§ 4. Homotopies
1. Definition of homotopies
Definition 1 01U3
Let G be a groupoid, H a quiver, and morphisms of quivers from H into G. A homotopy connecting to is a mapping of the set of vertices of H into the set of arrows of G having the following properties:
(i) For every vertex of H, the arrow has origin and end ;
(ii) For every arrow of H, with origin and end , one has .
One says that and are homotopic if there exists a homotopy connecting to .
Let G be a groupoid, let H be a quiver and let be morphisms of quivers from H into G. The mapping is a homotopy connecting to . If is a homotopy connecting to , the mapping is a homotopy connecting to . Let and be homotopies connecting to and to respectively. For every vertex of H, the arrows and are composable. The mapping is a homotopy connecting to .
It follows that the relation “ is homotopic to ” is an equivalence relation on the set of morphisms of quivers from H into G.
Let G be a groupoid, H a quiver and morphisms of quivers from H into G which are homotopic. By condition (i) of Definition 1, for every vertex of H, the vertices and belong to one and the same orbit of G.
Suppose moreover that H is a groupoid, and let be a homotopy joining to . The maps Orb( ) and Orb() deduced from and by passing to orbits are therefore equal. For every vertex of H and every arrow , one has Int(), by condition (ii) of Definition 1. In other words, the homomorphism is equal to Int(). In particular, if the homomorphism is injective (resp. bijective, resp. surjective), the same is true of the homomorphism .
Remark 1 01U4
Let G, be groupoids, H, quivers, let be a morphism of quivers and a morphism of groupoids. If morphisms of quivers of H into G are homotopic, the morphisms of quivers and of into are homotopic. More precisely, if is a homotopy joining to , the map Fl() Som() is a homotopy joining to .
Remark 2 01U5
Let G be a groupoid, let H be a quiver, and let be morphisms of quivers of H into G. Let us denote by the canonical morphism of H into Grp(H), and let and be the morphisms of groupoids of Grp(H) into G such that and .
Recall that Som(H) = Som(Grp(H)).
A homotopy Som(H) Fl(G) joining to is a homotopy joining to .
2. Homotopisms of groupoids
In this No, we shall use the notation to express that two morphisms of groupoids and are homotopic.
Definition 2 01U6
Let G, be groupoids and let be a morphism from G into . An inverse up to homotopy of is a groupoid morphism from into G such that the morphisms and are respectively homotopic to Id and to Id. One says that is a homotopism if there exists an inverse of up to homotopy.
An isomorphism of groupoids is a homotopism.
Let G and be groupoids. Let be groupoid morphisms from G into which are homotopic. If is a homotopism, the same is true of . In fact, if denotes an inverse of up to homotopy, one has Id and Id, which proves that is an inverse up to homotopy of .
Let G, be groupoids and let , be groupoid morphisms. Then, of the following conditions:
(i) is an inverse up to homotopy of ;
(ii) is an inverse up to homotopy of ;
(iii) is an inverse up to homotopy of ; any two imply the third. In fact, suppose first that (i) and (ii) are satisfied; one then has
Id Id
and, analogously, Id, whence (iii). If (i) and (iii) are satisfied,
Id
and
Id,
whence condition (ii). The proof that conditions (ii) and (iii) imply condition (i) is analogous.
In particular, if two of the morphisms are homotopisms, the same is true of the third.
Proposition 1 01U7
Let G, be groupoids, let be a morphism from G into , and let A be a subset of the set of vertices of G which meets every orbit of G. In order that be a homotopism, it is necessary and sufficient that the following conditions be satisfied:
(i) the mapping Orb( ) from Orb(G) into Orb(G), deduced from by passing to orbits, is bijective;
(ii) for every , the homomorphism is bijective.
