Resolutions of homology manifolds, and the topological characterization of manifolds (Q760691)

The following important result is presented: Every generalized n-manifold \((=ENR\) \({\mathbb{Z}}\)-homology n-manifold) X, \(n\geq 5\), admits a resolution \((=X\) is the cell-like image of an n-manifold). (Recently the author has extended this to dimension four [J. Differ. Geom. 17, 503-521 (1982; Zbl 0533.57009)].) The methods used in the proof are ''\(\epsilon\)-versions'' of algebraic topology, homotopy theory, algebraic K-theory, and surgery. It suffices to show that every point of \(X\times {\mathbb{R}}^ m\) \((n+m=4k)\) has a neighborhood U with a resolution \(f: M\to U\). Now, (M,f) is obtained by applying \textit{F. Quinn'}s end theorem [Ann. Math., II. Ser. 110, 275-331 (1979; Zbl 0394.57022)] to a sequence \(\Sigma =\{f_ n: M_ n\to X\}\) of proper \(2^{-n}\)-homotopy equivalences (where \(M_ n\) is an n-manifold and \(M_ n\approx M_{n+1})\). The surgery obstructions to the existence of \(\Sigma\) are \(\epsilon\)-versions of the ordinary ones. They are all reduced to a single integer obstruction \(\sigma\), defined by transversality on a manifold degree 1 normal map with the same form as \(f: M\to X.\) By transferring the problem to a torus, \(\sigma\) is recognized as a part of the ordinary surgery obstruction \(L({\mathbb{Z}}[{\mathbb{Z}}^ n])\) hence it depends only on the form (and not on the particular map) so it vanishes. Together with the well-known approximation theorem of \textit{R. D. Edwards} [Proc. Int. Congr. Math., Helsinki 1978, Vol. 1, 111-127 (1980; Zbl 0428.57004)], Quinn's resolution theorem yields the proof of an important conjecture of \textit{J. W. Cannon} [Bull. Am. Math. Soc. 84, 832-866 (1978; Zbl 0418.57005)] - a characterization of higher dimensional topological manifolds: A space X is an n-manifold (n\(\geq 5)\) if and only if X is a generalized n-manifold satisfying the disjoint disks property \((=maps\) f,g: \(B^ 2\to X\) can be approximated by maps \(f^*,g^*: B^ 2\to X\) such that \(f^*(B^ 2)\cap g^*(B^ 2)=\emptyset)\). As a corollary, the double suspension of every closed ENR \({\mathbb{Z}}\)-homology n-manifold with the \({\mathbb{Z}}\)-homology of the n-sphere is the \((n+2)\)-sphere. (This long outstanding conjecture was first proved by \textit{J. W. Cannon} [Ann. Math., II. Ser. 110, 83-112 (1979; Zbl 0424.57007)] before even Edward's theorem was known.)