On the summand theorem for A-groups (Q752857)

The paper is concerned with the question whether a direct summand of an \(A\)-group is again an \(A\)-group. \(A\)-groups are abelian \(p\)-groups defined as follows. Let \(\mu\) be a limit ordinal not cofinal with \(\omega\), the least countable ordinal. A group \(H\) is a \(\mu\)-elementary \(A\)-group and \((H,G)\) an \(A_{\mu}\)-pair if \(G\) is a totally projective group of length at most \(\mu\) such that (1) \(H\) is an isotype subgroup of \(G\), (2) \(p^{\lambda}(G/H)=(p^{\lambda}G+H)/H\) for all \(\lambda <\mu\), and (3) \(G/H\) is totally projective. A group \(H\) is an \(A\)-group if \(H\) is the direct sum of \(\mu\)-elementary \(A\)-groups for various limit ordinals \(\mu\) not cofinal with \(\omega\). Certain \(A\)-invariants \(f_ H(\alpha,\beta)\) were defined by \textit{P. Hill} [Trans. Am. Math. Soc. 288, 505-525 (1985; Zbl 0573.20053)], and it was shown that two \(A\)-groups are isomorphic if and only if they have the same \(A\)-invariants. The large cofinality dimension of the \(A\)-group \(H\) is defined as \[ LCD(H)=| \{\beta >0| \quad f_ H(\alpha,\beta)\neq 0\text{ for some } \alpha \leq \infty \}|. \] The main results of the paper under review are: Theorem 3.4. If \(H\) is an \(A\)-group with finite large cofinality dimension, then every summand \(B\) of \(H\) is also an \(A\)-group with \(LCD(B)\leq LCD(H)\). Theorem 3.5. If \((H,G)\) is an \(A_{\mu}\)-pair (i.e. \(LCD(H)=1)\) and \(H=B\oplus C\) then \(G=G_ 1\oplus G_ 2\) such that \((B,G_ 1)\) and \((C,G_ 2)\) are \(A_{\mu}\)-pairs. An example is given showing that the lifting of decompositions as in 3.5 does not always work for \(LCD(H)=2\).