basic localizer


Category theory

Homotopy theory

homotopy theory, (∞,1)-category theory, homotopy type theory

flavors: stable, equivariant, rational, p-adic, proper, geometric, cohesive, directed

models: topological, simplicial, localic, …

see also algebraic topology



Paths and cylinders

Homotopy groups

Basic facts


Basic localizers


By a basic localizer one means a localizer on the category Cat of categories, hence a choice of a class of functors to be called the weak equivalences, subject to some conditions.

These conditions ensure in particular that a basic localizer always contains the weak equivalences of the Thomason model structure on Cat (see Maltsiniotis 11, 1.2.1), the localization at which is equivalent to the standard homotopy category. On the other hand, the standard equivalences of categories, which are the weak equivalences in the canonical model structure on Cat, do not form a basic localizer.

Hence basic localizers are a tool for homotopy theory modeled on category theory. In fact, their introduction by Grothendieck was motivated from the study of test categories (see remark 2 below).


The definition is due to Grothendieck:


A basic localizer is a class WW of morphisms in Cat such that

  1. WW contains all identities, satisfies the 2-out-of-3 property and is closed under retracts (in the literature this is sometimes called being weakly saturated),

  2. If AA has a terminal object, then the functor A1A\to 1 is in WW, and

  3. Given a commutative triangle in CatCat:

    A u B v w C\array{ A & & \overset{u}{\to} & & B\\ & _v \searrow & & \swarrow_w \\ & & C }

    if each induced functor v/cw/cv/c \to w/c between comma categories is in WW, then uu is also in WW.


The term in French is localisateur fondamental, which is sometimes translated as fundamental localizer.


In Pursuing Stacks Grothendieck wrote about def. 1 the following:

These conditions are enough, I quickly checked this night, in order to validify all results developed so far on test categories, weak test categories?, strict test categories, weak test functors and test functors (with values in (Cat)(Cat)) (of notably the review in par. 44, page 79–88), provided in the case of test functors we restrict to the case of loc. cit. when each of the categories i(a)i(a) has a final object. All this I believe is justification enough for the definition above.



Asphericity and local equivalences

If WW is a basic localizer, we define the following related classes. We sometimes refer to functors in WW as weak equivalences.

  • A category AA is (WW-)aspherical if A1A\to 1 is in WW. Thus the second axiom says exactly that any category with a terminal object is aspherical.
  • A functor u:ABu\colon A\to B is (WW-)aspherical if for all bBb\in B, the comma category u/bu/b is aspherical.
  • When the hypotheses of the third axiom are satisfied, we say that uu is a local weak equivalence over CC. Thus the third axiom says exactly that every local weak equivalence is a weak equivalence.

If W=π 0W = \pi_0-equivalences, then a category is aspherical iff it is connected, and a functor is aspherical iff it is initial.


If W=W = nerve equivalences, then a category is aspherical iff its nerve is contractible, and a functor is aspherical iff it is homotopy initial.

We observe the following.

  • A category AA is aspherical iff the functor A1A\to 1 is aspherical, since the only comma category involved in the latter assertion is AA itself.

  • An aspherical functor is a weak equivalence. For if u:ABu\colon A\to B is aspherical, then consider the triangle

    A u B u B\array{ A & & \overset{u}{\to} & & B\\ & _u \searrow & & \swarrow \\ & & B }

    The third axiom tells us to consider, for a given bBb\in B, the functor u/bB/bu/b \to B/b. But u/bu/b is aspherical by assumption, while B/bB/b is aspherical by the second axiom since it has a terminal object. Thus, by 2-out-of-3, the functor u/bB/bu/b \to B/b is in WW, and thus by the third axiom uu is in WW.

  • If uu has a right adjoint, then it is aspherical. For in this case, each category u/bu/b has a terminal object, and thus is aspherical.

  • If II denotes the interval category, then for any category AA the projection A×IAA\times I\to A has a right adjoint, hence is aspherical and thus a weak equivalence. By 2-out-of-3, the two injections AA×IA \rightrightarrows A\times I are also weak equivalences, so A×IA\times I is a cylinder object for WW. It follows that if we have a natural transformation fgf\to g, then ff is in WW if and only if gg is. Moreover, if ff is a “homotopy equivalence” in the sense that it has an “inverse” gg such that fgf g and gfg f are connected to identities by arbitrary natural zigzags, then ff is a weak equivalence.

  • In particular, any left or right adjoint is a weak equivalence.


It is a non-obvious fact that the notion of basic localizer is self-dual.


A functor u:ABu : A \to B is in a basic localizer WW if and only if u op:A opB opu^{op} : A^{op} \to B^{op} is in WW.


See Proposition 1.2.6 in (Cisinski 04).

Cisinski’s theorem

Since the definition consists merely of closure conditions, the intersection of any family of basic localizers is again a basic localizer. It follows that there is a unique smallest basic localizer. The following was conjectured by Grothendieck and proven by Denis-Charles Cisinski.

Theorem (Cisinski)

The class of functors whose nerve is a weak homotopy equivalence is the smallest basic localizer.


See Théorème 2.2.11 in (Cisinski 04).

Note that this is a larger class than the class of “homotopy equivalences” considered above. For instance, the category generated by the graph

\bullet \leftarrow \bullet \rightarrow \bullet \leftarrow \bullet \rightarrow \bullet \leftarrow \bullet \rightarrow \cdots

has a contractible nerve, but its identity functor is not connected to a constant one by any natural zigzag.


See also at Cisinski model structure.

Revised on July 18, 2013 12:26:07 by Urs Schreiber (