INTRODUCTION TO INFINITY-CATEGORIES

MARKUS LAND

Abstract. These are lecture notes for the lecture courses “Introduction to ∞-categories”
and “Infinity-categories” taught at the university of Regensburg in the winter term 2018/2019
and the summer term 2019. Most of the material is not original and in many occasions I
follow the arguments of Rezk. Further source of inspiration are the standard notes: Lurie’s
Higher Topos theory, Cisinski’s book on higher categories and homotopical algebra, and
Joyal’s paper about quasicategories. Further sources I recommend are Haugseng, Groth.
I want to thank Hoang Kim Nguyn and Denis-Charls Cisinski for many helpful discussions
about the basics of higher category theory and Harry Gindi for many valuable questions and
comments during the lectures. Thanks also to Sebastian Wolf, Paul Bärnreuther, and Jonas
Kleinöder for spotting errors in earlier versions of this script.

Contents
Motivation and overview 2
1. Categories and Simplicial Sets 4
2. ∞-categories 17
3. Anodyne maps and fibrations 40
4. Joins and slices 57
5. Joyal lifting and applications 67
6. Pointwise criterion for natural equivalences 72
7. Fully faithful and essentially surjective functors 82
8. Localizations 93
9. Fat joins, fat slices and mapping spaces 99
10. (Co)Cartesian fibrations 114
11. Marked simplicial sets and marked anodyne maps 122
12. Straightening-Unstraightening 133
13. Terminal and initial objects 141
14. The Yoneda lemma 144
15. Limits and colimits 149
16. Cofinal and coinitial functors 163
17. Adjunctions 169
18. An adjoint functor theorem 180
Appendix A. Exercises 183
References 200

Date: July 21, 2019.

1
2 M. LAND

Motivation and overview

Lecture 1 – 15.10.2018. This lecture should be called “introduction to ∞-categories” as
opposed to “introduction to higher categories”: Let me briefly explain what kind of higher
category we will be talking about:
Informally, an n-category consists of
(1) a set of objects (these will be called 0-morphisms)
(2) a set of morphisms between objects (these will be called 1-morphisms), and
(3) a set of morphisms between (k − 1)-morphisms, (these will be called k-morphisms),
for every k ≤ n.
For every k-object, there should be an identity k + 1-morphism, and there should be compo-
sition maps that “compose” two composable n-morphisms to a new one.
The question is what the axioms are that are supposed to be satisfied. Informally, one defi-
nition could be that we define a 1-category to be an ordinary category, and define inductively
an n-category to be a category enriched in (n − 1)-categories:
For n = 2, we obtain the following examples:
Example 0.1. Consider the category of categories Cat. It is canonically enriched in Cat
itself, because for any two categories C and D, not only is there a set of functors from C to D,
in fact there is a category of functors Fun(C, D) from C to D: The 2-morphisms are precisely
natural transformations of functors.

Example 0.2. Consider the category of groups and group homomorphisms: This is also
canonically a 2-category, namely we say that a 2-morphism from ϕ → ψ : G0 → G is an
element g ∈ G such that ψ = g −1 ϕg. Notice that every 2-morphism is automatically an
isomorphism: the inverse being given by g −1 .

An (n, k)-category is then an n-category in which all `-morphisms are invertible if ` > k
(this can be read as saying that an n-category determines an (n, n)-category).
Example 0.3. Cat is really only a (2, 2)-category, whereas the 2-category of groups described
above is in fact a (2, 1)-category.

What we will consider in our lecture is the notion of (∞, 1)-categories, which will be called
∞-categories.
There are motivating examples:
Example 0.4. A category is a (1, 1)-category and should certainly give rise to an (∞, 1)-
category.

Example 0.5. Let X be a topological space. Then consider the following ∞-category:
(1) objects are the points of X,
(2) 1-morphisms are paths from x to y,
(3) 2-morphisms are homotopies (rel. endpoint) between paths from x to y,
(4) 3-morphisms are homotopies between homotopies between paths, and so on...

There is also another informal construction we can do: Associated to any (n, k)-category
C, and two objects x and y in C, we obtain an (n − 1, k − 1)-category of morphisms between
 INTRODUCTION TO INFINITY-CATEGORIES 3

them. Thus an (∞, 1)-category should have (∞, 0)-categories as hom-objects between any to
spaces.
(∞, 0)-categories deserve to be called ∞-groupoids. Observe now that the ∞-category
associated to a topological space X is in fact an ∞-groupoid.
One of the guiding principles of higher category theory is the Homotopy hypotheses,
which states that topological spaces provide all ∞-groupoids.
Thus in some sense, ∞-categories should give a theory which contains ordinary categories
and topological spaces as special cases. One can make such a formalism very precise, and the
idea is to perform category theory, where one has replaced the category of sets (which is the
basic building block for category theory) with a (to be defined) ∞-category of spaces. What
should come out of such a construction is just the theory of ∞-categories (this approach is
for instance something Cisinski, among others, likes to advertise).
Let us just spend a little time on thinking about the (to be defined) ∞-category of spaces,
and how it should behave. We first consider the following: Let CW be the category whose
objects are CW-complexes and whose maps are simply maps of topological spaces. Associated
to this category we can define new category: the homotopy category of CW-complexes, whose
objects are still CW complexes and whose morphisms are homotopy classes of morphisms of
topological spaces.
There is a canonical functor
CW → h(CW)
which is the initial functor from CW to any other category which sends homotopy equivalences
to isomorphisms. Any “homotopy invariant” functor descends through this functor and thus
one might be tempted to replace CW by h(CW) when dealing with homotopy invariant
notions of spaces.
Here is a problem with this approach: One great feature of topological spaces is that they
are very flexible, i.e. we can manipulate them in many ways: We can build new spaces from
old, by passing to subsets and forming quotients for instance. These constructions are spe-
cial cases of limits and colimits of spaces. A (co)limit is a “universal space” associated to a
diagram I → CW of spaces indexed over a small category I (e.g. pushouts and pullbacks, or
directed colimits, inverse limits). Now it may well happen that there is a natural transfor-
mation between two diagrams, which is a homotopy equivalence on each i ∈ I, but where the
associated map on colimits is not a homotopy equivalence. So, a priori, our flexibility is now
less: what are homotopy theoretic meaningful colimits? There is something like this, and
these are called homotopy colimits.
Unfortunately, homotopy colimits are (a priori) only a method of constructing something
that should behave like a colimit and is invariant under the above transformations, and in
fact they do not satisfy a universal property in the category CW. Maybe they do however in
the category h(CW)? This is also not true, and even worse, although any diagram I → CW
admits a homotopy colimit, the category h(CW) does not admit all colimits (this is a fun
exercise in homotopy theory, I recommend to do it!)
Here is all motivating fact for now that I want to give: The ∞-category of spaces, denoted
by An receives a functor
CW → An → h(CW)
which has the following properties:
(1) the functor CW → An is now initial among all functors CW → D, where D is an
∞-category, which map homotopy equivalences to equivalences,
4 M. LAND

(2) any ∞-category has a homotopy category, and the homotopy category of An is h(CW),
(3) there are ∞-categorical notions of colimits and limits, which have universal properties
as one would like, and An has all of them,
(4) a homotopy (co)limit in CW is mapped to a (co)limit in An.
Informally, this says that if we are interested in studying topological spaces up to homotopy
invariant notions, we are free to replace CW by An by (1), but we do not loose the flexibility
to construct new spaces from old ones, this is part (3), and moreover, the ad hoc construction
of homotopy colimits now actually have a universal property (4). Furthermore, if one liked
the category h(CW) a lot, we still know about this as much as before.
The fact that homotopy categories are typically not well behaved was observed a very long
time ago: This is where model categories are useful: They help us getting a concrete grasp
on the homotopy category. But model categories are still not quite optimal in some aspects:
For instance it is in general quite hopeless, for two model categories M and M0 , to define
a model category structure on the category of functors Fun(M, M0 ). In the context of ∞-
categories, defining this is not too hard (although it is not tautological!).
Replacing model categories with their associated ∞-categories (there is such a construc-
tion), has many technical advantages: There is an ∞-category of ∞-categories, which for
instance has all small (co)limits. This allows to study to which extend certain functors with
values in a category of categories satisfy nice “glueing properties” (are sheaves for some
Grothendieck topology). In principle, one can do the same thing with ordinary categories,
but it just turns out that in many situations the passage to ∞-categories has the advantage
of
(1) allowing very clean formulations of what it means to satisfy nice “glueing properties”
(being a sheaf), and
(2) that sometimes the ∞-categorical statements hold true, whereas the corresponding 1-
or 2-categorical statement is simply not correct.
I suggest, we leave it at that for the moment and just dive into the mathematics!

1. Categories and Simplicial Sets

We work in a model V of ZFC-set theory and assume the large cardinal axiom. This
implies that there exists an inaccessible cardinal κ larger than ℵ0 , which we can use to define
its associated Grothendieck universe U ∈ V. U itself satisfies all axioms of ZFC-set theory
and it is this model of ZFC that we will work in generally. We refer to the sets of U simply
as small sets, and call the elements of V (large) sets.
Definition 1.1. A category C consists of a (possibly large) set of objects ob(C), and for
any two objects x and y a (also possibly large) set HomC (x, y) of morphisms between them,
equipped with composition maps
HomC (x, y) × HomC (y, z) → HomC (x, z)
and identities ∗ → HomC (x, x) for all objects x, satisfying associativity and unitality.
A category is called locally small if all hom sets are small, it is called small if it is locally
small and the set of objects is also small. It is called essentially small if it is locally small and
the set of isomorphism classes of objects are small.

Remark. Think about this set theory business just as saying that a set corresponds to
a small set, and a class corresponds to a set. It is part of the above axiomatic that, for
 INTRODUCTION TO INFINITY-CATEGORIES 5

instance the collection of all vector spaces (which consist of small sets) is itself not small, but
an element of V.
If you do this, then what usually is called a category is now what we call a locally small
category.

Definition 1.2. A partially ordered set is a set P equipped with a reflexive, antisymmetric
and transitive relation ≤. That is, a ≤ a, if a ≤ b and b ≤ a then a = b, and if a ≤ b and
b ≤ c, then a ≤ c. A map of partially ordered sets is a map of sets f : P → Q such that x ≤ y
implies that f (x) ≤ f (y). This defines a category PoSet whose objects are posets an whose
morphisms are maps of posets.

Example 1.3. Finite linearly ordered sets: The set {0, 1, . . . , n} is linearly ordered: 0 ≤
1 ≤ · · · ≤ n and written as [n]. A general finite poset S is called linearly ordered if it is
isomorphic to one of the [n]’s. Morphisms of linearly ordered sets are just morphisms of the
underlying poset. We obtain a category LinOrdSet.

Example 1.4. The subset poset: Let S be a set. Consider its set P(S) of subsets: P(S) =
{I ⊆ S}. This is partially ordered by inclusion: I ≤ J ⇔ I ⊆ J.

Definition 1.5. The category ∆ is the full subcategory of the category PoSet of posets
consisting of the linearly ordered set [n] for all n ≥ 0. Notice that a morphism from [n] to
[m] is thus simply a weakly monotonic map.

Example 1.6. There are special maps in ∆: the face and degeneracy maps: For every n ≥ 0
and 0 ≤ i ≤ n, there are maps
di : [n − 1] → [n]
uniquely determined by the property that i ∈/ Im(di ) and that di is injective. Furthermore,
for n ≥ 1 and 0 ≤ i ≤ n − 1, there are maps

si : [n] → [n − 1]

uniquely determined by the property |s−1

i (i)| = 2 and that si is surjective. I.e. we have
si (i) = si (i + 1) = i.

Definition 1.7. Let C be a category. We denote the category of functors Cop → Set by
P(C) and call it the category of presheaves on C. An object x ∈ C determines a representable
presheaf, namely the presheaf HomC (−, x) which sends y ∈ C to the set of morphisms from y
to x. This determines a functor C → P(C) which is called the Yoneda embedding.

Definition 1.8. A simplicial set is a presheaf on ∆, i.e. a functor ∆op → Set. Given a
simplicial set X, its set of n-simplices is given by X([n]) and will be written Xn . An n-
simplex x is called degenerate if there exists a surjection α : [n] → [m] with m 6= n, and an
n-simplex y such that x = α∗ (y). Equivalently, x is degenerate if x = s∗i (y) for some y ∈ Xn−1
and some 0 ≤ i ≤ n − 1. An n-simplex is called non-degenerate if it is not degenerate.

Definition 1.9. We let ∆n be the simplicial set represented by [n] ∈ ∆. Concretely, we

have (∆n )m = Hom∆ ([m], [n]).
6 M. LAND

Lecture 2 – 18.10.2018.
Definition 1.10. Let X be a simplicial set. We define π0∆ (X) to be the following set.
We consider the equivalence relation ∼ on the set of 0-simplices X0 which is generated by
the relation x ∼ y if and only if there exists a 1-simplex f ∈ X1 such that d0 (f ) = x and
d1 (f ) = y. (This relation is reflexive but in general neither transitive nor symmetric). Then
we let π0∆ (X) = X0 /∼.

Lemma 1.11. The Yoneda lemma: Let F : Cop → Set be a functor and x ∈ C an object.
Then the map
HomP(C) (HomC (−, x), F ) → F (x)
given by sending η to η(idx ) is a bijection.
Proof. The inverse is given by sending an element s ∈ F (x) to the function HomC (y, x) → F (y)
sending f to f ∗ (s). It is an explicit check to see that this is a natural transformation and an
inverse the the above described map. 
Lemma 1.12. The Yoneda embedding C → P(C) is fully faithful.
Proof. This follows immediately from the Yoneda Lemma: The effect of the Yoneda embed-
ding on morphisms is the map
HomC (x, y) → HomP(C) (HomC (−, x), HomC (−, y))
given by sending f to
f∗
HomC (z, x) → HomC (z, y).
We claim that this map is inverse to the map described in the Yoneda lemma, which is given
by sending a map f : HomC (x, y) to the function HomC (z, x) → HomC (z, y) given by sending
ϕ to ϕ∗ (f ) = f∗ ϕ. 
Corollary 1.13. For a simplicial set, there is a canonical bijection
HomsSet (∆n , X) ∼= Xn .
Definition 1.14. A (co)limit of a functor F : I → C is an object of C, written colimI F ,
equipped with maps F (i) → colimI F for every i which are compatible in the sense that for
every morphism i → j in I, the diagram
F (i) colimI F

F (j)
commutes. This datum is required to satisfy the following universal property: Whenever
given a further object X ∈ C, also equipped with maps F (i) → X which are compatible in
the above way, then there exists a unique morphism colimI F → X making the diagrams
F (i) X

colimI F
commute.
 INTRODUCTION TO INFINITY-CATEGORIES 7

Dually, a limit of F is an object limI F , equipped with maps limI F → F (i), which are
again compatible, satisfying the dual universal property: Whenever we are given an object X
equipped with compatible morphisms X → F (i) for all i ∈ I, there exists a unique morphism
X → limI F making the obvious diagram commute.

Remark. Notice that such a universal property specifies an object up to unique isomor-
phism. Notice also that the universal property refers to more than just the object colimI F .
The reference maps are part of the data, and this is what makes the object unique up to
unique isomorphism.

Example 1.15. (1) A colimit of the empty diagram ∅ → C is an initial object: It is an

object which admits a unique morphism to any other object. Dually, A limit of the
empty diagram ∅ → C is a terminal object: It is an object which admits a unique
morphism from any other object.
(2) A colimit of the diagram • ← • → • is called a pushout.
(3) A limit of the diagram • → • ← • is called a pullback.

Example 1.16. The quotient vector space V /U is a pushout of the diagram

U V

0 V /U

Observation 1.17. One can phrase general (co)limits via inital and terminal objects. This
point of view will be used later when we discuss limits and colimits in ∞-categories. Given a
functor F : I → C we can consider the category of (co)cones of this functor. Given a category
I we consider a new category I / and I . , which are constructed from I by adding an initial
respectively a terminal object. There is an obvious functor I → I / and I → I . . We can thus
consider the functor categories
FunF (I / , C) and FunF (I . , C)
of functors which restrict to F along the above mentioned inclusion. These are called the
categories of cones and cocones over F , respectively.

The following lemma is immediate from the definition of (co)limits, and the fact established
in Exercise 1.5 that the category Set is bicomplete (else the statement does not make sense).
Lemma 1.18. Let C be a category and let F : I → C be an I-shaped diagram in C. Then,
for every object x ∈ C, there are canonical bijections
(1) HomC (colimI F, x) =∼ limI HomC (F (i), x), and
∼
(2) HomC (x, limI F ) = limI HomC (x, F (i)).
Moreover, this property characterizes (co)limits uniquely.
Definition 1.19. An adjunction consists of a pair of functors (F : C → D, G : D → C)
together with a natural isomorphism between the two functors Cop × D → Set given by
HomD (F (−), −) and HomC (−, G(−)).
8 M. LAND

Lemma 1.20. Left adjoints preserve colimits, right adjoints preserve limits.

Proof. Let F : C → D be a functor which admits a right adjoint, say G. Let X : I → C be a

diagram which has a colimit colimI X ∈ C. We claim that F sends that colimit to a colimit
of the diagram I → C → D. In formulas, we claim that the canonical map colimI F (X(i)) →
F (colimI X(i)) induced from the compatible maps F (X(i)) → F (colimI X(i)) that are part
of the datum of the colimit (and then applying F ) is an isomorphism. To see this, it suffices
to show that it induces a bijection on hom sets for all other objects y ∈ D:

HomD (F (colim X(i)), y) ∼

= HomC (colim X(i), Gy)
I I
∼
= lim HomC (X(i), Gy)
I
∼
= lim HomD (F (X(i)), y)
I
∼
= HomD (colim F (X(i)), y)
I

so we are done by the Yoneda lemma. The argument for the claim that right adjoints preserve
limits is similar. 

Lemma 1.21. Let F : C → D be a functor which admits right adjoints G and G0 . Then
there is a specified natural isomorphism between G and G0 . (Adjoints, if they exist, are unique
up to unique isomorphism).

Proof. Consider the following two natural bijections

HomC (Gx, G0 x) ∼
= HomD (F Gx, x) ∼
= HomC (Gx, Gx).
Then the identity of Gx corresponds to a natural transformation G → G0 . Applying the same
trick for HomC (G0 x, Gx) shows that this must be a natural isomorphism. 

Definition 1.22. A category is called (co)complete, if it admits (co)limits indexed over

arbitrary small (co)limits. It is called bicomplete if it is both complete and cocomplete.

Lemma 1.23. If C is bicomplete, then (co)lim is left/right adjoint to the constant diagram
functor. In particular, forming (co)limits determines a functor

Fun(I, C) → C.

Proof. Let’s spell out the colimit case. Consider the constant functor const : C → Fun(I, C).
Now we specify, for each functor F : I → C an object, namely colimI F . Part of the datum of
a colimit are compatible maps {F (i) → colimI F }{i∈I} which are easily seen to assemble into
a natural transformation
F → const(colim F ).
I

Then we consider the composite

HomC (colim F, X) → HomFun(I,C) (const(colim F ), constX) → HomFun(I,C) (F, constX)

I I

which is a bijection by the universal property of a colimit. The lemma thus follows from
Exercise 1.9. The case of limits is completely analogous. 
 INTRODUCTION TO INFINITY-CATEGORIES 9

Lecture 3 – 22.10.2018.
Lemma 1.24. Given an adjunction with F : C → D being left adjoint to G : D → C, and
given a further auxiliary small category I, then the functors
F∗ : Fun(I, C) Fun(I, D) : G∗
again form an adjoint pair.
Proof. The adjunction is determined by a counit map ε : F G → idD and a unit map η : idC →
GF that satisfy the snake identities of Exercise 1.8. We now use these to construct counit
and unit maps for the pair of functors (F∗ , G∗ ) as follows: Let ϕ ∈ Fun(I, D). We need to
specify a natural map ε∗ : F∗ (G∗ (ϕ)) → ϕ of functors I → D, so let x ∈ E. We define the new
counit ε∗ to be the map
εϕ(x)
F (G(ϕ(x))) −→ ϕ(x).
It is easy to see that this is natural in ϕ, since ε itself is a natural transformation. Similarly
we define a natural transformation η∗ : ψ → G∗ F∗ (ψ) to be given by
ηψ(y)
ψ(y) −→ G(F (ψ(y))).
It is then easy to see that the snake identities are satisfied, because (ε, η) satisfy the snake
identities. 
Proposition 1.25. Let C be a bicomplete category, then Fun(I, C) is bicomplete as well. A
(co)limit of a diagram X : J → Fun(I, C) is given by the functor sending i ∈ I to colimJ X(j)(i).
Proof. Let us argue that Fun(I, C) is cocomplete. The completeness argument is similar
(or can be formally deduced from this case by applying op correctly). We claim that the
composite
colimJ
Fun(J, Fun(I, C)) ∼
= Fun(I, Fun(J, C)) Fun(I, C)
is a colimit functor we wish to show exists. By Lemma 1.24 this functor has a right given by
const∗ : Fun(I, C) → Fun(I, Fun(J, C)) ∼
= Fun(J, Fun(I, C))
it the proposition is shown once we convince ourselves that this is itself the constant functor
(which is immediate from the definition), as then we allude to Lemma 1.23. 
Corollary 1.26. The category of simplicial sets sSet is bicomplete.
Definition 1.27. Suppose F : C → D is a functor and let x ∈ D be an object. Then the slice
category F/x has as objects pairs (c ∈ C, α : F (c) → x). For x/F the arrow goes the other
direction. Morphisms from (c, α) to (c0 α0 ) are given by a morphism β : c → c0 in C making
the obvious diagram commute.
If F is a subcategory, we also simply write C/x and x/C. For an object (c, α) ∈ C/x we
sometimes simply write c → x.

Lemma 1.28. Every presheaf is a colimit of representables. More precisely, every presheaf
F : Cop → Set satisfies that the tautological map
colim HomC (−, X) → F
X→F
is an isomorphism.
10 M. LAND

Proof. We prove this again by the Yoneda lemma, i.e. we show that this map induces a
bijection on maps to an auxiliary presheaf G. We calculate
HomP(C) (colim HomC (−, X), G) ∼
= lim HomP(C) (HomC (−, X), G)
X→F X→F
∼
= lim G(X)
X→F
and it is not hard to see that the latter is in fact the set of natural transformations from F
to G. 
Let i : C0 ⊆ C be a small subcategory of a category, and let D be a bicomplete category.
Fact 1.29. The restriction functor
i∗ : Fun(C, D) → Fun(C0 , D)
has a left adjoint denoted by i! and a right adjoint denoted by i∗ . They are given as follows
i! (F )(x) = colim F (c)
c∈C0 /x

and
i∗ (F )(y) = lim F (x).
c∈x/C0
Notice that the slices are small by assumption, so that the colimits and limits exist. I will
now not spell out why this formula produces adjoints for the restriction functor i∗ .

Definition 1.30. In the situation above, we call i! (F ) the left Kan extension of F along i
and i∗ (F ) the right Kan extension of F along i.

Observation 1.31. The statement that the tautological map is an isomorphism shows that
the identity of P(C) is left Kan extended from the Yoneda embedding (along the Yoneda
embedding).

Corollary 1.32. If D is a cocomplete category and C is a small category, then the canonical
functor
Funcolim (P(C), D) → Fun(C, D)
obtained by restriction along the Yoneda embedding is an equivalence.
Proof. Given a functor f : C → D, we want to construct a colimit preserving functor fˆ: P(C) →
D, such that fˆ(X) = f (X) for X ∈ C. By Lemma 1.28, given an object F ∈ P(C), we are
forced to define
F̂ (F ) = colim f (X).
X→F
One can check that this is in fact a functor: For F → G a morphism in P(C), there is an
induced functor of the category of X → F , to X → G given by postcomposition with the
given morphism. Then it is not hard to see that taking colimits produces a map
colim f (X) → colim f (X).
X→F X→G
Also, it is not hard to see that this is in fact a functor.
To see that it is colimit preserving, we observe that fˆ admits a right adjoint G given by
the following formula:
G(d)(X) = HomD (f (X), d).
 INTRODUCTION TO INFINITY-CATEGORIES 11

In the proof of Lemma 1.28 we saw that the set of natural transformations between F and
F 0 is given by
HomP(C) (F, F 0 ) = lim F 0 (X).
X→F
We thus see that
HomD (fˆ(F ), d) ∼
= lim HomD (f (X), d) ∼
= lim G(d)(X).
X→F X→F

Hence fˆ is left adjoint to G and thus preserves colimits. 

Corollary 1.33. Let X be a simplicial set. Then
X∼ = colim ∆n .
[n]∈∆/X

Lemma 1.34. Let X be a fixed simplicial set. Then the functor sSet → sSet sending Y to
X × Y admits a right adjoint Hom(X, −) determined by the formula
HomsSet (∆n , Hom(X, Z)) = HomsSet (∆n × X, Z).
Sometimes we will also write Z X for Hom(X, Z).
Proof. Mapping [n] to the set on the right hand side clearly determines a simplicial set which
we call Hom(X, Z). It satisfies the adjunction property on representable simplicial sets, so
we can extend the adjunction to all simplicial sets because every simplicial set is a colimit
of representables. Notice that we use here that the functor X × − : sSet → sSet preserves
colimits: This is certainly true in Set because again the hom set provides a right adjoint. 
Definition 1.35. Let n ≥ 0 be a natural number. Then the topological n-simplex ∆ntop is the
subspace of Rn+1
≥0 consisting of those points whose coordinates add up to 1. The topological
simplices form a cosimplicial space [n] 7→ ∆ntop : The induced maps are the unique affine linear
maps that do what they should on vertices: Precisely given α : [n] → [m], the induced map
α∗ : ∆ntop → ∆m
top is given by
α∗ (t0 , . . . , tn ) = (v0 , . . . , vm )
P
where vi = tj .
j7→i

Definition 1.36. The singular simplicial set of a topological space X is the simplicial set
S(X) = [n] 7→ HomTop (∆ntop , X) .

Definition 1.37. The geometric realization is the unique colimit preserving functor sSet →
Top which sends ∆n to ∆ntop . Concretely, the geometric realization of a simplicial set X is
the following topological space
|X| = colim
n
∆ntop .
∆ →X
An even more concrete formula is given by
a 
|X| = Xn × ∆ntop /((f ∗ (x), t) ∼ (x, f∗ (t))
n∈∆
for x ∈ Xn , t ∈ ∆m
top and f : [m] → [n] a morphism in ∆.

Proposition 1.38. The singular complex is right adjoint to geometric realization.

12 M. LAND

Proof. By definition of adjunctions we need to specify a natural isomorphism of functors

sSetop × Top → Set between
HomTop (|X|, Y ) ∼
= HomsSet (X, S(Y ))
But by the previous work we know that these functors are equivalent to
lim HomTop (∆ntop , Y ) and lim HomsSet (∆n , S(Y ))
∆n →X ∆n →X
and the latter two are already isomorphic (by definition of S(Y )) before forming the limit. 
Definition 1.39. (1) The boundary ∂∆n is the subsimplicial set of ∆n whose k-simplices
consist of the non-surjective maps [k] → [n].
(2) For any subset S ⊆ [n] the S-horn ΛnS ⊆ ∆n consists of those k-simplices f : [k] → [n]
where there exists a i ∈ [n] \ S such that i is not in the image of f . A horn Λnj is
called inner if 0 < j < n and it is called outer if j = 0, n (left, right horn).
(3) The spine I n ⊆ ∆n is given by those k-simplices f : [k] → [n] whose image is either of
the form {j} or {j, j + 1}.

Lecture 4 – 25.10.2018.
Definition 1.40. The n-skeleton skn (X) of a simplicial set X is given by the simplicial set
i! i∗ (X), where i : ∆≤n ⊆ ∆ is the inclusion of the full subcategory on objects of cardinality
≤ n + 1. Dually, the n-coskeleton coskn (X) of a simplicial set is given by the simplicial set
i∗ i∗ (X). This implies that the k-simplices of coskn (X) are given by
coskn (X)k = HomsSet (skn (∆k ), X).
Dually, coskn (X) is the largest simplicial set which contains X and whose k-simplices are the
same as those of X for k ≤ n.

Lemma 1.41. (1) The skeleton skn (X) is the smallest sub simplicial set of X whose set
of k-simplices coincides with the ones of X for k ≤ n.
(2) The functors skn and coskn are left and right adjoint to each other.
(3) There is the formula coskn (X) = HomsSet (skn (∆k ), X).
Proof. (1) It is easy to see that skn (X) is a sub simplicial set of X, and that skn (X)k = Xk
for k ≤ n. Given any other sub simplicial set Z with this property we have skn (X) =
skn (Z) ⊆ Z, so the claim follows.
(2) Obvious, since adjoints compose.
(3) Obvious, by (2).

Lemma 1.42. The geometric realization of the horn is a horn, the geometric realization of
the spine is a spine.
Proof. This follows from the fact that |∆n | = ∆ntop and the following observations.
(1) I n = I n−2 q∆0 ∆1 , and
(2) there is a coequalizer
a a
∆[n]\i ×∆n ∆[n]\j → ∆[n]\i → Λni .
0≤i≤j≤n 0≤i≤n,i6=k

how one obtains horns and spines as pushouts of ∆k ’s. 

 INTRODUCTION TO INFINITY-CATEGORIES 13

Corollary 1.43. The geometric realization of a simplicial set is a CW complex.

Proof. Given as simplicial set X, define a filtration on |X| through |skn (X)| ⊆ |X|. Since
geometric realization commutes with colimits, we see that this is in fact a filtration of |X| and
the the pushouts of above give pushouts of geometric realizations. Then use that |∂∆n | ∼ =
S n−1 and |∆n | ∼
= Dn . 
Observation 1.44. A poset determines a category in the following way: Objects are the
elements of the posets P and for each pair of elements x, y ∈ P we have
(
∗ if x ≤ y
Hom(x, y) =
∅ else.
Furthermore, a functor between categories associated to posets is the same thing as a map
of posets, i.e. a map of sets respecting the partial ordering. This determines a fully faithful
functor PoSet → Cat. It follows that we can view [n] a category. Sending [n] to this category
produces a cosimplicial small category.

Definition 1.45. The nerve of a category C is the simplicial set given by

[n] 7→ Fun([n], C)
i.e. given by taking functors out of the previous cosimplicial category to the given one.

Lemma 1.46. ∆n is isomorphic to the nerve of the category [n].

Proof. Unravelling the definitions we find that
(∆n )m = Hom∆ ([m], [n])
whereas
N([n])m = Fun([m], [n]).
It thus suffices to recall that the functor Posets → Cat is fully faithful. 
Definition 1.47. The classifying space BG of a group G is the geometric realization of the
nerve of the group considered as groupoid with one object.

Definition 1.48. A Kan complex is a simplicial set which has the extension property for
horn inclusions Λnj → ∆n for 0 ≤ j ≤ n.

Lemma 1.49. The singular complex of a topological space is a Kan complex.

Proof. By adjuction, there is an equivalence of lifting problems
Λnj S(X) |Λnj | X

∆n |∆n |
Then we recall that the topological horn inclusion has a retract, so that the right lifting
problem can be solved. 
Fact 1.50. A Kan complex satisfies the extension property for any monomorphism of
simplicial sets K → L which induces a weak equivalence on geometric realizations (these are
called anodyne maps).
14 M. LAND

Lemma 1.51. If the nerve of a category is a Kan complex then the category is a groupoid.
Proof. Because we can lift outer 2-horns, one can easily show that every morphism in C has
a right and a left inverse, so is itself invertible. 
Definition 1.52. Let f, g : X → Y be a map of simplicial sets. We say that f and g are
homotopic if there exists H : X × ∆1 → Y such that H restricts to f and g. Given a pointed
Kan complex (X, x) we define its simplicial homotopy groups to be the
πn∆ (X, x) = [(∆n , ∂∆n ), (X, x)]∗

Fact 1.53. The homotopy relation is in fact an equivalence relation if Y is a Kan complex.
The simplicial homotopy groups of a Kan complex agree with the ordinary homotopy groups
of the geometric realization. In particular, they are groups for n ≥ 1 and abelian groups for
n ≥ 2. For more about simplicial homotopies and simplicial homotopy groups see the book
of Goerss–Jardine, I.6 and I.7.

Lecture 5 – 29.10.2018.
Theorem 1.54. For a simplicial set X, the following three conditions are equivalent.
(1) X has unique extensions for Λnj → ∆n if 0 < j < n.
(2) X has unique extensions for I n → ∆n for n ≥ 2.
(3) X is isomorphic to the nerve of a category.
Proof. We will show that (1) ⇔ (2) ⇔ (3).
To show (3) ⇒ (2), we consider a category C and its nerve N(C). Recall that its n-simplices
are given by Fun([n], C), and thus are given by chains of composable morphisms. Face and
degeneracies are given by composition and inserting identities. In particular, the restriction
along the spine inclusion I n , picks out precisely the morphisms, so that restriction along the
spine induces an bijection between Fun([n], C) and HomsSet (I n , N(C)).
To show (2) ⇒ (3), consider a simplicial set X which has unique liftings against spines.
We define a category C as follows: The objects are given by X0 , the 0-simplices of X. The
morphisms from x to y are given by all 1-simplices f ∈ X1 such that d1 (f ) = x and d0 (f ) = y.
Identities are given by s0 (x).
We need to explain how to compose morphisms: Two composable morphisms determine a
map I 2 = Λ21 → X which we can extend over ∆2 and restrict to the new edge. We claim that
this indeed is a category: We have to check that identities are such and that composition is
associative (both follow from uniqueness).
(a) s0 (f ) : ∆2 → X has d0 s0 (f ) = f = d1 s0 (f ) and d2 s0 (f ) = s0 d1 (f ) = s0 (x). Similarly,
s1 (f ) : ∆2 → X is a 2-simplex whitnessing that idy ◦ f = f .
(b) Let f, g, h be composable 1-simplices. Consider the associated map I 3 → X. It can
be uniquely filled to a map Φ : ∆3 → X. The restriction of this map to ∆{0,2} is gf .
The 2-simplex d2 (Φ) is thus a composition of gf and h, i.e. Φ|∆{0,3} = h ◦ gf . On the
other hand, the 2-simplex d0 (Φ) gives hg, and thus the 2-simplex d1 (Φ) gives hg ◦ f .
Hence
h ◦ gf = Φ∆{0,3} = hg ◦ f
and associativity of composition is shown.
We claim that there is a preferred map X → N(C) given by the following construction: A
map ∆n → X can be restricted along the spine and thus determines a sequence of composable
 INTRODUCTION TO INFINITY-CATEGORIES 15

morphisms of C, so that we obtain a map I n → N(C). This map can be (uniquely) extended
over ∆n and thus gives an association mapping n-simplices of X to n-simplices of N(C). Using
that an n-simplex of X is determined by its restriction to the spine we now easily see that
the map X → N(C) satisfies:
(a) It is an isomorphism on 0- and 1-simplices,
(b) There is a commutative diagram

Xn N(C)n

X1 ×X0 · · · ×X0 X1 N(C)1 ×N(C)0 · · · ×N(C)0 N(C)1

where the vertical maps are bijections and where the lower fibre product is over the
source and target maps and has n-many factors on both sides.
By (a) the lower map is a bijection, so the upper horizontal map is a bijection as well. This
shows that the map X → N(C) is an isomorphism of simplicial sets, which shows (2) ⇒ (3).
We now show (1) ⇒ (2). We prove this via induction over n. For n = 2 this is clear, as the
2-spine is the inner 2-horn. So we may assume that one can uniquely lift maps I k → X to ∆k
for all k strictly smaller than n and consider a map I n → X which we wish to show extends
uniquely to ∆n . We will show that it extends uniquely to Λnj for some 0 ≤ j ≤ n, and then use
(1) to deduce the claim. We first observe that I n ∩ ∆n\{n} is the spine I n−1 of this simplex,
and likewise that I n ∩ ∆n\{0} is also the spine. Thus, by the inductive hypothesis, there is
are unique maps ∆n\{ε} → X extending the map from the spine to X for ε = 0, n. Since
the intersection of these two faces is given by ∆n\{0,n} , which intersects the spine again in
a smaller spine, these two extensions agree on this intersection, by the inductive hypothesis.
We hence obtain a map
I n ∪ ∆n\{0} ∪ ∆n\{n} = ∆n\{0} ∪ ∆n\{n} → X
where the union is in ∆n . We claim that there exists a unique extension to the union
∆n\{0} ∪ ∆n\{n} ∪ ∆n\{1} .
For this we claim that ∆n\{0} ∪ ∆n\{n} contains the spine of ∆n\{1} : The edges from i → i + 1
for 2 ≤ i ≤ n − 1 all lie in ∆n\{0} , and the edge from 0 → 2 lies in ∆n\{n} because n ≥ 3.
Hence there is a unique map from ∆n\{1} → X extending this map on the spine. We need to
argue that it agrees with the given one on
∆n\{0} ∪ ∆n\{n} ∩ ∆n\{1} = ∆n\{0,1} ∪ ∆n\{1,n} .

On both of these simplices, the map is determined by its restriction to the spine which shows
the claim. Inductively, we find that there exists a unique extension of the map in question to
a map Λnn−1 → X. This can now uniquely be extended to ∆n by assumption.
To see that (2) ⇒ (1), we consider an extension problem β : Λni → X which we want
to show to uniquely extend to ∆n . Clearly we may assume that n ≥ 3, because I 2 = Λ21 .
By assumption (and an exercise) there is an inclusion I n → Λni and we can consider the
restricted extension problem. This can be solved uniquely by assumption, so that we obtain
a map α : ∆n → X. This map can in turn be restricted to Λni , and we want to show that this
map is given by β. To do so, we may restrict to the faces of Λni , i.e. to the union of ∆n\{j}
16 M. LAND

for j 6= i. It is easy to see that

α|∆n\{0} = β|∆n\{0}
because the spine of that simplex is given by a subset of the big spine, and by its very
definition α|I n = β|I n . The same holds for
α|∆n\{n} = β|∆n\{n}
provided β is defined there. We need to show that
α|∆n\{j} = β|∆n\{j}
and may assume that j 6= 0, n. For this, we show that α|∆{j−1,j+1} = β|∆{j−1,j+1} (again, all
the other edges of the spine are contained in the big spine already). Since n ≥ 3 this edge is
contained in ∆n\{ε} for ε either 0 or 1.
Then we induct and see that this determines the map from Λni and thus that we are
done. 
One can in fact say slightly more:
Lemma 1.55. If n ≥ 4 and 0 ≤ j ≤ n and C is an ordinary category, then every lifting
problem
Λnj N(C)

∆n
can be solved uniquely.
Proof. It is an exercise to show that nerves of categories are 2-coskeletal, and we give an-
other argument in Corollary 2.21. It hence suffices to recall that sk2 (Λnj ) → sk2 (∆2 ) is an
isomorphism for n ≥ 4 and all 0 ≤ j ≤ n. 
Lemma 1.56. The nerve of a category C is a Kan complex if and only if C is a groupoid.
Proof. We have seen already the direction that N(C) being a Kan complex implies that C is a
groupoid, Lemma 1.51. So we need to show the other direction. Let us thus assume that C is
a groupoid, and let us show that N(C) is a Kan complex. By Theorem 1.54 we know already
that we can (uniquely) lift all inner horns and all horns of dimension greater or equal to 4.
So we need to prove that we can lift outer 2-horns and outer 3-horns. By passing to opposite
categories it suffices to show that every extension problem
Λn0 N(C)

∆n
has a (unique) solution for n = 2, 3. For n = 2, such a map is given by two maps f : x → y
and g : x → z. We can then choose f ◦ g −1 for the other edge. To show the claim for the left
outer 3-horn, we consider the restriction along the spine and obtain three composable maps
f , g, and h. We find an extension to ∆3 precisely if the edge ∆{1,3} → Λ30 → C is given by the
composite hg. Considering the 2-simplex ∆{0,1,3} → Λ30 → C, we find that this edge satisfies
that precomposition with f is given by hgf . Since f is an isomorphism, the claim follows. 
 INTRODUCTION TO INFINITY-CATEGORIES 17

2. ∞-categories
Definition 2.1. A composer is a simplicial set which has the extension property for spine
inclusions I n → ∆n .

Definition 2.2. In a composer (in fact in a general simplicial set) we call 0-simplices
objects, and a 1-simplex f is a called a morphism from d1 (f ) (the source) to d0 (f ) (the
target). We define the identity morphism of an object x to be s0 (x). For composers, we
define a composition of n-composable morphisms to be a choice of an extension to a ∆n ,
sometimes also just the restriction to the edge ∆{0,n} ⊆ ∆n .

The name composer thus comes from the fact that one can compose morphisms. To avoid
associativity questions, a composer is equipped with an “n-ary” composition law.

Example 2.3. The singular set of a topological space is a composer: Objects are the points,
morphisms from x to y are paths. A composition of morphisms is any path which is homotopic
relative endpoints to the concatenation of the paths.

Definition 2.4. Let X be a simplicial set. We call two 1-simplices f and g from x to y
equivalent if there exists a 2-simplex σ : ∆2 → X which satisfies the following:

(1) σ|∆{0,1} = f ,
(2) σ|∆{0,2} = g, and
(3) σ|∆{1,2} = idy

Observation 2.5. This relation is obviously reflexive, but again a priori neither transitive
nor symmetric. Can you find a further lifting criterion for a composer so that this becomes
in fact an equivalence relation? The inner 3-horn lifting condition.

Lecture 6 – 05.11.2018.

Definition 2.6. Let X be a simplicial set. We define a category hX by means of generators

and relations: The objects are given by X0 . Morphisms are generated by X1 , i.e. for every
1-simplex f : ∆1 → X there is a morphism from d1 (f ) to d0 (f ). The free composites will be
denoted by f ? g. Now we start imposing relations:

(1) The 1-simplex s0 (x) is the identity of x,

(2) for every 2-simplex σ : ∆2 → X with boundary given by a triple (f, g, h) we impose
the relation that h = g ? f ,
(3) if f ∼ f 0 , then f ? g ∼ f 0 ? g and g 0 ? f ∼ g 0 ? f 0 .

The category hX is called the homotopy category of X.

Remark. This construction is obviously functorial, i.e. we have a functor h : sSet → Cat.

Lemma 2.7. Let f : X → Y be a map of simplicial sets which induces an isomorphism

sk2 (X) → sk2 (Y ). Then the induced map hX → hY is an isomorphism of categories.
18 M. LAND

Proof. The whole construction only referred to the 2-skeleton of X. In other words, the
evident map h(sk2 (X)) → hX is an isomorphism. Then we use that the diagram

∼
=
h(sk2 X) h(sk2 Y )
∼
= ∼
=

hX hY

commutes. 

Observation 2.8. In general, the morphisms in hX are formal composites of 1-simplices

(with correct source and target). If X is a composer, we see that the set of morphisms of hX
is a quotient of the set of 1-simplices, and that any two composites of the same two morphisms
will be identified. In particular, equivalent morphisms are identified: If we have a 2-simplex

y
f idy

g
x y

we find that g ∼ idy ? f ∼ f .

Lemma 2.9. Suppose that a composer X has in addition the lifting property with respect to
inner 3-horn inclusions. Let f and g be composable 1-simplices in X. Then
(1) There exists a composite of f and g,
(2) The relation “equivalence” of morphisms in the sense of Definition 2.4 is an equiva-
lence relation,
(3) Any two composites of f and g are equivalent in the sense of Definition 2.4, and
(4) given a 2-simplex σ with σ∆{0,1} = idx , σ∆{1,2} = h and σ∆{0,2} = h0 , then h0 ∼ h.

Proof. (1) follows from the definition of a composer. For (2), we need to prove symmetry
and transitivity. Let us first prove symmetry. So let f, g : x → y be morphisms with f ∼ g.
Pick a 2-simplex σ with σ∆{0,1} = f , σ∆{0,2} = g, and σ∆{1,2} = idy . Together with s0 (f ) and
s0 (idy ), this determines a map Λ31 → X:

y
idy

f
idy y
f
idy

x g y

Since X has the extension property for inner 3-horns, there exists an extension to ∆3 , which
can be restricted to the face ∆{0,2,3} . This 2-simplex witnesses that g ∼ f .
 INTRODUCTION TO INFINITY-CATEGORIES 19

To show transitivity, suppose that f ∼ g ∼ h. Pick 2-simplices σ and σ 0 witnessing these

relations. Together with s0 (idy ), these define a map Λ32 → X:

y
idy

h
idy y
f
idy

x g y

Extending to ∆3 and restricting then to ∆{0,1,3} shows that f ∼ h. To prove (3), let f : x → y
and g : y → z be composable morphisms. Choosing compositions h and h0 , together with
s0 (g) determines a map Λ31 → X:

z
g

h0
idy y
f
g

x h
z

Extending to ∆3 and then restricting to ∆{0,1,3} shows that h ∼ h0 . To prove (4), we consider
the map Λ32 → X given by the diagram

y
h

h
idy
x
idx
h
x y
h0

and extend to ∆3 . Restricting the result to ∆{0,2,3} shows that h0 ∼ h. 

Lemma 2.10. Suppose X is a composer with the inner 3-horn extension property. Then
there is a category π(X) with objects given by 0-simplices of X and morphisms given by
equivalence classes (in the sense of Definition 2.4) of 1-simplices in X. Composition is defined
via lifting along I 2 → ∆2 . The uniqueness of composition (up to equivalence) shows that
composition in π(X) is associative.

Proof. It remains only to prove that composition is associative: For this, it suffices to show
that if f, g, h are composable morphisms, then h ◦ (gf ) is a composition of hg and f . Consider
20 M. LAND

the map Λ31 → X given by

z
hg

(hg)f
h y
f
g

x gf
z

Extending this to ∆3 and then restricting to ∆{0,2,3} shows that (hg)f is a composition of gf
and h. Since composition in X is unique up to equivalence, associativity of composition in
π(X) follows. 
Corollary 2.11. Let X be a composer which furthermore has the inner 3-horn lifting
property. Then hX is isomorphic to π(X). In particular, for composers with the additional
inner 3-horn lifting condition, there is a very explicit description of the homotopy category of
X.
Proof. There is a canonical functor hX → π(X) constructed as follows: It is the identity on
objects and induced by the identity on 1-simplices. Since all relations imposed in hX are
fulfilled in π(X) this in fact descends to a functor as needed. It suffices to prove now that for
any two objects x, y ∈ X, the canonical map
HomhX (x, y) → Homπ(X) (x, y)
is a bijection. To show this, we observe that there is a commutative diagram
X1 (x, y)

HomhX (x, y) Homπ(X) (x, y)

where X1 (x, y) denotes the set of 1-simplices f with d1 (f ) = x and d0 (f ) = y. Both maps
from X1 (x, y) are surjective, thanks to the definition of π and Observation 2.8. Thus the
horizontal map is surjective. It thus remains to show that it is injective as well. Hence
assume given two morphisms f, g ∈ X1 (x, y) with the same image in π(X). This means that
they are equivalent in the sense of Definition 2.4. But again by Observation 2.8, we know
that then f ∼ g. 
Remark. We observe that in the above corollary, X needed not really be a composer: Only
a composition of 2 composable morphisms was needed, we will call a simplicial set which
has the extension property for the 2-spine (which is the inner 2-horn) a weak composer. To
compose many morphisms at the same time, we compose inductively and obtain a well-defined
“n-fold composition” up to equivalence (provided the weak composer satsifies the extension
property for inner 3-horns as well). Now notice that I 2 = Λ21 . Thus we can reformulate the
above as

Corollary 2.12. Let X be a simplicial set which admits liftings for inner 2- and 3-horns.
Then hX is isomorphic to π(X).
Corollary 2.13. Let C be a category. Then h(N(C)) is canonically isomorphic to C.
 INTRODUCTION TO INFINITY-CATEGORIES 21

Proof. We have seen that N(C) admits (unique) lifts for many horns, including the ones
described in the previous corollary. It hence suffices to prove that π(N(C)) ∼
= C. But we recall
that the relation of “equivalence” for morphisms in N(C) ist the relation of “being equal”. 
Definition 2.14. A simplicial set is called an ∞-category if it has the extension property
for all inner horn inclusions Λnj → ∆n , n ≥ 2, 0 < j < n.

Definition 2.15. A functor between two ∞-categories is just a map of simplicial sets. I.e.
the category of ∞-categories is the full subcategory of sSet on objects which are ∞-categories.

It will take some time, but we will see that, informally, an ∞-category is a composer in
which the choice of a composition is unique up to a contractible space of choices: For each
pair of composable morphisms in a composer X, there is a simplicial set of compositions of f
and g, CompX (f, g), given by the pullback

CompX (f, g) Hom(∆2 , X)

.
(f,g)
∆0 Hom(Λ21 , X)
It will be an exercise next week that the inner 3-horn lifting condition tells us that CompX (f, g)
is connected (i.e. its π0∆ (−) vanishes). If we now demand that X in fact has the extension
property for inner horn inclusions for all Λnj → ∆n , we obtain the following two facts (which
we will prove later!)
(1) The simplicial set CompX (f, g) is a Kan complex,
(2) all simplicial homotopy groups πn∆ (CompX (f, g)) vanish.
Thus, an ∞-category is a composer in which composition is well-defined up to a contractible
space of choices.
Example 2.16. A Kan complex, and thus the singular set S(X) of a space X is an ∞-
category. Also, the nerve of any category is an ∞-category.

To make sure that we were not just smart enough we state the following proposition. The
proof will be given later, see Proposition 3.17 and Proposition 3.10.
Proposition 2.17. Every ∞-category is a composer. However, there exist composers which
are not ∞-categories.
Proposition 2.18. The pair of functors
h : sSet Cat : N
are adjoint, N being the right adjoint to h.
Proof. We need to specify unit and counit transformations: The counit is the isomorphism
h(N(C)) ∼ =C
of Corollary 2.13. To construct a natural map X → N(hX) (the unit of the adjunction) we
find that there are canonical maps on 0 and 1-simplices: Recall that the objects of hX are
the 0-simplices of X and that there is a canonical map from X1 to the morphisms of hX
(this map is a surjection if X is a composer). To construct the map on general n-simplices,
22 M. LAND

we observe that an n-simplex ∆n → N(hX) is uniquely determined by its restriction to the

spine, in other words consider the diagram

HomsSet (∆n , X) HomsSet (∆n , N(hX))

∼
=

HomsSet (I n , X) HomsSet (I n , N(hX))

But then simply use the bijection

HomsSet (I n , X) ' X1 ×X0 · · · ×X0 X1
and the corresponding one for N(hX) and the previous observation to obtain a map for general
n-simplices. Checking that these are in fact unit and counit of an adjunction is easy.
Alternatively, you can also show that the map induced by h and the above counit
HomsSet (X, N(C)) → Fun(hX, C)
is a bijection. This follows immediately from the definitions. 

Lecture 7 – 8.11.2018.
Lemma 2.19. Let (F, G, ε, η) be an adjunction with F : C → D the left adjoint, G the right
adjoint, ε : F G → id the counit and η : id → GF the unit. Then
(1) G is fully faithful if and only if ε is an isomorphism,
(2) F is fully faithful if and only if η is an isomorphism.
Proof. Consider the diagram
G F
Hom(X, Y ) Hom(GX, GY ) Hom(F GX, F GY )
τ
∼ (εY )∗
=
ε∗X Hom(F GX, Y )

The right triangle commutes by one of the first exercises (this is how adjunctions with a
binatural isomorphism τ are translated into adjunctions using unit and counits). We now
claim that the big diagram also commutes: Spelling this out, we need to see that for every
morphism f : X → Y , the diagram
F G(f )
F GX F GY
εX εY
f
X Y
commutes. This is true since ε : F G → id is a natural transformation of functors. Hence we
see that G is fully faithful if and only if for all X, Y ∈ D, the map ε∗X is an isomorphism. By
the Yoneda lemma, this is the case if and only if εX itself is an isomorphism for all X, which
shows claim (1). The argument for (2) is similar. 
Corollary 2.20. The nerve functor is in addition fully faithful. Its essential image is
described by Proposition 1.54.
 INTRODUCTION TO INFINITY-CATEGORIES 23

Proof. We proved in Proposition 2.18 that N is a right adjoint to h by constructing explicit

unit and counit maps. Thus, by Lemma 2.19 it suffices to see that the counit of this adjunction
is an isomorphism. But by construction, the counit is given by the canonical isomorphism
h(N(C)) → C. 
Let us record also the following consequence, which was a previous exercise:
Corollary 2.21. The nerve of a category is 2-coskeletal.
Proof. There is a commutative diagram
Fun(hX, C) HomsSet (X, N(C))

Fun(h(sk2 X), C) HomsSet (sk2 X, N(C)) HomsSet (X, cosk2 N(C))

where the diagonal arrow is induced by the canonical map N(C) → cosk2 N(C). By Propo-
sition 2.18, the two left horizontal maps are bijections, and by Lemma 2.7 the remaining
horizontal arrow is a bijection as well. Thus the vertical map is a bijection if and only if the
diagonal map is a bijection. Since X is arbitrary, the claim follows from Yoneda’s lemma. 
Definition 2.22. A morphism in an ∞-category is called an equivalence if its image in the
homotopy category is an isomorphism.

Lemma 2.23. A morphism f : x → y in an ∞-category C is an equivalence if and only if

there exist 2-simplices σ l : ∆2 → C and σ r : ∆2 → C such that
l r
σ|Λ 2 = (f, id) and σ|Λ2 = (f, id).
0 2

ff Draw pictures here:

Proof. If a 2-simplex σ l exists, then σ|∆
l
{1,2} is a left inverse of the image of f in hC. Similarly,
r
σ|∆{0,1} is a right inverse of the image of f in hC. For the converse, suppose that the image
of f is an equivalence in hC. This means that there exists a 1-simplex g : y → x such that
[f g] and [gf ] are the identity in hC, i.e. that there is a 2-simplex η which witnesses that h
is a composite of f and g, and that there is a further 2-simplex η 0 which witnesses that h is
equivalent to the identity. We can use these two 2-simplices (plus an degenerate 2-simplex on
g) to obtain a map Λ32 → C which can be filled because C is an ∞-category:
x
g

idx
idx y
f
g

x h
x

Restricting the resulting 3-simplex to the 2-simplex ∆{0,1,3} produces a 2-simplex σ l . The
argument for σ r is analogous. 
Definition 2.24. An ∞-category is called ∞-groupoid if every morphism is an equivalence.
24 M. LAND

Definition 2.25. The maximal sub-groupoid of an ordinary category is the subcategory

consisting of all isomorphisms. We denote this by C' ⊆ C. For an ∞-category C, we define
the maximal sub-∞-groupoid to be the pullback
C' C

N(hC' ) N(hC)

Remark. For an ordinary category C, we have that N(C' ) = N(C)' . This is because, by
definition there is a pullback
N(C)' N(C)

N(h(N(C))' ) N(h(N(C)))
∼
= ∼
=

N(C' ) N(C)
where the right vertical map composite is the identity, and the lower vertical maps are induced
by the canonical isomorphism h(N(C)) ∼ = C.

Lemma 2.26. An n-simplex x of an ∞-category C belongs to the maximal sub-∞-groupoid

if and only if all edges are equivalences.
Proof. It suffices to observe that an n-simplex in N(hC) is determined by its restriction to all
edges, and that this n-simplex lies in the N(hC' ) if and only if all edges are isomorphisms in
hC. 
Corollary 2.27. The maximal sub-∞-groupoid of an ∞-category is in fact an ∞-groupoid,
and it is the largest such which sits inside the given ∞-category.
Proof. Let us first prove that C' is an ∞-category. For this, consider a lifting problem
Λnj C' ⊆ C

∆n
for 0 < j < n. Since C is an ∞-category, this problem can be solved in C. The claim follows
if we can prove that if we are given a map ∆n → C which induces a map
Λnj → ∆n → C → N(hC)
having image contained in N(hC' ), then already the map ∆n → N(hC) factors through
N(hC' ). For this we again recall that an n-simplex of the nerve of a category is determined
by its restriction to the spine. Since the spine includes in the inner horns Λnj for all n, the
claim follows. This formalizes that C' is the sub-∞-category consisting of the equivalences
of C. In particular, it follows that h(C' ) = (hC)' , so that C' is an ∞-groupoid. It is then
clear that C' is the largest such sitting inside C. 
Lemma 2.28. A Kan complex X is an ∞-groupoid.
 INTRODUCTION TO INFINITY-CATEGORIES 25

Proof. It suffices to show that for every morphism in X, there is a left and a right inverse
σ l and σ r as in Lemma 2.23. But since X is a Kan complex, the maps (f, id) : Λ20 → X and
(f, id) : Λ22 → X can be filled to 2-simplices σ l and σ r . 

The converse is also true – this is a non-trivial and very important theorem in higher
categories. Proving it will be what we aim at next. But before that, we will spend some time
on more examples of ∞-categories, i.e. how can we produce ∞-categories.
Definition 2.29. Let V be a category. A monoidal structure on V consists of the following
data:
(1) A functor − ⊗ − : V × V → V , called the monoidal product,
(2) a unit object 1 ∈ V , together with natural isomorphisms ηl : X → 1 ⊗ X, and
ηr : X → X ⊗ 1, called left unit and right unit, and
(3) natural associativity isomorphisms αX,Y,Z : (X ⊗ Y ) ⊗ Z → X ⊗ (Y ⊗ Z).
These data are required to satisfy some axioms...