Suppose first that is a homotopism and let be an inverse of up to homotopy. Then,
Orb( ) Orb( ) Orb( ) Orb(Id) Id,
for two homotopic groupoid morphisms induce the same mapping by passing to orbits. Analogously, Orb( ) Orb( ) Id. The mapping Orb( ) is therefore bijective, whence assertion (i). We also have, for every vertex of G,
since is homotopic to Id, the homomorphism is bijective (cf. p. 180), so that is injective and is surjective. On interchanging the roles of and , we see that is also surjective, whence condition (ii).
Suppose now that conditions (i) and (ii) are satisfied, and let us prove that is a homotopism.
Let us first treat the case where each orbit of is reduced to a point.
For each vertex of , choose a vertex of G belonging to A and whose image under is . This is possible because the mapping Orb( ) is surjective and A meets each orbit of G. Let be an arrow of ; we have by hypothesis; put and . By condition (ii), there exists a unique arrow whose image under is . The pair is a groupoid morphism from into G. Let us prove that is inverse to up to homotopy. We already have Id by construction of .
Let be a vertex of G. Put ; this is an element of A such that . Since Orb( ) is injective, belongs to the orbit of in G and there exists an arrow in G joining to . The arrow then joins to and one has .
Let us show that the mapping Som(G) Fl(G) resulting is a homotopy joining to Id. Condition (i) of definition 1 is satisfied, by construction. Let be an arrow of G its origin and its term. One has ; put . The arrows and belong to and both have image in . Since the mapping is injective, one has . Condition (ii) of definition 1 is thus satisfied.
Let us now prove proposition 1 in the general case. Let X be a maximal oriented forest of II, p. 157, prop. 1). Let be the groupoid deduced from by contraction of the arrows of X and let be the canonical morphism. The morphism satisfies the conditions of prop. 1 (II, p. 170, remark 1 and p. 178, cor. 2), and hence the same is true of the morphism .
Since the orbits of are reduced to points, it follows from the special case already proved that and are homotopisms, so that is likewise one. This completes the proof of the proposition.
Corollary 1 01U8
Let G be a groupoid, let A be a set and let Som(G) be a mapping. If the image of meets each orbit of G, the canonical groupoid morphism of into G is a homotopism.
By definition of the inverse image groupoid (II, p. 166, example 4), A is the set of vertices of the groupoid and one has Fl Fl for every pair of elements of A. Moreover, one has Som( ) and Fl( ) induces the identical mapping of Fl into Fl. Consequently, the mapping Orb( ) is bijective and the homomorphism is an isomorphism, for every . The hypotheses of proposition 1 are therefore satisfied.
Corollary 2 01U9
Let G be a groupoid, let X be an oriented forest of G, let be the groupoid deduced from G by contraction of the arrows of X. The canonical morphism of G into is a homotopism.
It follows in fact from remark 1 of II, p. 170 and corollary 2 of II, p. 178 that the hypotheses of proposition 1 are satisfied.
3. Cohomotoper
Let H be a quiver, let G be a groupoid and let and be quiver morphisms of H into G.
Let denote the quiver defined as follows: the vertices of are those of G; the arrows of are the elements of the sum of the sets Fl(G) and Som(H); the origin mapping of coincides with that of G on Fl(G) and with Som( ) on Som(H); the term mapping of coincides with that of G on Fl(G) and with Som( ) on Som(H). Let denote the quiver morphism from G into defined by the identity mapping of Som(G) and by the canonical injection of Fl(G) into Fl(G). Let denote the canonical injection Som(H) Fl(G).
Consider the free groupoid Grp(G) constructed on II, p. 174, def. 9), and let denote the canonical quiver morphism of into Grp(G). Lastly, denote by Coh( ) the groupoid deduced from Grp(G) by contraction of the loops (at the origin of )
for every pair of composable arrows of G, as well as of the loops (at )
for in Som(H) and Fl. Let Grp(G) Coh( ) denote the canonical morphism; let = and Fl().
Proposition 2 01UA
The groupoid Coh( ) is generated by the subquiver whose set of vertices is Som(G) and whose set of arrows is the union of the images of the mappings Fl( ) and .
This subquiver being the image of the quiver under the quiver morphism , the proposition follows at once from the construction of Coh( ).
