---
title: Roux Scheme in Association Schemes
url: https://www.emergentmind.com/topics/roux-scheme
type: topic
---

# Roux Scheme in Association Schemes

to=arxiv_search.query  天天彩票怎么  天天中彩票官网json
{"query":"roux scheme association scheme Iverson Mixon Gavrilyuk Lansdown Munemasa Suda", "max_results": 5, "sort_by": "relevance"}

to=arxiv_search.query  彩神争霸高json
{"query":"roux scheme association scheme", "max_results": 10, "sort_by": "relevance"}
In algebraic combinatorics, a **roux scheme** is a commutative association scheme obtained from a special matrix over the group algebra of a finite abelian group. The notion was introduced in the study of doubly transitive lines, equiangular tight frames, and related Schurian association schemes [1806.09037]. Subsequent work showed that roux schemes can also be produced from ordinary association schemes, characterized by their eigenmatrices, and recognized from the association scheme induced on a local neighbourhood when that local scheme has as many relations as the thin radical [2507.18960].

## 1. Definition through a roux matrix

Let \(\Gamma\) be a finite abelian group, written multiplicatively, and let \(B\in \mathbb{C}[\Gamma]^{n\times n}\). A **roux** for \(\Gamma\) is an \(n\times n\) matrix satisfying the four axioms
\[
\text{(R1)}\quad B_{ii}=0,
\]
\[
\text{(R2)}\quad B_{ij}\in \Gamma \qquad (i\neq j),
\]
\[
\text{(R3)}\quad B_{ji}=(B_{ij})^{-1} \qquad (i\neq j),
\]
\[
\text{(R4)}\quad \{gI\}_{g\in\Gamma}\ \text{and}\ \{gB\}_{g\in\Gamma}\ \text{span an algebra } \mathscr{A}(B).
\]
The associated **roux scheme** is obtained by applying the Cayley regular representation entrywise to \(B\): if \([\cdot]\) denotes the injective \(*\)-algebra homomorphism from \(\mathbb{C}[\Gamma]^{n\times n}\) into ordinary complex matrices, then the matrices
\[
\{[gI]\}_{g\in\Gamma}\cup\{[gB]\}_{g\in\Gamma}
\]
are the adjacency matrices of the scheme [1806.09037].

A central simplification is the quadratic characterization. If \(B\) satisfies (R1)–(R3), then \(B\) is a roux if and only if
\[
B^2=(n-1)I+\sum_{g\in\Gamma} c_g\, gB
\]
for some coefficients \(\{c_g\}_{g\in\Gamma}\), called the **roux parameters**. In this case,
\[
c_g\in \mathbb{Z}_{\ge 0}, \qquad \sum_{g\in\Gamma} c_g=n-2,\qquad c_{g^{-1}}=c_g.
\]
This quadratic relation is the basic algebraic signature of the theory [1806.09037].

## 2. Schurian origin and Higman pairs

Roux schemes were introduced as a bridge between doubly transitive lines and Schurian association schemes. For a finite group \(G\) and subgroup \(H\leq G\), the Schurian scheme of \((G,H)\) is built from double cosets \(HaH\). The relevant group-theoretic input is a **Higman pair** \((G,H)\), defined by a subgroup \(H\), its normalizer \(K=N_G(H)\), and a key element \(b\in G\setminus K\) satisfying conditions denoted (H1)–(H5) in the original paper [1806.09037].

The classification theorem is exact: the Schurian scheme of \((G,H)\) is isomorphic to a roux scheme if and only if \((G,H)\) is a Higman pair [1806.09037]. Moreover, from a Higman pair one constructs a roux matrix for the abelian group \(K/H\). Choosing coset representatives \(\{x_j\}\) for \(K\) in \(G\) and \(\{a_g\}\) for \(H\) in \(K\), one defines \(B_{ii}=0\) and, for \(i\neq j\), lets \(B_{ij}\) be the unique \(g\in K/H\) such that
\[
x_i^{-1}x_j \in H a_g b H.
\]
The resulting matrix is a roux for \(K/H\), and its roux scheme is isomorphic to the Schurian scheme of \((G,H)\) [1806.09037].

This Schurian viewpoint clarifies why roux schemes were introduced in the first place. They abstract the combinatorial structure appearing in doubly transitive line systems while remaining inside the language of association schemes.

## 3. Internal characterization and spectral structure

Roux schemes also admit an intrinsic characterization inside association scheme theory. An association scheme is isomorphic to a roux scheme if and only if it is commutative, its thin radical acts regularly on the other adjacency matrices, and at least one of those other adjacency matrices is symmetric [1806.09037]. In this sense, a roux scheme is a commutative scheme obtained by “thickening” a thin scheme by a single regular orbit of non-thin relations.

The primitive idempotents of a roux scheme are explicitly controlled by the character theory of the underlying abelian group. If \(\widehat{\Gamma}\) is the character group and
\[
\widehat{c}_\alpha=\sum_{h\in\Gamma} c_h \alpha(h),
\]
then one defines
\[
\mu_\alpha^\epsilon = \frac{\widehat{c}_\alpha+\epsilon\sqrt{(\widehat{c}_\alpha)^2+4(n-1)}}{2(n-1)}, \qquad \epsilon\in\{+,-\}.
\]
The primitive idempotents are scalar multiples of
\[
\mathcal{G}_\alpha^\epsilon
=
\sum_{g\in\Gamma}\alpha(g)[gI]
+
\mu_\alpha^\epsilon
\sum_{g\in\Gamma}\alpha(g)[gB],
\]
and their ranks are
\[
d_\alpha^\epsilon=\frac{n}{1+(n-1)(\mu_\alpha^\epsilon)^2}.
\]
Thus the idempotent structure is parameterized by \((\alpha,\epsilon)\in \widehat{\Gamma}\times\{+,-\}\) [1806.09037].

