MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I asked this on, but didn't get any answer.

Let $\mathcal{A}$ be an abelian category and $\mathcal{C}$ a localizing subcategory in the sense of Gabriel. (A Serre/thick/dense subcategory, such that the quotient functor $T\colon \mathcal{A}\rightarrow\mathcal{A}/\mathcal{C}$ admits a right adjoint, the so called "section functor".) Then one can form the quotient category $\mathcal{A}/\mathcal{C}$.

Which properties inherits $\mathcal{A}/\mathcal{C}$ from $\mathcal{A}$? To be more precise:

  1. If $\mathcal{A}$ has enough injectives (resp. projectives), does $\mathcal{A}/\mathcal{C}$ too? If not, under which conditions?
  2. If $A\in \mathcal{A}$ is injective (resp. projective), is it $T(A)$, too? If not, under which conditions?
  3. If $A\in \mathcal{A}$ is a cogenerator, is it $T(A)$ too? If not, under which conditions?
  4. If $\mathcal{A}$ is complete, is it $\mathcal{A}/\mathcal{C}$ too? If not, under which conditions?

I know that:

  1. If $\mathcal{A}$ is cocomplete then so is $\mathcal{A}/\mathcal{C}$. ($T$ is a left adjoint and one can proove that $S$ is fully faithful.)
  2. If $\{U_i\}$ is a set of generators then so is $\{T(U_i)\}$.
  3. If $\mathcal{A}$ is AB5 then so is $\mathcal{A}/\mathcal{C}$. ($T$ commutes with filtered limits and one can prove that taking filtered limits is exact.)
  4. The second and the third point implies: If $\mathcal{A}$ is Grothendieck then so is $\mathcal{A}/\mathcal{C}$. (Gabriel–Popesco theorem even says that every Grothendieck category has this shape and one can choose $\mathcal{A}$ as a category of modules.)
  5. If $\mathcal{A}$ is complete with respect to finite limits then so is $\mathcal{A}/\mathcal{C}$. ($T$ is exact.)


  1. Jeremy Rickard gave counterexamples for the second point of my question.
  2. I answered the fourth point by myself: Let $F\colon \mathcal{D}\rightarrow \mathcal{E}$ be a functor with a fully faithful left adjoint $G$. Completeness of $\mathcal{D}$ implies completeness of $\mathcal{E}$ in general and this applies to the situation of quotient categories.
  3. Jeremy Rickard also answered the first question for projectives: There are Grothendieck categories with not enough projectives. They form a counterexample since they are quotients of a category of modules (Gabriel-Popescu) and every category of modules has enough projectives.
share|cite|improve this question
It just occurred to me that your comments on the Gabriel-Popescu theorem answer the "enough projectives" part of (1): every Grothendieck category is the quotient of a module category (which does have enough projectives) by a localizing subcategory, but not all Grothendieck categories have enough projectives. – Jeremy Rickard Jun 7 '13 at 12:25
Nice. Thanks! I didn't know Grothendieck categories without enough projectives. Thank you. Meanwhile, I solved more of my points. An update of the question will follow. – archipelago Jun 7 '13 at 16:55

Here's a simple counterexample for (2).

Let $\mathcal{A}$ be the category of sequences of linear maps $U\stackrel{\alpha}{\to}V\stackrel{\beta}{\to}W$ between vector spaces over a field $k$ such that $\beta\alpha=0$, and let $\mathcal{C}$ be the full subcategory of objects of the form $0\to0\to W$, so that $\mathcal{A}/\mathcal{C}$ is equivalent to the category of linear maps $U\to V$, with the quotient functor sending $U\to V\to W$ to $U\to V$ and the section functor sending $U\to V$ to $U\to V\to0$.

Then $0\to k=k$ is injective in $\mathcal{A}$, but $T(0\to k=k)$ is $0\to k$, which is not injective in $\mathcal{A}/\mathcal{C}$.

For the projective case, you can use the same $\mathcal{A}$, but take $\mathcal{C}$ to be the category of objects $U\to0\to0$, so that $\mathcal{A}/\mathcal{C}$ is again equivalent to the category of linear maps, the quotient functor sends $U\to V\to W$ to $V\to W$, and the section functor sends $V\stackrel{\beta}{\to}W$ to $\ker(\beta)\to V\stackrel{\beta}{\to}W$.

Then $k=k\to0$ is projective in $\mathcal{A}$ but $T(k=k\to0$) is $k\to 0$, which is not projective in $\mathcal{A}/\mathcal{C}$.

share|cite|improve this answer

A simple example of Grothendieck category without projectives is the category of torsion abelian groups. Another examples are the categories of sheaves.

About the other questions: If $\mathcal A$ is complete, then ${\mathcal A}/{\mathcal C}$ has to be also complete, since the fully faithful functor $S$ (which may be seen as an inclusion) has a left adjoint, hence it preserves limits. But, if $\mathcal A$ is complete then, an inclusion functor preserves limits iff the respective subcategory is closed under limits. See also Cocomplete but not complete abelian category

If $\mathcal A$ is Grothendieck, then it has enough injectives. The proof of this fact may be a consequence of Gabriel-Popescu theorem, as in B. Stenstrom, Rings of Quotients, Chap. X, Corollary 4.3, but may be shown directly by transfinite induction, as Grothendieck did in so called Tohoku paper:

Al. Grothendieck, "Sur quelques points d'algèbre homologique", The Tohoku Mathematical Journal. Second Series 9: 119–221

I think is a good question if the existence of injectives in $\mathcal A$ imply the same property for $\mathcal A/\mathcal C$.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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