1 Introduction

Let k be an algebraically closed field. Algebraic orbifolds over k are smooth separated irreducible Deligne-Mumford stacks of finite type over k, with trivial generic stabilizer. Birational invariants of algebraic orbifolds were introduced in [7]; they are invariant under birational projective morphisms of algebraic orbifolds. In this paper, we focus on toric Deligne-Mumford stacks, introduced in [3], and specialize to dimension 2, i.e., toric orbifold surfaces. The goal of this paper is to study their birational invariants over an algebraically closed field of characteristic 0.

2 Toric varieties

To fix notation, we recall basic definitions of toric varieties [6]. We fix a base field k, which will later be taken to be algebraically closed of characteristic 0, and a dimension n, which will later be taken to be 2. There are lattices

$$\begin{aligned} M:={\mathbb {Z}}^n\qquad \text {and}\qquad N:={{\,\mathrm{Hom}\,}}(M,{\mathbb {Z}}). \end{aligned}$$

The elements of M are characters on the algebraic torus

$$\begin{aligned} T:={{\,\mathrm{Spec}\,}}(k[M]). \end{aligned}$$

We view the lattice N as sitting inside its extension over the real numbers

$$\begin{aligned} N_{\mathbb {R}}:=N\otimes _{{\mathbb {Z}}}{\mathbb {R}}; \end{aligned}$$

we employ analogous notation, e.g., for an extension over \({\mathbb {Q}}\).

Cones in \(N_{\mathbb {R}}\) are always strongly convex rational polyhedral cones: A fan in N is a finite collection of cones in \(N_{\mathbb {R}}\), such that for every cone in the fan, its faces are also in the fan, and for every pair of cones in the fan, the intersection is a face of each. The support of a fan is the union of the cones.

A cone \(\sigma \) determines an affine toric variety

$$\begin{aligned} U_\sigma :={{\,\mathrm{Spec}\,}}(k[\sigma ^\vee \cap M]), \end{aligned}$$

where \(\sigma ^\vee \) denotes the dual cone in \(M_{\mathbb {R}}\).

The toric variety determined by a fan \(\Sigma \) is the union of \(U_\sigma \) over all \(\sigma \in \Sigma \), glued by identifying a common open subvariety \(U_{\sigma \cap \tau }\) of \(U_\sigma \) and \(U_\tau \), for all \(\sigma \), \(\tau \in \Sigma \). This yields a normal algebraic vareity \(X(\Sigma )\) over k. Every normal algebraic variety over k with T-action and equivariant open immersion of T is the toric variety of a uniquely determined fan in N.

Each \(U_\sigma \) is T-invariant and has a unique closed T-orbit, the image of

$$\begin{aligned} {{\,\mathrm{Spec}\,}}(k[\sigma ^\perp \cap M])\rightarrow {{\,\mathrm{Spec}\,}}(k[\sigma ^\vee \cap M]), \end{aligned}$$

where \(\sigma ^\perp \subset M_{\mathbb {R}}\) is defined by pairing to 0 with all elements of \(\sigma \). The closure is a closed subvariety of \(X(\Sigma )\), determined by \(\sigma \). This is a point when \(\sigma \) is an n-dimensional cone, and has codimension 1 when \(\sigma \) is a ray (1-dimensional cone).

A cone is smooth if it is generated by part of a \({\mathbb {Z}}\)-basis of N. We have \(X(\Sigma )\) smooth if and only if all of the cones of \(\Sigma \) are smooth.

For \(X(\Sigma )\) to be projective, it is necessary and sufficient for the support of \(\Sigma \) to be equal to \(N_{\mathbb {R}}\) and \(\Sigma \) to admit a piecewise linear convex support function. When \(n=2\), the second condition is automatic, so projective toric surfaces correspond to fans with support equal to \(N_{\mathbb {R}}\).

As well, when \(n=2\) we profit from the elementary structure of low-dimensional cones. The cones are 0, rays \({\mathbb {R}}_{\ge 0} v\), and 2-dimensional cones \({\mathbb {R}}_{\ge 0} v+{\mathbb {R}}_{\ge 0} w\). The generator \(v\in N_{{\mathbb {R}}}\) of a ray may be uniquely specified by the requirements to lie in N and not be a multiple of another generator in N; we call such \(v\in N\) the primitive ray generator. For a 2-dimensional cone, primitive generators v, \(w\in N\) are uniquely determined up to swapping.

