Higher category theory

higher category theory

Basic concepts

Basic theorems





Universal constructions

Extra properties and structure

1-categorical presentations



In higher category theory an (,n)(\infty,n)-category may be thought of as

Accordingly, the notion of (,n)(\infty,n)-categories is a joint generalization of categories, 2-categories, 3-categories, 4-categories, etc. and of ∞-groupoids / homotopy types and (∞,1)-categories. From the point of homotopy theory they are a generalization to directed homotopy theory, from the point of view of homotopy type theory they are a generalization to directed homotopy type theory.

There are two main recursive definitions of (,n)(\infty,n)-categories:

  1. by iterated (∞,1)-enrichment

    Cat (,n)(((Cat (,0)Cat)Cat))Cat Cat_{(\infty,n)} \simeq (\cdots((Cat_{(\infty,0)} Cat) Cat) \cdots) Cat
  2. by iterated (∞,1)-internalization

Cat (,n)Cat((Cat(Cat (,0)))). Cat_{(\infty,n)} \simeq Cat(\cdots(Cat(Cat_{(\infty,0)}))\cdots) \,.

There is also a fairly simple axiomatization of the (∞,1)-category Cat (,n)Cat_{(\infty,n)} itself, as something generated by strict n-categories.

Then there is also a plethora of model category structures that present the (∞,1)-category Cat (,n)Cat_{(\infty,n)} of all (,n)(\infty,n)-categories, which means that there are many (and many different) very explicit ways to describe them.

A central result of (,n)(\infty,n)-category theory is the proof of the cobordism hypothesis, which revolves around the (∞,n)-category of cobordisms. This turns out to be the free symmetric monoidal (∞,n)-category with duals and provides deep relations between algebraic topology, higher algebra and extended topological quantum field theory. Other fundamental examples of (,n)(\infty,n)-categories, also in this context, are (∞,n)-categories of spans and of (∞,n)-vector spaces.

While the subject is still young, visible at the horizon is its role in higher topos theory. Where (∞,1)-toposes regarded as (∞,1)-categories of (∞,1)-sheaves/∞-stacks are by now fairly well understood, it is clear that the (∞,2)-categories of (∞,2)-sheaves – such as the codomain fibration/self-indexing of an (∞,1)-topos – will form an (∞,2)-topos in generalization of the non-homotopic notion of 2-topos. And so on.


Here are some introductory words for readers unfamiliar with the general idea. Other readers should skip ahead.

For 1-category theorists

This section assumes that the reader is well familiar with category theory and maybe with strict omega-categories but in need of some introductory words on (,n)(\infty,n)-categories.

Ordinary category theory provides various powerful tools for generating higher order structures, among them notably

  1. enrichment

  2. internalization.

Here we are interested in higher order categories, so we consider Cat itself as a 1-categorical context for either of these procedures. Since Cat naturally a cartesian monoidal category

(𝒱,)(Cat,×) (\mathcal{V}, \otimes) \coloneqq (Cat, \times)

we may form the category of V-enriched categories 𝒱CatCatCat\mathcal{V}Cat \coloneqq Cat Cat. A CatCat-category consists of

  • a collection of objects;

  • for each pair of objects AA, BB a category of morphisms, hence to be thought of as collection of ordinary morphisms ABA \to B together with morphisms between these morphisms: 2-morphisms;

  • such that composition is a functor on these hom-categories.

This is the structure of a strict 2-category. We have that

CatCatStr2Cat. Cat Cat \simeq Str 2 Cat \,.

is the category of strict 2-categories.

By general results of enriched category theory (or by immediate inspection), this is still a cartesian monoidal category and so we may iterate this and consider now the enriching category

(𝒱,)(CatCat,×) (\mathcal{V}, \otimes) \coloneqq (Cat Cat, \times)

and construct again 𝒱Cat\mathcal{V}Cat, which now is

(CatCat)CatStr3Cat (Cat Cat) Cat \simeq Str 3 Cat

the category of strict 3-categories. It continues this way, and so for every nn \in \mathbb{N} the nn-fold iterated enrichment of CatCat is the category

StrnCat(((CatCat)Cat))Cat Str n Cat \simeq (\cdots ((Cat Cat)Cat) \cdots) Cat

of strict n-categories. The inductive limit of this construction finally is the category of strict omega-categories.

While this easily generates higher categorical structures, it does so, as the terminology indicates, only in a very restrictive way: while every 2-category still happens to be equivalent to a strict 2-category, already the general 3-category is no longer equivalent to a strict 3-category, and the discrepancy only increases with nn.

But inspection in the case of 2-categories already shows what the problem is: in a weak 2-category structural relations such as associativity and unitality no longer hold as equations but only up to an invertible 2-morphism, whereas objects in Str2CatCatCatStr 2 Cat \simeq Cat Cat, by definition of enriched category, satisfy these relations strictly – therefore the name.

But this problem directly corresponds to an evident shortcoming of the very starting point of the above recursive construction: that construction regarded Cat as a 1-category in order to fit it into the standard formulation of enriched category theory; however Cat is naturally rather a 2-category itself. The enrichment procedure should be allowed to make use of this extra structure. On the other hand, as we have just seen, the failure of CatCatCat Cat to model all of 2Cat is only in the lack of invertible 2-morphisms. Therefore what should really matter for the improved enrichment is just the (2,1)-category underlying Cat, which is the 2-category consisting of all categories, all functors between them, but only natural isomorphism instead of all natural transformations between those.

