betaFeedback

WorksMathematics

Coxeter Group Symmetries of Frieze Patterns

Abstract

Coxeter group symmetries of frieze patterns are investigated through their relation to 2×n2\times n matrices and the associated Plücker coordinates. For odd order nn, the symmetric group acts on the set of friezes, while for even nn, the dihedral subgroup acts. We further show that a larger group acts when n=6n = 6.

Introduction

The study of frieze patterns was initiated by Coxeter [6]. In his original formulation, a frieze pattern is represented as an infinite planar array of numbers arranged in staggered rows. A frieze pattern of order nn is a grid of numbers with n+1n+1 rows as in Figure 1, where the first and last rows (row 0 and row nn) are composed entirely of zeros, while the second and second-to-last rows (row 1 and row n−1n-1) consist entirely of ones. In other words, for any integer ii, we have

Figure 1. A frieze pattern of order nn arranged in staggered rows.

fi,i=fi,i+n=0,fi,i+1=fi,i+n−1=1.(1)f_{i,i}=f_{i,i+n}=0,\qquad f_{i,i+1}=f_{i,i+n-1}=1. \tag*{(1)}

The internal entries of the frieze are governed by the “unimodular rule”; given any four adjacent elements arranged in a diamond configuration,

fi+1,jfi,jfi+1,j+1fi,j+1\begin{array}{c} f_{i+1,j}\\ f_{i,j}\qquad f_{i+1,j+1}\\ f_{i,j+1} \end{array}

the relation

fi,jfi+1,j+1−fi+1,jfi,j+1=1(2)f_{i,j}f_{i+1,j+1}-f_{i+1,j}f_{i,j+1}=1 \tag*{(2)}

must hold for i<j<i+ni<j<i+n. By the unimodular rule (2), providing n−3n-3 diagonal entries f1,3,f1,4,…,f1,n−1f_{1,3}, f_{1,4},\ldots,f_{1,n-1} is sufficient to determine all other entries [6].

In their foundational work [5], Conway and Coxeter proved that every frieze pattern of order nn whose non-trivial entries {fi,i+k;i∈Z,k=2,3,…,n−2}\{f_{i,i+k}; i \in\mathbb{Z}, k=2,3,\ldots,n-2\} are positive integers corresponds uniquely to a triangulation of a convex nn-gon. The entries in the first non-trivial row (“quiddity” [5]) {fi,i+2; i=1,2,…,n}\{f_{i,i+2};\, i=1,2,\ldots,n\} represent the number of triangles incident to each vertex of the polygon. Since then, frieze patterns have been studied in connection with geometry and combinatorics [12, 14]. Recently, the introduction of cluster algebras [7] brought increasing attention to the subject, as Conway-Coxeter friezes give a natural realization of a cluster algebra of type An−3A_{n-3}.

As noted by Coxeter [6], if we assume that the entries fi,jf_{i,j} are non-zero for i<j<i+ni<j<i+n, the following property holds:

Proposition 1 (cf. [6]). For a frieze of order nn, let fi,jf_{i,j} be its entries. If fi,j≠0f_{i,j}\ne0 for all i<j<i+ni<j<i+n, then the Plücker relation

fi,jfk,ℓ−fi,kfj,ℓ+fi,ℓfj,k=0(3)f_{i,j}f_{k,\ell}-f_{i,k}f_{j,\ell}+f_{i,\ell}f_{j,k}=0 \tag*{(3)}

is satisfied for any integers i≤j≤k≤ℓ≤i+ni\le j\le k\le\ell\le i+n.

A proof of Proposition 1 can be found in [11] (answer to exercise 2.3.4.6-4). The Plücker relations (3) are fundamental to the geometry of Grassmannians. From this perspective, a frieze pattern can be viewed as a realization of the coordinate ring of the Grassmannian

Gr⁡(2,n):=GL⁡2(C)\{Z=(v1,v2,…,vn)∈Mat⁡2×n(C)∣rank⁡Z=2}.(4)\operatorname{Gr}(2,n):=\operatorname{GL}_{2}(\mathbb{C})\backslash\{Z=(v_1,v_2,\ldots,v_n)\in\operatorname{Mat}_{2\times n}(\mathbb{C})\mid\operatorname{rank}Z=2\}. \tag*{(4)}

where vi=(z1,i,z2,i)Tv_i=(z_{1,i},z_{2,i})^{T} (i=1,…,ni=1,\ldots,n). In the next section, we introduce an identification of a frieze with a 2×n2\times n matrix Z=(v1,…,vn)Z=(v_1,\ldots,v_n) [14]. This formulation enables us to analyze the symmetries of the frieze as geometric transformations of the vector sequence {vi∣i=1,…,n}\{v_i\mid i=1,\ldots,n\} induced by the action of the symmetric group Sn\mathfrak{S}_n. In this paper, we provide an explicit description of these induced transformations. If nn is odd, the entire symmetric group Sn\mathfrak{S}_n acts on the frieze; if nn is even, the dihedral group Dn\mathfrak{D}_n (⊂Sn\subset\mathfrak{S}_n) acts on it.

This paper is organized as follows. Section 2 establishes the algebraic foundations of frieze patterns, describes their correspondence with the matrix space M^n\widehat{\mathcal{M}}_n, and introduces the normal forms. Section 3 then investigates the induced action of the symmetric group Sn\mathfrak{S}_n. For odd nn, we demonstrate that the unique normal form yields a well-defined action of Sn\mathfrak{S}_n. For even nn, we analyze which subgroup can act.

Frieze Patterns and the Matrix Space M^n\widehat{\mathcal{M}}_n

In this section, we establish the algebraic foundations for friezes extended to the entire planar lattice Z×Z\mathbb{Z}\times\mathbb{Z}.

Definition 2 (cf. [15]). Let K\mathbb{K} be a subset of the field of rational functions with complex coefficients C(X)\mathbb{C}(X), and let n≥4n\ge4 be an integer. An array of numbers F=(fi,j)i,j∈ZF=(f_{i,j})_{i,j\in\mathbb{Z}} (fi,j∈Kf_{i,j}\in\mathbb{K}) is called a frieze of order nn over K\mathbb{K} if it satisfies the global Plücker relation,

fi,jfk,ℓ−fi,kfj,ℓ+fi,ℓfj,k=0,(5)f_{i,j}f_{k,\ell}-f_{i,k}f_{j,\ell}+f_{i,\ell}f_{j,k}=0, \tag*{(5)}

and the boundary conditions,

fi,i=fi,i+n=0,fi,i+1=fi+1,i+n=1.(6)f_{i,i}=f_{i,i+n}=0,\qquad f_{i,i+1}=f_{i+1,i+n}=1. \tag*{(6)}

In what follows, we denote the set of all friezes of order nn over C(X)\mathbb{C}(X) by CFn\mathcal{C}\mathcal{F}_n.

Remark 3. With the initial conditions in (6), Definition 2 is equivalent to the notion of a tame SL2\mathrm{SL}_2-tiling in [15].

