Quillen Homotopical-Algebra

Daniel Quillen
Homotopical
Algebra
A X
i f
B Y
Revised and modernized edition by

TEXromancers
TEXromancers
A collaborative typesetting project
This book is the product of a community effort. It was typeset by TEXromancers:

an enthusiast group of mathematicians (for the most part), consisting of people
organized on Discord. The reader should contact [email protected] if they are
interested in joining the group. Link to the group’s page: https://aareyanman
zoor.github.io/Texromancers.html.
More books in this series:

• J. F Adams, Stable Homotopy and Generalised Homology
– Available at https://people.math.rochester.edu/faculty/doug
/otherpapers/Adams-SHGH-latex.pdf
• Noel J. Hicks, Notes on Differential Geometry

– Available at https://aareyanmanzoor.github.io/assets/hicks.
pdf
• Hideyuki Matsumura, Commutative Algebra
– https://aareyanmanzoor.github.io/assets/matsumura-CA.pdf
• John Milnor and James Stasheff, Characteristic Classes
– Available at https://aareyanmanzoor.github.io/assets/books/
characteristic-classes.pdf
• J.C. Lagarias and A.M. Odlyzko, Effective Versions of the Chebotarev

Density Theorem
– Available at https://aareyanmanzoor.github.io/assets/articl
es/lagarias-odlyzko.pdf
• Daniel Quillen, Homotopical Algebra

– Available at https://aareyanmanzoor.github.io/assets/books/
homotopical-algebra.pdf
Daniel Quillen
Homotopical
Algebra
Revised and modernized edition by

TEXromancers
Typeset by TEXromancers.
This digital book is meant for educational and scholarly purposes only, itself
being a derivative work containing a transcription of a copyrighted publication.
We claim no ownership of the original work’s contents; all rights to the original
book’s contents belong to the respective publisher.
We believe that our use of the original work constitutes fair use, as defined
under section 107 of the United States Copyright Act of 1976, wherein allowance
is made for “fair use” for purposes such as criticism, comment, news reporting,
teaching, scholarship, education and research.
Commercialization of this digital book is prohibited.
Original edition: 1965

This edition: 2022
Contents
Contents 5
Preface 7
Editor’s Preface 9
1 Axiomatic homotopy theory 11

1.0 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.1 The axioms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.2 The loop and suspension functors . . . . . . . . . . . . . . . . . . . 29
1.3 Fibration and Cofibration Sequences . . . . . . . . . . . . . . . . . 38
1.4 Equivalences of homotopy theories . . . . . . . . . . . . . . . . . . 52
1.5 Closed model categories . . . . . . . . . . . . . . . . . . . . . . . . 59
2 Examples of simplicial homotopy theories 65

2.0 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.1 Simplicial categories . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.2 Closed simplicial model categories . . . . . . . . . . . . . . . . . . 76
2.3 Topological spaces, simplicial sets, and simplicial groups . . . . . . 83
2.4 sA as a model category . . . . . . . . . . . . . . . . . . . . . . . . 100
2.5 Homology and cohomology . . . . . . . . . . . . . . . . . . . . . . 109
2.6 Modules over a simplicial ring . . . . . . . . . . . . . . . . . . . . . 122
Bibliography 139
Index 141
5
6 CONTENTS
Preface
Homotopical algebra or non-linear homological algebra is the generalization

of homological algebra to arbitrary categories which results by considering a
simplicial object as being a generalization of a chain complex. The first step in
the theory was presented in [Dol58], [DP61], where the derived functors of a non-
additive functor from an abelian category A with enough projectives to another
category B were constructed. This construction generalizes to the case where A
is a category closed under finite limits having sufficiently many projective objects,
and these derived functors can be used to give a uniform definition of cohomology
for universal algebras. In order to compute this cohomology for commutative
rings, the author was led to consider the simplicial objects over A as forming
the objects of a homotopy theory analogous to the homotopy theory of algebraic
topology, then using the analogy as a source of intuition for simplicial objects.
This was suggested by the theorem of Kan [Kan58a] that the homotopy theory
of simplicial groups is equivalent to the homotopy theory of connected pointed
spaces and by the derived category ([Har66], [Ver]) of an abelian category. The
analogy turned out to be very fruitful; but there were a large number of arguments
which were formally similar to well-known ones in algebraic topology, so it was
decided to define the notion of a homotopy theory in sufficient generality to cover
in a uniform way the different homotopy theories encountered. This is what is
done in the present paper; applications are reserved for the future.
7
8 CONTENTS
The following is a brief outline of the contents of this paper; for a more
complete discussion see chapter introductions. Chapter 1 contains an axiomatic
development of homotopy theory patterned on the derived category of an abelian
category. In Chapter 2 we give various examples of homotopy theories that
arise from these axioms, in particular we show that the category of simplicial
objects in a category A satisfying suitable conditions gives rise to a homotopy
theory. Also in §2.5 we give a uniform description of homology and cohomology
in a homotopy theory as the “linearization” or “abelianization” of the non-linear
homotopy situation, and we indicate how in the case of algebras this yields a
reasonable cohomology theory.
The author extends his thanks to S. Lichtenbaum and M. Schlesinger who sug-
gested the original problem on commutative ring cohomology, to Robin Hartshorne
whose seminar [Har66] on Grothendieck’s duality theory introduced the author
to the derived category, and to Daniel Kan for many conversations during which
the author learned about simplicial methods and formulated many of the ideas
in this paper.
Preface to the new Typesetting
The book was TeX’d up by the Texromancers, a latexing group. The credits
for the typesetting of this book go to: Aareyan Manzoor, Jonas Hardt, Evelyn
Koo, Yohan Wittgenstein, Grisha Taroyan and others.
Here is a link to a dyslexic friendly version of the book: https://aareya
nmanzoor.github.io/assets/books/homotopical-algebra-dyslexic.pdf .
Here is a link to our LATEX: https://github.com/AareyanManzoor/Quillen-H
omotopical-Algebra
We added citations and references with hyperlinks. References to e.g. the-
orems/lemmas in the book are in blue, while citations to the bibliography is in
red. The bibliography also has URLs now, for easy access. Some of the books in
the bibliography had newer editions, so we went with those.
We also added a small index. This is minimalist as of now, if the reader
feels more words should be added to it, please let us know. Also please report
typos or anything else to us, readers taking the time to help the proofreading is
appreciated. This can be reported at our website: https://aareyanmanzoor.g
ithub.io/Texromancers.html.
We changed some notations, particularly for the name of categories. (sets) is
now Set for example. Underlining is also a relic of the typewritter era, so those
were replaced with bolded text, same for category names. For example category
C in the original book is now C.
9
Chapter 0: CONTENTS
All diagrams are redrawn in tikz or tikzcd.
10
1. Axiomatic homotopy theory
1.0 Introduction
Chapter 1 is an attempt to define what is meant by a “homotopy theory” in a

