The signature and cusp geometry of hyperbolic knots (Q6620573)

For a hyperbolic knot \(K\), the boundary of the maximal cusp is a Euclidean torus, and so it is isometric to \(\mathbb{C} / \Lambda\) for a lattice \(\Lambda\) in \(\mathbb{C}\). The meridian and longitude of \(K\) give generators \(\mu\) and \(\lambda\) for \(\Lambda\). In this paper, the authors introduce the natural slope \(\mathrm{slope}(K) = \mathrm{Re}(\lambda / \mu)\) and investigate its relations to other invariants.\N\NThe main theorem gives a relation between the natural slope \(\mathrm{slope}(K)\) and the signature \(\sigma (K)\) as follows. There exists a constant \(c_{1}\) such that any hyperbolic knot \(K\) satisfies\N\[\N| 2 \sigma (K) - \mathrm{slope}(K) | \leq c_{1} \mathrm{vol}(K) \mathrm{inj}(K)^{-3},\N\]\Nwhere \(\mathrm{vol}(K)\) is the hyperbolic volume of \(K\), and \(\mathrm{inj}(K)\) is the injectivity radius of \(S^{3} \setminus K\) defined as the infimum of the injectivity radii of the points outside the maximal cusp. It is conjectured that the constant \(c_{1}\) can be \(0.3\).\N\NThe authors show that the factor \(\mathrm{inj}(K)^{-3}\) in the main theorem cannot simply be dropped. Nevertheless, \(\mathrm{inj}(K)\) is typically not too small. Small injectivity radii are realized by highly twisted knots. The authors also refine the main theorem with a correction term for a twisting parameter that does not depend on the injectivity radius.\N\NThe main theorem shed light on a new connection between geometric and algebraic invariants of a knot. It has applications in low-dimensional topology. For example, the natural slope relates to exceptional Dehn surgery. The signature gives a lower bound of the 4-ball genus. Further progress can also be expected.\N\NTo prove the main theorem, the authors construct a spanning surface of a knot such that its genus is bounded above and its boundary slope is close to the natural slope. This construction is based on an appropriate triangulation of the complement. The signature is calculated using the Goeritz matrix and the boundary slope for a spanning surface.\N\NA notable feature of the main theorem is that it was obtained through suggestion by machine learning to compare various knot invariants. While it was initially conjectured that\N\[\N| 2 \sigma (K) - \mathrm{slope}(K) | \leq c_{1} \mathrm{vol}(K) + c_{2},\N\]\Nit was revised as in the main theorem. The background was explained in [\textit{A. Davies} et al., Nature, London 600, No. 7887, 70--74 (2021; Zbl 1505.57001)]. The same project also yielded a result for Kazhdan-Lusztig polynomials [\textit{C. Blundell} et al., Represent. Theory 26, 1145--1191 (2022; Zbl 1509.20057)]. This paper exhibits a new possibility for mathematical discoveries.