Several noteworthy features of this definition deserve immediate attention, particularly regarding how the global algebraic constraints recover and extend classical properties of friezes. By specializing the indices in the global Plücker relation (5) and applying the boundary conditions (6), we obtain

k=i+1, ℓ=j+1: fi,j=−fj,i(skew symmetry),(7)k=i+1,\ \ell=j+1:\ f_{i,j}=-f_{j,i}\quad\text{(skew symmetry),} \tag*{(7)}
k=i, ℓ=i+1: fi,jfi+1,j+1−fi+1,jfi,j+1=1(unimodular rule),(8)k=i,\ \ell=i+1:\ f_{i,j}f_{i+1,j+1}-f_{i+1,j}f_{i,j+1}=1\quad\text{(unimodular rule),} \tag*{(8)}
k=i+1, ℓ=i+n: fi,j=−fi+n,j(anti-periodicity in i),(9)k=i+1,\ \ell=i+n:\ f_{i,j}=-f_{i+n,j}\quad\text{(anti-periodicity in }i\text{),} \tag*{(9)}
k=j+1, ℓ=j+n: fi,j=−fi,j+n(anti-periodicity in j).(10)k=j+1,\ \ell=j+n:\ f_{i,j}=-f_{i,j+n}\quad\text{(anti-periodicity in }j\text{).} \tag*{(10)}

By combining the skew symmetry (7) with the anti-periodicity (9), we arrive at the identity:

fi,j=fj,i+n.(11)f_{i,j}=f_{j,i+n}. \tag*{(11)}

This relation is called a glide symmetry, a well-known classical property of frieze patterns [12, 14]. Figure 2 illustrates the anti-periodicity (9) and glide symmetry (11) described above.

Figure 2. The extended frieze pattern of order 5.

Henceforth, we let K\mathbb{K} be the field of rational functions with complex coefficients C(X)\mathbb{C}(X). Let Mn\mathcal{M}_n be the set of 2×n2 \times n matrices such that all 2×22 \times2 minors are non-zero:

Mn:={Z=(z1,1z1,2⋯z1,nz2,1z2,2⋯z2,n) | Di,j(Z)≠0 for 1≤i,j≤n}.(12)\mathcal{M}_n:=\left\{Z=\begin{pmatrix}z_{1,1}&z_{1,2}&\cdots&z_{1,n}\\z_{2,1}&z_{2,2}&\cdots&z_{2,n}\end{pmatrix}\,\middle|\,D_{i,j}(Z)\ne0\text{ for }1\le i,j\le n\right\}. \tag*{(12)}

where

Di,j(Z):=det⁡(z1,iz1,jz2,iz2,j)(13)D_{i,j}(Z):=\det\begin{pmatrix}z_{1,i}&z_{1,j}\\z_{2,i}&z_{2,j}\end{pmatrix} \tag*{(13)}

for 1≤i,j≤n1\le i,j\le n. We omit the argument ZZ when there is no risk of confusion. For a matrix Z∈MnZ\in\mathcal{M}_n as in (12), we define the coordinate function xi,jx_{i,j} by

xi,j(Z):=zi,j.(14)x_{i,j}(Z):=z_{i,j}. \tag*{(14)}

We define M^n\widehat{\mathcal{M}}_n and M^^n\widehat{\widehat{\mathcal{M}}}_n as subsets of Mn\mathcal{M}_n as follows:

M^n:={(01∗∗⋯∗1−101∗⋯∗∗)∈Mn | Di,i+1=1 for 1≤i≤n, and D1,n=1}.(15)\widehat{\mathcal{M}}_n:=\left\{\begin{pmatrix}0&1&*&*&\cdots&*&1\\-1&0&1&*&\cdots&*&*\end{pmatrix}\in\mathcal{M}_n\,\middle|\,D_{i,i+1}=1\text{ for }1\le i\le n,\text{ and }D_{1,n}=1\right\}. \tag*{(15)}
M^^n:={Z∈M^n∣z1,3=1}.(16)\widehat{\widehat{\mathcal{M}}}_n:=\left\{Z\in\widehat{\mathcal{M}}_n\mid z_{1,3}=1\right\}. \tag*{(16)}

We remark that M^n⊂M^n⊂Mn\widehat{\mathcal{M}}_n \subset\widehat{\mathcal{M}}_n \subset\mathcal{M}_n. When n=2mn=2m and c∈C×c\in\mathbb{C}^{\times}, we define a map η(c):M2m→M2m\eta(c):\mathcal{M}_{2m}\to\mathcal{M}_{2m}.

η(c)(Z)=diag⁡(c,1)⋅Z⋅diag⁡(1,c−1,…,1,c−1).(17)\eta(c)(Z)=\operatorname{diag}(c,1)\cdot Z\cdot\operatorname{diag}(1,c^{-1},\ldots,1,c^{-1}). \tag*{(17)}

For Z∈M^2mZ\in\widehat{\mathcal{M}}_{2m}, z1,3≠0z_{1,3}\ne0 because D1,3(Z)≠0D_{1,3}(Z)\ne0. Thus, η(z1,3−1)(Z)∈M^n\eta(z_{1,3}^{-1})(Z)\in\widehat{\mathcal{M}}_n. We define a map p:M^2m→M^2mp:\widehat{\mathcal{M}}_{2m}\to\widehat{\mathcal{M}}_{2m} by

p(Z)=η(z1,3−1)(Z).(18)p(Z)=\eta(z_{1,3}^{-1})(Z). \tag*{(18)}

For Z∈M^2mZ\in\widehat{\mathcal{M}}_{2m}, it is clear that p∘η(c)(Z)=Zp\circ\eta(c)(Z)=Z (c∈K×c\in\mathbb{K}^{\times}).

We now introduce a map α\alpha as follows:

α:CCF⁡n⟶M^n(fi,j)i,j∈Z⟼(fi,j)i=1,2, j=1,2,…,n.(19)\begin{aligned} \alpha:\quad&\operatorname{CCF}_n\longrightarrow\widehat{\mathcal{M}}_n\\ &(f_{i,j})_{i,j\in\mathbb{Z}}\longmapsto(f_{i,j})_{i=1,2,\ j=1,2,\ldots,n}. \tag*{(19)} \end{aligned}

Proposition 4. The inverse map α−1\alpha^{-1} can be defined as follows:

(zi,j)i=1,2, j=1,2,…,n⟼(fi,j)i,j∈Z,(20)(z_{i,j})_{i=1,2,\ j=1,2,\ldots,n}\longmapsto(f_{i,j})_{i,j\in\mathbb{Z}}, \tag*{(20)}

where

fi,j=(−1)k+ℓ(z1,i′z2,j′−z1,j′z2,i′),i=kn+i′,j=ℓn+j′,1≤i′,j′≤n.(21)f_{i,j}=(-1)^{k+\ell}(z_{1,i'}z_{2,j'}-z_{1,j'}z_{2,i'}),\qquad i=kn+i',\quad j=\ell n+j',\quad1\le i',j'\le n. \tag*{(21)}

Proof. For i=1i=1 and 1≤j≤n1\le j\le n, since k=ℓ=0k=\ell=0, formula (21) reduces to

f1,j=z1,1z2,j−z1,jz2,1=z1,j.(22)f_{1,j}=z_{1,1}z_{2,j}-z_{1,j}z_{2,1}=z_{1,j}. \tag*{(22)}

Similarly, for i=2i=2 and 1≤j≤n1\le j\le n, we obtain f2,j=z2,jf_{2,j}=z_{2,j}. Next, we consider the case 1≤i,j≤n1\le i,j\le n. By specializing the indices of the global Plücker relation (5) to k=1k=1 and ℓ=2\ell=2 and invoking the normalization (6), we can express fi,jf_{i,j} as a 2×22\times2 determinant:

fi,j=∣f1,if1,jf2,if2,j∣=∣z1,iz1,jz2,iz2,j∣.(23)f_{i,j}= \left|\begin{matrix} f_{1,i}&f_{1,j}\\ f_{2,i}&f_{2,j} \end{matrix}\right| = \left|\begin{matrix} z_{1,i}&z_{1,j}\\ z_{2,i}&z_{2,j} \end{matrix}\right|. \tag*{(23)}

