1
$\begingroup$

Localization of a category deals with the family of morphisms rendered invertible by a functor, i.e. the (saturated) family of weak equivalences, which I think resembles the kernel of a group or a monoid when the category being regarded as an algebraic structure as a group or a monoid.

Whereas, the family of morphisms f such that Qf=identity for some functor Q does not play any central role although it more looks like a kernel to me than the family of weak equivalences just because an identity behaves more like the identity element of a group than an isomorphism.

Being just a beginner for the category theory, I'd like for someone to tell me briefly why that family of morphisms I said does not matter concerning localizations.

$\endgroup$
6
  • 1
    $\begingroup$ If I am not mistaken, you are trying to understand localization of categories as some kind of generalization of quotient of groups. That approach does not seem quite right! Rather, you can see a localization of categories as a generalization of localization of rings! Or, if you only consider the trivial localization (localization w.r.t all the morphisms), it could be seen as a generalization of the adjunction between monoids and groups. $\endgroup$
    – Fernando
    Jan 4, 2015 at 13:12
  • 1
    $\begingroup$ thanks, Fernando! You're right. I thought as you said. But then I wonder what procedure in category theory corresponds to a generalization of quotient of groups. I thought the quotient category of "categories for the working mathematician" is basically similar to localization. $\endgroup$
    – Sn K
    Jan 4, 2015 at 13:20
  • $\begingroup$ Usuallly, we use quotient categories to deal with localization of categories. But, as far as I know, you can't say it's the same thing: there are quotient categories of a given category $X$ that can't be realized as a localization of $X$... $\endgroup$
    – Fernando
    Jan 4, 2015 at 13:53
  • $\begingroup$ I mean, although quotient and localization are important to each other, they are not the same thing (and I wouldn't say they are "similar"). $\endgroup$
    – Fernando
    Jan 4, 2015 at 14:05
  • $\begingroup$ You said "Usuallly, we use quotient categories to deal with localization of categories." If localization has some relation with quotient categories , to deal with quotient categories (so localization too), I think it could be also useful to check the family of morphisms that are sent to identities. But anyway I understand what you said and great thanks for your answers! $\endgroup$
    – Sn K
    Jan 4, 2015 at 14:11

2 Answers 2

1
$\begingroup$

Localization is a very different operation than taking the quotient of your category. When you localize you end up with all sorts of new maps because of zigzags (i.e. you can have compositions with the new maps you end up with). So it's not obvious to me that there is a functor $Q$ to capture these maps. Just a couple of days ago there was a MO thread about the difference, and the answers there should help you: Localizations or quotients of categories?

$\endgroup$
1
$\begingroup$

Let me give you some general thoughts, although I am not answering the question directly. As it was pointed out, quotient and localization are not similar constructions. Although, usually quotient categories are used to deal with localization problems - for example, in model category theory (where you construct a homotopy category (a special quotient category) which is equivalent to the localized category), or even in direct constructing a localization of a small category (as the localization of $ \mathbb{N} $).

To see the difference, you may observe the following: For simplicity, consider only functors bijective onto the objects and small categories. You can see that all such fully faithful functors could be seen as a quotient functor (projection of a category in a quotient category). But they are not always a localization!

On the other hand, if you think about categories in general, the unit component/projection $X\to Localization (X) $ is not fully faithful in general, while the quotient functors are always fully faithful.

Quotients, in general (in mathematics), are special types of colimits: this is the case, for instance, for groups, topology and for categories. That said, you can conjecture that this is the reason why such "quotients" are well used to construct other types of colimits or images of left adjoint functors.

Concerning about your question, although quotient categories are somehow similar to quotient of groups (as I said in the last paragraph), quotient of categories is not a horizontal categorification of the quotient of groups. Rather, I would say that quotient categories may be seen as a horizontal categorification of quotient monoids.

And if you take a look at quotient monoids, it cannot be seen easily as quotient in the case of groups. I mean, precisely, if I am not mistaken right now, quotient of monoids can't be controlled by the inverse image of the unit/identity (as you do with groups). Take, for instance, the morphism $ \mathbb{N}\ast \mathbb{N}\to \mathbb{N}\times \mathbb{N} $, in which $ \ast $ denotes the coproduct (free product).

This morphism, by definition, is induced by the two different canonical inclusions $ \mathbb{N}\to \mathbb{N}\times \mathbb{N} $.

This morphism is a projection of a quotient (quotient of categories): however the inverse image of the identity is trivial.

As I said, the best analogy for localization in dimension $0$ seems to be localization of rings (or localization of monoids). So, my thought is that you can consider the localization as a horizontal categorification of localization of monoids.

At last, formally we have the following: consider the category $ sGpd $ of groupoids posets - or the category of groupoids such that each connected component is simply connected. A category with a suitable relation is just a $ sGpd $ - enriched category. We have a canonical inclusion $ U: CAT\to sGpd $ induced by the inclusion $ Set\to sGpd $ which sends each set to the discrete groupoid. If you have a Gpd-category $X$, its quotient is just its left 2-reflection along $U$.

On the other hand, to consider localization in general, it seems harder. But avoiding technical problems, you may consider the 2-category $W$ of categories endowed with (suitable) subcategories of weak equivalences and functors preserving such subcategories (and natural transformations). Again, there is a canonical inclusion $ T: CAT\to W $ which sends each category to the category endowed with the subcategory of isomorphisms. And the localization of a category $ X $ w.r.t. a suitable subcategory would be (if it exists) the left 2-reflection along $T$ of $X$ (considering $X$ as an object of $W$).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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