Natural Transformations

Let C and D be categories, and let

F,G:CD

be functors.

A natural transformation η:FG assigns to each object XC a morphism in D,

ηX:F(X)G(X),

called the component at X.

These components satisfy the naturality condition: for every morphism f:XY in C, the following square commutes:

F(X)F(f)F(Y)ηXηYG(X)G(f)G(Y)

Equivalently,

G(f)ηX=ηYF(f).

Consequence: the 2-category Cat

With natural transformations included, categories, functors, and natural transformations do not merely form a category—they form a 2-category:

This expresses the idea that one can compare not only objects and morphisms, but also functors between them.

Example

Let Man be the category whose objects are smooth manifolds and whose morphisms are smooth maps. Define two functors F,G:ManMan as follows:

  1. F(M)=TM, the tangent bundle of M.
    For a smooth map f:MN, define

    F(f)=Tf:TMTN,

    the differential (pushforward) of f.

  2. G(M)=M×M.
    For a smooth map f:MN, define

    G(f)=f×f:M×MN×N.

There is a natural transformation η:FG, where for each manifold M,

ηM:TMM×M,

is the map

ηM(vx)=(x,x).

This simply sends each tangent vector vxTxM to the pair of its base point with itself.

Naturality check.
Given any smooth map f:MN, the naturality square

TMTfTNηMηNM×Mf×fN×N

commutes, since for any vxTxM,

(f×f)(ηM(vx))=(f(x),f(x))=ηN(Tf(vx)).

Thus,

(f×f)ηM=ηNTf,

so η is indeed a natural transformation between the tangent-bundle functor and the diagonal-pair functor.