3 Orbifold toric surfaces

Toric Deligne-Mumford stacks were introduced in [3]. The idea is to consider only fans whose cones are simplicial, and in this case to glue the stack quotients of smooth affine toric varieties by finite groups to obtain a smooth separated irreducible Deligne-Mumford stack with trivial generic stabilizer, i.e., an orbifold. The construction requires the given fan \(\Sigma \), with simplicial cones, to be supplemented by the additional data of a generator in N of every ray of \(\Sigma \), to make a so-called stacky fan \(\varvec{\Sigma }\).

We continue to work over a base field k, take \(n=2\), and fix a prime number p, not equal to the characteristic of k. Let \(\Sigma \) be a fan in N, such that every 2-dimensional cone is of the form \(\sigma ={\mathbb {R}}_{\ge 0}v+{\mathbb {R}}_{\ge 0}w\) with primitive generators v, \(w\in N\) such that the (well-defined) absolute value of the determinant of the \(2\times 2\) integer matrix of coordinates of v and w, with respect to a \({\mathbb {Z}}\)-basis of N, is equal to 1 or p. We consider a stacky fan \(\varvec{\Sigma }\), where the additional data of a choice of generator of each ray \(\varrho \in \Sigma \) is subject to the following conditions:

  • The chosen generator is equal to or is p times the primitive generator of \(\varrho \).

  • When the generator is p times the primitive generator, the 2-dimensional cones containing \(\varrho \) are smooth.

  • No pair of rays, for which the chosen generator is p times the primitive generator, generates a cone of \(\Sigma \).

The construction of the toric orbifold \({\mathcal {X}}(\varvec{\Sigma })\) of a stacky fan \(\varvec{\Sigma }\) proceeds by suitably modifying the construction of \(X(\Sigma )\); we only describe the construction in the present setting, where the conditions have been chosen to correspond precisely to toric orbifold surfaces \({\mathcal {X}}(\varvec{\Sigma })\), whose nontrivial stabilizer groups all have order p. For every 2-dimensional cone \(\sigma ={\mathbb {R}}_{\ge 0}v+{\mathbb {R}}_{\ge 0}w\) with primitive vectors v, \(w\in N\) and corresponding \(2\times 2\) integer matrix with determinant of absolute value p, we have an index p sublattice of N generated by v and w, and by duality, a lattice \(M'\subset M_{\mathbb {Q}}\) with \(M\subset M'\) and \(|M'/M|=p\). Then the image of

$$\begin{aligned} {{\,\mathrm{Spec}\,}}(k[M'/M])\rightarrow {{\,\mathrm{Spec}\,}}(k[M']) \end{aligned}$$

is a subgroup, isomorphic to \(\mu _p\), of \(T':={{\,\mathrm{Spec}\,}}(k[M'])\), which is the kernel of \(T'\rightarrow T\). The stack quotient

$$\begin{aligned} {[}{{\,\mathrm{Spec}\,}}(k[\sigma ^\vee \cap M'])/\mu _p] \end{aligned}$$

is used in place of \(U_\sigma \) in the construction; just with this modification, the result would be a smooth orbifold surface with finitely many points with \(\mu _p\)-stabilizer. This is further modified by passing to the pth root stack [4, Sect. 2] [1, App. B] of the union of divisors, that correspond to rays where p times the primitive ray generator is given in \(\varvec{\Sigma }\). The resulting orbifold \({\mathcal {X}}(\varvec{\Sigma })\) has locus with \(\mu _p\)-stabilizer equal to the union of finitely many points and finitely many pairwise disjoint rational curves.

The smooth Deligne-Mumford stack \({\mathcal {X}}(\varvec{\Sigma })\) has coarse moduli space \(X(\Sigma )\). We describe an orbifold as projective, when it has projective coarse moduli space. But \(n=2\), so \({\mathcal {X}}(\varvec{\Sigma )}\) is projective if and only if the support of \(\Sigma \) is equal to \(N_{\mathbb {R}}\).

4 Orbifold Burnside group

The Burnside group of stacks \(\overline{\mathrm{Burn}}_n\) and its weight module \(\overline{{\mathcal {B}}}=\bigoplus _{j\ge 0}\overline{{\mathcal {B}}}_j\) were introduced in [7]. The weight module is an abelian group, defined by explicit generators and relations.

