Chapter 0_III (suite)

Preliminaries

§8. Representable functors

8.1. Representable functors

8.1.1.

We denote by Set the category of sets. Let be a category; for two objects , of , we set ; for every morphism in , we denote by the map from into . It is immediate that with these definitions is a contravariant functor, that is, an object of the category, denoted , of covariant functors from the category dual to into the category Set (T, 1.7, d) and [29].

Translator's note. We render EGA's Ens as Set throughout, the standard modern English form. The 1961 original capitalizes the category as Ens; the typographical species ("category in bold") is rendered here by the upright ASCII string Set inside backticks.

8.1.2.

Now let be a morphism in ; for every and every , we have ; we denote by the map from into . It is immediate that for every morphism in , the diagram

              h_X(u)
   h_X(Y') ────────→ h_X(Y)

   h_w(Y')│           │h_w(Y)
          ↓           ↓

   h_{X'}(Y') ───────→ h_{X'}(Y)
                h_{X'}(u)

is commutative; in other words, is a functorial morphism (T, 1.2), or again a morphism in the category (T, 1.7, d). The definitions of and therefore constitute the definition of a canonical covariant functor

  h : 𝒞 → Hom(𝒞°, Set).                                                    (8.1.2.1)

8.1.3.

Let be an object of , a contravariant functor from into Set (an object of ). Let be a functorial morphism: for every ,

is a map such that for every morphism in , the diagram

              h_X(u)
   h_X(Y') ────────→ h_X(Y)

   g(Y')  │           │ g(Y)                                              (8.1.3.1)
          ↓           ↓

   F(Y') ────────────→ F(Y)
                F(u)

is commutative. In particular, we have a map , whence an element

and consequently a canonical map

  α : Hom(h_X, F) → F(X).                                                  (8.1.3.3)

Conversely, consider an element ; for every morphism in , is a map ; consider the map

from into ; if we denote this map by ,

is a functorial morphism, for we have, for every morphism in , , which verifies the commutativity of (8.1.3.1) for . We have thus defined a canonical map

  β : F(X) → Hom(h_X, F).                                                  (8.1.3.6)

Proposition (8.1.4).

The maps and are mutually inverse bijections.

Proof. Let us compute for ; for every , is the map from into . We therefore have

  α(β(ξ)) = (g_1(X))(1_X) = (F(1_X))(ξ) = 1_{F(X)}(ξ) = ξ.

Let us now compute for ; for every , is the map ; by the commutativity of (8.1.3.1), this map is none other than by the definition of ; in other words, it is equal to , which proves the proposition.

8.1.5.

Recall that a subcategory of a category is defined by the condition that its objects be objects of , and that if , are two objects of , the set of morphisms in is a subset of the set of morphisms in , the canonical map of "composition of morphisms"

  Hom_{𝒞'}(X', Y') × Hom_{𝒞'}(Y', Z') → Hom_{𝒞'}(X', Z')

being the restriction of the canonical map

  Hom_𝒞(X', Y') × Hom_𝒞(Y', Z') → Hom_𝒞(X', Z').

We say that is a full subcategory of if for every pair of objects of . The subcategory of formed by the objects of isomorphic to objects of is then also a full subcategory of , equivalent (T, 1.2) to , as one verifies without difficulty.

A covariant functor is said to be fully faithful if, for every pair of objects X_1, Y_1 of , the map from into is bijective; this entails that the subcategory of is full. Moreover, if two objects X_1, have the same image X_2, there exists a unique isomorphism such that . For each object X_2 of , let be one of the objects X_1 of such that ( being defined by means of the axiom of choice); for every morphism in , will be the unique morphism such that ; is then a functor from into ; is the identity functor on , and what precedes shows that there exists an isomorphism of functors such that , , and the identity define an equivalence of the category with the full subcategory of (T, 1.2).

8.1.6.

Apply Proposition (8.1.4) to the case where the functor is , being an arbitrary object of ; the map is here none other than the map defined in (8.1.2); this map being bijective, we see, with the terminology of (8.1.5), that:

Proposition (8.1.7).

The canonical functor is fully faithful.

8.1.8.

