Moduli space of non-negative sectional or positive Ricci curvature metrics on sphere bundles over spheres and their quotients (Q2161431)

Two recent results [\textit{M. J. Goodman}, Ann. Global Anal. Geom. 57, No. 2, 305--320 (2020; Zbl 1447.58012); \textit{A. Dessai}, ``On the moduli space of nonnegatively curved metrics on Milnor spheres'', Preprint, \url{arXiv:1712.08821}] concern metrics of non-negative sectional or positive Ricci curvature on total spaces of linear \(S^3 \to E \to S^4\) bundles, respectively: provided these total spaces are rational homology spheres, the corresponding moduli spaces of such metrics have infinitely many path components. Milnor's classical homology \(7\)-spheres are examples of such manifolds. This paper proves analogues of these results for total spaces of linear \(S^7\)-bundles over \(S^8\), including Shimada's homology \(15\)-spheres [\textit{N. Shimada}, Nagoya Math. J. 12, 59--69 (1957; Zbl 0145.20303)]. More precisely, it is proven that the space of positive Ricci curvature metrics on such a manifold has infinitely many components, provided the manifold is a rational homology sphere. However, it is not known whether these manifolds always support metrics of non-negative sectional curvature. The antipodal action on the fibers of any sphere bundle gives rise to an involution on the total space. For the examples referred to above, the resulting quotients are called Milnor/Shimada projective spaces. The paper recalls that there are 16 different diffeomorphism types of Milnor projective spaces. Then it is proven that the moduli spaces of metrics of positive sectional or positive Ricci curvature on each of them has infinitely many path components. It is also shown that there are at least 4096 different diffeomorphism types of Shimada projective spaces for which the moduli spaces of positive Ricci curvature metrics has infinitely many path components. The number of different diffeomorphism types of all Shimada projective spaces seems to be unknown, but the paper proves that there are only finitely many. The tools for the diffeomorphism classification are very classical. In particular, the Browder-Livesay invariant plays an important role, and using it, the Eells-Kuiper invariant [\textit{J. Eells jun.} and \textit{N. H. Kuiper}, Ann. Mat. Pura Appl. (4) 60, 93--110 (1962; Zbl 0119.18704)] is computed for the Shimada projective spaces. For Milnor projective spaces this was already done in [\textit{Z. Tang} and \textit{W. Zhang}, Adv. Math. 254, 41--48 (2014; Zbl 1290.51005)]. The bound of at least 4096 different diffeomorphism types alluded to above is obtained by means of a computer program, as the author did not find a way to solve the resulting arithmetic.