On the geometric boundaries of hyperbolic 4-manifolds (Q1583634)

A classical result of Rokhlin states that every closed orientable 3-manifold \(M^3\) can be identified with the boundary \(\partial W^4\) of an appropriate compact 4-manifold \(W^4\). The paper deals with two geometric incarnations of the above. Firstly, if \(M^3\) is a hyperbolic 3-manifold, is it necessarily the totally geodesic boundary of some compact, complete hyperbolic 4-manifold? Secondly, if \(M^3\) is a flat 3-manifold, is it, up to homeomorphism, the cusp cross-section of some complete finite volume 1-cusped hyperbolic 4-manifold? The answer to both questions turns out to be negative. The authors prove this by showing that in both cases, if \(M\) is bounding in the above geometric sense, then the eta invariant \(\eta(M)\) of \(M\) must be an integer (theorems 1.1 and 1.3). On the other hand, it has been shown that running through the surgeries on a hyperbolic knot in \(S^3\), the \(\eta\)-invariant of the resulting manifolds takes on a dense set of values in \(\mathbb{R}\). Similarly, it can also be shown that there exist flat 3-manifolds with non-integral \(\eta\)-invariants. The authors also show (theorem 1.5) that if \(M(n)\) is an infinite sequence of distinct closed hyperbolic 3-manifolds such that for each \(n\), \(M(n)\) is the totally geodesic boundary of a hyperbolic 4-manifold \(W(n)\), then \(\chi(W(n)) \to\infty\) as \(n\to \infty\) where \(\chi\) is the Euler characteristic. As a consequence, they deduce that there are hyperbolic integral homology 3-spheres that bound rationally acyclic 4-manifolds but cannot bound any rationally acyclic hyperbolic manifold. The proofs of theorems 1.1 and 1.3 use the formula of Atiyah, Patodi and Singer relating the signature of the 4-manifold \(W\) to the integral of the first Pontryagin class and the \(\eta\)-invariant of the boundary \(M\). The main idea is to prove that the term coming from the integral of the Pontryagin class is zero so that \(\text{sign} (W)=-\eta(M)\). The proof of theorem 1.5 involves more geometric arguments. We remark that the examples constructed by the authors above are the first examples of 3-manifolds which are known to be not geometrically cobordant. Also, the methods used by the authors provide obstructions in general for \(4n-1\) manifolds to be the totally geodesic or cusp cross-sections of \(4n\)-manifolds, however, to date, there are no known methods for computing the \(\eta\)-invariants in the higher dimensional case.