Let be a contravariant functor from into Set; we say that is representable if there exists an object such that is isomorphic to ; it follows from (8.1.7) that the data of an and an isomorphism of functors determines up to unique isomorphism. Proposition (8.1.7) also means that defines an equivalence of with the full subcategory of formed by the representable contravariant functors. It moreover follows from (8.1.4) that the data of a functorial morphism is equivalent to that of an element ; to say that is an isomorphism is equivalent, for , to the following condition: for every object of the map from into is bijective. When satisfies this condition, we shall say that the pair represents the representable functor . By abuse of language, we shall also say that the object represents if there exists such that represents , in other words if is isomorphic to .

Let , be two representable contravariant functors from into Set, and two isomorphisms of functors. Then it follows from (8.1.6) that there is a canonical one-to-one correspondence between and the set of functorial morphisms .

8.1.9. Examples. I: projective limits.

The notion of representable contravariant functor covers in particular the "dual" notion of the usual notion of "solution of a

universal problem". More generally, we shall see that the notion of projective limit is a particular case of that of representable functor. Recall that in a category , one defines a projective system by the data of a preordered set , a family of objects of , and, for every pair of indices such that , a morphism . A projective limit of this system in consists of an object of (denoted ), and, for each , a morphism , such that: 1° for ; 2° For every object of and every family of morphisms such that for , there exists a unique morphism (denoted ) such that for every (T, 1.8). This is interpreted as follows: the canonically define maps

  ū_{αβ} : Hom(X, A_β) → Hom(X, A_α)

which define a projective system of sets , and is by definition an element of the set ; it is clear that is a contravariant functor from into Set, and the existence of the projective limit is equivalent to saying that is an isomorphism of functors in

  lim Hom(X, A_α) ⥲ Hom(X, B)                                              (8.1.9.1)

in other words that the functor is representable.

8.1.10. Examples. II: final object.

Let be a category, a set reduced to a single element. Consider the contravariant functor which assigns to every object of the set , and to every morphism in the unique map . To say that this functor is representable means that there exists an object such that for every , is reduced to one element; we say that is a final object of , and it is clear that two final objects of are isomorphic (which allows us to define, in general by means of the axiom of choice, a final object of , then denoted ). For example, in the category Set, the final objects are the sets reduced to one element; in the category of augmented algebras over a field (where the morphisms are the algebra homomorphisms compatible with the augmentations), is a final object; in the category of -preschemes (I, 2.5.1), is a final object.

8.1.11.

For two objects , of a category , set , and for every morphism , let be the map from into ; is then a covariant functor , from which one deduces, as in (8.1.2), the definition of a canonical covariant functor ; a covariant functor from into Set, that is, an object of , is then said to be representable if there exists an object (necessarily unique up to unique isomorphism) such that is isomorphic to ; we leave to the reader the task of developing the "dual" considerations of the preceding ones for this notion, which this time covers that of inductive limit, and in particular the usual notion of "solution of a universal problem".

8.2. Algebraic structures in categories

8.2.1.

Given two contravariant functors , from into Set, recall that for every object , we set , and for every morphism in , we set , which is the map from into ; is therefore a contravariant functor (which is moreover none other than the product of the objects , in the category ). Given an object , we shall call an internal composition law on a functorial morphism

  γ_X : h_X × h_X → h_X.                                                   (8.2.1.1)

In other words (T, 1.2), for every object , is a map (so by definition an internal composition law on the set ) subject to the condition that, for every morphism in , the diagram

                    h_X(u) × h_X(u)
   h_X(Y') × h_X(Y') ──────────────→ h_X(Y) × h_X(Y)

   γ_X(Y')         │                          │ γ_X(Y)
                   ↓                          ↓

       h_X(Y') ───────────────────────→ h_X(Y)
                          h_X(u)

is commutative; this means that for the composition laws and , is a homomorphism from into .

In the same way, given two objects , of , one calls an external composition law on , having as domain of operators, a functorial morphism

  ω_{X,Z} : h_Z × h_X → h_X.                                               (8.2.1.2)

One sees as above that for every , is an external composition law on , having as domain of operators, and such that for every morphism , and form a di-homomorphism from into .

8.2.2.

Let be a second object of , and suppose given on an internal composition law ; we shall say that a morphism in is a homomorphism for these composition laws, if for every , is a homomorphism for the composition laws and . If X'' is a third

object of equipped with an internal composition law and a morphism in which is a homomorphism for and , it is clear that the morphism is a homomorphism for the composition laws and . An isomorphism in is called an isomorphism for the composition laws and if is a homomorphism for these composition laws, and if its inverse morphism is a homomorphism for the composition laws and .

