homotopy hypothesis


Homotopy theory

Higher category theory

higher category theory

Basic concepts

Basic theorems





Universal constructions

Extra properties and structure

1-categorical presentations



The homotopy hypothesis is the assertion that

or rather the stronger statement that

and moreover

What this actually means in detail depends on which definition of ∞Grpd is being used and to which precise incarnation of Top it is being compared to.

There are some definitions of ∞-groupoids for which the homotopy hypothesis is a proven theorem . Depending on where in the spectrum between geometric definitions of higher categories and algebraic definitions of higher categories a given definition of \infty-groupoids is located, the statement may be more or less obvious.

For instance there is some justification for defining an \infty-groupoid to be equivalently a topological space (considered modulo weak homotopy equivalence). For this definition the homotopy hypothesis is of course a tautology.

A definition of \infty-groupoid that is still very geometrical but much more combinatorial is that given by Kan complexes. For these the homotopy hypothesis has a (non-trivial but fairly tractable) proof. The equivalence between Kan complexes and CW-complexes obtained this way is at the heart of all traditional homotopy theory.

A genuine algebraic definition of \infty-groupoids for which the homotopy theory has a (non-trivial but tractable) proof is given by algebraic Kan complexes.

However for other algebraic definitions of \infty-groupoids not much indication for how to prove the homotopy hypothesis is known. The definition of Trimble ω-category stands out as an algebraic definition that has the notion of fundamental ∞-groupoid built right into it, but also here it seems unclear at the moment how to make progress with proving the homotopy hypothesis.

In fact, generally the homotopy hypothesis is regarded as a consistency condition for definitions in higher category theory:

One way to justify this condition is by recourse to the proven cases of the homotopy hypothesis: experience shows that the collections of all three models – topological spaces, Kan complexes, algebraic Kan complexes – provide a model for ∞Grpd that supports the general abstract higher category theory – specifically (∞,1)-category theory – that one expects in analogy to how Set supports ordinary category theory. Any other definition of \infty-groupoids is hoped/required to reproduce this, and hence is hoped/required to satisfy the homotopy hypothesis.

Apart from having different models of \infty-groupoids that lend themselves more or less to a comparison with topological spaces, there is also the issue as to how to conceive of the notion of equivalence between \infty-groupoids.

The usual, unstated, implication is that the notion of equivalence of n-groupoids used to model homotopy n-types is the appropriate n-category-theoretic notion of equivalence. It is in this way that, for instance, it is known that 1-groupoids model homotopy 1-types (see below).

The reason this is important to specify is that there are other notions of equivalence on categorical structures which model homotopy types in other ways. For example, if we declare a functor between categories to be a weak equivalence iff its nerve is a weak equivalence of simplicial sets, then all homotopy types can be modeled by 1-categories in this way; see the Thomason model structure for 1-categories.

Finally, in analogy to the homotopy hypothesis, there are also attempts to relate general (∞,n)-categories (not necessarily groupoidal) to directed topological spaces by a fundamental (∞,n)-category-construction. There have been claims that a directed homotopy hypothesis can be proven, but at the moment there does not seem to be a published statement.

Abstract statement

The following general abstract statement of the homotopy hypothesis is often useful to make explicit.


There is an equivalence of (∞,1)-categories

(Π||):(\Pi \dashv |-|) : Top \simeq ∞Grpd.

This statement can be formulated, holds true and is proven below at least for the standard definitions of these two (∞,1)-categories (see section on Kan complexes, section on algebraic Kan complexes).


We discuss various different definitions of n-groupoids and ∞-groupoids and the formulation and proof of the homotopy hypothesis for them, to the extent that it is available.

For groupoids


The 2-category Grpd of groupoids, functors, and natural transformations is equivalent to the 2-category of homotopy 1-types, continuous maps, and homotopy classes of homotopies.

See homotopy hypothesis for 1-types for more.

For 2-groupoids

(…) 2-groupoids model all homotopy 2-types (…)

strict 2-groupoids suffice (…) (but note that strict 2-functors are not sufficient to model all maps between 2-types)

For Gray-groupoids

(…) Gray-groupoids model all homotopy 3-types (…)

It is known that not all homotopy 3-types can be modeled by strict 3-groupoids, but that Gray-categories (semi-strict 3-categories) suffice; the obstruction is the Whitehead product which arises from a nontrivial interchanger.


For Kan complexes

We write sSet QuillensSet_{Quillen} for sSet equipped with the classical model structure on simplicial sets. The cofibrant-fibrant objects in sSet QuillensSet_{Quillen} are precisely the Kan complexes.