way sufficiently general for various applications. The basic definition is that of a
model category which is a category endowed with three distinguished families
of maps called cofibrations, fibrations, and weak equivalences satisfying certain
axioms, the most important being the following two:
M1 Given a commutative solid arrow diagram
A X
i p
B Y
where i is a cofibration, p is a fibration, and either i or p is also a weak

equivalence, there exists a dotted arrow such that the total diagram is
commutative.
M2 Any map f may be factored f = pi and f = p′ i′ where i, i′ are cofibrations,

where p, p′ are fibrations and where p and i′ are also weak equivalences. It
should be noticed that we do not assume the existence of a path or cylinder
functor; in fact the homotopy relation for maps may be recovered as follows:
11
Chapter 1: Axiomatic homotopy theory
Call an object cofibrant if the map ∅ ! X is a cofibration (hence in the

category of simplicial groups the cofibrant objects are the free simplicial
groups) and fibrant if the map X ! e is a fibration (hence in the category
of simplicial sets the fibrant objects are the Kan complexes). Then two
maps f, g from a cofibrant object A to a fibrant object B are said to be
homotopic if there exists a commutative diagram
f +g
A∨A B
id + id
i0 +i1
h (I)
A σ A′
where ∨ denotes direct sum, f + g is the map with components f and g,

and where σ is a weak equivalence.
Given a model category C, the homotopy category Ho C is obtained from

C by formally inverting all the weak equivalences. The resulting “localization”
γ : C ! Ho C is in general not calculable by left or right fractions [GZ67] but is
rather a mixture of both. The main result of §1.1 is that Ho C is equivalent to
the category πCcf whose objects are the cofibrant and fibrant objects of C and
whose morphisms are homotopy classes of maps in C. If C is a pointed category
then in §§1.2–1.3 we construct the loop and suspension functors and the families
of fibration and cofibration sequences in the homotopy category. If one defines
a cylinder object for a cofibrant object A to be an object A′ together with a
cofibration i0 +i1 and a weak equivalence σ as in diagram I, then the constructions
are the same as in the ordinary homotopy theory except that, since a cylinder
object of A is neither unique nor functorial in A, one has to be careful that things
are well–defined. This is done by defining operations in two ways using the left
(cofibration) structure and the right (fibration) structure, and showing that the
12
Section 1.1: The axioms
two definitions coincide.

The term “model category” is short for “a category of models for a homo-
topy theory”, where the homotopy theory associated to a model category C is
defined to be the homotopy category Ho C with the extra structure defined in
§1.2-1.3 on this category when C is pointed. The same homotopy theory may
have several different models, e.g. ordinary homotopy theory with basepoint is
([Kan58a], [Mil57]) the homotopy theory of each of the following model cate-
gories: 0–connected pointed topological spaces, reduced simplicial sets, and sim-
plicial groups. In section 1.4 we present an abstract form of this result which
asserts that two model categories have the same homotopy theory provided there
are a pair of adjoint functors between the categories satisfying certain conditions.
This definition of the homotopy theory associated a model category is ob-
viously unsatisfactory. In effect, the loop and suspension functors are a kind of
primary structure on Ho C, and the families of fibration and cofibration sequences
are a kind of secondary structure since they determine the Toda bracket (see §1.3)
and are equivalent to the Toda bracket when Ho C is additive. (This last remark
is a result of Alex Heller.) Presumably there is higher order structure ([Ger65],
[Spa63]) on the homotopy category which forms part of the homotopy theory of a
model category, but we have not been able to find an inclusive general definition
of this structure with the property that this structure is preserved when there
are adjoint functors which establish an equivalence of homotopy theories.
In section 1.5 we define a closed model category which has the desirable
property that a map is a weak equivalence if and only if it becomes an isomor-
phism in the homotopy category.
1.1 The axioms
All diagrams are assumed to be commutative unless stated otherwise.
13
Definition 1.1.1. By a model category we mean a category together with

three classes of maps in C , called the fibrations, cofibrations, and weak equiva-
lences, satisfying the following axioms.
M0 C is closed under finite projective and inductive limits.
M1 Given a solid arrow diagram
A X
i p (1)
B Y
where i is a cofibration, p is a fibration, and where either i or p is a weak

equivalence, then the dotted arrow exists.
M2 Any map f may be factored f = pi where i is a cofibration and weak

equivalence and p is a fibration. Also f = pi where i is a cofibration and p
is a fibration and weak equivalence.
M3 Fibrations are stable under composition and base change. Any isomorphism
is a fibration.
Cofibrations are stable under composition and co-base change. Any iso-
morphism is a cofibration.
M4 The bases extension of a map which is both a fibration and a weak equiva-
lence is a weak equivalence. The co-base extension of a map which is both
a cofibration and a weak equivalence is a weak equivalence.
M5 The bases extension of a map which is both a fibration and a weak equiva-
lence is a weak equivalence. The co-base extension of a map which is both
a cofibration and a weak equivalence is a weak equivalence.
14
f g
M6 Let X −! Y −! Z be maps in C . Then if two of the maps f, g, and gf are
weak equivalences, so is the third. Any isomorphism is a weak equivalence.
Examples. A. Let C be the category of topological spaces and continuous

maps. Let fibrations in C be fibrations in the sense of Serre, let cofibra-
tions be maps having the lifting property of axiom M1 whenever p is both
a Serre fibration and a weak homotopy equivalence, and finally let weak
equivalences in C be weak homotopy equivalences (maps inducing isomor-
phism for the functions [K, −] where K is a finite complex). Then the
axioms are satisfied. (This is proved in §2.3.)
B. Let A be an abelian category with sufficiently many projectives and let

C = C+ (A ) be the category of complexes K = {Kq , d : Kq −! Kq−1 } of
objects of A which are bounded below (Kq = 0 if q ≪ 0). Then C is a
model category where weak equivalences are maps inducing isomorphisms
on homology, where fibrations re the epimorphisms in C , and where the
cofibrations are maps i which are injective such that Coker i is a complex
having a projective object of A in each dimension.
C. Let C be the category of semi-simplicial sets and let fibrations in C be

the Kan fibrations, cofibrations be injective maps, and let the weak equiv-
alences be maps which become homotopy equivalences when the geometric
realizations functor is applied. Then C is a model category (§2.3).
For the rest of this section C will denote a fixed model category.
Definition 1.1.2. Let ∅ (resp. e) denote “the” initial (resp. final) object of the
category C ). (These exist by M0.) An object X will be called cofibrant if
∅ −! X is a cofibration and fibrant if X −! e is a fibration. A map which
is a fibration (resp. cofibration) and a weak equivalence will be called a trivial
fibration (resp. trivial cofibration.)
15
Remark. In example A. every object is fibrant and the class of cofibrant objects
include CW complexes, and more generally any spaces that is constructed by a
well ordered succession of attaching cells. In example B. every object is fibrant
and the cofibrant objects are the projective complexes (that is, complexes con-
sisting of projective objects – these are not projective objects in C+ (A )). In
example C. every object is cofibrant and the fibrant objects are those s.s. sets
satisfying the extension condition.
Before stating the next definition we recall some standard notation concern-
ing the fibre products and introduce some not-so-standard notation for cofibre
products. Given a diagram
β
A X
α γ (2)
δ
B Y
there is a unique map A −! B ×Y X denotes (α, β)Y or simply (α, β) such

