Eichler orders, quotient graphs and random walks

Luis Arenas-Carmona, Marco Godoy · 2025

Read Eichler orders, quotient graphs and random walks by Luis Arenas-Carmona, Marco Godoy (2025) 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-2503-12237.txt.

Introduction

Luis Arenas-Carmona and Marco Godoy study quotient graphs tied to arithmetic orders in a twenty twenty-five paper. Their objects are Eichler orders, a class of orders in quaternion algebras, and the Bruhat-Tits tree, a vast graph that codes the local structure at one prime place.

The question is how much of the quotient graph at one place can be computed from the same quotient at a distinct place. They show the computation works, apart from a small set of vertices that depends on the known place but not the unknown one. They also give geometric conditions that clear the exception. The work extends earlier formulas for maximal orders over the projective line.

Catalog record

CREATORLuis Arenas-Carmona, Marco Godoy
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE173 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-12237

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Eichler orders, quotient graphs and random walks Luis Arenas-Carmona, Marco Godoy Abstract . We study the extent to which the quotient of the Bruhat-Tits tree at one place Q Q , associated to a genus of orders of maximal rank, can be computed from the analogous quotient at a different place P P . We show that this computation can be carried out, except for a small set of vertices depending on P P , but not on Q Q . We give some geometrical conditions on the quotient at P P that ensure that this exceptional set is empty. This generalizes the formulas from a previous work that allow the computation of the quotient graph at all places, for the genus of maximal orders over the projective line. The methods presented here yield similar results for other genera or other curves. 1. Introduction In all of this work K is a global function field, i.e. the field of rational functions on a smooth irreducible projective curve X over a finite field \mathbb{F} . We let \mathcal{O}_{X} denote the structure sheaf of X , and we let |X| be the set of closed points in X . Furthermore, we write \mathfrak{A}=\mathbb{M}_{2}(K) for the 2 -by- 2 matrix algebra, while \mathfrak{R} , \mathfrak{D} and \mathfrak{E} denote X -orders of maximal rank in \mathfrak{A} , i.e. sheaves of \mathcal{O}_{X} -algebras whose generic fiber is \mathfrak{A} . Throughout this work, K_{P} is the completion of K for the valuation defined by the place P\in|X| . A similar convention applies to completions of vector spaces, algebras, and orders. For example \mathfrak{A}_{P}=\mathbb{M}_{2}(K_{P}) . The completion \mathfrak{D}_{P} of an order \mathfrak{D} in \mathfrak{A} is defined as the closure in \mathfrak{A}_{P} of the ring \mathfrak{D}(U)\subseteq K , for any affine open set U of X containing P . This definition is independent of the choice of U . Recall that the adele ring of an algebra is the restricted topological product (1) \mathfrak{A}_{\mathbb{A}}=\left\{a=(a_{P})_{P}\in\prod_{P\in|X|}\mathfrak{A}_{P}\Big|a_{P}\in\mathfrak{D}_{P}\textnormal{ for almost all }P\right\}. This definition is independent of \mathfrak{D} . The global algebra \mathfrak{A} is identified with the image of the diagonal embedding \mathrm{diag}:\mathfrak{A}\rightarrow\mathfrak{A}_{\mathbb{A}} defined by \mathrm{diag}(a)=(a)_{P} . Similarly, we define the adele ring \mathbb{A}=K_{\mathbb{A}} , replacing \mathfrak{D} in ( 1 ) by the order \mathcal{O}_{X} . The group J_{X}=\mathbb{A}^{*} is called the idele group of K in the sequel. More generally, we write R^{*} for the group of units of an arbitrary ring R . Two orders \mathfrak{D} and \mathfrak{E} are said to be locally conjugate, or to belong to the same genus, if any of the following equivalent statements hold: (1) The stalks \mathfrak{D}_{P}^{o},\mathfrak{E}_{P}^{o}\subseteq\mathfrak{A} are conjugates at every place P\in|X| . (2) The completions \mathfrak{D}_{P},\mathfrak{E}_{P}\subseteq\mathfrak{A}_{P} are conjugate at every place P\in|X| . (3) There is an element a\in\mathfrak{A}_{\mathbb{A}}^{*} satisfying a_{P}\mathfrak{D}_{P}a_{P}^{-1}=\mathfrak{E}_{P} at every place P . When (3) is satisfied we write simply a\mathfrak{D}a^{-1}=\mathfrak{E} by abuse of notation. Similar conventions apply to other linear algebraic groups acting on lattices. Being locally conjugate is an equivalence relation. As usual (see [ 2 ] or [ 4 ] ), by a