nLab
enriched category

Contents

Idea

The notion of enriched category is a generalization of the notion of category.

Very often instead of merely having a set of morphisms from one object to another, a category will have a vector space of morphisms, or a topological space of morphisms, or some other such thing. This suggests that we should take the definition of (locally small) category and generalize it by replacing the hom-sets by hom-objects , which are objects in a suitable category K. This gives the concept of ‘enriched category’.

The category K must be monoidal, so that we can define composition as a morphism

:hom(y,z)hom(x,y)hom(x,z)\circ : hom(y,z) \otimes hom(x,y) \to hom(x,z)

So, a category enriched over K (also called a category enriched in K, or simply a K-category), say C, has a collection ob(C) of objects and for each pair x,yob(C), a ‘hom-object

hom(x,y)K.hom(x,y) \in K .

We then mimic the usual definition of category.

We may similarly define a functor enriched over K and a natural transformation enriched over K, obtaining a strict 2-category of K-enriched categories. By general 2-category theory, we thereby obtain notions of K-enriched adjunction, K-enriched equivalence, and so on.

There is also an enriched notion of limit called a weighted limit, but it is somewhat more subtle (and in particular, it is difficult to construct purely on the basis of the 2-category K-Cat).

More generally, we may allow K to be a multicategory, a bicategory, a double category, or a virtual double category.

See also enriched category theory.

Definition

Ordinarily enriched categories have been considered as enriched over a monoidal category. This is discussed in the section

More generally, one may think of a monoidal category as a bicategory with a single object and this way regard enrichment in a monoidal category as the special case of enrichment in a bicategory . This is discussed in the section

Enriched categories and enriched functors between them form themselves a category, the category of V-enriched categories.

Enrichment in a monoidal category

Let V be a monoidal category with

  • tensor product :V×VV;

  • tensor unit IObj(V);

  • associator α a,b,c:(ab)ca(bc);

  • left unitor l a:Iaa;

  • right unitor r a:aIa.

A (small) V-category C (or V-enriched category or category enriched over/in V) is

  • a set Obj(C) – called the set of objects;

  • for each ordered pair (a,b)Obj(C)×Obj(C) of objects in C an object C(a,b)Obj(V) – called the hom-object or object of morphisms from a to b;

  • for each ordered triple (a,b,c) of objects of C a morphism a,b,c:C(b,c)C(a,b)C(a,c) in V – called the composition morphism;

  • for each object aObj(C) a morphism j a:IC(a,a) – called the identity element

  • such the following diagrams commute:

for all a,b,c,dObj(C):

(C(c,d)C(b,c))C(a,b) α C(c,d)(C(b,c)C(a,b)) b,c,dId C(a,b) Id C(c,d) a,b,c C(b,d)C(a,b) a,b,c C(a,d) a,c,d C(c,d)C(a,c)\array{ (C(c,d)\otimes C(b,c)) \otimes C(a,b) &&\stackrel{\alpha}{\to}&& C(c,d) \otimes (C(b,c) \otimes C(a,b)) \\ \downarrow^{\circ_{b,c,d}\otimes Id_{C(a,b)}} &&&& \downarrow^{Id_{C(c,d)\otimes \circ_{a,b,c}}} \\ C(b,d)\otimes C(a,b) &\stackrel{\circ_{a,b,c}}{\to}& C(a,d) &\stackrel{\circ_{a,c,d}}{\leftarrow}& C(c,d) \otimes C(a,c) }

this says that composition in C is associative;

and

C(b,b)C(a,b) a,b,b C(a,b) a,a,b C(a,b)C(a,a) j bId C(a,b) l r Id C(a,b)j a IC(a,b) C(a,b)I\array{ C(b,b)\otimes C(a,b) &\stackrel{\circ_{a,b,b}}{\to}& C(a,b) &\stackrel{\circ_{a,a,b}}{\leftarrow}& C(a,b) \otimes C(a,a) \\ \uparrow^{j_b \otimes Id_{C(a,b)}} & \nearrow_{l}&& {}_r\nwarrow& \uparrow^{Id_{C(a,b)}\otimes j_a} \\ I \otimes C(a,b) &&&& C(a,b) \otimes I }

this says that composition is unital.

