homotopy theory



In generality, homotopy theory is the study of mathematical contexts in which functions or rather (homo-)morphisms are equipped with a concept of homotopy between them, hence with a concept of “equivalent deformations” of morphisms, and then iteratively with homotopies of homotopies between those, and so forth.

A key aspect is that in such homotopy theoretic contexts the concept of isomorphism is relaxed to that of homotopy equivalence: Where a morphism is regarded as invertible if there is a reverse function such that both composites are equal to the identity morphism, for a homotopy equivalence one only requires the composites to be homotopic to the identity. Regarding objects in a homotopical context up to homotopy equivalence this way is to regard them as homotopy types.

The classical example is the classical homotopy theory of topological spaces, where one considers topological spaces with continuous functions between them, and with the original concept of topological homotopies between these continuous functions. The category whose objects are topological spaces and whose morphisms are homotopy equivalence-class of continuous functions is also called the classical homotopy category.

The classical homotopy theory of topological spaces has many applications, for example to covering space theory, to classifying space theory, to generalized (Eilenberg-Steenrod) cohomology theory and many more. (See also at shape theory.) Accordingly, homotopy theory has a large overlap with algebraic topology.

The classical homotopy theory of topological spaces may be abstracted to yield an “abstract homotopy theory” that applies to a large variety of contexts. There are several more or less equivalent formalizations of the concept of “abstract homotopy theory”, including

The terminology model category is short for “category of models of homotopy types”. The idea here is to consider categories equipped with suitable extra structure and properties that encodes the existence of homotopies between all morphisms and convient means to handle and control these, in particular a means to construct the corresonding homotopy category.

For a detailed introduction to homotopy theory from this perspective see at Introduction to Homotopy Theory.

The approach of (∞,1)-categories to homotopy theory is meant to be more truthful to the intrinsic nature of homotopy theory. Instead of equipping an ordinary category with a extra concept of homotopy between its morphisms, here one regards the resulting structure as a higher category where the homotopies themselves appear as a kind of higher order morphisms, called 2-morphisms and where higher homotopies of homotopies are regarded as k-morphisms for all kk.

The terminology “(∞,1)-category” signifies that homotopy theory is but one special case of general higher category theory, namely (∞,1)-categories (hence homotopy theories) are those infinity-categories in which all k-morphisms for k>1k \gt 1 are invertible up to homotopy. If one drops this constraint, so that homotopies become “directed” then one might still speak of “directed homotopy theory”.

The archetypical example of an (∞,1)-category is the (,1)(\infty,1)-category ∞Grpd of ∞-groupoids, just as Set is the archetypical 1-category.

This turns out to be equivalent, as homotopy theories, to the classical homotopy theory of topological spaces if restricted to those that admit the structure of CW-complexes. Another homotopy theory equivalent to this archetypical one is the classical homotopy theory of simplicial sets, see also at simplicial homotopy theory.

But there are many other homotopy theories besides (the various incarnations of) this classical one. Important sub-classes of homotopy theories include

The geometric homotopy theory of (∞,1)-toposes in particular serves as the foundation for higher geometry/derived geometry. This is relevant notably in the physics of gauge theory, where gauge transformations are identified with homotopies in geometric homotopy theory. For more on this see at geometry of physics -- homotopy types.

On the other hand, the incarnation of homotopy theory as homotopy type theory exhibits the remarkably foundational nature of homotopy theory. Contrary to its original appearance as a fairly complicated-looking theory built on top of classical set theory and classcal topology, homotopy theory turns out to be intrinsically simple: it arises from plain dependent type theory just by adopting a fully constructive attitude towards the concept of identity/equality, see at identity type for more on this. For exposition of this perspective see (Shulman 17).


