Bourbaki

§ 5. Coequalizer

1. Contraction of the arrows of a homotopy

Let H be a quiver, let G be a groupoid, let φ\varphi and ψ\psi be quiver morphisms of H into G, and let h:h: Som(H) →\rightarrow Fl(G) be a homotopy connecting φ\varphi to ψ\psi. Let G′G' be the groupoid deduced from G by contraction of the arrows of the image of h(h(II, p. 175, No.o11^o11) and let β:G→G′\beta : G\rightarrow G' be the canonical morphism.

Let us denote by Γ the quiver (Som(G), Som(H), Som(φ\varphi ), Som(ψ\psi )). By definition of a homotopy, the pair (IdSom(G),h_{Som(G)}, h) is a quiver morphism of Γ into G; it extends to a unique groupoid morphism η:\eta : Grp(Γ) →\rightarrow G. By construction, the set of vertices of G′G' is the set of connected components of the quiver Γ.

In all that follows in this No.o^o, we shall suppose that the groupoid G is transitive. According to remark 1 of II, p. 170, this amounts to supposing that the groupoid G′G' is so. We also fix a vertex a0a_0 of G.

Let us denote by Γ~\widetilde{\Gamma} the graph associated with the quiver Γ (cf. II, p. 156). The set of loops of length ⩾1\geqslant 1 in Γ~\widetilde{\Gamma} is identified with the set Ω(Γ~)\Omega (\widetilde{\Gamma}) of finite sequences (z1,...,zn)(z_1, . . . , z_n), where nn is an integer such that n⩾1n\geqslant 1 and z1,...,znz_1, . . . , z_n are elements of Fl(Γ~\widetilde{\Gamma}) such that t(zi)=o(zi+1)t(z_i) =o(z_{i+1}) for every integer ii such that 1⩽i<n1\leqslant i < n and t(zn)=o(z1)t(z_n) =o(z_1).

Let z=((b1,ε1),...,(bn,εn))\mathbf{z}= ((b_1, \varepsilon_1), . . . ,(b_n, \varepsilon_n)) be an element of Ω(Γ~)\Omega (\widetilde{\Gamma}). The conjugacy class of the element

Int(gg)(η(z))=gh(b1)ε1...h(bn)εng−1(\eta (\mathbf{z})) =gh(b_1)^{\varepsilon_1}. . . h(b_n)^{\varepsilon_n}g^{-1}

in Ga0G_{a_0} does not depend on the choice of the arrow gg of G joining the vertex a0a_0 to the origin of (b1,ε)(b_1, \varepsilon ). We denote this conjugacy class by c(z)c(\mathbf{z}).

Proposition 1 01UI

The group morphism βa0:Ga0→Gβ(a0)′\beta_{a_0}: G_{a_0}\rightarrow G'_{\beta(a_0)} is surjective, and its kernel is the smallest subgroup of Ga0G_{a_0} containing the conjugacy classes c(z)c(\mathbf{z}) for z∈Ω(Γ~)\mathbf{z}\in \Omega (\widetilde{\Gamma}).

Let K be the smallest distinguished subgroupoid of G whose set of arrows contains the image of hh. The morphism Ga0→Gβ(a0)′G_{a_0}\rightarrow G'_{\beta(a_0)} is surjective and its kernel is equal to Ka0(K_{a_0}(II, p. 170, prop. 2). The proposition then follows from prop. 8 of II, p. 176.

Definition 1 01UJ

A subset Z of Ω(Γ~)\Omega (\widetilde{\Gamma}) is said to be distinguished (relative to the pair (φ,ψ)(\varphi , \psi )) if it satisfies the following properties:

(i) For every arrow zz of Γ~\widetilde{\Gamma}, one has (z,z)∈Z(z, z)\in Z ;

(ii) For every (z1,...,zn)∈(z_1, . . . , z_n)\in Z, one has (z‾n,...,z‾2,z‾1)∈(\overline{z}_n, . . . ,\overline{z}_2,\overline{z}_1)\in Z and (zn,z1,...,zn−1)∈Z(z_n, z_1, . . . , z_{n-1})\in Z ;

(iii) Let z=(z1,...,zn)\mathbf{z}= (z_1, . . . , z_n) and z′=(z1′,...,zm′)\mathbf{z}'= (z'_1, . . . , z'_m) be elements of Z such that t(zn)=o(z1′)t(z_n) =o(z'_1). Put zz′=(z1,...,zn,z1′,...,zm′)\mathbf{z}\mathbf{z}'= (z_1, . . . , z_n, z'_1, . . . , z'_m). If two elements among z,z′,zz′\mathbf{z},\mathbf{z}',\mathbf{z}\mathbf{z}' belong to Z, the same is true of the third;

(iv) For every arrow ff of H, put φ~(f,1)=ψ~(f,−1)=\widetilde{\varphi}(f,1) =\widetilde{\psi}(f,-1) = φ(f)\varphi (f) and φ~(f,−1)=ψ~(f,1)=ψ(f)\widetilde{\varphi}(f,-1) =\widetilde{\psi}(f,1) =\psi (f). Let nn be an integer ⩾\geqslant 1 and (f1,ε1),...,(fn,εn)(f_1, \varepsilon_1), . . . ,(f_n, \varepsilon_n) a sequence of elements of Fl(H) ×{−1,1}\times \{-1,1\} such that ψ~(fi,εi)=φ~(fi+1,εi+1)\widetilde{\psi}(f_i, \varepsilon_i) =\widetilde{\varphi}(f_{i+1}, \varepsilon_{i+1}) for 1⩽i<n1\leqslant i < n and ψ~(fn,εn)=φ~(f1,ε1)\widetilde{\psi}(f_n, \varepsilon_n) =\widetilde{\varphi}(f_1, \varepsilon_1); let aia_i denote the origin of fif_i and bib_i its term. In order that ((a1,ε1),...,(an,εn))((a_1, \varepsilon_1), . . . ,(a_n, \varepsilon_n)) belong to Z, it is necessary and sufficient that ((b1,ε1),...,(bn,εn))((b_1, \varepsilon_1), . . . ,(b_n, \varepsilon_n)) belong to Z.

The intersection in Ω(Γ~)\Omega (\widetilde{\Gamma}) of every family of distinguished subsets (relative to (φ,ψ)(\varphi , \psi )) is again distinguished. In particular, there exists a smallest distinguished subset containing a given subset Z of Ω(Γ~)\Omega (\widetilde{\Gamma}).

Proposition 2 01UK

If N is a distinguished subgroup of GaG_a, the set of elements z∈Ω(Γ~)\mathbf{z}\in \Omega (\widetilde{\Gamma}) such that c(z)c(\mathbf{z}) is contained in N is a distinguished subset of Ω(Γ~)\Omega (\widetilde{\Gamma}).

Let Z denote the set of elements z∈Ω(Γ~)\mathbf{z}\in \Omega (\widetilde{\Gamma}) such that c(z)⊂Nc(\mathbf{z})\subset N.

Let z∈z\in Fl(Γ~\widetilde{\Gamma}). We have c(z,z)={ea}c(z, z) =\{e_a\} by definition, whence (z,z)∈Z(z, z)\in Z.

Let (z1,...,zn)∈Z(z_1, . . . , z_n)\in Z. By definition of cc, the conjugacy class c(z1,...,zn)c(z_1, . . . , z_n) is equal to c(zn,z1,...,zn−1)c(z_n, z_1, . . . , z_{n-1}) and is formed by the inverses of the elements of c(z‾n,...,z‾1)c(\overline{z}_n, . . . ,\overline{z}_1). This shows that Z satisfies condition (ii).

With the notation of condition (iii), one can choose elements u∈c(z),v∈c(z′)u\in c(\mathbf{z}),v\in c(\mathbf{z}') and w∈c(zz′)w\in c(\mathbf{z}\mathbf{z}') such that uv=wuv=w. If two of the elements u,vu,v and ww belong to N, the same is true of the third, therefore N satisfies (iii).

The notation being that of (iv), we have, for every integer ii such that 1⩽i⩽n1\leqslant i\leqslant n, the relation

φ(fi)h(bi)=h(ai)ψ(fi) \varphi (f_i)h(b_i) =h(a_i)\psi (f_i)

for hh is a homotopy joining φ\varphi to ψ\psi. This equality can also be written

φ~(fi,εi)h(bi)εi=h(ai)εiψ~(fi,εi) \widetilde{\varphi}(f_i, \varepsilon_i)h(b_i)^{\varepsilon_i}=h(a_i)^{\varepsilon_i}\widetilde{\psi}(f_i, \varepsilon_i)

Taking into account the relations ψ~(fi,εi)=φ~(fi+1,εi+1)\widetilde{\psi}(f_i, \varepsilon_i) =\widetilde{\varphi}(f_{i+1}, \varepsilon_{i+1}) for 1⩽i<n1\leqslant i < n and ψ~(fn,εn)=φ~(f1,ε1)\widetilde{\psi}(f_n, \varepsilon_n) =\widetilde{\varphi}(f_1, \varepsilon_1), one deduces

φ~(f1,ε1)h(b1)ε1...h(bn)εn=h(a1)ε1...h(an)εnφ~(f1,ε1) \widetilde{\varphi}(f_1, \varepsilon_1)h(b_1)^{\varepsilon_1}. . . h(b_n)^{\varepsilon_n}=h(a_1)^{\varepsilon_1}. . . h(a_n)^{\varepsilon_n}\widetilde{\varphi}(f_1, \varepsilon_1)

so that the conjugacy classes c((a1,ε1),...,(an,εn))c((a_1, \varepsilon_1), . . . ,(a_n, \varepsilon_n)) and c((b1,ε1),...,(bn,εn))c((b_1, \varepsilon_1), . . . ,(b_n, \varepsilon_n)) are equal. This shows that Z satisfies condition (iv) and completes the proof of the proposition.