that pr1 (α, β) = α and pr2 (α, β) = β, where pr1 : B ×Y X −! B and pr2 :
B ×Y X −! X are the canonical projections. Also (2) is said to be cartesian
if (α, β) is an isomorphism. We shall denote the cofibre product of B and X
under A by B ∨A X and the two canonical maps by in1 : B −! B ∨A X and
u
in2 : X −! B ∨A X. The unique map B ∨A X −! Y with u in1 = δ and u in2 = γ
will be denoted δ +A γ or simply δ + γ, and (2) will be called co-cartesian if
δ + γ is an isomorphism. Finally given a map f : X −! Y there is the diagonal
map
∆f = (idX , idX ) : X −! X ×Y X
16
and the codiagional map
∇f = idY + idY : Y ∨X Y −! Y
of f . We write ∆X (resp. ∇Y ) if Y = e (resp. X = ∅).
Definition 1.1.3. Let f, g : A ⇒ B be maps. We say that f is left-homotopic

l
to g notation f ∼ g) if there is a diagram of the form
f +g
A∨A B
∇
∂0 +∂1 h (3)
A σ A
e
where σ is a weak equivalence. Dually we say that f is right-homotopic to g

r
(notation: f ∼ g) if there is a diagram of the form
s
B
e B
k
(d0 ,d1 )
∆ (4)
A B×B
(f,g)
where s is a weak equivalence.
Remark. In example A. above two maps of spaces which are homtopic in the
usual sences are both left and right homotopic as one sees by taking A
e = A×I
and B
e = B I where I is the unit interval. In fact we have the implications:
homotopic =⇒ right homotopic =⇒ left homotopic (5)
where the last implication comes from the dual of lemma 1.1.5(i) below and the
fact that every space is fibrant. if A is cofibrant (e.g. a CW complex) then the
17
three notation coincide, but in general it seems that the implication (5) are strict.
Definition 1.1.4. By cylinder object for an object A we mean an object A × I

together with maps
∂ +∂ σ
A ∨ A −−0−−!
1
A × I −! A with σ(∂0 + ∂1 ) = ∇A
such that ∂0 + ∂1 is a cofibration and σ is a weak equivalence. Dually, a path

object for B shall be an object B I together with a factorization
s (d0 ,d1 )
B −! B I −−−−! B × B of ∆B
where s is a weak equivalence and (d0 , d1 ) is a fibration. By a left homotopy

from f : A −! B to g : A −! B we mean a diagram (3) where ∂0 + ∂1 is a
cofibration and hence A
e is a cylinder object for A. Similarly a right homotopy
from f to g is a diagram (4) where B

e is a path object for B.
Remark. 1. A × I is not the product of A and an object I nor is it a functor

of A. In example A., the product of a space A and the unit interval is not
necessarily a cylinder object of A unless A is cofibrant.
2. Since the dual of a model category is again a model category in an evident

way there is a corresponding dual assertion for every assertion we make. In
the following we will often give only one form and leave the formulation of
the dual assertion to the reader.
l
Lemma 1.1.1. If f, g ∈ Hom(A, B) and f ∼ g, then there is a left homotopy
h : A × I −! B from f to g.
′ ′
∂ +∂1 ρ
Proof. Given diagram (3) use M2 to factor ∂0 + ∂1 into A ∨ A −−0−−! A′ −! A
e
where ∂0′ + ∂1′ is a trivial cofibration and ρ is a trivial fibration. By M5
18
σ ′ = σρ : A′ −! A is a weak equivalences so A′ with ∂0′ , ∂1′ , and σ ′ is a cylinder

object for A. h′ = hρ : A′ −! B is the desired left homotopy from f to g.
Lemma 1.1.2. Let A be a cofibrant object and let A × I be a cylinder object

for A. Then
∂0 : A −! A × I and ∂1 : A −! A × I are trivial cofibrations.
Proof. in1 : A −! A ∨ A is a commutativ by the coase change assertion in M3,

hence ∂0 = (∂0 + ∂1 ) in1 is a cofibration. σ∂0 = idA ands M5 imply that ∂0 is
also a weak equivalence. Similarly ∂1 is a trivial cofibration.
Corollary (Covering Homotopy theorem). Let A be cofibrant and let

p : X −! Y be a fibration, let α : A −! X, and let h : A × I −! Y be a left
homotopy with h∂0 = pα. Then there is a left homotopy H : A × I −! X with
H∂0 = α and pH = h.
Proof. By M1, H exists in
α
A X
H p
∂0
A×I h
Y
The dual assertion is the homotopy extension theorem.
Lemma 1.1.3. Let A be cofibrant and let A × I and A × I ′ be two cylinder

objects for A. Then the result of “gluing” A × I to A × I ′ by the identification
19
∂1 A = ∂0′ A, defined precisely to be the object A

e in the co-Cartesian diagram
∂0′
A A × I′
∂1 in2 (6)
A×I in1
A
e
is also a cylinder object A × I ′′ for A with
∂0′′ = in1 ∂0 , ∂1′′ = in2 ∂1′ , σ ′′ in1 = σ, σ ′′ in2 = σ ′ .
Proof. M4 and Lemma 1.1.2 show that in1 and in2 are weak equivalences; as ∂0′′ =
in1 ∂0 , σ ′′ ∂0′′ = idA we have by M5 that σ ′′ : A
e −! A is a weak equivalence.
in1 ∨ idA
∂0′′ + ∂1′′ : A ∨ A −! A
e is the composition of A ∨ A −−−−−! (A × I) ∨ A, which is
′′
in1 in1 +∂0
the co-base extension of ∂0 by A −−! A ∨ A1 , and the map (A × I) ∨ A −−−−−! A,
e
∂ +id
which is the co-base extension of ∂0′ + ∂1′ by A ∨ A −−1−−−!
A
(A × I) ∨ A. By M3
∂0′′ + ∂1′′ is a cofibration and hence A
e is a cylinder object for A.
l
Lemma 1.1.4. If A is cofibrant, then ∼ is an equivalence relation in Hom(A, B).
Proof. The relation is reflexive since if f = g we may take A