A convenient, powerful, and traditional way to deal with (∞,1)-categories is to “present” them by 1-categories with specified classes of morphisms called weak equivalences : a category with weak equivalences or homotopical category. The idea is as follows. Given a category CC with a class WW of weak equivalences, we can form its homotopy category or category of fractions C[W 1]C[W^{-1}] by adjoining formal inverses to all the morphisms in WW. The (∞,1)-category presented by (C,W)(C,W)“ can be thought of as the result of regarding CC as an \infty-category with only identity kk-cells for k>1k\gt 1, then adjoining formal inverses to morphisms in WW in the \infty-categorical sense; that is, making them into equivalences rather than isomorphisms. It is remarkable that most (,1)(\infty,1)-categories that arise in mathematics can be presented in this way.

As with presentations of groups and other algebraic structures, very different presentations can give rise to equivalent (,1)(\infty,1)-categories. For example, several different presentations of the (,1)(\infty,1)-category of \infty-groupoids are:

The latter three can hence be regarded as providing “combinatorial models” for the homotopy theory of topological spaces.

Model Categories

The value of working with presentations of (,1)(\infty,1)-categories rather than the (,1)(\infty,1)-categories themselves is that the presentations are ordinary 1-categories, and thus much simpler to work with. For instance, ordinary limits and colimits are easy to construct in the category of topological spaces, or of simplicial sets, and we can then use these to get a handle on (,1)(\infty,1)-categorical limits and colimits in the (,1)(\infty,1)-category of \infty-groupoids. However, we always have to make sure that we use only 1-categorical constructions that are homotopically meaningful, which essentially means that they induce (,1)(\infty,1)-categorical meaningful constructions in the presented (,1)(\infty,1)-category. In particular, they must be invariant under weak equivalence.

Most presentations of (,1)(\infty,1)-categories come with additional classes of morphisms, called fibrations and cofibrations, that are very useful in performing constructions in a homotopically meaningful way. Quillen defined a model category to be a 1-category together with classes of morphisms called weak equivalences, cofibrations, and fibrations that fit together in a very precise way (the term is meant to suggest “a category of models for a homotopy theory”). Many, perhaps most, presentations of (,1)(\infty,1)-categories are model categories. Moreover, even when we do not have a model category, we often have classes of cofibrations and fibrations with many of the properties possessed by cofibrations and fibrations in a model category, and even when we do have a model category, there may be classes of cofibrations and fibrations, different from those in the model structure, that are useful for some purposes.

Unlike the weak equivalences, which determine the “homotopy theory” and the (,1)(\infty,1)-category that it presents, fibrations and cofibrations should be regarded as technical tools which make working directly with the presentation easier (or possible). Whether a morphism is a fibration or cofibration has no meaning after we pass to the presented (,1)(\infty,1)-category. In fact, every morphism is weakly equivalent to a fibration and to a cofibration. In particular, despite the common use of double-headed arrows for fibrations and hooked arrows for cofibrations, they do not correspond to (,1)(\infty,1)-categorical epimorphisms and monomorphisms.

In a model category, a morphism which is both a fibration and a weak equivalence is called an acyclic fibration or a trivial fibration. Dually we have acyclic or trivial cofibrations. An object XX is called cofibrant if the map 0X0\to X from the initial object to XX is a cofibration, and fibrant if the map X1X\to 1 to the terminal# object is a fibration. The axioms of a model category ensure that for every object XX there is an acyclic fibration QXXQ X \to X where QXQ X is cofibrant and an acyclic cofibration XRXX\to R X where RXR X is fibrant.