Corollary 01UL

Let Z be a subset of Ω(Γ~)\Omega (\widetilde{\Gamma}); in order that the conjugacy classes c(z)c(\mathbf{z}), for z∈\mathbf{z}\in Z, generate the kernel of the canonical homomorphism βa0:Ga0→Gβ(a0)′\beta_{a_0}: G_{a_0}\rightarrow G'_{\beta(a_0)}, it is enough that the smallest distinguished subset of Ω(Γ~)\Omega (\widetilde{\Gamma}) containing Z be equal to Ω(Γ~)\Omega (\widetilde{\Gamma}).

Let N be the smallest subgroup of Ga0G_{a_0} containing c(z)c(\mathbf{z}) for every z∈Z\mathbf{z}\in Z; it is distinguished. Let Z′Z' be the set of elements z\mathbf{z} of Ω(Γ~)\Omega (\widetilde{\Gamma}) such that c(z)∈Nc(\mathbf{z})\in N. By proposition 2, Z′Z' is a distinguished subset of Ω(Γ~)\Omega (\widetilde{\Gamma}). It contains Z. By hypothesis, it is therefore equal to Ω(Γ~)\Omega (\widetilde{\Gamma}). It then follows from prop. 1 (II, p. 197) that N is the kernel of the homomorphism βa\beta_a.

2. Definition of the coequalizer

Let H be a quiver, let G be a groupoid and let φ,ψ\varphi ,\psi be morphisms of quivers from H into G. Let us denote by Coh(φ,ψ\varphi , \psi ) the cohomotopizer of the pair (φ,ψ)(\varphi , \psi ) (II, p. 185, def. 3),α:G→,\alpha : G\rightarrow Coh(φ,ψ\varphi , \psi ) the canonical groupoid morphism and hh the canonical homotopy joining α∘φ\alpha \circ \varphi to α∘ψ\alpha \circ \psi.

Let Coeg(φ,ψ\varphi , \psi ) be the groupoid obtained from Coh(φ,ψ\varphi , \psi ) by contraction of the arrows belonging to the image of hh (II, p. 196, no^o 1), let us denote by β:\beta : Coh(φ,ψ\varphi , \psi )→\rightarrow Coeg(φ,ψ\varphi , \psi ) the canonical morphism and let us put γ=β∘α\gamma =\beta \circ \alpha.

Definition 2 01UM

The groupoid Coeg(φ,ψ\varphi , \psi ) is called the coequalizer of the pair (φ,ψ)(\varphi , \psi ); the groupoid morphism γ\gamma is called the canonical morphism from G into Coeg(φ,ψ\varphi , \psi ).

Proposition 3 01UN

The pair (Coeg(φ,ψ\varphi , \psi ),γ, \gamma ) has the following universal property:

a) We have γ∘φ=γ∘ψ\gamma \circ \varphi =\gamma \circ \psi.

b) Let G′G' be a groupoid and let θ:G→G′\theta : G\rightarrow G' be a groupoid morphism such that θ∘φ=θ∘ψ\theta \circ \varphi =\theta \circ \psi. There exists a unique groupoid morphism θ‾:\overline{\theta}: Coeg(φ,ψ\varphi , \psi )→G′\rightarrow G' such that θ‾∘γ=θ\overline{\theta}\circ \gamma =\theta.

Let aa be a vertex of H. The arrow h(a)h(a) of Coh(φ,ψ\varphi , \psi ) joins α(φ(a))\alpha (\varphi (a)) to α(ψ(a))\alpha (\psi (a)). By definition of the groupoid Coeg(φ,ψ\varphi , \psi ), the origin and the end of the arrow β(h(a))\beta (h(a)) are equal; therefore we have γ(φ(a))=γ(ψ(a))\gamma (\varphi (a)) =\gamma (\psi (a)).

Let ff be an arrow of H; if aa denotes its origin and bb its end, we thus have

α(φ(f))⋅h(b)=h(a)⋅α(ψ(b)) \alpha (\varphi (f))\cdot h(b) =h(a)\cdot \alpha (\psi (b))

Taking the image by β\beta of the two members of this equality, one obtains the relation γ(φ(f))=γ(ψ(f))\gamma (\varphi (f)) =\gamma (\psi (f)). This proves assertion a).

Let us prove b). The mapping η:\eta : Som(H) →\rightarrow Fl(G′') which associates with every vertex aa of H the arrow eθ(φ(a))e_{\theta(\varphi(a))} of G′G' is a homotopy joining θ∘φ\theta \circ \varphi to θ∘ψ\theta \circ \psi. By the universal property of cohomotopizers (II, p. 185, prop. 3), there exists a unique groupoid morphism θ1:\theta_1: Coh(φ,ψ\varphi , \psi )→G′\rightarrow G' such that θ1∘α=θ\theta_1\circ \alpha =\theta and θ1(h(a))=eθ(φ(a))\theta_1(h(a)) =e_{\theta(\varphi(a))} for every vertex aa of H. By prop. 7 of II, p. 176, this latter property implies the existence of a unique groupoid morphism θ‾:\overline{\theta}: Coeg(φ,ψ\varphi , \psi )→G′\rightarrow G' such that θ‾∘β=θ1\overline{\theta}\circ \beta =\theta_1. One then has θ‾∘γ=θ1∘α=θ\overline{\theta}\circ \gamma =\theta_1\circ \alpha =\theta.

Conversely, if θ‾′:\overline{\theta}': Coeg(φ,ψ\varphi , \psi )→G′\rightarrow G' is a groupoid morphism such that θ‾′∘γ=θ\overline{\theta}'\circ \gamma =\theta, one has (θ‾′∘β)∘α=(θ‾∘β)∘α(\overline{\theta}'\circ \beta )\circ \alpha = (\overline{\theta}\circ \beta )\circ \alpha, whence θ‾′∘β=θ‾∘β\overline{\theta}'\circ \beta =\overline{\theta}\circ \beta by II, p. 185, prop. 3, whence θ‾′=θ‾\overline{\theta}'=\overline{\theta} by prop. 7 of II, p. 176. This proves the uniqueness of θ‾\overline{\theta}, hence assertion b).

Corollary 01UO

The groupoid Coeg(φ,ψ\varphi , \psi ) is generated by the image of the morphism γ\gamma.

Let C be the subgroupoid of Coeg(φ,ψ\varphi , \psi ) generated by the image of G; let ii denote the canonical morphism of C into Coeg(φ,ψ\varphi , \psi ) and θ:G→C\theta : G\rightarrow C the morphism such that i∘θ=γi\circ \theta =\gamma. By proposition 2, there exists a unique groupoid morphism θ‾:\overline{\theta}: Coeg(φ,ψ\varphi , \psi )→C\rightarrow C such that θ‾∘γ=θ\overline{\theta}\circ \gamma =\theta. Then γ=i∘θ‾∘γ\gamma =i\circ \overline{\theta}\circ \gamma, hence i∘θ‾i\circ \overline{\theta} is the identity morphism of Coeg(φ,ψ\varphi , \psi ) (loc. cit.). In particular, the mappings Som(ii) and Fl(ii) are surjective, therefore C = Coeg(φ,ψ\varphi , \psi ).

Remark 1 01UP

The set of vertices of Coeg(φ,ψ\varphi , \psi ) is the set of connected components of the quiver

Γ = (Som(G), Som(H), Som(φ\varphi ), Som(ψ\psi ))

(II, p. 197). In other words, it is the quotient set of Som(G) by the finest equivalence relation such that φ(a)\varphi (a) is equivalent to ψ(a)\psi (a) for every a∈a\in Som(G).

Remark 2 01UQ

The mapping Orb(γ\gamma ), deduced from γ\gamma by passing to orbits, defines by passage to the quotient a bijection of the set of connected components of the framework of the pair (φ,ψ)(\varphi , \psi ) onto the set of orbits of the coequalizer. This results from II, p. 185, prop. 4 and from the fact that the mapping deduced from β\beta by passing to orbits is bijective ( II, p. 170, remark 1).

Consequently, the mapping of Som(G) into Orb(Coeg(φ,ψ\varphi , \psi )) deduced from γ\gamma identifies the set of orbits of Coeg(φ,ψ\varphi , \psi ) with the quotient set of Som(G) by the equivalence relation generated by the pairs (φ(x),ψ(x))(\varphi (x), \psi (x)) for x∈x\in Som(H) and the pairs (o(f),t(f))(o(f), t(f)) for f∈f\in Fl(G).

Remark 3 01UR

The mapping Fl(γ\gamma ) of Fl(G) into Fl(Coeg(φ,ψ\varphi , \psi )) is not in general surjective.