Example 2.30. Given a category C which admits finite products, then there is a cartesian
monoidal structure given by the product bifunctor (X, Y ) 7→ X × Y . The unit is given by
the terminal object (a product over the empty set). Dually, a category with finite coproducts
admits a cocartesian monoidal structure with (X, Y ) 7→ X q Y . The unit is given by the
initial object (the coproduct over the empty set).
Explicit examples we will care about are
(1) The category (Set, ×, ∗),
(2) the category (Cat, ×, [0]), and
(3) the category (sSet, ×, ∆0 ).

Definition 2.31. Let (V, ⊗V , 1V ) and (W, ⊗W , 1W ) be monoidal categories. A lax monoidal
functor consists of the following data
(1) a functor F : V → W ,
(2) a natural map 1W → F (1V ), and
(3) a natural map F X ⊗W F Y → F (X ⊗V Y ).
Dually, an oplax monoidal functor consists of the following data
(1) a functor F : V → W ,
(2) a natural map F (1V ) → 1W , and
(3) a natural map F (X ⊗V Y ) → F X ⊗W F Y .
The structure morphisms have to satisfy compatibility with respect to the associativity iso-
morphisms und left/right unit isomorphisms. Precisely, the following diagrams are required
to commute:
(F X ⊗ F Y ) ⊗ F Z F (X ⊗ Y ) ⊗ F Z
αW

F X ⊗ (F Y ⊗ F Z) F ((X ⊗ Y ) ⊗ Z)
F (αV )

F X ⊗ F (Y ⊗ Z) F (X ⊗ (Y ⊗ Z))
26 M. LAND

where αW and αV are the associativity isomorphism in (W, ⊗, 1) and αV and (V, ⊗, 1).
Likewise for the units:
F (ηrV )
FX F (X ⊗ 1)
ηrW

FX ⊗ 1 F X ⊗ F (1)

where ηrV and ηrW are the right unit isomorphisms of V and W . Similarly, the diagram
involving left units is required to commute as well.
A lax monoidal (or oplax monoidal) functor is called monoidal if the natural maps of (2)
and (3) are isomorphisms.

Lecture 8 – 12.11.2018.
Definition 2.32. Let F and G be lax monoidal functors between monoidal categories V
and W . A natural transformation τ : F → G is called lax monoidal if for all X, Y ∈ V , the
diagrams
FX ⊗ FY F (X ⊗ Y )

GX ⊗ GY G(X ⊗ Y )
and
1

F (1) G(1)

commute. We let Funlax (V, W ) be the category whose objects are the lax monoidal functors
and whose morphisms are lax monoidal transformations. We let
Fun⊗ (V, W ) ⊆ Funlax (V, W )
be the full subcategory on monoidal functors.

Remark. We thus see that the category MonCat of monoidal categories with monoidal
functors is canonically a 2-category: The Hom-category between V and W is given by
Funlax (V, W ). Of course, in order for this to make sense, we need to observe that the identity
of a monoidal category is canonically lax monoidal (in fact monoidal) and that the composition
of two lax monoidal functors is canonically lax monoidal.

Definition 2.33. Let (V, ⊗, 1) be a monoidal category. Then a V -enriched category C

consists of a set of objects, and for any two object x, y ∈ C an object HomC (x, y) ∈ V ,
together with “composition functors”
HomC (x, y) ⊗ HomC (y, z) → HomC (x, z)
and furthermore for every object an “identity” idx
1 → HomC (x, x)
 INTRODUCTION TO INFINITY-CATEGORIES 27

satisfying the obvious associativity and unitality conditions, namely that the following dia-
grams are required to commute:

(HomC (x, y) ⊗ HomC (y, z)) ⊗ HomC (z, w) HomC (x, z) ⊗ HomC (z, w)

α HomC (x, w)

HomC (x, y) ⊗ (HomC (y, z) ⊗ HomC (z, w)) HomC (x, y) ⊗ HomC (y, w)

where α denotes the associativity isomorphism of V. Furthermore

ηr ηl
HomC (x, y) HomC (x, y) ⊗ 1 HomC (x, y) 1 ⊗ HomC (x, y)
id

HomC (x, y) HomC (x, y)

Definition 2.34. A V -enriched functor between V -enriched categories f : C → D consists

of a map on objects x 7→ f (x), and for each two objects a morphism fx,y : HomC (x, y) →
HomD (f x, f y) in V such that the diagrams

HomC (x, y) ⊗ HomC (y, z) HomC (x, z)

fx,y ⊗fy,z fx,z

HomD (f x, f y) ⊗ HomD (f y, f z) HomD (f x, f z)

and
idx
1 HomC (x, x)

idf x fx,x

HomD (f x, f x)

commute.

Definition 2.35. Let V be a monoidal category. Then we let CatV be the category of V -
enriched categories, that is: objects are V -enriched categories and morphisms are V -enriched
functors.

Example 2.36. A Set-enriched category is just an ordinary category. A Cat-enriched

category is a 2-category.

Example 2.37. A Cat-enriched category with one object is the same datum as a monoidal
category which is strictly associative (i.e. where the associativity isomorphisms are the iden-
tity). An enriched functor between two such categories is the same datum as a monoidal
functor.
28 M. LAND

Lemma 2.38. If Φ : V → V 0 is a lax monoidal functor between monoidal categories, then

applying this functor to each hom-object produces a functor Φ∗ : CatV → CatV 0 . This con-
struction in fact determines a 2-functor
MonCat → Cat
from the 2-category of monoidal categories to the 2-category of categories. In particular, a
monoidal adjunction between V and V 0 determines an adjunction on the level of enriched
categories.
Proof. To construct this 2-functor, we consider the map on the level of objects first: It takes a
monoidal category V to the category CatV of V -enriched categories. To make this a 2-functor
we need to construct for every pair V, W of monoidal categories a functor
Funlax (V, W ) → Fun(CatV , CatW )
and then show that this construction is compatible with composition. To construct this
functor, again we first consider its effect on objects: Given a lax monoidal functor Φ : V → W ,
and a V -enriched category C, we consider the following W -enriched category Φ∗ (C): The
objects are the same as the objects of C, and for X, Y ∈ ob(C) we define
HomΦ∗ (C) (X, Y ) = Φ(HomC (X, Y )).
It is straight forward to show that Φ∗ (C) is a W -enriched category: For instance, to show
that composition satisfies the associativity constraint, one uses the compatibility of Φ with the
associativity isomorphisms of V and W . Next, we need to explain the effect on morphisms. So
let τ : Φ → Ψ be a monoidal transformation. We wish to construct a natural transformation
between Φ∗ and Ψ∗ . Concretely, we need to construct natural maps Φ∗ (C) → Ψ∗ (C) in CatW ,
i.e. natural W -enriched functors τ∗ : Φ∗ (C) → Ψ∗ (C). On objects this functor is defined to be
the identity, and on morphisms between X and Y we have to construct a map
τ∗ : Φ(HomC (X, Y )) → Ψ(HomC (X, Y ))
and we simply use the natural map τHomC (X,Y ) given by the natural transformation τ : Φ → Ψ.
To see that this is compatible with composition, we use that τ is a monoidal transformation,
so that the diagram
Φ(HomC (X, Y )) ⊗ Φ(HomC (Y, Z)) Φ(HomC (X, Y ) ⊗ HomC (Y, Z))

Ψ(HomC (X, Y )) ⊗ Ψ(HomC (Y, Z)) Ψ(HomC (X, Y ) ⊗ HomC (Y, Z))

commutes. Then we use naturality of τ to see that also the diagram

Φ(HomC (X, Y ) ⊗ HomC (Y, Z)) Φ(HomC (X, Z))

Ψ(HomC (X, Y ) ⊗ HomC (Y, Z)) Ψ(HomC (X, Z))

commutes. Glueing these two diagrams together shows that the map τ∗ is compatible with
composition, and thus is in fact a W -enriched functor Φ∗ (C) → Ψ∗ (C) as needed.
 INTRODUCTION TO INFINITY-CATEGORIES 29

It is then also clear by definition that for two composable lax monoidal functors Ψ and Φ,
we have Ψ∗ (Φ∗ (C)) = (Ψ ◦ Φ)∗ (C), so that compatibility with composition is immediate.
To see the in particular is now easy: An monoidal adjunction consists of lax monoidal
functors Φ and Ψ and unit and counit transformations which are itself monoidal transforma-
tions. By the previously established parts, these are sent to functors Φ∗ and Ψ∗ equipped with
candidates for the unit and counit. The only thing to check is the snake identities, but they
follow from the fact that they hold for Φ and Ψ, and that the constructed functor preserves
identities. 
Definition 2.39. Let V be a monoidal category and C a V -enriched category. Then Its
underlying category uC is obtained via the lax monoidal functor Hom(1, −) : V → Set. In
formulas, we have
uC = HomV (1, −)∗ (C) ∈ CatSet = Cat.

Definition 2.40. We will call a category enriched in simplicial sets simply a simplicial
category. Here sSet is viewed as a monoidal category via the cartesian product. We will write
Cat∆ for CatsSet .

Remark. There is some ambiguity in the above definition: Usually a simplicial category
would rather refer to a simplicial object in categories, i.e. a functor ∆op → Cat. Luckily we
have the following lemma about this:

Lemma 2.41. There is a canonical fully faithful embedding Cat∆ → Fun(∆op , Cat) deter-
mined by the family of functors (evn )∗ : Cat∆ → Cat, and the essential image can be charac-
terized as those simplicial objects in categories, whose underlying simplicial set of objects is
constant.
Lemma 2.42. Suppose that F : C → D is left adjoint to G : D → C. Suppose furthermore
that
(1) C is cocomplete, and
(2) G is fully faithful.
Then D is cocomplete as well.
Dually, if D is complete, and F is fully faithful, then C is complete as well.
Proof. Assume that (1) and (2) hold. Let X : I → D be a diagram and consider the object
F (colimI G(Xi )) which exists since C is cocomplete. We wish to prove that it satisfies the
universal property of a colimit:
HomD (F (colim G(Xi )), Y ) ∼
= HomC (colim G(Xi ), G(Y ))
I I
∼
= lim HomC (G(Xi ), G(Y )) ∼
= lim HomC (Xi , Y )
I I
where the last bijection holds by fully faithfulness of G.
Dually, let X : I → C be a diagram and consider the object G(limI F (Xi )) which exists
since D is complete. Then we calculate that
HomC (Y, G(lim F (Xi ))) ∼
= HomD (F (Y ), lim F (Xi ))
I I
∼
= lim HomD (F (Y ), F (Xi )) ∼
= lim HomC (Y, Xi )
I I
where the last bijection uses that F is fully faithful. 
30 M. LAND

Lecture 9 – 15.11.2018.
Proposition 2.43. The category Cat is bicomplete, and the functor ob(−) : Cat → Set
preserves limits and colimits.
Proof. The case of limits can be done by hand: Given a diagram C : I → Cat, sending i ∈ I
to Ci , we define its limit limI Ci as follows:
(1) ob(limI Ci ) = limI ob(Ci ), and
(2) for any two objects {Xi }i∈I and {Yi }i∈I we have
HomlimI Ci ({Xi }, {Yi }) = lim HomCi (Xi , Yi ).
I
To see that this is in fact a category, one uses that limits commute with products (when
defining composition) and that one has a canonical map from a limit indexed over I × I
to the limit indexed over the restriction to I along the diagonal. This is a reason why the
situation is more complicated with colimits. By definition then, the functor ob(−) : Cat → Set
commutes with limits (as it must, since it is right adjoint to the discrete category functor
d : Set → Cat).
To prove existence of colimits we make use of Lemma 2.42: We consider the adjunction
(h, N) of functors between sSet and Cat of Proposition 2.18. We have seen that
(1) sSet is cocomplete (in fact bicomplete by Corollary 1.26), and
(2) N : Cat → sSet is fully faithful, Corollary 2.20.
It follows that Cat is cocomplete and that a colimit of a diagram C : I → Cat is given by
h(colimI N(Ci )). With this we calculate its objects as follows:
ob(colim Ci ) ∼= ob(h(colim N(Ci )))
I I
∼
= (colim N(Ci ))0 ∼
= colim(N(Ci )0 )
I I
∼
= colim ob(Ci )
I
as needed. 
Lemma 2.44. Let C0 ⊆ C be a full subcategory of a category C and let X : I → C0 be a
diagram. If X has a (co)limit in C which happens to lie in C0 , then this is also a (co)limit in
C0 .
Proof. Obvious from the universal property and the fact that the inclusion is full. 
Corollary 2.45. The category Cat∆ is bicomplete.
Proof. Consider an I-shaped diagram of simplicial categories Ci . By Lemma 2.41 this gives
rise to an I-shaped diagram in Fun(∆op , Cat), whose associated I-shaped diagram of sim-
plicial sets of objects in constant, i.e. where for all i ∈ I, we have that ob(Ci ) is a constant
simplicial set. We wish to show that then also ob(colimI Ci ) is a constant simplicial set. By
Proposition 2.43, we have
ob(colim Ci ) ∼
= colim ob(Ci )
I I
and the latter is a colimit in simplicial sets, over constant simplicial sets. Since the constant
functor c : Set → sSet admits a right adjoint (ev0 ) it preserves colimits. Thus the colimit over
constant simplicial sets is itself constant (on the colimit of the sets involved). The argument
for limits works the same. 
 INTRODUCTION TO INFINITY-CATEGORIES 31

Remark. Nothing is special about ∆ here. In fact the same argument goes through to prove
that for any small category C, there is a fully faithful inclusion CatP(C) ⊆ Fun(Cop , Cat) with
essential image given by those functors whose presheaf of objects is constant. It follows com-
pletely analogously that CatP(C) is bicomplete. Here, we always use the pointwise (cartesian)
monoidal structure on P(C).

Example 2.46. The nerve functor N : Cat → sSet is monoidal: It preserves products since
it is a right adjoint. Hence, given a 2-category, we obtain a simplicial category by applying the
nerve functor to all hom-categories. Furthermore, for a 2-category C we have u(N∗ (C)) = uC:
This is simply because
HomsSet (∆0 , −)∗ ◦ N∗ = (HomsSet (∆0 , −) ◦ N) = HomsSet (∆0 , N(−)) = HomCat (h∆0 , −)
and the h∆0 is the unit of the monoidal structure on Cat given by cartesian product.

Lemma 2.47. The functors c : Set → sSet and π0 , ev0 : sSet → Set are canonically monoidal.
Proof. We claim that every functor F : (V, ×, ∗) → (W, ×, ∗) is canonically oplax monoidal.
The oplax monoidal structure maps are given by the maps
(1) F (∗) → ∗, the unique map to the terminal object of W , and
(2) F (X × Y ) → F (X) × F (Y ) given by the effect of F on the two projections
X ← X × Y → Y.
It hence suffices to see that the canonical oplax structure maps in our examples are isomor-
phisms. For c and ev0 this follows directly from the definitions. Only the functor π0 requires
an actual argument. We want to check that the map
π0 (X × Y ) → π0 (X) × π0 (Y )
is a bijection. By definition of π0 of a simplicial set, Definition 1.10, we have a commutative
square
∼
=
(X × Y )0 X0 × Y0

π0 (X × Y ) π0 (X) × π0 (Y )
in which the vertical maps are bijections. Since the top horizontal map is a bijection (this is
the monoidality of ev0 ), it follows that the oplax monoidal structure map is surjective. To
see that it is injective, it suffices to check that generators of the relations can be lifted. Since
these are given by 1-simplices, and ev1 is also monoidal, the claim follows. Also it follows
directly that the map π0 (∗) → ∗ is an isomorphism. 
Definition 2.48. We obtain a functors
(1) c = c∗ : Cat → Cat∆ , which sends a category to the simplicially enriched category
with constant simplicial enrichment,
(2) π = (π0 )∗ : Cat∆ → Cat, called the homotopy category of a simplicial category, and
(3) u = (ev0 )∗ : Cat∆ → Cat, called the underlying category.
Notice that ev0 = HomsSet (∆0 , −) so that this is the same underlying category as in Defini-
tion 2.39.
32 M. LAND

Definition 2.49. A simplicial functor f : C → D between simplicial categories is called a

weak equivalence if it induces
(1) a weak equivalence on all hom simplicial sets, (weakly fully faithful) and
(2) an essentially surjective functor π(C) → π(D) (weakly essentially surjective).

Lemma 2.50. Every functor between cartesian monoidal categories is canonically oplax
monoidal. Every natural transformation between two such functors is also canonically oplax
monoidal. In particular, the adjunctions (π0 , c) and (c, ev0 ) are monoidal adjunctions.
Proof. We have seen that every functor F : (V, ×, ∗) → (W, ×, ∗) is canonically oplax monoidal
in the proof of Lemma 2.47. So let τ : F → G be a natural transformation. We wish to show
that the diagram
F (X × Y ) F (X) × F (Y )
τX×Y τX ×τY

G(X × Y ) G(X) × G(Y )

commutes. For this it suffices to see that each of the two following squares commutes

F (X) F (X × Y ) F (Y )
τX τX×Y τY

G(X) G(X × Y ) G(Y )

which follows from naturality of τ .

It now suffices to show that an oplax monoidal transformation between monoidal functors is
also a monoidal transformation. This follows from the general fact that if a square commutes,
in which both vertical maps are isomorphisms, then the square with the inverse vertical maps
also commutes. 
Corollary 2.51. The two functors c : Cat → Cat∆ and π : Cat∆ → Cat form an adjoint pair,
with π being left adjoint to the constant functor c. Similarly, the two functors u : Cat∆ → Cat
and c : Cat → Cat∆ form an adjoint pair with c being left adjoint to the underlying functor.
Proof. This follows from Lemma 2.38 and Lemma 2.50. 
Observation 2.52. The adjunction gives rise to a canonical functor C → cπ(C) of simplicial
categories.

Lecture 10 – 19.11.2018.
Definition 2.53. Given a simplicial category C, and two objects x, y ∈ C, we say that a
morphism from x to y is a morphism in the underlying category uC. In other words, it is a
0-simplex of HomC (x, y). Such a morphism is called an equivalence if its image in π(C) is an
isomorphism.

We wish to extend the notion of the nerve of a category to simplicially enriched categories.
For this we need a version of the category [n] which is well suited for simplicially enriched
categories.
 INTRODUCTION TO INFINITY-CATEGORIES 33

Definition 2.54. Let J be a finite non-empty linearly ordered set, and let i, j ∈ J be
elements. We let Pi,j be the following set of subsets of J:
Pi,j = {I ⊆ J : i, j ∈ I and k ∈ I ⇒ i ≤ k ≤ j}
In words, Pi,j consists of all subsets of [i, j] ⊆ J which contain i and j.
Pi,j is partially ordered by inclusion: I ≤ I 0 ⇔ I ⊆ I 0 . Notice that Pi,j is only non-empty
if i ≤ j.

Observation 2.55. Given a triple i ≤ j ≤ k in J, there is a canonical map of partially

ordered sets
Pi,j × Pj,k → Pi,k
given by sending (I, I 0 ) to I ∪ I 0. This clearly defines an associative binary operation.

As a consequence we obtain the following:

Definition 2.56. Let J be a non-empty linearly ordered set. Then the following defines a
simplicially enriched category C[∆J ] ∈ Cat∆ . Objects of C[J] are given by the elements of J.
Furthermore (
∅ if j < i
HomC[∆J ] (i, j) =
N(Pi,j ) if i ≤ j.
Composition is defined via the previous observation.

Lemma 2.57. For every n ≥ 1, we have that N(P0,n ) is isomorphic to (∆1 )n−1 . Further-
more, Pi,j ∼
= P0,j−i .
Proof. The furthermore simply follows by choosing an isomorphism J ∼ = [n] for some n, and
then considering the unique order preserving isomorphism of [i, j] with [0, j − i]. To show
that N(P0,n ) ∼
= (∆1 )n−1 it suffices to find an isomorphism of posets
P0,n ∼
= [1] × · · · × [1]
where the latter product is has n − 1 many factors. This simply comes from the following
construction: A subset I ⊆ [n] containing {0, n} is determined by which of the elements
1, . . . , n − 1 is contained in I. We label an element contained in I with a 1 and elements not
contained in I with a 0. This constructs a map of posets P0,n → [1] × · · · × [1], which is clearly
an isomorphism. 
Lemma 2.58. Let C be a category with initial or terminal object. Then N(C) is contractible,
i.e. its identity map is homotopic to the constant map at the initial or terminal object.
Proof. Let us consider the case where C has an initial object ∅. The other case follows from
the fact that a simplicial set X is contractible if and only if X op is contractible. We claim
that the identity functor of C admits a natural transformation from the constant functor with
value ∅, simply given on an object X ∈ C by the unique map ∅ → X. The relevant diagrams
commute by the uniqueness of the maps. We obtain a functor C × [1] → C whose restriction
to 0 is the constant map at ∅ and whose restriction to 1 is the identity of C. Applying the
nerve functor N, we obtain a simplicial homotopy
N(C) × ∆1 → N(C)
from the constant map at the vertex given by ∅ to the identity of N(C). 
34 M. LAND

Corollary 2.59. For i ≤ j, the simplicial set HomC[∆J ] (i, j) is contractible.

Proof. We have just seen that HomC[∆J ] (i, j) = ∼ ∆j−i−1 = N([1] × · · · × [1]). The category
[1] × · · · × [1] has both an initial and a terminal object, so we can appeal to Lemma 2.58. 
Remark. The functor | − | : sSet → Top preserves products. To see this, it suffices to see
that |∆n × ∆m | ∼
= ∆ntop × ∆m top . This is a concrete calculation. Hence, a simplicial homotopy
induces a homotopy of geometric realizations. It follows that every contractible simplicial set
is weakly contractible (i.e. its geometric realization is contractible).

Lemma 2.60. There is a unique isomorphism π(C[∆n ]) ∼ = [n] which is the identity on
objects. By adjunction we obtain a canonical functor C[∆n ] → c[n], and this functor is a weak
equivalence of simplicial categories.
Proof. Everything follows from Corollary 2.59: It calculates π(C[∆n ]) to be [n]. It follows
that the induced functor C[∆n ] → c[n] is bijective on objects (and thus weakly essentially
surjective) and a weak equivalence on hom simplicial sets. 
Remark. ADJUST THIS REMARK. In fact it is a cofibrant replacement in a suitable
model structure on simplicial categories. In the former category we have dropped the strict
associativity for composition i → j → k in [n], now have instead a (contractible) space of
compositions.

Lemma 2.61. The association J 7→ C[∆J ] extends to a functor

[Link] → Cat∆ .
In particular, we obtain a functor
∆ → Cat∆ [n] 7→ C[∆n ]
which is a cosimplicial object in simplicially enriched categories.
Proof. We need to show that every map J → J 0 of linearly ordered sets induces a simplicially
0
enriched functor C[∆J ] → C[∆J ]. For this it suffices to explain how such a map J → J 0
induces, for every i ≤ j in J a map of posets Pi,j → Pf (i),f (j) . This is simply given by sending
a subset I to its image f (I). It is easy to see that this in fact produces a functor. 
Definition 2.62. Let C be a simplicially enriched category. Then we define its simplicial
nerve (or the homotopy coherent nerve) as the following simplicial set
N(C)n = HomCat∆ (C[∆n ], C).

Lemma 2.63. If C is an ordinary category, viewed as a simplicially enriched category cC

via the constant functor, then there is an isomorphism
N(C) ∼= N(cC).
Proof. This is because:
N(cC)n = HomCat∆ (C[∆n ], cC)
∼
= HomCat (π(C[∆n ]), C)
∼
= HomCat ([n], C) = N(C)n
The only thing that needs justification is the isomorphism π(C[∆n ]) ∼
= [n] which we proved
in Lemma 2.60. 
 INTRODUCTION TO INFINITY-CATEGORIES 35

Remark. The analog of such a statement with π(C) and uC is wrong!!

Lecture 11 – 22.11.2018.
Observation 2.64. Let C be a simplicially enriched category. Let us investigate the sim-
plicial set N(C) in more detail: We recall that
N(C)n = HomCat∆ (C[∆n ], C).
Unravelling the simplicial categories C[∆0 ] and C[∆1 ] we find the following
(1) C[∆0 ] is a simplicial category with a single object, and its endomorphism simplicial
set is given by N(P0,0 ) which is also given by ∆0 .
(2) C[∆1 ] is a simplicial category with two objects, 0, and 1, and all morphism simplicial
sets are given by ∆0 .
In other words, both C[∆0 ] and C[∆1 ] are given by c[0] and c[1], where we view [0] and [1] as
categories, and then apply the constant functor c : Cat → Cat∆ . As a consequence we obtain
(1) N(C)0 = HomCat∆ (C[∆0 ], C) ∼ = HomCat ([0], uC), and
(2) N(C)1 = HomCat∆ (C[∆1 ], C) ∼ = HomCat ([1], uC),
by of Corollary 2.51. In words, objects of N(C) are given by the objects of C, and morphisms
of N(C) are given by the morphisms (i.e. the 0-simplices of the mapping simplicial sets) of C.
Let us go one step further and analyze the simplicial category C[∆2 ]. Its objects are given
by {0, 1, 2}. All endomorphism simplicial sets are given by ∆0 . Furthermore
HomC[∆2 ] (0, 1) = ∆0 = HomC[∆2 ] (1, 2).
However, to analyze the simplicial set of morphism from 0 to 2 we have to investigate N(P0,2 ).
By definition, P0,2 is the partially ordered set of subsets of {0 < 1 < 2} which contain 0 and
2. There are precisely two such subsets, so that we obtain P0,2 = [1]. In particular,
HomC[∆2 ] (0, 2) = N(P0,2 ) = ∆1 .
Unravelling, we find that a 2-simplex in N(C) consists of the following data: Objects X, Y ,
and Z (associated to the three objects 0, 1, and 2). A morphism f : X → Y (associated to
the unique morphism from 0 to 1 in C[∆2 ]), and a morphism g : Y → Z (associated to the
unique morphism from 1 to 2 in C[∆2 ]), a 1-simplex ∆1 → HomC (X, Z) (associated to the
mapping simplicial set between 0 and 2 in C[∆2 ]) whose restriction to 0 is given by gf and
whose restriction to 1 is given by some other morphism.
Informally, a 2-simplex thus consists of the data of two composable morphisms X → Y
and Y → Z, a further morphism X → Z and a homotopy between the composite of the first
two to the last morphism.

Lemma 2.65. The category Cat∆ of simplicially enriched categories admits all small col-
imits. Hence there exists a unique colimit preserving functor
C[−] : sSet → Cat∆
which sends ∆n to C[∆n ]. This functor is automatically left adjoint to the simplicial nerve
functor.
Proof. The existence of colimits was done in Corollary 2.45. The rest is formal, we have seen
the argument several times. 
36 M. LAND

Fact 2.66. S Given a simplicial set X, consider subsimplicial sets Ai ⊆ X. Then also the
union A = Ai is a sub simplicial set of X. In this situation we have that C[A] is the sub
simplicial category of C[X] generated by the C[Ai ]. Recall that a sub simplicial category of
a simplicial category C is determined by a subset of the objects, and for any two such, a sub
simplicial set of the hom simplicial set in C.
This fact is for instance shown in [Joy07, Corollary 1.15]. It builds fundamentally on
the fact that the functor C[−] : sSet → Cat∆ sends monomorphisms of simplicial sets to
monomorphisms of simplicial categories.

Lemma 2.67. Let 0 < j < n and consider the horn Λnj . We have that C[Λnj ] is the sub
simplicial category of C[∆n ] with:
(1) The objects of C[Λnj ] are given by the vertices of Λnj , thus by all objects of C[∆n ].
(2) The morphism simplicial sets are given as follows:
HomC[Λnj ] (i, k) = HomC[∆n ] (i, k)
unless (i, k) = (0, n), and
HomC[Λnj ] (0, n) ⊆ HomC[∆n ] (0, n) = N(P0,n )
is given by the sub simplicial set of (∆1 )n−1 obtained by deleting the interior and the
bottom j-face.
Proof. Let us first show that the candidate for C[Λnj ] is in fact a sub simplicial category of
C[∆n ]. We know that Λnj =
S [n]\{i}
∆ . So let us first describe the sub simplicial category
i6=j

C[∆[n]\{i} ] ⊆ C[∆n ].
The objects of this subcategory are given by all objects except for the object corresponding
to i ∈ [n]. Now assume k, l ∈ [n] \ {i} with k ≤ l. If l < i, then the there is an obvious
isomorphism
[n]\{i} ∼ [n]
Pk,l = Pk,l
where the superscript indicates in which linearly ordered set to perform the construction Pk,l
of Definition 2.54. Thus we obtain an equality of simplicial hom sets
HomC[∆[n]\{i} ] (k, l) = HomC[∆n ] (k, l).
Likewise there is such an isomorphism if k > i.
Now let (k, l) 6= (0, n), i.e. either k 6= 0 or l 6= n. Assume first that k 6= 0. Then we have
the following chain on inclusions of simplicial sets
HomC[∆[n]\{0} ] (k, l) ⊆ HomC[Λnj ] (k, l) ⊆ HomC[∆n ] (k, l)
in which the composition is an equality by the previous arguments. We thus see that in fact
both inclusions must be equalities. If l 6= n, the same argument works. We summarize that
HomC[Λnj ] (k, l) = HomC[∆n ] (k, l)
unless (k, l) = (0, n).
Let us next determine HomC[Λnj ] (0, n). By Fact 2.66 we have for every i 6= j and i 6= 0, n
inclusions as follows:
HomC[∆[n]\{i} ] (0, n) ⊆ HomC[Λnj ] (0, n) ⊆ HomC[∆n ] (0, n).
 INTRODUCTION TO INFINITY-CATEGORIES 37

For 0 < i < n we find that

[n]\{i} [n]
P0,n ⊆ P0,n
[n]
is a sub poset, consisting precisely of those I ∈ Pk,l for which i ∈
/ I. In other words, it is
given by the map of posets
[1]×(n−2) → [1]×(n−1)
which includes a 0 at the ith spot of [1]×(n−1) . Upon taking the nerve, this gives the map
(∆1 )n−2 ⊆ (∆1 )n−1 which is the zero vertex of ∆1 in the ith coordinate.
It follows that for every i 6= j, the face (∆1 )n−2 ⊆ (∆1 )n−1 where the ith entry is 0 is
contained in HomC[Λnj ] (0, n). We call the face where the ith entry is 0 the bottom i-face. We
find that a priori, the bottom j-face is not contained in HomC[Λnj ] (0, n) as promised in the
statement of the lemma. It remains to show that the top k-face is contained in HomC[Λnj ] (0, n)
for all 0 < k < n. For this we consider the diagram

HomC[Λnj ] (0, k) × HomC[Λnj ] (k, n) HomC[Λnj ] (0, n)

∼
=

HomC[∆n ] (0, k) × HomC[∆n ] (k, n) HomC[∆n ] (0, n)

which encodes the composition in the respective categories. This diagram must commute, as
C[Λnj ] is a sub simplicial category of C[∆n ]. Since the left vertical map is an isomorphism it
hence suffices to show that the top i-face are contained in the image of the lower horizontal
map. This map is induced by the map of posets
P0,k × Pk,n → P0,n
which is induced by sending (I, I 0 ) to I ∪ I 0 . Since I and I 0 contain the element k, we see
that after the identification of these posets with cubes as in Lemma 2.57, we obtain the map
(∆1 )n−2 ∼
= (∆1 )k−1 × (∆1 )n−k−1 → (∆1 )n−1
which is induced by inserting a 1 in the k th slot. This shows that the top k-face is contained
in HomC[Λnj ] (0, n). Since we already know that this punctured cube gives rise to a sub sim-
plicial category, and C[Λnj ] is contained in it, and itself contains the sub simplicial categories
determined by the i-faces of ∆n for i 6= j, the lemma is shown. 

Lecture 12 – 26.11.2018.
Lemma 2.68. The coherent nerve of a simplicial category C is a composer. Furthermore,
it is an ∞-category if all hom-simplicial sets are Kan complexes, i.e. if it is in fact enriched
in the symmetric monoidal category of Kan complexes.
Proof. We consider an extension problem

In N(C)

∆n
38 M. LAND

which, by adjunction is equivalent to the extension problem

C[I n ] C

C[∆n ]

and to show that the latter can be solved it suffices to show that the map C[I n ] → C[∆n ]
has a retraction. To see this, we first claim that C[I n ] ∼
= c[n]. This follows simply from the
fact that it is true for n = 1, and induction, using that I n ∼= I n−1 q∆0 I 1 , and the fact that
the functor c : Cat → Cat∆ preserves colimits. In Lemma 2.60 we have seen that there is a
unique simplicial functor C[∆n ] → c[n] which is the identity on objects. It follows that the
composition
c[n] ∼
= C[I n ] → C[∆n ] → c[n]
is a functor which is the identity on objects, and therefore is an isomorphism.
To show that, in the case of a Kan enrichment, N(C) is an ∞-category, we need to solve
an extension problem
Λnj N(C)

∆n
for 0 < j < n.
By adjunction, we need to argue why every simplicially enriched functor C[Λnj ] → C extends
to a simplicially enriched functor C[∆n ] → C provided 0 < j < n. For this we use our analysis
of C[Λnj ] of Lemma 2.67.
To solve the extension problem we are interested in, it thus suffices to prove that there
exists an extension in the diagram
f
HomC[Λnj ] (0, n) HomC (f (0), f (n))

HomC[∆n ] (0, n)

It follows from the description of Lemma 2.67 that the vertical map is an anodyne map of
simplicial sets, i.e. it induces a weak equivalence on geometric realizations, so by Fact 1.50 the
dotted arrow exists. It now suffices to show that this construction in fact produces a simplicial
functor C[∆n ] → C. To see this, one uses that HomC[∆n ] (0, 0) = ∆0 = HomC[∆n ] (n, n) so that
no new relations for functoriality are to check: Since the diagram

HomC[Λnj ] (0, k) × HomC[Λnj ] (k, n) HomC[Λnj ] (0, n)

∼
=

HomC[∆n ] (0, k) × HomC[∆n ] (k, n) HomC[∆n ] (0, n)

commutes, any composition which factors through the object k with 0 < k < n, is already
contained in the sub simplicial category C[Λnj ]. 
 INTRODUCTION TO INFINITY-CATEGORIES 39

Observation 2.69. We find that a 2-category gives rise to a simplicial category, so that its
coherent nerve is a composer. Moreover, a (2, 1)-category gives rise to a simplicial category
where all hom simplicial sets are nerves of groupoids and thus Kan complexes by Lemma 1.56.
In particular, a (2, 1)-category gives rise to an ∞-category in our sense.

Definition 2.70. Consider the simplicial category with objects CW-complexes and mapping
simplicial sets given by the singular set of the mapping space. (Notice that the singular set
commutes with products). The simplicial nerve of this category gives an ∞-category which
we will call the ∞-category of spaces and denote by An.

Observation 2.71. Objects of An are CW complexes, and morphism are given by points
in the space map(X, Y ), i.e. by a continuous map from X to Y . The homotopy category is
what one would expect: Morphisms are homotopy classes of maps.

Remark. We would like to have a “purely simplicial” model of the ∞-category of spaces
(what a perverse thing to say – but it comes from the fact that we wish to think of Kan
complexes/spaces as ∞-groupoids and later want to have spaces and ∞-categories on equal
footing), i.e. where we directly construct a simplicial category whose objects are Kan com-
plexes. For this we will need to show that for a simplicial set K and a Kan complex X, the
simplicial set of maps Hom(K, X) is again a Kan complex. The things needed to show this
are also needed on the way of showing that ∞-groupoids are Kan complexes, and we will
start to develop these tools in the next section.

Lemma 2.72. The product of ∞-categories is an ∞-category. The coproduct of ∞-categories

is an ∞-category.
Proof. For products, one can solve the extension problem in every ∞-category individually.
This provides an extension for the product. For coproducts, we observe that both Λnj and
∆n are connected. Thus an extension problem for a coproduct of ∞-categories is in fact an
extension problem for a single one. 
Definition 2.73. A sub-∞-category C0 of an ∞-category C is a sub-simplicial set determined
by a subset X ⊆ C0 of 0-simplices and a subset S ⊆ C1 of 1-simplices between objects lying in
X, such that S contains identities and is closed under compositions and equivalences. Then an
n-simplex of C belongs to C0 if and only its restriction to the spine I n has its edges contained
in S. A subcategory is called full if S = C1 .

Lecture 13 – 29.11.2018.
Lemma 2.74. A sub-∞-category of an ∞-category is itself an ∞-category. Its homotopy
category is the subcategory of hC on the image of the morphisms lying in S. The diagram
C0 C

N(hC0 ) N(hC)
is a pullback. Furthermore, for any subcategory C0 ⊆ hC of the homotopy category, this
pullback defines a sub-∞-category of C with X and S given by the preimage of objects and
morphisms along the canonical map C → N(hC).
40 M. LAND

Proof. Let C be an ∞-category and let D ⊆ hC be a subcategory of its homotopy category. Let
C0 be the pullback N(D)×N(hC) C. As pullbacks preserve monomorphisms, C0 is a subsimplicial
set of C. Let us spell out what it means for an n-simplex of C to lie in C0 : It means precisely
that the induced n-simplex of N(hC) lies in N(D). This is the case if and only if the spine, and
thus every edge of the simplex lies in D. Now observe, that a subset S ⊆ C1 which contains
all identities of objects in X and is closed under compositions is in fact the preimage of the
1-morphisms of the nerve of a subcategory of hC. Notice that in the above definition, we have
that (C0 )1 = S. 

Corollary 2.75. There is a bijective correspondence between sub-∞-categories of C and

subcategories of hC induced by taking the homotopy category. Full sub-∞-categories of C
correspond precisely to full subcategories of hC.

Definition 2.76. A natural transformation between two functors f, g : C → D is a simplicial

map C × ∆1 → D which restricts to the given functors appropriately.

Observation 2.77. We observe that functors and natural transformations of functors be-
tween X and Y are precisely the 0- and 1-simplices of the hom-simplicial set Hom(X, Y ). As
in ordinary category theory we would like to have an ∞-category of functors between two
∞-categories. Also, one would expect that one has N(Fun(C, D)) = Fun(NC, ND). Since
one also has N(Fun(C, D)) = Hom(NC, ND) (use that the nerve functor is fully faithful) one
might hope that the hom-simplicial set is again an ∞-category. This will turn out to be true,
and like in the case of Kan complexes is not a triviality. It will be the objective of the next
lectures to prove (amongst other things) this fact.

3. Anodyne maps and fibrations

Definition 3.1. A map of simplicial sets X → Y is an (inner, left, right) fibration if it
satisfies the right lifting property (RLP) with respect to (inner, left, right) horn inclusions

Λnj X

∆n Y

Definition 3.2. A map of simplicial sets A → B is an (inner, left, right) anodyne map, if
it satisfies the left lifting property (LLP) with respect to (inner, left, right) fibrations

A X

B Y

Definition 3.3. Let S be a set of morphisms in a category C. We let χR (S) be the set of
morphisms having the RLP wrt to S and χL (S) be the set of morphisms having the LLP wrt
S. We let χ(S) = χL (χR (S)), i.e. χ(S) is the set of morphisms which have the LLP wrt to
morphisms having the RLP wrt S.
 INTRODUCTION TO INFINITY-CATEGORIES 41

Example 3.4. Let S be the set of (inner, left, right) horn inclusions. Then the (inner, left,
right) fibrations are given by χR (S), and the (inner, left, right) anodyne maps are given by
χ(S).

Definition 3.5. A saturated set of morphisms is a set of morphisms T which is closed under
taking pushouts (along arbitrary maps), arbitrary coproducts, countable compositions (i.e.
colimits along N), and retracts. Given an arbitrary set of morphisms S, we call the smallest
saturated set containing S the saturated closure of S and denote it by S.

Remark. Notice that the intersection of saturated sets is again saturated. Thus to see that
the saturated closure exists, it suffices to show that there is a saturated set containing S. One
can for instance take simply all morphisms: It is obviously saturated and contains S.

Definition 3.6. A morphism f : A → B is a retract of a morphism f 0 : A0 → B 0 if there is

a commutative diagram
id

A A0 A
f f0 f

B B0 B
id

Remark. Let us make more precise what the conditions of Definition 3.5 mean: Being
closed under arbitrary coproducts
` means that given a family {fi }i∈I such that each fi is an
element of T , then also fi is an element of T . Being closed under pushouts means that
i∈I
given a map f : A → B which belongs to T and any other map ϕ : A → A0 , then in the
following pushout diagram
ϕ
A A0
f f0

B B0
also the map f 0 belongs to T . Being closed under countable compositions means the following:
Consider the category N (the category associated to the poset N) and consider a functor
X : N → sSet. A set S is closed under countable compositions if the following condition
holds. Suppose that for every i ∈ N, the canonical map
X(i) → X(i + 1)
is contained is S, then also the map X(0) → colimN X is contained in S. Being closed under
retracts means that for f and f 0 as in Definition 3.6, if f 0 is an element of T , the so is f .

Example 3.7. The set of monomorphisms in sSet is a saturated set. A further description
of this set will be worked out in ??, namely that the monomorphisms are given by χ({∂∆n →
∆n }n≥0 ).

Lemma 3.8. Given a set of morphisms S, we have that χL (S) is a saturated set. In
particular χ(S) is a saturated set.
42 M. LAND

Proof. Let α : A → B be a morphism in χL (S) and let ϕ : A → A0 be an arbitrary morphism.

Consider the pushout B 0 = A0 qA B. We wish to show that the canonical map A0 → B 0 is
contained in χL (S). So let f : X → Y be a map in S and consider the diagram
A A0 X

B B0 Y
where we wish to construct the dashed map. We certainly have the dotted map, and hence
obtain the dashed map by the universal property of a pushout. Likewise, suppose that A → B
is a retract of A0 → B 0 and that A0 → B 0 is contained in χL (S). To show that then also
A → B is contained in χL (S) we consider a map f : X → Y in S and a diagram
A A0 A X

B B0 B Y
and we wish to show that the dashed arrow exists. Again, the dotted arrow exists, which we
may restrict to B along the map B → B 0 .
Suppose`now given a family {fi : Ai → Bi }i∈I of elements of χL (S). We then want to show
that also fi ∈ χL (S). So consider a lifting problem
i∈I
`
Ai X
i∈I
`
fi ϕ
i∈I
`
Bi Y
i∈I

with ϕ : X → Y being an element of S. Then the dashed arrow exists simply by the universal
property of coproducts and the assumption that each fi is an element of χL (S).
It remains to show that for any diagram A : N → C where each map Ai → Ai+1 is contained
in χL (S), then also the map A0 → A = colimi Xi is contained in χL (S). This follows simply
from the universal property of colimits: We consider again f : X → Y in S and a diagram
A0 X

A Y
where we need to show existence of the dashed arrow. This is a map out of a colimit, so
it suffices to show that there exists compatible maps Ai → X making everything commute.
This can be done inductively using that each map Ai → Ai+1 is contained in χL (S). 
The following is known as the small object argument and is a very useful tool to construct
non-trivial factorizations of maps.
Proposition 3.9. Let S be a set of morphisms {Ai → Bi }i∈I such that for every i ∈ I,
the simplicial set Ai has only finitely many non-degenerate simplices. If f is an arbitrary
morphism, then f can be factored as first a map contained in S followed by a map in χR (S),
i.e. one which has the RLP wrt S.
 INTRODUCTION TO INFINITY-CATEGORIES 43

Proof. Consider first the set ΘS which consists of triples (αi , ui , vi ) where αi : Ai → Bi is an
element of S, and where ui : Ai → X and vi : Bi → Y are maps such that the diagram
ui
Ai X
αi f
vi
Bi Y

commutes. We obtain a commutative diagram

`
Ai X
ΘS

E 1 (f )
`
Bi
ΘS

where E 1 (f ) is defined to be the pushout. Since S is closed under arbitrary coproducts and
pushouts, we deduce that the map X → E 1 (f ) is contained in S.
Doing the same with the map f : X → Y replaced by the map E 1 (f ) → Y , we obtain a
sequence
X → E 1 (f ) → E 2 (f ) → · · · → Y.
Let us define E ω (f ) = colimk E k (f ) so that we obtain a factorization
X → E ω (f ) → Y.

By construction every map E k (f ) → E k+1 (f ) is contained in S because S is closed under

pushouts and coproducts. Hence, since S is saturated, also the map X → E ω (f ) is contained
in S. It remains to show that the map E ω (f ) → Y is contained in χR (S). So let us consider
a map Ai → Bi in S and a diagram

Ai E ω (f )

Bi Y

where we wish to show the existence of the dashed arrow, making both triangles commute.
We now claim that the canonical map

colim HomsSet (Ai , E k (f )) → HomsSet (Ai , E ω (f ))

k∈N

is a bijection (one calls such Ai compact). This follows from the fact that there are only
finitely many non-degenerate simplices in Ai , and that any simplicial map is determined on
the non-degenerate simplices. Hence, we find a k ∈ N, such that the given map Ai → E ω (f )
factors as a composition
Ai → E k (f ) → E ω (f ).
44 M. LAND

Now the diagram

Ai E k (f ) E ω (f )

Bi Y
commutes, and thus by the very definition of E k+1 (f ), there exists a commutative diagram

Ai E k (f ) E ω (f )

Bi E k+1 (f ) Y
which solves our lifting problem. 
Remark. In all of our situations, the simplicial sets Ai for a small object argument will
have only finitely many non-degenerate simplices. In general, we would have to find a regular
cardinal κ which is larger that the cardinality of any of the Ai ’s appearing in S (this ensures
that all Ai are κ-compact). Then we could continue the above inductive process: for successor
ordinals consider pushouts as before, and for limit ordinals take a colimit as before. At some
point one has defined E κ (f ) which sits in a factorization X → E κ (f ) → Y , and a similar
argument as before will show that the first map is contained in S (if one defines saturated sets
as being closed under colimits over arbitrary ordinals (as opposed to ω) and that the latter
map is contained in χR (S). We chose to not deal with colimits over ordinals in this text since
we will not need it, but it is of course useful to know that the small object argument does
not depend on such size issues.

Lecture 14 – 03.12.2018.
Remark. The factorization described above is functorial: Whenever given a commutative
diagram
f
X Y

f0
X0 Y0
The small object argument in fact provides a commutative diagram
X E ω (f ) Y

X0 E ω (f 0 ) Y0

With the small object argument we can now give a brut-force proof of Proposition 2.17.
Proposition 3.10. The saturated set generated by spine inclusions is not equal to the inner
anodyne maps. Moreover, there exists a composer which is not an ∞-category.
Proof. Consider the inclusion Λ31 → ∆3 . By the small object argument we can factor this
map as a compostion
Λ31 → X → ∆3
 INTRODUCTION TO INFINITY-CATEGORIES 45

such that the first map is in the saturated class generated by the spine inclusions, and the
last map satisfies RLP wrt spine inclusions, and hence wrt the saturated set generated by
the spine inclusions. It follows that X is a composer because also the map ∆3 → ∆0 has the
RLP wrt spine inclusions. If the saturated set generated by the spine inclusions equals the
inner anodyne maps, it would follow that the map X → ∆3 is an inner fibration (as it has
the RLP wrt inner anodyne maps) and we will show that this is not the case.
We claim that the lifting problem

Λ31 X

∆3 ∆3

does not have a solution. One proves this via induction over the filtration on X obtained
from the small object argument using the following observation. Given a pushout

In A

∆n B

of simplicial sets with n ≥ 3, it follows that the image of Λnj ⊆ ∆n in B is not contained in A.
If X is an ∞-category, then the map X → ∆3 were an inner fibration, see Exercise 52, but
we have shown it is not. 

Remark. The proposition also follows from a different argument: Combining Exercise 44
and Lemma 2.68 we see that there exists a composer X which is not an infinity category.
Now the map X → ∆0 is contained in χR ({I n → ∆n }n∈N ) but is not an inner fibration. This
shows that the saturated closure of the spine inclusions cannot be given by the inner anodyne
maps, as else any map that satisfies the RLP with respect to the spine inclusions were also
an inner fibration.
Notice that the sheer fact that the saturated closure of the spine inclusions is not equal to
the inner anodyne maps does not formally imply that there exists a composer which is not
an ∞-category. However, the converse holds: The existence of a composer which is not an
∞-category shows that the saturated set generated by spine inclusions can not contain all
inner anodyne maps.

Remark. We will show in Proposition 3.17 that spine inclusions are inner anodyne, so that
Proposition 3.10 can be restated as saying that the saturated set generated by spine inclusions
is strictly contained in the inner anodyne maps.

Lemma 3.11. Consider a set of morphisms S = {Ai → Bi }i∈I such that all Ai have only
finitely many non-degenerate simplices. Then the saturated closure S of S is given by χ(S).

Proof. Obviously S ⊆ χ(S). In Lemma 3.8 we have shown that χ(S) is itself saturated so
S ⊆ χ(S). To prove the converse, consider a map f : x → y with f ∈ χ(S). By the small
46 M. LAND

object argument Proposition 3.9 we find a factorization of this map

j
x z
f p
α
y y
where j ∈ S and where p satisfies the RLP wrt S. Since f ∈ χ(S) it follows that there exists
a dashed arrow α making the diagram commute. We hence have a commutative diagram
x x x
f j f
α p
y z y
which shows that f is a retract of j. Since j ∈ S, so is f . 
Corollary 3.12. (Inner, left, right) anodyne maps precisely the saturated closure of the
(inner, left, right) horn inclusions. In particular, all of these are monomorphisms.
We have used this already, but let us again state the following fact.
Fact 3.13. A monomorphism is anodyne if and only if its geometric realization is a weak
equivalence. (This is part of the existence of the Kan–Quillen model structure on simplicial
sets, this is the one which is equivalent to the Quillen model structure on topological spaces).

The following corollary is the relative version of Lemma 2.68.

Corollary 3.14. Let F : C → D be a morphism of simplicial categories. Assume that for all
objects X, Y ∈ C, the induced map HomC (X, Y ) → HomD (F X, F Y ) is a Kan fibration. Then
N(F ) : NC → ND is an inner fibration.
Corollary 3.15. Let F : C → D be a functor between ordinary categories. Then the induced
functor N(F ) : NC → ND is an inner fibration.
Proof. Given any map between sets A → B, then the induced map of constant simplicial sets
cA → cB is a Kan fibration. This follows from the fact that for all n ≥ 1 and all 0 ≤ j ≤ n,
the map π0 (Λnj ) → π0 (∆n ) is an isomorphism. Then use that N(C) = N(cC) and the previous
corollary. 
Remark. In fact more holds true: In ?? we will show that a map of simplicial sets X →
N(D), where D is an ordinary category, is an inner fibration if and only if X is an ∞-category.

Recall the definition of the S-horn ΛnS of Definition 1.39, so that we have
[
ΛnS = ∆[n]\{s} .
s∈S
/

Lemma 3.16. Let S ⊆ [n] be a non-empty subset. The map ΛnS → ∆n is

(1) anodyne, provided S 6= [n],
(2) left anodyne, provided {n} ∈/ S,
(3) right anodyne, provided {0} ∈ / S,
(4) inner anodyne, provided S is not the complement of an intervall, i.e. there are a <
b < c ∈ [n] with a, c ∈
/ S but b ∈ S.
 INTRODUCTION TO INFINITY-CATEGORIES 47

Proof. Let S ⊆ [n] be a non-empty subset and let s ∈ S and let S 0 = S \ {s}. Consider the
pushout diagram
∆[n]\{s} ∩ ΛnS ∆[n]\{s}

ΛnS ΛnS 0 ∆n
[n]\{s}
Notice that the top horizontal arrow is a generalized horn inclusion ΛS 0 ⊆ ∆[n]\{s} . It
n n
follows that the inclusion ΛS → ∆ is contained in the smallest saturated set containing the
inclusions
[n]\{s}
ΛS 0 ⊆ ∆[n]\{s} and ΛnS 0 ⊆ ∆n .
We will prove the lemma by an induction over the size of S (for arbitrary [n]). Let us consider
(1). If S contains only one element, say S = {i} we have that ΛnS is a horn inclusion, and
hence anodyne. If S contains more than one element, then S 0 still contains at least one
[n]\{s}
element and is smaller than S, so that ΛnS 0 ⊆ ∆n and ΛS 0 ⊆ ∆n is anodyne by induction.
For (2), Suppose that S = {i}, then the horn inclusion is left anodyne, because i 6= n.
If S contains more than one element, then S 0 is smaller and still does not contain n. The
statement for (3) is similar.
To show (4), if S = {i} then 0 < i < n, else its complement is an interval, so that the
inclusion ΛnS → ∆n is inner anodyne. If S contains more than one element, then we claim that
there exists an element s in S such that S \ {s} is again not the complement of an interval:
By assumption, there are a < b < c such that b ∈ S and a, c ∈ / S. By assumption S 6= {b}, so
choose some other element s ∈ S \ {b}. Then S 0 = \{s} is again not the complement of an
intervall (because b ∈ S 0 ). 
Proposition 3.17. The spine inclusions I n → ∆n are inner anodyne.
Proof. The spine inclusion in : I n → ∆n can be factored as follows:
fn gn
I n −→ ∆[n]\{0} ∪ I n −→ ∆n
We will show by induction on n that both fn and gn are inner anodyne. The induction start
n = 1 and n = 2 is obvious. Now let n ≥ 3 and consider the pushout diagram
I [n]\{0} ∆[n]\{0}

In ∆[n]\{0} ∪ I n
in which the upper composite is inner anodyne by induction. Hence fn is inner anodyne as a
pushout of an inner anodyne map. It remains to show that gn is also inner anodyne. In this
case we consider the pushout diagram
gn−1
∆[n]\{0,n} ∪ I [n]\{n} ∆[n]\{n}

∆[n]\{0} ∪ I n ∆[n]\{0} ∪ ∆[n]\{n} ∆n

in which the top horizontal map is inner anodyne by induction. Hence so is the left lower
horizontal map. The right lower horizontal map is given by Λn[n]\{0,n} ⊆ ∆n , which is anodyne
by Lemma 3.16 part (4) because {0, n} is not an interval. 
48 M. LAND

Corollary 3.18. Every ∞-category is a composer.

Proof. Obvious from Proposition 3.17. 
Definition 3.19. A trivial fibration is a map which has the RLP wrt the boundary inclusions
∂∆n → ∆n for n ≥ 0.

Lecture 15 – 06.12.2018.
Definition 3.20. Let J be the nerve of the category consisting of two objects with a unique
isomorphism between them.

Observation 3.21. The category with two objects and a unique isomorphism between them
is a classifier for isomorphisms in a category: In other words the functor corepresented by
this category is the functor which assigns to a category its set of isomorphisms.
Given a morphism f in an ∞-category X, one can thus wonder when its classifying map
1
∆ → X extends over J. It is easy to see that if this is the case, then f is an equivalence
(Exercise). The converse turns out to be true and will be yet another application of the fact
that ∞-groupoids are Kan complexes: Note that the map ∆1 → J is anodyne.

Definition 3.22. A Joyal fibration between ∞-categories is an inner fibration which in

addition has the RLP wrt the map ∆0 → J.

Construction 3.23. Let f : X → Y and i : A → B be maps of simplicial sets. Then there

is a commutative diagram
XB XA

YB YA
and thus there is an induced map
hf, ii : X B → X A ×Y A Y B .

Construction 3.24. Dually, for morphisms i : A → B and g : S → T , we obtain a commu-

tative diagram
A×S A×T

B×S B×T
and thus there is an induced map
i  g : A × T qA×S B × S → B × T.

Lemma 3.25. The following two lifting problems are equivalent.