The formula (21) follows from the anti-periodicity relations (9) and (10). □

To relate Mn\mathcal{M}_n and M^n\widehat{\mathcal{M}}_n, we consider the quotient space

GL2(K)\Mn/(K×)n(24)GL_2(\mathbb{K})\backslash\mathcal{M}_n/(\mathbb{K}^{\times})^n \tag*{(24)}

by the equivalence relation defined as follows: for Z1,Z2∈MnZ_1,Z_2\in\mathcal{M}_n, we say Z1∼Z2Z_1\sim Z_2 if and only if Z1=LZ2RZ_1=LZ_2R for L∈GL2(K)L\in GL_2(\mathbb{K}) and R=diag⁡(d1,…,dn)R=\operatorname{diag}(d_1,\ldots,d_n) (d1,…,dn∈K×d_1,\ldots,d_n\in\mathbb{K}^{\times}). We remark that this quotient space coincides with that used in the study of hypergeometric functions [8, 10].

When nn is even, for any matrix Z∈M2mZ\in\mathcal{M}_{2m}, there exists an invariant mr⁡(Z)\operatorname{mr}(Z) with respect to the action of GL2(K)×(K×)2mGL_2(\mathbb{K})\times(\mathbb{K}^{\times})^{2m}:

mr⁡(Z):=D2,3D4,5⋯D2m−2,2m−1D1,2mD1,2D3,4⋯D2m−3,2m−2D2m−1,2m.(25)\operatorname{mr}(Z):=\frac{D_{2,3}D_{4,5}\cdots D_{2m-2,2m-1}D_{1,2m}}{D_{1,2}D_{3,4}\cdots D_{2m-3,2m-2}D_{2m-1,2m}}. \tag*{(25)}

Using this invariant, we define a subset M2m′\mathcal{M}'_{2m} of M2m\mathcal{M}_{2m} as follows:

M2m′:={Z∈M2m∣mr⁡(Z)=1}.(26)\mathcal{M}'_{2m}:=\{Z\in\mathcal{M}_{2m}\mid\operatorname{mr}(Z)=1\}. \tag*{(26)}

The following lemma is an immediate consequence of (15).

Lemma 5. For every Z∈M^2mZ\in\widehat{\mathcal{M}}_{2m}, we have mr⁡(Z)=1\operatorname{mr}(Z)=1.

We remark that M^2m⊂M^2m⊂M2m′\widehat{\mathcal{M}}_{2m}\subset\widehat{\mathcal{M}}_{2m}\subset\mathcal{M}'_{2m}.

Remark 6. When we specialize K\mathbb{K} to C\mathbb{C}, the quotient space (24) can be interpreted in two different ways, depending on the order in which the quotients are taken. On the one hand, taking the quotient by GL2(C)GL_2(\mathbb{C}) first yields the quotient of the Grassmannian Gr(2,n)Gr(2,n) by the right action of (C×)n(\mathbb{C}^{\times})^n. On the other hand, taking the quotient by (C×)n(\mathbb{C}^{\times})^n first yields the quotient of the configuration space of nn points on the complex projective line P1\mathbb{P}^1 by the left action of GL2(C)GL_2(\mathbb{C}) (cf. [2]). The quotient space M2m′/(C×)2m\mathcal{M}_{2m}'/(\mathbb{C}^{\times})^{2m} was introduced in [13], where it is denoted by C2m\mathcal{C}_{2m}. The invariant mr⁡(Z)\operatorname{mr}(Z) in (25) is known as the multi-ratio associated with a configuration of 2m2m points on the projective line P1\mathbb{P}^1. In the six-point case, this multi-ratio was studied by Bobenko et al. in their work on integrable circle patterns [3, 4].

Theorem 7. Let n≥4n \ge4 be an integer and Z∈MnZ \in\mathcal{M}_n.

(i) For odd nn, there exists a unique matrix Z^∈M^n\widehat{Z} \in\widehat{\mathcal{M}}_n such that Z∼Z^Z \sim\widehat{Z}.

(ii) For even nn, a unique matrix Z^∈M^n\widehat{Z} \in\widehat{\mathcal{M}}_n such that Z∼Z^Z \sim\widehat{Z} exists if and only if Z∈Mn′Z \in\mathcal{M}_n'.

We refer to this Z^\widehat{Z} as the normal form of ZZ.

Proof. We fix a matrix Z=(v1,v2,…,vn)∈MnZ=(v_1,v_2,\ldots,v_n)\in\mathcal{M}_n, where vi=(ai,bi)Tv_i=(a_i,b_i)^{\mathsf{T}} (i=1,…,n)(i=1,\ldots,n). Let L1,L2L_1,L_2 and RR be given by

L1=(−b1a1−b2a2),L2=diag⁡(μ1,μ2),R=diag⁡(λ1,…,λn),(27)L_1=\begin{pmatrix}-b_1 & a_1\\ -b_2 & a_2\end{pmatrix},\qquad L_2=\operatorname{diag}(\mu_1,\mu_2),\qquad R=\operatorname{diag}(\lambda_1,\ldots,\lambda_n), \tag*{(27)}
μ1,μ2,λ1,…,λn∈K×,\mu_1,\mu_2,\lambda_1,\ldots,\lambda_n\in\mathbb{K}^{\times},

Then we have

L2L1ZR=(0μ1λ2D1,2μ1λ3D1,3⋯μ1λnD1,n−μ2λ1D1,20μ2λ3D2,3⋯μ2λnD2,n),(28)L_2L_1ZR= \begin{pmatrix} 0 & \mu_1\lambda_2D_{1,2} & \mu_1\lambda_3D_{1,3} & \cdots& \mu_1\lambda_nD_{1,n}\\ -\mu_2\lambda_1D_{1,2} & 0 & \mu_2\lambda_3D_{2,3} & \cdots& \mu_2\lambda_nD_{2,n} \end{pmatrix}, \tag*{(28)}

where Di,j=Di,j(Z)D_{i,j}=D_{i,j}(Z). For simplicity, we write Z~:=L2L1ZR\widetilde{Z}:=L_2L_1ZR. To satisfy the normalization conditions (15) for the first three columns of M^n\widehat{\mathcal{M}}_n, the values of λ1\lambda_1, λ2\lambda_2 and λ3\lambda_3 are uniquely determined as λ1=1/(μ2D1,2)\lambda_1=1/(\mu_2D_{1,2}), λ2=1/(μ1D1,2)\lambda_2=1/(\mu_1D_{1,2}), and λ3=1/(μ2D2,3)\lambda_3=1/(\mu_2D_{2,3}). Furthermore, the conditions (15) for adjacent minors,

μ1μ2λiλi+1(D1,iD2,i+1−D1,i+1D2,i)=μ1μ2λiλi+1D1,iD2,i+1=1(29)\mu_1\mu_2\lambda_i\lambda_{i+1}(D_{1,i}D_{2,i+1}-D_{1,i+1}D_{2,i})=\mu_1\mu_2\lambda_i\lambda_{i+1}D_{1,i}D_{2,i+1}=1 \tag*{(29)}

allow us to recursively solve for λk\lambda_k. For k=4,…,nk=4,\ldots,n, the general form of λk\lambda_k is given by:

λk={1μ2⋅D3,4D5,6⋯Dk−2,k−1D2,3D4,5⋯Dk−1,k,(if k is odd),1μ1⋅D2,3D4,5⋯Dk−2,k−1D1,2D3,4⋯Dk−1,k,(if k is even).(30)\lambda_k= \begin{cases} \frac{1}{\mu_2}\cdot\frac{D_{3,4}D_{5,6}\cdots D_{k-2,k-1}}{D_{2,3}D_{4,5}\cdots D_{k-1,k}}, & \text{(if } k \text{ is odd)},\\[1.2em] \frac{1}{\mu_1}\cdot\frac{D_{2,3}D_{4,5}\cdots D_{k-2,k-1}}{D_{1,2}D_{3,4}\cdots D_{k-1,k}}, & \text{(if } k \text{ is even)}. \end{cases} \tag*{(30)}

By substituting these values into (28), the matrix Z~\widetilde{Z} is explicitly written as:

Z~={(01μ1D1,3μ2D2,3D2,3D1,4D1,2D3,4⋯μ1D3,4D5,6⋯Dn−2,n−1D1,nμ2D2,3D4,5⋯Dn−3,n−2Dn−1,n−101μ2D2,3D2,4μ1D1,2D3,4⋯D3,4D5,6⋯Dn−2,n−1D2,nD2,3D4,5⋯Dn−3,n−2Dn−1,n),(if n is odd),(01μ1D1,3μ2D2,3D2,3D1,4D1,2D3,4⋯mr⁡(Z)−101μ2D2,3D2,4μ1D1,2D3,4⋯μ2D2,3D4,5⋯Dn−2,n−1D2,nμ1D1,2D3,4⋯Dn−3,n−2Dn−1,n),(if n is even).(31)\widetilde{Z}= \begin{cases} \begin{pmatrix} 0 & 1 & \dfrac{\mu_1D_{1,3}}{\mu_2D_{2,3}} & \dfrac{D_{2,3}D_{1,4}}{D_{1,2}D_{3,4}} & \cdots& \dfrac{\mu_1D_{3,4}D_{5,6}\cdots D_{n-2,n-1}D_{1,n}}{\mu_2D_{2,3}D_{4,5}\cdots D_{n-3,n-2}D_{n-1,n}}\\ -1 & 0 & 1 & \dfrac{\mu_2D_{2,3}D_{2,4}}{\mu_1D_{1,2}D_{3,4}} & \cdots& \dfrac{D_{3,4}D_{5,6}\cdots D_{n-2,n-1}D_{2,n}}{D_{2,3}D_{4,5}\cdots D_{n-3,n-2}D_{n-1,n}} \end{pmatrix}, & \text{(if } n \text{ is odd)},\\[1.5em] \begin{pmatrix} 0 & 1 & \dfrac{\mu_1D_{1,3}}{\mu_2D_{2,3}} & \dfrac{D_{2,3}D_{1,4}}{D_{1,2}D_{3,4}} & \cdots& \operatorname{mr}(Z)\\ -1 & 0 & 1 & \dfrac{\mu_2D_{2,3}D_{2,4}}{\mu_1D_{1,2}D_{3,4}} & \cdots& \dfrac{\mu_2D_{2,3}D_{4,5}\cdots D_{n-2,n-1}D_{2,n}}{\mu_1D_{1,2}D_{3,4}\cdots D_{n-3,n-2}D_{n-1,n}} \end{pmatrix}, & \text{(if } n \text{ is even)}. \end{cases} \tag*{(31)}

where both cases satisfy the conditions in (15) except for the (1,n)(1,n)-entry. If nn is odd, we can set

μ1μ2=D2,3D4,5⋯Dn−3,n−2Dn−1,nD3,4D5,6⋯Dn−2,n−1D1,n(32)\frac{\mu_1}{\mu_2}=\frac{D_{2,3}D_{4,5}\cdots D_{n-3,n-2}D_{n-1,n}}{D_{3,4}D_{5,6}\cdots D_{n-2,n-1}D_{1,n}} \tag*{(32)}

to make the (1,n)(1,n)-entry of Z~\widetilde{Z} equal to 11. If nn is even, we require the condition mr⁡(Z)=1\operatorname{mr}(Z)=1, which means that Z∈Mn′Z\in\mathcal{M}'_n. Furthermore, we set

μ1μ2=D2,3D1,3(33)\frac{\mu_1}{\mu_2}=\frac{D_{2,3}}{D_{1,3}} \tag*{(33)}

to make the (1,3)(1,3)-entry of Z~\widetilde{Z} equal to 11, which means Z~∈M^n\widetilde{Z}\in\widehat{\mathcal{M}}_n.

We then address the uniqueness of the normal form. Suppose there exist two representatives Z^1,Z^2∈M^n\widehat{Z}_1,\widehat{Z}_2\in\widehat{\mathcal{M}}_n (resp. M^n\widehat{\mathcal{M}}_n) such that Z∼Z^1∼Z^2Z\sim\widehat{Z}_1\sim\widehat{Z}_2. This implies Z^2=LZ^1R\widehat{Z}_2=L\widehat{Z}_1R for some L∈GL⁡2(K)L\in\operatorname{GL}_2(\mathbb{K}) and R=diag⁡(λ1,…,λn)R=\operatorname{diag}(\lambda_1,\ldots,\lambda_n). From the first and second columns of (15), we see that LL must be a diagonal matrix diag⁡(1/λ2,1/λ1)\operatorname{diag}(1/\lambda_2,1/\lambda_1). Since Z^2∈M^n\widehat{Z}_2\in\widehat{\mathcal{M}}_n, we obtain the following nn relations:

1=Di,i+1(Z^2)=(λiλi+1/λ1λ2)Di,i+1(Z^1)=λiλi+1/λ1λ2(1≤i≤n−1),(34)1=D_{i,i+1}(\widehat{Z}_2)=(\lambda_i\lambda_{i+1}/\lambda_1\lambda_2)D_{i,i+1}(\widehat{Z}_1)=\lambda_i\lambda_{i+1}/\lambda_1\lambda_2\qquad(1\leq i\leq n-1), \tag*{(34)}
1=D1,n(Z^2)=(λ1λn/λ1λ2)D1,n(Z^1)=λn/λ2.(35)1=D_{1,n}(\widehat{Z}_2)=(\lambda_1\lambda_n/\lambda_1\lambda_2)D_{1,n}(\widehat{Z}_1)=\lambda_n/\lambda_2. \tag*{(35)}

Moreover, if Z^2∈M^n\widehat{Z}_2\in\widehat{\mathcal{M}}_n, we obtain the following relation from the condition on the (1,3)(1,3)-entry:

1=λ3λ2.(36)1=\frac{\lambda_3}{\lambda_2}. \tag*{(36)}

The equations (34) imply the recurrence λi+1=λ1λ2/λi\lambda_{i+1}=\lambda_1\lambda_2/\lambda_i, which yields

λ1=λ3=λ5=⋯ ,λ2=λ4=λ6=⋯ .(37)\lambda_1=\lambda_3=\lambda_5=\cdots,\qquad\lambda_2=\lambda_4=\lambda_6=\cdots. \tag*{(37)}

