Cobordism categories and moduli spaces of odd dimensional manifolds (Q2319884): Difference between revisions

This paper is a contribution to our understanding of moduli spaces of manifolds like \(B\mathrm{Diff}_\partial(W)\), the classifying space of the group of diffeomorphisms of a compact smooth manifold \(W\), restricting to the identity near the boundary. Studying such moduli spaces is a natural companion question to classification of manifolds, and is necessary for understanding bundles of manifolds. In particular, cohomology of \(B\mathrm{Diff}_\partial(W)\) is the ring of characteristic classes for certain smooth manifold bundles, The focus is on the case where the dimension of \(W\) is an odd number \(d = 2n+1\), with particular emphasis on those \(W\) which are \((n-1)\)-connected, \(n\)-parallelizable, and have boundary identified with \(S^{2n}\). For technical reasons it is assumed that \(n \geq 4\) and \(n \neq 7\). For fixed \(n\) such manifolds may be studied ``all at once'' by choosing a sequence of embeddings \[ W_1 \subset W_2 \subset \dots, \] of compact \(d\)-manifolds such that each \(W_i\) is \((n-1)\)-connected, \(n\)-parallelizable, and has boundary \(S^{2n}\), and which is exhaustive in the sense that any \(W\) with these properties admits an embedding into some \(W_i\). The paper under review constructs a map \[ \mathop{\mathrm{hocolim}}\limits_{i \to \infty} B\mathrm{Diff}_\partial(W_i) \longrightarrow \Omega^\infty_0 MT\mathcal{L}_{2n+1}\tag{1} \] inducing an isomorphism in integral homology, where \(\Omega^\infty_0 MT\mathcal{L}_{2n+1}\) denotes the basepoint component of the infinite loop space associated to a very interesting spectrum \(MT\mathcal{L}_{2n+1}\) defined by Hebestreit and Perlmutter. The point here is that the domain of (1) is a \textit{filtered} colimit and that filtered colimits commute with homology. Cohomology of the domain of (1) is a ring of characteristic classes defined for any \(W \subset W_\infty\), compatible under inclusions \(W \subset W' \subset W_\infty\). It follows from the main results of the paper that any non-zero \(\kappa \in H^*(\Omega^\infty_0 MT\mathcal{L}_{2n+1};\mathbb{Q})\) is detected on \(B\mathrm{Diff}_\partial(W)\) for some \(W \subset W_\infty\). At present the spectrum \(\mathcal{L}_{2n+1}\) is not well understood, even rationally, but the homology equivalence (1) is nevertheless an important structural result. It could a priori be expected that the stable homology is captured by a \((2n+1)\)-fold loop space, but the fact that there is an infinite loop structure is very surprising. The situation should be compared with the case of oriented 2-dimensional manifolds, where similar developments took place several decades ago. Here one gets an analogous filtered colimit by choosing \(W_i\) to be a connected oriented 2-manifold of genus \(i\) with one boundary component. In this case, an influential result of \textit{U. Tillmann} [Invent. Math. 130, No. 2, 257--275 (1997; Zbl 0891.55019)] established that the stable homology is the homology of an infinite loop space. Tillmann's paper did not give much information about the underlying spectrum, but that was later determined by \textit{I. Madsen} and \textit{M. Weiss} in their proof of the Mumford conjecture [Ann. Math. (2) 165, No. 3, 843--941 (2007; Zbl 1156.14021)]. Similar results were proved more recently for certain higher dimensional manifolds by \textit{S. Galatius} and \textit{O. Randal-Williams} [Acta Math. 212, No. 2, 257--377 (2014; Zbl 1377.55012)], but their methods are strangely biased towards the dimension being an even number and do not say much about the odd dimensional case. The paper under review gives strong evidence that a similar theory may nevertheless exist in odd dimensions. In comparison to the case of 2-manifolds, the main result seems analogous to Tillmann's theorem: in both cases an infinite loop space structure is constructed, but without determining the homotopy type of the underlying spectrum. It is now a very interesting problem to determine the homotopy type of Hebestreit and Perlmutter's spectrum \(MT\mathcal{L}_{2n+1}\): this would be an analogue of Madsen and Weiss' theorem. The strategy of proof is explained lucidly in the introduction. After summarizing the even-dimensional case, the authors explain which methods may be carried over from that case, and where new ideas are required. As in the even dimensional case, cobordism categories form the central organizing tool, and the method of parametrized surgery is used to compare subcategories of cobordism categories defined by connectivity requirements. An important innovation by Hebestreit and Perlmutter is a category \[ \mathrm{Cob}_{2n+1}^{\mathcal{L}}, \] a cobordism category whose objects are \(2n\)-dimensional closed manifolds \(M\) whose middle homology group \(H_n(M)\) is furthermore \textit{equipped} with the algebraic data which would arise from a null bordism of \(M\), namely a Lagrangian. This category is defined in Section 4 of the paper, and for a while the Lagrangians are mostly carried along. They are needed in Section 9 of the paper, where they are used to overcome obstructions to performing surgery on the \(2n\)-dimensional objects in the middle dimension. The presence of algebraic cobordism data in \(\mathrm{Cob}_{2n+1}^\mathcal{L}\) suggests a connection to algebraic \(L\)-theory in the sense of Wall and Ranicki. The introduction of the paper contains an intriguing suggestion about a more precise relation to the algebraic cobordism categories studied recently by \textit{B. Calmès} et. al. [``Hermitian K-theory for stable \(\infty\)-categories I--III''. Preprints, \url{arXiv:2009.07223}, \url{arXiv:2009.07224}, \url{arXiv:2009.07225}] in their work on hermitian \(K\)-theory.