S XB A × T qA×S B × S X

T X A ×Y A Y B B×T Y
 INTRODUCTION TO INFINITY-CATEGORIES 49

Proof. 
The crucial technical lemma about the maps i  g is the following.
Lemma 3.26. In the notation of above,
(1) i  g is inner anodyne if i or g is,
(2) i  g is left anodyne if i or g is,
(3) i  g is right anodyne if i or g is,
(4) i  g is anodyne if i or g is.
In order to prove this lemma, we will need the following steps:
Lemma 3.27. Let S and T be a sets of morphisms whose domains are all compact. Then
S  T ⊆ S  T ⊆ S  T . In particular S  T = S  T = S  T .
Proof. The very first inclusion is obvious. To see the second inclusion, we let F = χR (S  T ).
Then S  T = χL (F) by Lemma 3.11. Now consider the set A = {f : f  T ∈ χL (F)}. By
Lemma 3.25 we have that
A = χL (hF, T i)
and thus is a saturated set by a previous exercise. Since S ⊆ A by definition of A, it follows
that S ⊆ A. Thus S  T ⊆ S  T . Then consider the set B = {f : S  f ∈ χL (F)}. As before
we get that
B = χL (hF, Si)
so that B is also a saturated set. We see that T ⊆ B by the previous step, so that also T ⊆ B.
This proves the first part. For the in particular, we argue as follows: Since S  T is saturated
and contains S  T and S  T , we find
S  T ⊆ S  T ⊆ S  T.
On the other hand, S  T ⊆ S  T ⊆ S  T , so the other inclusion also holds. 
Lemma 3.28. For 0 < j < n, the inclusion Λnj → ∆n is a retract of the map
Λnj × ∆2 qΛn ×Λ21 ∆n × Λ21 → ∆n × ∆2 .
j

Proof. For the first part consider the maps

s r
[n] → [n] × [2] → [n]
where 
(i, 0)
 if i < j,
s(i) = (i, 1) if i = j,

(i, 2) if i > j

and where 
i
 if i < j and k = 0
r(i, k) = i if i > j and k = 2

j else

We now have to show that
(1) rs = id,
(2) s(Λnj ) ⊆ Λnj × ∆2 ∪ ∆n × Λ21 , and
(3) r(Λnj × ∆2 ∪ ∆n × Λ21 ) ⊆ Λnj .
50 M. LAND

(1) is an immediate check. To see (2) we observe that in fact s(Λnj ) ⊆ Λnj × ∆2 : For this it
suffices to see that composing s with the projection [n] × [2] → [n] is the identity. To further
see (3) we need to show two things:
(a) r(Λnj × ∆2 ) ⊆ Λnj , and
(b) r(∆n × Λ21 ) ⊆ Λnj .
To prove (a), consider a k-simplex of Λnj , i.e. f : [k] → [n] such that there exists an m ∈ [n] \ j
which is not in the image of f , and an arbitrary k-simplex α : [k] → [2] of ∆2 . The composite
r
[k] → [n] × [2] → [n]
is easily seen to send i ∈ [k] to either f (i) or j. Hence its image is contained in the image of
f union {j}. In particular, m is not in the image of this composite, and thus it represents
a k-simplex of Λnj . To see (b), consider again a general k-simplex β : [k] → [n] of ∆n , and a
k-simplex f : [k] → [2] of Λ21 , i.e. where either 0 or 2 is not in the image of f . For definiteness,
say that 2 is not in the image (the other case works similarly). We find that the composite
(β,f ) r
[k] −→ [n] × [2] → [n]
sends i ∈ [k] to β(i) if β(i) < j and to j else. Thus the image is contained in {0, . . . , j}. Now
since 0 < j < n, we see that n is not in the image, so that the above composite represents
a k-simplex of Λnj . In the case that 0 is not in the image of f , we find that 0 is not in the
image by a similar argument. This proves the lemma. 

Lecture 16 – 10.12.2018. The following is [Lur09, Proposition [Link]].

Lemma 3.29. The following sets of morphisms all generate the set of inner anodyne maps.
(1) The inner horn inclusions S1 = {Λnj → ∆n } for all n ≥ 2,
(2) the maps S2 = {(K → L)  (i : Λ21 → ∆2 )} for all monomorphisms K → L, and
(3) the maps S3 = {(∂∆n × ∆n )  (i : Λ21 → ∆2 )} for all n ≥ 0.
(4) the maps S4 = {(K → L)  (Λnj → ∆n )} for all monomorphisms K → L and all inner
horns.
Proof. Let us introduce some more notation: We denote the set of monomorphisms by T2
and the set of boundary inclusions by T3 . We hence have S2 = T2  i, S3 = T3  i, and
S4 = T2  S1 . It is an exercise that T2 = T3 , see Exercise 60. It thus follows from Lemma 3.27
that
S3 = T3  i = T3  i = T2  i = T2  i = S2 .
Tautologically, we have that S2 ⊆ S4 , since S2 ⊆ S4 . We also find that S1 ⊆ S2 = T2  i,
as any inner horn inclusion is a retract of a map in S2 by Lemma 3.28. Now notice that
T2  T2 = T2 , so that we also obtain

S4 = T2  S1 ⊆ T2  T2  i = T2  T2  i = T2  i = S2
so that S2 = S4 . The lemma is shown once we can show that S3 ⊆ S1 , for which it suffices to
show that S3 ⊆ S1 .
So let m ≥ 0 and consider the inclusion
∆m × Λ21 ∪ ∂∆m × ∆2 ⊆ ∆m × ∆2 .
 INTRODUCTION TO INFINITY-CATEGORIES 51

If m = 0, then this map is given by Λ21 → ∆n and thus is inner anodyne. So assume m ≥ 1.
We will construct a filtration of ∆m × ∆2 as follows. For 0 ≤ i ≤ j < m consider the
(m + 1)-simplices of ∆m × ∆2 given by

(k, 0)
 if 0 ≤ k ≤ i
σi,j (k) = (k − 1, 1) if i + 1 ≤ k ≤ j + 1 .

(k − 1, 2) if j + 2 ≤ k ≤ m + 1


For 0 ≤ i ≤ j ≤ m consider the (m + 2)-simplices of ∆m × ∆2 given by


(k, 0)
 if 0 ≤ k ≤ i
τi,j (k) = (k − 1, 1) if i + 1 ≤ k ≤ j + 1 .

(k − 2, 2) if j + 2 ≤ k ≤ m + 2


We observe that the non-degenerate k-simplices of ∆m × ∆2 correspond to paths of length

k in the grid [m] × [2] which do not take “a break at any point”, i.e. are precisely the injective
maps [k] → [m] × [2]. We claim that
(1) The simplices τi,j are the non-degenerate (m + 2)-simplices of ∆m × ∆2 : Necessarily,
the paths corresponding to non-degenerate (m + 2)-simplices have to start at (0, 0)
and end at (m, 2) in order for there to be an injective map [m + 2] → [m] × [2].
(2) The simplices σi,j are non-degenerate.
(3) The simplex σi,j is a face of τi,j and of τi,j+1 but of no other of the τ ’s just constructed.
(4) The simplices σi,j and τi,j are not contained in ∆m × Λ21 ∪ ∂∆m × ∆2 .
(1) and (2) are obvious from the previous observation. (3) follows immediately from the
definitions, and (4) is also an explicit check: the projection [m]×[2] → [m] sends the simplices
in question to surjections, so that they are not contained in ∂∆m ×∆2 . Likewise, the projection
[m] × [2] → [2] sends the simplices in question to surjections as well, so that they are also not
contained in ∆m × Λ21 .
Let us now inductively define simplicial sets X(j, i) for 0 ≤ i ≤ j < m as follows:
X(0, 0) = ∆m × Λ21 ∪ ∂∆m × ∆2 ,
For fixed j, we inductively define for i ≤ j < m:
X(i + 1, j) = X(i, j) ∪ σi,j
and we set
X(j + 1, j) = X(0, j + 1).
Then we define X(0, m) = Y (0, 0) and again inductively define for i ≤ j ≤ m:
Y (i + 1, j) = Y (i, j) ∪ τi,j
and set
Y (j + 1, j) = Y (0, j + 1).
Because the τi,j are the non-degenerate (m + 2)-simplices of ∆m × ∆2 , we find that Y (0, m +
1) = ∆m × ∆2 .
To finish the proof of the lemma we have to show the following statements:
(A) The simplicial set X(i, j) ∩ σi,j is an inner horn, and
(B) The simplicial set Y (i, j) ∩ τi,j is an inner horn.
52 M. LAND

From this it follows that all maps X(i, j) → X(i + 1, j) are inner anodyne, and that all maps
Y (i, j) → Y (i + 1, j) are inner anodyne: In either case, they are pushouts of inner horn
inclusions. Thus their composite X(0, 0) → Y (0, m + 1) is also inner anodyne, so we can
conclude the lemma.
To show (A), it we have to prove that for all 0 ≤ i ≤ j < m, the map
σi,j
[m + 1] −→ [m] × [2]
sends all m-dimensional faces to X(i, j), unless one inner m-dimensional face. We claim that
the (i+1)-face of σi,j is the one which is not contained in X(i, j). Notice that 0 < i+1 < m+1,
so that σi,j ∩ X(i, j) is an inner horn. There are only two faces of σi,j which do not lie in
∂∆m × ∆2 ⊆ X(0, 0) namely the ones which compose the horizontal edges adjacent to the
unique vertical edge. These are di σi,j and di+1 σi,j , so that it suffices to see that di σi,j is
contained in X(i, j) and that di+1 σi,j is not contained in X(i, j). To see the former, observe
that di (σi−1,j ) = di (σi,j ), provided i > 0 and that d0 (σ0,j ) is contained in ∆m × Λ21 , since after
projecting to [2], 0 is not in the image. It remains to prove the latter, namely that di+1 (σi,j )
is not contained in X(i, j). We have already seen that di+1 σi,j is not contained in ∂∆m × ∆2 .
We hence need to show
(1) di+1 (σi,j ) is not contained in ∆m × Λ21 ,
(2) di+1 (σi,j ) is not contained in σk,j for k < i, and
(3) di+1 (σi,j ) is not contained in σk,l for k ≤ l < j.
(1) follows from the fact that 0 < i+1 < m+1, which says that after projecting [m]×[2] → [2],
0 and 2 are in the image of di+1 σi,j ). For (2) we observe that the path corresponding to
di+1 (σi,j ) runs through the spot (i, 0), which is not the case for σk,j for k < i, and hence also
not for any face of it. Likewise, if i < j then di+1 σi,j runs through the spot (j, 1) which is
not the case for any σk,l with l < j. If i = j, then di+1 σi,j runs through again through (i, 0)
which is not the case for σk,l as k ≤ l < j = i, so that k < i as in the first case.
It remains to prove (B). Again, we claim that di+1 τi,j is the only face not contained in
Y (i, j). We first consider the case i < j:
Observe again that d` τi,j is contained in ∂∆m × ∆2 unless ` ∈ {i, i + 1, j + 1, j + 2}, and
that dj+1 τi,j = σi,j−1 and dj+2 τi,j = σi,j , so that they are contained in Y (i, j). Likewise,
if i > 0, then di τi,j = di τi−1,j , so that this face is also contained in Y (i, j). If i = 0, then
di τi,j is contained in ∆m ∪ Λ21 as its projection to [2] does not have 0 in its image. We now
show that di+1 τi,j is not contained in Y (i, j), so that again Y (i, j) ∩ τi,j is an inner horn since
0 < i + 1 < m + 2. As before we need to see that
(1) di+1 τi,j is not contained in τk,j for k < i, and
(2) di+1 τi,j is not contained in τk,l for k ≤ l < j.
As before, di+1 τi,j runs through the spot (i, 0) which is not the case for τk,j if k < i. Likewise,
di+1 τi,j runs through the spot (j, 1) which is not the case for τk,l if l < j.
Finally, we consider the case i = j, and claim that again di+1 τi,i is the only face not
contained in Y (i, i): If i = j = 0, then d0 τ0,0 is contained in ∆m × Λ21 , because its projection
to [2] misses 0. If i > 0, we have that di τi,i = di τi−1,i so di τi,i is contained in Y (i, i) in all
cases. Likewise, we have that di+2 τi,i = σi,i . If l < i or l > i + 2, then the projection to [m]
of dl τi,i is not surjective and thus dl τi,i is contained in ∂∆m × ∆2 . It remains to show that
(1) di+1 τi,i is not contained in τk,j for k < i, and
(2) di+1 τi,i is not contained in τk,l for k ≤ l < j.
 INTRODUCTION TO INFINITY-CATEGORIES 53

Note that in both cases, k < i and that di+1 τi,i runs through the spot (i, 0). But τk,l does
not run through (i, 0), no matter what l is. 
Lecture 17 – 17.12.2018.
Lemma 3.30. The following sets of morphisms all generate the set of left anodyne maps
(1) The left horn inclusions S1 = {Λnj → ∆n } for all n ≥ 1 and 0 ≤ j < n,
(2) the maps S2 = {(K → L)  (i : {0} → ∆1 )} for all monomorphisms K → L, and
(3) the maps S3 = {(∂∆n × ∆n )  (i : {0} → ∆1 ))} for all n ≥ 0.
(4) the maps S4 = {(K → L)  (Λnj → ∆n )} for all monomorphisms K → L, all n ≥ 1
and 0 ≤ j < n.
Proof. As in the proof of Lemma 3.29, the only thing which does not follow from previous
considerations are the following two statements:
(1) The maps in S3 are left anodyne, and
(2) the map Λnj → ∆n is a retract of the pushout product map
∆n × {0} qΛnj ×{0} Λnj × ∆1 → ∆n × ∆1 .
Both these statements are an exercise, see Exercise 63, we give the following hints: For (1)
consider a similar filtration of ∆n × ∆1 starting with the domain of the pushout product by
adding the missing non-degenerate (n + 1)-simplices in ∆n × ∆1 and run the same argument
as in Lemma 3.29. For (2), consider the following maps:
r s
[n] → [n] × [1] → [n]
where r is the inclusion k 7→ (k, 1) and

k
 if k 6= j + 1 and i = 0
s(k, i) = j if (k, i) = (j + 1, 0)

k if i = 1.



Corollary 3.31. The following sets of morphisms all generate the set of right anodyne maps
(1) The right horn inclusions S1 = {Λnj → ∆n } for all n ≥ 1 and 0 < j ≤ n,
(2) the maps S2 = {(K → L)  (i : {1} → ∆1 )} for all monomorphisms K → L, and
(3) the maps S3 = {(∂∆n × ∆n )  (i : {1} → ∆1 ))} for all n ≥ 0.
(4) the maps S4 = {(K → L)  (Λnj → ∆n )} for all monomorphisms K → L, all n ≥ 1
and 0 < j ≤ n.
Proof. We observes that a map is right anodyne if and only if its opposite if left anodyne,
because a map is a left fibration if and only if its opposite is a right fibration according to
Exercise 52. Then the first part follows from the fact that (Λnj )op ∼
= Λnn−j and that all other
morphisms of simplicial sets involved are “self-opposite”. 
Proof of Lemma 3.26. Part (1) follows from the equality S4 = S1 of Lemma 3.29, Part (2)
follows from the equality S4 = S1 of Lemma 3.30, and Part (3) follows from the equality
S4 = S1 of Corollary 3.31. To see Part (4) we simply observe that anodyne maps are generated
(as a saturated set) by left and right anodyne maps. It hence suffices to show that for i a left,
respectively right, anodyne map and g a monomorphism, then i  g is anodyne. Since left,
54 M. LAND

respectively right, anodyne maps are anodyne, this follows from statement (2), respectively
(3).

Theorem 3.32. Let f : X → Y be a (inner, left, right) fibration and let i : A → B be a
monomorphism. Then
(1) the map hf, ii is a (inner, left, right) fibration.
(2) If furthermore i is (inner, left, right) anodyne, then the map hf, ii is a trivial fibration.
Proof. We prove (1) first. Consider a lifting problem
S XB
g hf,ii

T X A ×Y A Y B
in which the map g : S → T is (inner, left, right) anodyne. By Lemma 3.25 solving this is
equivalent to solving the lifting problem
A × T qA×S B × S X
ig f

B×T Y
By Lemma 3.26 we know that i  g is (inner, left, right) anodyne, and by assumption f is an
(inner, right, left) fibration.
To prove (2), we need to consider again a lifting problem as above, where now g is a
monomorphism. Since i is (inner, left, right) anodyne, we again see that by Lemma 3.26 we
can solve this lifting problem. 
Corollary 3.33. Let K and X be simplicial sets. If X is an ∞-category, then so is X K . If
X is a Kan complex, then so is X K .
Proof. This is the special case of Theorem 3.32 part (1) where A is empty, B = K and
Y = ∆0 . 
Definition 3.34. Let C and D be ∞-categories. We define the ∞-category of functors from
C to D by Fun(C, D) = Hom(C, D), which is an ∞-category thanks to Corollary 3.33.

Observation 3.35. Notice again that the 0- and 1-simplices of Fun(C, D) are functors and
natural transformations as defined in Definition 2.15 and Definition 2.76.
If moreover D is a Kan complex, then Fun(C, D) is itself a Kan complex and hence in
particular an ∞-groupoid. This, together with the (to be proven) fact that ∞-groupoids
are Kan complexes, suggests that in the functor category, a morphism which is pointwise an
equivalence is in fact an equivalence. This will turn out to be true, but even more complicated
then showing that Kan complexes are precisely the ∞-groupoids, see Theorem 6.1.

We are now also in position to give a new definition of the ∞-category of spaces, which is
purely simplicial:
Definition 3.36. We consider the simplicial category Kan with objects given by Kan
complexes X and with hom simplicial sets from X to Y given by the internal hom simplicial
set Hom(X, Y ). We let An
c = N(Kan) be its coherent nerve.
 INTRODUCTION TO INFINITY-CATEGORIES 55

Observation 3.37. We claim that there is a canonical functor An → An c which is induced

by the following simplicial functor: Recall that An is the coherent nerve of the simplicial
category of CW-complexes as in Definition 2.70. We claim that sending a CW complex X to
its singular complex S(X) induces a simplicial functor as needed. For this we need to show
that there is a canonical map of simplicial sets
S(map(X, Y )) → Hom(S(X), S(Y ))
which is compatible with composition. This map is constructed as follows: By adjunction it
suffices to construct a map
S(map(X, Y )) × S(X) → S(Y ).
Using that S (as a right adjoint) commutes with products, this is equivalently provided by a
canonical map
S(map(X, Y ) × X) → S(Y )
and we use the continuous evaluation map map(X, Y ) × X → Y and apply the functor S to
it. It is not hard to see that the map obtained in this fashion is compatible with composition,
and hence determines a functor An → An c as claimed.
Furthermore, we claim that the map
S(map(X, Y )) → Hom(S(X), S(Y ))
is a weak equivalence, so that the simplicial functor from CW-complexes to Kan is a weak
equivalence in the sense of Definition 2.49. We will prove later that weak equivalences of
simplicial functors induce equivalences of ∞-categories upon applying the coherent nerve, so
that both ∞-categories of spaces we have defined are in fact equivalent.

Corollary 3.38. A simplicial set C is an ∞-category if and only if the canonical map
2 2
C∆ → CΛ1 is a trivial fibration. In particular, for an ∞-category C the fibre over a fixed
diagram Λ21 → C is a contractible Kan complex.
Proof. The only if part follows from the fact that Λ21 → ∆2 is inner anodyne, C → ∗ is an
inner fibration, and Theorem 3.32. To see the converse, we wish to show that C → ∗ is an
2 2
inner fibration if C∆ → CΛ1 is a trivial fibration.
To show that C → ∗ is an inner fibration, by Lemma 3.29 it suffices to show that it admits
2 2
the extension property for maps in S2 . By adjunction this holds if and only if C∆ → CΛ1
satisfies the extension property for all monomorphisms, which is the case if and only if it is a
trivial fibration.
For the in particular, consider two composable morphisms f and g in C, and view them as
a map Λ21 → C. Then in the pullback diagram

CompC (f, g) Fun(∆2 , C)

∆0 Fun(Λ21 , C)

the right vertical map is a trivial fibration, so the left vertical map is also a trivial fibration. In
particular, for any two composable morphisms, the simplicial set CompC (f, g) of compositions
is a contractible Kan complex. 
56 M. LAND

n n
Remark. The same argument shows that for an ∞-category C, the map C∆ → CI is a
trivial fibration for all n ≥ 2, because the maps I n → ∆n are inner anodyne.

Lemma 3.39. Any trivial fibration X → Y admits a section.

Proof. Consider the diagram
∅ X

id
Y Y
and use that trivial fibrations have the RLP wrt monomorphisms, of which ∅ → Y is an
example. 
We can thus choose a section of the above trivial fibration and obtain the following com-
posite
2 2 {0,2}
CΛ1 → C∆ → C∆
as functors of ∞-categories (i.e. maps of simplicial sets). Since the first category is equivalent
1 1
to C∆ ×C C∆ where the maps are target and source, we obtain a functor which encodes
composition in the ∞-category C:
1 1 1
C∆ ×C C∆ → C∆ .
Definition 3.40. Let C be an ∞-category and let x, y ∈ C be objects. Then we define the
mapping ∞-category between x and y to be the pullback
mapC (x, y) Fun(∆1 , C)

(x,y)
∗ C×C
where the right vertical map is source and target (i.e. evaluation at 0 and 1).

Notice that it is easy to see that mapC (x, y) is an ∞-category, since the map Fun(∆1 , C) →
C×C is an inner fibration by Theorem 3.32. In fact more is true, but the proof of the following
proposition has to be deferred to a later point.
Proposition 3.41. If C is an ∞-category, then for all objects x, y ∈ C, mapC (x, y) is an
∞-groupoid.
Taking it for granted for the moment, we thus obtain a functor
mapC (x, y) × mapC (y, z) → Fun(∆1 , C) ×C Fun(∆1 , C) → Fun(∆1 , C)
which makes the diagram
mapC (x, y) × mapC (y, z) Fun(∆1 , C)

(x,z)
∗ C×C
commute. We hence obtain a functor
mapC (x, y) × mapC (y, z) → mapC (x, z)
which we refer to as the composition in the ∞-category C.
 INTRODUCTION TO INFINITY-CATEGORIES 57

4. Joins and slices

Lecture 18 – 20.12.2018.
Definition 4.1. Let C and D be categories. Then the join C ? D is given by the following
category:
Ob(C ? D) = Ob(C) q Ob(D)
and hom sets are given by
x, y ∈ C


HomC (x, y) if
x, y ∈ D

Hom (x, y)
D if
HomC?D (x, y) = .
∗
 if x ∈ C, y ∈ D

∅ if x ∈ D, y ∈ C


Remark. Notice that the join is not symmetric.

Definition 4.2. Let C be a category, and let x ∈ C be an object. Then there are categories
C/x and Cx/ of objects over and under x, called slice categories. Objects are given by mor-
phisms y → x for C/x and x → y for Cx/ . Morphisms between to such objects are given by
commutative triangles.

Remark. If C is a cocomplete category and x an object of C, then the slice Cx/ of objects
under x is again cocomplete. However, the forgetful functor Cx/ → C does not preserve all
colimits, see ??.

Definition 4.3. Given a linearly ordered set J, we define the set of cuts of J, Cut(J) as
decompositions of J = J1 q J2 into two disjoint pieces J1 and J2 such that x < y whenever
x ∈ J1 and y ∈ J2 . The half empty cuts (∅, J) and (J, ∅) are allowed.

Lemma 4.4. Given linearly ordered sets J and J 0 with a map of such α : J → J 0 , and given
(J10 , J20 ) ∈ Cut(J 0 ), there exists a unique cut (J1 , J2 ) ∈ Cut(J) such that α restricts to maps
α1 : J1 → J10 and α2 : J2 → J20 .
Proof. We need to define Ji = α−1 (Ji0 ). The only thing to check is that this is in fact a cut
of J. This follows since α is order preserving. 
Observation 4.5. This implies that Cut(−) is a contravariant functor from linearly ordered
sets to sets. In fact, Cut(−) is representable by [1].

Definition 4.6. Let X and Y be simplicial sets. Define their join X ? Y to be the simplicial
set given by the following: For a finite linearly ordered set J, we set
a
(X ? Y )(J) = X(J1 ) × Y (J2 )
(J1 ,J2 )∈Cut(J)

where we declare that X(∅) = ∗ = Y (∅). Let α : J → J 0 be a morphism of linearly ordered

sets and consider a cut of J 0 . Consider the associated cut of J as in Lemma 4.4. Then there
is a map
X(J10 ) × Y (J20 ) → X(J1 ) × Y (J2 ).
58 M. LAND

This provides a unique map

(X ? Y )(J 0 ) → (X ? Y )(J)
restricting to the above on each cut of J. This makes X ? Y a simplicial set.

Example 4.7. Given two ordinary categories C and D, we have that

N(C ? D) ∼
= N(C) ? N(D).
In particular, we have that ∆n ? ∆m = ∆n+1+m .

Proof. We construct a map of simplicial sets

N(C ? D) → N(C) ? N(D)

by observing that any n-simplex in N(C ? D) determines a cut of [n]: at some point one
jumps from morphisms in C to morphisms in D. It is then easy to see that this map is an
isomorphism. 

Lemma 4.8. Given a simplicial set X, the join construction determines a functor X ?
− : sSet → sSetX/ , likewise it produces a functor − ? X : sSet → sSetX/ .

Proof. We need to show that for every Y ∈ sSet, the simplicial set X ? Y comes equipped
with a map X → X ? Y . This is obviously the case by the right half empty cut inclusion
a
X(J) × Y (∅) ⊆ X(J1 ) × Y (J2 ) = (X ? Y )(J).
(J1 ,J2 )∈Cut(J)

Furthermore, given a morphism Y → Y 0 of simplicial sets, we claim that the canonical diagram

X X ?Y

X ?Y0
commutes. This is immediate, and functoriality is also obvious. 

Lemma 4.9. Let K be a simplicial set equipped with a map p : K → ∆1 . Then there is a
functorial factorization into a composite
c
K → K0 ? K1 → ∆1

where Ki = p−1 (i) and the map c is the map K0 ? K1 → ∆0 ? ∆0 ∼

= ∆1 .
Proof. We need to construct the map K → K0 ? K1 . For any n ≥ 0, we have that

HomsSet (∆n , ∆1 ) = Hom([n], [1]) = Cut([n]).

Thus, for every n-simplex x : ∆n → K, the composite px : ∆n → K → ∆1 determines a cut

([i], [j]) ∈ Cut([n]), so that the map px : [n] → [1] sends the first i points to 0 and the rest to
1. This, by definition, determines a point of K0 ([i]) × K1 ([j]) which in turn determines an
n-simplex of K0 ? K1 .
 INTRODUCTION TO INFINITY-CATEGORIES 59

It is then easy to see that this in fact determines a map of simplicial sets K → K0 ? K1
and that this construction is functorial in sSet/∆1 . Concretely, for a commutative triangle

K ∆1

K0
the induced diagram
K K0 ? K1 ∆1

K0 K00 ? K10
commutes as well. 
Corollary 4.10. Given a map ϕ : K → X ? Y over ∆1 , i.e. a morphism in sSet/∆1 , there
f ?g
is a factorization into K → K0 ? K1 −→ X ? Y .
Proof. We observe that the factorization provided by Lemma 4.9 for the map X ? Y → ∆1 is
given by X ? Y → X ? Y → ∆1 . Since this factorization is functorial we obtain a commutative
diagram
ϕ
K X ?Y
∼
=
f ?g
K0 ? K1 X ?Y
so the claim follows. 
Proposition 4.11. X and Y are ∞-categories if and only if X ? Y is an ∞-category.
Proof. Consider a map
Λnj → X ? Y
for n ≥ 2 and 0 < j < n. We want to show that it extends over ∆n . We can post compose
this map with the canonical map X ? Y → ∆1 and obtain a factorization
Λnj → (Λnj )0 ? (Λnj )1 → X ? Y.
There are several possibilities for what this first map is: First recall that any map Λnj → ∆1
factors uniquely over ∆n (since ∆1 is the nerve of category and Λnj is an inner horn). There
are three cases we will consider now:
(1) The map Λnj → ∆1 is constant at 0,
(2) The map Λnj → ∆1 is constant at 1,
(3) The map Λnj → ∆1 is not constant.
In the first case, we find that the map Λnj → X ? Y factors through X → X ? Y , and thus
can be extended over ∆n if and only if X is an ∞-category: If an extension of the composite
Λnj → X → X ? Y
to ∆n exists, then the composite ∆n → X ? Y → ∆1 is constant at 0, so that the map
∆n → X ? Y in fact factors through the inclusion X → X ? Y . Similarly, in the second case
60 M. LAND

we find that the map Λnj → X ? Y factors through Y → X ? Y , and thus can be extended
over ∆n if and only if Y is an ∞-category.
Lastly, let us consider the case where the map Λnj → ∆1 is not constant. Observe that
this map factors uniquely through a non-constant map ∆n → ∆1 . The non constant maps
correspond precisely to the non half-empty cuts of [n], so there is a 0 ≤ k < n such that
the map ∆n → ∆1 is isomorphic to the canonical map ∆k ? ∆` → ∆1 . It follows that (Λnj )0
consists of all those m-simplices of Λnj , which are represented by maps [m] → [n] whose image
is contained in {0, . . . , k} (so that it lies in the fibre over 0) and such that there exists a number
different from j which is not in the image of [m] → [n] (so that it lies in the horn). Since
k < n, we can be sure that n does not lie in the image of the map [m] → [n] representing
an m-simplex of (Λnj )0 . In other words, we find that (Λnj )0 ∼ = ∆k . Likewise, we find that
(Λnj )1 ∼
= ∆` . The factorization of Corollary 4.10 hence reads as
Λnj → ∆k ? ∆` → X ? Y
which is the desired extension. The Proposition follows. 
Lecture 19 – 14.01.2019. We now want to come to the construction of slice categories of
∞-categories. For this we observe that if the functor S ? − : sSet → sSetS/ admits a right
adjoint sSetS/ → sSet
(p : S → X) 7→ Xp/
we obtain that a map from Y → Xp/ is the same thing as a map S ? Y → X in sSetS/ .
Specializing to Y = ∆n we obtain a simplicial set:
Definition 4.12. For p : S → X, the association n 7→ HomsSetS/ (X ? ∆n , S) determines a
simplicial set which we call Xp/ .

Lemma 4.13. If an ordinary category C is (co)complete and x ∈ C is an object, then Cx/

and C/x are (co)complete as well. The forgetful map Cx/ → C preserves limits and connected
colimits (i.e. colimits indexed over connected categories), and the forgetful map C/x → C
preserves colimits and connected limits.
Proof. We first observe that C/x ∼ op
= (C )op , so that it suffices to treat the case of colimits. We
x/
now show that the forgetful map C/x → C preserves colimits: Consider a diagram F : I → C/x .
The colimit of the underlying diagram I → C/x → C canonically comes with a map to x, and
it is easy to see that this produces a colimit of F .
The case of colimits of a diagram F : I → Cx/ is slightly more complicated. We observe that
any such diagram is equivalently given by a diagram G : I / → C whose restriction to the cone
point is given by the object x. We observe that there are canonical functors ∆0 → I / ← I
which are inclusions. In particular we have a canonical map
x = colim G|∆0 → colim
/
G.
∆0 I

We claim that this morphism is a colimit of F : Suppose given a functor F̄ : I . → C/x , i.e.
a compatible family of maps F (i) → (x → y) in Cx/ . We wish to show that there exists a
unique map
colim
/
G→y
I
compatible with both maps from x. This comes from the observation that G|I = F , so that
for i ∈ I, there is a canonical map G(i) = F (i) → y, and for the cone point, we have G(∗) = x,
 INTRODUCTION TO INFINITY-CATEGORIES 61

so that there is a canonical map to y as well. These are compatible since F takes values in
the slice Cx/ . This shows that F admits a colimit, namely the map
x → colim
/
G
I
described above.
To finish the proof of the lemma, we need to show that if I is connected, then the canonical
map
colim F → colim
/
G
I I
is an isomorphism, so that the functor Cx/ → C preserves connected colimits. We first
construct a canonical map in the other direction: For this it suffices to construct a map
G(j) → colim F for j ∈ I / , compatible in j. If j ∈ I, then G(j) = F (j) so that there is a
I
canonical map to colim F . So we need to construct a map x = G(∗) → colim F . For this we
I I
choose any object i ∈ I, and get a map
x = G(∗) → G(i) = F (i) → colim F
I
as wanted. We need to show that these maps assemble into a map
colim
/
G → colim F.
I I
In other words, we need to show that for any morphism in I / , the corresponding triangle
commutes. It suffices to treat morphisms of the form ∗ → j for some j ∈ I (for morphisms
in I, it holds by construction). Concretely we need to show that for any two objects i, j ∈ I,
the two maps
x = G(∗) → G(i) = F (i) → colim F
I
and
x = G(∗) → G(j) = F (j) → colim F
I
are the same maps. This is were the assumption that I is connected enters: We find a sequence
of morphisms connecting i to j in I. By induction on the length, we may assume that the
length is one, so that there is in fact a map i → j in I. In this case, it follows from the
property of colimits that the triangle
F (i) colim F
I

F (j)
commutes. On the other hand, by assumption on F , the triangle
x F (i)

F (j)
also commutes, so that the claim is shown.
By construction, the composite
colim F → colim
/
G → colim F
I I I
62 M. LAND

is the identity. The other composite gives a map

colim
/
G → colim F → colim
/
G
I I I

whose restriction to G(j) → colim

/
G for j ∈ I is the canonical map, and whose restriction
I
to the cone point is given by x → G(i) → colim
/
G. This map is the canonical map G(∗) →
I
colim G, so that the above composite is also the identity. 
Lemma 4.14. The functors S ? − and − ? S : sSet → sSetS/ preserve colimits.
Proof. It suffices to check that it preserves coequalizers, which are calculated underlying by
Lemma 4.13, and that it preserves coproducts. For the latter, we recall that the coproduct
(S → A) q (S → B) in sSetS/ is given by the canonical map to the pushout S → A qS B in
sSet, see again Lemma 4.13 for the description of colimits in such slices. Then observe that
a a a a
(S ∗ ( Ai ))n = Sn q (Ai )n q Sk × (Ai )l
i∈I i∈I k+l=n−1 i∈I

whereas a a a 
(S ∗ Ai )n = Sn q (Ai )n q Sk × (Ai )l / ∼
i∈I i∈I k+l=n−1
`
where the relation identitfies I Sn to Sn . For coequalizers, the statement follows similarly
from the explicit description of the simplices of the join. 
Corollary 4.15. The functors S ? − and − ? S : sSet → sSet preserve pushouts.
Proof. The functors S ? − and − ? S : sSet → sSetS/ preserve all colimits by Lemma 4.14 and
the forgetful functor sSetS/ → sSet preserves connected colimits by Lemma 4.13. Pushouts
are connected colimits, so the corollary follows. 
Corollary 4.16. The functor (p : S → X) 7→ Xp/ is right adjoint to S ? − : sSet → sSetS/ .
Likewise, the functor (p : S → X) 7→ X/p is right adjoint to − ? S : sSet → sSetS/ .
Proof. By definition, the adjunction property holds for representables. By Lemma 4.14, the
functors S ? − and − ? S preserve colimits, so that the adjunction bijection prolongs from
representables to all simplicial sets. 
Example 4.17. Let C be an ∞-category and x ∈ C an object, which we view as a functor
x : ∆0 → C. We obtain slices Cx/ and C/x . For a general simplicial set K we will write
K / = ∆0 ? K and K . = K ? ∆0 and call these constructions cone and cocone over K.

Observation 4.18. Let us spell out explicitly the unit and counit of the slice/join adjunc-
tion. For a fixed simplicial set S, the counit of the adjunction is given by a natural map as
follows: Let p : S → X be an object of sSetS/ , so we obtain the slice Xp/ . The counit is then
the map S ? Xp/ → X in sSetS/ given by
a
Sn q HomsSetS/ (S ? ∆n , X) q Sk × HomsSetS/ (S ? ∆l , X) → Xn
k+l=n−1

which is given by p in the first component, induced by precomposition with ∆n → S ? ∆n

on the second component, and induced by precomposition with ∆k ? ∆l → S ? ∆l for each
k-simplex of S on the last component.
 INTRODUCTION TO INFINITY-CATEGORIES 63

Likewise, the unit is the map X → (S ? X)can/ , where can : S → S ? X is the canonical
map. It is given by joining an n-simplex of X with S, which produces a map
Xn ∼
= HomsSet (∆n , X) → HomsSetS/ (S ? ∆n , S ? X).

Definition 4.19. Let C be an ordinary category. We define a new category Tw(C), the
twisted arrow category of C as follows: Objects are the morphisms of C. A morphism in
Tw(C) from f 0 : x0 → y 0 to f : x → y is given by a commutative diagram
f
x y
α β
f0
x0 y0

We also write that the pair (α, β) is a morphism from f 0 to f = βf 0 α. Composition is obtained
by glueing together such diagrams.

Lecture 20 – 17.01.2019.
Lemma 4.20. The slice construction induces a functor Tw(sSet) → sSet. In particular, for
i ϕ f
A→B→X→Y
there is an induced map
Xϕ/ → Xϕi/ ×Yf ϕi/ Yf ϕ/ .
The same holds true for the other slice.
Proof. The objects of Tw(sSet) are given by maps p : S → X of simplicial sets, and such an
object is sent to X/p . We need to argue how this is functorial in morphisms of the twisted
arrow category, i.e. we need to produce a canonical map
Xϕ/ → Yf ϕi/
as the pair (i, f ) is a morphism from ϕ to f ϕi in Tw(sSet).
First, we construct maps Xϕ/ → Xϕi/ and Xϕ/ → Yf ϕ/ which correspond the the mor-
phisms
ϕi fϕ
A X B Y
i f
ϕ ϕ
B X B X
in Tw(sSet). Using those constructions, we similarly obtain maps
Xϕi/ → Yf ϕi/ ← Yf ϕ/
and we will then show that the diagram
Xϕ/ Yf ϕ/

Xϕi/ Yf ϕi/
64 M. LAND

commutes. This is already part of functoriality in the twisted arrow category because the
pair (i, f ) satisfies
(i, id) ◦ (id, f ) = (i, f ) = (id, f ) ◦ (i, id)
as the following diagrams show:
f ϕi f ϕi
A Y A Y
f i
ϕi fϕ
A X B Y
i f
ϕ ϕ
B X B X
The map Xϕ/ → Xϕi/ is adjoint to a map A ? Xϕ/ → X under A which we define to be
the canonical composite
A ? Xϕ/ → B ? Xϕ/ → X
consisting of the map induced by i and the counit of the adjunction. Likewise, the map
Xϕ/ → Xf ϕ/ is adjoint to a map B ? Xϕ/ → Y under B which we define to be the composite
B ? Xϕ/ → X → Y
consisting of the counit, followed by f .
To see that the diagram
Xϕ/ Yf ϕ/

Xϕi/ Yf ϕi/
commutes, we observe that both composites are adjoint to the map
A ? Xϕ/ → B ? Xϕ/ → X → Y.
It is then easy to see that this construction is functorial in Tw(sSet). For the other slice,
the argument is similar. 
Lemma 4.21. The slice/join adjunction induces a bijection of lifting problems between
diagrams of the kind
S Xϕ/

T Xϕi/ ×Yf ϕi/ Yf ϕ/

and diagrams of the kind
ϕ

B A ? T qA?S B ? S X

B?T Y
Proof. Exercise 71. 
 INTRODUCTION TO INFINITY-CATEGORIES 65

An analogue of Lemma 3.26 and Theorem 3.32 holds for joins and slices in place of product
and mapping simplicial sets.
Lemma 4.22. Let i : A → B and g : S → T be monomorphisms. Then the induced map
?g : A ? T qA?S B ? S → B ? T
iˆ
is a monomorphism and in addition satisfies that it is
(1) inner anodyne if i is right anodyne or g is left anodyne,
(2) is left anodyne if i is left anodyne,
(3) is right anodyne if g is right anodyne.
Proof. For (1) let us prove the case where i is right anodyne. We claim that the set which
contains all monomorphisms i : A → B such that the map iˆ?g is inner anodyne (for any
monomorphism g : S → T ) is a saturated class: This is because it is the set of all monomor-
phisms which has the LLP wrt morphisms of the form
Xϕ/ → Xϕi/ ×Yf ϕi/ Yf ϕ/
for an inner fibration f : X → Y and an arbitrary map ϕ : T → X. It hence suffices to show
that the horn inclusions Λnj → ∆n for 0 < j ≤ n are in this set. Now we claim that the
set of monomorphisms g : S → T such that the map (Λnj → ∆n )ˆ?g is inner anodyne is also
a saturated set. It hence suffices to finally prove that the claim holds for g the boundary
inclusions ∂∆m → ∆m . In this case we have to see that
Λnj ? ∆m ∪ ∆n ? ∂∆m → ∆n ? ∆m
is inner anodyne. This follows from Exercise 72, which shows that the former is given by
Λn+1+m
j which is now an inner horn because j ≤ n < n + 1 + m.
The case where g is left anodyne follows from a similar calculation using that
∂∆m ? ∆n ∪ ∆m ? Λnj → ∆m+1+n
is isomorphic to the inclusion Λm+1+n
m+1+j → ∆
m+1+n and 0 ≤ j < n, so that this is again an

inner horn.
Let us now prove (2). Using the same reduction arguments as before, it suffices to treat the
case where i : Λnj → ∆n with 0 ≤ j < n and where g : ∂∆m → ∆m is the boundary inclusion.
Then we get, as before that the map iˆ?g is given by Λn+1+m
j → ∆n+1+m with 0 ≤ j < n
which is clearly a left anodyne map. The case (3) is analogues. 
i ϕ f
Theorem 4.23. Let A → B → X → Y be composable maps and assume that i is a
monomorphism and that f is an inner fibration.
(1) The induced map
Xϕ/ → Xϕi/ ×Yf ϕi/ Yf ϕ/
is a left fibration.
(2) If the map f : X → Y is a left fibration, then also the induced map
X/ϕ → X/ϕi ×Y/f ϕi Y/f ϕ
is a left fibration.
(3) If the map i : A → B is right anodyne, then the map
Xϕ/ → Xϕi/ ×Yf ϕi/ Yf ϕ/
is a trivial fibration.
66 M. LAND

(4) If the map f : X → Y is a trivial fibration, then the map

Xϕ/ → Xϕi/ ×Yf ϕi/ Yf ϕ/
is a trivial fibration.
Proof. Again we consider a general lifting problem
S Xϕ/

T Xϕi/ ×Yf ϕi/ Yf ϕ/

This lifting problem is equivalent (Exercise) to the following lifting problem

ϕ

B A ? T qA?S B ? S X

B?T Y
To prove (1) we thus need to see the the left vertical map is inner anodyne provided S → T
is left anodyne. This follows from Lemma 4.22 part (1). To prove (2) we one needs that the
map
S ? B qS?A T ? A → T ? B
is left anodyne provided the map S → T is left anodyne. This is the content of Lemma 4.22
part (2). To prove (3) we need to observe that if S → T is a monomorphism and A → B is
right anodyne, then the map A ? T qA?S B ? S → B ? T is inner anodyne, again by Lemma 4.22
part (1). To prove (4) we only need to use that the map A ? T qA?S B ? S → B ? T is always
a monomorphism and that trivial fibrations satisfy the RLP wrt monomorphisms. 
Let us spell out some explicit special cases:
Corollary 4.24. Suppose given maps A → B → X → Y as before where A → B is a
monomorphism and where X → Y is an inner fibration.
(1) If Y = ∗, so that X is an ∞-category we get that Xϕ/ → Xϕi/ is a left fibration, and
that X/ϕ → X/ϕi is a right fibration. In particular if A = ∅, then the map Xϕ/ → X
is a left fibration and X/ϕ → X is a right fibration. In particular, Xϕ/ and X/ϕ are
∞-categories if X is.
(2) If Y = ∗ so that X is an ∞-category, we see that Xϕ/ → Xϕi/ is a trivial fibration if
A → B is right anodyne and that X/ϕ → X/ϕi is a trivial fibration if A → B is left
anodyne.
(3) If f : X → Y is a trivial fibration, consider the case where A = ∅. Then the map
Xϕ/ → X ×Y Yf ϕ/ is a trivial fibration. Furthermore the map X ×Y Yf ϕ/ → Yf ϕ/ is
a pullback of X → Y and thus also a trivial fibration.
Yet another special case of this is the following. Suppose C is an ∞-category and that
f : x → y is a morphism in C. We can consider the situation ∆0 → ∆1 → C → ∗ and obtain
maps
C/x ← C/f → C/y
 INTRODUCTION TO INFINITY-CATEGORIES 67

corresponding to the two restrictions of f to ∆0 (likewise for the other slice). Since the
inclusion {0} → ∆1 is left anodyne, it follows from Corollary 4.24 part (2) that the map
Cf / → Cx/ is a trivial fibration so that we can choose a section to obtain a composite
C/x → C/f → C/y
which we informally think of as the functor of post composition with f . The same works for
the other slice to obtain a functor Cy/ → Cx/ which we think of as pre composition with f .

5. Joyal lifting and applications

Lecture 21 – 21.01.2019.
Definition 5.1. A functor F : C → D between ∞-categories is called conservative if it
detects equivalences, i.e. if whenever f : x → y is a morphism in C such that F (f ) : f x → f y
is an equivalence in D, then f itself is an equivalence.

Observation 5.2. A functor F : C → D is conservative if and only if its opposite functor

F op : Cop → Dop is conservative.

Proposition 5.3. Left and right fibrations between ∞-categories are conservative.
Proof. By passing to opposite categories it suffices to treat the case of left fibrations. Suppose
given a morphism f : ∆1 → C which becomes an equivalence in D. Consider the diagram
Λ20 C
p

∆2 D
where the map Λ20 → C is given by f on the edge ∆{0,1} and by the identity on ∆{0,2} . Since
the image in D is an equivalence, there exists a dashed arrow making the diagram commute.
Since C → D is a left fibration there also exists the dotted arrow. This proves that f admits a
left inverse in C which becomes a left inverse of p(f ) after applying p and thus an equivalence
after applying p. Running the same argument for this morphism proves that it itself admits
left inverse showing that the first constructed left inverse of f is an equivalence. Thus also f
is an equivalence. 
Definition 5.4. A inner fibration C → D of ∞-categories is called an isofibration if every
lifting problem
{0} C

f
∆1 D
in which f represents an equivalence of D has a solution which represents an equivalence of
C.

Lemma 5.5. An inner fibration C → D between ∞-categories is an isofibration if and only

if the induced functor N(hC) → N(hD) is an isofibration.
Proof. Exercise 80. 
68 M. LAND

Corollary 5.6. A functor C → D between ∞-categories is an isofibration if and only if

Cop → Dop is an isofibration.
Proof. Exercise 81. 
Proposition 5.7. Left and right fibrations between ∞-categories are conservative isofibra-
tions.
Proof. Left and right fibrations are conservative by Proposition 5.3. Now let p : C → D be
a left fibration and consider a lifting problem as in Definition 5.4. Since C → D is a left
fibration and {0} → ∆1 is left anodyne, a lift as needed exists. By conservativity, any such
lift is an equivalence. For right fibrations, use Corollary 5.6. 

The D = ∆0 case of the following theorem is already very interesting.

Theorem 5.8. Let C → D be an inner fibration between ∞-categories and let φ : ∆1 → C
be a morphism in C. Then a lifting problem

∆{0,1} Λn0 C

∆n D
in which the top composite is φ can be solved if φ is an equivalence in C.
Proof. To prove the lifting property provided φ is an equivalence, we consider a diagram

∆{0,1} Λn0 C

∆n D
where the top horizontal composite is an equivalence in C, say φ, and wish to show the
existence of the dashed arrow. We observe that the map Λn0 → ∆n is isomorphic to the
join-pushout product
{0} ? ∆−2+n q{0}?∂∆−2+n ∆1 ? ∂∆n−2 → ∆1 ? ∆−2+n

where ∆−2+n short for ∆{2,...,n} . This was done in an earlier exercise. The map ∆{0,1} → Λn0
identifies with the canonical composite
∆1 → ∆1 ? ∂∆−2+n → {0} ? ∆−2+n q{0}?∂∆−2+n ∆1 ? ∂∆n−2 .
The above diagram is, by adjunction, equivalent to the diagram

{0} C/∆−2+n

φ0
∆1 C/∂∆−2+n ×D/∂∆−2+n D/∆−2+n
φ

C
 INTRODUCTION TO INFINITY-CATEGORIES 69

We claim that all of the three top horizontal maps in the diagram
C/∆−2+n C/∂∆−2+n ×D/∂∆−2+n D/∆−2+n C/∂∆−2+n C

D/∆−2+n D/∂∆−2+n
are right fibrations and thus conservative by Proposition 5.3. The first one is the dual version
of Theorem 4.23 part (1), the last one is explicitly stated in Corollary 4.24 part (1) and the
middle map is a pullback of D/∆−2+n → D/∂∆−2+n which is a right fibration by the same
reasoning as the first map, hence also the pullback is a right fibration.
It thus follows from the assumption that φ is an equivalence that φ0 is also an equiva-
lence. Hence the dashed arrow exists by the fact that right fibrations are isofibrations by
Proposition 5.7. 
Remark. One can ask whether the statement of the theorem can be promoted to an “if
and only if”. In other words, if any such lifting problem can be solved, can one conclude
that φ is an equivalence. The answer to this question is no, as the following example shows.
Suppose that C is an ordinary category and φ is a morphism which admits a left inverse ψ,
but is not an isomorphism. Such categories exist, we leave explicit examples as an exercise
to the reader. Consider the functor N(C) → ∆0 . We claim that any lifting problem

∆{0,1} Λn0 N(C)

∆n ∆0
can be solved: If n ≥ 3 we have seen that any map from a horn extends (uniquely) to ∆n , ??.
If n = 2, this is possible since φ admits a left inverse: The map Λ20 → N(C) in the diagram
is uniquely determined by its restriction to ∆{0,2} , let us denote by g the corresponding
morphism of C. Then the string (φ, g ◦ ψ) determines a 2-simplex in N(C) which extends
the given map from Λ20 . But φ is not an isomorphism. However, we do have the following
statement.

Corollary 5.9. An inner fibration p : C → D between ∞-categories is conservative if and

only if for every n ≥ 2 and every lifting problem
φ

∆{0,1} Λn0 C
p

∆n D
in which p(φ) is an equivalence in D, there exists a solution.
Proof. Suppose that p is conservative. Then the assumption that p(φ) is an equivalence
implies that φ itself is an equivalence, and hence any such lifting problem can be solved by
Theorem 5.8. Conversely, suppose that any such lifting problem has a solution. To show that
p is conservative, let us consider a morphism φ : ∆1 → C such that p(φ) is an equivalence
in D. Consider the map Λ20 → C whose restriction to ∆{0,1} is φ and whose restriction to
70 M. LAND

∆{0,2} is the identity. Since φ becomes an equivalence in D, there exists the solid arrows in
the lifting problem
φ

∆{0,1} Λn0 C
p

∆n D
which can be solved by assumption. This provides a left inverse ψ of φ. It follows as in the
proof of Proposition 5.3 that p(ψ) is an equivalence. Running the same argument for ψ in
place of φ we again find that ψ itself admits a left inverse and hence is an equivalence. Thus
also φ is an equivalence and consequently, p is conservative. 
Remark. The opposite of the inclusion ∆{0,1} → Λn0 is given by the map ∆{n−1,n} → Λnn .
Since inner fibrations and conservative functors are invariant under passing to opposites, we
find that the analogues statements of Theorem 5.8 and Corollary 5.9 where we replace the
inclusion ∆{0,1} → Λn0 by the map ∆{n−1,n} → Λnn hold as well.

Notice that the important direction in Theorem 5.8 for us is that such lifting problems can
be solved provided p is conservative, as the next corollary shows.
Corollary 5.10. ∞-groupoids are Kan complexes.
Proof. By definition, ∞-groupoids are precisely the ∞-categories C where the canonical map
C → ∗ is conservative. Thus the claim follows from Corollary 5.9 
Using this we are now in the position to define the ∞-category of ∞-categories:
Definition 5.11. The ∞-category Cat∞ of ∞-categories is the coherent nerve of the sim-
plicial category with objects ∞-categories and hom simplicial sets given by the maximal
∞-groupoid inside the functor ∞-category Fun(C, D). (This uses that forming the maximal
∞-groupoid is a monoidal functor from ∞-categories to Kan complexes: it is right adjoint to
the inclusion and thus preserves products).

Definition 5.12. A functor f : C → D between ∞-categories is a Joyal (or categorical)

equivalence, if the corresponding 1-simplex in Cat∞ is an equivalence in the sense of Defini-
tion 2.22.

Remark. Concretely, this means that there is a 2-simplex σ : ∆2 → Cat∞ such that
σ|∆{0,1} = f and σ|∆{0,2} = idC . From Observation 2.64 we find that for g = σ|∆{1,2} we
have specified a 1-simplex in Fun(C, D)' from gf to idC . In other words, f is an equivalence
if and only if there is a functor g : D → C and natural equivalences gf ' idC and f g ' idD .

Definition 5.13. Two functors f, f 0 : C → D are called (naturally) equivalent if the corre-
sponding morphisms in Cat∞ are equivalent in the sense of Definition 2.22.

Remark. Unwinding the definitions, we find that f is equivalent to f 0 precisely if there

exists a natural equivalence τ : f → f 0 , i.e. τ is an equivalence between f and f 0 in the
∞-category Fun(C, D).
 INTRODUCTION TO INFINITY-CATEGORIES 71

We continue with more applications of Joyal’s extension theorem.

Corollary 5.14. Equivalences in an ∞-category C are represented precisely by those maps
∆1 → C which extend over the canonical map ∆1 → J.
Proof. The fact that any map ∆1 → C which extends over J is an equivalence is an exercise,
see Exercise 62. Conversely, an equivalence is represented by a map ∆1 → C' ⊆ C. To show
that this map extends over the extension ∆1 → J, it suffices to observe that C' is a Kan
complex by Corollary 5.10 and that the map ∆1 → J is anodyne (its geometric realization
is a homotopy equivalence as both are contractible). See also the proof of Lemma 8.4 for a
purely simplicial proof of the fact that this map is anodyne. 

Lecture 22 – 24.01.2019.
Corollary 5.15. The pullback of a conservative inner fibration C → D along any functor
D0 → D of ∞-categories is again a conservative inner fibration.
Proof. We use the lifting criterion for conservative inner fibrations established in Corollary 5.9
and consider the diagram

∆{0,1} Λn0 C0 C
q p

∆n D0 D
in which the composite ∆1 → D0 represents an equivalence. and want to show that a dashed
arrow exists. Since p is conservative, the dotted arrow exists. Hence the dashed arrow exists,
because the right square is a pullback, thus q is conservative. 
Proposition 5.16. A inner fibration p : C → D between ∞-categories is an isofibration if
and only if the induced functor C' → D' is a Kan fibration.
Proof. First, let us suppose that p is an isofibration. First, we show that the induced functor
p' : C' → D' is also an isofibration. It is again an inner fibration by a similar argument that
shows that C' is itself an ∞-category: Consider a lifting problem
Λnj C' C

∆n D' D
A dotted arrow exists since p is an inner fibration. We claim that the dotted arrow must
already land in C' giving rise to the dashed arrow. This simply follows from the fact that
its restriction to the spine lands in C' which implies that the whole n-simplex lies in C' .
Clearly, the further lifting property of isofibrations is satisfied by p' : The definition says that
any lifting problem
{0} C' C

∆1 D' D
has a solution (we simply spell out that certain 1-simplices are required to be equivalences).
72 M. LAND

Next, we observe that p' is clearly conservative as is any functor from an ∞-groupoid.
We thus now know that p' : C' → D' is a conservative inner fibration, and an isofibration.
To see that it is a Kan fibration, we first show that it is a left fibration. The left horn
∆0 → ∆1 can be extended because p' is an isofibration. To deal with the higher dimensional
left horns, we use the criterion for conservative inner fibrations given by Corollary 5.9: It tells
us that lifts exists provided certain edges of the horn map to equivalences. This condition is
tautologically fulfilled because C' is an ∞-groupoid. Running the same argument (using the
version of Joyal lifting with the right outer horn) we also find that p' is a right fibration, and
hence a Kan fibration.
The converse if obvious: The map {0} → ∆1 is a horn inclusion, and thus admits a lift for
C → D' because we assume it to be a Kan fibration. This shows that p is an isofibration. 
'

Proposition 5.17. An inner fibration C → D is an isofibration if and only if it has the

RLP wrt ∆0 → J. In particular, the isofibrations are precisely the Joyal fibrations between
∞-categories according to Definition 3.22.

Proof. It is clear that having the lifting property wrt ∆0 → J implies that the map is an
isofibration. To show the converse we observe that every diagram

∆0 C

J D

factors through the subcategories of equivalences of C and D, respectively. It thus suffices to

show that a map f : C → D which is an isofibration if and only if the induced map C' → D'
is a Kan fibration which was done in Proposition 5.16. 

Corollary 5.18. Isofibrations are stable under pullback.

6. Pointwise criterion for natural equivalences

In this section we aim to prove the following theorem.

Theorem 6.1. Let K → L be a map between simplicial sets which induces a bijection
K0 → L0 . Then, for every ∞-category C, the induced functor

Fun(L, C) → Fun(K, C)

is conservative.