This way one does arrive at a suitable refined notion of enrichment over the (2,1)-category CatCat, and interpreted this way one does finds that CatCatCat Cat then indeed produces all of 2Cat.

However, this only fixed the first step of the above recursive definition. In the next step we want (2Cat)Cat(2 Cat)Cat to produce all 3-categories, but their associativity and unitalness now involves invertible coherence 3-morphisms which do not appear in enriched (2,1)(2,1)-category theory. And so on, as the recursion proceeds.

This shows that the natural starting point for a construction of n-categories by recursive enrichment must be a conception of 1-category theory which knows already about invertible k-morphisms for all kk. The notion of category where all 1-categorical operations are relaxed up to invertible higher morphisms is that of (∞,1)-category. And this now turns out to be a good starting point for producing nn-categories by recursive enrichment.

If we then just replace in the above the naive Cat with (∞,1)Cat, then the simple formula

Cat (,n)∶−(((Cat (,1)Cat (,1))Cat (,1)))Cat (,1) Cat_{(\infty,n)} \coloneq (\cdots ((Cat_{(\infty,1)} Cat_{(\infty,1)})Cat_{(\infty,1)}) \cdots) Cat_{(\infty,1)}

does produce a good general notion of nn-categories, these are the (,n)(\infty,n)-categories discussed here.

There is also an alternative road to the same conclusion: another standard procedure for producing higher order structures from the 1-category Cat is to consider internal categories in CatCat. For EE a category with finite limits, write Cat(E)Cat(E) for the category of EE-internal categories, and hence Cat(Cat)Cat(Cat) for the category of CatCat-internal categories.

This gives double categories

DoubleCatCat(Cat) DoubleCat \simeq Cat(Cat)

and hence again not quite the 2-categories that we are after. But it is of interest to note that now there are two problems, not just the one above: while a CatCat-internal category again has strict associativity and unitality, instead of the desired version up to an invertible 2-morphism, in another direction it is more general than a strict 2-category: the latter only corresponds to those special double categories for which the “vertical” and the “horizontal” 1-morphisms come from the same 1-category and have sufficiently many degenerate 2-morphisms between them.

The first problem turns out to be solved as before: instead of working with the 1-category Cat we should already regard that as a (2,1)-category and then formulate internal (2,1)-categories in straightforward generalization of the ordinary notion. For the second problem it turns out that one needs to slightly enhance that straightforward generalization and add a condition known (somewhat undescriptively) as completeness. But if this is understood then (as discussed in detail at internal category in an (∞,1)-category) the simple idea of iterated internalization does work out and we obtain (,n)(\infty,n)-categories by

Cat (,n)Cat((Cat(Cat (,0)))). Cat_{(\infty,n)} \simeq Cat(\cdots(Cat(Cat_{(\infty,0)}))\cdots) \,.

For homotopy theorists

This section assumes that the reader is well familiar with homotopy theory and maybe with (∞,1)-category theory but in need of some introductory words on (,n)(\infty,n)-categories.

A fundamental insight of homotopy theory is, of course, that the cellular shape of simplices naturally serves to model paths and higher homotopies in “spaces”, which here really means: in homotopy types/∞-groupoids. In fact, the simplices see a bit more: since Δ[n]\Delta[n] is naturally identified with the linear category {012n}\{0 \to 1 \to 2 \to \cdots \to n\} on (n+1)(n+1)-objects, there is a direction on the paths which form the 1-skeleton of a map Δ nX\Delta^n \to X.

If XX is a topological space/simplicial set/homotopy type, then this directedness in a way “disappears up to equivalence”, in that for every such directed path there is also the reverse path, which is an inverse up to equivalence.

But it is straightforward to consider a slight generalization of this situation where we take XX to be such that not all paths in it have inverses. Still thinking of XX as a homotopy type this may be thought of as modelling a directed homotopy type. For XX instead modeled as a simplicial set, this has been formalized by the concept of a quasi-category or (∞,1)-category. These are combinatorial models for directed homotopy types in direct generalization of how Kan complexes are combinatorial models for ordinary homotopy types.

As the notation already suggests, the idea of (,n)(\infty,n)-category theory is that this generalization from ∞-groupoids (“(,0\infty,0)-categories”) to (∞,1)-categories is but the first step in a tower of higher generalizations, where in step nn one considers “directed homotopy” up to and including dimension nn.

It is natural that such (,n)(\infty,n)-categories should be probed by corresponding higher dimensional analogs of the objects in the simplex category, the linear categories Δ[n]={012n}\Delta[n] = \{0 \to 1 \to 2 \to \cdots \to n\} that support traditional homotopy theory. There are many such generalizations which one could consider. One which has proven to be useful are the objects in the nnth Theta-category Θ n\Theta_n. Where the linear categories as above arise from gluing – pasting – of cellular intervals, the objects of Θ n\Theta_n arise from pasting of nn-dimensional cellular globes (an interval being a 1-dimensional globe).