Also we write Top QuillenTop_{Quillen} for Top equipped with the classical model structure on topological spaces.


There is a Quillen equivalence

(||Sing):TopSing||sSet Quillen, (|-| \dashv Sing) : Top \stackrel{\overset{|-|}{\leftarrow}}{\underset{Sing}{\to}} sSet_{Quillen} \,,

where the left adjoint |||-| is geometric realization and the right adjoint Sing()Sing(-) is forming the singular simplicial complex Sing()Sing(-).

This induces an equivalence of (∞,1)-categories of the corresponding presented (∞,1)-categories

(Quillen 67), see e.g. (Goerss-Jardine 96, section I.11, Joyal-Tierney 05, chapter I)

For algebraic Kan complexes

An algebraic Kan complex is an algebraic definition of higher groupoids obtained by taking the ordinary definition of Kan complex and equipping these with choices of horn-fillers. These choice encode specified composition operations, specified associators for these, specified pentagonators and so on.

Algebraic Kan complexes constitute a genuine algebraic model in that they are precisely the algebras over a monad on sSet.


The transferred model structure on algebraic Kan complexes is Quillen equivalent to the standard model structure on simplicial sets

(FU):AlgKanUFsSet Quillen. (F \dashv U) : Alg Kan \stackrel{\overset{F}{\leftarrow}}{\underset{U}{\to}} sSet_{Quillen} \,.

The proof is spelled out at model structure on algebraic fibrant objects.


With the above homotopy hypothesis-theorem for Kan complexes this gives a zig-zag of Quillen equivalences between AlgKanAlg Kan and TopTop

AlgKanUFsSet QuillenSing||Top. Alg Kan \stackrel{\overset{F}{\leftarrow}}{\underset{U}{\to}} sSet_{Quillen} \stackrel{\overset{|-|}{\to}}{\underset{Sing}{\leftarrow}} Top \,.

This already yields the homotopy hypothesis for algebraic Kan complexes at the level of the corresponding presented (∞,1)-categories (as discussed there)

AlgC Top . Alg C^\circ \simeq Top^\circ \,.

But there is also a direct Quillen equivalence:


Write Δ n\Delta^n and Λ k n\Lambda^n_k for the topological nn-simplex and its kk-horn.

Fix any choice of retracts

R(n,k):Δ nΛ k n R(n,k) : \Delta^n \to \Lambda^n_k

for all topological horn inclusions Λ i nΔ n\Lambda^n_i \hookrightarrow \Delta^n.

For XX any topological space equip the singular simplicial complex SingXSing X with the stucture of an algebraic Kan complex by taking the filler of Λ k[n]SingX\Lambda_k[n] \to Sing X to be given by the (||Sing)(|-| \dashv Sing)-adjunct of Δ nR(n,k)Λ k nX\Delta^n \stackrel{R(n,k)}{\to} \Lambda^n_k \to X. Write Π (X)AlgKan\Pi_\infty(X) \in Alg Kan for the resulting algebraic Kan complex.

This construction constitutes a functor

Π ():TopAlgC, \Pi_\infty(-) : Top \to Alg C \,,

with UΠ =SingU \Pi_\infty = Sing.


The choices of fillers in Π (X)\Pi_\infty(X) may be thought of as explicit choice of reparameterizations of paths in XX. These choices are arbitrary, but by the general statement at model structure on algebraic fibrant objects, any two chocies yield equivalent objects.


Given choices R(,)R(-,-) of horn retracts as above, define a functor

|| r:AlgKanTop |-|_r : Alg Kan \to Top

called reduced geometric realization by taking it on an object AAlgKanA \in Alg Kan to be given by the coequalizer

|A| r:=lim ( Λ[n] kAΔ n|UA|), |A|_r := \lim_{\to}(\coprod_{\Lambda[n]_k \to A} \Delta^n \stackrel{\to}{\to} |U A|) \,,

where |UA||U A| is the ordinary geometric realization of the underlying simplicial set of AA and where the two maps are

  1. the image under |||-| of the distinguished fillers Δ[n]UA\Delta[n] \to U A of AA;

  2. the composite Δ nR(n,k)Λ k n|UX|\Delta^n \stackrel{R(n,k)}{\to} \Lambda^n_k \to |U X| .


The functor || r|-|_r is left adjoint to Π \Pi_\infty.