Corollary 6.2. The canonical functor

Y
Fun(K, C) → C
x∈K0

is conservative. In other words, Let f : ∆1 → Fun(K, C) be a natural transformation between

functors F, G : K → C. If for all x ∈ K, the induced morphism ∆1 → C is an equivalence,
then f is an equivalence.
 INTRODUCTION TO INFINITY-CATEGORIES 73

Lecture 23 – 28.01.2019. In order to prove Theorem 6.1 we will prove the following pre-
liminary lemma.
Lemma 6.3. Let S → T be a monomorphism such that S0 → T0 is a bijection and let C be
an ∞-category. Consider a diagram
Fun(T, C) C
Q
{0}
T0

∼
=
φ
∆1 Fun(S, C) C
Q
S0

in which the lower composite represents an equivalence. Then there exists a dashed arrow
making the square commute.
Proof. We first claim that the set of monomorphisms S → T which induce a bijection on
0-simplices is a saturated class and that this saturated class is generated by the boundary
inclusion ∂∆n → ∆n for n ≥ 1. This follows again from a relative skeletal argument, using
that we only have to attach higher dimensional simplices if the inclusion induces a bijection
on 0-simplices. Then we claim that the set of monomorphisms S → T which satisfy the
conclusion of the lemma is a saturated class. To see the pushout property, consider a pushout
of monomorphisms
S0 S

T0 T
and then a lifting problem
0
{0} CT CT

0
∆1 CS CS
in which the map ∆1 → CS represents a transformation which is objectwise an equivalence.
0 0
Then the same holds for the further composite to CS . So the dashed arrow to CT exists, and
since the right square is a pullback, the dotted arrow to CT exists as well. To show that the
composition S → T → U satisfies the conclusion of the lemma if S → T and T → U does,
consider a lifting problem
{0} CU

CT

∆1 CS
such that the lower map represents an natural transformation which is objectwise an equiva-
lence. By assumption, a dashed lift to CT exists, but to find the further dotted lift, we need to
know that the dashed lift to CT represents again a natural transformation which is objectwise
an equivalence. This follows from the (needed!) assumption that all monomorphisms in the
74 M. LAND

set under consideration induce bijections on 0-simplices. The case of transfinite compositions
follows from the one just considered: Just the case of limit ordinals needs to be done, and in
this case, any map from {0} to a colimit over an ordinal factors through some earlier stage,
which is then again dealt with by the successor step just done. Finally we need to argue that
retracts behave well. So let S → T be a retract of U → V and assume U → V satisfies the
conlusion of the lemma. We consider a diagram

{0} CT CV CT

∆1 CS CU CS
in which the right horizontal composites are the identity and where the map ∆1 → CS is
objectwise an equivalence. This property remains so after passing to CU . Thus a dashed
arrow exists, post composing this with the map CV → CT proves the claim.
It hence suffices to prove the conclusion of the lemma for the boundary inclusions of di-
mension ≥ 1. There is an induced lifting problem

∆1 × {j} {0} × ∆n q{0}×∂∆n ∆1 × ∂∆n C

∆1 × ∆n
in which the upper composite represents an equivalence in C, and where j is any 0-simplex
of ∂∆n . The proof of this fact is very similar to the proof we gave in Proposition 6.6: One
considers the same filtration of ∆1 × ∆n starting with
{0} × ∆n ∪ ∆1 × ∂∆n
by adding the missing simplices in a clever order, and then observing that all these simplices
are either attached along inner horns, or an outer horn, but where the outer edge is labeled
with an equivalence. Then one uses Joyal lifting for this case to deduce the claim. 
Remark. Let us spell out the situation for n = 1 and n = 2.

We recall that the following lemma was used in the proof of Lemma 3.30 and its proof was
deferred to Exercise 63 part (1). Since we will use the explicit retraction, we now include a
proof of this lemma:
Lemma 6.4. The inclusion Λn0 → ∆n is a retract of the map
{0} × ∆n ∪ ∆1 × Λn0 → ∆1 × ∆n .
Proof. Consider the maps
s r
[n] → [1] × [n] → [n]
given by s(x) = (1, x) and r(x, y) = y if (x, y) 6= (0, 1) and r(0, 1) = 0.
As in Lemma 3.28, we have to show that
(1) rs = id,
(2) s(Λn0 ) ⊆ {0} × ∆n ∪ ∆1 × Λn0 , and
(3) r({0} × ∆n ∪ ∆1 × Λn0 ) ⊆ Λn0 .
 INTRODUCTION TO INFINITY-CATEGORIES 75

Part (1) is obvious. (2) is obvious as well: s(Λn0 ) = {1} × Λn0 ⊆ ∆1 × Λn0 . To see (3), we first
observe that the composite
ι0 r
[n] → [1] × [n] → [n]
sends k to 0 if k = 0, 1 and k to k if k ≥ 2. In particular, 1 is not in the image, and thus
r({0} × ∆n ) ⊆ Λn0 . To see that also r(∆1 × Λn0 ) ⊆ Λn0 , consider a composite
(β,α) r
[m] −→ [1] × [n] → [n]
where α represents an m-simple of ∆1 × Λn0 . Obviously, the image of r ◦ (β, α) is contained
in Image(α) ∪ {0}, thus this composite represents again a simplex of Λn0 . 
Proof of Theorem 6.1. Since the functor we consider is an inner fibration (restriction along
a monomorphism) we can apply Corollary 5.9 to prove conservativity. I.e. we consider a
diagram
∆{0,1} Λn0 Fun(K, C)

∆n Fun(L, C)
and assume the composite ∆1 → Fun(L, C) to be an equivalence. By Lemma 6.4, the map
Λn0 → ∆n is a retract of the pushout product map
{0} × ∆n q{0}×Λn0 ∆1 × Λn0 → ∆1 × ∆n
so that we can consider the diagram
Λn0 {0} × ∆n q{0}×Λn0 ∆1 × Λn0 Λn0 Fun(K, C)

s r
∆n ∆1 × ∆n ∆n Fun(L, C)
so that it suffices to find the dashed arrow in this diagram. By adjunction this relates to the
lifting problem
{0} Fun(K × ∆n , C)

∆1 Fun(L × ∆n qL×Λn0 K × Λn0 , C)

Now notice that the right vertical map is restriction along a monomorphism which induces a
bijection on 0-simplices. To apply Lemma 6.3 we need to check that for every 0-simplex of
∆n , the composite
∆1 → ∆1 × ∆n → Fun(L, C)
represents an objectwise equivalence in C. This of course relies on the explicit choice of the
maps s and r which we have specified in Lemma 6.4.
The thing to observe is that for 0 ≤ k ≤ n, the inclusion ∆1 → ∆1 × ∆n given by sending
i to (i, k), followed by the map r : ∆1 × ∆n , is
(1) either the map which is constant at k (if k 6= 1) in which case clearly the map
∆1 → ∆1 × ∆n → ∆n → Fun(K0 , C)
is also constant, and thus objectwise an equivalence;
76 M. LAND

(2) or the map is given by the canonical inclusion ∆{0,1} → ∆n (if k = 1), in which case
the resulting morphism of Fun(L, C) is an objectwise equivalence by assumption. In
either case, a lift exists, so the theorem is proven.


Corollary 6.5. Proposition 3.41 holds true. More precisely, given a monomorphism K →
L of simplicial sets which is a bijection on 0-simplices, then the fibre of the induced map
Fun(L, C) → Fun(K, C) over any point ∆0 → Fun(K, C) is an ∞-groupoid. In particular, for
an ∞-category C, the ∞-category mapC (x, y) is an ∞-groupoid.

Proof. By Theorem 6.1 the functor Fun(L, C) → Fun(K, C) is an conservative inner fibration,
and so the pullback along ∆0 → Fun(K, C) is a conservative inner fibration as well (by
Corollary 5.15). But X → ∗ is a conservative inner fibration if and only if X is an ∞-
groupoid. 

Proposition 6.6. Let p : C → D be an inner fibration and let i : K → L be a monomorphism

of simplicial sets. Suppose that
(1) p is an isofibration, or
(2) i induces a bijection on 0-simplices.
Then the induced functor
CL → CK ×DK DL
is an isofibration.

Proof. By Theorem 3.32, we know that this map is an inner fibration. It thus suffices to show
that any lifting problem
{0} CL

∆1 CK ×DK DL

in which the bottom horizontal map is an equivalence, has a solution which is again an
equivalence. By Theorem 6.1 this is the case if for every object of L, the induced morphism
in C is an equivalence. In particular, we see that if K → L is a bijection on 0-simplices, the
above right vertical map is conservative, so that any lift of an equivalence is automatically an
equivalence. By adjunction, this is lifting problem is thus equivalent to the lifting problem

{0} × L q{0}×K ∆1 × K C

∆1 × L D

in which the composite of the dashed map with the map ∆1 × ∆0 → ∆1 × L given by an
arbitrary object of L is an equivalence. We claim that the set of morphisms K → L for
which the conclusion holds is a saturated class, a very similar argument was worked out in
Lemma 6.3, we leave this version as Exercise 86. It hence suffices to show it for the boundary
inclusions ∂∆n → ∆n in case (1) and the boundary inclusions with n ≥ 1 in case (2). Let us
 INTRODUCTION TO INFINITY-CATEGORIES 77

do case (1) first: If n = 0, then we have the lifting problem

{0} C

∆1 D
where the lower horizontal map is an equivalence. A dashed arrow representing an equivalence
in C exists because C → D is an isofibration by assumption. For the remaining cases n ≥ 1
we we need to consider diagrams of the form

{0} × ∆n q{0}×∂∆n ∆1 × ∂∆n C

∆1 × ∆n D
and we will not use that p is an isofibration so that the following argument settles both
the remaining cases of (1) and (2). One constructs a filtration on ∆1 × ∆n , starting with
{0} × ∆n ∪ ∆1 × ∂∆n by adding the missing simplices. As in the proof of Lemma 6.3, one
finds that the relevant simplices are either attached along inner horn inclusions, or along
outer horns where one outer edge is labelled with an equivalence. Using Joyal lifting, the
proposition follows. 
Corollary 6.7. Let K → L be a monomorphism of simplicial sets which induces a bijection
on 0-simplices. Then the map
{0} × L q{0}×K J × K → J × L
has the LLP with respect to inner fibrations between ∞-categories.
Proof. Let us consider an inner fibration p : C → D and a lifting problem

{0} × L q{0}×K J × K C

J ×L D
This lifting problem is, by adjunction equivalent to the lifting problem

{0} CL

J CK ×DK DL

and by Proposition 6.6 the right vertical map is an isofibration. The lifting problem can thus
be solved by Proposition 5.17. 
Corollary 6.8. Let f : C → D be a functor between ∞-categories and K → L be a monomor-
phism of simplicial sets. Then
(CK ×DK DL )' = (CK )' ×(DK )' (DL )' .
78 M. LAND

Proof. Both simplicial sets are subsets of the pullback CK ×DK DL , thus we easily find the
inlcusion “⊆”. To show the converse, it suffices to prove that the right hand side is in fact an
∞-groupoid, as then it certainly contains the smallest ∞-groupoid contained in CK ×DK DL ,
which is the left hand side. By Proposition 6.6 applied to the isofibration D → ∆0 , we find
that DL → DK is also an isofibration. By Proposition 5.16, the map (DL )' → (DK )' is
hence a Kan fibration between Kan complexes, so that any pullback along a map from a Kan
complex is again a Kan complex. 
Proposition 6.9. Let f : C → D be a functor between ∞-categories. Then f is a Joyal
equivalence if and only if for every ∞-category E, the induced map
f ∗ : Fun(D, E) → Fun(C, E)
is a Joyal equivalence.
Proof. To prove the “only if”, choose an inverse D → C and choose natural transformations
(η0 , η1 ) whitnessing that g is inverse to f . Then the quadruple (f ∗ , g ∗ , η0∗ , η1∗ ) determines an
equivalence between Fun(C, E) and Fun(D, E).
We now show that f is in fact a Joyal equivalence if for all ∞-categories E, the functor f ∗
is a Joyal equivalence. We obtain, by Exercise 88, a bijection
∼
=
f ∗ : π0 (Fun(D, E)' ) −→ π0 (Fun(C, E)' )
Now consider the case where E = C. Then this bijection shows the existence of a functor
g : D → C such that f ∗ (g) = gf is equivalent to idC . Now taking E = D, we see that
f ∗ (f g) = f gf ' f
and thus that f g ' idD as needed. 
Definition 6.10. A map f : X → Y between simplicial sets is called a Joyal (or categorical)
equivalence if for all ∞-categories C, the induced map
Fun(Y, C) → Fun(X, C)
is a Joyal equivalence between ∞-categories.

Observation 6.11. (1) This does not change the definition if X and Y are already
∞-categories by Proposition 6.9.
(2) A Joyal equivalence between Kan complexes is precisely a homotopy equivalence.

Proposition 6.12. A trivial Kan fibration f : X → Y is a Joyal equivalence.

Proof. Consider the following pullback squares of simplicial sets
Hom/Y (Y, X) Hom(Y, X) Hom/Y (X, X) Hom(X, X)
f∗ f∗
idY f
∆0 Hom(Y, Y ) ∆0 Hom(X, Y )
The right vertical maps in each square are trivial fibrations, thus so are their pullbacks.
But a trivial fibration over ∆0 has a source a contractible Kan complex. Choose a 0-simplex
s ∈ Hom/Y (Y, X) so that f∗ (s) = f s = idY : Recall that f∗ , as every trivial fibration, admits a
section, and that any 0-simplex s ∈ Hom(Y, X) with f s = idY lies in Hom/Y (Y, X). Observe
 INTRODUCTION TO INFINITY-CATEGORIES 79

that thus sf determines a 0-simplex in Hom/Y (X, X) and thus that there must be a 1-
simplex connecting it to the identity (again because these spaces of sections are contractible).
Explicitly, we can find a map
∆1 → Hom/Y (X, X)
whose restriction to 0 is sf and whose restriction to 1 is idX . For an arbitrary ∞-category
we can compose this with the canonical map
Hom/Y (X, X) → Hom(X, X) → Hom(CX , CX )
and see that the resulting map
∆1 → Hom/Y (X, X) → Hom(CX , CX )'
determines a natural equivalence between (sf )∗ and id. Since (f s)∗ = id we have shown that
f ∗ and s∗ determine inverse equivalences of CX and CY for any ∞-category Y , so that f ∗ is
a Joyal equivalence and thus f itself is a Joyal equivalence. 
Corollary 6.13. An inner anodyne map is a Joyal equivalence.
Proof. Let i : A → B be an inner anodyne map and C be an ∞-category. We need to show
that CB → CA is a Joyal equivalence. By Theorem 3.32 part (2) this map is a trivial fibration,
and thus by Proposition 6.12 the claim is shown. 
Corollary 6.14. Every simplicial set is Joyal equivalent to an ∞-category.
Proof. By the small object argument, Proposition 3.9, for any simplicial set X one can factor
the map X → ∗ into an inner anodyne map followed by an inner fibration. This produces
a map X → C where C is an ∞-category and X → C is inner anodyne and thus a Joyal
equivalence by Corollary 6.13. 
Lemma 6.15. A Kan fibration p : X → Y between Kan complexes which induces a surjection
on π0 is in fact surjective on 0-simplices. Likewise, a Kan fibration between Kan complexes
which induces an injection on π0 and a surjection on π1 has the property that any lifting
problem
∂∆1 X

∆1 Y
can be solved.
Proof. A homotopy equivalence induces a bijection on simplicial path components. Thus for
a 0-simplex y : ∆0 → Y , we find a commutative diagram
x0
{0} X

h
∆1 Y
where h|{1} = y. Since p is a Kan fibration, a dashed arrow exists. Its restriction to {1}
provides a preimage of y in X.
80 M. LAND

To see the second claim, pick two objects x, x0 of X which give rise to the top horizontal
map in the commutative diagram
∂∆1 X

h
∆1 Y
which we wish to show admits a dashed arrow. The assumption that p induces a bijection on
π1 implies that the same is true not for loops at a point of x but at homotopy classes of paths
from x to x0 (This is where we use that p induces an injection on path components: The
assumptions imply that there exists a path from x to x0 in X which we can use to compare
the set of homotopy classes of paths from x to x0 to the set of homotopy classes of loops at
x). One can thus find a path α : ∆1 → X which becomes equivalent to h after applying p.
This means that we can find a 2-cell σ : ∆2 → Y such that σ|∆{0,1} = p(α), σ|∆{0,2} = h and
σ|∆{1,2} = idx0 . Since we can lift both p(α) and the identity, we obtain a lifting problem

Λ21 X

σ
∆2 Y
which can be solved as p is a Kan fibration. The resulting map solves the original lifting
problem. 
Remark. In fact, a Kan fibration which is also a weak equivalence is a trivial fibration.
This is a classical fact in simplicial homotopy theory. We will deduce it from the previous
lemma together with the following observation:

Lecture 24 – 31.01.2019.
Lemma 6.16. A functor p : C → D between ∞-categories is a trivial fibration if and only if
it is a Joyal equivalence and an isofibration.
Proof. Trivial fibrations are Joyal equivalences by Proposition 6.12 and isofibrations because a
trivial fibration is a left fibration which in turn is a conservative isofibration by Proposition 5.7.
To see the converse, we want to show that for any monomorphism K → L and any lifting
problem
K C
p

L D
has a solution. This is equivalent to the lifiting problem

∅ CL
p

∆0 CK ×DK DL
which amounts to showing the surjectivity of the right hand map on 0-simplices. To prove
this, we may pass to the underlying groupoid cores, as this does not change the 0-simplices.
 INTRODUCTION TO INFINITY-CATEGORIES 81

Since p is an isofibration, so is p, by Proposition 6.6. In particular, the induced map

p
(CL )' → (CK ×DK CL )'
is a Kan fibration. Furthermore, by Corollary 6.8 the latter is further equal to (along the
canonical map) the pullback of the groupoid cores, so that we find that the map
(CL )' → (CK )' ×(DK )' (DL )'
is a Kan fibration. We wish to show that it is surjective on 0-simplices. By Lemma 6.15
it suffices to show that this map induces a surjection on π0 . For this we observe that since
C → D is a Joyal equivalence, so are the maps
CK → DK and CL → DL
by an exercise, Exercise 91. Hence, as they are isofibrations, passing to groupoid cores gives
us Kan fibrations, which are in addition Joyal equivalences.
We then consider the diagram

(CL )' (CK )' ×(DK )' (DL )' (DL )'

(CK )' (DK )'

and note that all maps are Kan fibrations and that both the composite and the lower horizontal
map are in Joyal equivalences, and thus homotopy equivalences. We wish to show that the
second horizontal top map induces a bijection on π0 , so that the first map induces a surjection
on π0 . For this we observe that the lower horizontal map has the RLP wrt ∂∆1 → ∆1 by
Lemma 6.15. As a pullback, so does the second top horizontal map. This implies that this
map is injective on π0 .
Using Lemma 6.15, we find that the map
(CL )' → (CK )' ×(DK )' (DL )'
induces a surjection on 0-simplices. This proves the lemma. 
Corollary 6.17. A Kan fibration p : X → Y between Kan complexes is a trivial Kan
fibration if and only if it is a homotopy equivalence. In particular, among Kan fibrations, the
trivial Kan fibrations satisfy the 3-for-2 property.
Proof. A Kan fibration which is a homotopy equivalence is an isofibration which is a Joyal
equivalence. The in particular follows from the 3-for-2 property for homotopy equivalences.

Remark. We hence find that given an isofibration which is in addition a Joyal equivalence,
p : C → D, the induced map
p
(CL )' → (CK ×DK CL )'
is in fact a trivial fibration: We claim that in the following composite
(CL )' → (CK )' ×(DK )' (DL )' → (DL )'
both the composite, and the latter map are trivial fibrations: The composite is a Kan fibration
which is a homotopy equivalence and thus a trivial fibration. The latter map is a pullback of
82 M. LAND

the map (CK )' → (DK ) ' which is a trivial Kan fibration by the same reasoning. By 3-for-2
for trivial fibrations (among Kan fibrations) it follows that the map

(CL )' → (CK )' ×(DK )' (DL )'

is also a trivial fibration.

7. Fully faithful and essentially surjective functors

The goal of this section is to prove that as in ordinary category theory, functor which are
essentially surjective and fully faithful are in fact invertible. The proof we present follows an
argument which we learned from Gijs Heuts. It uses some results from classical simplicial
homotopy theory.
We wish to emphasize however, that the proof of this fact in ordinary category theory, which
we presented in Exercise 37, goes through verbatim in ∞-categories once enough technology
is established: The main thing to prove is that an essentially surjective and fully faithful
functor admits a right adjoint. This will then be an inverse. Of course, for this argument to
make sense we will have to speak of adjunctions and the way we want to do it will require
the straightening–unstraightening equivalence. We will come back to this approach in a later
chapter.

Definition 7.1. A functor f : C → D is called fully faithful if for all objects x, y ∈ C, the
induced map mapC (x, y) → mapD (f x, f y) is a Joyal equivalence, i.e. a homotopy equivalence.

Definition 7.2. A functor f : C → D is called essentially surjective if the induced functor

hf : hC → hD is. In other words, if for every object d ∈ D, there exists an object x ∈ C and
an equivalence f x ' d in D.

Lemma 7.3. Let f, g : C → D be functors and let τ : ∆1 → Fun(C, D) be a natural transfor-

mation from f to g. Then the diagram

mapC (x, y) mapD (f x, f y)

mapD (gx, gy) mapD (f x, gy)

commutes up to homotopy. If τ is a natural equivalence, then the lower horizontal and right
vertical maps are equivalences. In particular, if D = C and g = id, we find that if f is
equivalent to idC , then the map

mapC (x, y) → map(f x, f y)

is a homotopy equivalence.
1 1
Proof. Recall that there is a functor Fun(C, D) → Fun(C∆ , D∆ ) induced by postcompos-
1 1
tion. The given transformation τ thus induces a functor ∆1 → Fun(C∆ , D∆ ) which is (by
adjunction) a functor
1 1 ×∆1
C∆ → D∆ .
 INTRODUCTION TO INFINITY-CATEGORIES 83

Unravelling this construction we find that this functor sends a morphism x → y to the diagram
τx
fx gx

τy
fy gy
We hence find a diagram
1 1 ×∆1 1 1
C∆ D∆ D∆ ×D D∆

1 1 1
D∆ ×D D∆ D∆
1 1
here the horizontal functor from D∆ ×∆ is given by restriction along the map Λ21 → ∆1 × ∆1
which singles out one corner of the square and the vertical functor is restriction along the map
Λ21 → ∆1 × ∆1 which singles out the other corner. The diagonal map is given by restriction
along the inclusion ∆1 → ∆1 × ∆1 sending 0 to (0, 0) and 1 to (1, 1). The remaining two
functors are given composition. Both of the triangles commute up to a natural equivalence
2 2
(one has to choose a section of the trivial fibration D∆ → DΛ1 ).
1
Now we fix two objects x and y of C and consider the inclusion mapC (x, y) → C∆ . By
inspection we see that the diagram
mapC (x, y) {τx } × mapD (gx, gy)

1 1 1
C∆ D∆ ×D D∆
commutes, i.e. that the lower composite factors as the dashed arrow indicates.
The same holds for the restriction of the vertical map to mapC (x, y) so that in total we
obtain a diagram
mapC (x, y) {τx } × mapD (gx, gy)

mapD (f x, f y) × {τy } mapD (f x, gy)

which commutes up to a natural equivalence as needed. 

Proposition 7.4. A Joyal equivalence between ∞-categories is fully faithful and essentially
surjective.
Proof. By Exercise 90 we see that the functor hC → hD is an equivalence and thus essentially
surjective, so that C → D is essentially surjective according to Definition 7.2. To show fully
faithfulness, choose an inverse g of f . Then, for pair of objects x, y ∈ C we get
mapC (x, y) → mapD (f x, f y) → mapC (gf x, gf y) → mapD (f gf x, f gf y)
Since the functor gf is naturally equivalent to idC and the functor f g is naturally equivalent
to idD we see that the first two maps compose to an equivalence and the latter two compose
to an equivalence. This implies that the middle map is itself an equivalence (it has both a left
84 M. LAND

and a right inverse) and thus that the first map is also an equivalence (it is a right-inverse of
an equivalence). 
Remark. Alternatively, one can argue as follows: We have seen that every ∞-category gives
rise to a category enriched in the homotopy category of Kan complexes because composition
is well-defined up to homotopy see ??. Having this, we find that a functor f : C → D is
a Joyal equivalence if and only if the induced functor of h(Kan) enriched categories is an
equivalence. Likewise, it is fully faithful and essentially surjective if and only if the induced
functor of h(Kan) enriched categories is fully faithful. From this analysis we deduce that if f
is naturally equivalent to g, then f is fully faithful if and only if g is. In the above argument,
we can apply this to gf which is equivalent to id and thus must be fully faithful itself, so that
the required map is in fact a homotopy equivalence.

Lemma 7.5. The inclusion of a full subcategory C0 ⊆ C is a fully faithful functor.

Proof. We first observe that for every ∞-category D, the ∞-category Fun(D, C0 ) is the full
subcategory of Fun(D, C) on those functors which factor through C0 ⊆ C, see Exercise 48.
Having this we deduce that the diagram

Fun(∆1 , C0 ) Fun(∆1 , C)

C 0 × C0 C×C
is a pullback. This implies the lemma by passing to fibres over objects (x, y) of C0 × C0 . 
Lemma 7.6. Let C be an ∞-category and let x, y be objects of C. Then the map
mapC' (x, y) → mapC (x, y)
is the inclusion of those path components whose points are equivalences of C.
Proof. We will show that any map ∆n → mapC (x, y) lifts to mapC' (x, y) if for every i ∈ ∆n ,
the corresponding morphism from x to y is an equivalence. For this we consider a map
∆n → mapC (x, y) such that for all i ∈ ∆n , the restriction of its adjoint map
{i} × ∆1 → ∆n × ∆1 → C
represents an equivalence in C. Then we observe that, for all  = 0, 1 we also have that
∆n × {} → ∆n × ∆1 → C
is constant (at either x or y) and hence also represents an equivalence of C. Since all morphisms
in ∆n × ∆1 are composites of ones of the previous form, we deduce that the map
∆n × ∆1 → C
factors through C' . This shows the claim. 
We now aim to prove the converse of this proposition. In order to do so we shall need the
following preparatory statements.
Lemma 7.7. If f : C → D is fully faithful and essentially surjective, then so is f ' : C' →
D' .
 INTRODUCTION TO INFINITY-CATEGORIES 85

Proof. Essentially surjectivity is clear as C' and C have the same objects. Fully faithfulness
is also fine: As mapC' (x, y) ⊆ mapC (x, y) is a collection of path components, it suffices to
know the conclusion of the lemma in the case where C and D are ordinary categories (by
means of the homotopy category). This is an explicit and easy check. 

We will make use of the following fundamental property of Kan fibrations, see for instance
[GJ09, Lemma 7.3].

Lemma 7.8. Let f : X → Y be a Kan fibration between Kan complexes. Let x be a point in
X and let F be the fibre of f over the point y = f (x). Then there exists a long exact sequence

∂ f∗ ∂
... πn+1 (Y, y) πn (F, x) πn (X, x) πn (Y, y) ... π0 (X) π0 (Y )

natural in morphisms of fibrations, which is an exact sequence of groups for n ≥ 1 and is an

exact sequence of pointed sets for n = 0.

We will make use of Whitehead’s theorem:

Proposition 7.9. Let f : X → Y be a map between Kan complexes which induces a bijection
on path components. Then f is a homotopy equivalence if and only if for all points x in X
and all n ≥ 1, the induced map

f∗ : πn (X, x) → πn (Y, y)

is a bijection.

Corollary 7.10. A fully faithful and essentially surjective functor f : X → Y between Kan
complexes is a homotopy equivalence.

Proof. The functor induces an equivalence of homotopy categories so we find that the map f
induces a bijection π0 (X) → π0 (Y ). We wish to show that for all x in X and all n ≥ 1, also
the induced map
πn (X, x) → πn (Y, y)
is a bijection, where y = f (x). For this we consider the following diagram

mapX (x, x) Px (X) Fun(∆1 , X)

(id,x) (x,idX )
∆0 ∆0 × X X ×X

Since right most vertical map is an isofibration between Kan complexes, it is a Kan fibration.
Hence so is the middle vertical map. This construction is clearly natural in X. We thus
obtain a diagram of Kan fibre sequences

mapX (x, x) Px (X) X

mapY (y, y) Py (Y ) Y
86 M. LAND

where the vertical maps are induced by the map f : X → Y . We will now show that Px (X)
is contractible: This is because it sits also in the pullback diagram

Px (X) Fun(∆1 , X)
s
x
∆0 X
where t stands for the target map, i.e. the map obtained from restriction along {0} → ∆1 .
Since this map is anodyne and X is a Kan complex, the resulting map Fun(∆1 , X) → X
is a trivial fibration, and hence so is the map Px (X) → ∆0 . This implies that Px (X) is
contractible.
We hence obtain that for all n ≥ 1 there is a commutative diagram
πn (X, x) πn−1 (mapX (x, x), idx )

πn (Y, y) πn−1 (mapY (y, y), idy )

in which the horizontal maps are isomorphisms by the long exact sequence of Lemma 7.8 and
the contractability of Px (X) and Py (Y ). Since f is fully faithful, the map mapX (x, x) →
mapY (y, y) is a homotopy equivalence, and hence induces bijections on all homotopy groups.
It follows that the left vertical map is also an isomorphism. 
Corollary 7.11. A fully faithful and essentially surjective functor between ∞-categories
induces a homotopy equivalence on groupoid cores.
Proof. By Lemma 7.7 the induced functor on groupoid cores is still essentially surjective and
fully faithful so that Corollary 7.10 applies. 
For Lemma 7.13 we will need Reedy’s lemma, which is the following statement. It holds in
fact in any model category. For a proof for general model categories we refer to [Hir03, Prop.
13.1.2].
Lemma 7.12. Consider a pullback diagram
C ×A B B
p

'
C f
A

in which p is a fibration and where f is a weak equivalence between fibrant objects. Then also
the map C ×A B → B is a weak equivalence.
Remark. For simplicial sets, let us assume that we have long exact sequences in homotopy
groups for Kan fibrations at our disposal: In this case we easily find from a diagram chase
using the long exact sequence for the vertical fibrations that the induced map
πi (C ×A B, x) → πi (B, x0 )
is a bijection for every basepoint x of C ×A B and for every i ≥ 1. To prove Reedy’s lemma
in the case of interest it hence remains to see that the map C ×A B → B induces a bijection
 INTRODUCTION TO INFINITY-CATEGORIES 87

on path components. The arguments here are very similar to the ones used in the proof of
Lemma 6.15.
Let’s prove surjectivity of this map first: pick a point b in B representing a class [b] ∈ π0 (B)
and consider the point p(b) in A. Since C → A induces a bijection on path components, we
can find a point c in C and a path ∆1 → A connecting f (c) to p(b). This gives, as in
Lemma 6.15, a lifting problem
{0} B

∆1 A
which can be solved as p is a Kan fibration. We thus find that there exists a b0 in B such that
p(b0 ) = f (c) and such that [b0 ] = [b] in π0 (B). The pair (c, b0 ) thus determines an element of
π0 (C ×A B) which is sent to [b] in π0 (B). This shows that the map π0 (C ×A B) → π0 (B) is
surjective.
To show injectivity, consider two points (c, b) and (c0 , b0 ) of C ×A B and assume that there
is a path α : ∆1 → B connecting b and b0 in B. Then p(α) : ∆1 → A connects p(b) = f (c) to
p(b0 ) = f (c0 ). Since the map f is a homotopy equivalence, there is a path β : ∆1 → C such
that f∗ (β) is equivalent to p∗ (α). Precisely, we find a 2-cell σ : ∆2 → A such that
(1) σ|∆{0,1} = f∗ (β),
(2) σ|∆{1,2} = idf (c0 ) , and
(3) σ|∆{0,2} = p∗ (α).
Since we can lift both p∗ (α) and idf (c0 ) along p, similarly as in Lemma 6.15, we find a diagram

Λ22 B

σ
∆2 A
which admits a dashed arrow as indicated since p is a Kan fibration. It follows that there
exists a path γ : ∆1 → B connecting b and b0 such that p∗ (γ) = f∗ (β). Hence γ and β combine
to a map ∆1 → C ×A B connecting (c, b) and (c0 , b0 ). This shows that the map in question is
also injective.

Lecture 25 – 04.02.2019.
Lemma 7.13. Consider a diagram of Kan complexes

C A B
' ' '

C0 A0 B0
in which the vertical maps are weak equivalences and the left horizontal maps are Kan fibra-
tions. Then the induced map on pullbacks
C ×A B → C 0 ×A0 B 0
is again a weak equivalence.
88 M. LAND

Proof. We first reduce the case in question to the situation where also the maps B → A and
B 0 → A0 are fibrations: By the small object argument we find a commutative diagram
A D ' B

A0 D0 ' B0
by functorially factoring the map B → A as a weak equivalence followed by a fibration (in
our case an anodyne map followed by a Kan fibration). We observe that it follows that both
D and D0 are fibrant. We obtain a commutative diagram
C ×A B C 0 ×A0 B 0
' '

C ×A D C 0 ×A0 D0
and wish to show that the top horizontal map is an equivalence. We claim that both vertical
maps are equivalences: For instance the left vertical map sits inside a pullback diagram
C ×A B C ×A D C

'
B D A
so that Reedy’s lemma will imply that the top horizontal map is an equivalence because the
map C ×A D → D is a fibration as it is pulled back from C → A which is a fibration by
assumption. The argument for the right vertical map above is analogous.
We may thus assume that in the statement of the lemma, all horizontal maps are in fact
[Link] observe that the map in question factors as the composite
C ×A B → (C 0 ×A0 A) ×A B = C 0 ×A0 B → C 0 ×A0 B 0 .
Now we use Reedy’s lemma three times:
(1) The map C → C 0 ×A0 A is an equivalence: By Reedy’s lemma, the map C 0 ×A0 A → C 0
is a weak equivalence as it sits in the following pullback
C 0 ×A0 A C0

'
A A0
Hence in the composite
C → C 0 ×A0 A → C 0
both the second map and the composite are equivalences. By 3-for-2 for equivalences
the claim follows.
(2) The map C ×A B → (C 0 ×A0 A) ×A B is an equivalence: It sits in the pullback square
C ×A B (C 0 ×A0 A) ×A B B

C C 0 ×A0 A A
 INTRODUCTION TO INFINITY-CATEGORIES 89

in which the middle vertical map is a fibration, as it is pulled back from the map
B → A which is now a fibration by assumption. The lower horizontal map is an
equivalence by the previous step, so we conclude again using Reedy’s lemma.
(3) The map C 0 ×A0 B → C 0 ×A0 B 0 is an equivalence: It sits in a pullback square

C 0 ×A0 B C 0 ×A0 B 0 C0

B B0 A0
in which the right vertical map is a fibration, as it is pulled back from C 0 → A0 which
is also a fibration by assumption. Now the map B → B 0 is an equivalence, so we
conclude again using Reedy’s lemma.


We will need a similar invariance statement for inverse limits, a more general version of
this result is [Hir03, Theorem 19.9.1]:
Lemma 7.14. Consider a natural transformation between functors Nop → sSet
... X3 X2 X1 X0
' ' ' ' '

... Y3 Y2 Y1 Y0
and assume that all horizontal maps are fibrations, that all vertical maps are weak equivalences
and that all objects are Kan. Then the induced map
lim Xi → lim Yi
i i

is an equivalence.
Remark. Again, one can give a proof of this using long exact sequences in homotopy groups:
It turns out that there is an exact sequence

0 lim1i πk+1 (Xi ) πk (lim Xi ) lim πk (Xi ) 0

i i

so applying (carefully!) a diagram chase argument one can show that in our situation, the
induced map on inverse limits induces a bijection on all homotopy groups. For this argument
see for instance [Hir15].

Lemma 7.15. Consider a commutative diagram of Kan complexes

f
Y0 Y
p0 p
f0
X0 X
in which the map p is a fibration. Suppose that f 0 is a homotopy equivalence, and that for each
0-simplex x0 of X 0 the induced map p0−1 (x0 ) → p−1 (x) between vertical fibres is an equivalence
as well. Then the map f is a homotopy equivalence.
90 M. LAND

Proof. By Reedy’s lemma we may assume without loss of generality that f 0 is the identity:
We can simply replace the map p be the canonical map X 0 ×X Y → X 0 and leave the fibres
unchanged while knowing that the map X 0 ×X Y → Y is an equivalence.
If we have long exact sequences in homotopy groups available, it is an easy diagram chase
to see that for every point y 0 in Y 0 and every i ≥ 1, the map
πi (Y 0 , y 0 ) → πi (Y, y)
is a bijection, where y = f (y 0 ). It hence remains to show that the map f induces a bijection
on path components.
To show injectivity assume given two points x, y in Y 0 whose images under f in Y are
connected by a path α : ∆1 → Y . In other words, we have α(0) = f (x) and α(1) = f (y). It
follows that pα : ∆1 → X is a path between p(f (x)) = p0 (x) and p(f (y)) = p0 (y). Consider
the lifting problem
x
{0} Y0
β
p
pα
∆1 X
which Kan be solved since is a fibration. We hence have p0 β = pα, and since p0 = pf we
p0
obtain pf β = pα. Furthermore we have f β(0) = f (x) = α(0). We hence obtain a lifting
problem
(f β,α)
Λ20 Y
τ p

σ
∆2 X
where σ is a degeneration of the path pα. This lifting problem can be solved since p is a Kan
fibration. Restricting the dashed arrow τ to ∆{1,2} we obtain a path from f (y) to f (β(1))
which is sent by p to the constant path at p0 (y). In other words, τ|∆{1,2} is a path in p−1 (p0 (y)).
Since f restricts to a homotopy equivalence on this fibre, we obtain that there is also a path
between y and β(1) inside p0−1 (p0 (y)). Since β(1) is connected (via β) with x, we deduce that
x and y are connected by a path in Y 0 .
To show surjectivity, consider a point y of Y and let x = p(y). By definition, y lies in the
fibre Fx of p over x. This shows that Fx is not empty. By assumption, the map f 0 restricts
to a homotopy equivalence Fx ' Fx0 where the latter denotes the fibre of p0 over x. Pick a
point y in Fx0 which corresponds to [y] under the bijection π0 (Fx0 ) → π0 (Fx ). Then the map
π0 (Y 0 ) → π0 (Y ) sends [y 0 ] to [y] so the map in question is also surjective. 

Lecture 26 – 07.02.2019. We are now in the position to prove the characterization of Joyal
equivalences as the essentially surjective and fully faithful functors.
Theorem 7.16. A fully faithful and essentially surjective functor f : C → D between ∞-
categories is a Joyal equivalence.
Proof. We will prove the theorem by showing that for any simplicial set X, the canonical
functor
f∗ : (CX )' → (DX )'
 INTRODUCTION TO INFINITY-CATEGORIES 91

is a homotopy equivalence. Once this is shown, one can consider X = D and, by inverting
the homotopy equivalence, we obtain a diagram
g
∆0 (CD )'
f∗

∆1 h
(DD )'

where h is a path from idD to f g, i.e. h provides a natural equivalence between f g and idD .
To see that also gf is naturally equivalent to idC we consider the homotopy equivalence
f∗ : (CC )' → (DC )'
and observe that both idC is sent to f and that gf is sent to f gf . But since f g is connected to
idD , we find that f gf is also connected to f through a natural equivalence. Since the above
map is a homotopy equivalence, this implies that there also must be a path between idC and
f g, so that any such path also provides a natural equivalence f g ' idC and thus that f and
g are mutually inverse functors. Hence, f is a Joyal equivalence.
We will now prove the remaining claim. We first consider the case where X = ∆0 . This is
equivalent to the statement that f induces a homotopy equivalence of groupoid cores which we
settled in Corollary 7.10. Next, we treat the case X = ∆1 . We recall that the source–target
1
map C∆ → C × C is an isofibration and hence the resulting map
1
(C∆ )' → C' × C'
is a Kan fibration which fits into the commutative square
1
(C∆ )' C' × C '
'
1
(D∆ )' D' × D'

in which the right vertical map is a homotopy equivalence by the previous step (and the
observation that products of Joyal equivalences are again Joyal equivalences). According
to Lemma 7.15, to show that the left vertical map is a homotopy equivalence it will hence
suffice to show that the induced map on fibres over a point (x, y) ∈ C' × C' is a homotopy
equivalence as well. We will argue momentarily that the fibre of the respective horizontal
maps over the point (x, y), respectively over the point (px, py), is given by the corresponding
mapping space, so that the induced map on fibres is given by
mapC (x, y) → mapD (px, py).
This map is a homotopy equivalence for all pairs (x, y) by the assumption that p is fully
faithful. We thus conclude the case X = ∆1 , once the claim about the fibres is justified. For
this we observe that the diagram
1 1
(C∆ )' C∆

C' × C' C×C

92 M. LAND

is a pullback, as the right vertical functor is conservative by Theorem 6.1 so that we can
allude to Exercise 79.
Next, we deal with the case X = I n , the n-dimensional spine. We will prove that the map
n n
(CI )' → (DI )'

is a homotopy equivalence by induction on n. The case n = 1 was done in the previous step.
Then we claim that there is a pullback diagram as follows.

n n−1
(CI )' (CI )'

1
(C∆ )' C'

As I n = I n−1 q ∆1 , we find that the diagram is a pullback before applying groupoid cores,
and the two maps with target C are isofibrations (as C is an ∞-category and the map C → ∆0
is an isofibration). As in the proof of Corollary 6.8, it hence suffices to observe that the
pullback of groupoid cores is itself an ∞-groupoid. This is the case because the right vertical
map is a Kan fibration (as it is an isofibration before applying the groupoid core). The map
C → D induces a map from this square to the corresponding square where C is replaced by
D throughout. On all spots except the top left spot, this map is a homotopy equivalence by
the inductive assumption. We thus conclude by Lemma 7.13.
Next, we deal with the case X = ∆n . For this we consider the diagram

n n
(C∆ )' (D∆ )'

n n
(CI )' (DI )'

induced by the functor C → D and the inclusion I n → ∆n . By the previous step, the lower
horizontal map is a homotopy equivalence, and by Theorem 3.32 the vertical maps are trivial
fibrations before applying the groupoid core, and thus remain so after applying the groupoid
core (the square obtained by restricting to groupoid cores is a pullback since trivial fibrations
are conservative). Since trivial fibrations are homotopy equivalences we conclude by 3-for-2
for homotopy equivalences.
Next we deal with an arbitrary but finite dimensional simplicial set X. We prove the
statement by induction over the dimension. For 0-dimensional X this follows again since
products of Joyal equivalences are Joyal equivalences. Let us prove the inductive step and
assume that X is an n-dimensional simplicial set. Consider its skelatal pushout

∂∆n
`
skn−1 (X)

∆n
`
X
 INTRODUCTION TO INFINITY-CATEGORIES 93

which induces a pullback square

(CX )' (Cskn−1 (X) )'

n n
(C∆ )' (C∂∆ )'
Q Q

in which the lower horizontal map is a product of Kan fibrations, and hence itself a Kan
fibration. We conclude this case by Lemma 7.13.
To prove the general case, we now write an arbitrary simplicial set X as the N-indexed
colimit over its skeleta. We then obtain an isomorphism
(CX )' ∼
= lim(Cskn (X) )'
n

and all transition maps in this diagram are Kan fibrations (as they are restrictions along
monomorphisms). We now conclude using Lemma 7.14. 

8. Localizations
Lecture 1 – 24.04.2019. Next we want to study a further construction of ∞-categories
which will play a role later as well: We wish to “universally invert” a chosen set of morphisms
in a given ∞-category. Such a construction will be called a Dwyer–Kan localization.
Definition 8.1. Let C be an ∞-category and let S ⊆ C1 be a subset of the morphisms of
C. For an auxiliary ∞-category D, we let FunS (C, D) ⊆ Fun(C, D) be the full subcategory
consisting of those functors f : C → D such that f (S) ⊆ D' , i.e. where f maps the morphisms
of S to equivalences in D. If S consists of all morphisms, we will write Fun' (C, D) for
FunC1 (C, D).

Definition 8.2. Let C be an ∞-category and let S ⊆ C1 be a subset of the morphisms of C.

A functor C → C[S −1 ] is called a Dwyer–Kan localization of C along S, if for every auxiliary
∞-category D, the restriction functor
Fun(C[S −1 ], D) −→ Fun(C, D)
is fully faithful and its essential image consists of those functors that send S to equivalences.

Remark. By Theorem 7.16 this is equivalent to saying that the restriction functor factors
through a Joyal equivalence Fun(C[S −1 ], D) → FunS (C, D).

Lemma 8.3. If a localization exists, it is uniquely determined up to Joyal equivalence.

Proof. Let i : C → X and j : C → Y be localizations of C along S. By the universal property
we obtain a diagram
C
j
i
F
X Y
G
where F is a functor such that F i ' j and G is such that Gj ' i. We now want to show
that F G ' idY and that GF ' idX . By the universal property, it again suffices to show that
94 M. LAND

these equations hold after precomposing with j and i respectively. There we find that
F Gj ' F i ' j
and likewise that
GF i ' Gj ' i.


In order to prove that localizations exist, we will first need the following lemma.
Lemma 8.4. The map ∆1 → J is a localization at the unique morphism from 0 to 1.
Proof. Let D be an ∞-category. We already know that the restriction map factors as follows:
Fun(J, D) → Fun' (∆1 , D) ⊆ Fun(∆1 , D)
and need to show that the first map is a Joyal equivalence. We will show that it is in fact a
trivial fibration and conclude the lemma using Proposition 6.12. For this we will consider the
following filtration Fk (J) of J with F1 (J) = ∆1 . We consider the non-degenerate k-simplex
νk : ∆k → J given by the string of composable morphisms
0 → 1 → 0 → ...
and we let Fk (J) be the smallest sub simplicial set of J which contains this k-simplex. We
observe that ν1 (∆1 ) ⊆ J is the canonical inclusion. In addition, for each k ≥ 2, we have that
νk (∆{0,1} ) = ν1 (∆1 ) ⊆ J. We claim that there is a pushout diagram as follows.

Λk0 Fk−1 (J)

∆k Fk (J)

To see this, we observe that clearly the image of νk union Fk−1 (J) equals Fk (J). It then
suffices to see that their intersection is given by Λk0 . For this we consider the composite
∆[k]\{i} → ∆k → Fk (J). For i = 0, this is given by the sequence of k − 1 composable maps
1 → 0 → 1 → ...
which is not contained in Fk−1 (J). However, if i 6= 0, then it is given by a sequence starting
with 0 of length k − 1 and is hence contained in Fk−1 (J) by definition.

Lecture 2 – 29.04.2019. We want to show that the map

Fun(J, D) → Fun' (∆1 , D)
is a Joyal equivalence. This map factors as follows
1 1
Fun(J, D) → Fun∆ (Fk (J), D) → Fun∆ (Fk−1 (J), D) → Fun' (∆1 , D).
We will show that the map in the middle is a trivial fibration for all k ≥ 2. It follows that
also the map
∼ lim Fun∆1 (Fk (J), D) → Fun' (∆1 , D)
Fun(J, D) =
k
 INTRODUCTION TO INFINITY-CATEGORIES 95

is a trivial fibration and hence a Joyal equivalence. We thus need to show that for every
commutative diagram
1 1
∂∆n Fun∆ (Fk (J), D) Fun∆ (∆k , D)

x 1 1
∆0 ∆n Fun∆ (Fk−1 (J), D) Fun∆ (Λk0 , D)
there exists a dashed arrow making everything commute. We claim that it suffices to find a
dotted arrow: It is clear that the right square is a pullback if we drop the superscript ∆1 , so
that the pullback consists of all functors Fk (J) → D, whose restriction to ∆k send ∆1 to an
equivalence. This shows that the right square is a pullback.
By adjunction, this lifting problem corresponds to the lifting problem
evx
∆1 Λk0 Fun(∆n , D) D

∆k Fun(∂∆n , D)
and we observe that the top horizontal composite is an equivalence for every object x of
∆n . Once we know this, the dashed arrow exists by Joyal’s extension theorem, because the
right vertical map is an inner fibration and the functor Fun(∆n , D) → D is conservative
Q
x
by Theorem 6.1. 
Lemma 8.5. For every ∞-category C, there exists a localization along all morphisms of C.
Proof. We first construct an anodyne map f : C → X to a Kan complex (an ∞-groupoid) X
as follows. We let Y be the pushout
C
` 1
∆
α∈C1
f

`
J Y
α∈C1

Since ∆1 → J is anodyne, as follows from the proof of Lemma 8.4, it follows that also f is
anodyne. Then we take an inner anodyne map g : Y → X with X an ∞-category: Via the
small object argument we can factor the map Y → ∗ through an inner anodyne map followed
by an inner fibration. Since g is inner anodyne, we see that the composite gf is anodyne
and we claim that X is in fact an ∞-groupoid. For this we need to show that its homotopy
category is a groupoid.
Since the map Y → X is inner anodyne, it is a Joyal equivalence, and thus induces an
equivalence on homotopy categories. Furthermore, taking homotopy categories is left adjoint
to the nerve functor, and hence preserves pushouts, so that there is a pushout of categories
qα∈C1 [1] hC

qα∈C1 J hY
96 M. LAND

Thus hY is obtained from hC by inverting all morphisms in C Exercise 97. In particular, it is

a groupoid, and thus so is hX.
Finally, we claim that for every ∞-category D, the restriction functor Fun(X, D) →
Fun(C, D) factors through a trivial fibration
Fun(X, D) → Fun' (C, D).
Since trivial fibrations are Joyal equivalences by Proposition 6.12, the map C → X is a
localization.
It is clear that Fun(X, D) → Fun(C, D) factors through Fun' (C, D) because every mor-
phism in C maps to an equivalence in X, since X is an ∞-groupoid. The map of interest now
factors as
Fun(X, D) → Fun(Y, D) → Fun' (C, D) ⊆ Fun(C, D).
The first map is a trivial fibration by Theorem 3.32 part (2) since Y → X is an inner fibration.
The second map sits inside a diagram
Fun(Y, D) Fun' (C, D) Fun(C, D)

'
Fun(J, C) Fun' (∆1 , C) Fun(∆1 , D)
Q Q Q
α∈C1 α∈C1 α∈C1

in which the big square is a pullback by definition of Y , and the right square is a pullback by
inspection. It follows that the left square is a pullback as well. By Lemma 8.4 the left lower
horizontal map is a trivial fibration, thus so is the upper horizontal map. This finishes the
proof of the lemma. 
Remark. We will see later that the association of sending C to the localization along all
morphisms is a left adjoint to the inclusion of ∞-groupoids into ∞-categories (as ∞-functors
between ∞-categories).

Proposition 8.6. For every S ⊆ C1 , there exists a localization of C along S.

Proof. We observe that for every subset S ⊆ C1 there is a smallest subcategory CS of C which
contains S: this is clear for categories, and the statement here follows by pulling back the
corresponding subcategory of the homotopy category of C. It follows easily that a localization
of C along CS is a localization of C at S, see also Exercise 94.
We thus take a localization of CS along all morphisms, more precisely an inner anodyne
map CS → X to an ∞-groupoid X as in Lemma 8.5. Then we consider the pushout
CS C

X W
and then consider an inner anodyne map g : W → D with D an ∞-category. Then, for an
auxiliary ∞-category E, consider the diagram
g∗
Fun(D, E) Fun(W, E) FunS (C, E)

Fun(X, E) FunS (CS , E).

 INTRODUCTION TO INFINITY-CATEGORIES 97

We claim that the right hand square is pullback diagram: To see this, it suffices to observe
that every functor W → E sends the morphisms in the image of S to equivalences, which
follows from the fact that they are sent to equivalences in X. Moreover, the lower horizontal
map is a trivial fibration by the previous step, thus the upper horizontal map is also a trivial
fibration. The map g ∗ is a trivial fibration, since W → D is inner anodyne. Thus the upper
composite is a trivial fibration, and thus a Joyal equivalence. This shows that the map C → D
is a localization along S. 
Apart from the fact that the procedure of “universally inverting” morphisms produces many
interesting examples of ∞-categories (even if the category we start out with is an ordinary
category), we will use it to prove a certain factorization property of functors between ∞-
categories later.
Example 8.7. Consider the 1-category Cat1∞ of ∞-categories, i.e. the full subcategory of
sSet whose objects are the ∞-categories. Recall that the ∞-category Cat∞ of ∞-categories
is given by the homotopy coherent nerve N(Cat1∞ ) of this category with its canonical Kan en-
richment given by Fun(C, D)' . The identity of Cat1∞ canonically refines to a functor between
simplicial categories, with constant simplicial enrichment on the domain and the canonical
simplicial enrichment on the target. In other words, we obtain a canonical functor of ∞-
categories
Cat1∞ −→ Cat∞ .
We observe that his functor sends Joyal equivalences to equivalences: By definition, a Joyal
equivalence is a map of ∞-categories which becomes an equivalence in the ∞-category Cat∞ ,
see Definition 5.12. It follows that this functor induces a functor
Cat1∞ [Joy−1 ] −→ Cat∞
where Joy denotes the set of Joyal equivalences.

Example 8.8. Likewise, there is a canonical functor Kan −→ An c which sends homotopy
equivalences to equivalences in An.
c Hence there is an induced functor

Kan[he−1 ] −→ An.
c

Lemma 8.9. The inclusions Kan → sSet and Cat1∞ → sSet induce equivalences
Kan[he−1 ] ' sSet[we−1 ] and Cat1∞ [Joy−1 ] ' sSet[Joy−1 ].
Proof. The small object argument provides functors F : sSet → Cat1∞ and G : sSet → Kan
by functorially factoring the map X → ∗ into an inner anodyne map followed by an inner
fibration, respectively by an anodyne map followed by a Kan fibration. We claim that these
functors send Joyal equivalences to Joyal equivalences, respectively weak homotopy equiva-
lences to weak homotopy equivalences: Suppose X → Y is such an equivalence, then we have
a commutative diagram
X C

Y D
where the horizontal maps are inner anodyne and hence Joyal equivalences. We find that
C → D is a Joyal equivalence if and only X → Y is, by 3-for-2 for Joyal equivalences. The
98 M. LAND

argument for weak homotopy equivalences is the same. Hence this functor induces a functor
sSet[Joy−1 ] −→ Cat1∞ [Joy−1 ]
which we claim to be an inverse to the canonical functor
Cat1∞ [Joy−1 ] −→ sSet[Joy−1 ]
induced by the inclusion.
We observe that the map X → C = F (X) determines a natural transformation from the
identity of sSet to the composite iF : sSet → sSet. More precisely it determines a map
∆1 −→ Fun(sSet, sSet) −→ Fun(sSet, sSet[Joy−1 ])
which we claim to land inside the full subcategory consisting of those functors that send
Joyal equivalences to equivalences: This is because we have just argued that F (and also
clearly i) send Joyal equivalences to Joyal equivalences. Hence, by the universal property of
localizations, we obtain a map
∆1 −→ Fun(sSet[Joy−1 ], sSet[Joy−1 ])
whose restriction to 0 and 1 are given by the identity and a functor, whose restriction to sSet
is induced by the composite F i. We observe that for fixed object X ∈ sSet, the resulting
morphism is given by X → F (X) which we argued to be a Joyal equivalence. In particular
the 1-simplex given above is a natural equivalence between the identity of sSet[Joy−1 ] and
the functor induced by F i.
It remains to show that also the functor induced by the composite iF is an equivalence. For
this we argue analogously and find that the map X → F X induces a transformation id → iF
which is pointwise a Joyal equivalence. This shows the corollary for Joyal equivalences, and
the argument for weak homotopy equivalences is the same. 

The following theorem is very important, but at the moment beyond the scope of this
lecture. It holds in greater generality for simplicial model categories, see [Lur17, Theorem
[Link]]. We will sketch a proof of this result later, see Corollary 11.26 and Corollary 11.27.
Theorem 8.10. The canonical functors sSet[we−1 ] → An and sSet[Joy−1 ] → Cat∞ are
equivalences of ∞-categories.
We finish this section with a useful factorization construction.
Definition 8.11. Let f : C → D be a functor between ∞-categories. We define the path-
fibration P (f ) of f by the pullback

P (f ) Fun(J, D)
s
f
C D
of simplicial sets, where s is the source map.

Consider the map C → Fun(J, D) which is adjoint to the map C × J → C → D. It

sends an object x to the identity morphism of f (x). Clearly the composition of this functor
with s is given by f so we obtain an induced map c : C → P (f ) which is a section of the
 INTRODUCTION TO INFINITY-CATEGORIES 99

canonical map P (f ) → C. In particular, c is a monomorphism. Furthermore, the composite

of P (f ) → Fun(J, D) with the target map Fun(J, D) → D produces a composite
c t
C → P (f ) → D.
This composite is given by f , so we have produced a factorization of f through P (f ).
Lecture 3 – 06.05.2019.
c t
Lemma 8.12. Let f : C → D be a functor between ∞-categories and let C → P (f ) → D
be the factorization just constructed. Then c is Joyal equivalence and t is an isofibration. In
particular, any map f : C → D between ∞-categories can be factored as a Joyal equivalence
which is a monomorphism, followed by an isofibration.
Proof. We claim that the source map s : Fun(J, D) → D is a trivial fibration. By Lemma 6.16
it suffices to show that it is a Joyal equivalence and an isofibration. We have already seen
that ∆0 → J is a Joyal equivalence, so it follows from Proposition 6.9 that also the induced
map Fun(J, D) → Fun(∆0 , D) ∼ = D is a Joyal equivalence. By Proposition 6.6 the restriction
along a monomorphism mapping into an ∞-category is an isofibration, so we conclude that
Fun(J, D) → D is a trivial fibration as claimed. Thus, as a pullback of this map, also the
functor P (f ) → C is a trivial fibration. Since the composite
C → P (f ) → C
is the identity (by construction), it follows from 3-for-2 for Joyal equivalences that the map
C → P (f ) is a Joyal equivalence as claimed.
To see that P (f ) → D is an isofibration we observe that the square
P (f ) Fun(J, D)

f ×id
C×D D×D
is also a pullback and that the right vertical map is an isofibration again by Proposition 6.6. It
follows that P (f ) → C×D is also an isofibration, Corollary 5.18. As the projection C×D → D
is also an isofibration, the lemma is proven. 

9. Fat joins, fat slices and mapping spaces

In this section we will construct an alternative join, and show that it is Joyal equivalent
to the construction of Definition 4.6. This is used to compare different models of mapping
spaces in an ∞-category.
Definition 9.1. Let X and Y be simplicial sets. We define a new simplicial set X  Y to
be the pushout
X × Y × ∂∆1 X qY

X × Y × ∆1 X Y
Here, the top horizontal arrow sends the triple (x, y, 0) to x and the triple (x, y, 1) to y.

