\(L\)-space knots have no essential Conway spheres (Q2103502)

A knot is called an \(L\)-space knot if it admits a non-trivial Dehn surgery yielding an \(L\)-space. Any \(L\)-space knot is known to be prime, and is expected to admit no essential Conway sphere. Here, a Conway sphere is the \(2\)-sphere which intersects the knot in four points transversely, and it is essential if the corresponding \(4\)-punctured sphere in the knot exterior is incompressible. The main result of the paper under review is to solve the above conjecture affirmatively. As an immediate corollary, Conway mutation preserves \(L\)-space knots. Also, if a knot \(K\) admits an essential Conway sphere, then any non-trivial surgery on \(K\) yields a \(3\)-manifold with infinite fundamental group. This recovers a result of \textit{Y.-Q. Wu} [J. Differ. Geom. 43, No. 1, 171--197 (1996; Zbl 0851.57018)]. \textit{K. L. Baker} and \textit{A. H. Moore} [J. Math. Soc. Japan 70, No. 1, 95--110 (2018; Zbl 1390.57002)] conjectured that an \(L\)-space knot is \(n\)-string prime. That is, it does not admit an essential \(n\)-string tangle decomposition. In this terminology, the conjecture is true for \(n\le 2\) at present. For satellite knots, it is also proved to hold for \(n\le 5\). In the argument, the main tool is the invariant \(\mathrm{HFT}(T)\) for a \(2\)-string tangle \(T\). It is a collection of immersed curves on the boundary of the tangle with \(4\) punctures. This invariant is known to detect rational tangles. As a major step, the authors prove that \(\mathrm{HFT}(T)\) detects split tangles.