Proposition 3 01UB
The quiver morphism is a groupoid morphism from G into Coh( ), and the mapping is a homotopy joining to .
The triple (Coh( )) has the following universal property: if is a groupoid, a morphism of groupoids from G into and Som(H) Fl(G) a homotopy joining to , there exists a unique morphism of groupoids Coh( ) such that
(3) and Fl( ).
In view of the definition of the groupoid deduced by contraction of arrows, the contraction of the loops (1) implies that is a morphism of groupoids, and the contraction of the loops (2) that is a homotopy joining to .
Let be as in the statement. Let be the morphism of quivers from into such that Som() is equal to Som() and such that Fl() coincides with Fl() on Fl(G) and with on Som(H). There exists a unique morphism of groupoids Grp(G) such that . Since is a morphism of groupoids and is a homotopy joining to defines by passing to the quotient a morphism of groupoids from Coh( ) into II, p. 170, proposition 3). This morphism satisfies relations (3), and it is the only one (II, p. 185, proposition 2).
Definition 3 01UC
The groupoid Coh( ) is called the cohomotopizer of the pair . One says that is the canonical morphism from G into Coh( ) and that is the canonical homotopy joining to .
The quiver whose set of vertices is the set of orbits of G, whose set of arrows is the set of connected components of H, and whose origin and end mappings are deduced from and by passing to the quotients is called the framework of the pair .
Proposition 4 01UD
The mapping Orb( ) Orb(G) Orb(Coh( )) is surjective; its fibres are the connected components of the framework of the pair .
The morphism is the composite of the morphisms , Grp(G) and Grp(G) Coh( ). The mapping induces a bijection of the set of connected components of the quiver onto the set of orbits of Grp(G), and the mapping Orb( ) Orb(Grp(G)) Orb(Coh( )) is bijective (II, p. 170, remark 1). It is therefore enough to prove that the mapping of Orb(G) into deduced from is surjective and that its fibres are the connected components of the framework of the pair . The surjectivity results from the fact that the mapping Som() is the identity mapping. The equivalence relation in Som(G) defined by “ and are in the same connected component of ” is generated by the relations “there exists an arrow of G joining to ” and “there exists a vertex of H such that and ”. This equivalence relation is compatible with the mapping of Som(G) into Orb(G), and the relation deduced from it in Orb(G) is generated by the relation “there exists an orbit of H such that Orb( ) and Orb( )”. It is therefore the relation “ and are in the same connected component of the framework of the pair ”.
Proposition 5 01UE
Let be a groupoid, let be groupoid morphisms of Coh( ) into , and let be a mapping of Som(G) into Fl(G). In order that be a homotopy joining to , it is necessary and sufficient that the following two conditions be satisfied:
(i) The mapping is a homotopy joining to ;
(ii) For every vertex of H, one has
By definition, in order that be a homotopy joining to , it is necessary and sufficient that the following two conditions be fulfilled (recall that Som(G) = Som(Coh( ))) :
a) For every vertex of Coh( ) joins to ;
b) For every pair of vertices of Coh( ) and every arrow Fl(Coh( )), one has .
By Prop. 2, it is enough to verify condition b) when belongs to the image of Fl( ) or to that of , so that b) is equivalent to the conjunction of the two conditions c) and d) below:
c) For Som(G), Som(G) and Fl, one has ;
d) For every Som(H), one has . Condition (i) is equivalent to the conjunction of a) and c), and condition (ii) is condition d), whence the corollary.
4. Comparison of two cohomotopisers
Consider a diagram
H G
where H, , G, are groupoids and are morphisms of groupoids such that and .
Let denote the canonical morphism of G into the cohomotopiser Coh( ) and the canonical homotopy joining to ; define and analogously. Then is a morphism of groupoids of G into Coh() and Som() is a homotopy joining to , that is to say to . By the universal property of cohomotopisers (II, p. 185, Prop. 3), there exists a unique morphism of groupoids of Coh( ) into Coh() such that and Fl() Som(). In particular, we have extended diagram (4) to a diagram
H G Coh( )
Coh()
in which the second square is commutative.