Lemma 9.2. For fixed Y , the association X 7→ X  Y extends to a functor sSet → sSetY / .
As such, it preserves colimits. The same holds for the association X 7→ Y  X.
100 M. LAND

Proof. The first statement is an easy calculation: we need to see that for any morphism
X → X 0 the diagram
Y X Y

X0  Y
commutes. This follows simply from the fact that the diagram
Y X qY

X0 q Y
commutes. To see that this functor commutes with colimits, it again suffices to show that it
commutes with coproducts and with coequalizers. So let X, X 0 and Y be simplicial sets. We
need to show that the diagram
Y X Y

X0  Y (X q X 0 )  Y
is a pushout in simplicial sets, which follows immediately from the definition. We recall that
coequalizers in sSetY / are calculated underlying, see ??. So let

X X0 C
be a coequalizer diagram. We need to show that applying −  Y gives again a coequalizer
diagram. Then functors − × Y × ∂∆1 and − × Y × ∆1 are left adjoint and hence preserve
this coequalizer diagram. We claim that the diagram
X qY X0 q Y C qY
is also a coequalizer: Suppose given a map X 0 q Y → T whose precomposition with the
two given maps give the same map X q Y → T , we find that there is a unique map from
C → T and from Y to T as needed. Now since two colimits always commute, we may take
the pushout over the coequalizers and obtain the coequalizer of the pushout. This proves the
lemma. 
Observation 9.3. There exists a canonical map of simplicial sets X  Y → ∆1 induced by
the commutative diagram
X × Y × ∂∆1 X qY

pr
X × Y × ∆1 ∆1
where the right vertical map is given by X → ∆{0} a combined with Y → ∆{1} .

Lemma 9.4. Let X and Y be simplicial sets. Then there exists a canonical map X  Y →
X ? Y which commutes with the projections to ∆1 and the inclusions of X and Y . This map
is functorial in X and Y .
 INTRODUCTION TO INFINITY-CATEGORIES 101

Proof. Recall from Lemma 4.9 that for every map p : K → ∆1 of simplicial sets, there exists a
canonical factorization into K → K0 ? K1 → ∆1 , where Ki = p−1 ({i}). We apply this to the
map X  Y → ∆1 we have just constructed. We find that (X  Y )0 is given by the pushout

X × Y × {0} X

X × Y × {0} X

and likewise that (X  Y )1 is given by Y . 

We will now need Lemma 7.12 (Reedy’s lemma) in the following context:
Proposition 9.5. Consider a pullback diagram of ∞-categories
f
C D
p0 p
f0
C0 D0
in which the map p is an isofibration and the map f 0 is a Joyal equivalence. Then the map f
is also a Joyal equivalence.
Proof. We will show that for every ∞-category E, the induced map
f∗
π0 (Fun(E, C)' ) −→ π0 (Fun(E, D)' )
is a bijection. Once this is shown, we can choose E = D and find a functor g : D → C such
that f∗ (g) = f g ' idD . It is then easy to see that g is an inverse of f .
We then observe that the diagram

Fun(E, C)' Fun(E, D)'

Fun(E, C0 )' Fun(E, D0 )'

is a pullback diagram of Kan complexes, in which the right vertical map is a Kan fibration
and the lower horizontal map is a homotopy equivalence. In Lemma 7.12 we have shown that
this implies that the top horizontal arrow is also a homotopy equivalence.
In fact, in the proof we have explicitly shown that the induced map on π0 is a bijection.
For sake of completeness, we reproduce the argument here. So let us assume given a pullback
diagram of Kan complexes
C ×A B B

C A
such that the map B → A is a Kan fibration and the map C → A is a homotopy equivalence.
We aim to show that the map π0 (C ×A B) → π0 (B) is a bijection.
Let’s prove surjectivity of this map first: pick a point b in B representing a class [b] ∈ π0 (B)
and consider the point p(b) in A. Since C → A induces a bijection on path components, we
102 M. LAND

can find a point c in C and a path ∆1 → A connecting f (c) to p(b). This gives, as in
Lemma 6.15, a lifting problem
{0} B

∆1 A
which can be solved as p is a Kan fibration. We thus find that there exists a b0 in B such that
p(b0 ) = f (c) and such that [b0 ] = [b] in π0 (B). The pair (c, b0 ) thus determines an element of
π0 (C ×A B) which is sent to [b] in π0 (B). This shows that the map π0 (C ×A B) → π0 (B) is
surjective.
To show injectivity, consider two points (c, b) and (c0 , b0 ) of C ×A B and assume that there
is a path α : ∆1 → B connecting b and b0 in B. Then p(α) : ∆1 → A connects p(b) = f (c) to
p(b0 ) = f (c0 ). Since the map f is a homotopy equivalence, there is a path β : ∆1 → C such
that f∗ (β) is equivalent to p∗ (α). Precisely, we find a 2-cell σ : ∆2 → A such that
(1) σ|∆{0,1} = f∗ (β),
(2) σ|∆{1,2} = idf (c0 ) , and
(3) σ|∆{0,2} = p∗ (α).
Since we can lift both p∗ (α) and idf (c0 ) along p, similarly as in Lemma 6.15, we find a diagram

Λ22 B

σ
∆2 A

which admits a dashed arrow as indicated since p is a Kan fibration. It follows that there
exists a path γ : ∆1 → B connecting b and b0 such that p∗ (γ) = f∗ (β). Hence γ and β combine
to a map ∆1 → C ×A B connecting (c, b) and (c0 , b0 ). This shows that the map in question is
also injective. 

Lemma 9.6. Consider a diagram of ∞-categories

C D E
' ' '

C0 D0 E0

in which the vertical maps are Joyal equivalences and the left horizontal maps are isofibrations.
Then the induced map on pullbacks

C ×D E → C0 ×D0 E0

is again a Joyal equivalence.

Proof. Copying the proof of Lemma 7.13 we again first explain how to reduce to the case
where the two maps E → D and E0 → D0 are also isofibrations. Here we simply functorially
factor the maps as in Lemma 8.12 as a monomorphism which is a Joyal equivalence followed
 INTRODUCTION TO INFINITY-CATEGORIES 103

by an isofibration to obtain a diagram

C D P ' E
' ' ' '

C0 D0 P0 ' E0

in which all vertical maps are Joyal equivalences. We notice that P and P0 are also ∞-
categories. We then obtain a commutative diagram

C ×D E C0 ×D0 E0

C ×D P C0 ×D0 P0
in which we claim the vertical maps to be Joyal equivalences: For instance, the left vertical
map sits inside a diagram
C ×D E C ×D P C

'
E P D
in which both squares are pullbacks. We may thus apply Proposition 9.5 to the left square.
The argument for the right vertical map above is the same. We have thus reduced to the case
where all horizontal maps in the diagram of the statement of the lemma are isofibrations.
Now it is a 3-fold application of Proposition 9.5 to obtain the lemma as in Lemma 7.13, by
considering the following factorization of the map in question:
C ×D E → (C0 ×D0 D) ×D E ∼
= C0 ×D0 E → C0 ×D0 E0 .

Corollary 9.7. Suppose given a diagram of simplicial sets
X Y Z

X0 Y0 Z0
in which all vertical maps are Joyal equivalences and the left horizontal maps are monomor-
phisms. Then the induced map
X qY Z → X 0 qY 0 Z 0
is again a Joyal equivalence.
Proof. Let C be an ∞-category. By definition, we need to show that the map
Fun(X 0 qY 0 Z 0 , C) → Fun(X qY Z, C)
is a Joyal equivalence. This map is isomorphic to the map
Fun(X 0 , C) ×Fun(Y 0 ,C) Fun(Z 0 , C) → Fun(X, C) ×Fun(Y,C) Fun(Z, C)
which is a Joyal equivalence by Lemma 9.6 and the assumptions. 
104 M. LAND

Corollary 9.8. For all monomorphisms A → B, the pushout product map

J × A q{0}×A {0} × B −→ J × B
is a Joyal equivalence.
Proof. We consider the following diagram
{0} × A {0} × B

J ×A J × A q{0}×A {0} × B

J ×B
The horizontal arrows are monomorphisms and the left vertical arrow and the right bended
arrow are Joyal equivalences: This is because {0} → J is a Joyal equivalence and Joyal
equivalences are closed under finite products as shown in Exercise 102. By Corollary 9.7 the
right vertical map is also a Joyal equivalence, so the lemma follows from the 3-for-2 property
for Joyal equivalences. 
Definition 9.9. A map of simplicial sets is called J-anodyne if it belongs the saturated set
generated by inner anodyne maps and maps of the form J × A q{0}×A {0} × B −→ J × B for
a monomorphism A → B.

Corollary 9.10. Every J-anodyne map is a Joyal equivalence.

Proof. We claim that monomorphisms which are Joyal equivalences themselves form a satu-
rated set, which we leave as Exercise 103. It then suffices to know that inner anodyne maps
are Joyal equivalences by ?? and that the maps of the form J × A q{0}×A {0} × B −→ J × B
are also Joyal equivalences, which is Corollary 9.8. 
Proof. It is closed under pushouts by because mapping a monomorphism an ∞-category gives
an isofibration. If the mono is in addition a Joyal equivalence, the isofibration will also be a
Joyal equivalence and thus a trivial fibration. Such are closed under pullbacks. Since both
monomorphisms and Joyal equivalences are closed under retracts, it remains to show that it is
closed under countable composition: Mapping a monomorphism which is a Joyal equivalence
to an ∞-category produces an isofibration which is also a Joyal equivalence and thus a trivial
fibration Lemma 6.16. Since trivial fibrations are closed under countable composition (the
map from the inverse limit to any stage is a trivial fibration) the claim follows. 

Lecture 4 – 08.05.2019. We can use Corollary 9.10 to give a smaller generating set for
J-anodyne maps, all of whose domains are finite simplicial sets. This will help to apply the
small object argument in the way presented here, which needs this technical assumption.
Proposition 9.11. The set of J-anodyne maps is the smallest saturated set containing inner
anodyne maps and the map {0} → J.
Proof. For notational convenience, let us call the smallest saturated set containing inner
anodyne maps and the map {0} → J the set of super-J-anodyne maps. It is clear that super-
J-anodyne maps are J-anodyne, so it suffices to show that for all monomorphisms A → B,
 INTRODUCTION TO INFINITY-CATEGORIES 105

the pushout product map

{0} × B q{0}×A J × A → J × B
super-J-anodyne. The set of monomorphisms A → B for which this is the case is itself
saturated, so it suffices to show the claim for A → B being the boundary inclusions ∂∆n → ∆n
for n ≥ 0. We use the small object argument for the set of super-J-anodyne maps and obtain
a factorization of the map we are interested in
i p
{0} × ∆n q{0}×∂∆n J × ∂∆n → C → J × ∆n

in which i is super-J-anodyne and the second map satisfies the RLP with respect to inner horn
inclusions and {0} → J. Since J × ∆n is an ∞-category, we deduce that the map C → J × ∆n
is an isofibration between ∞-categories. Since J-anodyne maps are Joyal equivalences by
Corollary 9.10, and super-J-anodyne maps are J-anodyne, we find that the map C → J × ∆n
is in fact a trivial fibration and thus admits a solution s : J × ∆n → C to the lifting problem
i
{0} × ∆n q{0}×∂∆n J × ∂∆n C
s
p

J × ∆n J × ∆n

Considering the diagram

{0} × ∆n q{0}×∂∆n J × ∂∆n J × ∆n

s

i
{0} × ∆n q{0}×∂∆n J × ∂∆n C
p

{0} × ∆n q{0}×∂∆n J × ∂∆n J × ∆n

we find that the map we are interested in is a retract of the super-J-anodyne map i and hence
is itself super-J-anodyne. 

If the target of a monomorphism is an ∞-category, we can prove the following strengthening

of Corollary 9.10.

Proposition 9.12. Let i : A → B be a monomorphism with B an ∞-category. Then i is

J-anodyne if and only if i is a Joyal equivalence.

Proof. By Corollary 9.10 it suffices to show the “if” part. We apply the small object argument
and factor the map A → B as a composition

A → B0 → B

where the map A → B0 is J-anodyne and the map B0 → B satisfies the RLP with respect to
J-anodyne maps. Since B is an ∞-category so is B0 and the map B0 → B is an isofibration.
By Corollary 9.10 and 3-for-2 for Joyal fibrations, the map B0 → B is an isofibration and a
Joyal equivalence and thus a trivial fibration by Lemma 6.16. Choosing a solution s : B → B0
106 M. LAND

of the lifting problem

A B0
s

B B
we find as in the proof of Proposition 9.11 that the map A → B is a retract of the J-anodyne
map A → B0 and thus is itself J-anodyne. 
We will later need the following theorem. Thanks to Hoang Kim Nguyen for the explanation
of the needed reduction steps.
Theorem 9.13. Let p : C → D be an isofibration between ∞-categories and let A → B be a
monomorphism which is in addition a Joyal equivalence. Then any lifting problem
A C
p

B D
admits a solution as indicated by the dashed arrow.
Proof. Using the small object argument, we can factor the map B → D through an inner
andoyne map B → B followed by a map B → D satisfying the RLP wrt J-anodyne maps. It
follows that B is an ∞-category. Since inner anodyne maps are Joyal equivalences we obtain
that the composite A → B is a monomorphism and a Joyal equivalence. By Proposition 9.12
this map is J-anodyne. Since isofibrations between ∞-categories have the RLP with respect
to J-anodyne maps, we can find a dashed arrow in the diagram
A C

B D
which also solves the original lifting problem. 
We finish this intermezzo on J-anodyne maps with a nice fact about inner anodyne maps.
The following lemma is taken from Stevenson [Ste18, Lemma 2.19]
Lemma 9.14. Let A → C be a monomorphism which is a bijection on 0-simplices, a Joyal
equivalence and where C is an ∞-category. Then i is inner anodyne.
j p
Proof. By the small object argument we may factor this map as a composite A → B → C
with j an inner anodyne map and p an inner fibration. Since C is an ∞-category, p satisfies
the assumptions of ?? and hence admits a section. As in the proof of Proposition 9.12 this
section shows that i is a retract of j and is thus inner anodyne as well. 
Corollary 9.15. Let K → L be a monomorphism which is a bijection on 0-simplices. Then
the map
{0} × L q{0}×K J × K → J × L
is inner anodyne.
 INTRODUCTION TO INFINITY-CATEGORIES 107

Proof. It suffices to prove the claim for the maps ∂∆n → ∆n with n ≥ 1. By construction, this
map is then a bijection on 0-simplices, a Joyal equivalence by Corollary 9.8, and has target
an ∞-category because ∆n and J are ∞-categories. Applying Lemma 9.14 we conclude the
corollary. 
We now come back to properties of the fat join which will be needed later.
Corollary 9.16. Let X → X 0 be a Joyal equivalence between simplicial sets and let Y be
a simplicial set. Then the map X  Y → X 0  Y is a Joyal equivalence. Likewise, the map
Y  X → Y  X 0 is a Joyal equivalence.
Proof. Since X  Y is the pushout
X × Y × ∂∆1 X qY

X × Y × ∆1 X Y
in which the left vertical map is a monomorphism, it suffices by Corollary 9.7 to show that
the maps induced by X → X 0 on the other three corners are Joyal equivalences. This follows
from Exercise 102. 
Definition 9.17. A map of simplicial sets f : X → Y is said to admit a pre-inverse if there
exists maps g : Y → X and τ : ∆1 → Hom(X, X) and τ 0 : ∆1 → Hom(Y, Y ) such that
(1) τε = idX and τ1+ε = gf , where ε ∈ {0, 1} ∼ = Z/2,
0
(2) τε0 = idY and τ1+ε = f g, where again ε ∈ {0, 1} ∼= Z/2,
(3) for all objects x of X, the morphism τ (x) : ∆1 → X represents a degenerate edge of
X, and for all objects y of Y , τ 0 (y) : ∆1 → Y represents a degenerate edge of Y .

Proposition 9.18. Let X and Y be simplicial sets. Then the canonical map X  Y → X ? Y
of Lemma 9.4 is a Joyal equivalence.
Proof. As noted in Lemma 9.2 and Lemma 4.14 both functors −  Y and − ? Y commute
with filtered colimits. We may therefore reduce the general situation to the case where X
has only finitely many non-degenerate simplices. In this case we can write X as a pushout
X 0 q∂∆n ∆n . Since pushouts are connected colimits, we also obtain isomorphisms
X Y ∼= X 0  Y q∂∆n Y ∆n  Y
and likewise for ? in place of . It hence suffices to show that claim for X 0 , ∂∆n and ∆n .
By the same reasoning, the statement for ∆n (for all n) implies the one for ∂∆n and X 0 by
induction. Hence it remains to show that
∆n  Y → ∆n ? Y
is a Joyal equivalence. Now the inclusion I n → ∆n is inner anodyne by Proposition 3.17, and
thus in the diagram
In  Y In ? Y

∆n  Y ∆n ? Y
both vertical maps are Joyal equivalences: For the left vertical map this is Corollary 9.16 and
for the right vertical map it is the fact that − ? Y preserves inner anodyne maps, Lemma 4.22
108 M. LAND

part (1) (in fact it even sends right anodyne maps to inner anodyne maps). It hence suffices
to prove the statement for I n . Since I n ∼
= I n−1 q∆0 ∆1 it finally suffices to treat the case
where X is either ∆ or ∆ . Now we observe that ∆0 is a retract of ∆1 , and so the map
0 1

∆0  Y → ∆0 ? Y
is a retract of the map
∆1  Y → ∆1 ? Y.
Since retracts of Joyal equivalences are Joyal equivalences, see Exercise 103, it suffices to
show that the latter is a Joyal equivalence for all Y . Performing the same reductions to Y , it
suffices to finally show that the map
∆1  ∆1 → ∆1 ? ∆1 ∼ = ∆3
is a Joyal equivalence. For this we will (almost) construct a pre-inverse for this map. In
fact, we will construct a zig-zag of such pre-inverses connecting the identity to a now to be
constructed inverse. To construct this map we first observe from the definitions that there is
a canonical quotient map (∆1 )×3 → ∆1  ∆1 . We claim that the composite
can : (∆1 )×3 → ∆1  ∆1 → ∆3
is given by the formula (
a if c = 0,
(a, b, c) =
b+2 if c = 1.
This is just an explicit check of the definitions. We then consider the 3-simplex σ of (∆1 )×3
represented by
(000) → (100) → (101) → (111)
1 1
and its image in ∆  ∆ . We first observe that the composite
∆3 → ∆1  ∆1 → ∆3
is the identity.
Lecture 5 – 20.05.2019. We then must construct a map ∆1 → Hom(∆1 ∆1 , ∆1 ∆1 ) which
exhibits the map σ as pre-inverse of the canonical map ∆1  ∆1 → ∆3 . We will construct
two maps ∆1 → Hom((∆1 )×3 , (∆1 )×3 ) connecting the identity to an auxiliary map Φ and
Φ to σ ◦ can. Then show that they descend to maps between ∆1  ∆1 and finally that the
resulting maps both have the property that for fixed object, the induced edge of ∆1  ∆1 is a
degenerate edge. In the following picture, the left cube represents the identity of (∆1 )3 , the
middle cube represents the map Φ and the right cube represents the composite σ ◦ can. There
are evident maps from the middle cube to both outer cubes.
011 111 011 111 111 111
010 110 000 100 000 100
001 101 001 001 101 101
000 100 000 100 000 100
Now one needs to check that these maps descend, after post composition with the projection
to the quotient ∆1  ∆1 . To do so, we first observe that all cubes restrict to endomorphisms of
∆1 × ∆1 × ∂∆1 . Concretely, this means that if we only look at the front and back layer (that
is we neglect the diagonal maps) and project to the third coordinate, only identity morphisms
 INTRODUCTION TO INFINITY-CATEGORIES 109

remain. Next we claim that there exist compatible endomorphisms of ∆1 q ∆1 : Both for Φ
and σ ◦ can the identity of ∆1 q ∆1 is compatible. This shows that Φ and σ ◦ can descend to
the pushout ∆1  ∆1 .
Alternatively, one sees that ∆1  ∆1 is the quotient of (∆1 )×3 where the subsimplicial
set ∆1 × ∆1 × {0} is collapsed (via the first projection) to ∆1 and the subsimplicial set
∆1 × ∆1 × {1} is collapsed (via the second projection) to ∆1 . We thus have to see that Φ
and σ ◦ can followed by this projection are suitably invariant, i.e. that they satisfy F (x, y, 0)
is independent of y and F (x, y, 1) is independent of x. This is an explicit check.
We observe that any morphism in ∆1 × ∆1 × ∂∆1 which is mapped to an identity (a
degenerate edge) of ∆1 q∆1 is also mapped to a degenerate edge in ∆1 ∆1 . This shows that all
maps from the middle to the left and right cube are degenerate edges in ∆1  ∆1 . This implies
that the horizontal maps between the cubes also descend 1-simplices of Hom(∆1 ∆1 , ∆1 ∆1 )
and furthermore that for any object x of ∆1  ∆1 , these maps are degenerate which finally
implies that the map σ is a pre-inverse to can. 
Corollary 9.19. Let X → X 0 be a Joyal equivalence and Y be a simplicial set. Then the
map X ? Y → X 0 ? Y is again a Joyal equivalence.
Proof. In the commutative diagram
X Y X0  Y

X ?Y X0 ? Y
the top horizontal map is a Joyal equivalence by Corollary 9.16 and the vertical maps are
Joyal equivalences by Proposition 9.18. 
Definition 9.20. Let p : Y → W be an object of sSetY / . We define the fat slice of p to be
the simplicial set W p/ defined by
(W p/ )n = HomsSetY / (Y  ∆n , W )
and the simplicial set W /p to be given by
(W /p )n = HomsSetY / (∆n  Y, W ).

Lemma 9.21. The functor sSetY / → sSet given by sending p : Y → W to W /p is right

7 W p/ is right adjoin to the functor
adjoint to the functor −  Y . Likewise, the functor p →
Y  −.
Proof. By definition the adjunction bijection holds for representables, and hence for all sim-
plicial sets since the functors −  Y and Y  − preserve colimits by Lemma 9.2. 
Lemma 9.22. Let Y be a simplicial set and p : Y → W a map of simplicial sets. Then
there are canonical maps W/p → W /p and Wp/ → W p/ .
Proof. On n-simplices, we have to give a map
HomsSetY / (Y ? ∆n , W ) → HomsSet/Y (Y  ∆n , W ).
For this, it suffices to recall that there is a map Y  ∆n → Y ? ∆n in sSetY / which is natural
with respect to maps in the simplex category ∆. Likewise for the other slice. 
110 M. LAND

As a consequence of Theorem 9.13 we can establish an analog of Theorem 4.23 for the fat
slices.
Lemma 9.23. Let
i f p
S −→ T −→ C −→ D
be maps of simplicial sets such that i is a monomorphism and p is an isofibration between
∞-categories. Then the functor
Cf / −→ Dpf / ×Dpf i/ Cf i/
is a left fibration. Likewise, the functor
C/f −→ D/pf ×D/pf i C/f i
is a right fibration.
Proof. Let A → B be a left anodyne map and consider a lifting problem

A Cf /

B Dpf / ×Dpf i/ Cf i/
By adjunction, this is equivalent to the lifting problem
T  A qSA S  B C

T B D
in which the left vertical map is a monomorphism. We wish to show that it is a Joyal
equivalence, so that we can allude to Theorem 9.13 to conclude the lemma. We claim that in
the diagram
T  A qSA S  B T ? A qS?A S ? B

T B T ?B
both horizontal maps are Joyal equivalences: For the lower horizontal map this is precisely
Proposition 9.18 and for the upper horizontal map we use that these are pushouts along
monomorphisms, so that a 3-fold application of Proposition 9.18 together with Corollary 9.7
gives the claim. Then we recall from Lemma 4.22 that the right vertical map is inner anodyne
and hence a Joyal equivalence. Again, the argument for the other slice is the same. 
Corollary 9.24. Let p : Y → C be a diagram. Then the functor Cp/ → C is a left fibration
and the functor C/p → C is a right fibration. In particular, both Cp/ and C/p are ∞-categories.
Proof. This is Lemma 9.23 in the special case ∅ → Y → C → ∆0 . 
Proposition 9.25. Let p : Y → C be a diagram. Then the canonical functor
Cp/ −→ Cp/
is a Joyal equivalence. The same is true for the other slice.
 INTRODUCTION TO INFINITY-CATEGORIES 111

Proof. By ?? it suffices to show that for every simplicial set K the canonical functor

Fun(K, Cp/ ) −→ Fun(K, Cp/ )

is a Joyal equivalence. By adjunction, this functor is isomorphic to

Funp/ (Y ? K, C) −→ Funp/ (Y  K, C).

By definition, the subscript p on the left hand side denotes the pullback of simplicial sets

Funp/ (Y ? K, C) Fun(Y ? K, C)

p
∆0 Fun(Y, C)

and likewise for Funp/ (Y  K, C). We observe that the right vertical map is an isofibration
because the map Y → Y ? K is a monomorphism and C is an ∞-category. The map Y  K →
Y ? K induces a map of the pullback diagrams which is an isomorphism on the lower corners,
so by Lemma 9.6 it suffices to show that the functor

Fun(Y ? K, C) −→ Fun(Y  K, C)

is a Joyal equivalence which follows from Proposition 9.18. Notice that we also use here that
the map Y → Y  K is a monomorphism which follows immediately from the definitions. 

Using ordinary slices one can also define right- and left mapping spaces in an ∞-category:

Definition 9.26. Let C be an ∞-category and let x and y be objects of C. We define the
right mapping space by the pullback

mapR
C (x, y) C/y

x
∆0 C

and the left mapping space by

mapL
C (x, y) Cx/

y
∆0 C

Remark. We notice that the map mapR 0

C (x, y) → ∆ is a right fibration and that the map
mapL 0
C (x, y) → ∆ is a left fibration. By ??, both are in fact Kan fibrations so that the left
and right mapping spaces are Kan complexes.

We wish to compare these to the mapping space we have already defined in Definition 3.40.
The following lemma will take care of this.
112 M. LAND

Lemma 9.27. The following two diagrams are pullback diagrams.

C/y Fun(∆1 , C) Cx/ Fun(∆1 , C)

(id,y) (x,id)
C C×C C C×C
Proof. We show that the left square is a pullback diagram. The argument for the right square
is analogues. We have to show that the diagram is a pullback on all n-simplices, so we consider
the following diagram
Hom/y (∆n  ∆0 , C) Hom(∆n  ∆0 , C) Hom(∆n × ∆0 × ∆1 , C)

(id,y)
Hom(∆n , C) Hom(∆n , C) × Hom(∆0 , C) Hom(∆n × ∆0 × ∂∆1 , C)

y
∆0 Hom(∆0 , C)
We wish to show that the big top square is a pullback. The right most square is a pullback by
the very definition of the fat join ∆n  ∆0 . The lower left square is a pullback by inspection.
Combining the left two squares is a pullback by definition. It follows that the top left square
is a pullback. Hence, combining the two top squares, we also obtain a pullback as needed. 
Passing to fibres we obtain the following corollary.
Corollary 9.28. The following diagrams are pullbacks
mapC (x, y) C/y mapC (x, y) Cx/

x y
∆0 C ∆0 C
Lecture 6 – 22.05.2019. We can then use the following lemma to compare the various
definitions of mapping spaces.
Lemma 9.29. Consider a diagram
f
E E0
p
p0
C
p0
where p and are isofibrations and where f is a Joyal equivalence. Then for all objects x in
C, the induced map on fibres Ex → E0x is also a Joyal equivalence.
Proof. This follows immediately from Lemma 9.6. 
We will see a partial converse of this lemma later.
Corollary 9.30. Let C be an ∞-category and let x and y be objects of C. Then the maps
mapR L
C (x, y) → mapC (x, y) ← mapC (x, y)
are homotopy equivalences.
 INTRODUCTION TO INFINITY-CATEGORIES 113

Proof. Apply Lemma 9.29 to the diagrams

C/y C/y Cx/ Cx/

C C
using Corollary 9.24 and Proposition 9.25. 
Finally, we wish to compare the mapping spaces of the coherent nerve of a Kan enriched
category with the mapping Kan complexes that are present in the simplicial category. The
proofs are beyond the scope of these lectures, but see for instance [Lur09, Section 2.2.2].
Theorem 9.31. Let C be a Kan enriched category and let x and y be objects of C. Then
there is a canonical map
HomC (x, y) −→ mapN(C) (x, y)
which is a homotopy equivalence. The homotopy class of this map is natural in x and y.
Corollary 9.32. Let F : C → C0 be a weak equivalence of Kan enriched categories. Then
then induced functor N(F ) : N(C) → N(C0 ) is a Joyal equivalence of ∞-categories.
Proof. The functor N(F ) is essentially surjective if and only if F is weakly essentially surjective
in the sense of ??. Furthermore, Theorem 9.31 shows that F is weakly fully faithful if and
only if N(F ) is fully faithful. Hence we conclude by Theorem 7.16. 
Recall that we have defined the simplicial category CW whose objects are CW-complexes
and whose simplicial set of maps is given by S(map(X, Y )). Its coherent nerve was denoted
by An. We have furthermore defined the simplicial category Kan whose objects are Kan
complexes and whose simplicial set of maps is given by the internal hom Hom(A, B). We
claim that there is a functor CW → Kan constructed as follows: On objects, it sends X to
S(X). On morphisms we have to give a simplicial map
S(map(X, Y )) −→ Hom(S(X), S(Y ))
compatible with composition. By adjunction, this is equivalent to giving a map
S(map(X, Y )) × S(X) −→ S(Y ).
We then recall that S is a right adjoint and hence preserves products. It hence suffices to
give a simplicial map
S(map(X, Y ) × X) −→ S(Y )
where we can use the ordinary composition map map(X, Y ) × X → Y and apply the functor
S.
Corollary 9.33. The previously described functor An = N(CW) → N(Kan) = An
c is a Joyal
equivalence.
Proof. We show that the functor CW → Kan is a weak equivalence of simplicial categories.
It is weakly essentially surjective because every Kan complex X is homotopy equivalent to
S(|X|) and thus up to equivalence in the image of the functor CW → Kan. To show that the
functor is fully faithful, we have to show that the map
S(map(X, Y )) −→ Hom(S(X), S(Y ))
114 M. LAND

is a homotopy equivalence. This is the case if and only if the composite

map(X, Y ) −→ |S(map(X, Y ))| −→ |Hom(S(X), S(Y ))|
is a homotopy equivalence. We will now need to use the fact that for any two Kan complexes
A and B, the canonical map
|Hom(A, B)| −→ map(|A|, |B|)
is a homotopy equivalence. 
Corollary 9.34. The canonical functor An → Cat∞ is fully faithful.
Proof. The functor is given by applying the coherent nerve the the functor of simplicial
categories Kan → Cat1∞ . We thus only need to show that for any two Kan complexes X and
Y , the canonical map
Hom(X, Y ) → Fun(X, Y )'
is a homotopy equivalence. In fact, it is an isomorphism of simplicial sets, so the proposition
follows. 
Remark. Later, we will be able to give a proof of this fact which is not based on comparing
the mapping anima in An and Cat∞ on the level of the simplicial categories which define
these ∞-categories.

10. (Co)Cartesian fibrations

Let p : X → Y be an inner fibration between simplicial sets.
Definition 10.1. A morphism f : ∆1 → X is called p-cartesian if for n ≥ 2 any lifting
problem
f

∆{n−1,n} Λnn X

∆n Y
admits a solution. Dually, it is called p-cocartesian if any lifting problem
f

∆{0,1} Λn0 X

∆n Y
admits a solution. One calls such an f a p-(co)cartesian lift of p(f ).

The definition of (co)cartesian morphisms can be rephrased in terms of slices as follows.

Lemma 10.2. Let p : X → Y be an inner fibration and let f : x → y be a morphism in Y .
Then f is p-cartesian if and only if the functor
X/f −→ X/y ×Y/p(y) Y/pf
 INTRODUCTION TO INFINITY-CATEGORIES 115

is a trivial fibration. Dually, f is p-cocartesian if and only if the functor

Xf / −→ Xx/ ×Yp(x)/ Ypf /
is a trivial fibration.
Proof. Let ∂∆n → ∆n for n ≥ 0 and consider a lifting problem
∂∆n Xf /

∆n X/y ×Y/p(y) Y/pf

which we wish to solve. By adjunction this corresponds to the lifting problem

∂∆n ? ∆1 q∂∆n ?∆0 ∆n ? ∆0 X

∆n ? ∆1 Y
where the restriction to ∆1 is given by f . We recall that the left vertical map is isomorphic
to Λn+2
n+2 → ∆
n+2 and that the inclusion of ∆1 into Λn+2 is the edge ∆{n+1,n+2} . The diagram
n+2
can thus be solved for all n ≥ 0. The cocartesian case is similiar. 
Remark. Suppose that p : X → Y is an inner fibration between ∞-categories. Then the
map X/p → X /p is a Joyal equivalence for any diagram p : W → X by Proposition 9.25.
Furthermore, the map
X /f −→ X /y ×Y /p(y) Y /pf
is also a right fibration by Lemma 9.23. Hence, f is cartesian if and only if this map is a
trivial fibration. Likewise, f is cocartesian if and only if the map
X f / −→ X x/ ×Y p(x)/ Y pf /
is a trivial fibration.

Lemma 10.3. Let p : E → D and q : D → C be inner fibrations between ∞-categories and let
f : ∆1 → E be a morphism in E such that p(f ) is q-(co)cartesian. Then f is p-(co)cartesian
if and only if f is qp-(co)cartesian.
Proof. Consider the diagram
'
E/f E/y ×D/p(y) D/pf E/y ×D/p(y) D/p(y) ×C/qp(y) C/qpf
∼
=

E/y ×C/qp(y) C/qpf

in which the second horizontal map is a trivial fibration by the assumption that pf is q-
cartesian. The first horizontal map and the diagonal map are both right fibrations and thus
trivial fibrations if and only if they are Joyal equivalences. The lemma follows. 
Remark. The previous lemma holds more generally without the assumption that the sim-
plicial sets involved are ∞-categories, see [Lur09, Prop. [Link]]: Since the composition of
trivial fibrations is again a trivial fibration we immediately see that if f is p-cartesian then it is
116 M. LAND

also qp-cartesian. To see the converse, we want to show that the map E/f → E/y ×D/p(y) D/pf
(which is a Joyal equivalence by 3-for-2) in fact has contractible fibres and then allude to the
general fact that a right fibration which has contractible fibres is a trivial fibration (or a Joyal
equivalence and thus a trivial fibration), see [Lur09, Lemma [Link]] (or Theorem 10.21). The
fibre we are interested in is the fibre of the map between fibres of the other two maps. These
are contractible by assumption so we are done.

Lemma 10.4. Let p : E → C be an inner fibration between ∞-categories. Let f : ∆1 → E

be a morphism. Then f is an equivalence if and only if it is p-(co)cartesian and p(f ) is an
equivalence.
Proof. Suppose that f is an equivalence. Then so is p(f ) and the Joyal lifting theorem ??
implies that f is both cartesian and cocartesian. Assume now conversely that p(f ) is an
equivalence so that it is q-cartesian where q : C → ∆0 is the projection, by Exercise 116. We
find that f is p-cartesian if and only if f is qp-cartesian by Lemma 10.3, which in turn is the
case if and only if f is an equivalence by another application of Exercise 116. 
Lemma 10.5. Let p : E → C be an inner fibration and let σ : ∆2 → E which we will depict
as the diagram
y
f ϕ

g
x z
Suppose that ϕ is p-cartesian. Then g is p-cartesian if and only if g is p-cartesian.
Proof. We observe that the inclusions ∆{0,1} → ∆2 and ∆{0,2} → ∆2 are left anodyne, so it
follows that in the two commutative squares
E/σ E/z ×C/p(z) C/pσ E/σ E/y ×C/p(y) C/pσ

E/g E/z ×C/p(z) C/pg E/f E/y ×C/p(y) C/pf

all vertical maps are trivial fibrations. We wish to show the the left lower horizontal map is
a trivial fibration if and only if the right lower horizontal map is. Since each are isofibrations
in any case, it suffices to show that one is a Joyal equivalence if and only if the other is. By
means of the commutative squares, it hence suffices to show that the left top horizontal map
is a Joyal equivalence if and only the right top horizontal map is. For this we consider the
diagram
E/σ

'
E/y ×C/p(y) C/pσ ' E/ϕ ×C/pϕ C/pσ E/z ×C/p(z) C/pσ
Here, the right horizontal map is an equivalence because ϕ is p-cartesian, so that the map
E/ϕ → E/z ×C/p(z) C/pϕ is an equivalence. The left horizontal map is an equivalence because
it is induced by the restriction {1} → ∆{1,2} which is left anodyne. We thus find that the left
bended map is an equivalence if and only if the right bended map is an equivalence, which
was left to show. 
 INTRODUCTION TO INFINITY-CATEGORIES 117

The following notion will be convenient to use.

Definition 10.6. A commutative diagram of ∞-categories
E E0
p

C C0
will be called homotopy cartesian if p is an isofibration and the induced map E → C ×C0 E0 is
a Joyal equivalence. If C = ∆0 we will also use the phrase homotopy fibre sequence for the
composition E → E0 → C0 .

Remark. More generally, one can neglect the condition that p be an isofibration as follows:
' q
One can factor p as a Joyal equivalence followed by an isofibration E0 → E00 → C0 and then
ask that the induced map E → C ×C0 E00 be a Joyal equivalence.

Lecture 7 – 27.05.2019.
Lemma 10.7. Let p : C → D be an isofibration between ∞-categories and let x and y be
objects of C. Then the induced map
mapC (x, y) → mapD (px, py)
is a Kan fibration. The same holds true for mapL and mapR , even if p is only an inner
fibration.
Proof. We consider the diagram
mapC (x, y) mapD (p(x), p(y)) ∆0
x

C/y D/p(y) ×D C C
p

D/p(y) D
respectively the ones with the ordinary slice and its variant using Cx/ instead of C/y . All
squares are pullbacks in this diagram: The lower one is by definition, as is the right large
one, so the small square in the top right corner is a pullback. The combined large horizontal
square is again a pullback by definition, so we deduce the the left small square is a pullback.
The lower horizontal map in this diagram is a right fibration (in the case of Cx/ is is a left
fibration) hence so is the upper horizontal map. Since mapD (px, py) is a Kan complex, it is
in fact a Kan fibration as claimed. 
Lemma 10.8. Let C → D be an isofibration between ∞-categories and let z be an object of
D, and x and y objects of C with px = py = z. Then the diagram
mapCz (x, y) mapC (x, y)

idz
∆0 mapD (z, z)
118 M. LAND

is a pullback.
Proof. To see the remaining claim we consider the diagram
Fx,y mapE (x, y) mapC (z, z)

1 1 1
Ez∆ E∆ C∆

∆0 ∆0 ∆0

E z × Ez E×E C×C
all horizontal composites are given by the fibre inclusion over the point corresponding to idz .
Since the middle and right vertical square are pullbacks, so is the left most vertical square. 
Corollary 10.9. Let C → D be an isofibration between ∞-categories and let z be an object
of D and x and y be objects of Cz . Then the diagram

map?Cz (x, y) map?E (x, y)

id
∆0 map?C (p(x), p(y))

is homotopy cartesian for ? = R, L or void.

Proof. For the ordinary mapping anima, this follows immediately since it is a pullback and
the right vertical map is a Kan fibration. To see the claim for the right and left mapping
anima, we use Lemma 10.7 and Corollary 9.30. 
Lemma 10.10. Let p : E → C be an inner fibration between ∞-categories and let f : ∆1 → E
be a p-cartesian morphism from x to y. Then, for all objects z of E, the induced map
E/f ×E {z} −→ E/y ×C/p(y) C/pf ×E {z}

is a trivial fibration as well.

Proof. We have a commutative diagram

E/f ×E {z} E/f

E/y ×C/p(y) C/pf ×E {z} E/y ×C/p(y) C/pf

∆0 E
in which both the big square and the lower square are pullbacks, and hence so is the upper
square. Since in this square the right vertical map is a trivial fibration by the assumption
that f is p-cartesian, the claim follows. 
 INTRODUCTION TO INFINITY-CATEGORIES 119

Remark. Replacing the fat slice with the ordinary slice, the same statement holds true for
inner fibrations between arbitrary simplicial sets.

Corollary 10.11. Let p : E → C be an inner fibration between ∞-categories and let f : ∆1 →

E be a p-cartesian morphism of E from x to y and let z be an object of E. Then the diagram

mapE (z, x) mapE (z, y)

mapC (p(z), p(x)) mapC (p(z), p(y))

is a homotopy cartesian diagram of ∞-groupoids. Here the horizontal maps are induced by
post composition with f and pf respectively.
Proof. We consider the diagram

mapE (z, x) ' E/f ×E {z} mapE (z, y)

mapC (p(z), p(x)) ' C/pf ×C {p(z)} mapC (p(z), p(y))

and observe that the left horizontal maps are trivial fibrations as the inclusion {0} → ∆1 is
left anodyne. It thus suffices to show that the right square is homotopy cartesian, and since
the right vertical map is a Kan fibration by Lemma 10.7 it suffices to recall that the map
E/f → E/y ×C/p(y) C/pf
is a trivial fibration, so that the same remains true after applying − ×E {z}. 
Remark. The statement of the corollary is not quite correct: The square we construct does
not a priori commute (only up to homotopy). Since the right vertical map is a Kan fibration,
it can always be replaced by a commutative diagram without changing the homotopy types
of the participants and a concrete way of doing this is to consider the right square in the
diagram appearing in the proof.

Remark. Given an inner fibration p : E → C and a morphism f : ∆1 → E such that for

every object z of E the diagram

mapE (z, x) mapE (z, y)

mapC (p(z), p(x)) mapC (p(z), p(y))

is homotopy cartesian, then f is p-cartesian. The needed input is that the map
E/f → E/y ×C/(y) C/pf
is a trivial fibration if and only if it induces an equivalence on each fibre over points in E. As
before, this holds in general for right fibrations via a combinatorial argument (without E and
C having to be ∞-categories). In the case where E and C are ∞-categories, we will prove this
later for so called cartesian fibrations and deduce the version for right fibrations from this.
120 M. LAND

Corollary 10.12. Let p : E → C be a cartesian fibration and let x and y be objects of E. Let
f : x0 → y be a p-cartesian morphism with p(x0 ) = p(x). Then the diagram

mapEp(x) (x, x0 ) mapE (x, y)

pf
∆0 mapC (p(x), p(y))
is homotopy cartesian.
Proof. By Corollary 10.11 we have a homotopy cartesian diagram
mapE (x, x0 ) mapE (x, y)

mapC (p(x), p(x0 )) mapC (p(x), p(y))

so that the induced map of vertical fibres is an equivalence over the point ∆0 → mapC (p(x), p(x0 ))
corresponding to idp(x) . The left vertical fibre is given by mapEp(x) (x, x0 ) by Corollary 10.9 so
the claim is shown. 
Inner fibrations p : E → C which have a sufficient supply of cartesian morphisms thus are
such that the mapping anima in E are controlled by those of C and all fibres.
Definition 10.13. An inner fibration p : X → Y is called a cartesian fibration if every lifting
problem
{1} X

∆1 Y
has a solution which is a p-cartesian morphism in X. Dually, p is called a cocartesian fibration
if every lifting problem
{0} X

∆1 Y
admits a solution which is a p-cocartesian morphism in X.

Informally, an inner fibration is a (co)cartesian fibration if it admits a p-(co)cartesian lift

of any morphism in C (which specified source or target).
Example 10.14. Right fibrations are cartesian fibrations and left fibrations are cocartesian
fibrations.

Lemma 10.15. Let p : E → C be a (co)cartesian fibration between ∞-categories. Then p is

an isofibration.
Proof. This follows from Lemma 10.4 which says that a p-cartesian lift of an equivalence is
an equivalence. 
 INTRODUCTION TO INFINITY-CATEGORIES 121

Lemma 10.16. A cartesian fibration E → C is a right fibration if and only if every morphism
in E is p-cartesian. Dually, a cocartesian fibration is a left fibration if and only if every
morphism in E is p-cocartesian.
Proof. By definition, in a cartesian fibration one can lift the right outer 1-horn. If furthermore
every morphism in E is p-cartesian, this simply says that one can also lift all right outer horns
of dimension greater or equals to two. Conversely, a right fibration admits some lift of the
diagram in the definition, and by definition of a right fibration, every morphism in p-cartesian.
The argument for cocartesian fibrations is the same. 
Proposition 10.17. Let p : E → C be a cartesian fibration between ∞-categories. Then p is
a right fibration if and only if for all objects x of C, the fibres Ex = E ×C {x} are ∞-groupoids.
Proof. Right fibrations are cartesian fibrations whose fibres are ∞-groupoids. Conversely,
assume that p : E → C is a cartesian fibration whose fibres are ∞-groupoids. We will show
that every morphism is p-cartesian and allude to Lemma 10.16. So let f : ∆1 → E be a
morphism from x to y and choose a p-cartesian lift ϕ of p(f ) with target y. We consider the
diagram
(f,ϕ)
Λ22 E

σ
∆2 C
where the map σ is given by the diagram
p(y)
id p(f )

p(f )
p(x) p(y)
Since ϕ is p-cartesian, there exists a dashed arrow in this diagram. The resulting 2-simplex
τ is given by the diagram
y
ψ ϕ

f
x y
where now ψ is a morphism in the fibre Ex over x and is hence invertible by the assumption
that all fibres are ∞-groupoids. We may thus apply Lemma 10.5 and see that f is cartesian
because ϕ is. 
Corollary 10.18. A cartesian fibration is conservative if and only if it is a right fibration.
Proof. Right fibrations are conservative by Proposition 5.7. Conversely, given a conservative
cartesian fibration p : E → C, the fibre Ex over each object x of C is an ∞-groupoid: Each
morphism in the fibre is sent to the identity of x by p. We may thus apply Proposition 10.17.

Definition 10.19. Let p : X → Y and p0 : X 0 → Y be (co)cartesian fibrations. We say that
a map f : X → X 0 is a morphism of (co)cartesian fibrations if p0 f = p and f sends p-cartesian
morphisms to p0 -cartesian morphisms.
122 M. LAND

Example 10.20. Suppose that p : X → Y is a cartesian fibration and that p0 : X 0 → Y is a

right fibration. Then any map f : X → X 0 with p0 f = p is a morphism of cartesian fibrations,
because all morphisms in X 0 are p0 -cartesian.

Lecture 8 – 29.05.2019.
Theorem 10.21. Let f : E → E0 be a morphism of (co)cartesian fibrations p : E → C and
p0 : E0 → C between ∞-categories. Then f is a Joyal equivalence if and only if for all objects
z of C, the induced map on fibres Ez → E0z is a Joyal equivalence.
Proof. The “only if” direction holds more generally for maps between isofibrations, see ??.
Let us hence assume that all induced maps Ex → E0x are Joyal equivalences. We wish to show
that f is a Joyal equivalence. We will show that f is fully faithful and essentially surjective and
conclude the theorem from Theorem 7.16. To see that f is essentially surjective, we consider
an object y 0 of E0 and let x = p0 (y 0 ). Since the map Ex → E0x is a Joyal equivalence, it is
in particular essentially surjective. Hence there exists an object y in Ex and an equivalence
between f (y) and y 0 in E0x . It follows that f is essentially surjective.
To see that f is fully faithful we consider two objects x and y in E and need to show that
the map
mapE (x, y) −→ mapE0 (f (x), f (y))
is a homotopy equivalence.
To see this, we choose a p-cartesian lift α̂ : x0 → y of α (which implies that p(x0 ) = p(x))
and note that f (α̂) is a p0 -cartesian lift of f (α) by assumption. Furthermore, by ?? we have
a diagram of homotopy fibre sequences

mapEp(x) (x, x0 ) mapE0 (f (x), f (x0 ))

p(x)

mapE (x, y) mapE0 (f (x), f (y))

mapC (p(x), p(y)) mapC (p(x), p(y))

and the horizontal map on the base and the fibre is an equivalence by the assumption that f
restricts to a fully faithful functor on the firbes. Thus the middle horizontal map is also an
equivalence by Lemma 7.15. 

11. Marked simplicial sets and marked anodyne maps

We have seen that a left/right fibration is a special kind of (co)cartesian fibration. Since
left fibrations are determined by a right lifting property (with respect to left anodyne maps)
one can ask whether (co)cartesian fibrations are also characterized by a lifting property. This
is not true on the nose, but in it is true in the context of marked simplicial sets.
Definition 11.1. A marked simplicial set is a pair (X, S) where S is a subset of the 1-
simplices of X which contains all degenerate 1-simplices. The elements of S will be called
marked edges in X. There is a corresponding category sSet+ of marked simplicial sets, where
morphisms are required to send marked edges to marked edges.
 INTRODUCTION TO INFINITY-CATEGORIES 123

Example 11.2. Let X be a simplicial set. Then we denote by X [ the marked simplicial set
where an edge is marked if and only if it is degenerate. We denote by X ] the marked simplicial
set in which all morphisms are marked. This produces functors (−)[ , (−)] : sSet → sSet+ .
There are also two functors sSet+ → sSet: The one forgets the marking, and the other takes
the smallest sub simplicial set spanned by the marked 1-simplices.

Example 11.3. Let p : X → Y be a map of simplicial sets. We denote by X \ the marked

simplicial set where an edge is marked if and only if it is p-cocartesian. Thus, if p : X → Y is
a cocartesian fibration, then the map X \ → Y ] is a map of marked simplicial sets.

Definition 11.4. We denote the smallest saturated set containing the following maps of
marked simplicial sets marked left anodyne.
(1) For all 0 < i < n, the maps (Λni )[ → (∆n )[ ,
(2) for every n > 0, the map (Λn0 )s[ → (∆n )s[ where the superscript s[ denotes all
degenerate edges and the special edge ∆{0,1} to be marked,
(3) the map (Λ21 )] q(Λ2 )[ (∆2 )[ → (∆2 )] , and
1
(4) for every ∞-groupoid X, the map X [ → X ] .

Remark. A different (but equivalent) generating set of the marked left anodyne maps is
given by the following maps:
(1’) For all 0 < i < n, the maps (Λni )[ → (∆n )[ ,
(2’) the maps (∆1 )] × (∆1 )[ ∪ {0} × (∆1 )] → (∆1 )] × (∆1 )] ,
(3’) the maps (∆1 )] × (∂∆n )[ ∪ {0} × (∆n )[ → (∆1 )] × (∆n )[ ,
(4’) the map J [ → J ] .

Remark. We could also equally well take the set generated by the maps (1), (2), (30 ), and
(4), we refer to [Lur09, Proposition [Link]] for details. This will be used in Lemma 11.8.

Proposition 11.5. A map of marked simplicial sets p : X → Y has the right lifting property
with respect to all marked left anodyne maps if and only if the following hold
(1) p is an inner fibration,
(2) An edge of X is marked if and only if it is p-cocartesian and its image is marked in
Y,
(3) any lifting problem of marked simplicial sets

{0} X

(∆1 )] Y
can be solved.
Proof. Since (−)[ is a left adjoint to the forgetful functor sSet+ → sSet, we find that p satisfies
the RLP wrt the maps (Λni )[ → (∆n )[ for 0 < i < n if and only if the underlying map of p is
an inner fibration. We now assume that p satisfies the RLP wrt marked left anodyne maps
and show that (2) and (3) hold: Since (∆1 )s[ = (∆1 )] we find that (3) holds. To see that (2)
holds, we first show any marked edge f : ∆1 → X is p-cocartesian. For this, consider a lifting
124 M. LAND

problem

∆{0,1} Λn0 X

∆n Y
which we want to show to admit a solution. Since the above composite is marked, this gives
rise to a diagram of marked simplicial sets

(Λn0 )s[ X

(∆n )s[ Y

which can be solved since the left vertical map is marked anodyne. To show the converse,
consider a p-cocartesian morphism f from x to y in X such that pf : ∆1 → Y is marked.
Consider the diagram
x
{0} X
g

pf
(∆1 )] Y

which can be solved by a previous argument. The resulting morphism g is marked and thus
p-cocartesian. Consider then the diagram
(g,f )
(Λ20 )s[ X
τ

(∆2 )s[ Y

which can again be solved by assumption. Here, the lower horizontal map is a degenerate
2-simplex on the morphism pf = pg. We denote τ|∆{1,2} by h. It follows from ?? that h is
cocartesian and since ph = id and that h is an equivalence. We consider the diagram
h
J[ X

id
J] Y

which can again be solved as J is an ∞-groupoid. It follows that h is marked. We then

observe that the RLP with respect to the map (Λ21 )] q(Λ2 )[ (∆2 )[ → (∆2 )] implies that a
1
composition of marked morphisms is marked. Since f is a composition of g and h, we find
that f is marked as needed.
We now prove that any map p : X → Y of marked simplicial sets with the properties (1)–
(3) of the statement satisfies the RLP with respect to marked left anodyne maps. The lifting
property with respect to (−)[ applied to inner horn inclusions is clear. Consider a lifting
 INTRODUCTION TO INFINITY-CATEGORIES 125

problem
(Λn0 )s[ X

(∆n )s[ Y

Since the special edge ∆{0,1} is marked in X and marked edges are p-cocartesian, we find
a lift by the definition of p-cocartesian edges. To see that p satisfies the RLP with respect
to the map (Λ21 )] q(Λ2 )[ (∆2 )[ → (∆2 )] we have to show that a composite of marked edges
1
is again marked. This follows from the assumptions and the fact that a composite of p-
cocartesian edges is again p-cocartesian Lemma 10.5. Finally we need to argue that p has the
RLP wrt K [ → K ] for any ∞-groupoid K. This follows from the fact that equivalences are
p-cocartesian and thus marked by assumption on p. 

Corollary 11.6. A map of marked simplicial sets p : (X, S) → Y ] has the right lifting
property with respect to marked left anodyne maps if and only if S equals all p-cocartesian
edges and p is a cocartesian fibration.

We find the following consequence which will be of use to us later.

Corollary 11.7. The map (Λ20 )] q(Λ2 )[ (∆2 )[ → (∆2 )] is marked anodyne.
0

Proof. We observe that the left hand side is the simplicial set ∆2 in which the edges ∆{0,1}
and ∆{0,2} are marked. We need to show that this map satisfies the LLP wrt maps p : X → Y
satisfying the properties (1)–(3) of Proposition 11.5. So we consider a diagram

∆{1,2} (Λ20 )] q(Λ2 )[ (∆2 )[ X

0

(∆2 )] Y

and need to show that in the top horizontal composite is marked. Since the marked edges in
X are precisely the p-cocartesian edges this follows again from Lemma 10.5. 

Lecture 9 – 03.06.2019.

Lemma 11.8. The pushout product of a marked left anodyne map with any monomorphism
is again marked left anodyne.

Proof. We refer to [Lur09, Prop. [Link].] for a full proof of this result. It is in spirit very
similar to the arguments we have given when showing that a left/right/inner anodyne map
pushout product with a monomorphism is again left/right/inner anodyne: The first thing to
observe are the following
(a) The set of monomorphisms such that the conclusion holds is saturated,
(b) The set of marked anodyne maps for which the conclusion holds is saturated,
(c) The monomorphisms in marked simplicial sets are generated by the maps (∂∆n )[ →
(∆n )[ and the map (∆1 )[ → (∆1 )] .
126 M. LAND

We will prove (c): It is clear that the boundary inclusions as described generate monomor-
phisms of the form K [ → L[ . For every marked simplicial set (K, S) and monomorphism of
simplicial sets K → L, the following is a pushout of marked simplicial sets

K[ (K, S)

L[ (L, S)
so that the right vertical map is generated by the boundary inclusions. To show that a
general monomorphism of marked simplicial set is generated by the above it suffices to see
that (L, S) → (L, S 0 ) is generated by the above for S ⊆ S 0 . For this we observe that there is
a pushout
(∆1 )[
`
(L, S)
s0 ∈S 0 \S

(∆1 )] (L, S 0 )
`
s0 ∈S 0 \S

so that the claim is proven.