We suppose that the base field k is algebraically closed of characteristic 0. Then a projective orbifold \({\mathcal {X}}\) of dimension n determines a class in an abelian group \(\overline{\mathrm{Burn}}_n\), generated by pairs \((K,\alpha )\) with K a finitely generated field over k, of some transcendence degree \(d\le n\), and \(\alpha \in \overline{{\mathcal {B}}}_{n-d}\). With suitable relations, this class is invariant under birational projective morphisms of projective orbifolds. In other words, \([{\mathcal {X}}]=[{\mathcal {X}}']\) in \(\overline{\mathrm{Burn}}_n\) if there exists an orbifold \({\mathcal {Y}}\) with birational projective morphisms

$$\begin{aligned} {\mathcal {Y}}\rightarrow {\mathcal {X}}\qquad \text {and}\qquad {\mathcal {Y}}\rightarrow {\mathcal {X}}'. \end{aligned}$$

Symbols with \(K=k\) generate a subgroup

$$\begin{aligned} \overline{{\mathcal {B}}}_n\subset \overline{\mathrm{Burn}}_n. \end{aligned}$$
(1)

We describe the group \(\overline{{\mathcal {B}}}_2\), or rather a particular subgroup \(\overline{{\mathcal {B}}}_2^{[p]}\), for which we have

$$\begin{aligned} {[}{\mathcal {X}}(\varvec{\Sigma })]\in \overline{{\mathcal {B}}}_2^{[p]}, \end{aligned}$$

in the subgroup (1). The locus with \(\mu _p\)-stabilizer of \({\mathcal {X}}(\varvec{\Sigma })\), as mentioned above, consists of finitely many divisors and isolated points, which we call \(\mu _p\)-divisors and \(\mu _p\)-points, respectively. Their complement is a quasi-projective variety.

The abelian group \(\overline{{\mathcal {B}}}_2^{[p]}\) is generated by symbols \((\mathrm{triv},(0,0))\), \(({\mathbb {Z}}/p{\mathbb {Z}},(0,1))\), and \(({\mathbb {Z}}/p{\mathbb {Z}},(1,a))\) for \(a\in ({\mathbb {Z}}/p{\mathbb {Z}})^\times \). These are subject to the relations

$$\begin{aligned} ({\mathbb {Z}}/p{\mathbb {Z}},(0,1))&=({\mathbb {Z}}/p{\mathbb {Z}},(1,1)), \\ (\mathrm{triv},(0,0))&=({\mathbb {Z}}/p{\mathbb {Z}},(1,p-1)), \\ ({\mathbb {Z}}/p{\mathbb {Z}},(1,a))&=({\mathbb {Z}}/p{\mathbb {Z}},(1,a^{-1})), \\ ({\mathbb {Z}}/p{\mathbb {Z}},(1,a))&=({\mathbb {Z}}/p{\mathbb {Z}},(1,a-1))+({\mathbb {Z}}/p{\mathbb {Z}},(1,a^{-1}-1))-(\mathrm{triv},(0,0)) \\&\qquad \qquad \text {for}\quad a\in \{2,\dots ,p-1\}, \end{aligned}$$

where \(a^{-1}\) denotes the multiplicative inverse to a mod p. The class \([{\mathcal {X}}(\varvec{\Sigma })]\) is the sum of contributions from the \(\mu _p\)-divisors, the \(\mu _p\)-points, and the quasi-projective variety that is the complement of the \(\mu _p\)-divisors and \(\mu _p\)-points. We let \(n_1\) denote the number of \(\mu _p\)-divisors, and \(n_2\), the number of \(\mu _p\)-points.

The quasi-projective variety contributes

$$\begin{aligned} (1-n_1-n_2)(\mathrm{triv},(0,0)). \end{aligned}$$

The \(\mu _p\)-divisors contribute

$$\begin{aligned} n_1({\mathbb {Z}}/p{\mathbb {Z}},(0,1)). \end{aligned}$$

The contribution from the \(\mu _p\)-points is the sum of individual contributions

$$\begin{aligned} ({\mathbb {Z}}/p{\mathbb {Z}},(1,a)) \end{aligned}$$

from each \(\mu _p\)-point. At a \(\mu _p\)-point, an identification of the stabilizer with \(\mu _p\) is made, so that one of the weights of the action on the tangent space is 1; then a is the second weight. But this involves a choice: if we reverse the roles of the two weights then the resulting weight would be \(a^{-1}\). Thanks to the corresponding relation in \(\overline{{\mathcal {B}}}_2^{[p]}\), the contribution \(({\mathbb {Z}}/p{\mathbb {Z}},(1,a))\in \overline{{\mathcal {B}}}_2^{[p]}\) of a \(\mu _p\)-point is well-defined.