Theorem 1 01UF
Make the following hypotheses:
(i) the morphism of groupoids is a homotopism;
(ii) the mapping Orb() Orb(H) Orb(H), deduced from by passing to orbits, is bijective;
(iii) there exists in each orbit of H a point such that the homomorphism is surjective.
Then, the groupoid morphism Coh( ) Coh() is a homotopism.
Let be the groupoid deduced from by contraction of the arrows of a maximal oriented forest. The canonical morphism is a homotopism (II, p. 184, corollary 2 of prop. 1). The two diagrams
H G
et Id
give rise to groupoid morphisms Coh( ) Coh() and Coh() Coh() ; we have . It is therefore enough to prove that and are homotopisms. Since the mappings Som() Som(G) Som(G) and Som() Som(G) Som(G) are surjective, it is consequently enough to prove the theorem under the additional assumption that the mapping Som() is surjective, an assumption that we shall make throughout the remainder of the proof.
Let us prove successively the following assertions:
– The mapping Orb() is bijective;
– For every vertex of G, the homomorphism is surjective;
– For every vertex of G, the homomorphism is injective.
a) By assumption, the mapping Orb() is bijective; the same is true of the mapping Orb() by II, p. 182, prop. 1, since is a homotopism. The morphism of quivers from the framework of the pair to that of the pair defined by the mappings Orb() and Orb() is therefore an isomorphism. In particular, the mapping deduced from it by passing to connected components is bijective. Prop. 4 of II, p. 185 then implies that the mapping Orb() is bijective.
(b) Let be an arrow of , and let us denote by its origin and by its term. Since the mapping Som() is surjective, there exist vertices and in G such that and . Since the morphism is a homotopism, there exist an arrow of G joining to and an element such that II, p. 182, prop. 1), whence . This proves that the mapping Fl() is surjective.
Let us then prove that the mapping Fl() is surjective. Its image contains that of Fl(), for one has and the mapping Fl() is surjective. Let be a vertex of ; let be a vertex of H such that and are in the same orbit of , and let be an arrow of joining to . Then,
since is a homotopy joining to . The arrow belong to the image of Fl(), and so do the two arrows and by what precedes. It follows that the arrow belong to the image of Fl(), which proves that the image of Fl() contains that of . By prop. 2 of II, p. 185, the mapping Fl() is surjective.
Let Coh(). Let be an arrow of Coh( ) such that . Let us denote by and the origin and the end of ; one has . Let (resp. ) be an arrow of G joining to (resp. to ) whose image under is . Then, is an element of Coh( ) whose image under is . Hence, for every vertex of G, the homomorphism is surjective.
c) Let us prove that, for every vertex of G, the homomorphism is injective. By considering successively the diagrams
one is thus reduced to treating the two following cases: 1) One has and Id One has and Id.
1) Suppose that one has and Id.
Consider a morphism of groupoids which is an inverse of up to homotopy and a homotopy Som(G) Fl(G) joining to Id. By Remark 1 of II, p. 181, the mappings and are homotopies joining respectively to and to . Hence (cf. II, p. 180), the mapping
Som(H) Fl(Coh( )),
is a homotopy joining to . By the universal property of cohomotopers (II, p. 185, prop. 3), there exists a unique groupoid morphism Coh() Coh( ) such that and Fl().
H G Coh( )
Id
H Coh()
Id
H G Coh( ) .
In particular one has
Since is a homotopy joining to Id is a homotopy joining to . Since Fl() Fl(), one has, for every vertex of H,
Fl(),
by definition of . By prop. 5 of II, p. 186, applied to the groupoid morphisms and Id, the mapping is a homotopy joining to Id. In particular, is a homotopism.
For every vertex of G, the group homomorphism is therefore bijective (II, p. 182, prop. 1). It follows that the homomorphism is injective, whence the result in case A).
