nLab
image

Contents

Idea

The familiar notion of the image of a map of sets may be formalized to yield a notion of image for morphisms in an arbitrary category.

Definitions

There are several definitions that are equivalent when they jointly apply.

In terms of subobjects

Let C be a category, and f:cd be a morphism. The image of f is the smallest subobject sd through which f factors (if it exists). There is a dual notion of co-image: the largest quotient (in the co-subobject sense) of c through which f factors.

Remarks

  • If C admits equalizers, and if i:kd represents the image of f:cd, then the unique map q:ck such that f=iq is an epimorphism. Thus, in a finitely complete category in which every morphism admits an image, one obtains in this way an epi-mono factorization, but the factorization may not have particularly good properties (in particular, the factorization through the image might not be stable with respect to pullback).

  • The notion of regular category formalizes a sense in which image factorizations do behave well: factorizations into a regular epimorphism followed by a mono which are stable under pullback.

  • In Cat, the image of a functor F:AB is the smallest subcategory of B which contains images through F of all morphisms in A. Some of the morphisms in the image may not be images of any morphism in A; all morphisms in the image of F are compositions in B of B-composable sequences of images of morphisms in A which themselves do not necessarily form A-composable sequences of morphisms in A. Sometimes the notion of essential image is more appropriate; as the essential image is only equivalent to the image, this is somewhat 2-category-theoretic point of view.

as a left adjoint functor

Alternatively, let C/d be the slice category over d, and let Mono(C)/d be the full subcategory whose objects are monos into d. Assuming images exist in C, taking the image of a map f:cd provides a left adjoint

C/dMono(C)/dC/d \to Mono(C)/d

to the inclusion Mono(C)/dC/d.

as an equalizer

If the category C admits finite limits and colimits, then the image Imf of a morphism f:cd my be expressed as

Imflim(dd cd),Im f \simeq lim (d \stackrel{\to}{\to} d \sqcup_c d) \,,

where d cd denotes the pushout

c f d f d d cd.\array{ c &\stackrel{f}{\to}& d \\ \downarrow^{f} && \downarrow \\ d &\to& d \sqcup_c d } \,.

In other words, the image is the equalizer of the cokernel pair?.

This is isomorphic to the pullback d× d cdd

imfd× d cdd.im f \simeq d \times_{d \sqcup_c d} d \,.

So in the diagram

c f d f imf d d cd\array{ c &&\stackrel{f}{\to}&& d \\ & \searrow && \nearrow \\ \downarrow^f && im f&& \downarrow \\ & \swarrow \\ d &&\to&& d \sqcup_c d }

the outer square is a pushout and the inner one is a pullback.

Since imf is an equalizer, the morphism

imfdim f \to d

is a monomorphism.

Lemma

There is a unique morphism

u:coimfimfu : coim f \to im f

from the coimage to the image of f such that

f=(ccoimfuimfd).f = (c \to coim f \stackrel{u}{\to} im f \to d) \,.
Proof

Because f coequalizes c× dcc, a morphism h in

c × dc c f d d cd epi h mono coimf imf\array{ c &\times_d c \stackrel{\to}{\to}& c &\stackrel{f}{\to}& d &\stackrel{\to}{\to}& d \sqcup_{c} d \\ && {}^{\;}\downarrow^{epi} &{}^{h}\nearrow& {}^{\;}\uparrow^{mono} \\ && coim f && im f }

exists uniquely.

Because ccoimf is epi it follows that h equalizes dd cd and hence u in the diagram

c × dc c f d d cd epi h mono coimf u imf\array{ c &\times_d c \stackrel{\to}{\to}& c &\stackrel{f}{\to}& d &\stackrel{\to}{\to}& d \sqcup_{c} d \\ && {}^{\;}\downarrow^{epi} &{}^{h}\nearrow& {}^{\;}\uparrow^{mono} \\ && coim f &\stackrel{u}{\to}& im f }

exists uniquely.

Remarks