Enrichment through lax monoidal functors

If V is a monoidal category, then an alternative way of viewing a V-category is as a set X together with a (lax) monoidal functor Φ=Φ d of the form

V opyon VSet VdSet X×XV^{op} \stackrel{yon_V}{\to} Set^{V} \stackrel{d}{\to} Set^{X \times X}

where the codomain is identified with the monoidal category of spans on X, i.e., the local hom-category hom(X,X) in the bicategory of spans of sets. Given an V-category (X,d:X×XV) under the ordinary definition, the corresponding monoidal functor Φ takes an object v of V op to the span

Φ(v) x,y:=hom V(v,d(x,y))\Phi(v)_{x, y} := \hom_V(v, d(x, y))

Under the composition law, we get a natural map

hom(v,d(x,y))×hom(v,d(y,z))hom(vv,d(x,y)d(y,z))hom(1,comp)hom(vv,d(x,z))\hom(v, d(x, y)) \times \hom(v', d(y, z)) \to \hom(v \otimes v', d(x, y) \otimes d(y, z)) \stackrel{\hom(1, comp)}{\to} \hom(v \otimes v', d(x, z))

which gives the tensorial constraint Φ(v)Φ(v)Φ(vv) for a monoidal functor; the identity law similarly gives the unit constraint.

Conversely, by using a Yoneda-style argument, such a monoidal functor structure on Φ=Φ d induces an M-enrichment on X, and the two notions are equivalent.

Alternatively, we can equivalently describe a V-enriched category as precisely a bicontinuous lax monoidal functor of the form

Set VSet X×XSet^V \to Set^{X \times X}

since bicontinuous functors of the form Set VSet X×X are precisely those of the form Set d for some function d:X×XV, at least if V is Cauchy complete.

Enrichment in a bicategory

Let B be a bicategory, and write for horizontal (1-cell) composition (written in Leibniz order). A B-enriched category consists of a set X together with

  • A function p:XB 0,
  • A function hom:X×XB 1, satisfying the typing constraint hom(x,y):p(x)p(y),
  • A function :X×X×XB 2, satisfying the constraint x,y,z:hom(y,z)hom(x,y)hom(x,z),
  • A function j:XB 2, satisfying the constraint j x:1 p(x)hom(x,x),

such that the associativity and unitality diagrams, as written above, commute. Viewing a monoidal category M as a 1-object bicategory ΣM, the notion of enrichment in M coincides with the notion of enrichment in the bicategory ΣM.

If X, Y are sets which come equipped with enrichments in B, then a B-functor consists of a function f:XY such that p Yf=p X, together with a function f 1:X×XB 2, satisfying the constraint f 1(x,y):hom X(x,y)hom Y(f(x),f(y)), and satisfying equations expressing coherence with the composition and unit data , j of X and Y. (Diagram to be inserted, perhaps.)

Enrichment in a double category

It is also natural to generalize further to categories enriched in a (possibly weak) double category. Just like for a bicategory, if D is a double category, then a D-enriched category X consists of a set X together with

  • for each xX, an object p(x) of D,

  • for each x,yX, a horizontal arrow hom(x,y):p(x)p(y) in D,

  • for each x,y,zX, a 2-cell in D:

    p(x) hom(x,y) p(y) hom(y,z) p(z) x,y,z p(x) hom(x,z) p(z)\array{p(x) & \overset{hom(x,y)}{\to} & p(y) & \overset{hom(y,z)}{\to} & p(z) \\ \Vert && \circ_{x,y,z} && \Vert\\ p(x) && \underset{hom(x,z)}{\to} && p(z)}
  • for each xX, a 2-cell in D:

    p(x) id p(x) p(x) hom(x,x) p(x)\array{p(x) & \overset{id}{\to} & p(x)\\ \Vert && \Vert\\ p(x) & \underset{hom(x,x)}{\to} & p(x)}

satisfying analogues of the associativity and unit conditions. Note that is is exactly the same as a category enriched in the horizontal bicategory of D; the vertical arrows of D play no role in the definition. However, they do play a role when it comes to define functors between D-enriched categories. Namely, if X and Y are D-enriched categories, then a D-functor f:XY consists of:

  • a function f:XY,

  • for each xX a vertical arrow f x:p X(x)p Y(f(x)) in D,

  • for each x,yX a 2-cell in D:

    p(x) hom X(x,y) p(y) f x f y p(f(x)) hom Y(f(x),f(y)) p(f(y))\array{p(x) & \overset{hom_X(x,y)}{\to} & p(y)\\ ^{f_x}\downarrow && \downarrow^{f_y}\\ p(f(x))& \underset{hom_Y(f(x),f(y))}{\to} & p(f(y))}

satisfying suitable equations. If D is vertically discrete, i.e. just a bicategory B with no nonidentity vertical arrows, then this is just the same as a B-functor as defined above. However, for many D this notion of functor is more general and natural.

Change of enriching category

Passage between ordinary categories and enriched categories

Every K-enriched category C has an underlying ordinary category, usually denoted C 0, defined by C 0(x,y)=K(I,hom(x,y)) where I is the unit object of K.

