Algorithms for experimenting with Zariski dense matrix groups over number fields
A. S. Detinko, D. L. Flannery, A. Hulpke · 2026
Read Algorithms for experimenting with Zariski dense matrix groups over number fields by A. S. Detinko, D. L. Flannery, A. Hulpke (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-2605-23798.txt.
Introduction
A. S. Detinko, D. L. Flannery and A. Hulpke turn a deep theorem about dense matrix groups into working code. For finitely generated Zariski dense groups inside special linear groups over number fields, with prime matrix size, their algorithms list all congruence quotients modulo maximal ideals of a suitable ring. The code runs in GAP, and degree two trials, centered on Bianchi groups, show what the experiments can reveal about these groups.
Catalog record
| CREATOR | A. S. Detinko, D. L. Flannery, A. Hulpke |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 61957 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.23798
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Algorithms for experimenting with Zariski dense matrix groups over number fields A. S. Detinko, D. L. Flannery, A. Hulpke Abstract . Let ℙ \mathbb{P} be an algebraic number field. We provide a computational analog of the strong approximation theorem for finitely generated Zariski dense groups H ≤ SL ( n , ℙ ) H\leq\mathrm{SL}(n,\mathbb{P}) , n n prime. That is, we present algorithms to find the set of congruence quotients of H H modulo all maximal ideals of a finitely generated subring R R of ℙ \mathbb{P} such that H ≤ SL ( n , R ) H\leq\mathrm{SL}(n,R) . The algorithms have been implemented in GAP . Potential applications are illustrated by a range of experiments in degree 2 2 , with a special focus on Bianchi groups. 2020 Mathematics Subject Classification 20-04, 20G15, 20H25, 22E40, 68W30 1. Introduction This work is a further contribution to an expanding area of computational group theory; namely, computing with matrix groups over an infinite field \mathbb{F} . Our algorithms take as input a finite subset of \mathrm{GL}(n,\mathbb{F}) . This generates a subgroup H of \mathrm{GL}(n,\mathbb{F}) such that H\leq\mathrm{GL}(n,R) for some finitely generated integral domain R\subseteq\mathbb{F} . Since H is residually finite [ 30 , 4.2] , it can be investigated efficiently via its finite quotients, in particular the congruence quotients arising from reduction modulo maximal ideals of R . A single congruence quotient of H is enough to solve some decision problems; e.g., testing finiteness [ 11 ] . Deeper study of H requires knowledge of a plurality of its congruence quotients. Two questions are foremost: can we find all congruence quotients of H ; how do the congruence quotients of H determine its structure and properties? By way of the strong approximation property for linear groups (see [ 23 , Window 9] and [ 25 , 31 ] ), we can describe the congruence quotients of H when H is a dense subgroup of \mathrm{SL}(n,\mathbb{C}) . (Here and from now on, ‘dense’ means ‘Zariski dense’.) Indeed, if \mathbb{F}=\mathbb{Q} then the strong approximation property implies that dense H surjects onto \mathrm{SL}(n,p) modulo almost all rational primes p . The congruence quotients of such H modulo all maximal ideals of R are therefore known once we have computed the finite set \Pi(H) of primes modulo which H and \mathrm{SL}(n,R) are not congruent. In [ 8 , 9 ] we gave practical algorithms to compute \Pi(H) for input dense H\leq\mathrm{SL}(n,\mathbb{Q}) : a computational analog of strong approximation. A hallmark of our methods is that they transfer much of the computing to matrix groups over finite fields. In this regard, the GAP package [ 24 ] , which realizes a categorization due to Aschbacher [ 2 ] of maximal subgroups of classical groups over finite fields, is a major component of our implementations. Strong approximation not only ensures that we can compute the congruence quotients of a dense group H ; it implies that H is well approximated by its congruence quotients. This affords a framework for examining the structure of H . In [ 10 ] we developed algorithms that accept dense H\leq\mathrm{SL}(n,\mathbb{Z}) and \Pi(H) as input and compute the arithmetic closure of H , i.e., the intersection of all finite-index subgroups of \mathrm{SL}(n,\mathbb{Z}) containing