Belyi map verification using certified path tracking
Alexandre Guillemot, John Voight · 2026
Read Belyi map verification using certified path tracking by Alexandre Guillemot, John Voight (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-15562.txt.
Introduction
Guillemot and Voight deliver a full workflow that works out Belyi map monodromy exactly from equations over number fields via certified homotopy continuation. Run at scale, it verifies monodromy triples for Belyi maps in the L-functions and Modular Forms Database. The checked pipeline turns a fiddly computation into a result with proof behind it.
Catalog record
| CREATOR | Alexandre Guillemot, John Voight |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 16761 |
| LICENSE | http://creativecommons.org/licenses/by-sa/4.0/ |
| RIGHTS | CC BY SA |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.15562
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Belyi map verification using certified path tracking Alexandre Guillemot, John Voight Abstract We provide an end-to-end workflow to rigorously compute the monodromy of Belyi maps from exact equations over number fields using certified homotopy continuation. We then apply this method at scale to certify the monodromy triples of Belyi maps in the L L -functions and Modular Forms Database (LMFDB). 1 Introduction Motivation and context Let X be a smooth projective algebraic curve over \mathbb{C} or equivalently a compact Riemann surface. A Belyi map is a nonconstant map \varphi\colon X\to\mathbb{P}^{1}_{\mathbb{C}} unramified away from \{0,1,\infty\} . [ 5 ] proved that X can be defined over \mathbb{Q}^{\textup{al}} if and only if it admits a Belyi map. Explicit methods for Belyi maps have seen many applications—see [ 25 ] for an overview, as well as the more recent work by [ 23 , 1 , 6 ] . These methods take as input a permutation triple encoding the monodromy around the branch points and return candidate equations over a number field K\subset\mathbb{C} . One can certify that the map is indeed a Belyi map with the correct ramification. In this paper, we tackle rigorously the certification that the monodromy triple of the map indeed coincides with the given input. Algebraic techniques [ 14 , 1 ] may suffice to compute the monodromy group and the triple when uniquely determined, but do not provide a general approach. On the other hand, monodromy permutations can be computed from the equations using certified path tracking , also called certified homotopy continuation : for a parametrized polynomial system F_{t} , one follows the zeros of F_{t} along loops in the base. For monodromy, a heuristic approach is not sufficient even with a posteriori end point certification, since path jumping or swapping may occur [ 3 , §6.2] . Certified methods instead isolate the tracked zero along the whole path, ensuring correctness. Previous work The idea of using certified path tracking to rigorously compute the monodromy of Belyi maps is not new. For instance, [ 11 ] and [ 9 , §2C] describe predictor-corrector loops whose certification rely on a step size bound. [ 10 ] and [ 2 ] present approaches that compute fibers above the points of a discretization of the parameter path and connect each consecutive fibers to recover the monodromy permutation. However, implementing these methods to obtain a rigorous result is difficult, as they may consider implicit bounds, that roots of polynomials can be obtained exactly, or assume a model of exact computation over the reals (e.g., the BSS model) for correctness. [ 4 ] provide the first implemented certified homotopy continuation algorithm in the Turing machine model, using exact arithmetic over the rationals and with generality well beyond Belyi maps. For rigorous numerical computations, other methods rely on interval arithmetic in some way. For the univariate case, see the algorithms and implementations by [ 20 , 19 , 27 ] . The multivariate case is tackled by [ 15 , 12 ] , based on ideas from [ 26 ] . The preprint by [ 13 ] addresses monodromy computations for general complex algebraic branched covers, with an emphasis on practical results and including Belyi maps. Contribution Using Algpath [ 17 , 15 ] , a certified path tracking software based on Krawczyk’s operator, interval arithmetic, and Taylor models, we rigorously compute the monodromy