Singly cogenerated annihilator classes (Q1821873)

''Group'' means abelian group in this review. An annihilator class is a class of groups closed under the formation of products and subgroups. If \({\mathcal A}\) is any class of groups and G some group, let \(R_{{\mathcal A}}G=\cap \{\ker f:\) \(f\in Hom(G,X)\), \(X\in {\mathcal A}\}\). Then \(R_{{\mathcal A}}\) is a radical, and \(\{\) \(G: R_{{\mathcal A}}G=0\}\) is the smallest annihilator class containing \({\mathcal A}\). If \({\mathcal A}\) is the one-element class \(\{\) \(X\}\), write \(R_ X\) for \(R_{{\mathcal A}}\), and \(\{\) \(G: R_ X=0\}\) is called singly cogenerated by X. If R is any radical, i.e. a subfunctor of the identity with \(R(G/RG)=0\) then \(\{\) \(G: RG=0\}\) is an annihilator class, and the question arises whether \(R=R_ X\) for some group X, or, equivalently, whether \(\{\) \(G: RG=0\}\) is singly cogenerated. Clearly the class of torsionfree groups is singly cogenerated by Q, \(p^ n=R_{Z(p^ n)}\), \(p^{\omega}=R_{\prod Z(p^ n)}=R_{\oplus Z(p^ n)}\), but it is shown (Corollary 2.2) that \(p^{\alpha}\) is not of the form \(R_ X\) for any \(\alpha \geq \omega +1\). The proof is based on the existence of a noteworthy group (Theorem 2.1): given any infinite cardinal \(\lambda\) there is a p-group H of cardinality \(\lambda^{\aleph_ 0}\) such that \(p^{\omega}H\cong Z(p)\) and whenever A is a group of cardinality \(<\lambda\) and \(\phi\in Hom(H,A)\) then \(\phi (p^{\omega}H)\) is contained in the divisible part of A. Every radical R gives rise to an idempotent radical \(R^{\infty}\) as follows. Given a group G let \(R^ 1G=RG\), \(R^{\alpha +1}G=R(R^{\alpha}G)\) and \(R^{\alpha}G=\cap_{\beta <\alpha}R^{\beta}G\) if \(\alpha\) is a limit ordinal. Then \(R^{\mu +1}G=R^{\mu}G\) for some \(\mu\) and \(R^{\infty}G=R^{\mu}G\) by definition. If \({\mathcal A}\) is a class of groups then the annihilator class of \(R^{\infty}_{{\mathcal A}}\) is the smallest class containing \({\mathcal A}\) and closed under the formation of products, subgroups and extensions. The annihilator class of \(p^{\infty}=R^{\infty}_{Z(p)}\) is not singly cogenerated [\textit{T. H. Fay}, \textit{E. P. Oxford}, \textit{G. L. Walls}, Lect. Notes Math. 1006, 671-684 (1983; Zbl 0522.20039)] and the annihilator class of all \(\aleph_ 1\)-free groups is not singly cogenerated [\textit{M. Dugas} and \textit{R. Göbel}, Pac. J. Math. 118, 79- 104 (1985; Zbl 0578.20050)]. It is shown here (Corollary 3.2) that for any cotorsion-free group G (i.e. \(Hom(J_ p,G)=0\) for each of the p-adic groups \(J_ p)\) the radical \(R^{\infty}_ G\) is not of the form \(R_ X\). Again the proof is based on a clever construction of a test group. Moreover (Theorem 3.5) the idempotent radicals of the form \(R_ X\) are exactly the radicals \(R_ Y\) where \(Y=Q\), \(\prod_{p\in \pi}Z(p^{\infty})\) or \(\prod_{p\in \pi}J_ p\) for some non-empty set \(\pi\) of primes. It is also shown (Proposition 3.6) that there exists a non-cotorsion-free group G with \(R^{\infty}_ G\) not of the form \(R_ X\) for any X. There are a number of further results.