Accordingly, just as an ∞-groupoid/homotopy type may be presented by a simplicial set, hence a presheaf on the simplex category – or more generally by a simplicial space– satisfying some (Kan filler-)condition that encodes the existence of composites and inverses, so an (∞,n)-category may be presented by a presheaf of spaces on the nnth Theta-category, similarly subject to some conditions that ensure the existence of composites and inverses – but only of inverses above dimension nn.


There are various different ways of defining (,n)(\infty,n)-categories, which are all natural in their own right, and all equivalent to each other.

There is an axiomatic characterization of the (∞,1)-category of (,n)(\infty,n)-categories by generation from strict n-categories:

Among the more direct definitions of (,n)(\infty,n)-categories one can roughly distinguish two flavors, those that build (,n)(\infty,n)-categories by enrichment over (,n1)(\infty,n-1)-categories

and those that build them by internalization in the collection of (,n1)(\infty,n-1)-categories

Via generation by strict nn-categories

We discuss a characterization of the (∞,1)-category of (,n)(\infty,n)-categories as an (,1)(\infty,1)-category generated by strict n-categories, due to (Barwick, Schommer-Pries).

The blueprint for the following construction is the traditional fact that a category is characterized by the fact that its nerve is a simplicial set which satisfies the Segal conditions, which reflect the existence of composition in a category. Since the simplicial nerve is induced from the linear categories Δ[n]={012n}\Delta[n] = \{0 \to 1 \to 2 \to \cdots \to n\} this can be taken as saying that these linear categories generate Cat, subject to the condition that there exists composites.

The following discussion takes this point of view and generalizes it to a similar presentation of (,n)(\infty,n)-categories by very simple strict n-categories.

Strict nn-categories

The main definition is def. 8 below, which roughly says that the collection of (,n)(\infty,n)-categories is generated from strict n-categories in a certain sense. Therefore we first need to fix some terminology and notions about strict nn-categories and about the relevant notion of generation.


Write StrnCatStr n Cat for the 1-category of strict n-categories.


StrnCat gauntStrnCat Str n Cat_{gaunt} \hookrightarrow Str n Cat

for the full subcategory on the gaunt nn-categories, those nn-categories whose only invertible k-morphisms are the identities.

This subcategory was considered in (Rezk). The term “gaunt” is due to (Barwick, Schommer-Pries). See prop. 11 below for a characterization intrinsic to (,n)(\infty,n)-categories.


For knk \leq n the kk-globe is gaunt, G kStrnCat gauntStrnCatG_k \in Str n Cat_{gaunt} \hookrightarrow \in Str n Cat.


𝔾 nStrnCat gaunt \mathbb{G}_{\leq n} \hookrightarrow Str n Cat_{gaunt}

for the full subcategory of the globe category on the kk-globes for knk \leq n.

Being a subobject of a gaunt nn-category, also the boundary of a globe G kG k\partial G_k \hookrightarrow G_k is gaunt, i.e. the (k1)(k-1)-skeleton of G kG_k.



σ k:Str(k)CatStr(k+1)Cat \sigma_k : Str (k) Cat \to Str (k+1) Cat

for the “categorical suspension” functor which sends a strict kk-category to the object σ(X)Str(k+1)Cat(StrkCat)Cat\sigma(X) \in Str (k+1) Cat \simeq (Str k Cat)Cat which has precisely two objects aa and bb, has σ(C)(a,a)={id a}\sigma(C)(a,a) = \{id_a\}, σ(C)(b,b)={id b}\sigma(C)(b,b) = \{id_b\}, σ(C)(b,a)=\sigma(C)(b,a) = \emptyset and

σ(C)(a,b)=C. \sigma(C)(a,b) = C \,.

We usually suppress the subscript kk and write σ i=σ k+iσ k+1σ k\sigma^i = \sigma_{k+i} \circ \cdots \circ \sigma_{k+1} \circ \sigma_k, etc.


The kk-globe G kG_k is the kk-fold suspension of the 0-globe (the point)

G k=σ k(G 0). G_k = \sigma^k(G_0) \,.

The boundary G k\partial G_k of the kk-globe is the kk-fold suspension of the empty category

G k=σ k(). \partial G_k = \sigma^k(\emptyset) \,.

Accordingly, the boundary inclusion G kG k\partial G_k \hookrightarrow G_k is the kk-fold suspension of the initial morphism G 0\emptyset \to G_0

(G kG k)=σ k(G 0). (\partial G_k \hookrightarrow G_k) = \sigma^k(\emptyset \to G_0) \,.

The category StrnCat gauntStr n Cat_{gaunt} is a locally presentable category and in fact a locally finitely presentable category.

(B-PS, lemma 3.5)

We are going to be interested in a full subcategory StrnCat genStrnCat gauntStr n Cat_{gen} \hookrightarrow Str n Cat_{gaunt}, given below in def. 5, which knows about the higher profunctors/correspondences between nn-categories.


For A,BA,B two categories, a profunctor A op×BSetA^{op} \times B \to Set is equivalently a category over the 1-globe functor, hence a functor

K G 1 =Δ[1] \array{ K \\ \downarrow \\ G_1 & = \Delta[1] }

equipped with an identification AK 0A \simeq K_0 and BK 1B \simeq K_1.

This motivates the following definition.


