Two-solvable and two-bipolar knots with large four-genera (Q2028609)

There are two potential approximations for the sliceness of knots. The first is in terms of the slice genus, i.e. the smaller the slice genus, the closer a knot is to being slice. The second is in terms of the 0-surgery manifold. If a knot is slice, the result of 0-framed Dehn surgery on the 3-sphere along the knot is the boundary of a 4-manifold, namely the slice disc exterior, with certain nice properties. The solvable filtration of the knot concordance group due to \textit{T. D. Cochran} et al. [Ann. Math. (2) 157, No. 2, 433--519 (2003; Zbl 1044.57001)] gives a systematic way of approximating sliceness in terms of the 4-manifolds bounded by 0-surgery manifolds for knots, where the deeper a knot is within the filtration, the farther it is from being slice. The bipolar filtration of \textit{T. D. Cochran} et al. [Geom. Topol. 17, No. 4, 2103--2162 (2013; Zbl 1282.57006)] follows a similar principle, but has the added advantage that it may successfully distinguish between smooth concordance classes of topologically slice knots. The paper under review considers the question of how the two above notions of approximations to sliceness interact with each other. In particular, it remains an open question whether there are knots of arbitrarily large slice genera deep within the solvable or bipolar filtrations of the knot concordance group. Knots with large slice genera in the zeroth and first terms were found by the first author in his previous work [Algebr. Geom. Topol. 8, No. 2, 885--909 (2008; Zbl 1162.57016)]. This paper continues the study of this conjecture by constructing the first examples of knots with high slice genera within the second term of the solvable (and bipolar) filtrations. The examples are necessarily algebraically slice, have vanishing Casson-Gordon obstructions, and cannot be detected using the known smooth slice genus bounds from gauge theory and Floer homology. As is common for the solvable and bipolar filtrations, the examples are satellite knots, with ribbon patterns. The novel lower bound for the slice genus comes from \(L^{(2)}\)-signature defects associated to meta-metabelian representations of the fundamental group of the knot complement. Here a group is said to be meta-metabelian if the third derived subgroup is trivial.