3. Comparison of the isotropy groups of the cohomotopizer and the coequalizer

Let H be a quiver, let G be a groupoid, and let φ\varphi and ψ\psi be morphisms of quivers from H into G. Let us denote by α:G→\alpha : G\rightarrow Coh(φ,ψ\varphi , \psi ), β:\beta : Coh(φ,ψ\varphi , \psi )→\rightarrow Coeg(φ,ψ\varphi , \psi ) and γ:G→\gamma : G\rightarrow Coeg(φ,ψ\varphi , \psi ) the canonical morphisms of groupoids, and by hh the canonical homotopy joining α∘φ\alpha \circ \varphi to α∘ψ\alpha \circ \psi. These notations are summarized by the following diagram

γ\gamma

H ψφ^{\varphi}_{\psi} G α^{\alpha} Coh(φ,ψ\varphi , \psi )β^{\beta} Coeg(φ,ψ\varphi , \psi ) .

In the whole remainder of this no^o, we suppose that the framework of the pair (φ,ψ)(\varphi , \psi ) is a connected quiver and we fix a vertex a0a_0 of G. Consequently, the groupoids Coh(φ,ψ\varphi , \psi ) and Coeg(φ,ψ\varphi , \psi ) are transitive.

Let Γ denote the quiver (Som(G), Som(H), Som(φ\varphi ), Som(ψ\psi )), Γ~\widetilde{\Gamma} the graph associated with Γ, and let Ω(Γ~)\Omega (\widetilde{\Gamma}) be the set of finite sequences (z1,...,zn)(z_1, . . . , z_n) of arrows of Γ~\widetilde{\Gamma}, with n⩾1n\geqslant 1, such that the terminus of ziz_i is the origin of zi+1z_{i+1} if 1⩽i<n1\leqslant i < n and that the terminus of znz_n is the origin of z1z_1. In No. 1 there was constructed a mapping z↦c(z)\mathbf{z}\mapsto c(\mathbf{z}) of the set Ω(Γ~)\Omega (\widetilde{\Gamma}) into the set of conjugacy classes of the group Coh(φ,ψ\varphi , \psi )a0_{a_0}.

Let H×GHH\times_GH denote the equalizer in H×HH\times H of the quiver morphisms ψ∘\psi \circpr1_1 and φ∘\varphi \circpr2_2. The set of its vertices is the set of pairs (a,b)(a, b) of vertices of H such that ψ(a)=φ(b)\psi (a) =\varphi (b); the set of its arrows is the set of pairs (f,g)(f, g) of arrows of H such that ψ(f)=φ(g)\psi (f) =\varphi (g); the origin of an arrow (f,g)(f, g) is the vertex (o(f),o(g))(o(f), o(g)) and its term is the vertex (t(f),t(g))(t(f), t(g)) (cf. II, p. 153).

Finally, let Ker(φ,ψ\varphi , \psi ) denote the equalizer of φ\varphi and ψ\psi; recall (II, p. 165, example 2) that this is the subquiver of H whose vertices aa are those such that φ(a)=ψ(a)\varphi (a) =\psi (a) and whose arrows are the f∈f\in Fl(H) such that φ(f)=ψ(f)\varphi (f) =\psi (f).

Proposition 4 01US

Let μ:H×GH→H\mu: H\times_GH\rightarrow H be a quiver morphism such that φ∘μ=φ∘\varphi \circ \mu=\varphi \circ pr1_1 and ψ∘μ=ψ∘\psi \circ \mu=\psi \circ pr2_2. Suppose that, for every pair (a,b)(a, b) of vertices of H such that φ(a)=φ(b)\varphi (a) =\varphi (b), there exists a vertex cc of H such that φ(c)=ψ(b)\varphi (c) =\psi (b) and a=μ(b,c)a=\mu(b, c).

Let A1A_1 be a set of vertices of Ker(φ,ψ\varphi , \psi ) meeting each of its connected components and let Z1Z_1 be the set of elements of Ω(Γ~)\Omega (\widetilde{\Gamma}) of the form ((a,1))((a,1)), where a∈A1a\in A_1. Let also A2A_2 be a set of vertices of H×GHH\times_GH meeting each connected component of H×GHH\times_GH and let Z2Z_2 be the set of triples of the form ((a,1),(b,1),(μ(a,b),−1))((a,1),(b,1),(\mu(a, b),-1)), as (a,b)(a, b) ranges over A2A_2.

Then Ω(Γ~)\Omega (\widetilde{\Gamma}) is the smallest distinguished subset of Ω(Γ~)\Omega (\widetilde{\Gamma}) containing Z1∪Z2Z_1\cup Z_2. In particular, the conjugacy classes c(z)c(\mathbf{z}) in Coh(φ,ψ\varphi , \psi )a0_{a_0}, where z\mathbf{z} ranges over Z1∪Z2Z_1\cup Z_2, generate the kernel of the canonical homomorphism βa\beta_a of Coh(φ,ψ\varphi , \psi )a0_{a_0} into Coeg(φ,ψ\varphi , \psi )β(a0)_{\beta(a_0)}.

Let Z1′,Z2′Z'_1,Z'_2 and Z′Z' denote the smallest distinguished subsets of Ω(Γ~)\Omega (\widetilde{\Gamma}) containing respectively Z1,Z2Z_1,Z_2 and Z1∪Z2(Z_1\cup Z_2(II, p. 197, definition 1). It is required to prove that Z′Z' is equal to Ω(Γ~)\Omega (\widetilde{\Gamma}); by II, p. 199, Corollary 1 of Proposition 1 of II, p. 197, the last assertion of the proposition will follow.

Lemma 1 01UT

a) For every vertex aa of Ker(φ,ψ\varphi , \psi ), ((a,1))((a,1)) belongs to Z1′Z'_1.

b) For every vertex (a,b)(a, b) of H×GHH\times_GH, ((a,1),(b,1),(μ(a,b),−1))((a,1),(b,1),(\mu(a, b),-1)) belongs to Z2′Z'_2.

a) Let A1′A'_1 be the set of vertices aa of Ker(φ,ψ\varphi , \psi ) such that ((a,1))((a,1)) belongs to Z1′Z'_1. We have A1⊂A1′A_1\subset A'_1, by definition of Z1′Z'_1. Let ff be an arrow of Ker(φ,ψ\varphi , \psi ), let aa be its origin and bb its end. It follows from property (iv) in the definition of a distinguished subset (II, p. 197, definition 1) applied to the sequence ((f,1))((f,1)) that a∈A1′a\in A'_1 is equivalent to b∈A1′b\in A'_1. Since A1A_1 meets every connected component of Ker(φ,ψ\varphi , \psi ), we have A1′=A'_1= Som(Ker(φ,ψ\varphi , \psi )), which was to be proved.

b) Let A2′A'_2 be the set of vertices (a,b)(a, b) of H×GHH\times_GH such that the triple ((a,1),(b,1),(μ(a,b),−1))((a,1),(b,1),(\mu(a, b),-1)) belongs to Z2′Z'_2. By hypothesis, we have A2⊂A2′A_2\subset A'_2. Since A2A_2 meets each connected component of H×GHH\times_GH, it is enough to establish that, if there exists an arrow (f,f′)(f, f') joining a vertex (a′,b′)(a', b') to a vertex (a,b)(a, b), then (a,b)(a, b) belongs to A2′A'_2 if and only if analogously for (a′,b′)(a', b').

Put f′′=μ(f,f′)f''=\mu(f, f'); this is an arrow joining μ(a′,b′)\mu(a', b') to μ(a,b)\mu(a, b), and one has φ(f′′)=φ(f)\varphi (f'') =\varphi (f) and ψ(f′′)=ψ(f′)\psi (f'') =\psi (f'). By condition (iv) in the definition of a distinguished subset of Ω(Γ~)\Omega (\widetilde{\Gamma}) (loc. cit.) applied to the sequence ((f,1),(f′,1),(f′′,−1))((f,1),(f',1),(f'',-1)), the two conditions

(i) the triple ((a,1),(b,1),(μ(a,b),−1))((a,1),(b,1),(\mu(a, b),-1)) belongs to Z′Z';

(ii) the triple ((a′,1),(b′,1),(μ(a′,b′),−1))((a',1),(b',1),(\mu(a', b'),-1)) belongs to Z′Z';

are equivalent. This shows that (a,b)(a, b) belongs to A2′A'_2 if and only if (a′,b′)(a', b') belongs to A2′A'_2 and completes the proof of the lemma.

Let us now prove by induction on the integer n⩾1n\geqslant 1 that every element ((a1,ε1),(a2,ε2),...,(an,εn))((a_1, \varepsilon_1),(a_2, \varepsilon_2), . . . ,(a_n, \varepsilon_n)) of Ω(Γ~)\Omega (\widetilde{\Gamma}) belongs to Z′Z'.

A) Case in which n=1n= 1.

Let ((a,ε))((a, \varepsilon )) be an element of Ω(Γ~)\Omega (\widetilde{\Gamma}) of length 1. We have φ(a)=ψ(a)\varphi (a) =\psi (a), whence ((a,1))∈Z′((a,1))\in Z' by lemma 1. Condition (ii) in the definition of a distinguished subset then implies that ((a,−1))((a,-1)) belongs to Z′Z'.

B) Case in which n⩾2n\geqslant 2.

