Casson towers and slice links (Q328677)

Freedman's celebrated work on topological 4-manifolds has been applied by many mathematicians, but on most occasions it was used as a ``blackbox'', in the sense that one takes the relevant theorem from Freedman and one does not look too much at the famously intricate arguments in Freedman's papers. This approach was used for example in [\textit{M. H. Freedman}, Invent. Math. 68, 195--226 (1982; Zbl 0504.57016); \textit{S. Friedl} and \textit{P. Teichner}, Geom. Topol. 9, 2129--2158 (2005; Zbl 1120.57004)] and to a somewhat lesser degree in [\textit{T. D. Cochran} et al., Comment. Math. Helv. 84, No. 3, 617--638 (2009; Zbl 1179.57008)] to show that certain knots are slice. Here recall that a knot \(K\subset S^3\) is called slice if there exists a locally flat disk \(D\) in \(D^4\) such that \(\partial D=K\). In this paper the authors dare to look under the hood of Freedman's proofs and they obtain significant new results. The key definition is the notion of a Casson tower which is a certain type of 4-manifold \(T\) with a framed embedded circle \(C(T)\) in its boundary. In the following we give a shortened definition of a Casson tower. A Casson tower of height 1 is a thickened disk \(D^2\times D^2\) with \(C(T) :=\partial D^2\times \{0\}\) and with some number of self-plumbings performed in the interior of \(D^2 \times \{0\}\). For \(n \geq 2\), a Casson tower of height \(n\) is constructed inductively by taking a height one Casson tower and, for each ``double point loop'', identifying a neighbourhood of the double point loop with the attaching region of some Casson tower of height \(n - 1\). A Casson tower of infinite height is called a \textit{Casson handle}. In [J. Differ. Geom. 17, 357--453 (1982; Zbl 0528.57011)] \textit{M. H. Freedman} showed that a Casson handle is homeomorphic to an open 2-handle, and consequently it contains a flat embedded disk with framed boundary \(C\). A key ingredient of the proof was Freedman's reimbedding theorem which says that a height 6 Casson tower contains within it a height 7 Casson tower. Iterating this, it follows that a height 6 Casson tower \(T\) contains a Casson handle, and consequently contains a flat embedded disc with framed boundary \(C(T)\). \textit{R. E. Gompf} and \textit{S. Singh} [Contemp. Math. 35, 277--309 (1984; Zbl 0575.57006)] showed that the same conclusion already holds for height 5 Casson towers. One of the main results of the paper says that the same statement holds even for height 4 Casson towers. Furthermore the authors also prove similar statements about Casson towers of height 3 and 2, albeit under increasingly strong extra hypotheses. The authors then use these results to give a plethora of knots which they can show are slice, but which were inaccessible with the previous methods. In particular they give an example of a slice knot that is conceivably a counter example to the topological version of the slice-ribbon conjecture. More precisely, they give a knot \(K\) that is slice where it is not clear whether it bounds a slice disk \(D\) such that the induced map \(\pi_1(S^3\setminus K)\to \pi_1(D^4\setminus D)\) is an epimorphism. To the authors' and to the reviewer's best knowledge this is the first potential candidate for a counter example to the slice-ribbon conjecture.