If nn is odd, it follows from (37) that λn=λ1\lambda_n=\lambda_1. Then, by (35), we have λ1=λ2=⋯=λn\lambda_1=\lambda_2=\cdots=\lambda_n. On the other hand, if Z^2∈M^n\widehat{Z}_2\in\widehat{\mathcal{M}}_n, then (36) yields λ2=λ3\lambda_2=\lambda_3, and again by (37), we obtain λ1=λ2=⋯=λn\lambda_1=\lambda_2=\cdots=\lambda_n. Therefore, in either case, LL and RR can be expressed as L=(1/λ)IL=(1/\lambda)I and R=λIR=\lambda I for some λ∈K×\lambda\in\mathbb{K}^{\times}, which implies Z^2=LZ^1R=Z^1\widehat{Z}_2=L\widehat{Z}_1R=\widehat{Z}_1. □

Let φodd:M2m+1→M^2m+1\varphi_{\mathrm{odd}}:\mathcal{M}_{2m+1}\to\widehat{\mathcal{M}}_{2m+1} (resp. φeven:M2m′→M^2m\varphi_{\mathrm{even}}:\mathcal{M}'_{2m}\to\widehat{\mathcal{M}}_{2m}) be the map that assigns Z∈M2m+1Z\in\mathcal{M}_{2m+1} (resp. M2m′\mathcal{M}'_{2m}) to its normal form in the sense of Theorem 7. More explicitly, for Z∈M2m+1Z\in\mathcal{M}_{2m+1}, we can write

φodd(Z):=(01z^1,3z^1,4⋯z^1,2m1−101z^2,4⋯z^2,2mz^2,2m+1),\varphi_{\mathrm{odd}}(Z):= \begin{pmatrix} 0&1&\widehat{z}_{1,3}&\widehat{z}_{1,4}&\cdots&\widehat{z}_{1,2m}&1\\ -1&0&1&\widehat{z}_{2,4}&\cdots&\widehat{z}_{2,2m}&\widehat{z}_{2,2m+1} \end{pmatrix},
z^1,k={Dk+1,k+2Dk+3,k+4⋯D2m,2m+1D1,kDk,k+1Dk+2,k+3⋯D2m−1,2mD1,2m+1(if k is odd),D2,3D4,5⋯Dk−2,k−1D1,kD1,2D3,4⋯Dk−3,k−2Dk−1,k(if k is even),\widehat{z}_{1,k}=\begin{cases}\frac{D_{k+1,k+2}D_{k+3,k+4}\cdots D_{2m,2m+1}D_{1,k}}{D_{k,k+1}D_{k+2,k+3}\cdots D_{2m-1,2m}D_{1,2m+1}}&\text{(if }k\text{ is odd)},\\[1.2em]\frac{D_{2,3}D_{4,5}\cdots D_{k-2,k-1}D_{1,k}}{D_{1,2}D_{3,4}\cdots D_{k-3,k-2}D_{k-1,k}}&\text{(if }k\text{ is even)},\end{cases}
z^2,k={D3,4D5,6⋯Dk−2,k−1D2,kD2,3D4,5⋯Dk−3,k−2Dk−1,k(if k is odd),Dk+1,k+2Dk+3,k+4⋯D2m−1,2mD1,2m+1D2,kDk,k+1Dk+2,k+3⋯D2m−2,2m−2D2m,2m+1D1,2(if k is even).(38)\widehat{z}_{2,k}=\begin{cases}\frac{D_{3,4}D_{5,6}\cdots D_{k-2,k-1}D_{2,k}}{D_{2,3}D_{4,5}\cdots D_{k-3,k-2}D_{k-1,k}}&\text{(if }k\text{ is odd)},\\[1.2em]\frac{D_{k+1,k+2}D_{k+3,k+4}\cdots D_{2m-1,2m}D_{1,2m+1}D_{2,k}}{D_{k,k+1}D_{k+2,k+3}\cdots D_{2m-2,2m-2}D_{2m,2m+1}D_{1,2}}&\text{(if }k\text{ is even)}.\end{cases} \tag*{(38)}

and for Z′∈M2m′Z' \in\mathcal{M}'_{2m},

φeven(Z′):=(011w^1,4⋯w^1,2m−11−101w^2,4⋯w^2,2m−1w^2,2m),w^1,k={D3,4D4,5⋯Dk−2,k−1D1,kD1,3D4,5⋯Dk−3,k−2Dk−1,k(if k is odd),D2,3D4,5⋯Dk−2,k−1D1,kD1,2D3,4⋯Dk−3,k−2Dk−1,k(if k is even),w^2,k={D3,4D5,6⋯Dk−2,k−1D2,kD2,3D4,5⋯Dk−3,k−2Dk−1,k(if k is odd),D1,3D4,5⋯Dk−2,k−1D2,kD1,2D3,4⋯Dk−3,k−2Dk−1,k(if k is even).(39)\begin{aligned} \varphi_{\mathrm{even}}(Z')&:=\left(\begin{array}{cccccccc} 0&1&1&\widehat{w}_{1,4}&\cdots&\widehat{w}_{1,2m-1}&1\\ -1&0&1&\widehat{w}_{2,4}&\cdots&\widehat{w}_{2,2m-1}&\widehat{w}_{2,2m} \end{array}\right),\\ \widehat{w}_{1,k}&=\begin{cases} \dfrac{D_{3,4}D_{4,5}\cdots D_{k-2,k-1}D_{1,k}}{D_{1,3}D_{4,5}\cdots D_{k-3,k-2}D_{k-1,k}}&\text{(if $k$ is odd)},\\[6pt] \dfrac{D_{2,3}D_{4,5}\cdots D_{k-2,k-1}D_{1,k}}{D_{1,2}D_{3,4}\cdots D_{k-3,k-2}D_{k-1,k}}&\text{(if $k$ is even)}, \end{cases}\\ \widehat{w}_{2,k}&=\begin{cases} \dfrac{D_{3,4}D_{5,6}\cdots D_{k-2,k-1}D_{2,k}}{D_{2,3}D_{4,5}\cdots D_{k-3,k-2}D_{k-1,k}}&\text{(if $k$ is odd)},\\[6pt] \dfrac{D_{1,3}D_{4,5}\cdots D_{k-2,k-1}D_{2,k}}{D_{1,2}D_{3,4}\cdots D_{k-3,k-2}D_{k-1,k}}&\text{(if $k$ is even)}. \end{cases} \tag*{(39)} \end{aligned}

Remark 8. A formula similar to the expressions for z^1,k\widehat{z}_{1,k} and z^2,k\widehat{z}_{2,k} in (38) has also appeared in the study of string theory (see [9], (2.10)).

Action of the symmetric group

Based on Theorem 7, the map φodd:M2m+1→M^2m+1\varphi_{\mathrm{odd}}:\mathcal{M}_{2m+1}\to\widehat{\mathcal{M}}_{2m+1}, which assigns each matrix ZZ to its unique normal form Z^\widehat{Z}, is well-defined. For σ∈S2m+1\sigma\in\mathfrak{S}_{2m+1}, define ρ(σ):M^2m+1→M^2m+1\rho(\sigma):\widehat{\mathcal{M}}_{2m+1}\to\widehat{\mathcal{M}}_{2m+1} as

ρ(σ)(Z)=φodd(ZPσ),(40)\rho(\sigma)(Z)=\varphi_{\mathrm{odd}}(ZP_\sigma), \tag*{(40)}

where PσP_\sigma is the permutation matrix corresponding to σ\sigma. We define a map ρ~(σ):CCF2m+1→CCF2m+1\widetilde{\rho}(\sigma):\mathcal{CCF}_{2m+1}\to\mathcal{CCF}_{2m+1} as

ρ~(σ):=α−1∘ρ(σ)∘α.(41)\widetilde{\rho}(\sigma):=\alpha^{-1}\circ\rho(\sigma)\circ\alpha. \tag*{(41)}

This action yields a representation of the symmetric group on CCF2m+1\mathcal{CCF}_{2m+1}. In what follows, let σij\sigma_{ij} denote the transposition (i,j)(i,j) for i,j=1,…,2m+1i,j=1,\ldots,2m+1.

