\(SU(2)\)-cyclic surgeries and the pillowcase (Q2152781)

This paper studies knots in the 3-sphere with infinitely many \(\mathrm{SU}(2)\)-cyclic surgeries and mainly characterizes slopes admitting such surgeries. A Dehn surgery is said to be \(\mathrm{SU}(2)\)-cyclic if the resulting manifold \(Y\) satisfies that all representations \(\pi_1(Y) \to \mathrm{SU}(2)\) have cyclic image [\textit{J. Lin}, Int. Math. Res. Not. 2016, No. 19, 6018--6033 (2016; Zbl 1404.57015)]. A slope giving such a surgery is called an \(\mathrm{SU}(2)\)-cyclic slope. In this paper, a non-trivial knot is said to be \(\mathrm{SU}(2)\)-averse if it admits infinitely many \(\mathrm{SU}(2)\)-cyclic slopes. In the following, let \(K\) be an \(\mathrm{SU}(2)\)-averse knot. Then it is known that the set of its \(\mathrm{SU}(2)\)-cyclic slopes has a unique limit point, called the limit slope \(r(K)\), in \(\mathbb{R}\cup\{ \infty \}\). The first main theorem in this paper states that \(r(K)\) is a rational number with \(|r(K)|>2\) and is a boundary slope for \(K\). Moreover, \(K\) is prime, and \(s\)-surgery yields an instanton L-space for all rational numbers \(s\) with \(s\ge \lceil r(K) \rceil - 1\) if \(r(K)>0\), or \(s\le \lfloor r(K) \rfloor + 1\) if \(r(K)<0\). It follows from this together with a recent result in a preprint by \textit{J. A. Baldwin} and \textit{S. Sivek} [``Instantons and L-space surgeries'', Preprint, \url{arXiv:1910.13374}] that \(K\) is also fibered, either \(K\) or its mirror image is strongly quasi-positive, and \(|r(K)|>2g(K)-1\). The second main theorem states that all \(\mathrm{SU}(2)\)-cyclic slopes \(m/n\) satisfy \(| m/n-r(K) | \le |r(K)|/n\), and there is an integer \(N\ge 1\) such that \(r(K)+1/(kn)\) is an \(\mathrm{SU}(2)\)-cyclic slope for all but finitely many \(k\in \mathbb{Z}\). Additionally, if \(K\) is small, then \(|r(K)|\ge 6\), and \(|r(K)|\ge 1+\sum_{j=-d}^{d}|a_j|\), where \(a_j\) is a coefficient of the Alexander polynomial \(\Delta_K(t)=\sum_{j=-d}^{d} a_j t^j\). The authors also propose a conjecture that any \(\mathrm{SU}(2)\)-averse knot is a torus knot. As supporting evidence, they verify the conjecture for algebraic knots, all alternating Montesinos knots with at most three rational tangles and all knots up to 11 crossings.