A kk-profunctor / kk-correspondence of strict nn-categories is a morphism KG kK \to G_k in StrnCatStr n Cat. The category of kk-correspondences is the slice category StrnCat/G kStr n Cat/ G_k.


The categories StrnCat gaunt/G kStr n Cat_{gaunt}/G_k of kk-correspondences between gaunt nn-categories are cartesian closed category.

(B-SP, cor. 5.4)


By standard facts, in a locally presentable category 𝒞\mathcal{C} with finite limits, a slice 𝒞/X\mathcal{C}/X is cartesian closed precisely if pullback along all morphisms f:YXf : Y \to X with codomain XX preserves colimits (see at locally cartesian closed category the section Cartesian closure in terms of base change and dependent product).


Without the restriction that the codomain of ff in the above is a globe, the pullback f *f^* in StrnCatStr n Cat will in general fail to preserves colimits. For a simple example of this, consider the pushout diagram in Cat Cat (,1)\hookrightarrow Cat_{(\infty,1)} given by

Δ[0] δ 1 Δ[1] δ 0 δ 0 Δ[1] δ 2 Δ[2]. \array{ \Delta[0] &\stackrel{\delta_1}{\to}& \Delta[1] \\ {}^{\mathllap{\delta_0}}\downarrow && \downarrow^{\mathrlap{\delta_0}} \\ \Delta[1] &\stackrel{\delta_2}{\to}& \Delta[2] } \,.

Notice that this is indeed also a homotopy pushout/(∞,1)-pushout since, by remark 5, all objects involved are 0-truncated.

Regard this canonically as a pushout diagram in the slice category Cat /Δ[2]Cat_{/\Delta[2]} and consider then the pullback δ 1 *:Cat /Δ[1]Cat /Δ[1]\delta_1^* : Cat_{/\Delta[1]} \to Cat_{/\Delta[1]} along the remaining face δ 1:Δ[1]Δ[2]\delta_1 : \Delta[1] \to \Delta[2]. This yields the diagram

Δ[1], \array{ \emptyset &\stackrel{}{\to}& \emptyset \\ {}^{}\downarrow && \downarrow^{} \\ \emptyset &\stackrel{}{\to}& \Delta[1] } \,,

which evidently no longer is a pushout.

(See also the discussion here).

The definition of Cat (,n)Cat_{(\infty,n)} below, def. 8, will take this property to be one of the characteristic properties. Therefore consider



StrnCat genStrnCat gaunt Str n Cat_{gen} \hookrightarrow Str n Cat_{gaunt}

for the smallest full subcategory that

  1. contains the globe category 𝔾 n\mathbb{G}_{\leq n}, example 1;
  2. is closed under retracts in StrnCat gauntStr n Cat_{gaunt};
  3. has all fiber products over globes (equivalently: such that all slice categories over globes have products).

(B-SP, def. 5.6)


The following categories are naturally full subcategories of StrnCat genStr n Cat_{gen}

This is discussed in more detail below in Presentation by Theta-spaces and by n-fold Segal spaces.


The following pushouts in StrnCatStr n Cat we call the fundamental pushouts

  1. Gluing two kk-globes along their boundary gives the boundary of the (k+1)(k+1)-globle

    G k C k1G kG k+1G_k \coprod_{\partial C_{k-1}} G_k \simeq \partial G_{k+1}
  2. Gluing two kk-globes along an ii-face gives a pasting composition of the two globles

    G k G iG kG_k \coprod_{G_i} G_k
  3. The fiber product of globes along non-degenerate morphisms G i+jG iG_{i+j} \to G_i and G i+kG iG_{i+k} \to G_i is built from gluing of globes by

    G i+j× G iG i+k(G i+j G iG i+k) σ i+1(G j1×G k1)(G i+k G iG i+j) G_{i+j} \times_{G_i} G_{i+k} \simeq (G_{i+j} \coprod_{G_i} G_{i+k}) \coprod_{\sigma^{i+1}(G_{j-1} \times G_{k-1})} (G_{i+k} \coprod_{G_i} G_{i+j})
  4. The interval groupoid (ab)(a \stackrel{\simeq}{\to} b) is obtained by forcing in Δ[3]\Delta[3] the morphisms (02)(0\to 2) and (13)(1 \to 3) to be identities and it is equivalent, as an nn-category, to the 0-globe

    Δ[3] {0,2}{1,3}(Δ[0]Δ[0])G 0 \Delta[3] \coprod_{\{0,2\} \coprod \{1,3\}} (\Delta[0] \coprod \Delta[0]) \stackrel{\sim}{\to} G_0

    and the analog is true for all suspensions of this relation

    σ k(Δ[3]) σ k{0,2}σ k{1,3}(G kG k)G k. \sigma^k(\Delta[3]) \coprod_{\sigma^k\{0,2\} \coprod \sigma^k\{1,3\}} (G_k\coprod G_k) \stackrel{\sim}{\to} G_k \,.

We say a functor ii on StrnCatStr n Cat preserves the fundamental pushouts if it preserves the first three classes of pushouts, and if for the last one the morphism i(σ k(Δ[3])) i(σ k{0,2})i(σ k{1,3})(i(G kG k))i(G k)i(\sigma^k(\Delta[3])) \coprod_{i(\sigma^k\{0,2\}) \coprod i(\sigma^k\{1,3\})} (i(G_k \coprod G_k)) \to i(G_k) is an equivalence.

