Splitting theorems for certain \(PD^3\)-groups (Q798437): Difference between revisions

The article discusses the question whether every \(PD^ 3\)-group (i.e. group of cohomological dimension 3 satisfying Poincaré-duality) is the fundamental group of a closed, aspherical 3-manifold. As evidence for a positive answer algebraic counterparts of theorems of Stallings and of Hempel-Jaco are given: Thm. 5 (p. 204). Let \(\Gamma\) be a \(PD^ 3\)-group which is an extension of a finitely presented by an infinite cyclic group. Then \(\Gamma\) is the fundamental group of a surface fibration over \(S^ 1\). Thm. 7(i). Let \(\Gamma\) be a \(PD^ 3\)-group which is an extension \(1\to K\to\Gamma @>\varphi>>\Gamma_ 2\to 1\) of a finitely presented group \(K\) by a non-abelian fundamental group of a closed surface \(M^ 2\). Then \(K\) is cyclic and there exists a bundle map \(g:N^ 3\to M^ 2\) such that \(N^ 3\simeq K(\Gamma,1)\) and \(g_*=\varphi\). In addition, results concerning the realization of \(PD^ 3\)-pairs as fundamental group pairs of a 3-manifold with boundary are established. The paper concludes with a weak loop theorem for Poincaré-complexes. Thm. 8. Let \((X,Y)\) be a finite 3-dimensional Poincaré-pair. Let \(F\) be some component of \(Y\). Then there are disjoint simple loops \(u_ 1,\dots,u_ k\) on \(F\) such that \(\ker(\pi_ 1F\to\pi_ 1X)\) is the normal closure of \(\{[u_ 1],\dots,[u_ k]\}\). Remarks. (i) The claim of Prop. 6 is correct, but the given argument does not suffice to establish it. (In the situation under consideration the direct limit of copies of \(F_{2g}*_{{\mathbb{Z}}}F_{2g}\) is of the form \((\varinjlim{_{\nu}F_{2g}^{(\nu)}})*_{{\mathbb{Z}}}(\varinjlim{_{\nu }F_{2g}^{(\nu)}})\) and hence has cohomological dimension at most 2.) (ii) The understanding of Thm. 7(ii) and of its proof is rendered difficult by two misstatements: \(\Delta\) is meant to be a non-empty family of \(PD^ 2\)-groups and the dualizing module of \(\Gamma\) is supposed to be isomorphic to \(\ker({\mathbb{Z}}(\Gamma /\Delta)@>\varepsilon>>{\mathbb{Z}})\).