ask: what can one obtain in a similar fashion, for just stable groups? The answer, for now, is: essentially nothing, without additional assumptions. There are various such additional assumptions possible. One concerns the ''geometry'' of forking: Assumption A. Th(G) is 1-based. - Under assumption A, it follows that G is essentially just an abelian structure, i.e., an abelian group with distinguished subgroups of \(G^ n\). The key to another kind of assumption is Poizat's study of stable groups with a regular generic type. Assumption B. The generic of G is nonorthogonal to some regular type p. - Under assumption B, and in addition assuming G to be either a simple group of a field, Hrushovski has shown that G is ''p-semiregular'', meaning roughly that the generic of G can be ''coordinatised'' by an independent sequence of realizations of p. The number of coordinates is the p-weight of G and one can use this as a substitute for rank.