(|| rΠ ). (|-|_r \dashv \Pi_\infty) \,.

This is (Nikolaus, prop. 3.4).


We check the hom-isomorphism. A morphism f:|A| rXf : |A|_r \to X is by definition of the coequalizer the same as a map f˜:|A|X\tilde f : |A| \to X such that for each horn h:Λ[n] kAh : \Lambda[n]_k \to A with distinguished filler h^:Δ[n]A\hat h : \Delta[n] \to A the composites

Δ nR(n,k)Λ k n|h||UA|f˜X \Delta^n \stackrel{R(n,k)}{\to} \Lambda^n_k \stackrel{|h|}{\to} |U A| \stackrel{\tilde f}{\to} X


Δ n|h^||UA|f˜X \Delta^n \stackrel{|\hat h|}{\to} |U A| \stackrel{\tilde f}{\to} X

are equal. This means equivalently that the (||Sing)(|-| \dashv Sing)-adjunct f˜˜:UASingX\tilde \tilde f : U A \to Sing X sends distinguished fillers in AA to distinguished fillers in Π (X)\Pi_\infty(X) and is hence a morphism in AlgKanAlg Kan.

the constructon shows that the map (f:|A| rX)(f˜˜:AΠ (X))(f : |A|_r \to X) \mapsto (\tilde \tilde f : A \to \Pi_\infty(X)) thus obtained is a bijection.


We have an identity

AlgKan U = Π sSet Sing Top \array{ && Alg Kan \\ & {}^{\mathllap{U}}\swarrow & \Downarrow^{=}& \nwarrow^{\mathrlap{\Pi_\infty}} \\ sSet &&\underset{Sing}{\leftarrow}&& Top }

and a natural isomorphism

AlgKan F || r sSet || Top. \array{ && Alg Kan \\ & {}^{\mathllap{F}}\nearrow & \Downarrow^{\simeq}& \searrow^{\mathrlap{|-|_r}} \\ sSet &&\underset{|-|}{\to}&& Top } \,.

This is (Nikolaus, corollary 3.5)


The identity is evident by definition of Π \Pi_\infty.

Using this, we have that

(|| rFUΠ =Sing). (|-|_r \circ F \dashv U \circ \Pi_\infty = Sing) \,.

So || rF|-|_r \circ F is another left adjoint to SingSing and hence naturally isomorphism to |||-|.


The adjunction

(|| rΠ ):AlgCTop (|-|_r \dashv \Pi_\infty) : Alg C \to Top

constitutes a Quillen equivalence

This is Nikolaus, corollary 3.6


By the above theorem and the 2-out-of-3-property of Quillen equivalences.

For cubical sets

Also cubical sets may serve as a model for homotopy theory.

There is an evident simplicial set-valued functor

sSet \Box \to sSet

from the cube category to sSet, which sends the cubical nn-cube to the simplicial nn-cube

1 n(Δ[1]) ×n. \mathbf{1}^n \mapsto (\Delta[1])^{\times n} \,.

Similarly there is a canonical Top-valued functor

Top \Box \to Top
1 n(Δ Top 1) n. \mathbf{1}^n \mapsto (\Delta^1_{Top})^n \,.

The corresponding nerve and realization adjunction

(||Sing ):TopSing ||Set op (|-| \dashv Sing_\Box) : Top \stackrel{\overset{|-|}{\leftarrow}}{\underset{Sing_\Box}{\to}} Set^{\Box^{op}}

is the cubical analogue of the simplicial nerve and realization discussed above.


There is a model structure on cubical sets Set opSet^{\Box^{op}} whose

  • weak equivalences are the morphisms that become weak equivalences under geometric realization |||-|;

  • cofibrations are the monomorphisms.

This is Jardine, sections 3.


The unit of the adjunction

ASing (|A|) A \to Sing_\Box(|A|)

is a weak equivalence in Set Set^{\Box} for every cubical set AA.

The counit of the adjunction

|Sing X|X |Sing_\Box X| \to X

is a weak equivalence in TopTop for every topological space XX.

It follows that we have an equivalence of categories induced on the homotopy categories

Ho(Top)Ho(Set op). Ho(Top) \simeq Ho(Set^{\Box^{op}}) \,.

This is Jardine, theorem 29, corollary 30.

For cat ncat^n groups and n+1n+1-fold groupoids

Loday’s notion of a cat-n-group corresponds to the connected version of an n+1n+1-fold groupoid. We will restrict our discussion to that connected case.