H . These algorithms facilitated the solution of problems for which no methods were previously available (e.g., see [ 10 , Section 4] ). Let \mathbb{P} be an algebraic number field. Our primary goal in this paper is to extend the techniques of [ 8 , 9 ] to handle finitely generated dense subgroups H of \mathrm{SL}(n,\mathbb{P}) . Now H\leq\mathrm{SL}(n,R) for R\subseteq\mathcal{O}[1/\mu]=\{a/\mu^{i}\,|\,a\in\mathcal{O},i\in\mathbb{Z}\} where \mathcal{O}=\mathcal{O}_{\mathbb{P}} is the ring of integers of \mathbb{P} and \mu\in\mathbb{Z} , \mu>0 . Specifically, we obtain practical algorithms to describe the set of congruence quotients of H modulo all maximal ideals I of R . As with the algorithms over \mathbb{P}=\mathbb{Q} , our approach is based on reduction to computing in \mathrm{SL}(n,R/I) . So we continue to rely on Aschbacher’s categorization of the maximal subgroups of \mathrm{SL}(n,p^{k}) , p prime; see [ 3 ] for an exhaustive treatment. Suppose that \mathrm{SL}(n,R/I) and the congruence image \widetilde{H} of H modulo a maximal ideal I of R are different. Then \widetilde{H} lies in a maximal subgroup of \mathrm{SL}(n,R/I) belonging to some Aschbacher class \mathscr{C}_{i} , 1\leq i\leq 9 . The idea is to say, for each i , precisely when there can be a maximal subgroup in \mathscr{C}_{i} containing a congruence image of H . We emphasize that the situation for groups over \mathbb{P} differs markedly from that for groups over \mathbb{Q} . Firstly, if \mathbb{P}\neq\mathbb{Q} then the image field R/I is not necessarily of prime size, and classes of maximal subgroups of \mathrm{SL}(n,R/I) that are not present in \mathrm{SL}(n,p) emerge. Another issue is that there might be infinitely many maximal ideals I of R\subseteq\mathcal{O}[1/\mu] such that dense H\leq\mathrm{SL}(n,R) does not surject onto \mathrm{SL}(n,R/I) modulo I . However, under certain conditions, H will surject onto \mathrm{SL}(n,R/I) modulo all but finitely many maximal ideals I . We give algorithms to decide whether H satisfies those conditions; if it does, then we compute the (finite) set of maximal ideals I\subset R such that H does not surject onto \mathrm{SL}(n,R/I) . Some of our procedures are valid for any degree n . Others are restricted to prime n , taking advantage of the simplification of linear group structure in prime degree. At the same time this suffices to cover several important applications; cf. [ 8 , Section 4] and see below. Our approach for arbitrary n entails computing with adjoint representations, and is dealt with in [ 7 ] . We are motivated by applications—e.g., in algebraic number theory and geometry—that feature linear groups over number fields other than \mathbb{Q} , especially in small degrees. As a sample application of our algorithms, we consider subgroups H of Bianchi groups; for these n=2 and \mathbb{P}=\mathbb{Q}(\sqrt{-d}\hskip 0.75pt) , where d\in\mathbb{Z} is positive and square-free. If H is dense then we compute \Pi(H) and test whether H has finite index. Subsequently we decide whether each finite-index H is a congruence subgroup, i.e., contains the full kernel of a congruence homomorphism defined on \mathrm{SL}(2,\mathcal{O}_{\mathbb{P}}) . This enables us to classify all subgroups of \mathrm{SL}(2,\mathbb{Z}[\omega]) of index up to 30 , for a primitive cube root of unity \omega , and find the congruence subgroups among them. We repeat the latter exercise to classify all normal subgroups of \mathrm{SL}(2,\mathbb{Z}[\omega]) of index up to 20000 . Our algorithms are the first to be developed for solving such problems. The paper is organized as follows. In Section 2 we explain how to compute congruence homomorphisms defined on finitely generated subgroups of \mathrm{GL}(n,\mathbb{P}) . In Section 3 we determine when a congruence image of H\leq\mathrm{SL}(n,\mathbb{P}) is in a subgroup from each Aschbacher class. In Section 4 we use a characterization of dense H in terms of its congruence images to design an algorithm for computing congruence quotients of H modulo maximal ideals of an ambient coefficient ring. Finally, in Section 5 we discuss computer experiments with subgroups of Bianchi groups performed using a GAP implementation of the algorithms, that demonstrate the effectiveness and applicability of our methods. 2.