Example 9. As an example, we consider the following F∈CCF5F\in\mathcal{CCF}_5:

F=⋯(01xy10−1−x−y−1−1011+yx1+x+yxy10−1−1+yx−1+x+yxy−x−1011+xyx10−1−1+xy−y−1+y1+x+y−101y1+y1+x+y10−1−1−1+x+yxy−1+xy−1011+x+yxy1+xy10)⋯F=\cdots \begin{pmatrix} 0&1&x&y&1&0&-1&-x&-y&-1\\ -1&0&1&\dfrac{1+y}{x}&\dfrac{1+x+y}{xy}&1&0&-1&-\dfrac{1+y}{x}&-\dfrac{1+x+y}{xy}\\ -x&-1&0&1&\dfrac{1+x}{y}&x&1&0&-1&-\dfrac{1+x}{y}\\ -y&-\dfrac{1+y}{1+x+y}&-1&0&1&y&\dfrac{1+y}{1+x+y}&1&0&-1\\ -1&-\dfrac{1+x+y}{xy}&-\dfrac{1+x}{y}&-1&0&1&\dfrac{1+x+y}{xy}&\dfrac{1+x}{y}&1&0 \end{pmatrix} \cdots

For simplicity, only the upper triangular part of the first 5×55 \times5 block of ρ~(σ).F\widetilde{\rho}(\sigma).F is shown below:

ρ~(σ12).F=(01xy1+x+y−(1+y)101−1+x+yx1x001−1+x1+x+y00010000),\widetilde{\rho}(\sigma_{12}).F = \begin{pmatrix} 0 & 1 & \frac{xy}{1+x+y} & -(1+y) & 1 \\ 0 & 1 & -\frac{1+x+y}{x} & \frac{1}{x} \\ 0 & 0 & 1 & -\frac{1+x}{1+x+y} \\ 0 & 0 & 0 & 1 \\ 0 & 0 & 0 & 0 \end{pmatrix},
ρ~(σ23).F=(01x1+y−y1+y1011x−(1+x)(1+y)xy001−1+x+yy00010000),\widetilde{\rho}(\sigma_{23}).F = \begin{pmatrix} 0 & 1 & \frac{x}{1+y} & -\frac{y}{1+y} & 1 \\ 0 & 1 & \frac{1}{x} & -\frac{(1+x)(1+y)}{xy} \\ 0 & 0 & 1 & -\frac{1+x+y}{y} \\ 0 & 0 & 0 & 1 \\ 0 & 0 & 0 & 0 \end{pmatrix},
ρ~(σ34).F=(01−(1+x)−(1+y)101y1+x−1+x+y(1+x)(1+y)001x1+y00010000),\widetilde{\rho}(\sigma_{34}).F = \begin{pmatrix} 0 & 1 & -(1+x) & -(1+y) & 1 \\ 0 & 1 & \frac{y}{1+x} & -\frac{1+x+y}{(1+x)(1+y)} \\ 0 & 0 & 1 & \frac{x}{1+y} \\ 0 & 0 & 0 & 1 \\ 0 & 0 & 0 & 0 \end{pmatrix},
ρ~(σ45).F=(01−x1+xy1+x101−1+x+yx−(1+x)(1+y)xy0011y00010000).\widetilde{\rho}(\sigma_{45}).F = \begin{pmatrix} 0 & 1 & -\frac{x}{1+x} & \frac{y}{1+x} & 1 \\ 0 & 1 & -\frac{1+x+y}{x} & -\frac{(1+x)(1+y)}{xy} \\ 0 & 0 & 1 & \frac{1}{y} \\ 0 & 0 & 0 & 1 \\ 0 & 0 & 0 & 0 \end{pmatrix}.

In the case of φeven:M^2m′→M^2m\varphi_{\mathrm{even}}:\widehat{\mathcal{M}}_{2m}' \to\widehat{\mathcal{M}}_{2m}, the action (40) does not work in general; even if mr⁡(Z′)=1\operatorname{mr}(Z')=1 (i.e. Z′∈M2m′Z' \in\mathcal{M}_{2m}'), the condition mr⁡(Z′Pσ)=1\operatorname{mr}(Z'P_{\sigma})=1 need not hold for every σ∈S2m\sigma\in\mathfrak{S}_{2m}. Define rr and tt as follows:

r=σ1,2mσ1,2m−1⋯σ12,t=σ2,2mσ3,2m−1⋯σm,m+2.(42)r=\sigma_{1,2m}\sigma_{1,2m-1}\cdots\sigma_{12},\qquad t=\sigma_{2,2m}\sigma_{3,2m-1}\cdots\sigma_{m,m+2}. \tag*{(42)}

They generate the dihedral group D2m⊂S2m\mathcal{D}_{2m}\subset\mathfrak{S}_{2m}. By the definition of the multi-ratio (25), if mr⁡(Z′)=1\operatorname{mr}(Z')=1, then

mr⁡(Z′Pr)=mr⁡(Z′Pt)=1.(43)\operatorname{mr}(Z'P_r)=\operatorname{mr}(Z'P_t)=1. \tag*{(43)}

Thus we can define the action of r,tr,t on M^2m\widehat{\mathcal{M}}_{2m}:

ρ(r)(Z′)=φeven(Z′Pr),ρ(t)(Z′)=φeven(Z′Pt),Z′∈M^2m.(44)\rho(r)(Z')=\varphi_{\mathrm{even}}(Z'P_r),\qquad\rho(t)(Z')=\varphi_{\mathrm{even}}(Z'P_t),\qquad Z'\in\widehat{\mathcal{M}}_{2m}. \tag*{(44)}

For an element σ\sigma of the dihedral group D2m\mathcal{D}_{2m} and a parameter c∈K×c\in K^{\times}, we define the map ρ(σ,c):M^2m→M^2m\rho(\sigma,c):\widehat{\mathcal{M}}_{2m}\to\widehat{\mathcal{M}}_{2m} as follows:

ρ(σ,c):=η(c)∘ρ(σ)∘p.(45)\rho(\sigma,c):=\eta(c)\circ\rho(\sigma)\circ p. \tag*{(45)}

Proposition 10. For any parameters c1,…,c2m∈K×c_1,\ldots,c_{2m}\in K^\times and Z∈M^2mZ\in\widehat{\mathcal{M}}_{2m}, the following relations hold:

ρ(r,c2m)⋯ρ(r,c1)(Z)=η(c2mx1,3(Z))(Z),(46)\rho(r,c_{2m})\cdots\rho(r,c_1)(Z)=\eta\left(\frac{c_{2m}}{x_{1,3}(Z)}\right)(Z), \tag*{(46)}
ρ(t,c2)ρ(t,c1)(Z)=η(c2x1,3(Z))(Z),(47)\rho(t,c_2)\rho(t,c_1)(Z)=\eta\left(\frac{c_2}{x_{1,3}(Z)}\right)(Z), \tag*{(47)}
ρ(r,c4)ρ(t,c3)ρ(r,c2)ρ(t,c1)(Z)=η(c4x1,3(Z))(Z).(48)\rho(r,c_4)\rho(t,c_3)\rho(r,c_2)\rho(t,c_1)(Z)=\eta\left(\frac{c_4}{x_{1,3}(Z)}\right)(Z). \tag*{(48)}