It hence now suffices to show that the pushout product of a map of the kind (1)–(4) of
Definition 11.4 with a map of the kind appearing in (c) above is marked anodyne. There are
thus eight cases to consider.
(1) We consider the pushout product of (Λni )[ ⊆ (∆n )[ with (∂∆n )[ ⊆ (∆n )[ . Since (−)[
preserves colimits the pushout product map is given by applying (−)[ to the pushout
product of the underlying maps of simplicial sets. This is again inner anodyne, so
that it becomes marked anodyne upon applying (−)[ .
(2) The pushout product of (Λni )[ ⊆ (∆n )[ with (∆1 )[ → (∆1 )] is an isomorphism and thus
marked anodyne: This uses that n ≥ 2 so that the map Λni → ∆n is an isomorphism
on vertices.
(3) The pushout product of K [ → K ] with (∂∆n )[ → (∆n )[ is an isomorphism if n > 0
and equals the map K [ → K ] for n = 0; in either case it is marked anodyne.
(4) The pushout product of K [ → K ] with (∆1 )[ → (∆1 )] is the map (K × ∆1 , S) →
(K × ∆1 )] where S is given by the pairs (a, b) where either a or b is degenerate. Since
every 1-simplex (a, b) in (K × ∆1 ) is a composite of (a, id) and (id, b) and identities
are degenerate, the map we are interested in is marked anodyne because adding a
composite of marked edges is marked anodyne.
(5) The pushout product of (Λ21 )] q(Λ2 )[ (∆2 )[ → (∆2 )] with (∂∆n )[ → (∆n )[ is an
1
isomorphism for n ≥ 1 and the given map for n = 0 and hence is marked anodyne in
any case.
(6) The pushout product of (Λ21 )] q(Λ2 )[ (∆2 )[ → (∆2 )] with (∆1 )[ → (∆1 )] is almost
1
an isomorphism: the only edge which is not marked in the domain (in the target all
edges are marked) is the edge (0 → 2, 0 → 1). It is however a composite of marked
edges, so that the needed map is again marked anodyne.
The remaining cases are easiest to to argue when using the alternative generating set of
marked left anodyne maps, i.e. working with the set (30 ) instead of (3). The key point is that
 INTRODUCTION TO INFINITY-CATEGORIES 127

the saturated set generated by the maps

(∆1 )] × (∂∆n )[ ∪ {0} × (∆n )[ → (∆1 )] × (∆n )[
is the same as that for the maps
(∆1 )] × A[ ∪ (∆1 )[ × B [ → (∆1 )] × B [
for monomorphisms A → B.
(7) The pushout product of (∆1 )] × (∂∆n )[ ∪ {0} × (∆n )[ → (∆1 )] × (∆n )[ with (∂∆n )[ →
(∆n )[ is of the latter kind: This follows from the associativity of pushout products:
It is given by the pushout product of {0} → (∆1 )] with the pushout product of
(∂∆n )[ → (∆n )[ with itself, which is clearly of the form A[ → B [ for a monomorphism
A → B.
(8) The pushout product of (∆1 )] × (∂∆n )[ ∪ {0} × (∆n )[ → (∆1 )] × (∆n )[ with (∆1 )[ →
(∆1 )] is, as before, an isomorphism if n > 0. For n = 0, the first pushout product
is simply {0} → (∆1 )] , so we obtain the map (∆1 × ∆1 , S) → (∆1 × ∆1 )] where S
consists of the degenerate edges and the edges {0} × ∆1 , and ∆1 × {ε} for ε = 0, 1.
Using property (3) and Corollary 11.7 this is a composition of marked left anodyne
maps.

We can then deduce the following result.
Proposition 11.9. Let p : E → C be a cocartesian fibration and let K be a simplicial set.
Then p∗ : EK → CK is again a cocartesian fibration, and an edge is p∗ -cocartesian if and only
if its image in E under the restriction along any object of K is p-cocartesian.
Proof. As a special case of Lemma 11.8, we find that for any marked left anodyne map A → B,
the map A × K [ → B × K [ is also marked left anodyne. Using Proposition 11.5 one can solve
any lifting problem
A × K[ E\

B × K[ C]
Since (−)[ is left adjoint to the forgetful functor, this means that (EK , S) → (CK )] has
the right lifting property wrt marked anodyne maps, where S consists of those edges whose
restriction to any object of K become p-cocartesian. By Corollary 11.6, S constists precisely
of the p∗ cocartesian edges and p∗ is a cocartesian fibration. 
Definition 11.10. Let p : E → C be a cocartesian fibration and K a marked simplicial
set. We denote by Funmcc (K, E) the full subcategory of Fun(K, E) on functors which send all
morphisms of K to p-cocartesian morphisms in E. If K is equipped with a map f : K → C] ,
f (K, E) the pullback
we denote by Funmcc

f (K, E)
Funmcc Funmcc (K, E)
p∗
f
∆0 Fun(K, C)
128 M. LAND

If K is an ordinary simplicial set, we will write Funcc (K, E), respectively Funcc
f (K, E) for
Fun (K , E), respectively Funf (K , E).
mcc ] mcc ]

Remark. In [Lur09] what we denote by Funmcc (K, E) is written as Map[ (K, E\ ). Likewise,
what we denote by Funmccf (K, E) is denoted by MapK (K, E ) in loc. cit.
[ \

The reason for this notation is the following: The category of marked simplicial sets is
cartesian closed: For every marked simplicial set K, the functor K × − admits a right adjoint,
denoted by X 7→ X K . We define a simplicial set Map[ (K, X) as follows. Its n-simplices are
given by HomsSet+ ((∆n )[ × K, X). Likewise, we define a simplicial set Map] (K, X) whose
n-simplices are given by HomsSet+ ((∆n )] × K, X). With our previous notation we have that
u(X K ) = Map[ (K, X) and m(X K ) = Map] (K, X).

We then find the following.

Proposition 11.11. Let p : E → C be a cocartesian fibration, let i : K → L be a marked

anodyne map and f : L → C] a morphism. Then the induced map

f (L, E) −→ Funf i (K, E)

Funmcc mcc

is a trivial fibration.

Proof. We need to show that any lifting problem

S f (L, E)
Funmcc

T f i (K, E)
Funmcc

can be solved if S → T is a monomorphism of simplicial sets. Unravelling definitions, this is

the case if and only if the lifting problem

S [ × L qS [ ×K T [ × K E\

T[ × L C]

can be solved. By Lemma 11.8, the left vertical map is marked anodyne, so the claim follows.


There is the following important lemma for us. We learned it from Hoang Kim Nguyen’s
thesis, [Ngu18, Lemma 3.2.3].

Lemma 11.12. Let K → L be a left anodyne map. Then the map K ] → L] of marked
simplicial sets is marked anodyne.

Proof. First we claim that the set of monomorphisms K → L of simplicial sets such that
K ] → L] is marked anodyne is saturated. It hence suffices to show that for 0 ≤ i < n, the
map (Λn0 )] → (∆n )] is marked anodyne. We observe that sk1 (Λni ) = sk1 (∆n ) once n is at
 INTRODUCTION TO INFINITY-CATEGORIES 129

least 3. Thus for n ≥ 3 and 0 < i < n we have a pushout

(Λni )[ (Λni )]

(∆n )[ (∆n )]
which shows that the right vertical map is marked anodyne. Likewise, there is a pushout

(Λn0 )s[ (Λn0 )]

(∆n )s[ (∆n )]

so that the right vertical map is again marked anodyne. It remains to treat the cases n < 3.
The case n = 1 is clear, so it remains to treat the case n = 2, in which we need to discuss the
cases i = 0 and i = 1. There are pushouts

(Λ20 )[ (Λ20 )s[ (Λ20 )]

(∆2 )[ (∆2 )s[ (∆2 )[ q(Λ2 )[ (Λ20 )]

0

so that the very right vertical map is marked anodyne. By Corollary 11.7 the further map
(∆2 )[ q(Λ2 )[ (Λ20 )] → (∆2 )]
0

is also marked anodyne, so that the map (Λ20 )] → (∆2 )] is also marked anodyne. For the
remaining case, we have the pushout

(Λ21 )[ (Λ21 )]

(∆2 )[ (Λ21 )] q(Λ2 )[ (∆2 )[

1

so that the right vertical map is marked anodyne. By definition, also the map
(Λ21 )] q(Λ2 )[ (∆2 )[ → (∆2 )]
1

is marked anodyne so the lemma is proven. 

With this at hand we have the following immediate consequence which will be very impor-
tant for us later.
Corollary 11.13. Let p : E → C be a cocartesian fibration, i : K → L a left anodyne map of
simplicial sets and f : L → C a morphism. Then the induced map
f (L, E) −→ Funf i (K, E)
Funcc cc

is a trivial fibration.
Proof. This is the special case of Proposition 11.11 where the marked left anodyne map is
i] : K ] → L] , using Lemma 11.12. 
130 M. LAND

If f : ∆1 → C is a morphism, then the ∞-category Funcc f (∆ , E) parametrizes all p-

1

cocartesian lifts of a given morphism in C. We thus find the following.

Corollary 11.14. Let p : E → C be a cocartesian fibration and f : ∆1 → C a morphism from
x to y in C. Then the map Funcc f (∆ , E) → Ex given by evaluating at {0} is a trivial fibration.
1

In particular, the simplicial set

f (∆ , E) ×Ex {z}
Funcc 1

is a contractible Kan complex for every object z of Ex .

Proof. We consider the pullback diagram

f (∆ , E) ×Ex {z}
Funcc f (∆ , E)
Funcc
1 1

z
∆0 x (∆ , E) ' Ex
Funcc 0

where the right vertical map is a trivial fibration because it is obtained by restriction along
{0} → ∆1 which is left anodyne so that we may apply Corollary 11.13. Thus also the left
vertical map is a trivial fibration. 
11.1. Digression – marked simplicial sets and localizations. In this section, we want
to indicate how one can make use of marked simplicial sets to study Dwyer–Kan localizations.
We start out with the following definition.
In what follows we will always view ∞-categories as a cocartesian fibration over ∆0 , so
that C\ denotes the marked simplicial set C with all equivalences marked.
Definition 11.15. Let X, Y be marked simplicial sets. A morphism f : X → Y is called a
marked equivalence if for any ∞-category, the induced map
Funmcc (Y, C) → Funmcc (X, C)
is an equivalence of ∞-categories.

Example 11.16. Let C be an ∞-category and S a set of morphisms containing all equiva-
lences. Let C[S −1 ] be a localization. Then the map (C, S) → C[S −1 ]\ is a marked equivalence.

Lecture 10 – 05.06.2019.
Example 11.17. Let A → B be a marked left anodyne map. Then A → B is a marked
equivalence. This is a special case of Proposition 11.11: We need to show that for every
∞-category E, the restriction functor
Funmcc (B, E) → Funmcc (A, E)
is a Joyal equivalence. In fact it is a trivial fibration, because Funmcc (B, E) = Funmcc
∗ (B, E)
for the cocartesian fibration E → ∆0 and the canonical map ∗ : B → ∆0 , likewise for A in
place of B.

As in Section 3, we expect to be able to factor any map as a marked anodyne map followed
by a map which satisfies the RLP wrt marked anodyne maps. For maps of the form X → ∆0 ,
we find that the resulting map C → ∆0 is an inner fibration, and the marked edges of C are
precisely the equivalences.
 INTRODUCTION TO INFINITY-CATEGORIES 131

Theorem 11.18. There exists a model structure on marked simplicial sets whose cofibrations
are monomorphisms, whose equivalences are marked equivalences, and where fibrant objects
are precisely ∞-categories with all equivalences marked.
Corollary 11.19. A Dwyer–Kan localization of C along S may thus be thought of as a
fibrant replacement of (C, S) in this model structure on marked simplicial sets.
Lemma 11.20. Let (C, S) be a marked ∞-category and let E be an ∞-category. Then there
is an equivalence Funmcc ((C, S), E) ' Fun(C[S −1 ], E) of ∞-categories.
Proof. This follows immediately from the definitions of localizations. 
More generally, we have:
Lemma 11.21. Let (C, S) and (D, T ) be ∞-categories equipped with sets of maps, viewed
as marked simplicial sets. A map f : (C, S) → (D, T ) of marked simplicial sets is a marked
equivalence if and only if the induced map
'
f¯: C[S −1 ] → D[T −1 ]
on Dwyer–Kan localizations is an equivalence.
Proof. Let E be an auxiliary ∞-category. Consider the commutative diagram
Funmcc ((D, T ), E) Funmcc ((C, S), E)

Fun(D[T −1 ], E) Fun(C[S −1 ], E)
in which the vertical maps are equivalences by Lemma 11.20. Hence the upper horizontal
map is an equivalence if and only if the lower one is. Since E is an arbitrary ∞-category, the
lemma follows. 
We wish to use this to obtain concrete examples of Dwyer–Kan localizations. The following
proposition will take care of this, see [Lur17, [Link]]. Its proof uses some techniques which we
develop at most later, and is beyond the scope of these notes. The statement and conclusion,
however, are easy to understand so we want to explain it here. We begin with a definition:
Definition 11.22. Let C be a simplicial category and x an object of C. An interval object
for x consists of the following data:
(1) an object Ix of C, equipped with
(2) a map h : ∆1 → HomC (x, Ix),
satisfying the following universal property: For every object y of C, composition with h
determines an isomorphism of simplicial sets
HomC (Ix, y) → Hom(∆1 , HomC (x, y)).

More concretely the map is given as follows. We consider the adjoint of the composition
map
HomC (Ix, y) → Hom(HomC (x, Ix), HomC (x, y))
and compose this wit the canonical restriction along h map
Hom(HomC (x, Ix), HomC (x, y)) → Hom(∆1 , HomC (x, y)).
132 M. LAND

Observation 11.23. An interval object Ix for x comes equipped with a canonical map
Ix → x: We have to exhibit a 0-simplex of HomC (Ix, x) which by assumption is isomorphic
to Hom(∆1 , HomC (x, x)) where we can take the map which is constant at the identity of x.

Proposition 11.24. Let C be a sinplicially enriched category and S a set of morphisms in

C. Let uC be the underlying ordinary category of C. Assume that the following conditions are
satisfied:
(1) Isomorphisms belong to S,
(2) S satisfies the 3-for-2 property
(3) For every object x of C, there exists an interval object Ix for x,
(4) the canonical maps Ix → x belong to S for all x of C.
Let C0 be a Kan enriched category, equipped with a weak equivalence C → C0 . We denote by
S 0 be image of S under this functor. Then the canonical map (N(uC), S) → (N(C0 ), S 0 ) is a
marked equivalence.
Corollary 11.25. Suppose that every morphism in S is sent to an equivalence in N(C0 ).
Then N(uC)[S −1 ] is equivalent to N(C0 ).
Corollary 11.26. The canonical functor sSet[we−1 ] → An is an equivalence of ∞-categories.
Proof. We consider the simplicial category sSet. It has interval objects given by ∆1 × X for
every X: We need to specify a map ∆1 → Hom(X, ∆1 × X) and we choose the identity. We
observe that sSet satisfy the assumptions of Proposition 11.24 where S is given by the set of
weak equivalences: We only need to recall that the map ∆1 × X → X is a weak equivalence
because ∆1 is contractible and geometric realization commutes with products. The small
object argument provides a functor sSet → Kan which is a weak equivalence: It is weakly
essentially surjective because every simplicial set is weakly equivalent to a Kan complex.
To see weakly fully faithfulness, we consider two simplicial sets A and B, and denote their
associated Kan complexes by XA and XB . Recall that there are anodyne maps A → XA and
B → XB . The map
Hom(XA , XB ) → Hom(A, XB )
is a trivial fibration and hence an equivalence. Furthermore, the composite
Hom(A, B) → Hom(XA , XB ) → Hom(A, XB )
is given by postcomposition with the map B → XB which is anodyne and hence a weak
equivalence. It hence suffices to show that postcomposition with a weak equivalence is itself a
weak equivalence. This follows from Exercise 107. We therefore apply Corollary 11.25: C0 can
be chosen to be the Kan enriched category Kan, S is the set of weak equivalences. Under the
functor C → N(C0 ) weak equivalences are sent to equivalences. The claim thus follows. 
Corollary 11.27. The functor sSet[Joy −1 ] → Cat∞ is an equivalence of ∞-categories.
Proof. We consider first the category sSet+ with simplicial enrichment given by Map] (X, Y ).
The fibrant replacement functor gives a simplicial functor sSet+ → Cat1∞ : We have to ar-
gue that for ∞-categories C and D, there is a canonical map Map] (C\ , D\ ) → Fun(C, D)'
which is a weak equivalence. In fact, we claim that these simplicial sets are isomorphic: In
both cases they are a subcategory of Fun(C, D), namely on those n-simplices C × ∆n → D,
all of whose edges are pointwise marked in D\ , respectively equivalences in D – this uses
again the pointwise criterion for natural equivalences we have developed in Theorem 6.1. In
 INTRODUCTION TO INFINITY-CATEGORIES 133

particular, we find that C0 can be chosen to be Cat1∞ with its canonical Kan enrichment as
discussed previously. We claim that the functor sSet+ → Cat1∞ sends marked equivalences
to Joyal equivalences: This is because marked (left) anodyne maps are marked equivalences
and marked equivalences satisfy the 3-for-2 property:
X X̄

Y Ȳ
where the horizontal maps are marked left anodyne, and the map X → Y is a marked
equivalence by assumption. We then let S be the set of marked equivalences, which satisfies (1)
and (2) of Proposition 11.24. We define an interval object by (∆1 )] × X, it is straightforward
to see that this is in fact an interval object according to ??. We claim that the canonical map
(∆1 )] × X → X is a marked equivalence: By exponentiating this follows from the fact that
(∆1 )] → ∆0 is a marked equivalence, thanks to Lemma 11.21.
We may thus apply Proposition 11.24 and Corollary 11.25 and obtain that the functor
N(u(sSet+ ))[me−1 ] → Cat∞ is an equivalence of ∞-categories. It remains then to show that
N(sSet+ )[me−1 ] ' N(sSet)[Joy −1 ].
This follows again because the latter can be identified with the localization of the subcategory
of fibrant objects using an argument similar to what we have seen earlier in Lemma 8.9. 

12. Straightening-Unstraightening
Lecture 11 – 12.06.2019. In this section we want to formulate and discuss in parts an
important correspondence: The Grothendieck construction. We begin with an informal con-
struction. Consider a cocartesian fibration p : E → C. We observe that we can extract the
following data from this:
(1) for each object x of C, we have the ∞-category Ex ,
(2) for each morphism f : x → y in C, and an object z in Ex , we can choose a p-cocartesian
lift z → w of f , we will denote w = f! (z)
(3) given a further object z 0 in Ex and a morphism α : z → z 0 , we can choose another
p-cocartesian lift z 0 → w0 = f! (z 0 ) and obtain a diagram
z z0

f! (z) f! (z 0 )
and since z → f! (z) is p-cocartesian, the space of dashed arrows making the diagram
commutative is contractible. We will denote any such dashed arrow by f! (α)
Summarizing, associated to a cocartesian fibration p : E → C, we wish to find a functor
C → Cat∞ , sending an object to the fibre of p, and sending a morphism f : x → y, the
“functor” f! .
Proposition 12.1. Given a cocartesian fibration p : E → C and a morphism f : ∆1 → C
from x to y, there exists a functor Ex × ∆1 → E whose restriction to every object z of Ex
provides a p-cocartesian morphism α : z → z 0 over f . Restricting this functor to Ex × {1}
gives a functor f! : Ex → Ey .
134 M. LAND

Proof. We begin by constructing for each cocartesian fibration p : E → C and each morphism
f : ∆1 → C from x to y a functor f! : Ex → Ey .
We recall that in this situation, the canonical map Funccf (∆ , E) → Ex given by taking
1

the source of a morphism is a trivial fibration. Choosing a section of this trivial fibration
produces the composite
f! : Ex → Funcc
f (∆ , E) → Ey
1

where the latter map is given by taking the target of a morphism. Furthermore, we find that
the first map is adjoint to a map
E x × ∆1 → E
with the following properties: Its restriction to Ex × {1} is given by f! and it makes the
diagram
Ex × ∆ 1 E
p
f
∆1 C
commute and furthermore, for each object z in Ex , the resulting morphism ∆1 → E is a
p-cocartesian morphism with source equal to z and target equal to f! (z). 
We wish to show that the association f 7→ f! is “functorial in f ”. For this we consider
a 2-simplex σ : ∆2 → C inside C, which exhibits h as a composition of f and g. We then
consider the diagram

f!

Funcc {0,1} , E) Ey
f (∆

' '
'
Ex ' σ (∆ , E)
Funcc 2 Funcc
g (∆
{1,2} , E)g!

' '

Funcc {0,2} , E) Ez
h (∆

h!

The maps labelled with a ' are trivial fibrations because they arise as restrictions along left
anodyne maps. This shows that there is a natural isomorphism between g! f! and h! .
We now explain the general construction. For this, we fix a cocartesian fibration p : E → C.
Lemma 12.2. Associating to σ : ∆n → C the ∞-category Funcc
σ (∆ , E) extends to a functor
n
op
Θ(p) : ∆/C → sSet.
 INTRODUCTION TO INFINITY-CATEGORIES 135

Proof. We need to show that a commutative diagram

f
∆n ∆m
σ τ

C
τ (∆ , E) → Funσ (∆ , E) which is clear from the
induces a well-defined a functorial map Funcc m cc n

definition. 
Definition 12.3. Let X be a simplicial set. We denote by WX the set of all morphisms
f : [n] → [m] in ∆/X such that f (0) = 0.

Lemma 12.4. For a cocartesian fibration p : E → C, the functor Θ(p) sends any morphism
in WC to a Joyal equivalence.
f σ
Proof. Consider a morphism f in ∆/C , represented by the composite ∆n → ∆m → C. We will
write τ = σf . By assumption, the composite
f
∆{0} → ∆n → ∆m
picks out the object 0 in ∆m . We thus have a commutative diagram
f∗
σ (∆ , E)
Funcc τ (∆ , E)
Funcc
m n

' '

{0} (∆ , E)
Funcc 0

in which both diagonal maps are trivial fibrations by Corollary 11.13 because for any k ≥ 0,
the inclusion ∆{0} → ∆k is left anodyne. Hence also the map f ∗ is a Joyal equivalence as
claimed. 
Corollary 12.5. For every cocartesian fibration p : E → C, we obtain a functor
Θ(p) : N(∆op −1
/C )[WC ] −→ Cat∞ .

Proof. By the previous lemma, Θ(p) induces a functor between the localizations
N(∆op −1 −1
/C )[WC ] −→ sSet[Joy ]

and the latter admits a further functor to Cat∞ (which is in fact an equivalence by Corol-
lary 11.27). 
Lemma 12.6. Let X be a simplicial set. There is a canonical map of simplicial sets
N(∆op
/X ) → X called the initial vertex map.

Proof. Recall that a k-simplex of the nerve is given by a sequence

α α α α
[n0 ] →1 [n1 ] →2 [n2 ] →0 · · · →k [nk ]
together with a map σ : ∆nk → X. We observe that the association α : [k] → [nk ] given by
sending 0 to 0 and i to αk ◦ · · · ◦ αk−i+1 (0) is a map of linearly ordered sets. We hence obtain
α σ
a k-simplex of X by the composite ∆k → ∆nk → X. It is straightforward to check that this
is compatible with the simplicial structure maps. 
136 M. LAND

Lemma 12.7. The initial vertex map sends all morphisms in WX to degenerate edges of X.
f σ
Proof. Recall that a 1-simplex in WX is represented by the composite [n] → [m] → X where
the map f sends 0 to 0. The resulting 1-simplex of X is given by restricting the map σ along
the map [1] → [m] given by sending 0 to 0 and 1 to f (0) = 0. This is a degenerate edge in
[m] and thus remains degenerate after applying the map σ. 
We will now need the following result, which is due to Joyal, Dwyer-Kan, and has also been
proved by Stevenson [Ste17, Theorem 1.3].
Theorem 12.8. For every ∞-category C, the initial vertex map induces a Joyal equivalence
'
N(∆op −1
/C )[WC ] −→ C.

Proof. The proof will consist of two steps: First, one shows that one can reduce the claim to
showing it only for C = ∆n , and then one has to show the claim in this case.
As a first step, we need a slightly more general version of the above: We want to show
that these maps make sense for an arbitrary simplicial set X in place of C: Clearly, the initial
vertex map defines a map ∆op /X → X. Now, for the moment, let us define for a marked
simplicial set (X, S) a simplicial set L(X, S) by the pushout
` 1
∆ X
f ∈S

`
J L(X, S)
f ∈S

and observe that if X = C is an ∞-category, then L(X, S) is Joyal equivalent to C[S −1 ], which
was defined by choosing an inner anodyne map L(C, S) → C[S −1 ] to obtain an ∞-category.
The initial vertex map takes a morphism in WX to a degenerate edge in X, so one can clearly
extend the corresponding map ∆1 → X over the inclusion ∆1 → J. In particular, we obtain
an induced map L(N(∆op /X ), WX ) → X and we claim that this is a Joyal equivalence for every
simplicial set X. Once this is shown, so is the theorem, by the above observation.
We denote the functor from simplicial sets to marked simplicial sets, sending X to (N(∆op
/X ), WX )
by F . We will use the following properties, whose verification we leave as an exercise:
(1) The functor LF preserves colimits,
(2) the functor LF preserves monomorphisms,
(3) the initial vertex maps assemble into a natural transformation LF ⇒ id.
Let us now suppose that the theorem is shown for C = ∆n and let X be an arbitrary
simplicial set. We write X as the colimit over its skeleta skn (X) and obtain the map
L(N(∆op ∼ op ∼
/X ), WX ) = colim L(N(∆/skn (X) ), Wskn (X) ) → colim skn (X) = X
n n

where we use the initial vertex map at each step. The fact that the needed diagrams commute
is an exercise. If we can show that each initial vertex map
L(N(∆op
/skn (X) ), Wskn (X) ) → skn (X)

is a Joyal equivalence, then so is the above map by yet another exercise.

 INTRODUCTION TO INFINITY-CATEGORIES 137

We then perform an induction over the dimension n. The induction start forces X to be
a disjoint union of ∆0 ’s, and since the initial vertex map commutes with disjoint union, this
map is a Joyal equivalence by assumption.
For the induction step, we consider the pushout
∂∆n
`
skn−1 (X)
Jn

∆n
`
skn (X)
Jn

and recall that the initial vertex map commutes with colimits. By induction and assumption,
the initial vertex map is a Joyal equivalence on the corners except a priori the lower right
corner. However, since the functor L(N(∆op/(−) ), W(−) ) preserves colimits and monomorphisms,
it follows from Exercise 101 that this map is also a Joyal equivalence.

Lecture 12 – 24.06.2019. We hence now need to show the statement of the theorem for
∆n . We observe that in this case, the initial vertex map
N(∆op
/∆n ) → ∆
n

is the map induced on nerves of the functor ∆op /∆n → [n] sending f : [m] → [n] to f (0): Its
effect on morphisms is given by the following: Suppose given a composite [k] → [m] → [n]
where the composite is g and the latter map is f . We then need to find a morphism in
[n] from f (0) to g(0). In other words we need to show that f (0) ≤ g(0). But we have
g(0) = f (h(0)) ≥ f (0) because f is monotone increasing and h is also monotone increasing.
We now construct a functor in the other direction: [n] → ∆op
/∆n given as follows: The object
i of [n] is sent to the map ∆{i,...,n} → ∆n . Clearly, if i ≤ j, there is a commutative diagram

∆{i,...,n} ∆n

∆{j,...,n}
so this gives a functor as desired. Its composition with the initial vertex map is given by the
identity, as one checks immediately. We now consider the composite
IV i
∆op op
/∆n → [n] → ∆/∆n .

We claim that there is a canonical natural transformation from this composite i ◦ IV to the
identity functor. Indeed, the composite is given by sending f : [m] → [n] to the canonical
inclusion {f (0), . . . , n} → [n]. The canonical commutative triangle
f
{0, . . . , m} {0, . . . , n}
f

{f (0), . . . , n}

gives the components of this natural transformation (the left vertical map). We observe that
these components are all contained in the set W∆n .
138 M. LAND

This construction provides a map of simplicial sets

∆1 → Fun(N(∆op op
/∆n ), N(∆/∆n ))

restricting to i ◦ IV on 0 and to the identity on 1. Postcomposing with the localization map

N(∆op op −1
/∆n ) → N(∆/∆n )[W∆n ], and recalling that i ◦ IV sends W∆ to equivalences, we obtain
n

a map
∆1 → Fun(N(∆op −1 op −1
/∆n )[W∆n ], N(∆/∆n )[W∆n ]).
We claim that this is a natural equivalence, which follows from the fact that the components
of the transformation above are contained in the set W∆n . We have thus constructed functors
N([n]) → N(∆op −1 op −1
/∆n )[W∆n ] → N([n]) → N(∆/∆n )[W∆n ]

such that both composites are naturally equivalent to the identity functor. Hence, the initial
vertex map is a Joyal equivalence as desired. 
Corollary 12.9. Let p : E → C be a cocartesian fibration. Inverting the above equivalence,
we obtain a functor
'
C ←− N(∆op −1
/C )[WC ] −→ Cat∞ .
We call this the straightening of the cocartesian fibration p.
We end this section with the straightening-unstraightening equivalence of Lurie. Informally,
it says that the straightening construction of Corollary 12.9 induces an equivalence of suitable
∞-categories. To state it precisely, let us denote by CoCart(C) the subcategory of the slice
(Cat∞ )/C on objects which are cocartesian fibrations E → C and whose morphisms are the
morphisms of cocartesian fibrations according to Definition 10.19.
Theorem 12.10. For every ∞-category C, there is an equivalence of ∞-categories
CoCart(C) ' Fun(C, Cat∞ ).
On objects, this equivalence implements our previous construction.
Definition 12.11. Consider the ∞-category C = Cat∞ and the identity functor. By
Theorem 12.10, this corresponds to a cocartesian fibration over Cat∞ , called the universal
cocartesian fibration. This is a functor (Cat∞ )∗ → Cat∞ , and (Cat∞ )∗ is an ∞-category
whose objects are pairs (C, x) where x is an object of C, and morphisms from (C, x) to (D, y)
consist of pairs (F, α) where F : C → D and α : y → F x is a morphism in D.

Remark. Constructing the ∞-category (Cat∞ )∗ is not easy: It involves the composition in
an ∞-category which is not strict. There are ways to work around this, but we will not get into
the details here, see [RV18, Remark 6.1.19]. The idea is to consider the coherent nerve of the
simplicial category of ∞-categories N(Cat1∞ ) without passing to the groupoids of the functor
categories. This is a simplicial set, and one can form its slice under the point. This gives
a map of simplicial sets N(Cat1∞ )∆0 → N(Cat1∞ ). By construction, there is also a functor
Cat∞ → N(Cat1∞ ), and the pullback of the slice projection turns out to be a cocartesian
fibration.

Remark. In general, one would like to have for each cocartesian fibration E → D, and each
∞-category C a functor Fun(C, D) → CoCart(C) given on objects by pulling back the given
cocartesian fibration. The statement that E → D is universal then translates to the property
 INTRODUCTION TO INFINITY-CATEGORIES 139

that this functor is an equivalence of ∞-categories. Such a construction is done in [RV18,

Theorem 6.1.13].

Remark. By means of the universal cocartesian fibration we can also say what the the
equivalence
CoCart(C) ' Fun(C, Cat∞ )
does to a functor F : C → Cat∞ . It sends it to the pulled pack cocartesian fibration
F ∗ p : F ∗ (Cat∞ )∗ → C.

We will now need the following lemma.

Lemma 12.12. Let f : C → D be a fully faithful functor between ∞-categories. Then for
any simplicial set K, the induced functor Fun(K, C) → Fun(K, D) is again fully faithful.
Proof. We first observe that a functor f is fully faithful if and only if the diagram

Fun(∆1 , C) Fun(∆1 , D)

C×C D×D
induces Joyal equivalences on all fibres over points of C × C. We can apply the functor
Fun(K, −) to this diagram. It preserves fibres, and Joyal equivalences so that also the resulting
diagram has the property that it induces Joyal equivalences on vertical fibres. Using the
equivalence Fun(K, Fun(∆1 , C)) ∼
= Fun(∆1 , Fun(K, C)) the lemma follows. 
Recall from Corollary 9.34 that the canonical functor An → Cat∞ is fully faithful. It
follows from Lemma 12.12 that for any ∞-category C, the functor
Fun(C, An) → Fun(C, Cat∞ )
is also fully faithful. In particular, under the above equivalence, the ∞-category Fun(C, An)
must correspond to some full subcategory of CoCart(C). This is given by the following:
Theorem 12.13. For every ∞-category C, the straightening-unstraightening equivalence
restricts to an equivalence
LFib(C) ' Fun(C, An)
where LFib(C) denotes the full subcategory of the slice (Cat∞ )/C on left fibrations.
Proof. Under the straightening-unstraightening equivalence, the functor C → An → Cat∞
corresponds to a cocartesian fibration E → C whose fibres are ∞-groupoids. Hence E → C is
a left fibration by the dual version of Proposition 10.17. Since any morphism E → E0 over C
preserves cocartesian edges, as every edge is cocartesian by Lemma 10.16, this is in fact the
full subcategory of the slice category as claimed. 
The following is almost a corollary of the above.
Theorem 12.14. Let X be an ∞-groupoid. Then the straightening-unstraightening equiva-
lence restricts to an equivalence
An/X ' Fun(X, An)
140 M. LAND

Proof. By Theorem 12.13, we need to show that there is a canonical equivalence LFib(X) '
An/X . Since X is an ∞-groupoid any left fibration E → X is in fact a Kan fibration. In
particular, E is itself a Kan complex. We thus find that the category LFib(X) is the full
subcategory of the slice (Cat∞ )/X whose objects consist of Kan fibrations. We obtain the
following diagram
LFib(X) An/X

(Cat∞ )/X
and claim that the functor An/X → (Cat∞ )/X is fully faithful because the functor An → Cat∞
is, see ??. It follows that the functor LFib(X) → An/X is also fully faithful Furthermore, any
map Y → X between ∞-groupoids is equivalent to a Kan fibration. This implies that the
inclusion LFib(X) → An/X is essentially surjective and fully faithful and thus an equivalence
as needed. 
To finish the proof, we will need the following analysis of the mapping anima in slice
∞-categories. This will become important again later. The dual version is [Lur09, Lemma
[Link]].
Proposition 12.15. Let C be an ∞-category and f : x → z and g : y → z be morphisms in
C, viewed as objects of C/z . Then the diagram

mapC/z (f, g) mapC (x, y)

g∗

f
∆0 mapC (x, z)
is homotopy cartesian.
Proof. Recall that the map g∗ : mapC (x, y) → mapC (y, z) is constructed as follows. We have
the two canonical restriction functors
'
C/y ← C/g → C/z
the first of which is an equivalence. Inverting this equivalence, and taking fibres over x in C,
we obtain
C/y ×C {x} ' C/g ×C {x} → C/z ×C {x}
where the first and last term are given by mapR R
C (x, y) and mapC (x, z) respectively. We then
consider the diagram
C/g ×C/z {f } C/g ×C {x} C/g

f
∆0 C/z ×C {x} C/z

in which both the right square and the big square are pullbacks. Hence all squares are
pullbacks. Furthermore, the very right vertical map is a right fibration, hence the middle
vertical map is a right fibration whose target is an ∞-groupoid. Hence the middle vertical
 INTRODUCTION TO INFINITY-CATEGORIES 141

map is a Kan fibration and models the map given by postcomposition with g. It hence remains
to show that there is a canonical equivalence
mapC/z (f, g) ' C/g ×C/z ×{f }.
For this we observe that (C/z )/g ∼
= C/g , so the claim follows. 
Lecture 13 – 26.06.2019.
Corollary 12.16. Let C ⊆ D be a full subcategory and let z be an object of C. Then the
canonical functor C/z → D/z is again fully faithful.
Proof. Let f : x → z and g : y → z be objects of C/z . We wish to show that the map
mapC/z (f, g) → mapD/z (f, g)
is a homotopy equivalence. By Proposition 12.15 it therefore suffices to prove that in the
diagram
mapC (x, y) mapD (x, y)

mapC (x, z) mapD (x, z)

both horizontal maps are equivalences. This follows from the assumption that f is fully
faithful. 

13. Terminal and initial objects

Definition 13.1. Let C be an ∞-category. An object x is said to be initial if for all objects
y of C, the mapping space mapC (x, y) is contractible. Likewise x is said to be terminal if it is
initial in Cop , i.e. if for all other objects y, the mapping space mapC (y, x) is contractible.

It will be useful to have the following characterizations:

Lemma 13.2. Let C be an ∞-category and x in C an object. Then the following conditions
are equivalent.
(1) x is terminal,
(2) the functor C/x → C is a trivial fibration, and
(3) for every n ≥ 1, every lifting problem
x

∆{n} ∂∆n C

∆n
admits a solution.
Proof. To show that (1) and (2) are equivalent, we consider the following diagram
C/x C

C
142 M. LAND

and we wish to show that the horizontal map is a trivial fibration. We already know that
it is a right fibration, so it suffices to show that it is a Joyal equivalence if and only if x
is terminal. By Theorem 10.21 this map is a Joyal equivalence if and only if it is a Joyal
equivalence fibrewise, which amounts to saying that for all objects y of C the canonical map
mapR 0
C (y, x) → ∆ is a Joyal equivalence.
To see that (2) and (3) are equivalent, we observe that the map
∂∆n−1 ? ∆0 q∂∆n−1 ?∅ ∆n−1 ? ∅ −→ ∆n−1 ? ∆0
is isomorphic to the map
∂∆n → ∆n .
Hence the lifting problem
x

∆{n} ∂∆n C

∆n
is equivalent to the lifting problem
∂∆n−1 C/x

∆n−1 C
so the lemma follows. 
The following tells us that initial and terminal objects, if they exist, are unique up to
contractible choices.
Proposition 13.3. Let C be an ∞-category and let Cterm be the full subcategory spanned by
all terminal objects. Then Cterm is either empty or a contractible Kan complex.
Proof. Suppose that Cterm is not empty. We need to show that any lifting problem
∂∆n Cterm

∆n
has a solution. If n = 0, this exists by the assumption that Cterm is not empty. If n ≥ 1, we
use (3) of Lemma 13.2 which is possible as in particular the object ∆{n} of ∂∆n is mapped
to a terminal object. 
Lemma 13.4. Let C be an ∞-category. Then an object x of C is initial if and only if the
map x : ∆0 → C is left anodyne. Dually x is terminal if the map ∆0 → C is right anodyne.
Proof. We prove the statement for initial objects. The case for terminal objects is obtained
by passing to opposite categories. We first observe that for any monomorphism S → T , the
map S ? ∆0 → T ? ∆0 is right anodyne, and the map ∆0 ? S → ∆0 ? T is left anodyne. To see
this it suffices to treat the case where S → T is a boundary inclusion ∂∆n → ∆n in which
case the maps in question become Λn+1 n+1 → ∆
n+1 and Λn+1 → ∆n+1 . Now let us assume that
0
 INTRODUCTION TO INFINITY-CATEGORIES 143

x is initial. By the version of Lemma 13.2 for initial objects, we find that Cx/ → C is a trivial
fibration. We choose a section s : C → Cx/ and consider the diagram

∆0 ∆0 ? ∆0 ∆0
x ∆0 ?x x
ŝ
C ∆0 ? C C

where ŝ is the adjoint map of s : C → Cx/ . It follows that both horizontal composites are
x
the identity. We thus find that the map ∆0 → C is a retract of the map ∆0 ? x which is left
anodyne by our first observation.
x
Conversely, assume that ∆0 → C is left anodyne. We can consider the diagram

idx
∆0 Cx/
x s

C C

and find a dashed arrow s making the diagram commute. We will show that x is initial by
establishing (3) of Lemma 13.2. We thus consider a diagram

s
∆{0} ∂∆n C Cx/ C

∆n

By construction s(x) = idx and is thus an initial object of Cx/ by Exercise 133. Hence the
dotted arrow exists and thus also a dashed arrow. 

The following is also useful to know.

Proposition 13.5. Let C be an ∞-category and let K be a simplicial set. Suppose given
a functor F : K → C such that for all objects x of K, the object F (x) is initial, respectively
terminal, in C. Then F is initial, respectively terminal, in Fun(K, C).

Proof. We show the case of terminal objects. We need to prove that for all n ≥ 1, any lifting
problem
F

F̂
∆{n} ∂∆n Fun(K, C)

∆n
144 M. LAND

admits a solution. By adjunction, this corresponds to the lifting problem

F

K × ∆{n} K × ∂∆n C

K × ∆n
where now by assumption the further restriction of the top horizontal composite along any
object x : ∆0 → K is a terminal object of C. We consider the filtration Fk (K) = skk (K ×
∆n ) ∪ K × ∂∆n and inductively wish to solve the extension problem
ai fk−1
∂∆k C
`
Fk−1 (K)
i∈I(k)

∆k
`
Fk (K)
i∈I(k)

For this it suffices to observe that for all i ∈ I(k) the composite fk−1 ◦ ai sends the vertex
{k} to a terminal object in C. By definition I(k) consists of those non-degenerate k-simplices
of K × ∆n which are not contained in K × ∂∆n , in other words consists in particular of pairs
(αi , βi ) ∈ Kk × ∆nk such that b : [k] → [n] is surjective. In particular, βi sends the object {k}
to {n}. Hence fk−1 (ai ({k})) = F̂ (αi ({k}, {n})) which is a terminal object by assumption.
Hence the dashed arrow exists. Passing to the colimit over k then gives the proposition. 
The converse of Proposition 13.5 is almost true.
Lemma 13.6. Suppose that Cterm is not empty. Then any terminal object of Fun(K, C)
takes values in Cterm .
Proof. Let x be a terminal object and consider the constant functor cx with value x. By
Proposition 13.5, cx is a terminal object of Fun(K, C). By Proposition 13.3, any other terminal
object of Fun(K, C) is equivalent to cx . In particular, any other terminal object T evaluates
on an object k to T (k) which is equivalent to cx (k) = x and is thus terminal. 

14. The Yoneda lemma

Lecture 14 – 01.07.2019. We now wish to show that for every ∞-category C, there is
a Yoneda functor C → Fun(Cop , An) which should send an object x of C to the “functor”
y 7→ mapC (y, x). Then we will establish that this functor is fully faithful, which is the ∞-
categorical version of the Yoneda lemma. The fact that mapC (−, x) is a functor for every
single x is something we now know.
Definition 14.1. Let x be an object of an ∞-category C. Then the functor C/x → C is a
right fibration and hence by Theorem 12.13 (applied to Cop ) is equivalently given by a functor
Cop → An sending y to C/x ×C {y} ' mapC (y, x). We shall denote this functor by mapC (−, x).

The task now is to make precise that the functors mapC (−, x) in turn are functorial in x.
We begin with a construction.
 INTRODUCTION TO INFINITY-CATEGORIES 145

Lemma 14.2. The association [n] 7→ [n] ? [n]op ∼

= [2n + 1] extends to a functor ∆ → ∆. In
n n op
particular, sending [n] to ∆ ? (∆ ) is a cosimplicial object in simplicial sets, i.e. a functor
∆ → sSet.
Definition 14.3. Let C be an ∞-category. We define its twisted arrow category Tw(C) to
be the following simplicial set
Tw(C)n = HomsSet (∆n ? (∆n )op , C)
where the simplicial structure comes from the cosimplicial object [n] 7→ ∆n ? (∆n )op . The
inclusions ∆n → ∆n ? (∆n )op ← (∆n )op determine a functor
Tw(C) −→ C × Cop .

The following proof is taken from [Lur17, Proposition [Link]].

Proposition 14.4. For an ∞-category C, the functor
Tw(C) −→ C × Cop
is a right fibration. In particular, Tw(C) is again an ∞-category.
Proof. Let 0 < k ≤ n, and consider a lifting problem
Λkn Tw(C)

∆n C × Cop
Unravelling the definition of the twisted arrow category, this corresponds to the following
lifting problem
K C

∆2n+1 ∆0
where K is the subsimplicial set of ∆2n+1 consisting of those faces σ satisfying either of the
following three properties:
(1) σ is contained in ∆{0,...,n} ⊆ ∆2n+1 ,
(2) σ is contained in ∆{n+1,...,2n+1} ⊆ ∆2n+1 ,
(3) There exists j 6= k, with 0 ≤ j ≤ n such that neither j nor 2n + 1 − j is a vertex of σ.
Since C is an ∞-category, it suffices to show that the inclusion K → ∆2n+1 is inner anodyne.
We call a simplex σ primary, if it is not contained in K and its vertices are contained in the
set {k, . . . , 2n + 1}. We call σ secondary if it is not contained in K and not primary. We let
S be the set containing the following simplices τ of ∆2n+1 :
(1) τ is primary and k is not a vertex of τ ,
(2) τ is secondary and 2n + 1 − k is not a vertex of τ .
Given a simplex τ in S, we let τ 0 be the simplex obtained by adding the vertex k if τ is
primary, and by adding 2n + 1 − k is τ is secondary. We observe that each simplex of ∆2n+1 is
either contained in K, in S, or is of the form τ 0 for a unique τ in S: If it is neither contained in
K or S, then it must be either primary and contain k as vertex, or be secondary and contain
2n + 1 − k as vertex. In either case, one can remove the vertex k or 2n + 1 − k and obtain a
primary and secondary simplex as needed.
146 M. LAND

We now choose an ordering {σ1 , σ2 , . . . , σm } of S with the following two properties:

(1) If p ≤ q, then dim(σp ) ≤ dim(σq ),
(2) If p ≤ q and dim(σp ) = dim(σq ) and σq is primary, then also σp is primary.
For 0 ≤ q ≤ m, we let Kq be the sub simplicial set of ∆2n+1 obtained from K by adding the
simplices σp and σp0 for 1 ≤ p ≤ q. Clearly Km = ∆2n+1 , so we obtain a filtration
K → K1 → K2 → · · · → Km = ∆2n+1
and it will suffices to show that for each q, the map Kq−1 → Kq is inner anodyne. Since Kq
is obtained from Kq−1 by adding the simplices σq and σq0 , and σq0 contains σq , it suffices to
show that there is a pushout
Λdj Kq−1

σq0
∆d Kq
where d is the dimension of σq0 and 0 < j < d. For this, we need to see which of the faces of
σq0 are already contained in Kq−1 , and argue that precisely one inner face is not contained in
Kq−1 . 
For every ∞-category, we denote by mapC (−, −) : Cop × C → An the functor associated
with the right fibration Tw(C) → C × Cop .
Remark. We could also go a different route here: If we only wanted to construct a mapping
functor mapC (−, −) : Cop × C → An, we can observe that by definition An = N(Kan), so such
a functor is equivalently given by a simplicial functor C[Cop × C] → Kan. We can consider the
composite
C[Cop × C] → C[C]op × C[C] → Kan
where the first is given by the canonical map and the second by the simplicial mapping space
followed by a functorial Kan-replacement. The fact that this is (at least pointwise) equivalent
to our approach is [Lur09, Theorem [Link]].

The following is then contained in the proof of [Lur17, Proposition [Link]].

Lemma 14.5. For an ∞-category C and every object x of C, there is a canonical commutative
diagram
C/x Tw(C)

C × {x} C × Cop
This diagram is homotopy cartesian. In other words, the induced map C/x → Tw(C)x is a
Joyal equivalence between right fibrations over C.
Proof. We recall that the n-simplices of Tw(C)x are given by those maps ∆n ? (∆n )op → C
whose restriction along the inclusion (∆n )op → ∆n ? (∆n )op are the map which is constant at
the object x of C. One defines an auxiliary simplicial set M, whose n-simplices are given by
maps ∆n ? ∆0 ? (∆n )op → C, whose restriction along ∅ ? ∆0 ? (∆n )op → ∆n ? ∆0 ? (∆n )op are
constant at x. The obvious inclusions define maps
C/x ← M → Tw(C)x
 INTRODUCTION TO INFINITY-CATEGORIES 147

and the left map admits a section induced from the map ∆0 ?(∆n )op → ∆0 . It hence suffices to
show that both of the above maps are Joyal equivalences. This is done by showing that both
are trivial fibrations, by explicitly exhibiting the lifting property with respect to boundary
inclusions. Let us treat the case M → C/x . We consider a lifting problem

∂∆n M

∆n C/x

and unravel definitions to see that this is equivalently given by a lifting problem
K C

∆n ? ∆0 ? ∆n ∆0
where K denotes the smallest sub simplicial set containing ∆n ? ∆0 ? ∅, ∅ ? ∆0 ? ∆n , and
∆I ? ∆0 ? (∆I )op , for every proper subset I ⊆ [n]. By Theorem 9.13, it suffices to show that
the inclusion K → ∆n ? ∆0 ? ∆n is a Joyal equivalence. First, we claim that the composite
∆ n ? ∆ 0 q∆ 0 ∆ 0 ? ∆ n → K → ∆ n ? ∆ 0 ? ∆ n
is a Joyal equivalence. To see this we consider the diagram

I n+1 q∆0 I n+1 I 2n+2

∆n+1 q∆0 ∆n+1 ∆2n+2

in which the upper horizontal map is an isomorphism and the lower horizontal map is the map
under investigation. It hence suffices to show that the vertical maps are Joyal equivalences.
For the right hand this follows since the spine is inclusion is inner anodyne by Proposition 3.17
and thus a Joyal equivalence by Corollary 6.13, and for the left vertical map we argue likewise
for the maps I n+1 → ∆n+1 and then use Corollary 9.7 to conclude that the map on pushouts
is also a Joyal equivalence. It now suffices to show that the map
∆n ? ∆0 q∆0 ∆0 ? ∆n → K
is a Joyal equivalence. We denote by K0 the sub simplicial set of K spanned by the simplices
of the form ∆I ? ∆0 ? (∆I )op for I ⊆ [n] a proper subset. The following diagram
j
∂∆n ? ∆0 q∆0 ∆0 ? ∂∆n K0

∆n ? ∆0 q∆0 ∆0 ? ∆n K
is a pushout, so it suffices to show that the map j is a Joyal equivalence. This map is a colimit
of maps of the form
∆I ? ∆0 q∆0 ∆0 ? (∆I )op → ∆I ? ∆0 (∆I )op
which we have just argued to be Joyal equivalences. We thus conclude that the map M → C/x
is a Joyal equivalence.
148 M. LAND

It remains to prove that the map M → Tw(C)x is also a Joyal equivalence. The proof is
very similar in spirit to the one of Proposition 14.4, so we will refrain from spelling out the
details and refer to [Lur17, Proposition [Link]] instead. 
Corollary 14.6. For every object x, the composite C × {x} → C × Cop → An is equivalent
to the functor mapC (−, x).
Proof. The first functor is the one associated to the right fibration Tw(C)x → C, whereas
the other one is associated to the right fibration C/x → C. By Lemma 14.5 and Proposi-
tion 9.25, these right fibrations are equivalent. Hence the claim follows from the straightening-
unstraightening equivalence Theorem 12.13. 
Definition 14.7. Let C be an ∞-category. The functor mapC (−, −) : Cop × C → An is
adjoint to a functor Y : C → Fun(Cop , An) which we call the Yoneda functor.

The following is the ∞-categorical version of the Yoneda lemma.

Proposition 14.8. Let F : C → An be a functor and let x be an object of C. Then the
canonical map
mapFun(C,An) (mapC (x, −), F ) → F (x)
given by evaluation at the identity is an equivalence.
Proof. We let p : E → C be the left fibration corresponding to the functor F . Under the
straightening-unstraightening equivalence Theorem 12.13
full
Fun(C, An) ' LFib(C) ⊆ Cat∞ /C ,
the left hand mapping anima corresponds to mapCat∞ /C (Cx/ , E) and we claim that there is a
canonical equivalence
mapCat∞ /C (Cx/ , E) ' Funq (Cx/ , E)
where q : Cx/ → C is the canonical forgetful functor and the latter is as in Definition 11.10
without the superscripts.
Taking this for granted for the moment, we need to show that the map
Funq (Cx/ , E) → Funid (∆0 , E) ∼
x = Ex
is an equivalence: Recall that by construction of the straightening-unstraightening equiva-
lence, there is a canonical equivalence Ex ' F (x). Here, the map is induced by the canonical
map ∆0 → Cx/ specifying the identity of x. By Exercise 133 the identity of x is an initial
object of Cx/ so that the map ∆0 → Cx/ is left anodyne by Lemma 13.4. It then follows from
Corollary 11.13 that the map in question is a trivial fibration and thus a Joyal equivalence:
Since p : E → C is a left fibration we find that Funq (Cx/ , E) = Funcc
q (Cx/ , E), and likewise for
∆0 in place of Cx/ .
It remains to prove the claim about the mapping anima in Cat∞ /C . For this we invoke
Proposition 12.15, which says that the following diagram is homotopy cartesian
mapCat∞ /C (q, p) mapCat∞ (Cx/ , E)
p∗
q
∆0 mapCat∞ (Cx/ , C)
 INTRODUCTION TO INFINITY-CATEGORIES 149

We now recall from Theorem 9.31 that the mapping anima in Cat∞ are canonically equivalent
to the groupoid cores of the functor categories. Hence the above right vertical map identifies
with the left vertical map in the diagram

Fun(Cx/ , E)' Fun(Cx/ , E)

p∗ p∗

Fun(Cx/ , C)' Fun(Cx/ , C)

Since p is a right fibration, so is p∗ by Theorem 3.32. Thus, by Proposition 5.3 the map p∗
is conservative, so that the diagram is cartesian by Exercise 79. Pasting together the two
diagrams, we find that the square

mapCat∞ /C (q, p) Fun(Cx/ , E)

p∗
q
∆0 Fun(Cx/ , C)

is homotopy cartesian. Since Funq (Cx/ , E) is the pullback of this diagram, we find that the
canonical map
mapCat∞ /C (q, p) → Funq (Cx/ , E)
is a Joyal equivalence as claimed. 

With this we find the usual consequence of the Yoneda lemma, namely that the Yoneda
functor is fully faithful. Henceforth, it will be called the Yoneda embedding.

Proposition 14.9. Let C be an ∞-category. Then the Yoneda functor Y : C → Fun(Cop , An)
is fully faithful.

Proof. We simply calculate that the evaluation map

mapFun(Cop ,An) (mapC (−, x), mapC (−, y)) −→ mapC (x, y)

is an equivalence by Proposition 14.8 and that the composite

Y
mapC (x, y) −→ mapFun(Cop ,An) (mapC (−, x), mapC (−, y)) −→ mapC (x, y)

is also an equivalence: The first map is equivalently described by the map

mapC (x, y) −→ mapCat∞ /C (Cx/ , Cy/ )

given by post composition on the slices. Restricting this to the identity of x clearly induces an
equivalence as needed. Hence also the map induced by the Yoneda functor is an equivalence,
which precisely says that the Yoneda functor is fully faithful. 

15. Limits and colimits

The following definition of colimits is taken from Krause–Nikolaus.
150 M. LAND

Definition 15.1. Let F : K → C be a functor and let x be an object of C. We define a

simplicial set MapC (F, x) by the pullback

MapC (F, x) Fun(K ? ∆0 , C)

(F,x)
∆0 Fun(K, C) × C

where the right vertical map is given by restriction along the canonical inclusion K ∪ {∞} ⊆
K ? ∆0 .

Lemma 15.2. In the above situation, MapC (F, x) is an ∞-groupoid. If F = y : ∆0 → C,

then MapC (y, x) = mapC (y, x).
Proof. Restriction along a monomorphism is a conservative inner fibration. Such are stable
under pullbacks. The second part is clear from the definition, as ∆0 ? ∆0 = ∆1 . 
Proposition 15.3. Let F : K → C be a functor and i : L → K be a map of simplicial sets.
Then for all objects x of C, i induces a map MapC (F, x) → MapC (F i, x). If i is right anodyne,
this map is a homotopy equivalence.
Proof. To see the first statement, we observe that there is a commutative diagram
(F,x)
∆0 Fun(K, C) × C Fun(K ? ∆0 , C)

(F i,x)
∆0 Fun(L, C) × C Fun(L ? ∆0 , C)

which induces the map of interest on pullbacks. To see that this map is a homotopy equiva-
lence if i is right anodyne, we consider the following. We claim that the diagram

Fun(K ? ∆0 , C) Fun(L ? ∆0 , C)

Fun(K ∪ {∞}, C) Fun(L ∪ {∞}, C)

is homotopy cartesian. To see this, we calculate the pullback to be given by

Fun(L ? ∆0 qL∪{∞} K ∪ {∞}, C).

The comparison map is then induced by the canonical inclusion L ? ∆0 qL∪{∞} K ∪ {∞} →
K ? ∆0 which is inner anodyne by ??, so that the comparison map is a Joyal equivalence as
needed. It follows that also the diagram

MapC (F, x) MapC (F i, x)

∆0 ∆0
is homotopy cartesian, so that the upper map is a homotopy equivalence as desired. 
 INTRODUCTION TO INFINITY-CATEGORIES 151

Definition 15.4. Let F : K → C be a functor and F̄ : K ? ∆0 → C a cone over F . We say

that F̄ is a colimit cone if for all objects x of C, the canonical map
MapC (F̄ , x) → MapC (F, x)
is a homotopy equivalence.

Remark. Since the map {∞} → K ? ∆0 is right anodyne, we find that for any cone
F̄ : K ? ∆0 → C, the canonical map
MapC (F̄ , x) → MapC (F̄ (∞), x) = mapC (F̄ (∞), x)
is a homotopy equivalence. Hence F̄ is a colimit cone if and only if for all objects x of C, the
above maps assemble to a homotopy equivalence
MapC (F, x) ' mapC (F̄ (∞), x).

Definition 15.5. Dually, for a functor F : K → C, one defines a simplicial set MapC (x, F )
as the pullback
MapC (x, F ) Fun(∆0 ? K, C)

(x,F )
∆0 C × Fun(K, C)
As before, a map i : L → K induces a map MapC (x, F ) → MapC (x, F i) which is a homotopy
equivalence if i is left anodyne. A cocone F̄ : ∆0 ? K → C of F is then called a limit cocone
if the canonical map
MapC (x, F̄ ) → MapC (x, F )
is a homotopy equivalence for all x in C.

Remark. Let us compare this definition to the definition of initial and terminal objects.
The goal is to see that an initial object is a colimit of the empty functor ∅ → C: A cone
over the empty functor is simply a functor y : ∆0 → C. Furthermore, MapC (∅, x) ∼ = ∆0 ,
and MapC (y, x) = mapC (y, x) by Lemma 15.2. Thus we find that an object y, viewed as
cone over the empty functor, is a colimit cone if and only if for all objects x, the mapping
anima mapC (y, x) is Joyal equivalent to ∆0 , i.e. is contractible. Thus, a colimit cone over the
empty functor is precisely an initial object. Likewise, a limit cocone over the empty functor
is precisely a terminal object.