2) Suppose that one has and Id.
Let be a vertex of . Since the mapping Orb() is surjective, there exist a vertex of H and an arrow of joining to . The arrows and join respectively to to and to , and therefore are composable in Coh( ). Put
Let us verify that the resulting arrow does not depend on the chosen elements and . Let be a vertex of H and let be an arrow of joining to . Since the mapping Orb() is injective, the vertices and of H belong to the same orbit and there exists an arrow Fl(H) joining to . Then is a loop at in . By hypothesis (iii), there exists a vertex of H such that the homomorphism is surjective and an arrow of H joining to ; then, Int() is a loop at in H, and is therefore the image under of a loop at . Consequently, the arrow of H joins the vertex to the vertex and satisfies . We then have
since is a homotopy joining to . This proves the announced independence.
By construction, we have for all Som(H). Thus we have defined a mapping of Som(H) into Fl(Coh( )). Let be an arrow of , and let denote its origin and its end. Let be a vertex of H and an arrow of joining to . Then is an arrow of joining to . By definition of , we thus have
This proves that is a homotopy joining to .
By the universal property of cohomotopizers, there exists a unique morphism of groupoids Coh() Coh( ) such that and = Fl(). We have and Fl() Fl() Som() by definition of . By the universal property of cohomotopizers, this implies that we have Id. In particular, for every Som(G), the homomorphism is injective.
It then follows from Prop. 1 of II, p. 182 that the morphism is a homotopism, whence the theorem.
5. Isotropy Groups of a Cohomotopizer
Let G and H be groupoids, and let be morphisms of groupoids from H into G. The purpose of this n is to calculate the isotropy groups of the cohomotopizer Coh( ). We again take up the notations of n.
Let denote the framework of the pair ; let us recall (II, p. 185, Def. 3) that this is the quiver (Orb(G), Orb(H)), where and are the mappings of Orb(H) into Orb(G) deduced from the mappings and by passing to orbits.
In all that follows in this No., we shall moreover suppose that the quiver is connected and nonempty; by II, p. 185, Prop. 4, this amounts to supposing that the groupoid Coh( ) is transitive, or again that the quiver is connected and nonempty (II, p. 185, Prop. 2).
Definition 4 01UG
An underlying equipment of the pair is said to be given by the data:
(i) For every Orb(G), of a vertex in the orbit of G;
(ii) For every Orb(H), of a vertex in the orbit of H;
(iii) For every Orb(H), of arrows and of G joining respectively to and to ;
(iv) Of a subquiver T of the framework whose associated graph is a maximal tree of the graph ;
(v) Of an orbit of G.
Let us choose an underlying equipment of the pair . We define a quiver morphism of into Grp(G) by setting for Som(Γ) Orb(G) and
for Fl(Γ) Orb(H). We shall denote by the composite of and the canonical morphism of Grp(G) into Coh( ) ; it is a quiver morphism of into Coh( ).
For Orb(G), let us denote by Coh( ) the group homomorphism deduced from the morphism Coh( ) by restriction to the isotropy groups at .
For Orb(H), let us denote
Int()
and
Int(),
so that one has, for every element of ,
(6) and .
For each vertex of , let us again denote by the unique path class joining to in the tree ; it is regarded as an arrow of Grp(Γ). Let us then denote by the arrow of Coh( ) which is the image of under the canonical morphism of Grp(Γ) into Coh( ) deduced from ; the origin of is , its end is .
The quiver morphism , the group homomorphisms (for Orb(G)) and (for Orb(H)), and the arrows in Coh( ) (for Orb(G)) will be said to be deduced from the basic equipment.
If is a family of groups, we denote by their free product; the image of an element under the canonical mapping of into will be denoted by , or even by if there is no possible confusion. If S is a set, we denote by F(S) the free group constructed on S (A, I, p. 84).
Proposition 6 01UH
There exists a unique group homomorphism
F(Orb(H)) Coh( )
Orb(G)
such that
(7) for Orb(G) and ,
(8) for Orb(H).
