An upper bound for the pseudoisotopy stable range (Q2402836)

For a smooth manifold \(M\) one defines the space \(P(M)\) of pseudoisotopies of \(M\) as those diffeomorphisms of \(M\times I\) that fix \(M\times \{0\}\). By crossing with the interval, one obtains a map \[ \lambda_M : P(M) \longrightarrow P(M\times I). \] The \textit{pseudoisotopy stable range} is then the function \[ \phi(n) = \max \{ k \in \mathbb{N} : \lambda_M \text{ is \(k\)-connected for all \(n\)-dimensional manifolds } M \}. \] So far, the main theorem about the pseudoisotopy stable range is a lower bound for \(\phi(n)\) proven by \textit{K. Igusa} [\(K\)-Theory 2, No. 1--2, 1--355 (1988; Zbl 0691.57011)]. It says that \[ \phi(n) \geq \min \{ \tfrac{n-7}{2}, \tfrac{n-4}{3} \}. \] The author of this paper contributes to the knowledge of the pseudoisotopy stable range by providing also an upper bound: The main result of this paper states that for \(2n\) at least \(6\) one has \(\phi(2n) \leq 2n-2\). The strategy to prove this result is as follows. One considers the fibre sequence \[ \frac{\widetilde{\mathrm{Diff}}_\partial(M)}{\mathrm{Diff}_\partial(M)} \longrightarrow \mathrm{BDiff}_\partial(M) \longrightarrow\mathrm B\widetilde{\mathrm{Diff}}_\partial(M) \] and studies the associated Serre spectral sequence in rational cohomology. An important input into this is the following claim. One has \[ \mathrm{ker} \Big( H^{2n}(\mathrm B\widetilde{ \mathrm{Diff}}_\partial(W_{g,1});\mathbb{Q}) \to H^{2n}(\mathrm{BDiff}_\partial(W_{g,1});\mathbb{Q}) \Big) \neq 0 \] For \(g\) very large compared to \(2n\) this follows from explicitly calculating both groups. The right hand side has been calculated by the author and \textit{S. Galatius} [J. Am. Math. Soc. 31, No. 1, 215--264 (2018; Zbl 1395.57044); Acta. Math. 212, No. 2, 257--377 (2014; Zbl 1377.55012)]. The left hand side has been calculated by \textit{A. Berglund} and \textit{I. Madsen} [Pure Appl. Math. Q. 9, No. 1, 1--48 (2013; Zbl 1295.57038); ``Rational homotopy theory of automorphisms of highly connected manifolds'', Preprint, \url{arXiv:1401.4096}]. The author of this paper gives a very nice proof of the above claim (for any \(g \geq 1\)) by providing an explicit class in the kernel. He generalises results of \textit{J. Ebert} and \textit{O. Randal-Williams} [Algebr. Geom. Topol. 14, No. 2, 1181--1204 (2014; Zbl 1291.57018)] and shows that for each smooth oriented block bundle \(E \to |K|\) with fibre a closed \(d\)-dimensional manifold \(M\), there exists a cohomology class \(e \in H^d(E;\mathbb{Z})\) which takes the role of the Euler class of the vertical tangent bundle of a smooth fibre bundle with fibre \(M\). This implies the existence of tautological classes \[ \kappa_{e^2}, \kappa_{p_n} \in H^{2n}(\mathrm B\widetilde{ \mathrm{Diff}}_\partial(W_{g,1});\mathbb{Q}) \] which agree upon pulling back to \(\mathrm{BDiff}_\partial(W_{g,1})\). It is then shown that they do not agree in the cohomology of \(\mathrm B\widetilde{\mathrm{Diff}}_\partial(W_{g,1})\) -- this occupies Section 3 of this paper. This proves in particular that in general, a smooth block bundle does not have a canonical \textit{unstable} vertical tangent bundle. The construction of tautological classes has been pushed further also to topological block bundles in a more recent paper, see [\textit{F. Hebestreit} et al., ``A vanishing theorem for tautological classes of aspherical manifolds'', Preprint, \url{arXiv:1705.06232}]. On the other hand, the fibre of the above sequence appears in work of \textit{M. Weiss} and \textit{B. Williams} [\(K\)-Theory 1, No. 6, 575--626 (1988; Zbl 0664.57012)], where for any compact \(n\)-manifold \(M\), a comparison map \[ \frac{\widetilde{\mathrm{Diff}}_\partial(M)}{\mathrm{Diff}_\partial(M)} \longrightarrow \Omega^\infty\big( \Omega(\mathbf{Wh}^{\mathrm{Diff}}(M))_{hC_2}\big) \] is constructed and shown to be \(\phi(n)+1\)-connected. Using results of \textit{W. C. Hsiang} and \textit{R. E. Staffeldt} [Invent. Math. 68, 227--239 (1982; Zbl 0505.57011)] the author calculates the right hand side's rational cohomology for \(M = W_{g,1}\) in degrees smaller or equal to \(2n-1\). Assuming that \(\phi(2n) = 2n-1\), and the Weiss-Williams comparison map, this can only be compatible with the non-triviality of the class \(\kappa_{e^2} - \kappa_{p_n}\) if a certain differential in the Serre spectral sequence associated to the above fibre sequence is non-zero. This is shown not to be the case and hence \(\phi(2n) \leq 2n-2\). In the very last section of this paper, it is shown how similar techniques can be used to show that the space \[ \frac{\mathrm{Top}(2n)}{\mathrm{O}(2n)} \] is rationally \((4n-5)\)-connected, provided \(n\) is at least 3. On the other hand, by results of \textit{M. Weiss} [``Dalian notes on Pontryagin classes'', Preprint, \url{arXiv:1507.00153}] one knows that this space is \textit{not} rationally \((4n-1)\)-connected.