Generation by strict nn-categories

Def. 8 considers an (,1)(\infty,1)-category generated from StrnCat genStr n Cat_{gen} in the following sense


For 𝒟\mathcal{D} an (∞,1)-category with all small (∞,1)-colimits, say that an (∞,1)-functor

f:𝒞𝒟 f : \mathcal{C} \to \mathcal{D}

strongly generates 𝒟\mathcal{D} if its (,1)(\infty,1)-Yoneda extension on the (∞,1)-category of (∞,1)-presheaves

f:𝒞yPSh (𝒞)Lan y𝒟 f : \mathcal{C} \stackrel{y}{\hookrightarrow} PSh_\infty(\mathcal{C}) \stackrel{Lan_y}{\to} \mathcal{D}

is the reflector of a reflective sub-(∞,1)-category

𝒟Lan yPSh (𝒞). \mathcal{D} \stackrel{\overset{Lan_y}{\leftarrow}}{\hookrightarrow} PSh_\infty(\mathcal{C}) \,.

By definition, a strongly generated (,1)(\infty,1)-category is in particular a presentable (∞,1)-category.


An (,1)(\infty,1)-category of (,n)(\infty,n)-categories Cat (,n)Cat_{(\infty,n)} is an (∞,1)-category equipped with a full and faithful functor

i:StrnCat genτ 0Cat (,n) i : Str n Cat_{gen} \hookrightarrow \tau_{\leq 0}Cat_{(\infty,n)}

from the generating strict nn-categories, def. 5 into its category of 0-truncated objects, such that

  1. StrnCat genτ 0Cat (,n)Cat (,n)Str n Cat_{gen} \to \tau_{\leq 0} Cat_{(\infty,n)} \hookrightarrow Cat_{(\infty,n)} strongly generates 𝒞\mathcal{C};

  2. ii preserves the fundamental pushout relations;

  3. the base change adjoint triple in Cat (,n)Cat_{(\infty,n)} exists along morphisms with codomain a globe;

and such that 𝒞\mathcal{C} is universal with respect to these properties in that for any other j:StrnCat gen𝒞j : Str n Cat_{gen} \hookrightarrow \mathcal{C} satisfying these three conditions it factors through ii

j:StrnCatiCat (,n)L𝒞 j : Str n Cat \stackrel{i}{\to} Cat_{(\infty,n)} \stackrel{L}{\to} \mathcal{C}

by an (∞,1)-functor LL which is the reflector of a reflective inclusion 𝒞Cat (,n)\mathcal{C} \hookrightarrow Cat_{(\infty,n)}.

(B-SP, def. 6.8)


By the first axiom, the localization demanded in the universal property is essentially unique. In particular, therefore, Cat (,n)Cat_{(\infty,n)} is defined uniquely, up to equivalence of (∞,1)-categories. For more on this see prop. 12 below.


The gaunt nn-categories, def. 1 are indeed among the 0-truncated objects: since we are looking at just the (∞,1)-category of (,n)(\infty,n)-categories, instead of more generally the (,n+1)(\infty,n+1)-category the non-invertible transfors between nn-categories are disregarded and so if an object XCat (,n)X \in Cat_{(\infty,n)} has no non-trivial invertible cells, then for every other objeyt YY, the hom-\infty-groupoid Cat (,n)(Y,X)Cat_{(\infty,n)}(Y,X) is 0-truncated, hence is a set.


The first axiom in particular says that Cat (,n)Cat_{(\infty,n)} is a presentable (∞,1)-category, and hence so are all its slices. In view of this the adjoint (∞,1)-functor theorem says that the third condition is equivalent to (∞,1)-pullbacks

f *:Cat (,n)/ i(G k)Cat (,n)/X f^* : Cat_{(\infty,n)}/_{i(G_k)} \to Cat_{(\infty,n)}/X

along morphisms of the form Xi(G k)X \to i(G_k) preserving (∞,1)-colimits.

Universal presentation

By def. 8 Cat (,n)Cat_{(\infty,n)} is equivalent to a localization of the (∞,1)-category of (∞,1)-presheaves on StrnCat genStr n Cat_{gen}. In fact, various subcategories of StrnCat genStr n Cat_{gen} are already sufficient, notable the Theta-category Θ nStrnCat\Theta_n \hookrightarrow Str n Cat (discussed below in \ref{PresentationByThetaSpaces}). Here we discuss these presentations.


Let S 0Mor(PSh (StrnCat gen))S_{0} \subset Mor(PSh_\infty(Str n Cat_{gen})) be the class of morphism generated under fiber product X× G k()X \times_{G_k} (-) with objects XStrnCat genX \in Str n Cat_{gen} over globes by

  1. the morphisms that witness the fundamental pushout relations

  2. the initial morphism i()\emptyset \to i(\emptyset) into presheaf represented by the empty category (which coincides with the initial presheaf on all objects except on the empty category, where it is the singleton).

Write SS for the strongly saturated class of morphisms (see reflective sub-(∞,1)-category) generated by S 0S_0.