Let i:M^^2m↪M^2mi:\widehat{\widehat{\mathcal{M}}}_{2m}\hookrightarrow\widehat{\mathcal{M}}_{2m} be the natural inclusion map defined by i(Z)=Zi(Z)=Z. We define an equivalence relation on M^2m\widehat{\mathcal{M}}_{2m} as follows: for Z1,Z2∈M^2mZ_1,Z_2\in\widehat{\mathcal{M}}_{2m}, Z1∼matZ2Z_1\sim_{\mathrm{mat}}Z_2 if and only if there exists c∈K×c\in K^\times such that η(c)(Z1)=Z2\eta(c)(Z_1)=Z_2. We introduce the quotient space M^2m/∼mat\widehat{\mathcal{M}}_{2m}/\sim_{\mathrm{mat}} and a projection map π:M^2m′→M^2m/∼mat\pi:\widehat{\mathcal{M}}'_{2m}\to\widehat{\mathcal{M}}_{2m}/\sim_{\mathrm{mat}}. The following diagram illustrates these maps:

M^^2m→iM^2m→πM^2m/∼mat.(49)\widehat{\widehat{\mathcal{M}}}_{2m}\xrightarrow{i}\widehat{\mathcal{M}}_{2m}\xrightarrow{\pi}\widehat{\mathcal{M}}_{2m}/\sim_{\mathrm{mat}}. \tag*{(49)}

It follows that M^2m/∼mat≅M^2m\widehat{\mathcal{M}}_{2m}/\sim_{\mathrm{mat}}\cong\widehat{\mathcal{M}}_{2m}. The action of ρ(σ,c)\rho(\sigma,c) on M^2m/∼mat\widehat{\mathcal{M}}_{2m}/\sim_{\mathrm{mat}} is independent of c∈K×c\in K^\times. It follows that ρ(σ,c)\rho(\sigma,c) induces a representation of D2mD_{2m} on M^2m/∼mat\widehat{\mathcal{M}}_{2m}/\sim_{\mathrm{mat}}.

For m=3m=3, the set CCF6CCF_6 exhibits enhanced symmetry associated with the Weyl group W(B3)W(B_3) (≅W(C3)\cong W(C_3)):

D6⊂W(B3)⊂S6.(50)D_6\subset W(B_3)\subset S_6. \tag*{(50)}

We define

g1=σ14,g2=σ16σ34,g3=σ26σ35=t,(51)g_1=\sigma_{14},\qquad g_2=\sigma_{16}\sigma_{34},\qquad g_3=\sigma_{26}\sigma_{35}=t, \tag*{(51)}

which satisfy

gi2=id⁡ (i=1,2,3),(g1g3)2=(g1g2)3=(g2g3)4=id⁡,r=g2g1g2g3g2,(52)g_i^2=\operatorname{id}\ (i=1,2,3),\qquad(g_1g_3)^2=(g_1g_2)^3=(g_2g_3)^4=\operatorname{id},\qquad r=g_2g_1g_2g_3g_2, \tag*{(52)}

and hence g1g_1, g2g_2, and g3g_3 generate W(B3)W(B_3).

Proposition 11 (Affolter, George and Ramassamy [1], Lemma 4.2). Let ZZ be a matrix in M6′\mathcal{M}'_6. For every permutation σ\sigma in S6S_6 such that

σ(i+3)≡σ(i)+3(mod6)(1≤i≤6),(53)\sigma(i+3)\equiv\sigma(i)+3\pmod6\qquad(1\leq i\leq6), \tag*{(53)}

we have mr⁡(ZPσ)=1\operatorname{mr}(ZP_\sigma)=1, i.e. ZPσ∈M6′ZP_\sigma\in\mathcal{M}'_6.

Remark 12. An equivalent discussion can also be found in Proposition 2.1 of [3].

Since g1g_1, g2g_2, and g3g_3 in (51) satisfy the condition (53), Proposition 11 implies that mr⁡(ZPg)=1\operatorname{mr}(ZP_g)=1 holds for all Z∈M2m′Z\in\mathcal{M}'_{2m} and g∈W(B3)g\in W(B_3). Therefore, for m=3m=3, the Weyl group W(B3)W(B_3), which is larger than the dihedral group D6D_6, acts on M^6/∼mat≅M^6\widehat{\mathcal{M}}_6/\sim_{\mathrm{mat}}\cong\widehat{\mathcal{M}}_6.

Conjecture 13. If m≠3m\ne3, the symmetry group of M^2m\widehat{\mathcal{M}}_{2m} coincides with the dihedral group D2mD_{2m}.

Using the computer algebra systems SageMath, Maxima, and GAP, we have verified that Conjecture 13 holds for m=2m=2, 44, 55, and 66.

In what follows, we define the map ρˇ(σ,c):CCF2m→CCF2m\check{\rho}(\sigma,c):CCF_{2m}\to CCF_{2m} by

ρˇ(σ,c):=α−1∘ρ(σ,c)∘α.(54)\check{\rho}(\sigma,c):=\alpha^{-1}\circ\rho(\sigma,c)\circ\alpha. \tag*{(54)}

The construction of ρˇ\check{\rho} can be summarized as follows:

CCF2m→αM^2m→pM^^2mρ~(σ,c)↓ρ~(σ,c)↓↓ρ(σ)CCF2m←α−1M^2m←η(c)M^^2m\begin{CD} CCF_{2m} @>\alpha>> \widehat{\mathcal{M}}_{2m} @>p>> \widehat{\widehat{\mathcal{M}}}_{2m}\\ @V\tilde{\rho}(\sigma,c)VV @V\tilde{\rho}(\sigma,c)VV @VV\rho(\sigma)V\\ CCF_{2m} @<\alpha^{-1}<< \widehat{\mathcal{M}}_{2m} @<\eta(c)<< \widehat{\widehat{\mathcal{M}}}_{2m} \end{CD}

Example 14. As an example, we consider the following F∈CCF6F \in CCF_{6}:

F=⋯(01xyz1−1011+yxx+z+yzxy(x+z)(1+y)xyz−x−101x+zyx+xy+zyz−y−1+yx−1011+yz−z−x+z+yzxy−x+zy−101−1−(x+z)(1+y)xyz−x+z+xyyz−1+yz−10)⋯F=\cdots \begin{pmatrix} 0 & 1 & x & y & z & 1\\ -1 & 0 & 1 & \frac{1+y}{x} & \frac{x+z+yz}{xy} & \frac{(x+z)(1+y)}{xyz}\\ -x & -1 & 0 & 1 & \frac{x+z}{y} & \frac{x+xy+z}{yz}\\ -y & -\frac{1+y}{x} & -1 & 0 & 1 & \frac{1+y}{z}\\ -z & -\frac{x+z+yz}{xy} & -\frac{x+z}{y} & -1 & 0 & 1\\ -1 & -\frac{(x+z)(1+y)}{xyz} & -\frac{x+z+xy}{yz} & -\frac{1+y}{z} & -1 & 0 \end{pmatrix} \cdots