If K(I,):KSet has a left adjoint I:SetK (taking a set S to the tensor or copower SI, viz. the coproduct of an S-indexed set of copies of I), then any ordinary category C can be regarded as enriched in K by forming the composite

Ob(C)×Ob(C)homSetIKOb(C) \times Ob(C) \stackrel{\hom}{\to} Set \stackrel{-\cdot I}{\to} K

These two operations form adjoint functors relating the 2-category Cat to the 2-category K-Cat.

Lax monoidal functors

More generally, any (lax) monoidal functor F:KL between monoidal categories can be regarded as a “change of base”. By applying F to its hom-objects, any category enriched over K gives rise to one enriched over L, and this forms a 2-functor from K-Cat to L-Cat, and in fact from K-Prof to L-Prof; see profunctor and 2-category equipped with proarrows.

Moreover, this operation is itself functorial from MonCat to 2Cat. In particular, any monoidal adjunction KL gives rise to a 2-adjunction KCatLCat (and also for profunctors). The adjunction CatKCat described above is a special case of this arising from the adjunction I:SetK:K(I,).

This and further properties of such “change of base” are explored in Geoff Cruttwell’s thesis.

Internalization versus Enrichment

The idea of enriched categories is not unrelated to that of internal categories, but is different. One difference is that in a K-enriched category, the objects still form a set (or a proper class) while the arrows are replaced by objects of K, while in a category internal to K, both the set of objects and the set of arrows are replaced by objects of K.

Another difference is that for K-enriched categories, K can be any monoidal category, while for K-internal categories, it must have pullbacks, which can be thought of as a generalization of cartesian monoidal structure. In particular, a K-internal category with one object (that is, whose object-of-objects is a terminal object) is a monoid in K with respect to the cartesian product, whereas a one-object K-enriched category is a monoid in K with respect to whatever monoidal structure we use to define enriched categories.

Nevertheless, internalization and enrichment are related in several ways. On the one hand, internal categories and enriched categories are both instances of monads in bicategories (the bicategory of spans and the bicategory of matrices, respectively). On the other hand, when K is an -extensive category, such as Set or simplicial sets (or more generally any Grothendieck topos), (small) K-enriched categories can be identified with K-internal categories whose object-of-objects is discrete (that is, a coproduct of copies of the terminal object).

The page here mentions internalization and enrichment being the result of applying two different interpretation techniques to the same theory.

Examples

  • A category enriched in Set is a locally small category.

  • A category enriched in chain complexes is a dg-category.

  • A category enriched in simplicial sets is called a simplicial category, and these form one model for (,1)-categories. But beware: the term ‘simplicial category’ is also used to mean a category internal to simplicial sets. In fact, a category enriched in simplicial sets is a special case of a category internal to simplicial sets, namely one where the simplicial set of object is discrete.

  • A category enriched in Top is a topological category. These are also a model for (,1)-categories. But again, beware: the term ‘topological category’ is perhaps more commonly used to mean a category internal to Top. And again: a category enriched in Top is a special case of one internal to Top, namely one where the space of objects is discrete.

  • A category enriched in Cat is a strict 2-category.

  • A strict n-category is a category enriched over strict (n1)-categories. In the limit n this leads to strict omega-categories.

  • An algebroid, or linear category, is a category enriched over Vect. Here Vect is the category of vector spaces over some fixed field K, equipped with its usual tensor product. It is common to emphasize the dependence on K and call a category enriched over Vect a K-linear category.

  • More generally, if K is any commutative ring, a category enriched over KMod is sometimes called a K-linear category.

  • In particular, taking K to be (the ring of integers), a ringoid (or Ab-enriched category) is a category enriched over Ab.

  • A (Lawvere) metric space is a category enriched over the poset ([0,],) of extended positive real numbers, where is +.

  • An ultrametric space is a category enriched over the poset ([0,],) of extended positive real numbers, where is max.

  • A poset is a category enriched over the category of truth values, where is conjunction.

  • An apartness space is a groupoid enriched over the opposite of the category of truth values, where is disjunction.

  • A group torsor? (over a group G) can be modeled by a category enriched over the discrete category on the set G, where is the group operation. Not every such category determines a torsor, however; it must be nonempty as well as Cauchy complete.

References

The standard reference on enriched categories is

  • Max Kelly, Basic concepts of enriched category theory, London Math. Soc. Lec. Note Series 64, Cambridge Univ. Press 1982, 245 pp.; remake: TAC reprints 10, tac,pdf

Vista of some modern generalizations is in

Change of base is discussed in

Blog entries

Revised on October 30, 2012 13:27:27 by Urs Schreiber (131.174.40.126)