The localization of the (∞,1)-category of (∞,1)-presheaves over StrnCat genStr n Cat_{gen}, def. 5 at the class of morphism SS from def. 9 is a presentation of Cat (,n)Cat_{(\infty,n)}, def. 8:

Cat (,n)PSh (StrnCat gen)[S 1]. Cat_{(\infty,n)} \simeq PSh_\infty(Str n Cat_{gen})[S^{-1}] \,.

(B-SP, theorem 7.6).


The three axioms of def. 8 are satisfied effectively by construction of SS (…). Conversely, every localization satisfying the second and third axiom must invert the morphisms in SS, hence must be a sub-localization.


This construction shows that the fundamental pushout relations encode the composition of k-morphisms in an (,n)(\infty,n)-category.

Let XPSh (StrnCat)X \in PSh_\infty(Str n Cat) be some object.

Firts, by the (∞,1)-Yoneda lemma the value of this (∞,1)-presheaf on a strict nn-category CC is the \infty-groupoid of (,n)(\infty,n)-functors CXC \to X, natural equivalences between them, and so on.

And if XX is an SS-local object then it has in particular the property that all the morphisms

Cat (,n)(i(G k) i(G j)i(G k)i(G jk G jG k),X) Cat_{(\infty,n)}( i(G_k) \coprod_{i(G_j)} i(G_k) \to i(G_jk \coprod_{ G_j } G_k) , X )

are equivalences of \infty-groupoids. So by the (∞,1)-Yoneda lemma this is equivalent to

X(G k)× X(G i)X(G k)Cat (,n)(i(G jk G jG k),X) X(G_k) \times_{X(G_i)} X(G_k) \to Cat_{(\infty,n)}(i(G_jk \coprod_{ G_j } G_k), X)

being an equivalence. On the left this is the collection of all those pairs of kk-globes in XX that touch at an ii-boundary. On the right this is the collection of all k-morphisms in XX equipped with a choice of decomposing them into two kk-morphisms touching at an ii-boundary. So the statement that this morphism is an equivalence says that composition of kk-morphisms along ii-boundaries exists in XX.

Various other presentations of Cat (,n)Cat_{(\infty,n)} are obtained by localizations over subcategories of
i:StrnCat restrStrnCat geni : Str n Cat_{restr} \hookrightarrow Str n Cat_{gen} at a set of morphisms TMor(PSh (R))T \subset Mor(PSh_\infty(R)). Write

PSh (StrnCat restr)i *i *i !PSh (StrnCat gen) PSh_\infty(Str n Cat_{restr}) \stackrel{\overset{i_!}{\to}}{\stackrel{\overset{i^*}{\leftarrow}}{\underset{i_*}{\to}}} PSh_\infty(Str n Cat_{gen})

for the induced essential geometric morphism.


The following conditions are sufficient in order that

Cat (,n)PSh (StrnCat gen)[S 1]i *PSh (StrnCat restr)[T 1] Cat_{(\infty,n)} \simeq PSh_\infty(Str n Cat_{gen})[S^{-1}] \stackrel{i^*}{\to} PSh_\infty(Str n Cat_{restr})[T^{-1}]

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

  1. i *(S 0)Ti^*(S_0) \subset T

  2. i !(T 0)Si_!(T_0) \subset S

  3. the counit idi *i !id \to i^* i_! has components in TT;

  4. the kk-globe G kG_k is in the essential image of ii, for each 0kn0 \leq k \leq n.

(B-SP, theorem 9.2)

Presentation by Θ n\Theta_n-spaces and nn-fold complete Segal spaces

We discuss now presentations of Cat (,n)Cat_{(\infty,n)} over subcategories of StrnCat genStr n Cat_{gen}, according to prop. 3.


The nnth Theta category is a full subcategory

Θ nStrnCat gen \Theta_n \hookrightarrow Str n Cat_{gen}

and the localization of PSh (Θ n)PSh_\infty(\Theta_n) that defines the (,1)(\infty,1)-category Θ nSpace\Theta_n Space of (,n)(\infty,n)-Theta-spaces satisfies the conditions of prop. 3.

Hence (,n)(\infty,n)-Theta-spaces are a model for (,n)(\infty,n)-categories, in the sense of def. 8:

Θ nSpaceCat (,n). \Theta_n Space \simeq Cat_{(\infty,n)} \,.

(B-SP, theorem 11.15)

There is a further restriction from the objects of Θ n\Theta_n to nn-fold simplices regarded as grid object, under the canonical embedding

δ n:Δ ×nΔ nΘ n \delta_n : \Delta^{\times n} \to \Delta^{\wr n} \simeq \Theta_n

induced by the identification of the nnthTheta-category (see there) with the nn-fold categorical wreath product of the simplex category with itself.


The inclusion

Δ ×nδ nΘ nStrnCat gen \Delta^{\times n} \stackrel{\delta_n}{\to} \Theta_n \hookrightarrow Str n Cat_{gen}

and the localization of PSh (Δ ×n)PSh_\infty(\Delta^{\times n}) that defines the (,1)(\infty,1)-category CSS(Δ ×n)CSS(\Delta^{\times n}) of n-fold complete Segal spaces satisfies the conditions of prop. 3.

Hence n-fold complete Segal spaces are a model for (,n)(\infty,n)-categories, in the sense of def. 8:

CSS(Δ ×n)Cat (,n). CSS(\Delta^{\times n}) \simeq Cat_{(\infty,n)} \,.

