Iterated localizations are, to a first approximation, functors right orthogonal to conservative functors, i.e. the class of functors obtained from factoring any functor as some functor followed by a conservative functor.
For the correct statement one needs to regard as a 2-category, or at least as a model category.
Given a small category and a set of morphisms , we can always construct the localization category which is universal in the sense that every functor which inverts all the morphisms in factors through it (see localization):
On the other hand, it is often interesting to look at which morphisms of become isomorphisms under the action of a functor (e.g. might be a cohomology theory and we want to assess which maps in are quasi-isomorphisms).
Call the class of maps in which are inverted by (ndr: this isn’t standard notation). Then we can localize at and thus factor as a localization followed by a functor :
It turns out that this , in general, still inverts further morphisms. Therefore it makes sense to iterate this process to obtain the following sequence of functors:
where .
Now consider the colimit
of this sequence of localizations. Clearly there is a canonical functor and also a mediating map
such that . In fact, one can prove that is conservative.
A functor is an iterated localization if (1) is an equivalence.
This makes a(n homotopy) factorization system on (considered as a 1-category with weak equivalences given by equivalences of categories).
The above is pretty much directly taken from:
Joyal‘s CatLab, Ex. 6.12 in: Factorisation systems
André Joyal, §11.14 and pp. 70 in: Notes on quasi-categories (2008) [pdf, pdf]
Last revised on August 29, 2024 at 15:24:53. See the history of this page for a list of all contributions to it.