e = A and h = f in
(3) and it is symmetric since given (3) we may interchange ∂0 and ∂1 . Finally
given f0 , f1 , f2 ∈ Hom(A, B) and a left homotopy h : A × I −! B from f0 to
f1 and a left homotopy h′ : A × I ′ −! B from f0 to f1 and a left homotopy
h′ : A × I ′ −! B we obtain by Lemma 1.1.3 a left homotopy h′′ : A × I ′′ −! B
from f0 to f2 by setting h′′ in1 = h and h′′ in2 = h′ .
Lemma 1.1.5. Let A be cofibrant and let f, g ∈ Hom(A, B). Then
l r
(i) f ∼ g =⇒ f ∼ g.
20
r
(ii) f ∼ g =⇒ there exists a right homotopy k : A −! B I from f to g with
s : B −! B I a trivial cofibration.
r r
(iii) If u : B −! C, then f ∼ g =⇒ uf ∼ ug.
Proof. (i) By Lemma 1.1.1 there is a left homotopy h : A × I −! B from f to

g and by M2 there is a path object B I for B. By Lemma 1.1.2 and M1 the
dotted arrow K exists in
sf
A BI
∂0
K
(d0 ,d1 ) (7)
A×I B×B
(f σ,h)
and k = K∂1 : A −! B I is the desired right homotopy from f to g.
s e ρ ′
(ii) Let k ′ : A −! B I be a right homotopy from f to g and let B −! B −! B I
′
be a factorization of s′ : B −! B I into a trivial cofibration followed by a
fibration. By M5 ρ is a weak equivalence. Let
e −! B × B so that (d0 , d1 ) is a fibration by M3 and
(d0 , d1 ) = (d′0 , d′1 )ρ : B
hence B
e with d0 , d1 , and s is a path object for B. By M1 there is a dotted
arrow k in
∅ BI
k ρ (8)
′
A BI
k′
and k gives the desired homotopy from f to g.
(iii) Let k be as in (ii) and let C I be a path object for C. By M1 it is possible
21
to lift in
su
B CI
s
ϕ
(d0 ,d1 ) (9)
BI C ×C
(d0 u,d1 u)
and kϕ : A −! C I is a right homotopy from uf to ug.
If A and B are objects of C we let π r (A, B) (resp. π l (A, B)) be the set of
equivalence classes of Hom(A, B) with respect to the equivalence relation gener-
r l l
ated by ∼ (resp. ∼). When A cofibrant and B is fibrant, in which case ∼ and
r
∼ coincide and are already equivalence relations by Lemmas 1.1.4, 1.1.5(i) and
their duals, we shall denote the relation by ∼, call it homotopy and let π0 (A, B)
or simply π(A, B) be the set of equivalence classes.
Lemma 1.1.6. If A is cofibrant, then composition in C induces a map

π r (A, B) × π r (B, C) −! π r (A, B).
r
Proof. It suffices to show that if f, g ∈ Hom(A, B), u ∈ Hom(B, C) and f ∼ g
r
then uf ∼ ug, which is Lemma 1.1.5(iii), and that if u, v ∈ Hom(B, C),
r r
f ∈ Hom(A, B), and u ∼ v, then uf ∼ vf , which is immediate from the definition.
Lemma 1.1.7. Let A be cofibrant and let p : X −! Y be a trivial fibration.

∼
Then p induces a bijection p∗ : π l (A, X) −! π l (A, Y ).
l l
Proof. The map is well–defined since f ∼ g =⇒ pf ∼ pg is immediate from the
definition. The map is surjective by M1. By Lemma 1.1.4 if f, g ∈ Hom(A, X)
and pf, pg represent the same element of π l (A, Y ), then there is a left homotopy
22
h : A × I −! Y from pf to pg. If H is a lifting in
f +g
A∨A X
∂0 +∂1
H p (10)
h
A×I Y
then H is a left homotopy from f to g. This shows that p∗ is injective.
Let Cc , Cf , and Ccf be the full subcategories consisting of the cofibrant,

fibrant, and both fibrant and cofibrant objects of C respectively. By Lemma 1.1.6
we may define a category πCc with the same objects as Cc , with
HomπCc (A, B) = π r (A, B) and with the composition induced from that of C .
If we denote the right homotopy class of a map f : A −! B by f we obtain a
functor Cc −! πCc given by X −! X, f −! f . Similarly largely by the dual of
Lemma 1.1.6 we may define πCf (resp. πCcf ) to be the category with the same
objects as Cf and with π l (A, B) (resp. π(A, B)) as maps from A to B.
Definition 1.1.5. Let C be an arbitrary subcategory and let S be a subclass

of the class of maps of C . By the localization of C with respect to S we mean
a category S −1 C together with a functor γ : C −! S −1 C having the following
universal property: For every s ∈ S, γ(s) is an isomorphism; given any functor
F : C −! B with F (s) an isomorphism for all s ∈ S, there is a unique functor
θ : S −1 C −! B such that θ ◦ γ = F .
Except for set–theoretic difficulties the category S −1 C exists and may be

constructed by mimicking the construction of the free group (see Gabriel–Zisman
[GZ67]).
Definition 1.1.6. Let C be a model category. Then the homotopy category

of C is the localization of C with respect to the class of weak equivalences and
23
is denoted by γ : C −! Ho C . γc : Cc −! Ho Cc (resp. γf : Cf −! Ho Cf )
will denote the localization of Cc with respect to the class of maps in Cc (resp.
Cf ) which are weak equivalences in C . We sometimes use the notation [X, Y ] for
HomHo C (X, Y ).
Lemma 1.1.8. (i) Let F : C −! B carry weak equivalences in C into iso-

l r
morphisms in B. If f ∼ g or f ∼ g, then F (f ) = F (g) in B.
(ii) Let F : Cc −! B carry weak equivalences in Cc into isomorphisms in B.

r
If f ∼ g, then F (f ) = F (g) in B.
Proof. (i) Let h : A × I −! B be a left homotopy from f to g. As σ is a weak

equivalence, F (σ) is an isomorphism. As
F (σ)F (∂0 ) = F (σ)F (∂1 ) = idA =⇒ F (∂0 ) = F (∂1 )
and so
F (f ) = F (h)F (∂0 ) = F (h)F (∂1 ) = F (g).
(ii) The proof is the same same as (i) since by Lemma 1.1.4 (ii) we may assume
that s : B −! B I is a cofibration and hence B I is in Cc .
By Lemma 1.1.8 the functors γc , γf , γ induce functors γ c : πCc −! Ho Cc , γ f :