(B-SP, theorem 12.6)


Below in Via ∞-Internalization – Presentation by complete Segal spaces is discussed that nn-fold complete Segal spaces also naturally model an alternative definition of (,n)(\infty,n)-categories by iterated ∞-internalization. Then prop. 5 serves to show that this is equivalent to def. 8 above.

Via \infty-enrichment


There should be a general notion of enriched (∞,1)-category (see there) over a monoidal (∞,1)-category 𝒱\mathcal{V}. Write 𝒱Cat\mathcal{V}Cat for the (∞,1)-category of 𝒱\mathcal{V}-enriched (,1)(\infty,1)-categories.


For nn \in \mathbb{N} write

Cat (,n)(((GrpdCat)Cat))Cat. Cat_{(\infty,n)} \coloneqq (((\infty Grpd Cat) Cat) \cdots) Cat \,.

Presentation by Segal nn-categories

The notion of Segal n-categories is a realization of the idea of weak enrichment in a suitable model category. For nice enough model categories this can be further strictfied to just the notion of enriched model category, discussed below


Presentation by enriched model categories


Via \infty-internalization


There is a general notion of internal category in an (∞,1)-category 𝒞\mathcal{C} provided that

  1. 𝒞\mathcal{C} has finite (∞,1)-limits – in order to formulate the Segal condition;

  2. 𝒞\mathcal{C} is equipped with a “choice of internal ∞-groupoids” – in order to formulate the completeness condition.

We can use this to define Cat (,n)Cat_{(\infty,n)} by iterative internalization.


Write Grpd(Cat (,0))Grpd(Cat_{(\infty,0)}) for the catgeory of groupoid objects in an (∞,1)-category in Cat (,0)Cat_{(\infty,0)} \simeq ∞Grpd.

Assume we have already defined Cat (,n)Cat_{(\infty,n)}, either by one of the methods above, or by the induction in the following. Then the canonical inclusion

Grpd(Cat (,0))PreCat Grpd(Cat (,0))(Cat (,n)) Grpd(Cat_{(\infty,0)}) \hookrightarrow PreCat_{Grpd(Cat_{(\infty,0)})} (Cat_{(\infty,n)})

into the (,1)(\infty,1)-category of simplicial objects X X_\bullet in Cat (,n)Cat_{(\infty,n)} that

  1. satisfy the Segal conditions

  2. such that X 0Cat (,0)X_0 \in Cat_{(\infty,0)}

has a right adjoint (∞,1)-functor CoreCore.

(Lurie, prop. 1.1.14).


An (,n+1)(\infty,n+1)-category is an object XPreCat Grpd(Cat (,0))X \in PreCat_{Grpd(Cat_{(\infty,0)})} such that Core(X)GrpdGrpd(Cat (,0))Core(X) \in \infty Grpd \hookrightarrow Grpd(Cat_{(\infty,0)}).

For n𝒩n \in \mathcal{N} the (,1)(\infty,1)-category of (,n)(\infty,n)-categories is

Cat (,n)Cat n(Grpd)Cat(Cat(Cat (,0))). Cat_{(\infty,n)} \coloneqq Cat^n(\infty Grpd) \coloneqq Cat(\cdots Cat(Cat_{(\infty,0)}) \cdots) \,.

(Lurie, prop. 1.1.14).


The (,1)(\infty,1)-category Cat (,n)Cat_{(\infty,n)} given by def. 12 is equivalent to that given by def. 8.

This is prop. 5 in view of the presentation discussed below.

Presentation by nn-fold complete Segal spaces

By the discussion here at category object in an (∞,1)-category we have


Write cSegal 0sSet QuillencSegal_0 \coloneqq sSet_{Quillen} for the standard model structure on simplicial sets. Then recursively for nn \in \mathbb{N}, n1n \geq 1, there is a model structure on

cSegal n[Δ op,cSegal n1] cSegal_n \coloneqq [\Delta^{op}, cSegal_{n-1}]

which presents Cat n(Grpd)Cat^n(\infty Grpd).


cSegal ncSegal_n is equivalent to the CSS(Δ ×n)CSS(\Delta^{\times n}) from prop. 5.





The (∞,1)-category Cat (,n)Cat_{(\infty,n)} is generated under (∞,1)-colimits from the kk-globes G kG_k for knk \leq n: every object is the (∞,1)-colimit over a diagram of globes.

(B-SP, cor. 8.4)


Equivalences in the (∞,1)-category Cat (,n)Cat_{(\infty,n)} are detected on globes: a morphism f:XYf : X \to Y in Cat (,n)Cat_{(\infty,n)} is an equivalence precisely if for all globes G knG_{k \leq n} the induced morphism on (∞,1)-categorical hom-spaces

Cat (,n)(G k,f):Cat (,n)(Y,f)Cat (,n)(X,f) Cat_{(\infty,n)}(G_k, f) : Cat_{(\infty,n)}(Y, f) \to Cat_{(\infty,n)}(X, f)

is an equivalence of ∞-groupoids.

