Half-integral finite surgeries on knots in \(S^3\) (Q318708)

The starting point of this paper is a conjecture by \textit{S. Boyer} and \textit{X. Zhang} [J. Am. Math. Soc. 9, No. 4, 1005--1050 (1996; Zbl 0936.57010)]: If the \(\frac{p}{q}\)-surgery \(M=S^3_K(\frac{p}{q})\) of a hyperbolic knot \(K\) has finite fundamental group, then \(\frac{p}{q}\) is an integer. What they proved is that \(\frac{p}{q}\) is integral or half-integral. In view of the Geometrization Theorem, \(M\) is one of the well-known five spherical space forms. To prove Boyer and Zhang's conjecture, the tetrahedral and the icosahedral space forms would have to be ruled out; for only these result from half-integral surgery. Therefore, the authors look for candidates of half-integral knot surgeries such that the knot has the ``right'' knot Floer homology. Their main result is a list -- knot type / surgery slope \(\frac{p}{2}\) / resulting \(3\)-manifold -- with \(10\) entries such that, for any knot \(K\) and half-integer \(\frac{p}{2}\) as above, there is a case in the list, matching in the following sense: the corresponding knot type has the same knot Floer homology as \(K\) and the resulting manifold is homeomorphic to \(M=S^3_K(\frac{p}{2})\). The knots occurring in the list are the torus knot \(T(5,2)\) or cables of \(T(3,2)\) or \(T(5,2)\); the resulting manifolds are all surgeries on the righthand trefoil. The proof of the main result uses the correction term of Ozsváth and Szabó's Heegaard Floer homology. Large \(p\) cannot occur due to a lengthy calculation; the finitely many remaining \(p\) are taken care of under support of a computer program.