πCf −! Ho Cf and γ : πCcf −! Ho C , provided these localizations exist. The
following result shows that the homotopy category Ho C as defined in Definition 6
is equivalent to the more concrete category πCcf .
Theorem 1’. Ho C exists and the functor γ : πCcf −! Ho C is an equivalence

of categories.
This is included in the following more complex assertion which is presented

for the purpose of comparison with (Gabriel–Zisman [GZ67]).
24
Theorem 1.1. The categories Ho C , Ho Cc , Ho Cf exist and there is a diagram

of functors
γc
πCc Ho Cc
γ
πCcf Ho C (11)
γf
πCf Ho Cf
∼
where ,−! denotes a full embedding and −! denotes an equivalence of categories.
Furthermore, if (γ)−1 is a quasi–inverse for γ, then the fully faithful functor
∼ (γ)−1
Ho Cc −! Ho C −−−! πCcf ,−! πCc
∼
is right adjoint to γ c and the fully faithful functor
∼ (γ)−1
Ho Cf −! Ho C −−−! πCcf ,−! πCf
∼
is left adjoint to γ f .
Proof. For each object X choose a trivial fibration pX : Q(X) −! X with Q(X)
cofibrant and a trivial fibration iX : X −! R(X) with R(X) fibrant. We assume
that Q(X) = X and pX = idX (resp. X = R(X) and iX = idX ) if X is already
cofibrant (resp. fibrant). For each map f : X −! Y we may choose by M1 a
map Q(f ) : Q(X) −! Q(Y ) (resp. R(f )iX = iY f ) which is unique up to left
l
(resp. right) homotopy by Lemma 1.1.7. It follows that Q(gf ) ∼ Q(g)Q(f ) and
l r r
Q(idX ) ∼ idQ(X) , hence Q(gf ) ∼ Q(g)Q(f ) and Q(idX ) ∼ idQ(X) by Lemma 1.1.4(i)
and therefore X −! Q(X), f 7! Q(f ) is a well–defined functor which we shall
denote Q : C −! πCc . Similarly there is a functor R : C −! πCf .
25
r
If X is cofibrant, f, g ∈ Hom(X, Y ), and f ∼ g, then by Lemma 1.1.4(iii)
r
iY f ∼ iY g and hence R(f ) ∼ R(g) by the dual of Lemma 1.1.7. It follows that R
restricted to Cc induces a functor πCc −! πCcf and that there is a well–defined
functor RQ : C −! πCcf given by X −! RQX, f 7! RQ(f ).
Let Ho C be the category having the same objects as C with
HomHo C (X, Y ) = HomπCcf (RQX, RQY ) = π(RQX, RQY )
and the obvious composition. Let γ : C −! HoC be given by γ(X) = X,

γ(f ) = RQ(f ). As RQ(X) = X if X is in Ccf , it is clear that the functor
γ : πCcf −! Ho C induced by γ is fully faithful. By Lemma 1.1.7 and its dual,
trivial fibrations and trivial cofibration in Ccf become isomorphisms in πCcf ;
hence any weak equivalence in Ccf becomes an isomorphism in πCcf by M2 and
M5. If f : X −! Y is a weak equivalence in C , then f pX = pY Q(f ) and
M5 imply that Q(f ) is a weak equivalence in Cc and similarly RQ(f ) is a weak
equivalence in Ccf ; hence γ(f ) = RQ(f ) is an isomorphism. It follows that for
any X the maps
pX iQ(X)
X −− Q(X) −−−! RQX
γ
yield an isomorphism of X and RQ(X) in Ho C and hence πCcf −! Ho C is an
equivalence of categories.
We now show that γ : C −! Ho C has the required universal property of

Definition 1.1.5. As mentioned above γ carries weak equivalences in C into
isomorphisms in Ho C . Let F : C −! B do the same. Define θ : Ho C −! C
by θ(X) = F (X) and for α ∈ HomHo C (X, Y ) choose f : RQ(X) −! RQ(Y )
26
representing α and let θ(α) be given by the diagram
θ(α)
F (X) F (Y )
F (pX ) ∼ ∼ F (pY )
F (QX) F (QY ) (12)
F (iQX ) ∼ ∼ F (iQY )
F (f )
F (RQX) F (RQY )
By Lemma 1.1.8(i), θ(α) is independent of the choice of f and it is then clear

that θ is a functor, in fact the unique functor with θ ◦ γ = F . This proves the
existence of Ho C and also the horizontal equivalence in (11).
∼
The existence of Ho Cc and the equivalence πCcf −! Ho Cc can be proved in
the same way using the functor Cc −! πCcf induced by R and Lemma 1.1.8(ii).
The last assertion of the theorem results from the fact that the inclusion functor
πCcf ,−! πCc is right adjoint to the functor R′ : πCc −! πCcf , since
π r (X, Y ) ∼
= π(RX, Y ) if X is in Cc and Y is in Ccf by Lemma 1.1.7, and from
the fact that up to the equivalence Ho Cc ∼
= Ho C ∼ = πCcf , γ c : Cc −! Ho Cc “is”
the functor R′ .
Corollary 1. If A is cofibrant and B is fibrant, then
HomHo C (A, B) = π(A, B)
Proof.
HomHo C (A, B) = π(RQA, RQB) = π(RA, QB) ∼

= π(A, QB) ∼
= π(A, B)
by Lemma 1.1.7 and its dual.
27
Corollary 2. The functor γ c : πCc −! Ho Cc permits calculations by left frac-

tions and the functor γ f : πCc f −! Ho Cf permits calculation by right fractions.
Proof. This follows from the first chapter of [GZ67], since γ c has a fully faithful
right adjoint.
Remarks. 1. In general the localization C −! Ho C cannot be calculated by

either left or right fractions.
2. In example A., C = Cf and the usual homotopy relation on maps coincides

with homotopy in the sense of Definition 1.1.5 on Cc . Thus πCcf = πCc
is the homotopy category of cofibrant spaces which in turn is equivalent to
the usual homotopy category of CW complexes. In example B., C = Cf
and homotopy on Cc coincides with the chain homotopy relation. Hence
πCc = πCcf is what is denoted by K − (P) is Harshorne [Har66] where P
is the additive sub–category of projectives in A , while Ho C is the derived
category D− (A ) or D+ (A ).
3. The following example shows that although Ho C is determined by the

category C and the class of weak equivalences, the model structure on C
isn’t. Let A be an abelian category of finite homological dimension hav-
ing enough projectives and injectives. Then C = cb (A ) the category of
bounded complexes is what one should call a full sub–model category of
c+ (A ) as in example B.. The dual of example B. gives the structure of a
model category on c− (A ), the category of complexes bounded above, where
cofibrations are injections, fibrations are surjective maps with injective ker-
nels, and weak equivalences are homology isomorphisms. Again cb (A ) is a
full–sub–model category of c− (A ) and we obtain different model structures
on cb (A ) with the same family of weak equivalences.
28
Section 1.2: The loop and suspension functors
1.2 The loop and suspension functors
Homotopy theory is concerned not only with the category Ho C as a category

but also with certain extra structure which comes from performing constructions
in C . In this section we will be concerned with one aspect of this extra structure–
the loop and suspension functors.
C denotes a mixed model category and f, g : A ⇒ B two maps in C where A is
cofibrant and B is fibrant.
Definition 1.2.1. Let h : A × I −! B and h′ : A × I ′ −! B be two left homo-

topies from f to g. By a left homotopy from h to h′ we mean a diagram
h+h′
A × I ∨ A × I′ B
A∨A
j0 +j1 H (1)
σ+σ ′
A τ A×J
where j0 +j1 is a a cofibration and τ is a weak equivalence. (Here A×I ∨ A×I ′