For a (higher) category theorist, the following examples of model categories are perhaps the most useful to keep in mind:

  • C=C= sets, W=W= isomorphisms. All morphisms are both fibrations and cofibrations. The (,1)(\infty,1)-category presented is again the 1-category SetSet.
  • C=C= categories, W=W= equivalences of categories. The cofibrations are the functors which are injective on objects, and the fibrations are the isofibrations. The acyclic fibrations are the equivalences of categories which are literally surjective on objects. Every object is both fibrant and cofibrant. The (,1)(\infty,1)-category presented is the 2-category CatCat. This is often called the folk model structure.
  • C=C= (strict) 2-categories and (strict) 2-functors, W=W= 2-functors which are equivalences of bicategories. The fibrations are the 2-functors which are isofibrations on hom-categories and have an equivalence-lifting property. Every object is fibrant; the cofibrant 2-categories are those whose underlying 1-category is freely generated by some directed graph. The (,1)(\infty,1)-category presented is the (weak) 3-category 2Cat2Cat. This model structure is due to Steve Lack.

Generalized Morphisms

The morphisms from AA to BB in the (,1)(\infty,1)-category presented by (C,W)(C,W) are zigzags \stackrel{\simeq}{\leftarrow} \to \stackrel{\simeq}{\leftarrow} \to \cdots ; these are sometimes called generalized morphisms. Many presentations (including every model category) have the property that any such morphism is equivalent to one with a single zag, as in \stackrel{\simeq}{\leftarrow} \to \stackrel{\simeq}{\leftarrow}. In a model category, a canonical form for such a zigzag is XQXRYYX \stackrel{\simeq}{\leftarrow} Q X \to R Y \stackrel{\simeq}{\leftarrow} Y where QXQ X is cofibrant and RYR Y is fibrant. In this case we can moreover take QXXQ X\to X to be an acyclic fibration and YRYY\to R Y to be an acyclic cofibration.

Often it suffices to consider even shorter zigzags of the form \stackrel{\simeq}{\leftarrow} \to or \to \stackrel{\simeq}{\leftarrow}. In particular, this is the case if every object is fibrant or every object is cofibrant. For example:

  • If XX and YY are strict 2-categories, then pseudofunctors XYX\to Y are equivalent to strict 2-functors QXYQ X \to Y, where QXQ X is a cofibrant replacement for XX.
  • anafunctors are zigzags \stackrel{\simeq}{\leftarrow} \to in the folk model structure on 1-categories whose first factor is an acyclic (i.e. surjective) fibration.
  • Morita morphisms in the theory of Lie groupoids are generalized morphisms of length one where both maps are acyclic fibrations.

If XX is cofibrant and YY is fibrant, then every generalized morphism from XX to YY is equivalent to an ordinary morphism. For example, if XX is a cofibrant 2-category, then every pseudofunctor XYX\to Y is equivalent to a strict 2-functor XYX\to Y

Quillen Equivalences

Quillen also introduced a highly structured notion of equivalence between model categories, now called a Quillen equivalence, which among other things ensures that they present the same (,1)(\infty,1)-category. Quillen equivalences are now being used to compare different definitions of higher categories.

“The ultimate aim of algebraic homotopy is to construct a purely algebraic theory, which is equivalent to homotopy theory in the same sort of way that ‘analytic’ is equivalent to ‘pure’ projective geometry.”

This theme was taken up by Baues, (1988), using a type of abstract homotopy theory closely related to Ken Brown’s categories of fibrant objects. Whitehead’s own work was extended by Ronnie Brown and Phil Higgins, see nonabelian algebraic topology.


An exposition with an eye towards homotopy type theory is in

The original axiomatization of homotopy theory by model categories is due to

  • Daniel Quillen, Homotopical algebra, Lecture Notes in Mathematics 43, Berlin, New York, 1967

The similar axiomatization involving the weaker structure of a calculus of fractions is due to

A standard account of the modern form of simplicial homotopy theory is in

Formulation of abstract homotopy theory as the theory of (∞,1)-toposes is due to

and the formalization of this in the internal language of homotopy type theory is due to

A foundational set of lecture notes is developing in

See also

Brief indications of open questions and future directions (as of 2013) of algebraic topology and stable homotopy theory are in

and in

  • Problems in homotopy theory (wiki)

Revised on April 12, 2017 07:37:00 by Urs Schreiber (