Counting roots of unity on the graphs of Laurent series over non-Archimedean local fields
Christoph Pütz
TL;DR
The paper classifies Laurent series over non-Archimedean local fields that map infinitely many roots of unity to roots of unity, showing such series are forced to be 'special' (simple monomial form in characteristic 0 or finite-field-like in positive characteristic). It develops a ramified/unramified analytic framework and leverages the auxiliary function $f(X^q)-f(X)^q$ to bound torsion on the graph, yielding effective, computable bounds tied to zeros of unit-valued functions. In positive characteristic, the bound is governed by the count of zeros of $f(X)^q-f(X^q)$; in characteristic 0, ramification and a computable constant $c_f$ appear, producing explicit finite bounds and applications to Manin–Mumford-type problems in $\mathbb{G}_m^2$. The results extend Schmidt’s ideas to Laurent series on unit spheres and provide practical bounds for torsion configurations relevant to arithmetic dynamics and algebraic geometry over $p$-adic fields.
Abstract
We completely classify Laurent series converging on the unit circle over a non-Archimedean local field (of any characteristic) that map infinitely many roots of unity to roots of unity. For a given Laurent series $f$ over a field of positive characteristic with residue field $\mathbb{F}_q$, we prove effective bounds for the number of possible roots of unity in terms of the number of zeroes of the auxilliary function $f(x^q)-f(x)^q$ on the unit circle. In characteristic $0$ our bound is still effective but also depends on the ramification degree of the base field over $\mathbb{Q}_p$ as well as the size of the coefficients of $f$. This has applications to the Manin-Mumford conjecture in $\mathbb{G}_m^2$. In characteristic $0$, this work builds upon a pigeon-hole based method by Schmidt.