A∨A
is the cofibre product of the maps ∂0 + ∂1 : A ∨ A −! A × I and
l
∂0′ + ∂1′ : A ∨ A −! A × I ′ .) We say h is left homotopic to h′ (notation ∼) if
such a left homotopy exists.
Remarks. 1. As in §1.1, the symbol A×J will denote an object of C together

with a cofibration j0 + j1 and weak equivalence τ as in (1). A × J is not
generally the product of A and an object “J”.
2. There is a dual notion of right homotopy of right homotopies whose

formulation we will leave to the reader.
Definition 1.2.2. Let h : A × I −! B be a left homotopy from f to g and let

k : A −! B I be a right homotopy from f to g. By a correspondence between h
29
and k we mean a map H : A × I −! B I such that H∂0 = k, H∂1 = sg, d0 H = h,

and d1 H = gσ. We say that h and k correspond if such a correspondence exists.
It will be useful to use the following diagrams to indicate a left homotopy h, a
right homotopy k, and a correspondence H between h and k respectively.
g g gσ g
f h g
k k H sg
f f h g
(2)
Lemma 1.2.1. Given A × I and a right homotopy k : A −! B I there is a left

homotopy h : A × I −! B corresponding to k. Dually given B I and h, there is a
k corresponding to h.
Proof. Same of that as Lemma 1.1.5(ii).
Lemma 1.2.2. Suppose that h : A × I −! B and h′ : A × I ′ −! B are two

left homotopies from f to g and that k : A −! B I is a right homotopy from f
to g. Suppose that h and k correspond. Then h′ and k correspond iff h′ is left
homotopic to h.
Proof. Let H : A × I −! B I be a correspondence between h and k, and let

H ′ : A × I −! B I be a correspondence between h′ and k. Let A × J, j0 + j1 , and
τ be as in Remark 1. The dotted arrow K exists in
30
H+H ′
A × I ∨ A × I′ BI
A∨A
K
j0 +j1 d1
gτ
A×J B
and d0 K : A × J −! B is a left homotopy from h to h′ . Conversely suppose
given H : A × I −! B I and a left homotopy K : A × J −! B from h to h′ . Then
j0 : A × I −! A × J is a cofibration by M3 since it’s the composition of j0 + j1
and
in1 : A × I −! A × I ∨ A × I ′
A∨A
which is the cobase extension of ∂0 + ∂1 . Also j0 is trivial by M5 since τ j0 = σ.

Therefore the dotted arrow φ exists in
H
A×I BI
φ
j0 (d0 ,d1 )
(K,gτ )
A×J B×B
and φj1 : A × I ′ −! B I is a correspondence between h′ and k.
Corollary. “is left homotopic to” is an equivalence relation on the class of left
homotopies from f to g and the equivalence classes form a set π1l (A, B; f, g).
Dually right homotopy classes of right homotopies form a set π1r (A, B; f, g). Cor-
respondence yields a bijection π1l (A, B; f, g) ≃ π1r (A, B; f, g)
Proof. Lemma 1.2.2 yields the equivalence relation assertion while Lemma 1.2.1
shows that every h is equivalent to a k : A −! B I with fixed B I and hence the
equivalence classes form a set. The last assertion is clear from Lemma 1.2.2 and
its dual.
By the corollary we may drop the “l” and “r” and write π1 (A, B; f, g) and
refer to an element of this set as a homotopy class of homotopies from f to g.
31
Again let C be a fixed model category, let A be a cofibrant object of C , and

let B be a fibrant object.
Definition 1.2.3. Let f1 , f2 , f3 ∈ Hom(A, B), let h : A × I −! B be a left

homotopy from f1 to f2 and let h′ : A × I ′ −! B be a left homotopy from f2
to f3 . By the composition of h and h′ , denoted h · h′ , we mean the homotopy
h′′ : A × I ′′ −! B given by h′′ in1 = h, h′′ in2 = h′ where A × I ′′ is the path object
constructed in Lemma 1.1.3. If f, g ∈ Hom(A, B) and h : A × I −! B is a left
homotopy from f to g, then by the inverse of h, denoted h−1 we mean the left
homotopy h′ : A × I ′ −! B from g to f , where A × I ′ is the path object for A
given by A × I ′ = A × I, ∂0′ = ∂1 , ∂1′ = ∂0 , σ ′ = σ and where h′ = h.
The following pictures for h · h′ and h−1 will be used.
f1 h f2 h′ f3
g h−1 f
(3)
Composition and inverses for right homotopies are defined dually and will be
pictured by diagrams like (3) but where the lines run vertically.
Proposition 1.2.1. Composition of left homotopies induces maps

π1l (A, B; f1 , f2 ) × π1l (A, B; f2 , f3 ) −! π1l (A, B; f1 , f3 ) and similarly for right ho-
motopies. Composition of left and right homotopies is compatible with the cor-
respondence bijection of the corollary of Lemma 1.2.2. Finally the category with
objects Hom(A, B), with a morphism from f to g defined to be an element of
π1 (A, B; f, g), and with composition of morphisms defined to be induced by com-
position of homotopies, is a groupoid, the inverse of an element of π1l (A, B; f, g)
represented by h being represented by h−1 .
32
Proof. Let h (resp. k) be a left (resp. right) homotopy from f1 to f2 , let h′ (resp.
k ′ ) be a left (resp. right) homotopy from f2 to f3 , and let H (resp. H ′ ) be a
correspondence between h and k (resp h′ and k ′ ). Then we have the following
correspondence between h · h′ and k · k ′ .
f3 σ f3 σ
k′ k′ σ k′ H′ s′ f3
f2 σ h′
k H sf2 sh′ sf3
h h′
Taking Lemma 1.2.2 into consideration this proves the first two assertions of the
proposition.
Composition is associative because (h · h′ ) · h′′ and h · (h′ · h′′ ) are both

represented by the picture
h h′ h′′
If h : A × I −! B from f to g and H : A × I −! B I is a correspondence of h

with some right homotopy k then the diagrams
33
gσ gσ gσ gσ
k kσ k H sg k H sg sgσ sg
fσ h h gσ
and Lemma 1.2.2 give f σ · h ∼ h, h · gσ ∼ h, proving the existence of iden-

tities and hence Hom(A, B) is a category. Finally let H ′ : A × I ′ −! B I be
H : A × I −! B I , where A × I ′ is A × I with ∂0′ = ∂1 , ∂1′ = ∂0 , and σ ′ = σ,
and let H ′′ : A × I ′ −! B I be a correspondence of h−1 : A × I ′ −! B with
some k ′′ : A −! B I , and let H̃ : A × I −! B be H ′′ . Then the diagrams
gσ gσ fσ fσ
sg H′ k H sg sf H
e k ′′ H ′′ sf
g h−1 f h g f h g h−1 f
show that h−1 ·h ∼ gσ and h·h−1 ∼ f σ providing the last assertion of Proposition
1.2.1.
It is clear that if i : A′ −! A is a map of cofibrant objects, then there is