The homotopy category of cat-n-groups is equivalent to that of pointed connected homotopy n+1-types.

This is proven in (Loday). (There are some glitches in his proof and these were fixed by various authors (Steiner, Gilbert, ..) and then detailed proofs were given by Bullejos, Cegarra, Duskin and separately, using the equivalent formulation of crossed n-cubes, by Porter. Detailed references and some more commentary is at cat-n-group.)

For Segal-groupoids


There is realization/singular complex adjunction

(||sSing):TopSegalGrpd (|-| \dashv sSing) : Top \to SegalGrpd

for Segal groupoids,

Its unit is an equivalence of Segal categories and its counit a weak homotopy equivalence of topological spaces.


This is lemma 6.3.21 and corollary 6.3.24 in (Pellissier)

For strict ω\omega-categories with weak inverses

While strict omega-groupoids in the sense of strict omega-categories with strict inverses are far from modelling all homotopy types, strict ω\omega-categories with all weak inverses come closer. In (Kapranov-Voevodsky) it was argued that these are in fact sufficient, but a mistake in the argument is claimed in (Simpson, cor. 5.2) (see also here).

The issue however is somewhat subtle, as very much highlighted by Voevodsky here. For more on this see at Simpson's conjecture.

For stratified spaces

See Ayala-Francis-Rozenblyum.

References and a bit of history.

The equivalence between the classical homotopy theory of topological spaces and the homotopy theory of Kan complexes is due to

Later in

it was argued that there were algebraic definitions of higher groupoids that could be put forward, so that the resulting objects ought to have a homotopy theory equivalent to the classical homotopy theory of topological spaces, and, moreover, would provide useful tools in the interpretation of non-abelian cohomology classes. This extended ideas that Grothendieck had explored in letters to Larry Breen in the mid 1970s in which he had given a sketch of a theory of nn-stacks and their relation with the homotopy nn-type of a space or more generally a topos.

At this stage, (in the 1970s and early 1980s) more geometric or combinatorial definitions of infinity categories were not yet available, or, perhaps more accurately, had been discovered, but were not recognised as having such an infinity categoric interpretation. These models included Kan complexes which now are interpreted as being one model for infinity groupoids, and weak Kan complexes, as put forward by Boardman and Vogt, which give one of the main models for (weak) (,1)(\infty,1)-categories. From this perspective, Quillen’s result can be seen as being a precursor of one form of the Homotopy Hypothesis.

The name “homotopy hypothesis” for this statement is due to

Technical reviews of Quillen’s proof of the homotopy hypothesis includes

The homotopy hypothesis for strict ω\omega-categories with weak inverses is discussed in

  • Mikhail Kapranov, Vladimir Voevodsky, \infty-groupoids and homotopy types, Cahiers de Topologie et Géométrie Différentielle Catégoriques, 32 no. 1 (1991), p. 29-46

but a mistake in the argument is claimed in cor 5.2 of

The homotopy hypothesis for algebraic Kan complexes is established and discussed in

The homotopy hypothesis for Segal groupoids is formulated in section 6.3.4 of

  • Regis Pellissier, Weak enriched categories - Categories enrichies faibles PhD Thesis (2002) (arXiv:math/0308246)

Models of homotopy nn-types by Cat nCat^n-groups are discussed in

  • Jean-Louis Loday, Spaces with finitely many non-trivial homotopy groups, J. Pure Appl. Algebra 24 (1982), 179-202.

More literature on models of homotopy types by strict higher groupoids is at

The first paper, as its title suggests, has an emphasis on using higher groupoids for computation of homotopical invariants, in fact by applying higher homotopy van Kampen Theorems. These theorems lead to algebraic colimit arguments in algebraic topology, implying results, often nonabelian, not obtainable by other methods. It is also remarkable that the precision of these results requires the use of strict structures, whereas the current emphasis in higher category theory is on non strict structures.

The homotopy theory of cubical sets is discussed in

  • Jardine, Model structure on cubical sets (pdf)
  • Maltsiniotis, G. La catégorie cubique avec connexions est une catégorie test stricte. Homology, Homotopy Appl. 11, (2) (2009) 309–326.

Cubical methods are also essential in

  • R. Brown, P.J. Higgins, R. Sivera, Nonabelian algebraic topology: filtered spaces, crossed complexes, cubical homotopy groupoids, EMS Tracts in Mathematics Vol. 15, 703 pages. (August 2011).

A version for stratified spaces is discussed in

Revised on May 19, 2017 11:53:47 by Urs Schreiber (