By virtue of condition (ii) for a distinguished subset, it is enough to treat the case in which ε2=1\varepsilon_2= 1. Suppose that ε1=1\varepsilon_1= 1; then (a1,a2)(a_1, a_2) is a vertex of H×GHH\times_GH; put a=μ(a1,a2)a=\mu(a_1, a_2); by lemma 1, the triple ((a1,1),(a2,1),(a,−1))((a_1,1),(a_2,1),(a,-1)) therefore belongs to Z′Z'. In the case in which ε1=−1\varepsilon_1=-1, we have φ(a1)=φ(a2)\varphi (a_1) =\varphi (a_2); we can therefore choose a vertex aa of H such that μ(a2,a)=a1\mu(a_2, a) =a_1 and the triple ((a2,1),(a,1),(a1,−1))((a_2,1),(a,1),(a_1,-1)) belongs to Z′Z', hence the triple ((a1,−1),(a2,1),(a,1))((a_1,-1),(a_2,1),(a,1)) also, by virtue of condition (ii) in the definition of a distinguished subset.

Then ((a,ε1),(a3,ε3),...,(an,εn))((a, \varepsilon_1),(a_3, \varepsilon_3), . . . ,(a_n, \varepsilon_n)) belongs to Ω(Γ~)\Omega (\widetilde{\Gamma}) and is of length n−1n-1. By induction, it is an element of Z′Z'. By virtue of conditions (i), (ii) and (iii) in the definition of a distinguished subset, we deduce successively that the elements

((a1,ε1),(a2,1),(a,−ε1),(a,ε1),(a3,ε3),...,(an,εn)) ((a_1, \varepsilon_1),(a_2,1),(a,-\varepsilon_1),(a, \varepsilon_1),(a_3, \varepsilon_3), . . . ,(a_n, \varepsilon_n))
((a,−ε1),(a,ε1),(a3,ε3),...,(an,εn),(a1,ε1),(a2,1)) ((a,-\varepsilon_1),(a, \varepsilon_1),(a_3, \varepsilon_3), . . . ,(a_n, \varepsilon_n),(a_1, \varepsilon_1),(a_2,1))
((a3,ε3),...,(an,εn),(a1,ε1),(a2,1)) ((a_3, \varepsilon_3), . . . ,(a_n, \varepsilon_n),(a_1, \varepsilon_1),(a_2,1))
((a1,ε1),(a2,1),(a3,ε3),...,(an,εn)) ((a_1, \varepsilon_1),(a_2,1),(a_3, \varepsilon_3), . . . ,(a_n, \varepsilon_n))

belong to Z′Z'. This completes the proof of the proposition.

Corollary 01UU

a) With the notation of proposition 4, suppose moreover that the mapping (Som(φ\varphi ), Som(ψ\psi )) of Som(H) into Som(G) ×\times Som(G) is injective and that its image is the graph of an equivalence relation in Som(G). Then, for every vertex (a,b)(a, b) of H×GHH\times_GH, there exists a unique vertex ca,bc_{a,b} of H such that φ(ca,b)=φ(a)\varphi (c_{a,b}) =\varphi (a) and ψ(ca,b)=ψ(b)\psi (c_{a,b}) =\psi (b).

b) Suppose further that for every arrow (f,f′)(f, f') of H×GHH\times_GH, there exists an arrow f′′f'' of H such that φ(f′′)=φ(f)\varphi (f'') =\varphi (f) and ψ(f′′)=ψ(f′)\psi (f'') =\psi (f').

Let A be a set of vertices of H×GHH\times_GH meeting each of its connected components. Then, the kernel of the morphism βa0\beta_{a_0} is the smallest distinguished subgroup of Coh(φ,ψ\varphi , \psi )a0_{a_0} containing the conjugacy classes c((a,1),(b,1),(ca,b,−1))c((a,1),(b,1),(c_{a,b},-1)), for (a,b)∈A(a, b)\in A.

Let R be the equivalence relation in Som(G) whose graph is the image of the mapping (Som(φ\varphi ), Som(ψ\psi )). Let (a,b)(a, b) be a vertex of H×GHH\times_GH. Then we have R{φ(a),ψ(a)}R\{\varphi (a), \psi (a)\} and R{φ(b),ψ(b)}R\{\varphi (b), \psi (b)\}, whence R{φ(a),ψ(b)}R\{\varphi (a), \psi (b)\} since ψ(a)=φ(b)\psi (a) =\varphi (b). Consequently, there exists a unique vertex cc of H such that φ(c)=φ(a)\varphi (c) =\varphi (a) and ψ(c)=ψ(b)\psi (c) =\psi (b); we denote it by μ(a,b)\mu(a, b).

For every arrow (f,f′)(f, f') of H×GHH\times_GH, choose an arrow f′′f'' of H such that φ(f′′)=φ(f)\varphi (f'') =\varphi (f) and ψ(f′′)=ψ(f)\psi (f'') =\psi (f) and denote it by μ(f,f′)\mu(f, f').

Thus a quiver morphism μ:H×GH→H\mu: H\times_GH\rightarrow H has been defined such that φ∘μ=φ∘\varphi \circ \mu=\varphi \circ pr1_1 and ψ∘μ=ψ∘\psi \circ \mu=\psi \circ pr2_2.

Let (a,b)(a, b) be a pair of vertices of H such that φ(a)=φ(b)\varphi (a) =\varphi (b). We have R{φ(a),ψ(a)}R\{\varphi (a), \psi (a)\} and R{φ(b),ψ(b)}R\{\varphi (b), \psi (b)\}, hence R{ψ(b),ψ(a)}R\{\psi (b), \psi (a)\}. Consequently there exists a unique vertex cc of H such that φ(c)=ψ(b)\varphi (c) =\psi (b) and ψ(c)=ψ(a)\psi (c) =\psi (a). The vertex μ(b,c)\mu(b, c) satisfies φ(μ(b,c))=φ(b)=φ(a)\varphi (\mu(b, c)) =\varphi (b) =\varphi (a) and ψ(μ(b,c))=\psi (\mu(b, c)) = ψ(c)=ψ(a)\psi (c) =\psi (a). Hence μ(b,c)=a\mu(b, c) =a because the mapping (Som(φ\varphi ), Som(ψ\psi )) of Som(H) into Som(G) ×\times Som(G) is injective.

Let A1A_1 be the set of vertices of Ker(φ,ψ\varphi , \psi ) and let Z1Z_1 be the set of elements of Ω(Γ~)\Omega (\widetilde{\Gamma}) of the form ((a,1))((a,1)), for a∈A1a\in A_1. Put A2=AA_2= A and let Z2Z_2 be the set of elements ((a,1),(b,1),(μ(a,b),−1))((a,1),(b,1),(\mu(a, b),-1)) of Ω(Γ~)\Omega (\widetilde{\Gamma}), for (a,b)∈A2(a, b)\in A_2. Let Z~\widetilde{Z} (resp. Z~1\widetilde{Z}_1, resp. Z~2\widetilde{Z}_2) denote the smallest distinguished subset of Ω(Γ~)\Omega (\widetilde{\Gamma}) containing Z1∪Z2Z_1\cup Z_2 (resp. Z1Z_1, resp. Z2Z_2).

Let a∈A1a\in A_1; one has φ(a)=ψ(a)\varphi (a) =\psi (a) ; let us put x=φ(a)x=\varphi (a). The vertex aa is the unique vertex of H such that φ(a)=ψ(a)=x\varphi (a) =\psi (a) =x; in particular, one has μ(a,a)=a\mu(a, a) =a. The triplet ((a,1),(a,1),(a,−1))((a,1),(a,1),(a,-1)) therefore belong to Z2Z_2. It then follows from conditions (i) and (iii) in the definition of a distinguished subset that ((a,1))((a,1)) belong to Z~2\widetilde{Z}_2. Consequently, one has Z1⊂Z~2Z_1\subset \widetilde{Z}_2, whence Z~=Z~2\widetilde{Z} =\widetilde{Z}_2.

By Proposition 4, Ω(Γ~)\Omega (\widetilde{\Gamma}) is the smallest distinguished subset of Ω(Γ~)\Omega (\widetilde{\Gamma}) containing Z2Z_2 and the kernel of βa0\beta_{a_0} is the smallest distinguished subgroup of Coh(φ,ψ\varphi , \psi )a0_{a_0} containing the conjugacy classes c(z)c(\mathbf{z}), for z∈Z2\mathbf{z}\in Z_2, whence the corollary.

Example 01UV

Let X and R be topological spaces, let oo and tt be continuous mappings of R into X, let C be the set of pairs (f,g)∈R2(f, g)\in R^2 such that t(f)=o(g)t(f) =o(g) in R, and let mm be a continuous mapping of C into R making the quiver (X,R,o,t)(X,R, o, t) into a groupoid; suppose moreover that the mapping f↦f−1f\mapsto f^{-1} of R into R is continuous. Then the hypotheses of the proposition are satisfied if one puts H=ϖ(R)H =\varpi (R), G=ϖ(X),φ=ϖ(o),ψ=ϖ(t)G =\varpi (X),\varphi =\varpi (o),\psi =\varpi (t) and μ=ϖ(m)\mu=\varpi (m).

Suppose further that R is the graph of an equivalence relation in X, the mappings oo and tt being deduced from the projections of X×XX\times X into X by passing to the subspaces. Then the hypotheses of the corollary are satisfied.

