Random knots on closed surfaces and the Neuwirth conjecture
Description
Let $S$ be a closed surface of genus at least two embedded in the $3$-sphere in an arbitrary way, and let $K_0$ be a non-separating simple closed curve on $S$. We move $K_0$ by a random walk $w_n$ on the mapping class group of $S$ and regard $w_n(K_0)\subset S\subset S^3$ as a random knot. Assume that the step distribution has finite support. If the semigroup generated by its support contains a complete subgroup, then, almost surely, $S$ is a Neuwirth surface for $w_n(K_0)$ for all sufficiently large $n$. That is, $S\cap E(w_n(K_0))$ is connected, incompressible and boundary incompressible in the knot exterior. If the support of the walk generates the whole mapping class group as a semigroup, the probability that $S$ fails to be a Neuwirth surface decays exponentially, and the representativity of $(S,w_n(K_0))$ grows exponentially in $n$. Consequently, the bridge number of $w_n(K_0)$ grows exponentially, and the probability that $w_n(K_0)$ is trivial or a torus knot decays exponentially. Under the full semigroup-generation hypothesis, when $S$ is a Heegaard surface of $S^3$ of genus at least $\max\{2,t(K)+1\}$, where $t(K)$ is the tunnel number, every knot $K$ occurs in this model with positive probability. The proofs combine a characterization of Neuwirth surfaces by distances to disc sets in the curve complex with results of Kaimanovich--Masur, Maher, and Lubotzky--Maher--Wu on random walks.