The quotient group

$$\begin{aligned} \overline{{\mathcal {B}}}_2^{[p]}/\langle (\mathrm{triv},(0,0)), ({\mathbb {Z}}/p{\mathbb {Z}},(0,1)) \rangle \end{aligned}$$
(2)

is the group denoted by \(\overline{{\mathcal {B}}}_2^{[p]}/ ({\mathcal {C}}\cap \overline{{\mathcal {B}}}_2^{[p]})\) in [7, Sect. 4]. In [7, Lemma 5.3], an explicit isomorphism is given from this group to a certain quotient of the group \(H_1(X_0(p)_{\mathrm{orb}}, {\mathbb {Z}})\), the first orbifold homology of the modular curve \(X_0(p)\). This group (2) is also mentioned as providing an obstruction to being linked by blow-ups of points to an orbifold surface without \(\mu _p\)-points. For the group (2) the following isomorphism types were found:

$$\begin{aligned} p=2:\,0,\quad p=3:\,0,\quad p=5:\,{\mathbb {Z}}/2{\mathbb {Z}}, \quad p=7:\,0,\quad p=11:\,{\mathbb {Z}}. \end{aligned}$$

As mentioned in [7, Sect. 4], the vanishing for \(p=2\) and \(p=3\) is the \(n=2\) case of a more general phenomenon, related to destackification [2]. By [7, Prop. 4.4], the group (2) is nonzero for all \(p\ge 5\), except \(p=7\).

5 Refined invariant

We continue to work over an algebraically closed base field k of characteristic 0. A natural class of birational projective morphisms of projective toric Deligne-Mumford stacks is the class of morphisms associated with the stacky star subdivisions (of toric Deligne-Mumford stacks) of [5]. All toric stacks will be projective toric orbifold surfaces with \(\mu _p\) as the only possible nontrivial stabilizer group. We emphasize that the torus T is regarded as fixed. All stacky fans will satisfy the corresponding conditions, stated in Sect. 3, and have support equal to \(N_{\mathbb {R}}\).

A stacky star subdivision is an operation that, given a stacky fan \(\varvec{\Sigma }\) and a chosen 2-dimensional cone \(\sigma \in \Sigma \), yields a new stacky fan \(\varvec{\Sigma }_\sigma \) and a T-equivariant birational projective morphism

$$\begin{aligned} f_\sigma :{\mathcal {X}}(\varvec{\Sigma }_\sigma )\rightarrow {\mathcal {X}}(\varvec{\Sigma }). \end{aligned}$$

Let us write \(\sigma ={\mathbb {R}}_{\ge 0}v+{\mathbb {R}}_{\ge 0}w\) with primitive vectors v, \(w\in N\); then the stacky fan \(\varvec{\Sigma }_\sigma \) is constructed by deleting the cone \(\sigma \), adding the cones

$$\begin{aligned} {\mathbb {R}}_{\ge 0}(v+w),\quad {\mathbb {R}}_{\ge 0}v+{\mathbb {R}}_{\ge 0}(v+w), \quad {\mathbb {R}}_{\ge 0}(v+w)+{\mathbb {R}}_{\ge 0}w, \end{aligned}$$

and endowing the ray \({\mathbb {R}}_{\ge 0}(v+w)\) with the choice of generator \(v+w\). Geometrically, \(f_\sigma \) is the blow-up of \({\mathcal {X}}(\varvec{\Sigma })\) at the point, corresponding to \(\sigma \); this is a point of the scheme locus when \(\sigma \) is a smooth cone whose rays come with integer multiple 1, a \(\mu _p\)-point when \(\sigma \) is not smooth, a point on a \(\mu _p\)-divisor when \(\sigma \) is smooth and one of its rays comes with multiple p.

By the classical factorization of birational projective morphisms of smooth surfaces as compositions of blow-ups, for any T-equivariant birational projective morphism

$$\begin{aligned} f:{\mathcal {X}}(\varvec{\Sigma }')\rightarrow {\mathcal {X}}(\varvec{\Sigma }), \end{aligned}$$

we obtain \(\varvec{\Sigma }'\) from \(\varvec{\Sigma }\) by performing a finite sequence of stacky star subdivisions, and f is the composite of the corresponding T-equivariant birational projective morphisms.