Remark 01UW

More generally, the hypotheses of the proposition are satisfied when (G,H,φ,ψ,μ)(G,H, \varphi , \psi , \mu) is a "groupoid of groupoids." This means that G and H are groupoids, φ\varphi and ψ\psi are groupoid morphisms of H into G, making the quadruplet (G,H,φ,ψ)(G,H, \varphi , \psi ) into a "quiver in groupoids"; finally, μ\mu is a law of composition in this "quiver" given by a groupoid morphism H×GH→HH\times_GH\rightarrow H, satisfying a certain number of properties expressing the associativity of the law and the fact that every "arrow" is invertible. The reader will moreover remark that the mappings to the fundamental groupoid studied in this work all fall under this case.

4. Isotropy Group of a Coequalizer

Let G and H be groupoids; let φ\varphi and ψ\psi be groupoid morphisms of H into G. The purpose of this No. is to summarize the calculus of the isotropy groups of the cohomotopizer (II, p. 193, prop. 6) and the comparison of the isotropy groups of the cohomotopizer and of the coequalizer made in the preceding No. in order to deduce from it the calculus of the isotropy groups of the coequalizer Coeg(φ,ψ\varphi , \psi ).

Let us denote by α:G→\alpha : G\rightarrow Coh(φ,ψ\varphi , \psi ),β:,\beta : Coh(φ,ψ\varphi , \psi )→\rightarrow Coeg(φ,ψ\varphi , \psi ) and γ=β∘\gamma =\beta \circ α:G→\alpha : G\rightarrow Coeg(φ,ψ\varphi , \psi ) the canonical groupoid morphisms. Let us denote by h:h: Som(H) →\rightarrow Fl(Coh(φ,ψ\varphi , \psi )) the canonical homotopy; let us recall that the set of vertices of Coh(φ,ψ\varphi , \psi ) is equal to Som(G).

Let Γ\Gamma denote the quiver (Som(G), Som(H), Som(φ\varphi ), Som(ψ\psi )); let then φ0\varphi_0 and ψ0\psi_0 denote the mappings deduced from φ\varphi and ψ\psi by passing to orbits and Γ0\Gamma_0 the framework (Orb(G), Orb(H),φ0,ψ0, \varphi_0, \psi_0) of the pair (φ,ψ)(\varphi , \psi ). We shall assume that the quiver Γ0\Gamma_0 is connected and that the set of its vertices is not empty, which amounts to assuming that the groupoids Coh(φ,ψ\varphi , \psi ) and Coeg(φ,ψ\varphi , \psi ) are transitive.

Recall (cf. II, p. 192, definition 4) that a basic equipment consists in the data of a family (a,b,c1,c2,T,i0)(a, b, c_1, c_2,T, i_0) where: for every i∈i\in Orb(G),a(i),a(i) is a vertex in the orbit ii of G; for every j∈j\in Orb(H),b(j),b(j) is a vertex in the orbit jj of H,c1(j),c_1(j) and c2(j)c_2(j) are arrows of G joining respectively φ(b(j))\varphi (b(j)) to a(φ0(j))a(\varphi_0(j)) and ψ(b(j))\psi (b(j)) to a(ψ0(j))a(\psi_0(j)); T is a subquiver of Γ0\Gamma_0 whose associated tree is a maximal tree of the graph Γ~0\widetilde{\Gamma}_0; finally, i0i_0 is an orbit of G and one sets a0=a(i0)a_0=a(i_0).

One then defines a morphism of quivers τ0\tau_0 from Γ0\Gamma_0 into Coh(φ,ψ\varphi , \psi ) such that τ0(i)=β(a(i))\tau_0(i) =\beta (a(i)) and τ0(j)=α(c1(j))−1h(b(j))α(c2(j))\tau_0(j) =\alpha (c_1(j))^{-1}h(b(j))\alpha (c_2(j)) for i∈i\in Orb(G) and j∈j\in Orb(H). If ii belongs to Orb(G), let did_i be the unique class of paths in the graph T~\widetilde{T} joining i0i_0 to ii, and let δi\delta_i denote its image in Coh(φ,ψ\varphi , \psi ) under the morphism of groupoids τ0\tau_0.

From these data, Prop. 6 (II, p. 193) yields a surjective homomorphism

Λ:(i∈π∗0(G)Ga(i))∗\Lambda :(_{i\in\pi}*_{_0(G)}G_{a(i)})* F(Orb(H)) →\rightarrow Coh(φ,ψ\varphi , \psi )a(i0)_{a(i_0)}

and describes generators of its kernel, thus providing a presentation of the group Coh(φ,ψ\varphi , \psi )a0_{a_0}.

The set of loops of length ⩾1\geqslant 1 in the graph Γ~\widetilde{\Gamma} associated to Γ is identified with the set Ω(Γ~)\Omega (\widetilde{\Gamma}) of sequences (z1,...,zn)(z_1, . . . , z_n) of arrows of Γ~\widetilde{\Gamma}, indexed by Z/nZ\mathbf{Z}/n\mathbf{Z}, where nn runs through the set of integers ⩾1\geqslant 1, such that for every k∈Z/nZk\in \mathbf{Z}/n\mathbf{Z}, the terminus of zkz_k is the origin of zk+1z_{k+1}. Let Z be a subset of Ω(Γ~)\Omega (\widetilde{\Gamma}) such that the conjugacy classes c(z)c(z), for z∈Zz\in Z, generate the kernel of the surjective homomorphism βa0\beta_{a_0}. The homomorphism βa0∘Λ\beta_{a_0}\circ \Lambda is surjective; to deduce its kernel, that is to say a presentation of the group Coeg(φ,ψ\varphi , \psi )β(a0)_{\beta(a_0)}, it remains to choose, for every z∈Zz\in Z, an element C(z)C(z) in the group (∗Ga(i))∗F(π0(H))(*G_{a(i)})*F(\pi_0(H))

i∈i\inOrb(G)

such that Λ(C(z))\Lambda (C(z)) belongs to the conjugacy class c(z)c(z).

Let therefore z=(z1,...,zn)z= (z_1, . . . , z_n) be an element of Ω(Γ~)\Omega (\widetilde{\Gamma}); put zk=(yk,εk)z_k= (y_k, \varepsilon_k), where yk∈y_k\in Fl(Γ) = Som(H) and εk∈{±1}\varepsilon_k\in \{\pm 1\}. By definition, c(z)c(z) is the conjugacy class of the element

gh(y1)ε1...h(yn)εng−1 gh(y_1)^{\varepsilon_1}. . . h(y_n)^{\varepsilon_n}g^{-1}

of the groupoid Coh(φ,ψ\varphi , \psi )a0_{a_0}, where gg is an arbitrary arrow in Coh(φ,ψ\varphi , \psi ) joining a0a_0 to the origin of z1z_1. For k∈Z/nZk\in \mathbf{Z}/n\mathbf{Z}, let jkj_k be the orbit of yky_k in H and choose an arrow fkf_k of H joining yky_k to the vertex b(jk)b(j_k). By definition of a homotopy, one then has the relation

h(yk)(α∘ψ)(fk)=(α∘φ)(fk)h(b(jk)) h(y_k)(\alpha \circ \psi )(f_k) = (\alpha \circ \varphi )(f_k)h(b(j_k))

in the groupoid Coh(φ,ψ\varphi , \psi ). Consequently, using the definition of the quiver morphism τ0\tau_0, one has

h(yk)=(α∘φ)(fk)⋅h(b(jk))⋅(α∘ψ)(fk)−1 h(y_k) = (\alpha \circ \varphi )(f_k)\cdot h(b(j_k))\cdot (\alpha \circ \psi )(f_k)^{-1}
=(α∘φ)(fk)⋅(α∘c1)(jk)⋅τ0(jk)⋅(α∘c2)(jk)−1⋅(α∘ψ)(fk)−1 = (\alpha \circ \varphi )(f_k)\cdot (\alpha \circ c_1)(j_k)\cdot \tau_0(j_k)\cdot (\alpha \circ c_2)(j_k)^{-1}\cdot (\alpha \circ \psi )(f_k)^{-1}
=(α∘φ)(fk)⋅(α∘c1)(jk)⋅δφ0(jk)−1⋅ = (\alpha \circ \varphi )(f_k)\cdot (\alpha \circ c_1)(j_k)\cdot \delta_{\varphi_0(j_k)}^{-1}\cdot
⋅δφ0(jk)⋅τ0(jk)⋅δψ0(jk)−1⋅ \cdot \delta_{\varphi_0(j_k)}\cdot \tau_0(j_k)\cdot \delta_{\psi_0(j_k)}^{-1}\cdot
⋅δψ0(jk)⋅(α∘c2)(jk)−1⋅(α∘ψ)(fk)−1 \cdot \delta_{\psi_0(j_k)}\cdot (\alpha \circ c_2)(j_k)^{-1}\cdot (\alpha \circ \psi )(f_k)^{-1}
=ukΛ(jk)vk =u_k\Lambda (j_k)v_k

where we have put

uk=α(φ(fk)c1(jk))⋅δφ0(jk)−1u_k=\alpha (\varphi (f_k)c_1(j_k))\cdot \delta_{\varphi_0(j_k)}^{-1} and vk=δψ0(jk)⋅α(ψ(fk)c2(jk))−1v_k=\delta_{\psi_0(j_k)}\cdot \alpha (\psi (f_k)c_2(j_k))^{-1}. For every element k∈Z/nZk\in \mathbf{Z}/n\mathbf{Z}, define arrows u~k,v~k\widetilde{u}_k,\widetilde{v}_k in G by

h(yk)εk=u~kΛ(jk)εkv~k h(y_k)^{\varepsilon_k}=\widetilde{u}_k\Lambda (j_k)^{\varepsilon_k}\widetilde{v}_k

so that

(uk,vk)(u_k, v_k) if εk=1\varepsilon_k= 1 ;

(u~k,v~k)=−− (\widetilde{u}_k,\widetilde{v}_k) =_{--}

(vk1,uk1)(v_k^1, u_k^1) if εk=−1\varepsilon_k=-1.