Later work gave a converse spectral characterization. If the first eigenmatrix has the block form
\[
P=
\begin{bmatrix}
\mathbb{T} & \mathsf{D}_+\mathbb{T}\\
\mathbb{T} & \mathsf{D}_-\mathbb{T}
\end{bmatrix},
\]
where \(\mathbb{T}\) is the character table of the thin radical and \(\mathsf{D}_\pm\) are diagonal matrices, and if at least one entirely real column corresponding to a thick relation occurs, then the scheme is a roux scheme [2507.18960]. This makes roux schemes spectrally recognizable.

## 4. Relation to equiangular lines, ETFs, and DRACKNs

A decisive feature of the theory is that a roux matrix simultaneously encodes an association scheme and, after character evaluation, a system of equiangular lines. If \(B\) satisfies (R1)–(R3), then \(B\) is a roux if and only if for every \(\alpha\in\widehat{\Gamma}\), the evaluated matrix \(\widehat{\alpha}(B)\) is the signature matrix of an equiangular tight frame [1806.09037]. This is the basic mechanism behind **roux lines**.

The associated line systems can also be characterized intrinsically. A sequence of linearly dependent complex lines is a roux line set if and only if it is equiangular, admits unit-norm representatives whose signature matrix entries are \(r\)th roots of unity for some \(r\), and the Gram matrix of all phased copies carries an association scheme; in that case the carried scheme is the corresponding roux scheme, and a scalar multiple of the Gram matrix is a primitive idempotent of that scheme [1806.09037].

Roux schemes also interact with graph-theoretic objects. The symmetric matrix \([B]\) is the adjacency matrix of the **roux graph**. Roux graphs generalize regular abelian DRACKNs: every regular abelian \((n,r,c)\)-DRACKN is a roux graph for a roux with parameters
\[
c_g=
\begin{cases}
n-c(r-1)-2,& g=\mathrm{id},\\
c,& g\neq \mathrm{id},
\end{cases}
\]
and, conversely, a roux graph has diameter \(3\) if \(c_g\neq 0\) for every \(g\in\Gamma\), in which case it is a regular antipodal cover of the complete graph [1806.09037]. This places roux schemes at the intersection of association schemes, complex line packings, and antipodal graph coverings.

## 5. Construction from local association schemes

A major later development was the realization that roux matrices can be produced directly from ordinary association schemes. Let \(G\) be an abelian group of order \(r\), and let \(Y\) be an \(r\)-class association scheme whose non-diagonal relations are indexed by \(G\). From the adjacency matrices of \(Y\), one forms a matrix
\[
B=\sum_{g\in G} g\,\mathsf{B}_g.
\]
This matrix is a roux matrix if and only if the local relations satisfy
\[
R_g^\top = R_{g^{-1}}
\]
and the intersection numbers satisfy
\[
\sum_{g\in G} p_{g^{-1},\,gh}^{\,m}=k_{h^{-1}m}-\delta_{h,1}
\qquad \text{for all } h,m\in G.
\]
When these conditions hold, the roux parameters are exactly the valencies \(\{k_g\}_{g\in G}\) of \(Y\) [2507.18960].

This leads to a canonical construction, denoted
\[
Y\widehat\otimes G,
\]
whose thin relations are
\[
T_k=I\otimes P_k
\]
and whose thick relations are
\[
S_k=\sum_{\ell\in G}\mathsf{B}_\ell\otimes P_{k\ell},
\]
where \(P_k\) is the regular permutation matrix of \(k\in G\) [2507.18960].

The local theory is equally rigid. If \(X\) is a roux scheme with thin radical \(G\), and the neighbourhood of some vertex with respect to some thick relation induces a \(|G|\)-class local association scheme \(Y\), then
\[
X=Y\widehat\otimes G.
\]
More generally, if a commutative association scheme has thin radical \(G\) acting regularly on the thick relations, then it is uniquely recoverable from \(G\) and the neighbourhood structure of any thick relation about any vertex [2507.18960]. This is the sense in which some roux schemes “carry association schemes locally.”

## 6. Examples, uniqueness, and later developments

The theory includes both Schurian and non-Schurian examples. Early constructions include roux matrices arising from antisymmetric conference matrices over \(C_4\), symplectic examples over finite fields, and matrices related to Hoggar lines [1806.09037]. Later work used the local-construction theorem to produce new families from amorphic pseudocyclic association schemes with elementary abelian \(2\)-groups, as well as cyclic-group examples [2507.18960].

A distinguished case is provided by the \(64\) equiangular lines in \(\mathbb{C}^8\) constructed by Hoggar. The later paper identifies the corresponding roux scheme as an important example and proves that it is unique, determined by its parameters up to isomorphism [2507.18960]. This is significant because Hoggar’s lines are among the most rigid and symmetric ETF configurations, and the roux-scheme formalism captures that rigidity at the level of association schemes.

In contemporary usage, a roux scheme is therefore best understood as a highly structured commutative association scheme built from an abelian thin radical and one regular orbit of thick relations, with three equivalent faces. It can be defined algebraically through a roux matrix, recognized internally through thin-radical regularity and spectral form, and interpreted geometrically through ETFs, roux lines, and abelian covers of complete graphs [1806.09037]. The later local theory shows that the concept is not merely a reformulation of Higman-pair constructions, but a framework in which global structure, eigenmatrices, and induced neighbourhood schemes can all determine one another [2507.18960].

Source: https://www.emergentmind.com/topics/roux-scheme