Inspired by [2, Example 4.2] and [7, Section 4], we might ask: For a given pair of toric orbifold surfaces \({\mathcal {X}}(\varvec{\Sigma })\) and \({\mathcal {X}}(\varvec{\Sigma }')\), does there exist a toric orbifold surface \({\mathcal {X}}(\varvec{\Sigma }'')\) with T-equivariant birational projective morphisms

$$\begin{aligned} {\mathcal {X}}(\varvec{\Sigma }'')\rightarrow {\mathcal {X}}(\varvec{\Sigma })\quad \text {and} \quad {\mathcal {X}}(\varvec{\Sigma }'')\rightarrow {\mathcal {X}}(\varvec{\Sigma }')\,? \end{aligned}$$

When the answer is affirmative, we say that \({\mathcal {X}}(\varvec{\Sigma })\) and \({\mathcal {X}}(\varvec{\Sigma }')\) are equivalent under T-equivariant birational projective morphisms.

As mentioned in Sect. 4, the group \(\overline{{\mathcal {B}}}_2^{[p]}/\langle (\mathrm{triv},(0,0)), ({\mathbb {Z}}/p{\mathbb {Z}},(0,1))\rangle \) supplies an obstruction to the equivalence under T-equivariant birational projective with a toric orbifold surface whose coarse moduli space is smooth.

We record the observation, that in the toric setting the abelian group \(\overline{{\mathcal {B}}}_2^{[p]}\) admits a refinement.

Definition 1

The refined toric orbifold surface weight group, for \(\mu _p\)-stabilizers, is the abelian group

$$\begin{aligned} \overline{{\mathcal {T}}}_2^{[p]}, \end{aligned}$$

generated by \((\mathrm{triv},(0,0))\), \(({\mathbb {Z}}/p{\mathbb {Z}},(0,1))\), and \(({\mathbb {Z}}/p{\mathbb {Z}},(1,a))\) for \(a\in ({\mathbb {Z}}/p{\mathbb {Z}})^\times \), with relations

$$\begin{aligned} ({\mathbb {Z}}/p{\mathbb {Z}},(0,1))&=({\mathbb {Z}}/p{\mathbb {Z}},(1,1)),\\ (\mathrm{triv},(0,0))&=({\mathbb {Z}}/p{\mathbb {Z}},(1,p-1)),\\ ({\mathbb {Z}}/p{\mathbb {Z}},(1,a))&=({\mathbb {Z}}/p{\mathbb {Z}},(1,a-1)) +({\mathbb {Z}}/p{\mathbb {Z}},(1,(a^{-1}-1)^{-1}))-(\mathrm{triv},(0,0)) \\&\qquad \qquad \text {for}\quad a\in \{2,\dots ,p-1\}. \end{aligned}$$

With the additional relations \(({\mathbb {Z}}/p{\mathbb {Z}},(1,a))=({\mathbb {Z}}/p{\mathbb {Z}},(1,a^{-1}))\) for \(a\in ({\mathbb {Z}}/p{\mathbb {Z}})^\times \), we recover the presentation of \(\overline{{\mathcal {B}}}_2^{[p]}\). So, we have a canonical surjective homomorphism

$$\begin{aligned} \overline{{\mathcal {T}}}_2^{[p]}\rightarrow \overline{{\mathcal {B}}}_2^{[p]}. \end{aligned}$$
(3)

Proposition 2

The assignment, to \({\mathcal {X}}(\varvec{\Sigma })\), of the class

$$\begin{aligned} (1-n_1-n_2)(\mathrm{triv},(0,0))+n_1({\mathbb {Z}}/p{\mathbb {Z}},(0,1))+ \sum _{\begin{array}{c} \mathrm{non}\text {-}\mathrm{smooth}\ \sigma \in \Sigma \\ \sigma ={\mathbb {R}}_{\ge 0}v+{\mathbb {R}}_{\ge 0}w\\ v_1w_2-w_1v_2=p \end{array}}({\mathbb {Z}}/p{\mathbb {Z}},(1,-w_i^{-1}v_i)) \end{aligned}$$

in \(\overline{{\mathcal {T}}}_2^{[p]}\), where in the sum we have \(v=(v_1,v_2)\), \(w=(w_1,w_2)\in N\), and \(i\in \{1,2\}\) is chosen so that \(p\not \mid w_i\), gives rise to an invariant of toric orbifold surfaces with \(\mu _p\) as only possible nontrivial stabilizer group, under torus-equivariant birational projective morphisms. The assignment is compatible with the homomorphism (3).