Lecture 15 – 03.07.2019.
Example 15.6. (1) A colimit over a set (viewed as a discrete category) is called a co-
product. A limit over a set is called a product.
(2) A colimit of Λ20 → C is called a pushout. A limit of Λ22 → C is called a pullback.
Notice that Λ20 ? ∆0 ∼
= ∆1 × ∆1 and likewise that ∆0 ? Λ22 ∼= ∆1 × ∆1 .

In Lemma 13.2, we characterized initial and terminal objects in terms of certain maps
between slices to be trivial fibrations. We will now work towards a similar description for
general colimits and limits. To get started, we have the following lemma.
152 M. LAND

Lemma 15.7. For any two simplicial sets X and S, there is a canonical isomorphism
K S ∼ = [(K  ∆0 ) × S] qK×S K
compatible with the maps from K and S.
Proof. Consider the following diagram
K × ∆1 × S K × ∂∆1 × S K ×SqS

K × ∆1 × S K × ∂∆1 × S K qS
whose pushouts are given by (K  ∆0 ) × S and K  S, respectively. We thus find that the the
right of the small squares in the diagram
K ×S K ×SqS (K  ∆0 ) × S

K K qS K S
is a pushout. The left square is a pushout by inspection, so also the combined square is a
pushout. 
Lemma 15.8. Let F : K → C be a functor and let x be an object of C. Then the diagrams
MapC (F, x) CF/ MapC (x, F ) C/F

x x
∆0 C ∆0 C
are homotopy cartesian.
Proof. We argue for the left hand square, the other case is analogous. We first show that
there is a pullback diagram
CF/ Fun(K  ∆0 , C)

F
∆0 Fun(K, C)
For the time being, let us call the pullback Φ(F ). For a simplicial set S, a map to CF/
corresponds to a map K  S → C, whereas a map to Φ(F ) corresponds to a map
(K  ∆0 ) × S qK×S K → C.
Thus we conclude by Lemma 15.7. We then define a simplicial set C F be the pullback

CF Fun(K ? ∆0 , C)

F
∆0 Fun(K, C)
From the canonical map K  ∆0 → K ? ∆0 we obtain maps
CF → CF/ ← CF/
 INTRODUCTION TO INFINITY-CATEGORIES 153

and we claim that both are Joyal equivalences: the right map was dealt with in Proposi-
tion 9.25 and for the left map it follows again from the fact that the map K  ∆0 → K ? ∆0 is
a Joyal equivalence, and that the pullbacks involving CF/ and CF are invariant under Joyal
equivalences by Lemma 9.6. We then consider the diagram

MapC (F, x) CF Fun(K ? ∆0 , C)

∆0 C Fun(K, C) × C

∆0 Fun(K, C)
consisting of pullback diagrams. The claim then follows from the fact that there is a Joyal
equivalence CF ' CF/ . 
Remark. One could define a variant of MapC (F, x) using the fat join instead of the ordinary
join. The resulting ∞-groupoid Map
g C (F, x) will be canonically equivalent to MapC (F, x) and
we will freely exchange the two when useful. The proof of Lemma 15.8 then shows that for
this variant the following diagram is a pullback.

Map
^C (F, x) CF/

∆0 C

The following theorem says that our definition of limits and colimits coincides with the one
given usually, for instance in [Lur09].
Theorem 15.9. Let F : K → C be a diagram in an ∞-category C. A cone F̄ : K ? ∆0 → C of
F is a colimit cone if and only if it is an initial object of CF/ . Dually, a cocone F̄ : ∆0 ?K → C
of F is a limit cocone if and only if it is a terminal object of C/F .
Proof. A cocone F̄ gives rise to a commutative diagram
CF̄ / CF/

C
of left fibrations. By Exercise 140, F̄ is an initial object if and only if the horizontal map is a
trivial fibration, which is the case if and only if it is a Joyal equivalence. By Theorem 10.21,
this is the case if and only if the induced map on fibres over objects x of C is a homotopy
equivalence. By Lemma 15.8, the induced map on fibres is equivalent to the map
MapC (F̄ , x) → MapC (F, x).
We thus find that the map CF̄ / → CF/ is a Joyal equivalence if and only if F̄ is a colimit cone
of F . The argument for limits is the same, using the dual version of Lemma 15.8. 
The following will be very useful later,
154 M. LAND

Proposition 15.10. Let C be an ∞-category and consider a pushout of simplicial sets

i
A B

C K
in which the map i is a monomorphism. Let F : K → C be a functor, and denote by FA , FB ,
and FC its restriction to A, B and C. Then, for each object x of C, the diagram
MapC (F, x) MapC (FB , x)

MapC (FC , x) MapC (FA , x)

is homotopy cartesian.
Proof. To prove this, we may replace MapC (F, x) with Mapg C (F, x), the version defined using
the fat slice. We observe that the diagram is a pullback, because Map
g C (F, x) is defined as a
fibre which commutes with pullbacks, and both functors Fun(−, C) and Fun(−, C) × C send
pushouts to pullbacks. Then we can use that −  ∆0 also preserves pushouts. It hence suffices
g C (FB , x) → Map
to show that the map Map g C (FA , x) is a Kan fibration. By a previous remark
there is a pullback diagram
Map
g C (FB , x) CFB /

Map
g C (FA , x) CFA /
so it suffices to recall that the map CFB / → CFA / is a left fibration, so that its pullback is a
left fibration between Kan complexes, and thus a Kan fibration. 
We again find that (co)limits, if they exist, are unique up to contractible choice.
Lemma 15.11. Let p : K → C be a diagram. Let (Cp/ )colim ⊆ Cp/ and (C/p )lim ⊆ C/p be
the full subcategories spanned by colimit cones and limit cocones. Then (Cp/ )colim , respectively
(C/p )lim , are either empty or contractible Kan complexes.
Proof. This is merely a reformulation of the case for initial and terminal objects Proposi-
tion 13.3. 
The following will be used later to show that forming colimits (if possible) is a functor.
Proposition 15.12. Let K be a simplicial set, F : K → C a functor, and x an object of an
∞-category C. There is a canonical homotopy equivalence
MapC (F, x) ' mapFun(K,C) (F, constx ).
Proof. We claim that the following diagram is cartesian
CF/ Fun(K, C)F/

const
C Fun(K, C)
 INTRODUCTION TO INFINITY-CATEGORIES 155

where we also view F as a functor ∆0 → Fun(K, C). Thus, pulling back along a map x : ∆0 →
C, we obtain a Joyal equivalence
'
MapC (F, x) → mapL
Fun(K,C) (F, constx )

as claimed. To see the claim, we find that a map X → CF/ corresponds to a map X  K → C
whose restriction to K is F . Likewise a map to the pullback of the above diagram corresponds
to a map
(X  ∆0 ) × K qX×K X → C
restricting also appropriately. We thus again conclude from Lemma 15.7. 
Remark. In fact, the proof shows that there is an isomorphism of simplicial sets
g C (F, x) ∼
Map = mapFun(K,C) (F, constx ).

Lemma 15.13. Let C be an ∞-category. Let id ¯ : C ? ∆0 → C be a cone over the identity of

¯ is a colimit cone if and only if id(∞)
C. Then id ¯ is a terminal object. In particular, C has a
terminal object if and only if the identity functor has a colimit.
Proof. Suppose t is a terminal object of C. Consider the composite {t} → C → C where the
latter functor is the identity. We obtain an induced functor on slices Cid/ → Ct/ , which is a
trivial fibration since the inclusion {t} → C is right anodyne by Lemma 13.4. Since Ct/ has
an initial object, so does Cid/ . Hence by Theorem 15.9, t is a colimit of the identity functor.
¯ We will now need to show
Conversely, let us assume that the identity has a colimit cone id.
¯
that x = id(∞) is a terminal object. We will allude to Lemma 13.2 and consider a lifting
problem
∆{n} ∂∆n C

∆n
in which the upper composite is given by the object x. We apply the functor − ? ∆0 to this
diagram and obtain

∆{n} ∆{n,n+1}

¯
id
∂∆n Λn+1
n+1 C ? ∆0 C

∆n ∆n+1
The composite ∆{n,n+1} → C is morphism in C from x to x and we claim that this morphism
is an equivalence. Hence a dashed arrow exists by Joyal’s lifting theorem Theorem 5.8. We
claim that this morphism extends to a morphism in Cid/ from id ¯ to itself. Since id
¯ is a
colimit cone, it is an initial object of Cid/ , so that all its endomorphisms are invertible by
Proposition 13.3. To see the claim, we consider the following composite
¯ 0 ¯
C ? ∆1 ∼
= C ? ∆0 ? ∆0
id?∆
C ? ∆0 id
C
156 M. LAND

¯ It hence suffices to show

which determines a 1-simplex in Cid , whose source and target are id.
that the induced map on cone points is the one considered above. By construction, we thus
have to analyze the map
∆1 → C ? ∆1 → C ? ∆0 → C
and observe that the map ∆1 → C ? ∆0 sends 0 to id(∞)¯ = x and 1 to the cone point ∞. It
is thus given by the map
x?id
∆1 ∼ = ∆0 ? ∆0 −→ C ? ∆0 .
Composing this map with id¯ is precisely the bended map in the above diagram. 
Lecture 16 – 08.07.2019. We will need the following refined version of Proposition 15.10.
Proposition 15.14. Let C be an ∞-category and consider a pushout of simplicial sets
i
A B

C K
in which the map i is a monomorphism. Let F : K → C be a functor, and denote by FA , FB ,
and FC its restriction to A, B and C. Then the square

CF/ CFB /

CFC / CFA /
is a cartesian and homotopy cartesian square of left fibrations over C.
Remark. Upon passing to fibres over objects x of C we obtain the cartesian and homotopy
cartesian square of Proposition 15.10.

We now interpret the definitions and properties of colimits as follows. First we observe
that the left fibration
CF/ → C
corresponds by straightening-unstraightening to a functor C → An. As it takes on an object
x in C the value MapC (F, x), we simply denote this functor by MapC (F, −).
Definition 15.15. A functor C → An is called representable if it is equivalent to the functor
mapC (x, −) for some x in C. Any such x will be called a representing object. Equivalently,
a functor is representable by x, if the associated left fibration E → C is equivalent to the left
fibration Cx/ → C.

Proposition 15.16. Let F : K → C be a functor. Then F admits a colimit if and only if

the functor MapC (F, −) is representable, and any representing object is a colimit of F .
Proof. Suppose F admits a colimit cone F̄ : K ? ∆0 → C and let x = F̄ (∞). We have seen
previously that this implies that both functors
CF/ ← CF̄ / → Cx/
are equivalences of left fibrations over C: For the right map this is always the case, and the
left one is an equivalence if and only if F̄ is a colimit cone.
 INTRODUCTION TO INFINITY-CATEGORIES 157

It hence remains to show that if MapC (F, −) is representable by an object x, then F admits
a colimit. By the assumption we find an equivalence
Cx/ ' CF/
of left fibrations over C. Pick an initial object of Cx/ for instance the identity of x. Under the
above equivalence this is transported to an initial object of CF/ . Any such is a colimit cone,
whose colimit point is x by construction. 
Corollary 15.17. Every diagram F : ∆n → C admits both a limit and a colimit: A limit is
given by evaluation on 0, and a colimit is given by evaluation on n.
Proof. It suffices to recall that these inclusions induce equivalences of left, resp. right fi-
brations CF/ ' CF (n)/ and C/F ' C/F (0) , so that both CF/ and C/F are representable left
fibrations. 
We will need the following technical proposition.
Proposition 15.18. Let F : K → C be a functor, and let i : L → K be a map of simplicial
sets. Let F̄ be a colimit cone of F , and let F i be a colimit cone of F i. Let G : ∆1 → CF i/
be a map with G(0) = F i and G(1) = F̄ i. Let g be the composite ∆1 → CF i/ → C, and let
g(0) = x and g(1) = y. Then there is a commutative diagram
'
CF/ CF i/ ' CF i/ Cx/
'

CF̄ / CF̄ i/ ' (CF i/ )G/

'

Cy/ Cy/ ' Cg/

and y is a colimit of F and x is a colimit of F i. The resulting map
Cy/ → Cx/
is given by precomposition with the map g which we call the induced map x = colimL F i →
colimK F = y.
Remark. The obvious dual situation for limiting cocones holds as well.

Observation 15.19. Informally, we summarize the above situation by saying that there is
a commutative diagram
CF/ CF i/
' '
g
Cy/ Cx/
where the vertical maps are the equivalences coming from the fact that y is a colimit of
F and x is a colimit of F i. Of course, this diagram does not actually commute, but it
commutes up to an invertible natural transformation. On the other hand, we could also take
this diagram as a definition for the morphism g : x → y: Inverting the map CF/ → Cy/ , we
obtain a functor Cy/ → Cx/ making the diagram commute. By the Yoneda lemma, such a
functor is equivalently given by an object of Cx/ ×C {y} ' mapC (x, y), this is going to be a
158 M. LAND

morphism equivalent to the morphism g of the previous proposition. The advantage of the
above approach is that it shows that the morphism g canonically lifts to a morphism between
cones over F i.

Lemma 15.20. Let I be a discrete category and let F : I → C be a diagram, i.e. a collection
xi of objects of C. Show that there is a canonical equivalence
Y
MapC (F, y) ' mapC (xi , y).
i∈I

between functors C → An in y.
Proof. The following diagram is a pushout:

∆1
`
I × {0} q I × {1}
I

I q {∞} I ? ∆0

Thus in the following diagram, both small squares are pullbacks

CF/ Fun(I ? ∆0 , C) Fun(∆1 , C)

Q
I

(F,id) (s,∆)
C Fun(I, C) × C C×C
Q
I

and thus so is the big square. This should imply the lemma.
Since pullbacks commute with arbitrary products, we find that there is an isomorphism

CF/ ∼
Y
= C xi /
I

of left fibrations. This implies the lemma. 

Corollary 15.21. Let F : Λ20 → C be a diagram, depicted as
f
x y
g

z
and let t be another object. Then there is a homotopy cartesian square of ∞-groupoids

MapC (F, t) mapC (y, t)

mapC (z, t) mapC (x, t)

which is natural in t.
 INTRODUCTION TO INFINITY-CATEGORIES 159

Proof. By ?? and the fact that Λ20 is the pushout ∆1 q∆0 ∆1 , we find that the diagram

CF/ Cf /

Cg/ Cx/
is cartesian and homotopy cartesian and consists of left fibrations over C. Furthermore,
Cf / ' Cy/ and Cg/ ' Cz/ . Passing to fibres over a point t we obtain the statement. 
Proposition 15.22. Let C be an ∞-category and K a simplicial set, written as a pushout
B qA C, where the map A → B is a monomorphism. Let F : K → C be a diagram. Suppose
F|B has a colimit y, F|A has a colimit x and F|C has a colimit y. If C has pushouts, then a
pushout y qx z is a colimit of F .
Proof. [Lur09, [Link]]. Let G : Λ20 → C be the diagram given by y ← x → z, where the maps
come from Proposition 15.18. Let Ḡ : ∆1 × ∆1 → C be a colimit cone of G. We wish to show
that w = Ḡ(1, 1) is a colimit of F : K → C. By Proposition 15.16, we have to show that there
is an equivalence of functors between mapC (w, −) and MapC (F, −). Rephrasing in terms of
left fibrations, we need to show that there is an equivalence of left fibrations Cw/ ' CF/ . We
have that Cw/ ' CG/ since w is a colimit of G. Furthermore, there is a homotopy cartesian
square
CG/ Cy/

Cz/ Cx/
Since x, y, and z are themselves colimits, we find that there is commutative diagram, see
Observation 15.19
CFB / CFA / CFC /
' ' '

Cy/ Cx/ Cz/

in which the vertical comparison maps are Joyal equivalences. We thus find that there is an
induced equivalence on homotopy pullbacks
CG/ ' CF/
so that CF/ is indeed representable, with w a representing object. 
Lecture 17 – 10.07.2019.
Proposition 15.23. Let F : K → C be a functor, and let K = colim Ki be an N-indexed
i≥0
decomposition with each map Ki → Ki+1 a monomorphism. Let Fi be the restriction of F
to Ki . Suppose that for all i the functor Fi admits a colimit and that C admits colimits over
1-dimensional simplicial sets. Then F admits a colimit.
Proof. First, we find that there is an isomorphism of left fibrations CF/ → lim CFi / , because
i≥0
for every simplicial set X, the functor X  − preserves connected colimits. By assumption,
all CFi are corpresentable left fibrations, say CFi / ' Cxi / . We thus obtain canonical maps
Cxi−1 / → Cxi / making the comparison diagrams commute. By the Yoneda lemma, all of these
160 M. LAND

maps corresponds to morphisms αi : xi−1 → xi , and they assemble into a functor G : I ∞ → C.

Since I ∞ = colimn I n , we find that CG/ → lim CGi . Since I n has a terminal object, we find
i≥0
that CGi ' CG(i)/ = Cxi / . We obtain commutative diagrams

CFi / C xi / CG i /

CFi−1 / Cxi−1 / CGi−1 /

in which all horizontal maps are Joyal equivalences, and the outer vertical maps are isofibra-
tions. We thus find that the induced map on vertical limits is a Joyal equivalence CG/ ' CF/ .
Since I ∞ is 1-dimensional, we find that C admits I ∞ -indexed colimits. Hence CG/ is corepre-
sentable, and thus so is CF/ . Thus F admits a colimit. 
Proposition 15.24. If an ∞-category C admits small coproducts and pushouts, then it
admits all small colimits.
Proof. [Lur09, [Link]]. We first show that C admits colimits indexed over finite dimensional
simplicial sets K by induction over the dimension of K. If K is zero dimensional, it is simply
a discrete set, so that colimits over such are coproducts and hence exist by assumption. Now
suppose K is n-dimensional, and consider its skeletal pushout
∂∆n
`
skn−1 (K)
i∈I

∆n
`
K
i∈I

and consider a functor F : K → C. By Proposition 15.22 it suffices to argue that ` each

restriction of F to any of the other three corners admits a colimit. For skn−1 (K) and ∂∆n
` i∈I
this follows by the induction hypothesis. It remains to show that any functor ∆n → C
i∈I
admits a colimit. By Corollary 15.17 we know that every single functor ∆n → C admits a
colimit. By the same argument we find that the restriction along all terminal objects gives a
Joyal equivalence
∆n / '
`

Ci∈I → CI/
because the coproduct of right anodyne maps is again right anodyne. It hence suffices to
argue that CI/ is equivalent to a representable left fibration, which again follows from the
assumption that C admits coproducts.
We then use Proposition 15.23 to conclude that C admits K-shaped colimits for all small
simplicial sets K by writing K as the colimit over its skeleta. 
In the presence of finite coproducts, having pushouts is in fact equivalent to having co-
equalizers.
Lemma 15.25. If an ∞-category C admits finite coproducts, then it admits pushouts if and
only if it admits coequalizers. In particular, C admits small coproducts and coequalizers, then
it admits all small colimits.
 INTRODUCTION TO INFINITY-CATEGORIES 161

Proof. The coequalizer category is the pushout of the diagram ∆1 ← ∂∆1 → ∆1 . Hence by
Proposition 15.22 C admits coequalizers if it admits pushouts and finite coproducts. To see
the converse, one argues similarly as in Proposition 15.22: Suppose K 0 is the coequalizer of
two morphisms L → K of simplicial sets, and suppose K 0 → C is a functor such that the
restrictions to K and to L admit a colimit. Then the coequalizer of the colimits is a colimit of
the functor K 0 → C, see [Lur09, [Link]] or Exercise 141. Having this, we observe that there
is a coequalizer diagram
∆0 ∆1 q ∆1 Λ20

where the two maps are the two inclusions as vertex 0. We deduce that colimits over Λ20 are
given by a coequalizer of two maps between colimits indexed over ∆0 and ∆1 q ∆1 which
exist if C admits finite coproducts. 
Definition 15.26. Let f : C → D be a functor between ∞-categories and let F : K → C be
a diagram. Suppose that F admits a colimit in C. We say that f preserves this colimit, if for
some (and hence any) colimit cone F̄ : K ? ∆0 → C, the resulting diagram K ? ∆0 → C → D
is a colimit cone over f F .
We say that F preserves K-shaped colimits, if for every functor F : K → C which admits
a colimit, F preserves this colimit.

Remark. A functor f : C → D thus preserves K-shaped colimits, if for every functor

F : K → C, the induced functor CF/ → Df F/ preserves initial objects.

Proposition 15.27. Let F : C → D be a functor between ∞-categories. Then F preserves

small colimits if and only if it preserves small coproducts and pushouts. The same holds true
if one replaces pushouts by coequalizers.
Proof. Exercise. 

We will collect the following properties of colimits, and leave proofs for later or as exercises.
Proposition 15.28. If D is (co)complete and K is a small simplicial set, then Fun(K, D)
is again (co)complete and colimits are calculated pointwise, i.e. for every object x of K, the
evaluation functor Fun(K, D) → D preserves (co)limits.
Proposition 15.29. Let C be a (co)complete ∞-category and p : K → C a diagram. Then
C/F admits colimits and the functor C/F → C preserves colimits. Dually, CF/ admits limits
and CF/ → C preserves limits.
Proof. [Lur09, [Link]]. 
Proposition 15.30. Let C be a cocomplete ∞-category and let F : K → C be a diagram.
Then CF/ is again cocomplete. Dually, if C is complete, then C /F is complete. The forgetful
functors, however, in general do not preserve these (co)limits.
Proposition 15.31. Let f : E → C be a left fibration and let K be a weakly contractible
simplicial set. Then f preserves K-shaped colimits. Likewise, right fibrations preserve con-
tractible limits.
Proof. [Lur09, [Link] & [Link]]. 
162 M. LAND

We will later give an independent proof of the following theorem, making use of the
straightening-unstraightening equivalence. The advantage of the following proof is to see
that limits and colimits are given as we expect.

Theorem 15.32. The ∞-categories Cat∞ and An admit all small limits and colimits.

Proof. By Proposition 15.24 it suffices to show that these ∞-categories admit small products,
coproducts, pullbacks and pushouts. Coproducts and products are quite easy, as we have
seen earlier. We indicate that Cat∞ admits pullbacks, all other cases are similar in flavour.
Consider a diagram F : Λ22 → Cat∞ , given by two functors

p f
D → C ← C0 .

Without loss of generality, we may assume that D → C is an isofibration. We let D0 be the

pullback of the above diagram of simplicial sets and let F̄ : ∆1 × ∆1 → Cat∞ be the whole
/F
pullback diagram. We wish to show that Cat∞ is a representable right fibration. By the
dual argument of Corollary 15.21, we know that there is a cartesian and homotopy cartesian
square

/F /f
Cat∞ Cat∞

/p /C
Cat∞ Cat∞

/D0 /F̄ /F
Furthermore, there is a canonical map Cat∞ ' Cat∞ → Cat∞ and we wish to show that
this functor is an equivalence. We will show that it is essentially surjective and fully faithful.
An object of C/F can be represented (up to equivalence) by a commutative diagram of ∞-
categories

E D
p
f
C0 C

by the universal property, E comes with a unique map to D0 which gives the resulting object
/D0 /D0
of Cat∞ . To see fully faithfulness it suffices to show that for any two objects E, E0 of Cat∞
the following diagram of mapping anima is homotopy cartesian.

map /D0 (E, E0 ) map /C0 (E, E0 )

Cat∞ Cat∞

mapCat/D (E, E0 ) mapCat/C (E, E0 )

∞ ∞

This follows from the description of mapping anima in slice-categories, Proposition 12.15, the
fact that D0 ∼
= D ×C C0 , and that D → C is an isofibration. 
 INTRODUCTION TO INFINITY-CATEGORIES 163

16. Cofinal and coinitial functors

Definition 16.1. Let f : K → L be a map of simplicial sets and p : X → L an inner
fibration. As before, we define an ∞-category by the pullback
Funf (K, X) Fun(K, X)
p∗
f
∆0 Fun(K, L)
If f = id : L → L we simply write FunL (L, X) instead of Funid (K, X).

Definition 16.2. Let f : K → L be a map of simplicial sets. Then f is called cofinal if for
all right fibrations p : X → Y , the canonical map
FunL (L, X) → Funf (K, X)
induced by f is a Joyal equivalence. Likewise, it is called coinitial, if for all left fibrations
p : X → Y , the canonical map
FunL (L, X) → Funf (K, X)
is a Joyal equivalence.

Remark. By construction, Funf (K, X) is an ∞-groupoid if p : X → L is a right or left

fibration, as then its canonical map to ∆0 is a right or left fibration and hence a Kan fibration.
In previous notation, if p : E → C is a left fibration between ∞-categories, this was written as
f (C, E ) because for left fibrations, any morphism in E is p-cocartesian.
Funcc \

Proposition 16.3. Let f : K → L and g : L → M be maps of simplicial sets.

(1) If f is cofinal, then gf is cofinal if and only if g is.
(2) If f is cofinal, then f is a weak equivalence.
(3) If f is a monomorphism, then f is cofinal if and only if it is right anodyne.
Proof. We prove (1) first and consider a right fibration D → M and the diagram
FunM (M, D) Fung (L, D) ∼
=
FunL (L, g ∗ D)

Fungf (K, D) ∼
=
Funf (K, g ∗ (D))
in which the right horizontal maps are isomorphisms, and the right most vertical map is a
Joyal equivalence by the assumption that f is cofinal and the fact that pullbacks of right
fibrations are right fibrations. We thus conclude the statement by the 3-for-2 property. To
show (2) it suffices to prove that for any Kan complex X, the canonical map Fun(L, X) →
Fun(K, X) is a homotopy equivalence. Consider the map L×X → L which is a Kan fibration,
and in particular a right fibration. We find that FunL (L, L × X) ∼ = Fun(L, X) and that
Funf (K, L × X) ∼ = Fun(K, X) as needed. One direction of (3) has been done previously:
A right anodyne map is cofinal, in fact the restriction map one has to analyze is a trivial
fibration: By adjunction it suffices to prove that for any monomorphism S → T , the induced
map
S×L∪T ×K →T ×L
164 M. LAND

is right anodyne as well. Conversely, suppose that f is a cofinal monomorphism and let
X → Y be a right fibration, and consider a lifting problem to see whether f is right anodyne.
By pulling back, we may assume that Y = L and get a diagram
K X

L L
and wish to show the existence of the dashed arrow. By assumption we know that the
morphism
FunL (L, X) → Funf (K, X)
is a Joyal equivalence, and we wish to show that it is in fact a trivial fibration (as then it is
surjective on 0-simplices). It hence suffices to show that it is an isofibration. To see that it
is an inner fibration, we again use that the pushout product of an inner anodyne map with a
monomorphism is inner anodyne, and that right fibrations are in particular inner fibrations.
For the remaining property, we have to show that any diagram

{1} FunL (L, X)

∆1 Funf (K, X)

admits a dashed arrow: This is because the right vertical map is conservative since it is a
functor between ∞-groupoids. Now we use that the pushout product of a right anodyne map
and a monomorphism is again right anodyne, so that a lift exists since X → L is a right
fibration. 

Lecture 18 – 15.07.2019.
Corollary 16.4. Let f : K → L and g : L → M be maps of simplicial sets.
(1) If f is coinitial, then gf is coinitial if and only if g is.
(2) If f is a monomorphism, then f is coinitial if and only if f is left anodyne
Proof. This follows immediately from Exercise 142 
Corollary 16.5. Among monomorphisms, the left and right anodyne maps satisfy the right
cancellation property: If f and g are composable morphisms, and both f and gf are left, resp.
right anodyne, then so is g.
Next we want to prove an important characterization of cofinal maps, which builds on the
following lemma.
Lemma 16.6. Let F : L → C be a diagram and x an object of C. Then there is a canonical
cartesian (and homotopy cartesian) diagram as follows.

Map
g C (F, x) Fun(L, C/x )

F
∆0 Fun(L, C)
 INTRODUCTION TO INFINITY-CATEGORIES 165

Proof. Again, for the time of the proof, let us call the pullback Φ(F ). We claim that a map
from a simplicial set X to Φ(F ) corresponds to a map
[(X × L)  ∆0 ] qX×L L → C
whose restriction to ∆0 is given by x and whose restriction to L is F . On the other hand, a
map from X to Mapg C (F, x) corresponds to a map
[X × (L  ∆0 )] qX×(L∪{∞}) L ∪ {∞} → C
whose restriction to L ∪ {∞} is the pair (F, x). We claim that there is an isomorphism of
simplicial sets
∼ [X × (L  ∆0 )] qX×(L∪{∞}) L ∪ {∞}.
[(X × L)  ∆0 ] qX×L L =
For this we calculate as follows:
h i
[(X × L)  ∆0 ] qX×L L ∼ = (X × L × ∆ 1
) q X×L×∂∆1 (X × L) q ∆ 0
qX×L L
∼
= (X × L × ∆1 ) qX×L×∂∆1 (X × L) q ∆0 qX×L L

∼ (X × L × ∆1 ) qX×L×∂∆1 L q ∆0
=
∼
= (X × L × ∆1 ) qX×L×∂∆1 X × (L q ∆0 ) qX×(Lq∆0 ) L q ∆0



∼
= (X × L × ∆1 ) qX×L×∂∆1 X × (L q ∆0 ) qX×(Lq∆0 ) L q ∆0

∼
= X × (L  ∆0 ) qX×(Lq∆0 ) L q ∆0

which shows the lemma, once we convince ourselves that the inclusions of ∆0 and L correspond
to each other which is a simple matter of checking the maps. 
Theorem 16.7. Let f : K → L be a map of simplicial sets. Then f is cofinal if and only if
for each ∞-category C and each diagram p : L → C, the induced map Cp/ → Cpf / is a Joyal
equivalence.
Proof. First we assume that f is cofinal and show that then the map Cp/ → Cpf / is a Joyal
equivalence. Since this is a map of left fibrations over C it suffices to show that the induced
map on fibres over object x of C is an equivalence. By Lemma 16.6, this map identifies up to
homotopy equivalence with the map Funp (L, C/x ) → Funpf (K, C/x ) which is an equivalence
because C/x → C is a right fibration and f is cofinal. Conversely, assume that Cp/ → Cpf /
is a Joyal equivalence for any diagram p : L → C and let X → L be a right fibration. By
the straightening-unstraightening equivalence, there is a functor p : L → Anop whose pullback
of the universal right fibration is equivalent to X → L. Now we use that the universal
right fibration is given by (An∗/ )op → Anop which is a representable right fibration since
(An∗/ )op ' (Anop )/∗ . Thus, using Lemma 16.6, we find
FunL (L, X) ' Funp (L, (An∗/ )op ) ' MapAnop (p, ∗) ' (Anop )p/ ×Anop {∗}
and likewise that
Funf (K, X) ' (Anop )pf / ×Anop {∗}.
The map we have to investigate is the map induced by the map (Anop )p/ → (Anop )pf / of left
fibrations over Anop by taking the fibre over ∗ ∈ Anop . By assumption this map is a Joyal
equivalence, and thus so is the induced map on fibres. 
166 M. LAND

Corollary 16.8. Any Joyal equivalence f : K → L is cofinal.

Proof. By Theorem 16.7 it suffices to show that for each ∞-category C and each diagram,
the induced map Cp/ → Cpf / is a Joyal equivalence. For this it suffices to show that for each
further ∞-category D, the induced map
Fun(D, Cp/ ) → Fun(D, Cpf / )
is an equivalence. By adjunction, these are isomorphic to
Funp (D  L, C) → Funpf (D  K, C)
which are in turn given by the pullbacks in the diagram
p
∆0 Fun(L, C) Fun(D  L, C)

pf
∆0 Fun(K, C) Fun(D  K, C)
in which the right horizontal maps are isofibrations and all vertical maps are Joyal equiva-
lences: For the right hand side this follows from the fact that D  K → D  L is again a Joyal
equivalence by Corollary 9.16. Thus the induced map on pullbacks is a Joyal equivalence as
well by Lemma 9.6. 
Corollary 16.9. Let f : K → L be a cofinal map and let p : L → C be a diagram with C an
∞-category. Then f admits a colimit if and only if pf admits a colimit, and in either case,
f preserves this colimit.
Lemma 16.10. Let f : K → L be a map of simplicial sets. Then f is a trivial fibration if
and only if it is a cofinal right fibration.
Proof. The only if follows from the fact that trivial fibrations are right fibrations and Joyal
equivalences. To see the converse, we will show that the fibres are contractible and allude to
the dual version of Exercise 112. So let K → L be a cofinal right fibration. We obtain that
the map
FunL (L, K) → Funf (K, K)
is an equivalence. The right hand side contains the functor idK : K → K, so there exists an
object ϕ : ∆0 → FunL (L, X) whose image in Funf (K, K) is equivalent to idK . Spelling this
out, we obtain that f ϕ = idL and that there exists a 1-simplex ∆1 → Funf (K, K) connecting
idK to ϕf . This corresponds to a commutative diagram
h
∆1 × K K
pr f
f
K L
One can then restrict the map h to ∆1
× Kx for any 0-simplex x of L. The resulting map
is easily seen to give a homotopy between idKx and ϕf restricted to Kx . The latter map
is constant, since f is constant on the fibres. Thus each fibre Kx is a contractible Kan
complex. 
Proposition 16.11. A map is cofinal if and only if it is a composite of a right anodyne
map followed by a trivial fibration.
 INTRODUCTION TO INFINITY-CATEGORIES 167

Proof. The if case is clear: Both right anodyne maps and trivial fibrations are cofinal, and
compositions of cofinal maps are cofinal. Conversely, given a cofinal map f : K → L, we may
factor it as a right anodyne map K → K 0 followed by a right fibration K 0 → L. Since right
anodyne maps are cofinal, we find that the right fibration K 0 → L is cofinal by Proposition 16.3
part (1). We then conclude the proposition from Lemma 16.10. 
Definition 16.12. Let p : Y → X be a map of simplicial sets. We call p smooth, if for every
pullback diagram
j
B Y
p
i
A X
where i is cofinal, the map j is again cofinal. Dually, it is called proper if for every such
pullback diagram where i is coinitial, the map j is again coinitial.

Definition 16.13. A map of simplicial sets p : Y → X is called universally smooth if the

pullback along any map X 0 → X is smooth. Likewise, it is called universally proper if the
pullback along any map is proper.

Remark. A word of warning is in order. In [Lur09] what we call universally smooth is simply
called smooth, and likewise for proper and universally proper. The reason to favour universally
proper over what we call proper ist that universally proper maps are closed under pullback,
whereas proper maps are not closed under pullback, see Exercise 146. In [Cis19, Ngu18] what
we call universally smooth is called proper, and what we call universally proper is called
smooth. The terminology in fact breaks a little earlier: What we call cofinal is called final in
loc. cit, and what we call coinitial is called cofinal. What we call smooth or proper does not
have a separate name in loc. cit.

For the following proposition, we follow the proof given in [Ngu18, 2.3.23] and [Cis19].
Proposition 16.14. Consider a pullback diagram
j
B Y
p
i
A X
where p is a left fibration and i is right anodyne. Then the map j is again right anodyne.
Proof. The first step is to see that it suffices to prove the claim for i contained in a generating
set of right anodyne maps. To see this, we claim that the set S of right anodyne maps
satisfying the conclusion of the lemma is saturated, see Exercise 144. For compositions and
retracts this is easy, one needs to work a little harder to see that these maps are closed
under pushouts. It hence suffices to show that the maps {1} × ∆n ∪ ∆1 × ∂∆n → ∆1 × ∆n
are contained in this set by Corollary 3.31. We observe also that the set S has the right
cancellation property, since right anodyne maps have this property, see again Exercise 144.
Thus it suffices to show that for any simplicial set K, the map
{1} × K → ∆1 × K
168 M. LAND

is contained in S: Consider the diagram

{1} × ∂∆n ∆1 × ∂∆n

{1} × ∆n {1} × ∆n ∪ ∆1 × ∂∆n

∆1 × ∆n

Using the claim, we find that the top horizontal map is contained in S, hence as S is closed
under pushouts, so is the lower horizontal map. Also, the the lower bended map is contained
in S. Since S satisfies the right cancellation property, also the diagonal map is contained in
S.
Now we observe that the maps {1}×K → ∆1 ×K are particular instances of right anodyne
extensions: They are right deformation retracts. Such are closed under pullbacks along left
fibrations, as we show in Exercise 145. 

Remark. By applying the opposite functor, we find that if p is a right fibration and i is left
anodyne, then j is again left anodyne.

Remark. The conclusion of Proposition 16.14 holds more generally for cocartesian fibrations
p : Y → X, see [Lur09, Proposition [Link]].

Corollary 16.15. Left fibrations are universally smooth. In fact, cocartesian fibrations are
universally smooth. Right fibrations, in fact cartesian fibrations, are universally proper.

Proof. Since left and right fibrations are closed under pullbacks, it suffices to show that a left
fibration is smooth. Consider a diagram as in the definition of smooth maps. Factor the map
i as
i0 p
A → A0 → X
with i0 right anodyne and p a trivial fibration and consider the enlarged diagram
j0 q
B B0 Y

A A0 X

We find that q is a trivial fibration and that j 0 is right anodyne by Proposition 16.14. Thus
j is cofinal as a composition of cofinal maps. 

Using that left fibrations are smooth and right fibrations are proper, we obtain a nice proof
of an ∞-categorical version of Quillen’s Theorem A. This is yet another characterization of
cofinality for maps whose target is an ∞-category.

Theorem 16.16. Let f : C → D be a map of simplicial sets with D an ∞-category. Then f

is cofinal if and only if for all objects d of D, the simplicial set Cd/ is weakly contractible.
 INTRODUCTION TO INFINITY-CATEGORIES 169

Proof. First assume that f is cofinal. In the pullback diagram

Cd/ Dd/

C D
the right vertical map is a left fibration, and thus smooth. It follows that Cd/ → Dd/ is cofinal
and thus a weak equivalence. Since Dd/ is weakly contractible (it has an initial object) so is
Cd/ . Conversely, we consider a factorization of f as
i p
C→E→D
where i is right anodyne and p is a right fibration. We aim to show that p is a trivial fibration,
so that f is cofinal by Proposition 16.11. Consider the diagram

Cd/ Ed/ Dd/

C E D
In which all squares are pullbacks. Since the very right vertical map is a left fibration, so is
the middle vertical map. By Proposition 16.14, the map Cd/ → Ed/ is again right anodyne
and hence a weak equivalence, and hence Ed/ is weakly contractible.
Now consider the diagram
Ed Ed/ E

idd
∆0 Dd/ D
Again, all squares are pullbacks. This time the right most vertical map is a right fibration,
hence so is the middle vertical map. Furthermore, the map ∆0 → Dd/ is left anodyne by
Lemma 13.4 as idd is an initial object of Dd/ , see Exercise 133. Hence, by the dual version of
Proposition 16.14, the map Ed → Ed/ is left anodyne and thus a weak equivalence. Since Ed/
is weakly contractible by the first step and Ed is an ∞-groupoid, this means that the fibres
Ed of the right fibration p : E → D are contractible. Hence p is a trivial fibration, for instance
by Exercise 112. 

Here is the actual statement Quillen proved:

Corollary 16.17. Let F : C → D be a functor between ordinary categories. If all slices Cd/
are weakly contractible, then the functor N(C) → N(D) is a weak equivalence.
Proof. By Theorem 16.16, the functor N(F ) : N(C) → N(D) is cofinal, and hence a weak
equivalence by Proposition 16.3. 

17. Adjunctions
Definition 17.1. An adjunction is a bicartesian fibration E → ∆1 , i.e. a functor which is
both a cartesian and a cocartesian fibration.
170 M. LAND

Definition 17.2. Given an adjunction, we can use the straightening theorem to obtain
a functor f : E0 → E1 , classified by the cocartesian fibration, and also to obtain a functor
g : E1 → E0 , classified by the underlying cartesian fibration of the adjunction. We refer to f
as the left adjoint and to g as the right adjoint of the adjunction.

Remark. We say that a functor f : C → D admits a right adjoint, if there exists an

adjunction E → ∆1 , whose associated functor is equivalent to f , i.e. where one specifies
equivalences C ' E0 and D ' E1 , such that the composite C ' E0 → E1 ' D is equivalent to
f . In general, an adjunction between two ∞-categories C and D hence consists of a bicartesian
fibration E → ∆1 together with specified equivalences C ' E0 and D ' E1 .

Remark. We directly want to show that an adjunction E → ∆1 gives rise to an equivalence

of anima
mapE1 (f (x), z) ' mapE0 (x, g(z))
if x is an object of E0 and z is an object of E1 . For this we consider the anima mapE (x, z).
Choosing a cartesian lift of the unique map 0 → 1 with target z, Corollary 10.12 shows that
there is a fibre sequence
mapE0 (x, g(z)) → mapE (x, z) → map∆1 (0, 1) ' ∗
so that the first map is a homotopy equivalence. Likewise, choosing a cocartesian lift with
domain x, we obtain a fibre sequence
mapE1 (f (x), z) → mapE (x, z) → map∆1 (0, 1) ' ∗
so we find the desired equivalence as the zig-zag
' '
mapE1 (f (x), z) → mapE (x, z) ← mapE0 (x, g(z)).

We now want to promote this to a natural equivalence of functors Eop

0 × E1 → An.

Proposition 17.3. Let E → ∆1 be an adjunction, and f : E0 → E1 and g : E1 → E0 the

associated functors. Then there is a natural equivalence of functors
mapE0 (−, g(−)) ' mapE1 (f (−), −).
Proof. We claim that both functors are equivalent to the composite
Eop
0 × E1 → E × E → An
op

where the latter is the bivariant mapping anima functor for the ∞-category E. We notice
that there is a natural transformation of functors τg : i0 ◦ g → i1 and τf : i0 → i1 ◦ f which
picks out the required (co)cartesian maps: One considers the diagrams

E0 × {0} E E1 × {1} E
τf τg

E0 × ∆ 1 ∆1 E1 × ∆ 1 ∆1

where the dashed arrows exist because EEi → ∆Ei is again bicartesian so that we find such
lifts as desired. We then consider the composite
Eop op
0 × E1 × ∆ → E0 × E → E × E → An
1 op
 INTRODUCTION TO INFINITY-CATEGORIES 171

which is a natural transformation from mapE (−, g(−)) to mapE (−, −) which is pointwise an
equivalence, since τg is pointwise a cartesian edge. Using that E0 → E is fully faithful (∆1
has trivial anima of self-maps) and natural transformations which are pointwise equivalences
are themselves equivalences, Corollary 6.2, we conclude that τg induces a natural equivalence
mapE0 (−, g(−)) ' mapE (−, −) : Eop
0 × E1 → An.
Likewise, τf induces a natural equivalence
mapE1 (f (−), −) ' mapE (−, −) : Eop
0 × E1 → An
which shows the proposition. 
Next we wish to define unit and counit transformations associated to an adjunction. As in
the proof of Proposition 17.3 we consider the transformations τf and τg and observe that the
two maps
τg
E1 × ∆ 1 E
g×id τf

E0 × ∆ 1
agree when restricted to E1 × {0}: By definition, the horizontal map is the functor i0 ◦ g,
whereas the composite is the composite of g with the inclusion i0 .
Hence these two maps combine to a map E1 ×Λ20 → E such that the restriction to E1 ×∆{0,1}
is given by τf ◦ (g × id), and the restriction to E1 × ∆{0,2} is given by τg . This gives the top
horizontal map in the diagram
E1 × Λ20 E

E1 × ∆ 2 ∆1
and the lower horizontal map is given by the composite E1 × ∆2 → ∆2 → ∆1 in which the
latter map sends 0 to 0 and both 1 and 2 to 1. To see that the diagram commutes, it suffices
to recall that both p ◦ τf and p ◦ τg are the projections. We now find that this lifting problem
can be solved, since the composite
E1 × ∆{0,1} → E1 × Λ20 → E
is pointwise cocartesian. For an object z of E1 , the resulting 2-simplex of E is given by
f (g(z))
cocart

cart
g(z) z

Finally, we see that the restriction E1 × ∆{1,2} → E1 × ∆2 → E factors through the inclusion
E1 → E by construction.
A similar constructions provides a functor E0 × ∆2 → E0 which can be depicted as follows
g(f (x))
cart

cocart
x f (x)
172 M. LAND

Definition 17.4. We refer to the resulting functor ε : E1 × ∆1 → E1 as the counit of the

adjunction. Dually, we refer to the resulting functor η : E0 × ∆1 → E0 as the unit of the
adjunction.

Remark. We notice that in an adjunction E → ∆1 , the unique morphism 0 → 1 has both a

cartesian and a cocartesian lift. However, in general a cartesian lift need not be cocartesian
and vice versa. In fact, this can be controlled very nicely.

Proposition 17.5. Let p : E → ∆1 be an adjunction, with left adjoint f : E0 → E1 and right

adjoint g : E1 → E0 . Then p-cartesian edges are p-cocartesian if and only if g is fully faithful.
Conversely, p-cocartesian edges are p-cartesian if and only if f is fully faithful.
In particular, f and g are mutually inverse equivalences if and only if the set of p-cartesian
edges equals the set of p-cocartesian edges.
Proof. We first show that cartesian edges are cocartesian if and only if the counit is an
equivalence. Consider the 2-simplex as above
f (g(z))
cocart

cart
g(z) z
and assume that cartesian edges are cocartesian. By the dual version of Lemma 10.5 we
find that the counit map f (g(z)) → z also cocartesian and thus is an equivalence for every z
because its image is invertible in ∆1 . Hence the counit is a natural equivalence. Conversely,
if the counit is an equivalence, then it is also cocartesian so that the cartesian edge g(z) → z
is also cocartesian as a composition of such.
Now we show that g is fully faithful if and only if the counit is an equivalence. By con-
struction, we have that the diagram
mapE1 (z, w) mapE0 (g(z), g(w))

mapE1 (f (g(z)), w) mapE (g(z), w)

commutes. The lower horizontal map and the right vertical map are equivalences because they
are induced by post composition with a cartesian edge, respectively with pre composition with
a cocartesian edge. Hence g is fully faithful if and only if the counit is an equivalence.
The argument for the unit is similar and the in particular follows since f and g are mutually
inverse if and only if both are fully faithful. 
Definition 17.6. Let f : C → D be a functor between ∞-categories. Then f is said to admit
a right adjoint if the cocartesian fibration E → ∆1 associated to f is cartesian. Conversely,
g : D → C is said to admit a left adjoint if the cartesian fibration associated to g is cocartesian.

The following is a useful way of constructing functors:

Proposition 17.7. Let f : C → D be a functor between ∞-categories. Specify for each object
x of D an object gx of C and maps f (gx) → x in D. If the induced composite
f ε
mapC (z, gx) → mapD (f (z), f (gx)) → mapD (f (z), x)
 INTRODUCTION TO INFINITY-CATEGORIES 173

is an equivalence, then there exists a functor g : D → C sending x to gx which is right adjoint to

f . Furthermore, the counit of the adjunction is then equivalent to the chosen map f (gx) → x.
Proof. Let p : E → ∆1 be the cocartesian fibration associated to the functor f . We aim to
show that p is cartesian. In other words, we must specify for each object x of E1 ' D a
p-cartesian morphism over the unique non-identity morphism of ∆1 . We consider the object
gx of E0 ' C and choose a p-cocartesian morphism gx → f (gx). Composing this with
the specified morphism f (gx) → x we obtain a map gx → x over the unique non-identity
morphism of ∆1 , and we wish to show that this map is p-cartesian, as then the first part of
the proposition follows. We recall that a morphism α : u → v in E over 0 → 1 is p-cartesian
if and only if the map
α
mapE (w, u) →∗ mapE (w, v)
is a homotopy equivalence for all w ∈ E0 (Exercise). In other words, we must show that the
composite
mapE (z, gx) → mapE (z, f (gx)) → mapE (z, x)
is an equivalence for all z ∈ E0 . For this we choose a p-cocartesian edge z → f (z) and consider
the diagram
mapE (z, gx) mapE (z, f (gx)) mapE (z, x)
' '
f
mapE (z, gx) mapE (f (z), f (gx)) mapE (f (z), x)
in which the middle and right vertical maps are equivalences since z → f (z) is p-cocartesian
and x and f (gx) are objects of E1 . The lower composite is an equivalence by assumption,
and hence so is the upper composite. We conclude that the above constructed map gx → x
is p-cartesian.
It remains to prove the claim about the counit of the adjunction, but this follows from the
construction: We need to see that there is a 2-simplex in E
f (gx)
cocart ε
cart
gx x
and we have just verified that any composite is a cartesian edge gx → x. 
Remark. Likewise, specifying gx for each object x of D and maps x → f (gx) such that the
composite
mapC (gx, z) → mapD (f (gx), f (z)) → mapD (x, f (z))
is an equivalence for all z ∈ C gives a functor g which is left adjoin to f and the specified map
being equivalent to the unit of the adjunction.

Proposition 17.8. The association C 7→ C[W −1 ] where W consists of all morphisms of C

extends to a left adjoint of the inclusion An → Cat∞ .
Proof. By Proposition 17.7 and its variant for the existence of left adjoints it suffices to specify
for each ∞-category C the ∞-groupoid C[W −1 ], together with the map C → C[W −1 ] and show
that the for each ∞-groupoid X, the composite
mapAn (C[W −1 ], X) → mapCat∞ (C[W −1 ], X) → mapCat∞ (C, X)
174 M. LAND

is an equivalence. As we have already seen, the functor An → Cat∞ is fully faithful, so the
first map is an equivalence. Then we recall that for arbitrary ∞-categories C and D we have
mapCat∞ (C, D) ' Fun(C, D)' .
However, if D is an ∞-groupoid, so is Fun(C, D) so that we it suffices to show that the map
Fun(C[W −1 ], X) ' Fun(C, X)
is an equivalence for all ∞-groupoids. The universal property of the localization shows that the
former is canonically equivalent to Fun' (C, X), the full subcategory of Fun(C, X) on functors
sending all morphisms to equivalences. Since every morphism in X is an equivalence, the
inclusion Fun' (C, X) → Fun(C, X) is an equivalence, so the proposition is shown. 
The next proposition makes sure that if we already have a candidate right adjoint functor,
then it is really a right adjoint.
Proposition 17.9. Let f : C → D and g : D → C be functors and let ε : f g → id be a natural
transformation such that the induced map
map(x, g(y)) → map(f (x), f (g(y))) → map(f (x), y)
is an equivalence for all x and y. Then g is right adjoint to f .
Proof. By Proposition 17.7 there exists a functor g 0 which is right adjoint to f , which is
pointwise equivalent to g and such that the counit map is equivalent to the chosen map. We
now need to show that g 0 is equivalent to g. We first construct a natural transformation g → g 0
as follows: We recall that the functor Fun(D, C) → Fun(D, P(C)), given by postcomposition
with the Yoneda functor, is fully faithful. It hence suffices to construct an equivalence between
the images of g and g 0 . These images are given by the two functors
(
mapD (f (−), d) for g 0
d 7→
mapC (−, g(d)) for g
We find that these two functors are equivalent by Exercise 1, and hence deduce that g and g 0
are also equivalent. 
Exercise 1. Let f : C → D and let g : D → C be functors. Suppose that ε : f g → id is a
natural transformation. Show that the map
map(x, g(y)) → map(f (x), f (g(y)) → map(f (x), y)
is natural in x.

Corollary 17.10. Let f : C → D and g : D → C be functors. Then g is right adjoint to f if

and only if there exist unit and counit transformations that satisfy the snake identities.
Proof. The fact that the snake identities are satisfied for an adjunction is left as an exercise.
The converse follows from the previous proposition, as satisfying the snake identities implies
that the canonical map
map(x, g(y)) → map(f (x), f (g(y))) → map(f (x), y)
is an equivalence: An inverse is given by the composite
map(f x, y) → map(gf x, gy) → map(x, gy).

 INTRODUCTION TO INFINITY-CATEGORIES 175

Having this we can prove the following result about the compatibility of adjunctions with
Dwyer–Kan localizations and functor categories.
Proposition 17.11. Let f : C → D and g : D → C. Suppose C is equipped with a set S of
morphisms and D is equipped with a set T of morphisms. If f (S) ⊆ T and g(T ) ⊆ S, then
there are induced functors F : C[S −1 ] → D[T −1 ] and G : D[T −1 ] → C[S −1 ]. If furthermore f
is left adjoint to g, then F is left adjoint to G.
Proof. By Corollary 17.10 it suffices to construct transformations F G → id and id → GF
satsifying the snake identities. We construct first the map id → GF , which is a 1-simplex of
Fun(C[S −1 ], C[S −1 ]) from id to the composition GF . We know that the restriction functor
`∗ : Fun(C[S −1 ], C[S −1 ]) → Fun(C, C[S −1 ])
is fully faithful, so it suffices to construct the desired 1-simplex in the latter category, namely
from ` to GF ◦ `. There is also a functor
Fun(C, C) → Fun(C, C[S −1 ])
given by postcomposition with the localization map ` : C → C[S −1 ]. The unit of the adjunction
η is a 1-simplex from id to gf in the former category, hence this functor gives rise to a 1-
simplex in Fun(C, C[S −1 ]) from ` to ` ◦ gf . By definition of G and F , there is an equivalence
` ◦ gf ' GF ◦ `. Hence we find a transformation from ` to GF ◦ ` as needed. Likewise, one
obtains the counit transformation F G → id. To see that the snake identities are fulfilled, we
consider the 2-simplex whitnessing the snake identity for f in the diagram
'
∆2 mapFun(C,D) (f, f ) mapFun(C,D[T −1 ]) (`D ◦ f, `D ◦ f ) mapFun(C[S −1 ],D[T −1 ]) (F, F )

Then we observe that the resulting 2-simplex of map(F, F ) whitnesses the sake identity for
F . The argument for G is similar. 
Proposition 17.12. Let f : C → D be a functor and let K be a simplicial set and E an
auxiliary ∞-category. If f admits a right, resp. a left adjoint, then so does f∗ : Fun(K, C) →
Fun(K, D). If f admits a right, resp. a left adjoint, then f ∗ : Fun(D, E) → Fun(C, E) admits
a left, resp. a right adjoint.
Proof. Let g be a right adjoint of f . We prove that g∗ is right adjoint to f∗ and that g ∗ is
left adjoint to f ∗ . The other cases are similar. Let ε : f g → id be the counit and η : id → gf
be the unit of the adjunction, viewed as morphisms
ε η
∆1 Fun(D, D) ∆1 Fun(C, C)

we can then postcompose with the canonical functor Fun(C, D) → Fun(CK , DK ) and obtain
new transformations ε∗ : f∗ g∗ → id∗ and η∗ : id∗ → g∗ f∗ . These are easily checked to satisfy
the snake identities, and so form an adjunction. Likewise, one can compose with the functor
Fun(C, D) → Fun(ED , EC ) and obtain transformations ε∗ : (f g)∗ = g ∗ f ∗ → id∗ and η ∗ : id∗ →
(gf )∗ = f ∗ g ∗ . Again, these satisfy the snake identities, and thus g ∗ is left adjoint to f ∗ . 
Corollary 17.13. The functor Cat∞ → An given by taking the maximal sub-∞-groupoid is
a right adjoint of the inclusion An → Cat∞ .
Proof. The inclusion functor Kan → Cat1∞ has a right adjoint given by Cat1∞ → Kan sending
C to C' . The inclusion sends homotopy equivalences to Joyal equivalences, and the maximal
176 M. LAND

subgroupoid functor sends Joyal equivalences to homotopy equivalences. Hence we may apply
Proposition 17.11. 
Remark. One can of course also prove Corollary 17.13 using Proposition 17.9. Then one
has to show that for an ∞-groupoid X and an ∞-category C, the canonical map
Fun(X, C' ) → Fun(X, C)'
is a homotopy equivalence.

Proposition 17.14. Let ` : C → D be a Dwyer-Kan localization. Suppose that ` admits a

right adjoint r. Then r is fully faithful.
Proof. We claim that the there is a commutative diagram
r
D C
yD yC
`∗
P(D) P(C)
for which we have to see that there is a natural equivalence between the functors
(
c 7→ mapC (c, r(d))
d 7→
c 7→ mapD (`(c), d)
which is a consequence of the fact that r is right adjoint to `. Since ` is a localization, the
functor `∗ is fully faithful. The proposition thus follows from the fact that the Yoneda functors
are fully faithful. 
Definition 17.15. A Dwyer-Kan localization which admits a right adjoint is called a
Bousfield localization.