a functor i∗ : Hom(A, B) −! Hom(A′ , B) which sends f into f i and a right
homotopy k : A −! B I into ki : A′ −! B I . Similarly if j : B −! B ′ is a map of
fibrant objects there is a function j∗ : Hom(A, B) −! Hom(A, B).
Lemma 1.2.3. The diagram

i∗
π1 (A, B; f, g) π1 (A, B; f i, gi)
j∗ j∗
i∗
π1 (A, B; jf, jg) π1 (A′ , B ′ , jf i, jgi)
34
commutes.
Proof. Let α ∈ π1 (A, B; f, g) and represent α with h : A×B −! B, k : A −! B I ,

and let H be a correspondence between h and k. By Lemma 1.1.5(ii) and Lemma
1.2.1 we may assume that σ : A × I −! A is a trivial cofibration. By M1 we can
choose dotted arrows in
∂0 i+∂1 i s′ j
A′ ∨ A′ A×I B (B ′ )I
φ ψ
∂0′ +∂1′ σ s (d′0 ,d′1 )
iσ ′ (jd0 ,jd1 )
A′ ∨ I A BI B′ × B′
Then H is a correspondence between jh and ψk; hence ψk represents j∗ α and so
ψki represents i∗ j∗ α. Similarly Hφ is a correspondence between ki and hφ; hence
hφ represents i∗ α and so jhφ represents j∗ i∗ α. Finally ψHφ is a correspondence
between ψki and jhφ which shows that i∗ j∗ α = j∗ i∗ α.
Definition 1.2.4. A pointed category is a category A in which “the” initial

object and final object exist and are isomorphic. We shall denote this object by
∗ and call it the null-object of A . If X and Y are arbitrary objects of A we
denote by 0 ∈ HomA (X, Y ) the composition X −! ∗ −! Y . If f : X −! Y is a
map in C , then we define the fibre of f to be the fibre product ∗ ×Y X and the
cofibre of f to be the fibre product ∗ ∨X Y .
By a pointed model category we mean a model category C which is also a
pointed category. If A is in C0 and B in Cf , then we will abbreviate π1 (A, B; 0, 0)
to π1 (A, B). π1 (A, B) is a group by the above proposition.
Theorem 1.2. Let C be a pointed model category. Then there is a functor

A, B −! [A, B]1 from (Ho C )◦ × Ho C to Grp1 which is determined up to canon-
ical isomorphism by [A, B]1 = π1 (A, B) if A is cofibrant and B is fibrant. Fur-
thermore, there are two functors from Ho C to Ho C , the suspension functor Σ
1 category of groups and homomorphisms
35
and the loop functor Ω and canonical isomorphisms
[ΣA, B] ≃ [A, B]1 ≃ [A, ΩB]
of functors (Ho C )◦ × Ho C −! Set2 where [X, Y ] = HomHo (X, Y )
π
Proof. Let A be cofibrant; choose a cylinder object A × I and let A × I −! ΣA
be the cofibre of ∂0 + ∂1 : A ∨ A −! A × I. By M3 ΣA is cofibrant. We shall
define a bijection
∼
ρ : π(ΣA, B) −! π1 (A, B) (4)
which is a natural transformation of functors to Set as B runs over Cf . Let

φ : ΣA −! B be a map and let ρ(φ) be the element of π1 (A, B) represented
by φπ : A × I −! B. If φ, φ′ ∈ Hom(ΣA, B) and φ ∼ φ′ , then there is a right
homotopy h : ΣA −! B I from φ to φ′ . Let H : A×I −! B I be a correspondence
of φ′ π with some right homotopy k from 0 to 0 and consider the diagram
0σ
k H s0
ϕ′ π
s0 hπ s0
ϕπ
2 Category of sets and functions
36
This shows that φπ commutes with s0·k and φ′ π corresponds to k, as s0·k and k
represents the same element of π1 (A, B) so do φπ and φ′ π and hence ρ(φ) = ρ(φ′ ).
This shows that ρ (4) is well-defined. ρ is surjective by Lemma 1.2.1. Finally,
if ρ(ϕ) = ρ(ϕ′ ), then, with the notation from Definition 1.2.1, there is a left
homotopy H : A × J −! B from φπ to φ′ π. Let H ′ : A × J −! B be given by
H ′ j0 = H ′ j1 = φπ and let K be the dotted arrow in
sφπ
A×I BI
K
j0 (d0 ,d1 )
(H,H ′ )
A×J (B, B)
(j0 was shown to be a trivial cofibration in proof of Lemma 1.2.2.) Then

Kj1 : A×I −! B I is a right homotopy from φπ to φ′ π such that Kj1 (∂0 +∂1 ) = 0
and so induces a right homotopy σA −! B I from φ to φ′ . This shows ρ is
injective and proves (4).
Dually if we choose a path object B I and let ΩB be the fibre of
(d0 , d1 ) : B I −! B × B, then ΩB is fibrant and there is a bijection
∼
π(A, ΩB) π1 (A, B) (5)
which is a natural transformation of functors as A runs over CC .

Lemma 1.1.3 shows that A, B −! π1 (A, B) is a functor (CC )◦ × Cf to Grp.
(4) and (5) combined with Theorem 1.1 and its first corollary show that this
functor induces a functor (Ho CC )◦ × Ho Cf to Grp, which then by Theorem 1.1
may be extended to a functor A, B −! [A, B]1 from (Ho C )◦ × Ho C to groups,
not uniquely but unique up to canonical isomorphism. By the first corollary of
Theorem 1.1 and (4) and (5) the bifunctor [·, ·]1 is representable in the first and
second variables which proves the theorem.
37
Remark. 1. Σ and Ω are adjoint functors on Ho C and are unique up to

canonical isomorphism. Also for any X, Σn X n ≥ 1 is a cogroup object
(resp. Ωn X is a group object) in Ho C , which is commutative for n ≥ 2.
2. We shall indulge in the abuse of notation of writing Σ for both the functors
on Ho C of Theorem 1.2 and writing σA for the cofibre of A ∨ A −! A × I
when A is in CC . If we should encounter a situation where this would lead
to confusion we shall denote the former use of Σ by LΣ because it’s kind of
a left-derived functor in the sense of §1.4 below. Similarly RΩ will be used
for the loop functor on Ho C if necessary.
1.3 Fibration and Cofibration Sequences
In this section we develop another part of the extra structure on Ho C , namely

the long exact sequences for fibrations and cofibrations and the Toda bracket
operation.
C denotes a fixed pointed model category in the following.
Let p : E −! B be a fibration where B is fibrant and let i : F −! E be the

inclusion of the fibre of p into E. F and E are fibrant by M3. Let
sB (dB ,dB )
B −! B I −−0−−−
1
!B×B
be a factorization of ∆B into a weak equivalence followed by a fibration. We shall

construct an object E I which is nicely related to B I .
Let E ×B B I (resp. B I ×B E) denote the fibre product of p : E −! B and
0 : B −! B (resp. d1 : B −! B), and let the fibre product sign ×B B to

dB I B I I
the left (resp. B I ×B to the right) of B I denote fibre products with dB