Let xkx_k denote the origin of the arrow h(yk)εkh(y_k)^{\varepsilon_k}; its end is then xk+1x_{k+1}; let iki_k be the orbit of xkx_k. We then define a loop λk(z)\lambda_k(z) at a(ik)a(i_k) in the groupoid G by the formula

(1)

c2(jk−1)−1ψ(fk−1)−1φ(fk)c1(jk)c_2(j_{k-1})^{-1}\psi (f_{k-1})^{-1}\varphi (f_k)c_1(j_k) if (εk−1,εk)=(1,1)(\varepsilon_{k-1}, \varepsilon_k) = (1,1);

c2(jk−1)−1ψ(fk−1)−1ψ(fk)c2(jk)c_2(j_{k-1})^{-1}\psi (f_{k-1})^{-1}\psi (f_k)c_2(j_k) if (εk−1,εk)=(1,−1)(\varepsilon_{k-1}, \varepsilon_k) = (1,-1); λk(z)=\lambda_k(z) =

c1(jk−1)−1φ(fk−1)−1φ(fk)c1(jk)c_1(j_{k-1})^{-1}\varphi (f_{k-1})^{-1}\varphi (f_k)c_1(j_k) if (εk−1,εk)=(−1,1)(\varepsilon_{k-1}, \varepsilon_k) = (-1,1);

c1(jk−1)−1φ(fk−1)−1ψ(fk)c2(jk)c_1(j_{k-1})^{-1}\varphi (f_{k-1})^{-1}\psi (f_k)c_2(j_k) if (εk−1,εk)=(−1,−1)(\varepsilon_{k-1}, \varepsilon_k) = (-1,-1). By construction, one has

Λ(λk(z))=v~k−1u~k \Lambda (\lambda_k(z)) =\widetilde{v}_{k-1}\widetilde{u}_k

so that the image under the homomorphism Λ of the element

C(z)=λ1(z)(j1)ε1λ2(z)(j2)ε2...λn(z)(jn)εn(2) C(z) =\lambda_1(z)(j_1)^{\varepsilon_1}\lambda_2(z)(j_2)^{\varepsilon_2}. . . \lambda_n(z)(j_n)^{\varepsilon_n} \tag{2}

belongs to the conjugacy class c(z)c(z).

Definition 3 01UX

Let G and H be groupoids; let φ\varphi and ψ\psi be morphisms of groupoids from H into G. Suppose that the groupoid Coeg(φ,ψ\varphi , \psi ) is transitive. Let (a,b,c1,c2,T,i0)(a, b, c_1, c_2,T, i_0) be a basic equipment of the pair (φ,ψ)(\varphi , \psi ).

A complementary equipment is the data of a subset Z of Ω(Γ~)\Omega (\widetilde{\Gamma}) such that the conjugacy classes c(z)c(\mathbf{z}) for z∈\mathbf{z}\in Z generate the kernel of the homomorphism βa(i0)\beta_{a(i_0)} and, for every element z=((y1,ε1),...,(yn,εn))\mathbf{z}= ((y_1, \varepsilon_1), . . . ,(y_n, \varepsilon_n)) of Z, of a sequence f(z)=(f1,...,fn)f(\mathbf{z}) = (f_1, . . . , f_n) of arrows in H such that fkf_k joins yky_k to the vertex b(jk)b(j_k), jkj_k denoting the orbit of yky_k in H.

A complete equipment of the pair (φ,ψ)(\varphi , \psi ) is the data of a basic equipment and of a complementary equipment.

Proposition 5 01UY

Let G and H be groupoids; let φ\varphi and ψ\psi be morphisms of groupoids from H into G. Suppose that the groupoid Coeg(φ,ψ\varphi , \psi ) is transitive. Let γ:G→\gamma : G\rightarrow Coeg(φ,ψ\varphi , \psi ) denote the canonical morphism of groupoids.

Let us endow the pair (φ,ψ)(\varphi , \psi ) with a complete equipment

(a,b,c1,c2,T,i0,Z,(f(z))z∈Z) (a, b, c_1, c_2,T, i_0,Z,(f(\mathbf{z}))_{\mathbf{z}\in Z})

For j∈j\in Orb(H), let φj:Hb(j)→Ga(φ0(j))\varphi_j: H_{b(j)}\rightarrow G_{a(\varphi_0(j))} and ψj:Hb(j)→Ga(ψ0(j))\psi_j: H_{b(j)}\rightarrow G_{a(\psi_0(j))} be the group homomorphisms Int(c1(j)c_1(j))−1∘φb(j)^{-1}\circ \varphi_{b(j)} and Int(c2(j)c_2(j))−1∘ψb(j)^{-1}\circ \psi_{b(j)} respectively, where φ0\varphi_0 and ψ0\psi_0 denote the canonical mappings deduced from φ\varphi and ψ\psi by passing to orbits. Let τ\tau be the morphism of the framework Γ0\Gamma_0 of the pair (φ,ψ)(\varphi , \psi ) into Coeg(φ,ψ\varphi , \psi ) defined by Som(τ\tau )(i)=γ(a(i))(i) =\gamma (a(i)) if i∈i\in Orb(G) and such that Fl(τ\tau )(j)(j) is the path γ(c1(j))−1γ(c2(j))\gamma (c_1(j))^{-1}\gamma (c_2(j)) in Coeg(φ,ψ\varphi , \psi ). If ii is an orbit of G, let cic_i be the unique class of paths in T~\widetilde{T} joining i0i_0 to ii and put δi=τ~(ci)\delta_i=\widetilde{\tau}(c_i), where τ~:\widetilde{\tau}: Grp(G) →\rightarrow Coeg(φ,ψ\varphi , \psi ) is the canonical groupoid morphism deduced from τ\tau.

There then exists a unique group homomorphism

λ:(∗Ga(i))∗\lambda :(*G_{a(i)})* F(Orb(H)) →\rightarrow Coeg(φ,ψ\varphi , \psi )γ(a(i0))_{\gamma(a(i_0))}

i∈i\inOrb(G)

such that

λ(f)=δiγa(i)(f)δi−1\lambda (f) =\delta_i\gamma_{a(i)}(f)\delta_i^{-1} for i∈i\in Orb(G) and f∈Ga(i)f\in G_{a(i)},

λ(j)=δφ0(j)τ(j)δψ0(j)−1\lambda (j) =\delta_{\varphi_0(j)}\tau (j)\delta_{\psi_0(j)}^{-1} for j∈j\in Orb(H).

The homomorphism λ\lambda is surjective; its kernel is the smallest normal subgroup containing the following elements:

(R1R_1)r1(j)=jr_1(j) =j for jj in Fl(T) ;

(R2R_2)r2(j,f)=φj(f)jψj(f)−1j−1r_2(j, f) =\varphi_j(f)j\psi_j(f)^{-1}j^{-1}

for j∈j\in Orb(H) and f∈Hb(j)f\in H_{b(j)};

(R3R_3)r3(z)=λ1(z)j1ε1λ2(z)j2ε2...λn(z)jnεnr_3(z) =\lambda_1(z)j_1^{\varepsilon_1}\lambda_2(z)j_2^{\varepsilon_2}. . . \lambda_n(z)j_n^{\varepsilon_n}

for z=((y1,ε1),...,(yn,εn))∈Zz= ((y_1, \varepsilon_1), . . . ,(y_n, \varepsilon_n))\in Z,

where the loops λi(z)\lambda_i(z) are defined by formula (1), p. 208.

The existence and uniqueness of such a morphism result from the universal property of the free product of a family of groups (A, I, p. 85, prop. 8). Let us denote by α:G→\alpha : G\rightarrow Coh(φ,ψ\varphi , \psi ) and β:\beta : Coh(φ,ψ\varphi , \psi )→\rightarrow Coeg(φ,ψ\varphi , \psi ) the canonical morphisms, so that γ\gamma = β∘α\beta \circ \alpha. Let also h:h: Som(H) →\rightarrow Fl(Coh(φ,ψ\varphi , \psi )) be the canonical homotopy joining α∘φ\alpha \circ \varphi to α∘ψ\alpha \circ \psi. Since β\beta is the groupoid morphism obtained by contraction of the arrows in the image of hh, we have

Fl(τ\tau )(j)=γ(c1(j))−1γ(c2(j))=γ(c1(j)−1)β(h(j))γ(c2(j))=β(τ0(j))(j) =\gamma (c_1(j))^{-1}\gamma (c_2(j)) =\gamma (c_1(j)^{-1})\beta (h(j))\gamma (c_2(j)) =\beta (\tau_0(j)) in Coeg(φ,ψ\varphi , \psi ), where τ0:Γ0→\tau_0: \Gamma_0\rightarrow Coh(φ,ψ\varphi , \psi ) denotes the quiver morphism deduced from the basic equipment (a,b,c1,c2,T,i0)(a, b, c_1, c_2,T, i_0). Consequently, the homomorphism λ\lambda is the composite of the homomorphism Λ defined in prop. 6 (II, p. 193) and the surjective homomorphism βa(i0)\beta_{a(i_0)}. In particular it is surjective.