(B-SP, cor. 8.5)

Truncated objects


The truncated objects in the (∞,1)-category Cat (,n)Cat_{(\infty,n)} are precisely the gaunt strict n-categories

(B-SP, cor. 8.6)


That 0-truncated objects in the Cat (,n)Cat_{(\infty,n)} regarded as an (,1)(\infty,1)-category are gaunt is effectively the definition of 0-truncation in the absence of non-invertibles transfors. That these gaunt (,n)(\infty,n)-categories are then necessarily strict reflects the fact that all the weakening, namely all the associators and unitors as well as all there coherences need to be invertible k-morphisms, and hence must be trivial if there are no non-trivial such.



Let Models (,n)Cat^ (,1)Models_{(\infty,n)} \hookrightarrow \hat Cat_{(\infty,1)} be the core (maximal ∞-groupoid inside) the full sub-(∞,1)-category of (∞,1)Cat on those that satisfy the definition 8.

This is equivalent to

Models (,n)B( 2) n, Models_{(\infty,n)} \simeq B (\mathbb{Z}_2)^n,

the delooping groupoid of the group ( 2) n(\mathbb{Z}_2)^n, the nn-fold product of the group of order 2 with itself.

The nontrivial element σ 2\sigma \in \mathbb{Z}_2 in the kkth slot acts by passing to the kk-opposite (,n)(\infty,n)-category.

(B-SP, theorem 8.13)


This means that

  1. the (,1)(\infty,1)-category Cat (,n)Cat_{(\infty,n)} from def. 8 is uniquely defined, up to equivalence of (∞,1)-categories;

  2. the automorphism ∞-group of Cat (,n)Cat_{(\infty,n)} in Cat^ (,1)\hat Cat_{(\infty,1)} is ( 2) n(\mathbb{Z}_2)^n, hence the only auto-equivalences are given by forming the nn analogs of forming an opposite (∞,1)-category.


The idea is this:

One first observes that StrnCat gauntStr n Cat_{gaunt} from def. 1 has ( 2) ×n(\mathbb{Z}_2)^{\times n} worth of automorphisms, given by reversing the directions of the k-morphisms.

For this,

  1. observe that the identity is the only natural transformation endomorphism on Id:StrnCat gauntStrnCat gauntId : Str n Cat_{gaunt} \to Str n Cat_{gaunt}: this can be checked on globes for which one observes that if a functor G nG nG_n \to G_n is the identity on G n\partial G_n, then it is so also on the unique nn-cell. (B-SP, lemma 4.1)

  2. observe that every autoequivalence of StrnCat gauntStr n Cat_{gaunt} restricts to one on the globe category 𝔾 n\mathbb{G}_n (B-SP, lemma 4.4).

  3. observe that the only autoequivalences of 𝔾 n\mathbb{G}_n are those that reverse the direction of the kk-morphisms for 1kn1 \leq k \neq n, which with the above implies the same for all of StrnCat gauntStr n Cat_{gaunt} (B-SP, lemma 4.5).

Now us that, by the above discussion, StrnCat gauntStr n Cat_{gaunt} generates all of Cat (,n)Cat_{(\infty,n)} under (∞,1)-colimits.

Web of Quillen equivalent model category presentations

We list model category structures that present Cat (,n)Cat_{(\infty,n)} and Quillen equivalences between them.

In the following AA is an model category presenting Cat (,n1)Cat_{(\infty,n-1)} that is an “absolute distributor” in the sense discussed at category object in an (∞,1)-category. (That includes most of the model structures in the table, so that one can recurse over these constructions.)

model categoryQuillen equivalencemodel categorynnLab pageliterature
projective structure for AA-Segal categoriesidentity\stackrel{identity}{\leftrightarrow}injective structure for AA-Segal categories(Lurie, prop. 2.3.9)
projective structure for AA-Segal categoriesinclusion\stackrel{inclusion}{\leftarrow}AA-enriched categoriesLurie, theorem 2.2.16
injective structure for AA-Segal categoriesUnPre\stackrel{UnPre}{\to}complete Segal space objects in AALurie, prop 2.3.1
Theta-(n-1)-space-Segal categoriesTheta-(n-1)-space-enriched categories(Bergner-Rezk, prop. 7.2)


Special cases

In addition,

  • (m,n)-categories can be obtained as particular (,n)(\infty,n)-categories whose kk-cells are trivial for k>mk\gt m.
  • In particular, n-categories = (n,n)(n,n)-categories can be so obtained.

Specific examples

Extra structure and properties

We discuss extra structure that an (∞,n)-category can carry and extra properties that it may enjoy.

𝒪\mathcal{O}-Monoidal (,n)(\infty,n)-categories

(,n)(\infty,n)-Categories with all adjoints


Definition in terms of n-fold complete Segal spaces and Segal n-categories are due to the (unpublished) thesis

  • Clark Barwick, (,n)(\infty,n)-CatCat as a closed model category PhD (2005)

The definition in terms of Theta spaces is due to

An iterartive definition in terms of n-fold complete Segal spaces is given in

A summary of definitions and some known comparison results can be found in

An axiomatic characterization is in

Comparison of models is in

A model for (,n)(\infty,n)-categories in terms of (∞,1)-sheaves on variant of a site of nn-dimensional manifolds with embeddings between them is discussed in

previewed in

This lends itself to a model of (∞,n)-category with adjoints. See there for more.

Revised on October 8, 2013 20:27:46 by Urs Schreiber (