Product category

From HandWiki
Short description: Product of two categories, in category theory

In the mathematical field of category theory, the product of two categories C and D, denoted C × D and called a product category, is an extension of the concept of the Cartesian product of two sets. Product categories are used to define bifunctors and multifunctors.[1]

Definition

The product category C × D has:

  • as objects:
    pairs of objects (A, B), where A is an object of C and B of D;
  • as arrows from (A1, B1) to (A2, B2):
    pairs of arrows (f, g), where f : A1A2 is an arrow of C and g : B1B2 is an arrow of D;
  • as composition, component-wise composition from the contributing categories:
    (f2, g2) o (f1, g1) = (f2 o f1, g2 o g1);
  • as identities, pairs of identities from the contributing categories:
    1(A, B) = (1A, 1B).

A product of a family of categories is defined exactly the same way.

Universal property

Just like for sets, a product of a family of categories is characterized by the following universal property. Given categories Ci indexed by a set I, P=Ci,pj:PCj,jI satisfy:

given a family of functors fi:DCi, there exists a unique functor f:DP such that fj=pjf for each jI.

Put in another way, a product of a family of small categories is exactly the categorical product of them in the category of small categories Cat. Thus, for example,

Fct(A,iBi)iFct(A,Bi).[2]

Functoriality

Given two functors f:CD,g:CD, the product f×g:C×CD×D is defined component-wise; that is,

(f×g)(x,x)=(f(x),f(x))

for a pair of objects x,x and a pair of morphisms x,x.[3] (This product may also be characterized by the universal property similar to that for categories.) This way, we get the functor

×:Cat×CatCat.

It satisfies the tensor-hom adjunction in the sense

HomCat(A×B,C)HomCat(A,Fct(B,C))

where Fct denotes a functor category.[4]

Example: C × 2

Let f,g:CD be functors. Suppose there is a natural transformation φ:fg. Then φ determines the functor

h:C×2_D

such that

h(,0)=f,h(,1)=g,

where 2_={0,1} is the category with two objects and the non-identity morphism :01.[3] Intuitively, h is a non-invertible homotopy from f to g. Indeed, define h by, for x:ab in C,

h(x,id0)=f(x),h(x,id1)=g(x),h(x,)=g(x)φa=φbf(x).

Conversely, given h:C×2_D, we get f,g,φ by f=h(,0),g=h(,1) and φa=h(ida,).[5]

Bifunctor

A functor whose domain is a product category is called a bifunctor. A bifunctor can be defined in each variable separately in the following sense:

Proposition — [6] Each bifunctor

F:A×BC

determines the families of the functors, for objects a in A and b in B,

Fb:AC,Fa:BC

given by

Fba=F(a,b) and Fbf=F(f,idb)

for f:aa and similarly for Fa. They commute in the sense:

FagFbf=FbfFag.

Conversely, given families of functors Fb,Fa as above, if they commute, they define the bifunctor F:A×BC by

F(f,g)=FbfFag.

For example, consider (a,b)Hom(a,b):Cop×CSet. For each fixed b in B, we have the functor

Hom(,b):CopSet

by pullback; i.e., f:aa goes to the function

f*:Hom(a,b)Hom(a,b)

defined by f*g=gf. On the other hand, Hom(a,):CSet is defined by pushforward; i.e., ff*=f. Clearly, these two functors commute (the associativity of composition) and so, by the proposition, we get the functor called the Hom functor

Hom(,):Cop×CSet,

which is explicitly given as: (f,g)(hghf).

There is a similar result for natural transformations between bifunctors:

Proposition — [7] Let F,G:A,BC be bifunctors and

α={αa,b:F(a,b)G(a,b)aOb(A),bOb(B)}

a family of morphisms. Then α:FG is a natural transformation if and only if it is natural in the first variable and the second variable separately; i,e., for each object b in B,

α,b:F(,b)G(,b)

is a natural transformation and similarly in the second variable.

References

  1. Mac Lane 1978, p. 37.
  2. Mac Lane 1978, Ch. II., § 5., Exercise 2.
  3. 3.0 3.1 Mac Lane 1978, Ch. II., § 3.
  4. Mac Lane 1978, Ch. II., § 5., Exercise 1.
  5. Mac Lane 1978, Ch. II., § 4., Exercise 8.
  6. Mac Lane 1978, Ch. II., § 3., Proposition 1.
  7. Mac Lane 1978, Ch. II., § 3., Proposition 2.