Proof

From the explicit description of stacky star subdivision, it is easy to check that \(\varvec{\Sigma }\) and any stacky star subdivision \(\varvec{\Sigma }_\sigma \) have the same class in \(\overline{{\mathcal {T}}}_2^{[p]}\). \(\square \)

Example

We may form a quotient

$$\begin{aligned} \overline{{\mathcal {T}}}_2^{[p]}/\langle (\mathrm{triv},(0,0)),({\mathbb {Z}}/p{\mathbb {Z}},(0,1))\rangle , \end{aligned}$$

analogous to (2). Isomorphism types for some p are:

$$\begin{aligned} p=2:\,0,\quad p=3:\,0,\quad p=5:\,{\mathbb {Z}}/2{\mathbb {Z}}, \quad p=7:\,{\mathbb {Z}}/3{\mathbb {Z}},\quad p=11:\,{\mathbb {Z}}^2. \end{aligned}$$

In particular, we obtain a nontrivial refinement for \(p=7\) of the trivial obstruction group mentioned in Sect. 4.

Comparison with the isomorphism types of \(H_1(X_0(p)_{\mathrm{orb}}, {\mathbb {Z}})\) from [7, Sect. 5] suggests that this refinement reproduces the full first orbifold homology of which (2) is a quotientFootnote 1. As recalled in [7, Lemma 5.2], the group \(H_1(X_0(p)_{\mathrm{orb}}, {\mathbb {Z}})\) is generated by Manin symbols

$$\begin{aligned} \Big \{ 0, \frac{1}{a} \Big \}, \quad 2 \le a \le p-2, \end{aligned}$$

with relations

$$\begin{aligned}&\Big \{ 0, \frac{1}{a}\Big \} + \Big \{ 0, \frac{1}{p-a^{-1}}\Big \} = 0, \end{aligned}$$
(4)
$$\begin{aligned}&\Big \{ 0, \frac{1}{(p-1)/2} \Big \} + \Big \{ 0, \frac{1}{p-2} \Big \} = 0, \end{aligned}$$
(5)
$$\begin{aligned}&\Big \{ 0, \frac{1}{a} \Big \} + \Big \{ 0, \frac{1}{a'} \Big \} +\Big \{ 0, \frac{1}{a''} \Big \} = 0 \quad (a \notin \{ p-2, (p-1)/2 \}), \end{aligned}$$
(6)

where

$$\begin{aligned} a' \equiv -a^{-1} -1 \, \mathrm{mod} \, p, \quad a'' \equiv -(a+1)^{-1} \, \mathrm{mod} \, p. \end{aligned}$$

Theorem 3

There is an isomorphism

$$\begin{aligned} \overline{{\mathcal {T}}}_2^{[p]}/\langle (\mathrm{triv},(0,0)), ({\mathbb {Z}}/p{\mathbb {Z}},(0,1))\rangle \rightarrow H_1(X_0(p)_{\mathrm{orb}}, {\mathbb {Z}}) \end{aligned}$$

sending \(({\mathbb {Z}}/p{\mathbb {Z}},(1,a))\) to the Manin symbol \(\{ 0, 1/a\}\) for \(2 \le a \le p-2\).

Proof

We are working in the quotient \(\overline{{\mathcal {T}}}_2^{[p]}/\langle (\mathrm{triv},(0,0)), ({\mathbb {Z}}/p{\mathbb {Z}},(0,1))\rangle \). We start by obtaining the relation

$$\begin{aligned} ({\mathbb {Z}}/p{\mathbb {Z}},(1,-b)) = -({\mathbb {Z}}/p{\mathbb {Z}},(1,b^{-1})). \end{aligned}$$
(7)

The relation follows by induction on b, starting with the base case \(b = 2\), which is a consequence of the last relation from Definition 1 for \(a = 2^{-1}\) and \(a = p-1\). The inductive step uses the same relation with \(a = b^{-1}\) and \(a = p+1-b\). This recovers (4). Relation (5) is recovered by once again using the last relation from Definition 1 with \(a = p-1\). Using (7) we may rewrite the last relation from Definition 1 in the form

$$\begin{aligned} 0 = ({\mathbb {Z}}/p{\mathbb {Z}},(1,-a^{-1})) + ({\mathbb {Z}}/p{\mathbb {Z}},(1,a-1)) + ({\mathbb {Z}}/p{\mathbb {Z}},(1,((a^{-1}-1)^{-1}))). \end{aligned}$$

Replacing a by \(a+1\) we reproduce (6). \(\square \)

The isomorphism of Theorem 3 is compatible with that of [7, Lemma 5.3].