The homomorphism Λ is surjective; its kernel is the smallest normal subgroup of F(Orb(H)) containing the elements of Fl(T) and the elements , for Orb(H) and .
The existence and uniqueness of the homomorphism Λ result from the universal property of free products and free groups (A, I, p. 85, prop. 8).
Let A denote the set of the for Orb(G) and the full subgroupoid of G whose set of vertices is A. For every Som(G), let denote the orbit of in G and choose an arrow of G joining to . The pair formed by the mapping of Som(G) into A and the mapping which to Fl associates the element of Fl is a groupoid morphism. It follows from Proposition 1 of II, p. 182 that is a homotopism. Let and , and then let be the canonical morphism of Coh( ) into Coh() ; it is a homotopism (II, p. 187, Theorem 1).
The orbits of are the sets , for , and the injection induces a bijection of Orb(G) onto Orb(G), by means of which we shall identify these two sets. One defines a basic equipment of the pair by setting for Orb(G) for Orb(H) and . The group homomorphisms and (for Orb(H)), the quiver morphism , the arrows (for Orb(G)), and therefore the group homomorphism , deduced from this basic equipment, are the composites with of the corresponding homomorphisms , of the quiver morphism , of the corresponding arrows and of the homomorphism Λ.
Let B be the set of the for Orb(H), let be the full subgroupoid of H with set of vertices B; let us denote by the canonical injection; put and . The morphism induces a bijection B Orb(H) by which we shall identify these two sets. From theorem 1 of II, p. 187 one further deduces a canonical homotopism Coh() Coh(). Moreover, the pair is endowed with a base equipment , so that Orb(G), Orb(H)) are the composites with of .
The various morphisms of groupoids introduced are summarized by the following diagram:
Coh()
(9) H Coh()
H G Coh( ) .
To prove the proposition, we may therefore suppose that A = Som(G) and B = Som(H), in other words that the canonical mappings Som(G) Orb(G) and Som(H) Orb(H) are bijective, hypotheses under which we shall place ourselves in the rest of the proof.
The quiver then has for set of vertices A and for arrows the sum set of the sets , and of the set B. The arrows of are loops at ; if , the arrow joins to , the arrows and are loops respectively at and . The quiver T will be identified with an oriented tree of ; it is a maximal oriented tree since the set of its vertices is equal to the set of vertices of II, p. 157, prop. 1). We put . The set of arrows of the framework of the pair being identified with B, the morphism of quivers Grp(G) associates to the arrow the class of paths in the graph .
Recall that denotes the canonical morphism of quivers of into Grp(G). Let
Grp(G)
be the unique group homomorphism such that one has
if and ;
if .
We thus have , where denotes the canonical group homomorphism from into Grp(G) defined by the maximal oriented tree T (II, p. 179, prop. 9), and where is the unique automorphism of the group such that for and , and for . According to remark 2 of II, p. 179, the homomorphism is surjective and its kernel is the smallest normal subgroup of containing the arrows of T.
Let us denote by Grp(G) Coh( ) the canonical groupoid morphism. According to II, p. 177, cor. 1 of prop. 8, the group homomorphism from Grp(G) into Coh( ) is surjective, and its kernel is the smallest normal subgroup of Grp(G) containing the loops Int(), for and , and the loops Int(), for and .
If denotes the canonical surjective homomorphism, we thus have . This formula implies that the homomorphism Λ is surjective; it remains to determine its kernel.
For and , let us denote by the image of in the group . For , we then have
Int().
Analogously, for and , the definition of the homomorphisms and (formula (6) of II, p. 193) implies that one has
Int()
Consequently, the kernel of the homomorphism is the smallest normal subgroup of which contains the elements for and , the elements for and , and the elements , for Fl(T).
Finally, the kernel of the homomorphism Λ is the smallest normal subgroup of which contains the images under of the preceding elements, in other words, the elements , for Fl(T), and the elements , for and . The proposition is thus proved.