Corollary 17.16. Let C be a locally small ∞-category and let C → D be a Bousfield

localization. Then D is locally small.
Proof. By Proposition 17.14 D identifies with a full subcategory of C and is thus also locally
small. 
Proposition 17.17. Let f : C → D be any functor which has a fully faithful right adjoint r.
Then f is a Bousfield localization.
Proof. By Proposition 17.14 it suffices to show that f is a Dwyer-Kan localization. Let E be
an auxiliary ∞-category. We need to show that the functor f ∗ : Fun(D, E) → Fun(C, E) is
fully faithful, and characterize the essental image. To see that f ∗ is fully faithful, we observe
that r∗ is a left adjoint to f ∗ by ??. Furthermore, by assumption f r → id is an equivalence.
From the construction, we find that also r∗ f ∗ → id is an equivalence, so that f ∗ is fully
faithful. It remains to show that f is a Dwyer–Kan localization. If this is the case, then it
must be a Dwyer–Kan localization along the set of morphisms S which are those morphisms
which become equivalences after applying f . We thus need to consider a functor a : C → E
with the property that it sends f -equivalences to equivalences and show that this equivalent
to a composite C → D → E for some functor b : D → E. We claim that b = ar works: We have
to show that there is an equivalence between a and bf = arf . The unit of the adjunction is a
map id → rf , which we claim to consist of f -equivalences: applying f to the map x → rf (x)
 INTRODUCTION TO INFINITY-CATEGORIES 177

gives a map f x → f rf (x), we may postcompose with the counit to obtain the composite
f x → f rf (x) → f (x). The snake identity says that the composite is an equivalence, and
the fact that r is fully faithful says that the counit is an equivalence. Hence the unit is an
f -equivalence. It follows then from the fact that a sends f -equivalences to equivalences, that
the canonical map a → arf is an equivalence as needed.

Example 17.18. Consider the functor Cat∞ → An given by inverting all morphisms. It
has a fully faithful right adjoint given by the inclusion An → Cat∞ and is hence a Bousfield
localization.

Proposition 17.19. Let C be an ∞-category and K a simplicial set. If C admits K-indexed

colimits, then the formation of such assembles into a functor colimK : Fun(K, C) → C which is
left adjoint to the constant functor. Conversely, if the constant functor const : C → Fun(K, C)
admits a left adjoint F , then F (p) is a colimit of p for any diagram p : K → C.
Proof. We employ Proposition 17.7, rather the remark thereafter, for the existence of a left
adjoint. Thus, first we have to specify for each object p of Fun(K, C) an object t of C and
a map p → const(t). As object we choose a colimit colimK p. We thus need to construct
a morphism p → const colimK p in Fun(K, C). By adjunction, such a morphism is a map
K × ∆1 → C. Choosing a colimit cone p̄ : K  ∆0 → C, we can restrict it along the canonical
map K × ∆1 → K  ∆0 . By construction, restricted to K × {0} one obtains p, and restricted
to K × {1} one obtains the constant functor with value p̄(∞) as needed.
We then have to see that the composite as in ... is an equivalence. For this we consider
the commutative diagram, which shows, together with Proposition 15.12, that the map in
question is in fact an equivalence.
Conversely, assume that the constant functor admits a left adjoint F : Fun(K, C) → C
and consider a functor p : K → C. We wish to show that F (p) is a colimit of p. The
unit of the adjunction gives a map p → constF (p) in Fun(K, C). This map is adjoint to a
map K × ∆1 → C, and as before, one checks that this map factors through the projection
K × ∆1 → K  ∆0 . The resulting map p̄ : K  ∆0 → C is a cone over p, and it remains to
show that it is an initial cone. This is the case if and only if the canonical map
MapC (p̄, x) → MapC (p, x) ' mapCK (p, constx ) ' mapC (F (p), x)
is an equivalence. Again, we claim that the composite is equivalent to the canonical map
given by restriction along the inclusion {∞} → K  ∆0 , which is an equivalence. 
Remark. Dually, the same statement holds for limits: It is then a right adjoint to the
constant functor.

Proposition 17.20. Let C be a cocomplete, resp. a complete ∞-category and K a simplicial

set. Then Fun(K, C) is again cocomplete, resp. complete.
Proof. We claim that the composite
Fun(L, Fun(K, C)) ' Fun(K, Fun(L, C)) → Fun(K, C),
where the latter functor is post composition with the colimit functor colimL : Fun(L, C) → C,
is left adjoint to the constant functor. This follows immediately from the fact that (colimL )∗
is right adjoint to const∗ by Proposition 17.12. 
178 M. LAND

Corollary 17.21. The constant functor C → Fun(K, C) preserves limits and colimits.
Proof. We show this only in the case where C admits all small limits, resp. colimits. We wish
to show that if p : L → C is a diagram with colimit x, then the constant functor with value x
is a colimit of the diagram L → C → Fun(K, C).
We first observe that for any functor ϕ : D → E and any simplicial set S, the diagram
ϕ∗
Fun(S, D) Fun(S, E)
const const
ϕ
D E
commutes. Applying this for the functor colimL : Fun(L, C) → C, and the simplicial set K,
we obtain the following commutative diagram

Fun(L, Fun(K, C))

colimL
'
(colimL )∗
Fun(K, Fun(L, C)) Fun(K, C)
const const
colimL
Fun(L, C) C

where the left vertical composite is given by post composition with the constant functor
C → Fun(K, C). Notice that the upper triangle commutes by the proof of Proposition 17.20.
The commutativity of the diagram then implies the statement of the corollary. 
Proposition 17.22. Let f : C → D be a left adjoint. Then f preserves colimits. Likewise,
right adjoints preserve limits.
Proof. We prove that left adjoints preserve colimits. The other case follows by passing to
opposite categories. So let F : K → C be a diagram and F̄ a colimit cone. We wish to show
that f F̄ is a colimit cone of f F . For this we consider an object z of D and need to show that
the canonical map
MapD (f F̄ , z) → MapD (f F, z)
is an equivalence. To prove this, we claim that for any functor G : L → C, there is a canonical
equivalence
MapD (f G, z) ' MapC (G, gz)
where g is the right adjoint of f . Taking this claim for granted for the moment, we then
consider the commutative diagram

MapD (f F̄ , z) MapD (f F, z)
' '
'
MapC (F̄ , gz) MapC (F, gz)

in which the vertical maps are equivalences by the claim, and where the lower horizontal map
is an equivalence by the assumption that F̄ is a colimit cone. Hence also the upper map is
an equivalence as needed.
 INTRODUCTION TO INFINITY-CATEGORIES 179

It remains to prove the claim. For this we consider the following chain of equivalences
MapD (f G, z) ' mapDK (f∗ (G), constz ) by Proposition 15.12
' mapCK (G, g∗ (constz )) by Proposition 17.12
' mapCK (G, constgz ) by inspection
' MapC (G, gz) by Proposition 15.12
where the inspection is to observe that g∗ (constz ) = constgz . 

Finally, we need the following proposition.

Proposition 17.23. Consider a pullback diagram of ∞-categories
f
C C0
q p
g
D D0
in which the map p is an isofibration. Suppose that p preserves colimits and let F : K → C be
a diagram.
(1) A cone F̄ : K  ∆0 → C is a colimit cone if its image under f and q is a colimit cone.
(2) If C0 and D are cocomplete and g preserves colimits, then C is also cocomplete. Fur-
thermore f and q preserve colimits.
Proof. The first thing we observe is that for any object x of C, there is a homotopy cartesian
diagram
MapC (F, x) MapC0 (f F, f x)

MapD (qF, qx) MapD0 (pf F, pf x)

This follows from Proposition 15.12 and the fact that the mapping anima in a pullback
are given by the pullback of the mapping anima. Note that we use here that applying the
functor Fun(K, −) to the above diagram again gives a pullback diagram where one leg is an
isofibration.
We now wish to analyze whether F̄ is a colimit cone. For this we consider the above squares
for F̄ and F , and obtain a canonical commutative cube. The assumption that f F̄ and q F̄
are colimit cones implies that the comparison maps are equivalences on the left lower and
right upper corner. Using that p preserves colimits, we find that pf F̄ is also a colimit cone,
so that the comparison map is also an equivalence there. We find that the comparison map
is an equivalence also in the upper left corner, so that F̄ is a colimit cone. This proves (1).
To prove (2) it will suffice to show that any diagram F : K → C admits a cone F̄ whose
image in D and C0 is a colimit cone. Then we apply (1) to see that F̄ is a colimit cone, and
by construction q and f send F̄ again to a colimit cone.
To do this, we consider the composite F1 : K → C → C0 and choose a colimit cone F̄1 : K 
∆ → C0 . Likewise, we consider the composite F2 : K → C → D and choose a colimit cone
0

F̄2 : K  ∆0 → D. The images g F̄2 and pF̄1 are then also colimit cones by the assumption that
both p and g preserve colimits. Hence there is an equivalence τ between these two cones, say
180 M. LAND

τ is an equivalence from pF̄1 to g F̄2 . We thus obtain a lifting problem

F̄1
∆0 C0f F/
τ̂

τ
∆1 D0pf F/
which admits a solution since the right vertical map is an isofibration, since p is an isofibration.
Furthermore, the dashed arrow is again an equivalence in C0 and hence τ̂ (1) is another cocone
of f F . Unravelling the definitions, we obtain a commutative diagram
τ̂ (1)
K  ∆0 C0
F̄2 p
g
D D0
so this gives a unique map K  ∆0 → C which is a cone over F . We are now in the situation
that we may apply (1) and deduce the proposition. 

18. An adjoint functor theorem

The goal of this section is to prove a general adjoint functor theorem, following the argument
given in [NRC19].
Definition 18.1. A full subcategory C0 ⊆ C of an ∞-category C is called colimit dense, if
every object of C can be written as a colimit of a diagram p : K → C0 ⊆ C.

Theorem 18.2. Let C be a locally small ∞-category which is cocomplete and contains an
essentially small, colimit dense full subcategory C0 ⊆ C. Let D be a locally small ∞-category
and let F : C → D be a functor. Then F admits a right adjoint if and only if F preserves
colimits.
Remark. Particular examples of locally small ∞-categories which admit a small colimit
dense full subcategory are accessible ∞-categories. An ∞-category is called accessible if it is κ-
accessible for some regular cardinal κ. A κ-accessible ∞-category is a locally small ∞-category
C which admits κ-filtered colimits, and which contains an essentially small subcategory C0
such that every object of C0 is κ-compact and such that every object of C is a κ-filtered
colimit of objects in C0 . An accessible ∞-category which is in addition cocomplete is called
presentable. The above theorem can hence be applied to functors between presentable ∞-
categories, so that any colimit preserving functor between presentable categories admits a
right adjoint.

Corollary 18.3. Let C be a locally small ∞-category which is cocomplete and contains an
essentially small colimit dense full subcategory C0 ⊆ C. Then C is complete.
Proof. Let K be a simplicial set. Consider the functor const : C → Fun(K, C). It preserves
colimits, see Corollary 17.21, moreover Fun(K, C) is again locally small: this needs justifica-
tion, but see [Lur09, Example [Link]]. Hence, by Theorem 18.2, the constant functor admits
a right adjoint. By Proposition 17.19 this functor takes a diagram p : K → C to a limit of
p. 
 INTRODUCTION TO INFINITY-CATEGORIES 181

To prove the theorem we need some preliminaries. To obtain the right adjoint we will
employ the following criterion.
Proposition 18.4. Let F : C → D be a functor. Then F admits a right adjoint if and only
if for all objects d of D, the ∞-category C/d admits a terminal object.
Proof. Let d be an object of D and let consider a terminal object of C/d . This is given by a
pair (Gd, f ) where f is a morphism F Gd → d in D. We wish to use Proposition 17.7 to show
that F admits a right adjoint. We must thus consider the lower composite in the diagram

mapC/d ((c, α), (Gd, f )) mapD/d (α, f ) ∆0

α

f∗
mapC (c, Gd) mapD (F c, F Gd) mapD (F c, d)

We observe that the left square is the pullback of mapping spaces induced from the pullback

C/d D/d

C D
of ∞-categories. Furthermore, the right square is a homotopy pullback by Proposition 12.15.
Thus, the big square is a homotopy pullback as well, and the upper composite is an equivalence
by the assumption that (Gd, f ) is a terminal object of C/d . Hence the lower composite is also
an equivalence so that we conclude the proposition from Proposition 17.7. 
Remark. Likewise, a functor admits a left adjoint if and only if for all objects d of D, the
category Cd/ admits an initial object.

We thus need to find criteria that ensure that specific categories admit terminal objects.
For this we will make use of the notion of weakly terminal sets:
Definition 18.5. Let C be an ∞-category and S a (small) set of objects. S is said to be
weakly terminal, if for every object x of C, there exists an object s in S such that the anima
mapC (x, s) is not empty. An object t is called weakly terminal if the set {t} is a weakly
terminal set.

Lemma 18.6. Let C0 ⊆ C be an essentially small, full subcategory of a cocomplete category

which is colimit dense. Then C has a weakly terminal object.
Proof. Consider the functor C0 → C and pick a Joyal equivalence C0 ' C0 with C0 a small
simplicial set. As Joyal equivalences are cofinal by Corollary 16.8 and C is cocomplete, we
find that the functor C0 → C admits a colimit t. We claim that t is a weakly terminal object.
To see this, let x be another object of C. By assumption, there is a functor K → C0 such that
the colimit over the composite K → C0 → C is given by x. We obtain a canonical map x to t
on colimits. In particular, the anima of maps from x to t is not empty. 
Proposition 18.7. Let C be a locally small and cocomplete ∞-category and let S be a weakly
terminal set. Let C0 be the full subcategory spanned by S. Then C0 → C is cofinal.
182 M. LAND

Proof. By Theorem 16.16 it will suffice to show that for any object x of C, the slice (C0 )x/ is
weakly contractible. We will show that for any small simplicial set K, any functor K → (C0 )x/
factors through the inclusion K ? ∆0 which is contractible. It follows that (C0 )x/ is weakly
contractible as needed. So consider a functor K → (C0 )x/ and the composite
K → (C0 )x/ → Cx/ .
As C is cocomplete, so is Cx/ by Proposition 15.30, so we may choose a colimit cone of the
above functor
K (C0 )x/ Cx/

K ? ∆0
and consider µ(∞) : x → t. Pick an object s in S for which there exists a map t → s, and
pick such a map. Choosing a composite of x → t and t → s gives a 2-simplex σ : ∆2 → C,
which is adjoint to a map ∆1 → Cx/ . We then consider the following lifting problem
(µ,σ)
K ? ∆ 0 q∆ 0 ∆ 1 Cx/
µ0

K ? ∆1
which can be solved as the vertical map is inner anodyne by Lemma 4.22 and Cx/ is an ∞-
category. Restricting µ0 along the inclusion K ? ∆{1} → K ? ∆1 gives a functor K ? ∆0 → Cx/
which factors through (C0 )x/ : Since (C0 )x/ ⊆ Cx/ is a full subcategory, it suffices to see that
all objects of K ? ∆0 go to (C0 )x/ : On K it is true by assumption and on the cone point {∞},
by construction, one obtains the map x → s which is in (C0 )x/ , again by construction. Hence
the proposition is proven. 
Corollary 18.8. Let C be a locally small ∞-category which is cocomplete. Assume that
there exists a weakly terminal set. Then C admits a terminal object.
Proof. Let S be a weakly terminal set and consider the full subcategory C0 spanned by S.
By Proposition 18.7 the inclusion C0 → C is cofinal. Since C0 is small, the functor C0 → C
admits a colimit. From Corollary 16.9 we thus deduce that also the identity functor C → C
admits a colimit. Such a colimit is a terminal object, by Lemma 15.13. 
Proof of Theorem 18.2. The fact that left adjoints preserve colimits was dealt with in Propo-
sition 17.22. Let us therefore prove that F admits a right adjoint if it preserves colimits. By
Proposition 18.4, it suffices to show that for every object d of D, the slice C/d has a terminal
object.
We then observe that C/d is again locally small and cocomplete. The cocompleteness follows
from Proposition 17.23 because the functor C → D preserves colimits by assumption, and the
functor D/d → D preserves colimits by Proposition 15.29. To see that C/d is again locally
small, we calculate the mapping anima in terms of those in C, D and D/d . Those in C and
D are essentially small by assumption, and those in D/d are then also essentially small by
Proposition 12.15. Thus the pullback is also essentially small.
Hence by Corollary 18.8 it suffices to establish the existence of a weakly terminal object,
which we will deduce by means of Lemma 18.6. In other words, we have to show that C/d
 INTRODUCTION TO INFINITY-CATEGORIES 183

contains an essentially small full subcategory which is colimit dense. We claim that (C0 )/d is
such a subcategory. It is essentially small since D is locally small, and colimit dense since F
preserves colimits. 

Appendix A. Exercises
A.1. Introduction.
Exercise 2. Let h(CW) be the homotopy category of CW-complexes. Show that this
category does not have all pushouts. More concretely, show that the diagram
·2
∗ ←− S 1 −→ S 1
does not admit a pushout.

A.2. Section 1.
Exercise 3. Work out at least three of the following simplicial identities
(1) d∗i d∗j = d∗j−1 d∗i if i < j
(2) d∗i s∗j = s∗j−1 d∗i if i < j
(3) d∗i s∗j = id if i = j, j + 1
(4) d∗i s∗j = s∗j d∗i−1 if i > i + 1
(5) s∗i s∗j = s∗j+1 s∗i if i ≤ j.
Here, for any simplicial set X : ∆op → Set, we denote the map X(di ) by d∗i . Hint: Think
about what this means for the maps di and sj in ∆ and prove the corresponding identities
there.

Exercise 4. Show that every map in ∆ can be uniquely factored as a composition of si ’s

followed by a composition of dj ’s. Thus a simplicial set is equivalently described by a sequence
of sets Xn equipped with face and degeneracy maps satisfying the simplicial identities.

Exercise 5. Give examples of simplicial sets where the relation of Definition 1.10 of the
lecture, leading to π0∆ (X), is not symmetric and not transitive.

Exercise 6. Show that every simplex x ∈ Xn is of the form α∗ (y) for a surjection α : [m] →
[n] and a non-degenerate n-simplex y, and show that the pair (α, y) is uniquely determined
by x.

Exercise 7. Show that the category Set is bicomplete. Hint: General colimits are con-
structed as quotients of disjoint unions, and general limits are constructed as subsets of
products.

Exercise 8. Let F : I → C be a functor. Show that a colimit of F can equivalently be

described as an initial cocone over F , and that a limit of F can be equivalently described as
a terminal cone over F .

Exercise 9. Calculate the limit and colimit of a simplicial set X : ∆op → Set.

Exercise 10. Show that the datum of an adjunction is equivalent to the datum of a pair of
functors (F, G) as above together with natural transformations ε : F G → id and η : GF → id
184 M. LAND

satisfying the snake identities, that is, the obvious composites

F (X) → F (GF (X)) ∼ = F G(F X) → F (X)
and
G(X) → GF (G(X)) ∼ = G(F G(X)) → G(X)
are the identity of F (X) and G(X) respectively.

Exercise 11. Show that a functor F : C → D admits a right adjoint if you can speficy
objects Gy for all y ∈ D and maps εy : F Gy → y, which have the property that the induced
map on hom-sets
F εy
HomC (x, Gy) −→ HomD (F x, F Gy) −→ HomD (F x, y)
is a bijection. There is an obvious dual notion which shows that F admits a left adjoint if
one can specify objects Gy for all y ∈ D and maps ηy : y → F Gy which make the induced
map on hom-sets
F ηy
HomD (Gy, x) −→ HomC (F Gy, F x) → HomC (y, F x)
a bijection.

Exercise 12. Prove that if a simplicial set X has at most n-dimensional non-degenerate
simplices, and Y has at most m-dimensional non-degenerate simplices, then their product
X × Y has at most (n + m)-dimensional non-degenerate simplices.

Exercise 13. Show that for every simplicial set X, there is a canonical bijection π0∆ (X) ∼
=
π0 (|X|).

Exercise 14. Show that there are inclusions I n ⊆ Λnj provided 0 < j < n or n ≥ 3, and
Λnj ⊆ ∂∆n ⊆ ∆n for all n ≥ 0.

Exercise 15. Let I be a category with an initial object i and let J be a category with a
terminal object t. Show that a limit of a functor F : I → C is given by F (i) (together with
the canonical maps F (i) → F (x) for all x ∈ I). Similarly, show that a colimit of a functor
G : J → C is given by G(t) (together with its maps G(x) → G(t) for all x ∈ J).

Exercise 16. Show that for every n ≥ 0, there is a pushout

∂∆n
`
skn−1 (X)
Jn

∆n
`
skn (X)
Jn

where Jn is the set of non-degenerate n-simplices. Furthermore X ∼

= colimn skn (X).

Exercise 17. Show that the following simplicial sets are not nerves of categories:
(1) ∂∆n for n ≥ 2,
(2) Λnj for n ≥ 2 and 0 ≤ j ≤ n,
(3) I n for n ≥ 2.
 INTRODUCTION TO INFINITY-CATEGORIES 185

Exercise 18. Suppose X is a Kan complex. Show that for all n ≥ 0, the simplicial set
coskn (X) is again a Kan complex. Prove that the canonical map X → coskn (X) induces a
bijection
πk∆ (X) → πk∆ (coskn (X))
for k < n and that πk∆ (coskn (X)) = 0 for k ≥ n.

Exercise 19. Show that a natural transformation between functors f, g : C → D induces

a homotopy between N(f ), N(g) : N(C) → N(D). Use this to show that conjugation with
an element determines a self map of BG which is homotopic to the identity. What does
conjugation induce on π1 (BG)? Why does this not show that every group is abelian?

Exercise 20. Show that the nerve of a category C is 2-coskeletal, i.e. that the canonical
map N(C) → cosk2 (N(C)) is an isomorphism of simplicial sets.

Exercise 21. Let X be a simplicial set and let n ≤ m. Show that skn (skm (X)) =
skn (X) = skm (skn (X)). Deduce that coskn (coskm (X)) ∼ = coskm (coskn (X)). Is it also true
that skn (coskm (X)) ∼
= coskm (skn (X)) (if not provide a counter example)? Is there a preferred
map between these two simplicial sets?

Exercise 22. Let C be a category and X a simplicial set. Recall that X op is the simplicial
set with: Xnop = Xn and dop op
i : Xn → Xn−1 is given by dn−i , likewise that si = sn−i as a map
Xn → Xn+1 . Prove the following assertions:
(1) N(Cop ) ∼ = N(C)op ,
n op ∼
(2) (∆ ) = ∆n ,
(3) (Λni )op ∼= Λnn−i ,
(4) (∂∆ ) ∼
n op
= ∂∆n .

Exercise 23. Let G be a group and let BG be the category with one object and G as
endomorphisms of that object. Show that N(B(G)) has only one non-trivial homotopy group,
namely π1∆ (N(BG)) and that this group is canonically isomorphic to G.

A.3. Section 2.
Exercise 24. Let X be a composer and let f : x → y be a morphism in X. Show that f is
a composition of idx with f and of f with idy .

Exercise 25. Consider the map [0] → [n] in ∆ with image {0}. Show that this determines
a map 0 : ∆0 → ∂∆n . Calculate the simplicial homotopy sets πi∆ (∂∆n , 0) for i ≥ 1 and n ≥ 2.
Deduce that ∂∆n is not a Kan complex.

Exercise 26. Show that the following simplicial sets are not ∞-categories:
(1) ∂∆n for n ≥ 2,
(2) Λnj for n ≥ 3 and 0 ≤ j ≤ n,
(3) I n for n ≥ 2.

Exercise 27. Determine the homotopy category of the following simplicial sets:
(1) ∂∆n for n ≥ 1,
186 M. LAND

(2) Λnj for n ≥ 2 and 0 ≤ j ≤ n, and

(3) I n for n ≥ 0.

Exercise 28. Let f : X → Y be a map of simplicial sets. Prove or give a counter example
to the following statements.
(1) If f is a monomorphism, then hX → hY is fully faithful,
(2) If f is a degree-wise surjection, then hX → hY is surjective and full i.e. induces a
surjection on objects and on hom-sets.
(3) If f induces a surjection on 0- and 1-simplices, then hX → hY is surjective and full.

Exercise 29. Prove or disprove the following statement: For any two simplicial sets X and
Y , the canonical map h(X × Y ) → hX × hY is an isomorphism of categories.

Exercise 30. A category C is called connected if π0∆ (N(C)) consists only of one element.
Show that a groupoid G is connected if and only if for every two objects x, y ∈ G, the set
HomG (x, y) is non-empty. Show that a connected groupoid is equivalent to BG for a group
G. Show however, that the category of connected groupoids is not equivalent to the category
of groups.

Exercise 31. Let X be a topological space. Describe the category h(S(X)). Show that the
endomorphisms of each object form a group. Which group is it?

Exercise 32. Suppose X is a composer with the inner 3-horn extension property. Let
σ : ∆1 × ∆1 → X be a map such that
(1) σ|∆1 ×{0} = f,
(2) σ|∆1 ×{1} = g,
(3) σ|{0}×∆1 = idx , and
(4) σ|{0}×∆1 = idy ,
for morphisms f, g : x → y. Show that f ∼ g in the sense of Definition 2.4.

Exercise 33. Let X be a composer with the extension property for inner 3-horns. Show that
for any two composable morphisms f : x → y and g : y → z, the simplicial set CompX (f, g)
is connected, i.e. that π0∆ (CompX (f, g)) consists only of one element.

Exercise 34. Let X be a simplicial set and consider the canonical map X → N(hX).
(1) Show that this map factors through the canonical map X → cosk2 (X).
(2) Show that the induced map cosk2 (X) → N(hX) is an isomorphism if X is isomorphic
to the nerve of a category.
(3) Show that the map cosk2 (X) → N(hX) is in general not an isomorphism. Hint: Find
an X which is 2-coskeletal, but not the nerve of a category.
(4) Prove or disprove the following statement: The map cosk2 (X) → N(hX) is an iso-
morphism if and only if X is isomorphic to the nerve of a category.

Exercise 35. Let (V, ⊗, 1) be a monoidal category. Then the functor HomV (1, −) : V → Set
is lax monoidal. Is it monoidal? Can you find a condition on (V, ⊗, 1) which ensures that it
is?
 INTRODUCTION TO INFINITY-CATEGORIES 187

Exercise 36. Let C be a category with finite products and finite coproducts. We say that C is
pointed if the canonical map ∅ → ∗ from the inital to the terminal object is an isomorphism.
Show that the identity canonically refines to a lax monoidal functor (C, ×, ∗) → (C, q, ∅).
When is this functor monoidal? Furthermore, show that any functor F : C → D refines
canonically to a lax symmetric monoidal functor (C, q, ∅) → (D, q, ∅). When is it monoidal?

Exercise 37. The goal of this exercise is to show that any essentially surjective and fully
faithful functor F : C → D between ordinary categories is an equivalence.

(1) Show that F admits an adjoint G. Hint: Use Exercise 11.

(2) Show that G is itself fully faithful.
(3) Show that an adjoint pair (F, G) of fully faithful functors makes F an equivalence
with G an inverse.

Exercise 38. Let F : C → D be a functor with right adjoint G : D → C. Show that they are
mutually inverse if F is fully faithful and G is conservative. Here, conservativity means that
if f : x → y is a morphism in D and G(f ) is an isomorphism, then f is an isomorphism.

Exercise 39. Let F : C → D be left adjoint to G : D → C. Show that if G is lax monoidal,

then F canonically refines to an oplax monoidal functor. Vice versa, show that if F is oplax
monoidal, then G canonically refines to a lax monoidal functor.

Exercise 40. Show that the left adjoint of a monoidal adjunction is in fact monoidal. Recall
that a monoidal adjunction consists of lax monoidal functors F and G, which are witnessed
to be adjoint by a unit η and a counit ε where both η and ε are monoidal transformations.

Exercise 41. Suppose F is left adjoint to G, witnessed by a unit and counit (η, ε). Show that
if F is monoidal, the induced lax monoidal structure on G of Exercise 6.1 makes (F, G, η, ε)
a monoidal adjunction.

Exercise 42. Show that the coherent nerve functor N : Cat∆ → sSet commutes with co-
products. Show that C is not right adjoint to N. Does N have a right adjoint at all? Hint:
Does the ordinary nerve functor N : Cat → sSet have a right adjoint? How are these two
questions related?

Exercise 43. Show that if the coherent nerve N(C) of a simplicial category is isomorphic
to the nerve of an ordinary category, then the underlying category uC is isomorphic to the
homotopy category π(C). Make explicit the coherent nerve of the following simplicial category
Bsimp (G): There is only one object, and the simplicial set of endomorphisms of this object is
given by N(G), where G is a group (of a monoid if you wish). Deduce from the explicit analysis
that N(Bsimp (G)) is not isomorphic to the nerve of a category although uN(Bsimp (G)) ∼ =
π(N(Bsimp (G))).

Exercise 44. Prove or disprove the following statements:

(1) There exists a simplicial category C, whose coherent nerve N(C) is not an ∞-category,
(2) There exists a simplicial category C, whose coherent nerve N(C) is an ∞-category, but
not a Kan complex.
188 M. LAND

Exercise 45. Suppose that C is a simplicial category all whose hom simplicial sets are Kan
complexes. Work out what it means concretely that a morphism f : x → y in a simplicial
category is an equivalence. Rephrase the condition of weakly fully faithful functors via this.

Exercise 46. Show that there exists monomorphisms X → Y where both X and Y are
∞-categories, but X is not a sub-∞-category in the sense of Definition 2.73.

Exercise 47. Let Y → X be an inclusion of topological spaces. When is the induced map
S(Y ) → S(X) a subcategory? When is it a full subcategory?

Exercise 48. Let C0 ⊆ C be a full sub-∞-category and let D be an ∞-category. Show that
the functor category Fun(D, C0 ) is the full sub-∞-category of Fun(D, C) on those functors
f : D → C which factor through the inclusion C0 ⊆ C.

Exercise 49. Show that a simplicial set X is an ∞-category if and only if X op is an ∞-

category and likewise that X is a Kan complex if and only if X op is. Show that if X is an
∞-category then X is an ∞-groupoid if and only if X op is.

Exercise 50. Let C be a simplicial category. Show that the coherent nerve N(C) is iso-
morphic to the nerve of an ordinary category if and only if C is in the image of the functor
c : Cat → Cat∆ .

Exercise 51. Show that the composite

N h
Cat∆ → sSet → Cat

is isomorphic to the functor π : Cat∆ → Cat.

A.4. Section 3.

Exercise 52. Show the following assertions.

(1) Let C be an ordinary category and X a simplicial set. Then X is an ∞-category if
and only if every map X → N(C) is an inner fibration.
(2) A map f : X → Y is an inner fibration if and only if f op : X op → Y op is.
(3) A map f : X → Y is a left fibration if and only if the map f op : X op → Y op is a right
fibration.

Exercise 53. Let S ⊆ S 0 be sets of morphisms. Show that

(1) χR (S 0 ) ⊆ χR (S),
(2) S ⊆ χ(S), and
(3) χR (S) = χR (χ(S)).

Exercise 54. A category I is called filtered if every functor K → I from a finite category
K extends over the inclusion K → K . . Show that a poset (viewed as a category) is filtered
if and only if
(1) for every finite collection of objects X1 , . . . , Xn of I, there exists an object X of I
equipped with maps Xk → X for all k = 1, . . . , n.
 INTRODUCTION TO INFINITY-CATEGORIES 189

(2) Any two morphisms f, g : X → Y can be equalized, i.e. there exists a morphism
h : Y → Z such that hf = hg.

Exercise 55. Let I be a finite category.

(1) Show that I is filtered if it has a terminal object.
(2) Show that there are examples where I is filtered but does not have a terminal object.
(3) Show that I is a poset and filtered if and only if it has a terminal object.
In particular, notice that this shows that there are many filtered categories which are not
posets.

Exercise 56. Show that every simplicial set A with only finitely many non-degenerate
simplices is compact, i.e. that the canonical map
colim HomsSet (A, Xi ) → HomsSet (A, colim Xk )
i∈I i∈I
is an isomorphism, provided I is a filtered category.

Exercise 57. We call a set S semi-saturated if it is closed under pushouts, retracts and
countable compositions. Show that a semi-saturated set
(1) contains isomorphisms, if it contains the identity of an inital object ∅,
(2) is closed under finite coproducts, if it contains the identity of an initial object ∅,
(3) is closed under composition, i.e. if f : A → B and g : B → C are elements of S, then
so is gf : A → C, and
(4) is closed under countable coproducts if it is closed under finite coproducts,
` i.e. if
{fi : Ai → Bi }i∈I is a countable family of elements if S, then the map : Ai → Bi is
i∈I
an element of S as well.

Exercise 58. Show that a saturated set S in a category C contains all isomorphisms. Find
an example of a category C and a semi-saturated set S of morphisms in C which is non-empty
and does not contain all isomorphisms.

Exercise 59. Show the following assertions.

(1) The map ∅ → {∗} in Set generates the set of injections. What is χR (∅ → ∗)? Spell out
the factorization obtained by the small object argument for a general map f : M → N
of sets.
(2) The map {∗, ∗} → {∗} generates the class of surjections. What is χR ({∗, ∗} → ∗)?
Spell out the factorization obtained by the small object argument for a general map
f : M → N of sets.

Exercise 60. Consider the set S = {∂∆n → ∆n }n≥0 given by the boundary inclusions.
Show that χ(S) is given by all monomorphisms of simplicial sets.

Exercise 61. Show that J is not a compact simplicial set, i.e. that there are infinitely many
non-degenerate simplices in J.

Exercise 62. Show that if a morphism f : ∆1 → C in an ∞-category extends over the

inclusion ∆1 → J, then f is an equivalence.
190 M. LAND

Exercise 63. Fill in the missing steps of Lemma 3.30. More precisely show the following
assertions:
(1) The horn inclusion Λnj → ∆n for 0 ≤ j < n is a retract of the pushout product map
∆n × {0} qΛnj ×{0} Λnj × ∆1 → ∆n × ∆1 .
(2) The pushout product map
∂∆n × ∆1 ∪ ∆n × {0} → ∆n × ∆1
is left anodyne.

Exercise 64. Show that a trivial fibration f : X → Y between Kan complexes induces
an isomorphism in the category π(Kan). Hint: Show that a trivial fibration between Kan
complexes is a homotopy equivalence.

Exercise 65. Show that if f : y → z is an equivalence, then

f
mapC (x, y) ' mapC (x, y) × ∆0 → mapC (x, y) × mapC (y, z) → mapC (x, z)
is a homotopy equivalence.

Exercise 66. Show that that composition as defined in the lecture is associative up to
homotopy, i.e. that composition in an ∞-category determines a category enriched in π(Kan),
the homotopy category of Kan complexes. Hint: Consider the diagram
?
∆{0,1} ∪ ∆{1,2} ∪ ∆{2,3} ∆{0,1,2} ∪ ∆{2,3} ∆{0,2} ∪ ∆{2,3}
? ? ?

?
∆{0,1} ∪ ∆{1,2,3} ∆3 ∆{0,2,3}

?
∆{0,1} ∪ ∆{1,3} ∆{0,1,3} ∆{0,3}
and show that all maps labelled with a ? are inner anodyne.

Exercise 67. Let f : C → D be a functor between ∞-categories. Let a : x → x0 and b : y → y 0

be morphisms in C. Then there is a homotopy commutative diagram of Kan complexes

mapC (x0 , y) mapD (f x0 , f y)

mapC (x0 , y 0 ) mapD (f x0 , f y 0 )

mapC (x, y 0 ) mapD (f x, f y 0 )

induced by precomposition with a, respectively f a and postcomposition with b, respectively
f b.
Deduce that there is a canonical functor F from Cat1∞ (the 1-category of ∞-categories) to
Catπ(Kan) , the category of categories enriched in the homotopy category of Kan complexes,
 INTRODUCTION TO INFINITY-CATEGORIES 191

where F (C) has objects the same as C and the hom object from x to y is given by the image
of mapC (x, y) in π(Kan).

A.5. Section 4.
Exercise 68. Show that [n] ? [m] = [n + m + 1]. Further show that C ? [0] = C. and
[0] ? C = C/ .

Exercise 69. For categories C and D we have that N(C) ? N(D) ∼

= N(C ? D). In particular,
there is a canonical isomorphism ∆i ? ∆j ∼
= ∆i+1+j .

Exercise 70. The functors X ? − and − ? X as functors sSet → sSet preserve pushouts.
Find an example of a colimit that is not preserved by X ? −.

Exercise 71. The slice/join adjunction induces a bijection of lifting problems between
diagrams of the kind
S Xϕ/

T Xϕi/ ×Yf ϕi/ Yf ϕ/

and diagrams of the kind

ϕ

B A ? T qA?S B ? S X

B?T Y

Exercise 72. Show that

Λnj ? ∆m ∪ ∆n ? ∂∆m = Λn+1+m
j

and that
∂∆m ? ∆n ∪ ∆m ? Λnj = Λn+1+m
m+1+j

For this, determine explicitly the following sub simplicial sets of ∆n+1+m :
(1) ∂∆n ? ∆m ,
(2) Λnj ? ∆m ,
(3) ∆m ? ∂∆n ,
(4) ∆m ? Λnj .

Exercise 73. For an object x in an ∞-category C, show that the canonical map Cx/ → C is
a left fibration and that C/x → C is a right fibration.

Exercise 74. Show that for an object x of a general simplicial set X, the canonical map
Xx/ → X is not in general a left fibration.
192 M. LAND

Exercise 75. Show that an inner fibration f : X → Y is inner anodyne if and only if it is
an isomorphism.

Exercise 76. Show that ∆0 → J is not inner anodyne.

Exercise 77. Show that the intersection of left and right anodyne maps strictly contains
the inner anodyne maps.

Exercise 78. Show that a functor F : C → D between ∞-categories is conservative if and

only if the induced functor hF : hC → hD between the homotopy categories is conservative.
Furthermore, show that the canonical functor C → N(hC) is conservative.

Exercise 79. Show that a functor p : C → D is conservative if and only if the following
diagram is a pullback.
C' C

D' D.

Exercise 80. An inner fibration C → D between ∞-categories is an isofibration if and only

if the induced functor N(hC) → N(hD) is an isofibration.

Exercise 81. A functor C → D between ∞-categories is an isofibration if and only if

Cop → Dop is an isofibration.

Exercise 82. Let C0 ⊆ C be a full subcategory. Show that the inclusion C0 → C is an

isofibration if C0 is closed under equivalences in C, i.e. that if x ∈ C0 and y ∈ C is equivalent
to x, then y is also in C0 .

Exercise 83. Show that if f : x → y is an equivalence, then the maps constructed in the
last lecture C/x → C/y and Cy/ → Cx/ are Joyal equivalences.

Exercise 84. Show that there exists a functor f : C → D between ∞-categories which is
conservative, but does not satisfy the RLP wrt ∆1 → J. This might be a hard one. Hint:
Consider the map J → S(|J|); as a map between Kan complexes it is clearly conservative.
The idea now should be that given a 1-simplex in S(|J|) there should be more than extension
to J. choose two distinct such extensions. They provide a commutative diagram

∆1 J

J S(|J|)

If there exists a lift in this extension problem, then the map J → J restricts to the identity
on ∆1 and thus must be the identity of J. This would imply that for any map ∆1 → S(|J|)
there is a unique extension to J. This is simply not correct.

A.6. Section 5.
 INTRODUCTION TO INFINITY-CATEGORIES 193

Exercise 85. Show that a left fibration p : C → D is a Kan fibration provided that D is an
∞-groupoid.

Exercise 86. Let p : C → D be an isofibration. Show that the set of all monomorphisms
K → L such that the induced map
CL → CK ×DK DL
is again an isofibration is a saturated set.

A.7. Section 7.
Exercise 87. The class of essentially surjective and fully faithful functors satisfies the 3-for-2
property.

Exercise 88. Suppose that f : C → D is a Joyal equivalence. Then show that also the
restricted map C' → D' is a Joyal equivalence.

Exercise 89. Show that two functors f, g : C → D are naturally equivalent if and only if f
and g represent the same element of π0 (Fun(C, D)' ).

Exercise 90. Show that a Joyal equivalence f : C → D induces an equivalence of ordinary

categories hC → hD.

Exercise 91. Show that a functor f : C → D between ∞-categories is a Joyal equivalence if

and only if for every simplicial set K, the induced map
f∗ : Fun(K, C) → Fun(K, D)
is a Joyal equivalence.

Exercise 92. Let F : C → D be a functor between ordinary categories. Show that the
induced map on nerves is inner anodyne if and only if F is an isomorphism.

A.8. Section 8.
Exercise 93. Let C be an ∞-category and S a set of morphisms of C. Then S is called
saturated if it coincides with the set S̄ of all morphisms that are sent to equivalences under
the functor C → C[S −1 ]. Show that
(1) If S ⊆ T and T is saturated, then S̄ ⊆ T as well.
(2) Two sets of morphisms S and T of C give rise to Joyal equivalent localizations (com-
patible with the map from C) if and only if S̄ = T̄ .

Exercise 94. Prove or disprove the following statements:

(1) For every ∞-category C, a set S of morphisms of C and a set T of morphisms of
C[S −1 ], the functor C → C[S −1 ][T −1 ] is a localization of C.
(2) For every ∞-category C, a set S of morphisms of C and a set T of morphisms of C,
the functor C → C[S −1 ][T −1 ] is a localization of C. Here, we view morphisms of C as
morphisms of C[S −1 ] via the canonical functor C → C[S −1 ].
194 M. LAND

Exercise 95. Let C be an ordinary category and S a set of morphisms. Show that C →
hC[S −1 ] is the initial functor C → D (up to natural isomorphism) between ordinary categories
sending S to isomorphisms.

Exercise 96. Let C be an ordinary category and S a set of morphisms. Show that every
morphism in hC[S −1 ] can be represented by a zig zag of morphisms in C, such that the maps
pointing in the wrong direction are contained in S.

Exercise 97. Let C be an ordinary category. Consider a pushout of categories

C
`
[1]
f ∈Mor(C)

D
`
J
f ∈Mor(C)

where Mor(C) denotes the set of all morphisms of C. Show that D is a groupoid.

Exercise 98. In this exercise you may use the fact that the unit map K → S(|K|) is a
homotopy equivalence for any Kan complex K. Recall that a map of simplicial sets is a weak
equivalence if its geometric realization is a homotopy equivalence and let X be a simplicial
set. Prove or disprove the following statements.
(1) the unit map X → S(|X|) is a monomorphism,
(2) the unit map X → S(|X|) is a weak equivalence,
(3) the unit map X → S(|X|) is anodyne.

Exercise 99. Let C be an ∞-category and let C → S(|C|) be the unit map of the adjunction
(S, | − |). Show that this is a localization of C along all morphisms.

Exercise 100. Show that the factorization C → P (f ) → D for a functor f : C → D between

∞-categories is functorial, i.e. that for every solid commutative diagram

C P (f ) D

C0 P (f 0 ) D0

a dashed arrow exists making both small squares commute.

A.9. Section 9.
Exercise 101. Consider a pushout diagram of simplicial sets
X Y

X0 Y0
in which X → X 0 is a monomorphism and X → Y is a Joyal equivalence. Show that the map
X 0 → Y 0 is also a Joyal equivalence.
 INTRODUCTION TO INFINITY-CATEGORIES 195

Exercise 102. Let X → X 0 and Y → Y 0 be a Joyal equivalences between simplicial sets.

Show that both maps X q Y → X 0 q Y 0 and X × Y → X 0 × Y 0 are Joyal equivalences.

Exercise 103. Show that

(1) A retract of a Joyal equivalence is a Joyal equivalence,
(2) The set of monomorphisms which are also Joyal equivalences is saturated
Prove or disprove that the set of Joyal equivalences saturated.

Exercise 104. Recall that a map f : X → Y is said to admit a pre-inverse if there exists
maps g : Y → X and τ : ∆1 → Hom(X, X) and τ 0 : ∆1 → Hom(Y, Y ) such that
(1) τε = idX and τ1+ε = gf , where ε ∈ {0, 1} ∼ = Z/2,
(2) τε0 = idY and τ1+ε
0 = f g, where again ε ∈ {0, 1} ∼= Z/2,
1
(3) for all objects x of X, the morphism τ (x) : ∆ → X represents a degenerate edge of
X, and for all objects y of Y , τ 0 (y) : ∆1 → Y represents a degenerate edge of Y .
Show that a map f : X → Y which admits a pre-inverse is a Joyal equivalence.

Exercise 105. Let p : C → D be an inner fibration between ∞-categories which induces a

surjection on 0-simplices and is a Joyal equivalence. Show that p is a trivial fibration.

Exercise 106. Show that for any two simplicial sets, X ? Y is a retract of X  Y .

Exercise 107. Show that the canonical map

|Hom(A, B)| −→ map(|A|, |B|)
is a homotopy equivalence. You may use the fact that both the unit map A → S(|A|) and
the counit map |S(X)| → X are weak equivalences.

The goal of the following exercises is to (almost) give a proof of the fact that anodyne maps
are precisely those monomorphisms which are weak equivalences. Precisely, we will show that
it is implied by the following statement. Let p : X → Y be a Kan fibration. Then there exists
factorization of p as
α β
X −→ Z −→ Y
where β is a trivial fibration (i.e. has the RLP with respect to monomorphisms) and α is a
minimal fibration. The following is what we need to know about minimal fibrations:
• a minimal fibration is a Kan fibration,
• a minimal fibration α : X → Z is locally trivial, i.e. for every simplex ∆n → Z, the
pulled back fibration is isomorphic (over ∆n ) to a projection ∆n × B → ∆n .
Using this, exercises 127–131 will prove our main result:
Exercise 108. Show that a Kan fibration p : X → Y is a trivial fibration if and only if its
fibres are contractible. Hint: For the interesting direction, consider a lifting problem
a
∂∆n X

b
∆n Y
196 M. LAND

and for future reference we let let y = b(n). Consider the map g : ∆n × ∆1 → ∆n determined
by g(k, 0) = k and g(k, 1) = n. Argue that the map ∂∆n × {0} → ∂∆n × ∆1 is anodyne, and
use g to obtain a map
n n
∆1 → X ∂∆ ×Y ∂∆n Y ∆
sending 0 to the original square and 1 to the square
∂∆n X

y
∆n ∆0 Y
where the map ∆0 → Y is the object y. Show that this lifting problem can be solved. Finally,
show that this implies that also the original lifting problem can be solved.

Exercise 109. In this exercise we will use minimal fibrations to show that the geometric
realization of a Kan fibration p : X → Y is a Serre fibration, i.e. has the RLP wrt the inclusions
Dn × {0} → Dn × D1 . Here, Dn is the n-dimensional topological cube.
(1) Show that the geometric realization of a minimal fibration is a Serre fibration whose
fibre is given by the geometric realization of the fibre of the Kan fibration.
(2) Show that the geometric realization of a trivial fibration is a Serre fibration.
Hints: For (1) show that the realization of a locally trivial map is also locally trivial. Then
show that a locally trivial map of spaces is a Serre fibration. For (2) show that there exists
a monomorphism X → W with W a contractible Kan complex. Consider the maps X →
W × Y → Y . Show that the latter is a trivial fibration and deduce that p is a retract of
W ×Y →Y.

Exercise 110. Show that a Kan fibration p : X → Y which is in addition a weak equivalence
has contractible fibres.

Exercise 111. Show that a monomorphism i : A → B is a weak equivalence if and only if

it is anodyne.

Exercise 112. Show that a cocartesian fibration p : X → Y whose fibres are Joyal equivalent
to ∆0 is a trivial fibration.

Exercise 113. Let P be a poset and C an ∞-category. Suppose given a function f : P →

ob(C) having the following property: Whenever x ≤ y are elements of P , then the anima of
maps mapC (f x, f y) is contractible.
(1) Show that there exists a functor F : P → C extending the given function f on objects,
(2) Show that any two such extensions are equivalent.
The “correct” version of (2) of course is the following: Show that there is a contractible anima
parametrizing all possible choices of such extensions, i.e. that in the pullback diagram
Ext(f ) Fun(P, C)

∆0 Fun(P0 , C)
 INTRODUCTION TO INFINITY-CATEGORIES 197

the simplicial set Ext(f ) is a contractible Kan complex. Can one replace P by an arbitrary
1-category and obtain the same results?

Exercise 114. Let C be an ∞-category and consider for each i ≥ 0 a morphism xi → xi+1
between objects in C. Show that these maps assemble into a functor N → C.

Exercise 115. Prove or disprove that an isofibration p : C → D between ∞-categories is a

trivial fibration if and only if its fibres are Joyal equivalent to ∆0 .

A.10. Section 10.

Exercise 116. Let C be an ∞-category and consider the inner fibration p : C → ∆0 . Show
that a morphism f in C is p-(co)cartesian if and only if f is an equivalence.

Exercise 117. Show that given a morphism between two squares of ∞-categories in which
each comparison map is a Joyal equivalence, then the one square is homotopy cartesian if and
only if the other is.

Exercise 118. Suppose given a pullback diagram

X0 X
p0 p

Y0 Y

in which the map p is a (co)cartesian fibration. Show that also the map p0 is a (co)cartesian
fibration.

A.11. ??.

Exercise 119. Show that the functors (−)[ , (−)] : sSet → sSet+ and the functors u, m : sSet+ →
sSet are involved in various adjunctions. Here, u is the forgetful functor and m is the func-
tor which sends a marked simplicial set (X, S) to the smallest sub simplicial set X0 ⊆ X
containing S.

Exercise 120. Let p : E → C be a cocartesian fibration and K a marked simplicial set

equipped with a map f : K → C] . Consider the sub-∞-category of Funmcc f (K, E) on those
1-simplices whose corresponding map K × ∆ → E is a map of marked simplicial sets K ×
1

(∆1 )] → E\ , i.e. we consider only those transformations of functors whose components are
f (K, E) .
pointwise p-cocartesian. Then this sub-∞-category is given by Funmcc '

Exercise 121. Show that if K = ∆0 and f : ∆0 → C picks out an object z of C, then

Funcc (K, E) ∼
= E and Funcc ∼
f (K, E) = Ez .

Exercise 122. Let X → Y be a cocartesian fibration and f : K → Y ] a map of marked

simplicial sets. Show that Map[f (K, X \ ) is an ∞-category and that Map]f (K, X \ ) is the largest
sub ∞-groupoid inside Map[f (K, X \ ).
198 M. LAND

Exercise 123. Show that the functor (−)[ : sSet → sSet+ sends Joyal equivalences to
marked equivalences.

Exercise 124. Show that the category of marked simplicial sets is canonically enriched in
simplicial sets.

Exercise 125. Show that the functor LF : sSet → sSet of Theorem 12.8 preserves colimits
and monomorphisms.

Exercise 126. Show that the initial vertex maps assemble to a natural transformation
LF ⇒ id.

Exercise 127. Suppose given a commutative diagram

A0 A1 A2 ...
f0 f1 f2

B0 B1 B2 ...
in which the maps fi are Joyal equivalences and all horizontal maps are monomorphisms.
Then the induced map f : A = colim Ai → colim Bi = B is a Joyal equivalence.

Exercise 128. Show that for an ∞-groupoid X, there is a canonical Joyal equivalence
X op ' X.

Exercise 129. Show that the functor N(∆op /C ) → C is full in the sense that every morphism of
C is the image of a morphism under this functor. Use this to show the composition C → Cat∞
induces on objects and morphisms the constructions we have done earlier.

Exercise 130. Suppose given a cocartesian fibration p : E → C. Recall that we have

constructed for every edge f : ∆1 → C a functor Ex × ∆1 → E, whose restriction to Ex × {1}
is given by f! . Show that this functor may equivalently be constructed as follows. Consider
the diagram
Ex × ∆{0} E

f ◦pr
Ex × {1} Ex × ∆ 1 C
and show that a dashed arrow exists having all the above properties.

Exercise 131. Given a cocartesian fibration E → C × ∆1 , construct a functor E0 → E1

which commutes with the projections to C. Here, Ei is the pulled back cocartesian fibration
along the inclusion C × {i} → C × ∆1 . Show that this functor is a morphism of cocartesian
fibrations. Likewise, construct for a cocartesian fibration E → C × ∆2 a 2-simplex in the
∞-category (Cat∞ )/C . If you are eager, do this for general n instead of 2.

Exercise 132. Let p : C → D be an inner fibration between ∞-categories. Suppose that for
every map f : ∆1 → D, the induced map ∆1 ×D C → ∆1 is a Joyal equivalence. Show that p
is a Joyal equivalence.
 INTRODUCTION TO INFINITY-CATEGORIES 199

Exercise 133. Let C be an ∞-category and x and object of C. Show that the object
represented by ∆0 → C/x which is adjoint to the map ∆0 ? ∆0 → C given by idx is a terminal
object.

Exercise 134. Let C be an ∞-category and x and initial object of C. Show that if y is
equivalent to x, then y is also initial in C.

Exercise 135. Suppose that Cterm is not empty. Show that any terminal object of Fun(K, C)
takes values in Cterm .

Exercise 136. Show that there exists C and K such that Fun(K, C) has a terminal object
but C does not.

Exercise 137. Show that there exists a simplicial set X such that the above described map
Tw(X) → X × X op is not a right fibration.

Exercise 138. Prove or disprove that the diagram

Fun(∆1 , C) Fun(∆1 , D)

C×C D×D

is homotopy cartesian if and only if f is fully faithful.

Exercise 139. Let p : K → C be a diagram. Suppose that there is a colimit cone p̃ with
colimit x in C. Let y be an object of C which is equivalent to x. Show that y is also the
colimit of a colimit cone of p.

Exercise 140. Let p : K → C be a diagram and let q : I → Cp/ be a further diagram. Let
q̄ : I ? K → C be the associated map. Then there is an isomorphism (Cp/ )q/ ∼ = Cq̄/ . Likewise,
there is an isomorphism (C/p )q/ ∼ (C )
= q / /p
0 0 where q 0 is the restriction of q : I ? K → C to I

and p0 is adjoint to q̄.

Exercise 141. Let K be the coequalizer of two monomorphisms f, g : A → B and let

F : K → C be a functor. Suppose that the restrictions of F to B and A admit colimits and
that C admits coequlizers. Show that then F admits a colimit.

Exercise 142. A map f : K → L is cofinal if and only if f op : K op → Lop is coinitial.

Exercise 143. Show that left anodyne maps do not satisfy the left cancellation property
among monomorphisms.

Exercise 144. Show that the set of right anodyne maps i : A → B whose pullback along
any left fibration is again right anodyne is a saturated set and satisfies the right cancellation
property.
200 M. LAND

Exercise 145. A right deformation retract is a monomorphism i : A → B such that there

exists a retraction p : B → A and a simplicial homotopy H : ∆1 × B → B with H(0) = idB ,
H(1) = ip and whose restriction to ∆1 × A is constant the identity of A.
(1) Show that for every simplicial set, the map {1} × K → ∆1 × K is a right deformation
retract.
(2) Show that a right deformation retract is a right anodyne map.
(3) Show that the pullback of a right deformation retract along a left fibration is again a
right deformation retract.

Exercise 146. Give an example of a proper (or smooth) map which is not universally proper
(or smooth).

Adjunctions.
Exercise 147. Let p : E → D be a cartesian fibration. Show that the canonical map
Ed → Ed/ admits a right adjoint. Deduce that for a cartesian fibration, the canonical map
Ed → Ed/ is a weak equivalence.

Exercise 148. Show that a fully faithful and essentially surjective functor is invertible.

Exercise 149. Let F : K → Cat∞ be a functor. Let p : E → K be the associated cocartesian

fibration. Show that the colimit of F is given by E[cc−1 ], where the set cc is the set of p-
cocartesian edges. Likewise, show that the limit is given by Funcc K (K, E) is given by the
category of cocartesian sections of p. Deduce the analogs for functors with values in An.

Exercise 150. Consider for a diagram F : K → C and an object x of C the functor

mapC (F (−), x) : K → Anop .
Show that its limit is given by MapC (F, x). Deduce that for a functor f : C → D, with right
adjoint g, there exists a canonical equivalence MapD (f F, z) ' MapC (F, gz).

Exercise 151. Let f : C → D and f 0 : D → E be composable functors which admit right

adjoints g and g 0 . Show that then gg 0 is a right adjoint of f 0 f .

References
[Cis19] D.-C. Cisinski, Higher categories and homotopical algebra, Cambridge studies in advanced mathe-
matics, vol. 180, Cambridge university press, 2019.
[GJ09] Paul G. Goerss and John F. Jardine, Simplicial homotopy theory, Modern Birkhäuser Classics,
Birkhäuser Verlag, Basel, 2009, Reprint of the 1999 edition [MR1711612]. MR 2840650
[Hir03] Philip S. Hirschhorn, Model categories and their localizations, Mathematical Surveys and Mono-
graphs, vol. 99, American Mathematical Society, Providence, RI, 2003. MR 1944041
[Hir15] , The homotopy groups of the inverse limit of a tower of fibrations, arXiv:1507.01627 (2015).
[Joy07] A. Joyal, Quasi-categories vs simplicial categories.
[Lur09] J. Lurie, Higher topos theory, Annals of Mathematics Studies, vol. 170, Princeton University Press,
Princeton, NJ, 2009. MR 2522659
[Lur17] , Higher algebra, available at [Link] lurie/papers/[Link] (2017).
[Ngu18] H. K. Nguyen, Theorems in higher categories and applications, Phd thesis (2018).
[NRC19] H. K. Nguyen, G. Raptis, and Schrade C., Adjoint functor theorems for ∞-categories, Journal of the
LMS, to appear (2019).
 INTRODUCTION TO INFINITY-CATEGORIES 201

[RV18] E. Riehl and D. Verity, The comprehension construction, High. Struct. 2 (2018), no. 1, 116–190.
MR 3917428
[Ste17] D. Stevenson, Covariant model structures and simplicial localization, arXiv:1512.04815v3 (2017).
[Ste18] , Model structures for correspondences and bifibrations, arXiv:1807.08226 (2018).

University of Regensburg, NWF I, Universitätsstr. 31, 93053 Regensburg

E-mail address: [Link]@[Link]