For simplicity, only the upper triangular part of the first 6×66 \times6 block of ρ~(σ).F\tilde{\rho}(\sigma).F is shown below:

ρ~(r,c).F=(01c(1+y)xx+z+yzxyc(x+z)(1+y)xyz101x+zcyx+z+xyyzxc01c(1+y)zyz01zc01),\tilde{\rho}(r,c).F= \begin{pmatrix} 0&1&\frac{c(1+y)}{x}&\frac{x+z+yz}{xy}&\frac{c(x+z)(1+y)}{xyz}&1\\ 0&1&\frac{x+z}{cy}&\frac{x+z+xy}{yz}&\frac{x}{c}&\\ 0&1&\frac{c(1+y)}{z}&\frac{y}{z}&\\ 0&1&\frac{z}{c}&\\ 0&1 \end{pmatrix},
ρ~(g3,c).F=ρ~(t,c).F=(01−czy−cx101−1+yczx+z+xyyz−(x+z)(1+y)cxyz01−c(x+z)yx+z+yzxy−1+ycx011−1+ycx01),\tilde{\rho}(g_{3},c).F=\tilde{\rho}(t,c).F= \begin{pmatrix} 0&1&-cz&y&-cx&1\\ 0&1&-\frac{1+y}{cz}&\frac{x+z+xy}{yz}&-\frac{(x+z)(1+y)}{cxyz}&\\ 0&1&-\frac{c(x+z)}{y}&\frac{x+z+yz}{xy}&-\frac{1+y}{cx}&\\ 0&1&1&-\frac{1+y}{cx}&\\ 0&1 \end{pmatrix},
ρ~(g1,c).F=(01−c−y1+y−cxz101−1c(1+y)−x+z+yzyz−1+xzcxy01−(1+x)(1+y)xy−1+xz1+x011x1+x01),\tilde{\rho}(g_{1},c).F= \begin{pmatrix} 0&1&-c&-\frac{y}{1+y}&-\frac{cx}{z}&1\\ 0&1&-\frac{1}{c(1+y)}&-\frac{x+z+yz}{yz}&-\frac{1+xz}{cxy}&\\ 0&1&-\frac{(1+x)(1+y)}{xy}&-\frac{1+xz}{1+x}&\\ 0&1&1&\frac{x}{1+x}&\\ 0&1 \end{pmatrix},
ρ~(g2,c).F=(01−cxz−x+z+xyx+zcxy(x+z)(1+y)101−1c(x+z)−yzc(x+z)x+z+yz(x+z)(1+y)1c01−cx(1+y)zx+zz011−1+yc01).\tilde{\rho}(g_{2},c).F= \begin{pmatrix} 0&1&-\frac{cx}{z}&-\frac{x+z+xy}{x+z}&\frac{cxy}{(x+z)(1+y)}&1\\ 0&1&-\frac{1}{c(x+z)}&-\frac{yz}{c(x+z)}&\frac{x+z+yz}{(x+z)(1+y)}&\frac{1}{c}\\ 0&1&-\frac{cx}{(1+y)z}&\frac{x+z}{z}&\\ 0&1&1&-\frac{1+y}{c}&\\ 0&1 \end{pmatrix}.

Remark 15. It is a natural direction to extend our matrix realization approach to SLkSL_k-frieze patterns by adopting the framework introduced by Surmann [16]. Representing SLkSL_k-frieze patterns via k×nk \times n matrices (or elements in Gr(k,n)Gr(k,n)) provides a foundation for analyzing actions of Sn\mathbb{S}_n and its subgroups. Generalizing our classification of symmetry groups to the SLkSL_k case remains an interesting topic for future research.

Acknowledgments

The authors acknowledge Yasuyuki Hatsuda, Masatoshi Noumi and Yoshihisa Saito for discussions on normal forms of elements in M^n\widehat{\mathcal{M}}_n. Thanks are also due to Rei Inoue, Takeyoshi Kogiso and Kyo Nishiyama for providing information on frieze patterns.

During the preparation of this work, the authors used OpenAI Prism to assist in the preparation and editing of this manuscript, and Google Gemini to improve the clarity and style of the English, as well as to double-check the mathematical computations and ensure accuracy. After using these tools, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Department of Mathematics, Graduate School of Science, Rikkyo University, Toshima-ku, Tokyo 171-8501, Japan.

Email address: [email protected]

Department of Mathematics, College of Science, Rikkyo University, Toshima-ku, Tokyo 171-8501, Japan.

Email address: [email protected]

References

  1. [1]N.C. Affolter, T. George and S. Ramassamy, Integrable dynamics in projective geometry via dimers and triple crossing diagram maps on the cylinder, SIGMA 21 (2025), 040.DOI
  2. [2]K. Aomoto, On the structure of integrals of power product linear functions, Sci. Papers. Gen. Univ. Tokyo 27 (1977), 49–61.DOI
  3. [3]A.I. Bobenko and T. Hoffmann, Conformally symmetric circle packings: a generalization of Doyle’s spirals, Experiment. Math. 10 (2001), 141–150.DOI
  4. [4]A.I. Bobenko, T. Hoffmann, and Yu.B. Suris, Hexagonal circle patterns and integrable systems: Patterns with the multi-ratio property and Lax equations on the regular triangular lattice, Internat. Math. Res. Notices 2002, 111–164.
  5. [5]J.H. Conway and H.S.M. Coxeter, Triangulated polygons and frieze patterns, Math. Gaz. 57 (1973), 87–94.
  6. [6]H.S.M. Coxeter, Frieze patterns, Acta Arithmetica 18 (1971), 297–310.DOI
  7. [7]S. Fomin and A. Zelevinsky, Cluster algebras I: foundations, J. Amer. Math. Soc. 15 (2002), 497–529.DOI
  8. [8]I.M. Gel’fand, M.I. Graev and V.S. Retakh, General hypergeometric systems of equations and series of hypergeometric type, Russian Math. Surveys, 47:4 (1992), 1–88.DOI
  9. [9]Y. Hatsuda, K. Ito and Y. Satoh, T-functions and multi-gluon scattering amplitudes, J. High Energy Phys. 1202 (2012) 003.
  10. [10]H. Kimura, Introduction to Hypergeometric Functions, Saieinsu-sha, 2007 [in Japanese].
  11. [11]D.E. Knuth, The Art of computer programming, Vol. 1: fundamental Algorithms, Third Edition, Addison-Wesley Professional, 1997.
  12. [12]S. Morier-Genoud, Coxeter’s frieze patterns at the crossroads of algebra, geometry and combinatorics, Bull. London Math. Soc. 47 (2015), 895–938.
  13. [13]S. Morier-Genoud, Counting Coxeter’s friezes over a finite field via moduli spaces, Algebra. Comb. 4 (2021), 225–240.
  14. [14]K. Nishiyama, Mathematics of frieze patterns — A glimpse of geometry and algebra, Kyoritsu Shuppan Co., Ltd., 2022 [in Japanese].
  15. [15]I. Short, Classifying SL₂-tilings, Trans. Amer. Math. Soc. 376 (2022), 1–38.
  16. [16]L. Surmann, Counting tame SL₃- and SL₄-frieze patterns over finite fields, arXiv:2505.04563 [math.CO], 2025.arxiv.org/abs/2505.04563

Paper details

vlnc:W_pammj2qq
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

OpenAI Prism, Gemini (multiple versions)

View statement
Looking for related papers…
Contents