Let z∈Zz\in Z; by construction, Λ(r3(z))\Lambda (r_3(z)) belongs to the conjugacy class c(z)c(z). By definition of a complementary equipment, the kernel of the homomorphism βa(i0)\beta_{a(i_0)} is therefore the smallest invariant subgroup of Coh(φ,ψ\varphi , \psi )a(i0)_{a(i_0)} containing the elements Λ(r3(z))\Lambda (r_3(z)) for z∈Zz\in Z. Since the homomorphism Λ is surjective, the kernel of the homomorphism λ=βa(i0)∘Λ\lambda =\beta_{a(i_0)}\circ \Lambda is therefore the smallest invariant subgroup of the group (∗Ga(i))∗F(π0(H))(*G_{a(i)})*F(\pi_0(H)) containing the generators of the kernel

i∈i\inOrb(G)

of Λ given by formulae (R1R_1),(R2), (R_2), together with the elements defined by formulae (R3R_3). The proposition is thus proved.

5. Quotient of a groupoid by the action of a group

Let G be a transitive groupoid, let K be a group, and let θ:K→\theta : K\rightarrow Aut(G)∘^{\circ} be a group homomorphism of K into the opposite group of the automorphism group of the groupoid G. We shall say that the group K acts on the right on G. If k∈Kk\in K, we shall sometimes denote by k∗xk^*x (resp. k∗fk^*f) the image of a vertex xx (resp. of an arrow ff) of G under the groupoid automorphism θ(k)\theta (k).

Let ∣K∣|K| be the groupoid whose set of vertices is K and whose set of arrows joining two vertices is empty if these vertices are distinct, and is reduced to one element otherwise. Let H be the product groupoid G×∣K∣G\times |K|; a vertex of H is a pair (a,k)(a, k), where aa is a vertex of G and kk is an element of K; if ff is an arrow of G joining a vertex aa to a vertex bb, we shall denote by (f,k)(f, k) the unique arrow of H joining (a,k)(a, k) to (b,k)(b, k). Let φ:H→G\varphi : H\rightarrow G be the groupoid morphism given by the first projection and let ψ:H→G\psi : H\rightarrow G be the groupoid morphism such that Som(ψ\psi )((a,k))=k∗a((a, k)) =k^*a and Fl(ψ\psi )((f,k))=k∗f((f, k)) =k^*f if k∈K,a∈k\in K,a\in Som(G) and f∈f\in Fl(G).

Let us denote by G/KG/K the coequalizer Coeg(φ,ψ\varphi , \psi ) and let γ:G→G/K\gamma : G\rightarrow G/K be the canonical groupoid morphism.

Let oo be a vertex of G. For k∈Kk\in K, choose an arrow ckc_k joining k∗ok^*o to oo in G; there exists one since G is transitive. For k∈Kk\in K, let us denote by Fix(kk) the subgroupoid of G whose vertices (resp. arrows) are the elements of Som(G) (resp. of Fl(G)) fixed by kk; choose a set AkA_k of vertices of Fix(kk) meeting all the orbits of this groupoid. For k∈Kk\in K and a∈Aka\in A_k, choose also an arrow f(a,k)f_{(a,k)} in G joining the vertex aa to the vertex oo.

Proposition 6 01UZ

The unique group homomorphism

λ:Go∗F(K)→(G/K)γ(o) \lambda : G_o*F(K)\rightarrow (G/K)_{\gamma(o)}

such that λ(f)=γo(f)\lambda (f) =\gamma_o(f) for f∈Gof\in G_o and λ([k])=γ(ck)\lambda ([k]) =\gamma (c_k) for k∈Kk\in K is surjective. Its kernel is the smallest normal subgroup of Go∗F(K)G_o*F(K) containing the following elements:

(R2R_2)r2(k,f)=[k]−1f[k](ck−1k∗(f)−1ck)r_2(k, f) = [k]^{-1}f[k](c^{-1}_kk^*(f)^{-1}c_k)

for k∈Kk\in K and f∈Gof\in G_o;

(R3′R'_3)r3′(k,a)=[k](ck−1k∗(f(a,k))−1f(a,k)))r'_3(k, a) = [k](c^{-1}_kk^*(f_{(a,k)})^{-1}f_{(a,k)}))

for k∈K−{e}k\in K-\{e\} and a∈Aka\in A_k

(R3′′R''_3)r3′′(k,h)=[kh]−1[k][h](ch−1h∗(ck−1)ckh)r''_3(k, h) = [kh]^{-1}[k][h](c^{-1}_hh^*(c^{-1}_k)c_{kh})

for kk and h∈Kh\in K.

Since G is transitive, the mapping which to an element k∈Kk\in K associates the orbit of (o,k)(o, k) is a bijection of K onto the set of orbits of H. Thus Orb(G) is identified with {o}\{o\} and Orb(H) with K. The framework Γ0\Gamma_0 of the pair (φ,ψ)(\varphi , \psi ) is then identified with the quiver having a unique vertex oo and whose set of arrows is K. Let T be the subquiver of Γ0\Gamma_0 with set of vertices {o}\{o\} and whose set of arrows is empty; the associated graph is the unique maximal tree of the graph Γ~0\widetilde{\Gamma}_0.

The family (o,(o,k)k∈K,(eo)k∈K,(ck)k∈K,T,o)(o,(o, k)_{k\in K},(e_o)_{k\in K},(c_k)_{k\in K},T, o) is a basic equipment of the pair (φ,ψ)(\varphi , \psi ).

The mapping f↦(f,k)f\mapsto (f, k) defines an isomorphism of the isotropy group GoG_o onto the isotropy group H(o,k)H_{(o,k)}; by means of this isomorphism, the homomorphisms φk\varphi_k and ψk\psi_k of H(o,k)H_{(o,k)} into GoG_o defined by Prop. 5 of II, p. 208 are given by

(3) φk(f,k)=f\varphi_k(f, k) =f et ψk(f,k)=ck−1k∗(f)ck\psi_k(f, k) =c^{-1}_kk^*(f)c_k,

for k∈Kk\in K and f∈Gof\in G_o.

The quiver H×GHH\times_GH is the equalizer in H×HH\times H of the quiver morphisms ψ∘\psi \circ pr1_1 and φ∘\varphi \circ pr2_2. Its vertices are the pairs ((a,k),(b,h))((a, k),(b, h)) where aa and bb are vertices of G and kk and hh elements of K such that b=k∗ab=k^*a, and its arrows are the pairs ((f,k),(g,h))((f, k),(g, h)) where ff and gg are arrows of G, and kk and hh elements of K such that g=k∗fg=k^*f; the origin of the arrow ((f,k),(g,h))((f, k),(g, h)) is equal to ((o(f),k),(o(g),h))((o(f), k),(o(g), h)); its end is equal to ((t(f),k),(t(g),h))((t(f), k),(t(g), h)).

One then defines a quiver morphism μ\mu from H×GHH\times_GH into H by putting μ((a,k),(b,h))=(a,kh)\mu((a, k),(b, h)) = (a, kh) and μ((f,k),(g,h))=(f,kh)\mu((f, k),(g, h)) = (f, kh). We have φ∘μ=φ∘\varphi \circ \mu=\varphi \circ pr1_1 and ψ∘μ=ψ∘\psi \circ \mu=\psi \circ pr2_2.

Let xx and yy be vertices of H such that φ(x)=φ(y)\varphi (x) =\varphi (y). Then there exist a vertex aa of G and elements kk and hh of K such that x=(a,k)x= (a, k) and y=(a,h)y= (a, h). Put z=(h∗a,h−1k)z= (h^*a, h^{-1}k); then μ(y,z)=x\mu(y, z) =x. This shows that the pair (φ,ψ)(\varphi , \psi ) satisfies the hypotheses of Prop. 4 (II, p. 202).

A vertex (a,k)(a, k) of H belong to the groupoid Ker(φ,ψ\varphi , \psi ), equalizer of φ\varphi and ψ\psi, if and only if k∗a=ak^*a=a, that is to say, if kk belong to the stabilizer of the vertex aa in the group K. Let A1A_1 be the set of pairs (a,k)(a, k), for k∈Kk\in K and a∈Aka\in A_k; it meets all the orbits of the subgroupoid Ker(φ,ψ\varphi , \psi ) of H. Let Z1Z_1 be the subset of Ω(Γ~)\Omega (\widetilde{\Gamma}) formed by the sequences of the form (((a,k),1))(((a, k),1)), for (a,k)∈A1(a, k)\in A_1. For z=((a,k),1)∈Z1,(f(a,k),k)\mathbf{z}= ((a, k),1)\in Z_1,(f_{(a,k)}, k) is an arrow of H joining (a,k)(a, k) to (o,k)(o, k). Put f(z)=((f(a,k),k))f(\mathbf{z}) = ((f_{(a,k)}, k)).

