Homotopy invariance of 4-manifold decompositions: connected sums (Q714747): Difference between revisions

If \(f:M\rightarrow N_{1}\#N_{2}\) is a homotopy equivalence between closed manifolds, with the range a nontrivial connected sum, it is natural to speculate that \(M\) can be split into a connected sum \(M^{n}=M_{1}\#M_{2}\) with \(M_{1}\) homotopy equivalent to \(N_{1}\) and \(M_{2}\) homotopy equivalent to \(N_{2}\). Unfortunately, that is not always the case. In this paper, the author obtains an ``approximate'' version of this hoped-for splitting for certain classes of 4-manifolds. In particular, he shows that, if \(\pi_{1}\left( N_{1}\right) \) and \(\pi_{1}\left(N_{2}\right) \) each fall into the class of \textit{good }groups (A group \(G\) is ``good'' if Freedman's famous disk embedding lemma holds in 4-manifolds with fundamental group \(G\).), then \(M\) is \(h\)-cobordant to a 4-manifold \(M^{\prime}=M_{1}^{\prime}\#M_{2}^{\prime}\), with \(f\) \(h\)-cobordant to a connected sum of homotopy equivalences \(f_{1}^{\prime }\#f_{2}^{\prime}\), where \(f_{i}^{\prime}:M_{i}^{\prime}\rightarrow N_{i}\). For comparable results in higher dimensions, see \textit{S. E. Cappell} [Invent. Math. 33, 69--170 (1976; Zbl 0348.57017)].