Desingularizing homology manifolds (Q2464822): Difference between revisions

The main theorem of this paper is: Every compact homology \(n\)-manifold (\(n\geq 6\)) is the cell-like image of a compact homology \(n\)-manifold \(Y\) with the disjoint disks property (DDP). Here, homology \(n\)-manifolds are ANRs with the local homology of \(\mathbb{R}^n\) and DDP is the very minimal general position property expected of \(2\)-dimensional objects in spaces of dimension \(5\) or greater. To put this result in context, recall that a homology \(n\)-manifold \(X\) is resolvable if \(X\) is the cell-like image of a manifold. The present authors [Ann. Math. (2) 143, No. 3, 435--467 (1996; Zbl 0867.57016)] previously proved that homology manifolds need not have resolutions, thereby providing examples with nontrivial Quinn obstruction [see \textit{F. Quinn}, Mich. Math. J. 34, 285--291 (1987; Zbl 0652.57011)]. The DDP is important because \textit{R. D. Edwards} [Proc. int. Congr. Math., Helsinki 1978, Vol. 1, 111--127 (1980; Zbl 0428.57004)] proved that a resolvable homology \(n\)-manifold \(X\) (\(n\geq 6\)) has the DDP if and only if \(X\) is a manifold. Thus, DDP homology manifolds are very close to being manifolds. As part of the proof of the main theorem, the authors establish the DDP for the homology manifold \(Y\) by constructing \(Y\) so that it has the \(UV^1\)-approximation property. This property is hard to describe briefly, but it is a version of Chapman's approximation property for approximate fibrations between manifolds [see \textit{T. A. Chapman}, Mem. Am. Math. Soc. 251, 64 p. (1981; Zbl 0478.57009)]. The authors prove that their \(UV^1\)-approximation property implies the DDP and the question arises whether the two properties are equivalent for homology manifolds in general. A second, more technical, main result is also related to Chapman's theorem. To partially describe the new result, suppose \(p: Y\to Z\) is a \(UV^1\)-map between ENRs so that inverse images of open sets of \(Z\) are \(n\)-dimensional Poincaré spaces. The result says that there exists \(\varepsilon_0 > 0\) so that if \(X\) is a DDP homology \(n\)-manifold (\(n\geq 6\)) and \(f: X\to Y\) is a \(p^{-1}(\varepsilon_0)\)-homotopy equivalence, then \(f\) may be replaced by a map that is a \(p^{-1}(\varepsilon)\)-homotopy equivalence for every \(\varepsilon >0\). Several of the arguments in this paper represent refinements and simplifications of constructions in their previous paper. Thus, a reader who is primarily interested in nonresolvable homology manifolds will do well to study this paper.