Acyclic resolutions for arbitrary groups (Q1424036)

The Edwards-Walsh cell-like resolution theorem [\textit{R. D. Edwards}, Notices Am. Math. Soc. 25, No. 2, A-259 (1978); \textit{J. Walsh}, Shape theory and geometric topology, Proc. Conf., Dubrovnik 1981, Lect. Notes Math. 870, 105--118 (1981; Zbl 0474.55002)] is a turning point of cohomological dimension theory. Namely a compactum \(X\) has integral cohomological dimension, \(\dim_{\mathbb{Z}} X \leq n\), if and only if there exists a compactum \(Z\) with \(\dim Z \leq n\) and a cell-like map \(f:Z \to X\). \textit{A. N. Dranishnikov} [Mat. Sb., N. Ser. 132(174), No. 3, 420--433 (1987; Zbl 0622.55002)] showed that a compactum \(X\) has cohomological dimension \(\dim_{\mathbb{Z}_p} X \leq n\) if and only if there exists a compactum \(Z\) with \(\dim Z \leq n\) and a map \(f:Z \to X\) such that \(\widetilde{\check{H}}^{\ast} (f^{-1}(x);\mathbb{Z}_p) = 0\) for all \(x\in X\). Then it was expected that for an arbitrary abelian group \(G\), compacta with cohomological dimension \(\dim_G \leq n\) could be characterized in terms of \(G\)-acyclic maps. Here a map is \(G\)-acyclic if all of its fibers have trivial reduced Čech cohomology with respect \(G\). However in the case \(G = \mathbb{Q}\) or \(\mathbb{Z}_{p^{\infty}}\) we know examples which cannot be characterized in terms of \(G\)-acyclic maps [see \textit{A. Koyama} and \textit{K. Yokoi}, Tsukuba J. Math. 18, No. 2, 247--282 (1994; Zbl 0851.55001); Topology Appl. 120, No. 1--2, 175--204 (2002; Zbl 1113.55301)]. However \textit{A. N. Dranishnikov} [Topology Appl. 94, No. 1--3, 75--86 (1999; Zbl 0929.55001)] noticed that for a compactum \(X\) with \(\dim_{\mathbb{Q}} X \leq n\) there exists a compactum \(Z\) with \(\dim Z \leq n+1\) and \(\dim_{\mathbb{Q}} Z \leq n\) and a \(\mathbb{Q}\)-acyclic map \(f:Z \to X\), and the reviewer and \textit{K. Yokoi} [loc. cit.)] succeeded to show the existence of such acyclic maps for compacta with cohomological dimension for a large class of groups. Thereby we conjectured if, for an arbitrary abelian group \(G\) and a compactum \(X\) with cohomological dimension, \(\dim_G X \leq n\), there exists a compactum \(Z\) with \(\dim Z \leq n+1\) and \(\dim_G Z \leq n\) and a \(G\)-acyclic map \(f:Z \to X\). The main result of this paper is the affirmative answer to the conjecture. Theorem: Let \(G\) be an abelian group and \(X\) a compactum with \(\dim_G X\leq n\), \(n\geq 2\). Then there is a compactum \(Z\) with \(\dim_G Z\leq n\) and \(\dim Z\leq n+1\) and a \(G\)-acyclic map \(r: Z\to X\) from \(Z\) onto \(X\). An affirmative answer to the conjecture was independently obtained by L. Rubin and P. Shapiro as follows: Let \(K\) be a connected CW-complex and \(X\) be a compactum. If \(\dim X\leq K\) (in the sense of extension theory) \(n\) and the first element of the natural number such that \(G=\pi_n(K)\neq 0\) and \(\pi_{n+1}(K)=0\), then there exists a compactum \(Z\) and a surjective map \(\pi:Z \to X\) such that \(\pi\) is \(G\)-acyclic, \(\dim Z \leq n+1\), and \(\dim Z \leq K\). If additionally, \(\pi_{n+2}(K)=0\), then \(\pi\) can be \(K\)-acyclic. Here \(\pi\) being \(K\)-acyclic means that each map of each fiber \(\pi^{-1}(x)\) to \(K\) is nullhomotopic.