nLab
model structure for quasi-categories

Idea

A quasi-category is a simplicial set satisfying weak Kan filler conditions that make it behave like the nerve of an (∞,1)-category.

There is a model category structure on the category SSet – the Joyal model structure or model structure on quasi-categories – such that the fibrant objects are precisely the quasi-categories and the weak equivalences precisely the correct categorical equivalences that generalize the notion of equivalence of categories.

Definition

Definition

The model structure for quasi-categories or Joyal model structure $sSet_{Joyal}$ on sSet has

• cofibrations are the monomorphisms

• weak equivalences are those maps that are taken by the left adjoint of the homotopy coherent nerve to a weak equivalence in the model structure on simplicial categories.

Properties

As a Cisinski model strucre

The model structure for quasi-categories is the Cisinski model structure on sSet induced by the localizer which consists of the spine inclusions $\{Sp^n \hookrightarrow \Delta^n\}$. See (Ara).

General properties

Proposition

The model structure for quasi-categories is

• a left proper model category,

• a combinatorial model category.

Remark

It is also a monoidal model category and is naturally an enriched model category over itself, hence is $sSet_{Joyal}$-enriched (reflecting the fact that it tends to present an (infinity,2)-category). It is however not $sSet_{Quillen}$-enriched and thus not a “simplicial model category”.

Proposition

For $p \colon \mathcal{C} \to \mathcal{D}$ a morphism of simplicial sets such that $\mathcal{D}$ is a quasi-category. Then $p$ is a fibration in $sSet_{Joyal}$ precisely if

1. it is an inner fibration;

2. it is an “isofibration”: for every equivalence in $\mathcal{D}$ and a lift of its domain through $p$, there is also a lift of the whole equivalence through $p$ to an equivalence in $\mathcal{C}$.

This is due to Joyal. (Lurie, cor. 2.4.6.5).

So ever fibration in $sSet_{Joyal}$ is an inner fibration, but the converse is in general false. A notably exception are the fibrations to the point:

Proposition

The fibrant objects in $sSet_{Joyal}$ are precisely those that are inner fibrant over the point, hence those simplicial sets which are quasi-categories.

(Lurie, theorem 2.4.6.1)

Relation to the model structure for $\infty$-groupoids

The inclusion of (∞,1)-catgeories ∞ Grpd $\stackrel{i}{\hookrightarrow}$ (∞,1)Cat has a left and a right adjoint (∞,1)-functor

$$(grpdfy \dashv i \dashv Core) \;\; : \;\; (\infty,1)Cat \stackrel{\overset{grpdfy}{\to}}{\stackrel{\overset{i}{\leftarrow}}{\overset{Core}{\to}}} \infty Grpd \,,$$

where

• $Core$ is the operation of taking the core, the maximal $\infty$-groupoid inside an $(\infty,1)$-category;

• $grpdfy$ is the operation of groupoidification that freely generates an $\infty$-groupoid on a given $(\infty,1)$-category

(see HTT, around remark 1.2.5.4)

The adjunction $(grpdfy \dashv i)$ is modeled by the left Bousfield localization

$$(Id \dashv Id) \; :\; sSet_{Joyal} \stackrel{\leftarrow}{\to} sSet_{Quillen} \,.$$

Notice that the left derived functor $\mathbb{L} Id : (sSet_{Joyal})^\circ \to (sSet_{Quillen})^\circ$ takes a fibrant object on the left – a quasi-category – then does nothing to it but regarding it now as an object in $sSet_{Quillen}$ and then producing its fibrant replacement there, which is Kan fibrant replacement. This is indeed the operation of groupoidification .

The other adjunction is given by the following

Proposition

There is a Quillen adjunction

$$(k_! \dashv k^!) \;\; : sSet_{Quillen} \stackrel{\overset{k^!}{\leftarrow}}{\overset{k_!}{\to}} sSet_{Joyal}$$

which arises as nerve and realization for the cosimplicial object

$$k : \Delta \to sSet : [n] \mapsto \Delta'[n] \,,$$

where $\Delta^'[n] = N(\{0 \stackrel{\simeq}{\to} 1 \stackrel{\simeq}{\to} \cdots \stackrel{\simeq}{\to} n\})$ is the nerve of the groupoid freely generated from the linear quiver $[n]$.

This means that for $X \in SSet$ we have

• $k^!(X)_n = Hom_{sSet}(\Delta'[n],X)$.

• and $k_!(X)_n = \int^{[k]} X_k \cdot \Delta'[k]$.

This is (JoTi, prop 1.19)

The following proposition shows that $(k_! \dashv k^!)$ is indeed a model for $(i \dashv Core)$:

Proposition

• For any $X \in sSet$ the canonical morphism $X \to k_!(X)$ is an acyclic cofibration in $sSet_{Quillen}$;

• for $X \in sSet$ a quasi-category, the canonical morphism $k^!(X) \to Core(X)$ is an acyclic fibration in $sSet_{Quillen}$.

This is (JoTi, prop 1.20)

Related concepts

A similar model for (∞,n)-categories is disucssed at

• model structure on cellular sets

References

The original construction of the Joyal model structure is in

• Andre Joyal, Theory of quasi-categories I

Unfortunately, this is still not publically available.

A proof that proceeds via homotopy coherent nerve and simplicially enriched categories is given in detail following theorem 2.2.5.1 in

• Jacob Lurie, Higher Topos Theory

The relation to the model structure for complete Segal spaces is in

• Andre Joyal, Myles Tierney, Quasi-categories vs. Segal spaces (arXiv)

Discussion with an eye towards Cisinski model structures and the model structure on cellular sets is in

• Dimitri Ara, Higher quasi-categories vs higher Rezk spaces (arXiv)

See also

• Denis-Charles Cisinski, Algèbre homotopique et catégories supérieures, course notes, 2009, pdf.

A model structure for (infinity,2)-sheaves with values in quasicategories is discussed in

• Nicholas Meadows, The Local Joyal Model Structure (arXiv:1507.08723)