In group theory, we know that a group homomorphism $f:G \to H$ induces an isomorphism of groups ${G \over \ker(f)} \simeq \operatorname{im}(f)$. I've searched a lot for a categorical version of this theorem in the following sense but I cannot find anything.

Let $F:\mathcal{A \to B}$ be a functor between abelian categories. Does this functor induce an isomorphism of categories ${\mathcal{A} \over \mathcal{Ker}(F)} \simeq \mathcal{B}$? (The quotient here is serre quotient as I mentioned in commnet to answer by Max)

Here, by $\mathcal{Ker}(F)$ I mean the full subcategory of $\mathcal{A}$ consisting of objects $A$ so that $F(A)=0$.

  • 3
    $\begingroup$ It's rather $G/\ker (f) \simeq \mathrm{im} f$. $\endgroup$ – Max Aug 12 '18 at 8:16

I don't know what you mean by quotienting by a subcategory, but here's an idea:

if $F: C\to D$ is a functor bewteen arbitrary categories, then you have a relation on morphisms (which we might denote $\ker(F)$) that is, for $f,g: A\to B$, $(f,g)\in\ker(F)\iff F(f)=F(g)$ (note that we do ask that $f,g$ have the same domain and codomain).

This relation is very clearly a congruence on $C$, that is : if $f\sim f', g\sim g'$, then $f\circ g \sim f'\circ g'$. This allows us to form the quotient category $C/\ker (F)$ whose objects are those of $C$ and morphisms are equivalence classes of morphisms.

Together with this we have a canonical functor $\Pi: C\to C/\ker(F)$ that satisfies the universal property you probably expect (note that you can do the same thing with any congruence, which shows that congruences and "kernels" are the same; just as for groups)

Now what's interesting here is that it shows that $F$ can be factored as $\tilde{F}\circ \Pi$ where $\tilde{F}$ is faithful, and $\Pi$ is full and surjective on objects, which should remind you of similar factorizations in various categories.

In particular, $\tilde{F}: C/\ker(F) \to D$ is faithful and so makes $C/\ker(F)$ equivalent to the essential image of $F$ (but of course we can't expect $\tilde{F}$ to be full if $F$ wasn't ! However if $F$ is full, and essentially surjective, then we do get that $C/\ker(F)$ is equivalent to $D$)

Until here, no need for abelian categories. What changes a bit in abelian categories is essentially the same thing as what happens when you go from sets to groups: you can exchange your congruence for a normal subgroup (or here indeed a subgroup). That is, if you have an additive functor $F:C\to D$, then instead of having a congruence you can have, for each $A,B\in \mathrm{Ob}C$ a subgroup of $\hom(A,B)$ (the subgroup of those $f$ such that $F(f) = 0$: it is indeed a subgroup if $F$ is additive), and you can again construct $C/\ker(F)$; but you still won't have an equivalence with $D$ unless $F$ is full and essentially surjective to begin with.

I think (though I haven't checked that thoroughly) that $C/\ker(F)$ will be abelian as well and $\Pi$ additive.

I can somehow relate this to quotienting by the subcategory of the $A$'s with $F(A) = 0$ in the following sense. If $F(A) = 0$, then for any $B$ and $f,g: B\to A$, $F(f)=F(g)= 0_{F(B),F(A)}$ and so $f\sim g$ and so in $C/\ker(F)$, $A$ is terminal (and similarly initial) so it's isomorphic to $0$: in a sense you've turned $A$ into $0$ (perhaps that's what you meant by quotienting by it)

  • 1
    $\begingroup$ Thanks for your time. Here, the quotient is (serre) quotient category as defined in [link] (arxiv.org/pdf/1210.1425.pdf). In this paper, I finally understand when the functor F induces such equivalence. $\endgroup$ – math16 Aug 12 '18 at 8:54
  • $\begingroup$ Ah, sorry then I completely missed the point ! You should probably add some clarification in your question so that future answers may be more on point ! (however it's not clear to me why your subcategory $\ker(F)$ is thick in the sense defined in the article) $\endgroup$ – Max Aug 12 '18 at 9:21
  • $\begingroup$ You may use the stronger version of definition as in ncatlab.org/nlab/show/thick+subcategory said. $\endgroup$ – math16 Aug 12 '18 at 9:57
  • $\begingroup$ I still don't see it, but it must be me (why, if $B$ is a subobject of $A$ and $F(A)=0$, must $F(B)=0$ ? ) $\endgroup$ – Max Aug 12 '18 at 10:12
  • 2
    $\begingroup$ Oh, $F$ is assumed to be exact, ok ! you should probably mention that in your question as well $\endgroup$ – Max Aug 12 '18 at 10:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.