triples of the 1111 Belyi maps present in the LMFDB [ 22 ] . The results prove for the first time the correctness of the monodromy groups and, up to minor issues with permuting the branch points \{0,1,\infty\} , of the monodromy triples attached to all entries in the database. Code is available online [ 16 ] . This demonstrates that recent advances make certified homotopy continuation a practical tool to compute the monodromy of finite algebraic maps. 2 Setup Monodromy representation Let X be a smooth projective algebraic curve over \mathbb{C} and let \varphi\colon X\to\mathbb{P}^{1}_{\mathbb{C}} be a Belyi map of degree d . Put U\mathrel{\mathchoice{\vbox{\hbox{$\displaystyle:$}}}{\vbox{\hbox{$\textstyle:$}}}{\vbox{\hbox{$\scriptstyle:$}}}{\vbox{\hbox{$\scriptscriptstyle:$}}}{=}}\mathbb{P}^{1}_{\mathbb{C}}\setminus\{0,1,\infty\} , so that \varphi restricts to a topological cover \varphi^{-1}(U)\to U of degree d . Fix a basepoint b\in U and label the fiber \varphi^{-1}(b)=\{x_{1},\dots,x_{d}\} . Analytic continuation along loops in U defines a monodromy representation \rho\colon\pi_{1}(U,b)\to S_{d} . Choosing standard loops \gamma_{0},\gamma_{1},\gamma_{\infty} around 0,1,\infty with \gamma_{0}\gamma_{1}\gamma_{\infty}=1 , we obtain permutations \sigma_{0},\sigma_{1},\sigma_{\infty} satisfying \sigma_{0}\sigma_{1}\sigma_{\infty}=1 , called the monodromy triple of \varphi , well-defined up to simultaneous conjugation. The subgroup G=\langle\sigma_{0},\sigma_{1},\sigma_{\infty}\rangle\subseteq S_{d} is the (geometric) monodromy group . The cycle structure of \sigma_{s} records the ramification above s\in\{0,1,\infty\} . Analytic continuation via polynomial systems To compute this triple from equations, let Y\subseteq\mathbb{C}^{n} be an affine chart cut out by polynomials g_{1},\dots,g_{n-1} , and suppose \varphi is represented on this chart by a rational function p/q . Consider the polynomial system F_{t}(x,z)=\bigl(g_{1}(x),\dots,g_{n-1}(x),\,p(x)-tq(x),\,q(x)z-1\bigr). (2.1) For each t\in U , the regular zeros of F_{t} are in bijection with the points of \varphi^{-1}(t) , since q(x)z-1=0 excludes the common vanishing locus of p and q . Over U , these zeros are regular and vary analytically with t , so analytic continuation of zeros of ( 2.1 ) agrees with path lifting for \varphi . Therefore, if \gamma\colon[0,1]\to U is a loop based at b and each regular zero of F_{b} is tracked correctly along \gamma , then the induced permutation on the zeros of F_{b} is the monodromy permutation of \varphi associated to \gamma . In particular, it is enough to compute the permutations associated to loops around 0 and 1 . 3 Monodromy computations using path tracking We briefly explain the underlying algorithm of Algpath and how it applies to the computation of monodromy permutations. Data structure for regular zeros of polynomial systems Let f\colon\mathbb{C}^{n}\to\mathbb{C}^{n} be a polynomial map. We review Moore’s root isolation criterion, based on Krawczyk’s operator, and the associated data structure for regular zeros. We denote by B the unit ball for the real \infty -norm on \mathbb{C}^{n} , and by I_{n} the n\times n identity matrix. Theorem 3.1 (Moore’s criterion) Let \rho\in(0,1) , x\in\mathbb{C}^{n} , r>0 and A\in\mathbb{C}^{n\times n} . Suppose that for all u,v\in rB , we have -Af(x)+(I_{n}-Adf(x+u))v\in\rho rB. (3.2) Then there exists a regular zero \zeta of f in x+\rho rB , and it is the unique zero of f in x+rB . The proof can be found in [ 24 , 15 ] and relies on Banach’s fixed-point theorem. Definition 1 (Moore boxes) Let \rho\in(0,1) . A \rho -Moore box for f is a triple (x,r,A)\in\mathbb{C}^{n}\times\mathbb{R}_{>}\times\mathbb{C}^{n\times n} satisfying ( 3.2 ). A Moore box is a \rho -Moore box for some \rho\in(0,1) . By Theorem 3.1 , a Moore box represents a unique regular zero of f . Condition ( 3.2 ) is checked effectively by interval arithmetic [ 21 ] . Given a \rho -Moore box m=(x,r,A) and \tau\in(0,1) , one can refine m to a \tau -Moore box with the same associated zero by quasi-Newton iterations and reductions of r [ 15 , Algorithm 2] . Equality of the zeros represented by two Moore boxes is decided by refining both to 1/3 -Moore boxes and checking whether the shrunken boxes intersect. Path tracking algorithm We focus on the algorithm implemented in Algpath; a precise