One defines in the same way the di-homomorphisms for the pairs of objects of equipped with external composition laws.

8.2.3.

When an internal composition law on an object is such that is a group law on for every , we say that , equipped with this law, is a -group or a -object in groups. One defines in the same way -rings, -modules, etc.

8.2.4.

Suppose that the product of an object by itself exists in ; by definition, we then have up to canonical isomorphism, since this is a particular case of projective limit (8.1.9); an internal composition law on may therefore be considered as a functorial morphism , and so canonically determines (8.1.6) an element such that ; in this case, the data of an internal composition law on is therefore equivalent to that of a morphism ; when is the category Set, one recovers the classical notion of internal composition law on a set. One has an analogous result for an external composition law when the product exists in .

8.2.5.

With the preceding notations, suppose in addition that exists in ; the characterization of the product as an object representing a functor (8.1.9) entails the existence of canonical isomorphisms

  (X × X) × X ⥲ X × X × X ⥲ X × (X × X);

if one canonically identifies with , the map identifies with for every . It is therefore equivalent to say that for every , the internal law is associative, or that the diagram of maps

                         γ_X(Y) × 1
   h_X(Y) × h_X(Y) × h_X(Y) ────────→ h_X(Y) × h_X(Y)

   1 × γ_X(Y) │                              │ γ_X(Y)
              ↓                              ↓

   h_X(Y) × h_X(Y) ──────────────────────→ h_X(Y)
                          γ_X(Y)

is commutative, or that the diagram of morphisms

                         c_X × 1_X
       X × X × X ────────────────→ X × X

   1_X × c_X │                       │ c_X
             ↓                       ↓

         X × X ──────────────────→ X
                          c_X

is commutative.

8.2.6.

Under the hypotheses of (8.2.5), if one wants to express that for every , the internal law is a group law, it is necessary, on the one hand, to express that it is associative, and on the other that there exists a map having the properties of the inverse in a group; since for every morphism in , we have seen that must be a group homomorphism , one sees first that must be a functorial morphism. One can on the other hand express the characteristic properties of the inverse in a group without making the neutral element intervene: it suffices to write that the two composed maps

  (s, t) ↦ (s, s^{−1}, t) ↦ (s, s^{−1} t) ↦ s(s^{−1} t)
  (s, t) ↦ (s, s^{−1}, t) ↦ (s, t s^{−1}) ↦ (t s^{−1}) s

are equal to the second projection from into . By virtue of (8.1.3), we have , where ; the first preceding condition then expresses that the composed morphism

                  (1_X, a_X) × 1_X         1_X × c_X            c_X
   X × X ──────────────────────→ X × X × X ────────→ X × X ────────→ X

is the second projection in , and the second condition is translated similarly.

8.2.7.

Suppose now that there exists in a final object (8.1.10). Suppose still that is a group law on for every , and denote by the neutral element of . Since, for every morphism in , is a group homomorphism, we have ; taking in particular , in which case is the unique element of , we see that the element completely determines for every . Set , the neutral element of the group ; the commutativity of the diagram

              h_ε(X)
   h_X(e) ───────────→ h_X(Y)

   h_{e_X}(e) │           │ h_{e_X}(Y)
              ↓           ↓

   h_X(e) ───────────→ h_X(Y)
              h_ε(X)

(cf. 8.1.2) shows that, in the set , the map is none other than

, transforming every element into the neutral element. One then verifies that the fact that is the neutral element of for every is equivalent to saying that the composed morphism

              (1_X, 1_X)         1_X × e_X            c_X
   X ──────────────────→ X × X ──────────→ X × X ──────→ X,

and the analogous one where one permutes 1_X and , are both equal to 1_X.

8.2.8.

One could of course multiply without difficulty the examples of algebraic structures in categories. The example of groups has been treated with sufficient detail, but in what follows we shall generally leave to the reader the task of developing analogous considerations in the examples of algebraic structures that we shall encounter.


Translator's footnote (from p. 5). To facilitate reference lookup, we shall henceforth refer to the paragraphs of Chapter 0 published with Chapter I by the sign 0_I.