genus of orders of maximal rank, we mean an equivalence class for this relation. The study of genera is an important part of the theory of orders, since local classification is much simpler than global classification. On the other hand, over a Dedekind domain, like A=\mathcal{O}_{X}(U) for an affine set U\subseteq X , two A -orders in \mathfrak{A} , like \mathfrak{D}(U) and \mathfrak{E}(U) , are conjugate precisely when they are in the same spinor genus. Two X -orders in a genus are said to be in the same spinor genus if the adelic element a in (3) above can be chosen of the form a=bc , where b\in\mathfrak{A} , and the local coordinate c_{P} of c=(c_{P})_{P} has trivial reduced norm at each place P . Spinor genera of A -orders are defined analogously. Determining whether two X -orders in the same spinor genus are conjugate can be studied through classifying graphs. Fix a place P\in|X| , let U=\{P\}^{c}\subset X be its complement, and set A=\mathcal{O}_{X}(U) as before. Fix a genus \mathbb{O} whose orders are maximal at P . The classifying graph C_{P}(\mathbb{O}) is a graph, with a vertex for each conjugacy class in \mathbb{O} , whose connected components C_{P}(\mathfrak{D}) are in correspondence with the conjugacy classes of A -orders of the form \mathfrak{D}(U) for \mathfrak{D}\in\mathbb{O} . Furthermore, each connected component is the quotient of the Bruhat-Tits tree \mathfrak{t}(K_{P}) for \mathrm{PSL}_{2}(K_{P}) by the normalizer in \mathrm{PSL}_{2}(K) of the corresponding \mathcal{O}_{X}(U) -order \mathfrak{D}(U) . Fix one such A -order \mathfrak{D}(U) . For any local maximal order \mathfrak{D}^{\prime}(P)\subseteq\mathfrak{A}_{P} , there is an order \mathfrak{D}^{\prime}\in\mathbb{O} satisfying \mathfrak{D}^{\prime}(U)=\mathfrak{D}(U) and \mathfrak{D}^{\prime}_{P}=\mathfrak{D}^{\prime}(P) . Such orders are called P -variants of \mathfrak{D} . The vertices in C_{P}(\mathfrak{D}) are in correspondence with the conjugacy classes of these P -variants. Computing the classifying graph for a given genus \mathbb{O} at a single place P is sufficient for the purpose of determine all conjugacy classes. There are, however, additional reasons to compute these classifying graphs at different places: (1) Basse-Serre’ Theory allows us to determine generators for the associated arithmetic subgroups (c.f. [ 14 , §5] ). (2) Genera of Eichler orders can be studied in terms of genera of simpler Eichler orders via suitable quotient graphs (c.f. [ 7 ] and [ 10 ] ). (3) Selectivity of certain suborders is related to the structure of the graphs in several ways (c.f. [ 4 ] , [ 5 ] ). For these reasons, it would be desirable to be able to compute the classifying graph C_{P}(\mathbb{O}) in terms of the corresponding graph C_{Q}(\mathbb{O}) at a different place. In [ 4 ] , we were able to compute all quotient graphs for maximal orders over the projective line \mathbb{P}^{1} using this idea. An alternative approach, for the same curve and genus appears in [ 12 ] . In general terms, the strategy used in the former reference can be summarized as follows: • From the classifying graph C_{P}(\mathbb{O}) we compute a “neighborhood matrix N_{P}(\mathbb{O}) ”. • The neighborhood matrix N_{Q}(\mathbb{O}) , at a different place, can be computed from N_{P}(\mathbb{O}) . • The neighborhood matrix N_{Q}(\mathbb{O}) can be used to obtain the quotient graph C_{Q}(\mathbb{O}) . Above, the neighborhood matrix N_{P}(\mathbb{O})=(m_{i,j})_{i,j} is an infinite matrix whose entry m_{i,j} is the number of neighbors in \mathfrak{t}(K_{P}) of \mathfrak{D}_{j} that are isomorphic to \mathfrak{D}_{i} , for some set \{\mathfrak{D}_{1},\mathfrak{D}_{2},\dots\} of representatives of all conjugacy classes in \mathbb{O} . Our ability to compute the classifying graph from the corresponding neighborhood matrix depends on our understanding on the action of the stabilizer of a vertex in its neighbor set. This is simple for some vertices, namely those corresponding to split orders, as defined in Proposition 4.1 , to the point where we can give explicit formulas for the number of edges joining two vertices from the corresponding coefficient m_{i,j} and conversely, but potentially much harder for vertices corresponding to non-split orders (c.f. [ 4 , §6] ). In this work we concentrate in the central step, i.e., computing one neighborhood matrix in terms of another. With this tool alone, we can compute which pairs of vertices are neighbors, but not the multiplicity of edges, and we cannot distinguish

Read the complete text