The set A2A_2 of vertices of H×GHH\times_GH of the form ((o,k),(k∗o,h))((o, k),(k^*o, h)), for (k,h)∈K2(k, h)\in K^2, meets all the orbits of H×GHH\times_GH. Observe also that one has μ((o,k),(k∗o,h))=(o,kh)\mu((o, k),(k^*o, h)) = (o, kh). The arrows (eo,k)(e_o, k), (ck,h)(c_k, h), (eo,kh)(e_o, kh) in H join respectively (o,k)(o, k), (k∗o,h)(k^*o, h), (o,kh)(o, kh) to (o,k)(o, k), (o,h)(o, h) and (o,kh)(o, kh). Let Z2Z_2 denote the set of sequences of the form (((o,k),1),((k∗o,h),1),((o,kh),−1))(((o, k),1),((k^*o, h),1),((o, kh),-1)) in Ω(Γ~)\Omega (\widetilde{\Gamma}); for such an element z\mathbf{z} of Z2Z_2, put f(z)=((eo,k),(ck,h),(eo,kh))f(\mathbf{z}) = ((e_o, k),(c_k, h),(e_o, kh)).

By Prop. 4 of II, p. 202, the set Z=Z1∪Z2Z = Z_1\cup Z_2 and the family (f(z))z∈Z(f(\mathbf{z}))_{z\in Z} is a complementary equipment.

Let k∈Kk\in K and let a∈Aka\in A_k. The element C((a,k),1)C((a, k),1) of the group Go∗F(K)G_o*F(K) defined by formula (2) of II, p. 208 is equal to

ck−1ψ((f(a,k),k))−1φ(f(a,k))eo[k]=(ck−1k∗(f(a,k))−1f(a,k))[k](4) c^{-1}_k\psi ((f_{(a,k)}, k))^{-1}\varphi (f_{(a,k)})e_o[k] =(c^{-1}_kk^*(f_{(a,k)})^{-1}f_{(a,k)})[k] \tag{4}

Let kk and h∈Kh\in K be given. One verifies that the element

C(((o,k),1),((k∗o,h),1),((o,kh),−1)) C(((o, k),1),((k^*o, h),1),((o, kh),-1))

of the group Go∗F(K)G_o*F(K) defined by formula (2) of II, p. 208 is equal to

[k][h](ch−1h∗(ck−1)ckh)[kh]−1(5) [k][h](c^{-1}_hh^*(c^{-1}_k)c_{kh})[kh]^{-1} \tag{5}

The elements r3′(e,a)r'_3(e, a) given by the relations (R3′R'_3) for k=ek=e are all equal to [e]ce−1[e]c^{-1}_e, an element of the group Go∗F(K)G_o*F(K) which one obtains by applying the relation (R3′′R''_3) to k=h=ek=h=e. Taking account of relations (3), (4), and (5), the proposition therefore follows from II, p. 208, Prop. 5.

Corollary 1 01V0

Suppose that the group K is generated by the stabilisers of the vertices of G. Then the group homomorphism γo:Go→(G/K)γ(o)\gamma_o: G_o\rightarrow (G/K)_{\gamma(o)} is surjective. Moreover, if the groupoid G is simply transitive, the same is true of the groupoid G/KG/K.

The relation (R3′′R''_3) implies λ([e])=λ(ce)=γo(ce)\lambda ([e]) =\lambda (c_e) =\gamma_o(c_e). The relations (R3′′R''_3) then imply that the set of k∈Kk\in K such that λ([k])\lambda ([k]) belong to the image of γo\gamma_o is a subgroup of K. Finally, the relations (R3′R'_3) show that, for every element k∈Kk\in K whose stabiliser is not empty, λ([k])\lambda ([k]) belong to the image of γo\gamma_o. The corollary follows.

Remark 1 01V1

One can give another description, sometimes more convenient, of the group (G/K)γ(o)(G/K)_{\gamma(o)}. For this, let M=K×GoM = K\times G_o and define a law of composition in M by the formula

(k,a)⋅(h,b)=(kh,ckh−1h∗(cka)chb) (k, a)\cdot (h, b) = (kh, c^{-1}_{kh}h^*(c_ka)c_hb)

for k,h∈Kk,h\in K and a,b∈Goa,b\in G_o. One verifies that this law of composition is associative, that (e,ce−1)(e, c^{-1}_e) is an identity element, and that the element (k−1,ck−−11(k−1)∗(cka)−1ce)(k^{-1}, c^{-1_1}_{k^-}(k^{-1})^*(c_ka)^{-1}c_e) is the inverse of (k,a)(k, a). It therefore endows M with a group structure. Moreover, the mapping λ′:M→(G/K)γ(o)\lambda ': M\rightarrow (G/K)_{\gamma(o)} defined by (k,a)↦λ([k]a)(k, a)\mapsto \lambda ([k]a) is a group homomorphism. Let α′\alpha ' be the unique group morphism of Go∗F(K)G_o*F(K) into M such that α′(f)=\alpha '(f) = (e,f)(e, f) if f∈Gof\in G_o and α′([k])=(k,eo)\alpha '([k]) = (k, e_o) if k∈Kk\in K; one has λ′∘α′=λ\lambda '\circ \alpha '=\lambda.

The relations (R2R_2) and (R3′′R''_3) show that every element of (G/K)γ(o)(G/K)_{\gamma(o)} is the image under the homomorphism λ\lambda of an element of Go∗F(K)G_o*F(K) of the form [k]f[k]f, with f∈Gof\in G_o and k∈Kk\in K. Hence the homomorphism λ′\lambda ' is surjective. One verifies moreover that the image under α′\alpha ' of an element of Go∗G_o* F(K) of the form (R2)(R_2) or (R3′′)(R''_3) is equal to the identity element. Consequently, the kernel of the homomorphism λ′\lambda ' is the smallest normal subgroup of M containing the images under α′\alpha ' of the elements of Go∗F(K)G_o*F(K) of the form (R3′R'_3).

Corollary 2 01V2

Suppose that the group K operates freely on Som(G). There then exists a unique group homomorphism π:(G/K)γ(o)→\pi : (G/K)_{\gamma(o)}\rightarrow K whose kernel contains the image of γo\gamma_o and such that π(λ([k]))=k\pi (\lambda ([k])) =k for every k∈Kk\in K. Moreover, Go⟶γo(G/K)γ(o)−→πG_o\longrightarrow^{\gamma_o}(G/K)_{\gamma(o)}-\rightarrow^{\pi} K is an extension of K by GoG_o.

If such a group homomorphism π\pi exists, the group homomorphism π∘λ\pi \circ \lambda is necessarily equal to the unique group homomorphism pp from Go∗F(K)G_o*F(K) into K such that p(f)=ep(f) =e for f∈Gof\in G_o and p([k])=kp([k]) =k. It is immediate to verify that the elements of Go∗F(K)G_o*F(K) defined by formulas (R2R_2) and (R3′′R''_3) belong to the kernel of pp. By assumption, there is no element of type (R3′R'_3). Thus, the kernel of the morphism λ\lambda contains that of pp. Consequently, there exists a unique group homomorphism π:(G/K)γ(o)→K\pi : (G/K)_{\gamma(o)}\rightarrow K such that π∘λ=p\pi \circ \lambda =p.

It is evident that the homomorphism π\pi is surjective. To prove that the homomorphism γo\gamma_o is injective and that its image is exactly the kernel of π\pi, let us remark that the homomorphism λ′:M→(G/K)γ(o)\lambda ': M\rightarrow (G/K)_{\gamma(o)} (II, p. 213, remark 1) is an isomorphism, since it has been assumed that K operates freely on Som(G). The composite homomorphism (λ′)−1∘γo(\lambda ')^{-1}\circ \gamma_o from GoG_o into M is given by f↦(e,f)f\mapsto (e, f), whereas the homomorphism π∘λ′:M→K\pi \circ \lambda ': M\rightarrow K maps (k,f)(k, f) to kk. The corollary follows.

Corollary 3 01V3

Suppose that the groupoid G is simply transitive. Let K0K_0 be the subgroup of K generated by the stabilizers of the vertices of G. The mapping from K into (G/K)γ(o)(G/K)_{\gamma(o)} which to k∈Kk\in K associates γ(ck)\gamma (c_k) is a surjective group homomorphism, with kernel K0K_0.

If an element k∈Kk\in K fixes a vertex aa of G, the element g−1kgg^{-1}kg fixes the vertex g∗ag^*a; this implies that K0K_0 is a normal subgroup of K.

By proposition 5, the unique homomorphism λ:F(K)→\lambda : F(K)\rightarrow (G/K)γ(o)(G/K)_{\gamma(o)} such that λ([k])=γ(ck)\lambda ([k]) =\gamma (c_k) is surjective, and its kernel is the smallest normal subgroup of F(K) containing the elements [k][k], where kk is an element of K which fixes a vertex of G, and the elements [kh]−1[k][h][kh]^{-1}[k][h], where (k,h)∈K2(k, h)\in K^2. In particular, the mapping λ′:K→(G/K)γ(o)\lambda ': K\rightarrow (G/K)_{\gamma(o)}, defined by λ′(k)=γ(ck)=λ([k])\lambda '(k) =\gamma (c_k) =\lambda ([k]) for k∈Kk\in K, is a group homomorphism. One has λ=λ′∘p\lambda =\lambda '\circ p, where p:F(K)→Kp: F(K)\rightarrow K denotes the canonical surjective group homomorphism. Consequently, the homomorphism λ′\lambda ' is surjective and its kernel is the smallest normal subgroup of K which contains the elements kk, for kk fixing a vertex of G, that is to say K0K_0. This proves the proposition.

Exercises

See the exercises for § 5.

Machine translation of the English, by gpt-5.4, against glossary version 34. Not checked by a person.