0 (resp.
38
Section 1.3: Fibration and Cofibration Sequences
1 ) in what follows. Let

dB
sE (dE ,pI ,dE )

E −! E I −−−
0
−−−−
1
! E ×B B I ×B E
be a factorization of (idE , sB p, idE ) into a weak equivalence followed by a fibra-

tion. The notation E I , sE , etc. is justified because sE is a weak equivalence
and (dE
0 , d1 ) is a fibration by M3 since it is the composition of (d0 , p , d1 ) and
E E I E
(pr1 , pr3 ) : E ×B B I ×B E −! E × E, which is the base extension of (dB

0 , d1 ) by
B
p × p. A similar argument shows that (dE

0 , p ) : E −! E ×B B and (p , d1 )
I I I I E
are fibrations.
The map pr1 : E ×B B I −! E is the base extension of dB

0 by p and hence is
a trivial fibration by M3 and M4. Hence by M5 the fibration
0 , p ) : E −! E ×B B is trivial since idE = pr1 ◦(d0 , p ) ◦ sE . The diagram

(dE I I I E I
pr2
F ×E E I ×E F EI
π (d0 ,pI ) (1)
i×j
F × ΩB E ×B B I
is Cartesian where π = (pr1 , j −1 pI pr2 ) and where j : ΩB ,−! B I is as in §1.2

the fiber of (dB
0 , d1 ). Here we are using the following convention which will be
B
used many times in this section.
Convention. If a : X −! Y is a monomorphism in a category and β : Z −! Y

is a map, then by α−1 β we mean the unique map γ : Z −! X with α ◦ γ = β, if
such a map exists.
Returning to the cartesian diagram (1) we have that π is a trivial fibration
39
by M4 and hence in Ho C (in fact in Ho Cf ) there is a map
m : F × ΩB −! F (2)
γ(π)−1 γ(pr )
given by the composition F × ΩB −−−−! F ×E E I ×E F −−−−3! F .
Proposition 1.3.1. The map m is independent of the choice of P I : E I −! B I

and is a right action of the group object ΩB on F in Ho C .
We first show that m may be defined in another way.

Recall that [X, Y ] = HomHo C (X, Y ) and [X, Y ]1 = [
P
X, Y ] = [X, ΩY ]
where these are the same respectively as π(X, Y ) and π1 (X, Y ) if X and cofibrant
and Y is fibrant.
Proposition 1.3.2. Let A be cofibrant and let the map

m∗ : [A, F ] × [A, ΩB] −! [A, F ] be denoted by α, λ −! α · λ. If α ∈ [A, F ]
is represented by u : A −! F , if λ ∈ [A, ΩB] = [A, B]1 is represented by
h : A × I −! B with h(∂0 + ∂1 ) = 0, and if h′ is a dotted arrow in
iu
A E
h′ p (3)
∂0
h
A×I B
then α · γ is represented by i−1 h′ ∂1 : A −! F .
Proof. Let H : A × I −! B I be a correspondence of h with k : A −! B I . Let

K be a lifting in
sE h′ ∂1
A EI
K
∂1 (dE I
0 ,p )
(h′ ,H)
A×I E ×B B I
40
Picture:
0 0σ d1 K∂0 d1 K h′ ∂1
pI
k H sB 0 K∂0 K sE h′ ∂1
0 h iu h′ h′ ∂1
Now K∂0 : A −! E I induces a map K∂0 : A −! F ×E E I ×E F such that

πK∂0 = (u, j −1 k) (see (1)) and hence by the definition of m we have that α · λ is
represented by i−1 dE
1 K∂0 : A −! F . But i d1 K : A × I −! F is a homotopy
−1 E
from i−1 dE
1 K∂0 to i h ∂1 and this proves the Proposition.
−1 ′
Proof of Prop. 1.3.1. Diagram (3) is clearly independent of pI so m is indepen-

dent of pI by Prop 1.3.2. On the other hand, let α, λ, u, h, h′ be as in Prop 1.3.2,
let λ1 ∈ [A, B]1 be represented by h1 : A × I −! B and let h′1 be a dotted arrow
in the first diagram
h′ ∂1 iu
A E A E
′
h′1 h ·h′1
∂0 p ∂0′ p
h h·h1
A×I B A × I′ B
so that i−1 h′1 ∂1 represents (α · λ) · λ1 by Prop. 1.3.2. As the composite homotopy

h·h1 represents λ·λ1 , the second diagram and Prop. 1.3.2 show that i−1 (h′ ·h′1 )∂1′
represents α · (λ · λ1 ). But (h′ · h′1 )∂1′ = h′1 ∂1 hence (α · λ) · λ1 = α · (λ · λ1 ) and
m is an action as claimed.
41
Definition 1.3.1. By a fibration sequence in Ho C we mean a diagram in

Ho C of the form
X −! Y −! Z X × ΩZ −! X
which for some fibration p : E −! B in Cf is isomorphic to the diagram
i p m
F −! E −! B F × ΩB −! F (4)
constructed above.
Remark 1. By dualizing the above construction one may construct a diagram
A −! X −! C C −! C ∨ ΣA
starting from a cofibration u in Cc , where v : X −! C is the cofibre of u and n is

a right co-action of the cogroup ΣA on C, and define the notion of a cofibration
sequence in Ho C .
Proposition 1.3.3. If (4) is a fibration sequence so is
∂ i n
ΩB −! F −! E ΩB × ΩE −! ΩB (5)
0,id m
where ∂ is the composition ΩB −−! F × ΩB −! F and where
n∗ : [A, ΩB] × [A, ΩE] −! [AΩB] is given by (λ, u) −! ((Ωp)∗ u)−1 · λ.
Proof. We may assume that (4) is the sequence constructed above from a fibration
p. Let pI : E I −! B I be as in the definition of m. Then
pr1 : E ×B B I ×B (∗) −! E is the base extension of (dB
0 , d1 ) by
B
(p, 0) : E −! B × B and hence is a fibration; so we get a fibration sequence
(0,j,0) pr n
ΩB −−−−! E ×B B ×B (∗) −−!
1
E ΩB × ΩE −! ΩB. (6)
42
We calculate n by Proposition 1.3.2; let λ ∈ [A, ΩB] be represented by

u : A −! ΩB, let µ ∈ [A, ΩE] be represented by h : A × I −! E and let (h, H, 0)
be a lifting in
(0,ju,0)
0σ
A E ×B B I ×B (∗)
∂0 pr1 ju H∂1
(h,H,0)
h
A×I E
ph
where H : A × I −! B I is pictured at the right. By Prop. 1.3.2, j −1 H∂1

represents n∗ (λ, µ) in [A, ΩB]. Letting H ′ : A × I −! B I be a correspondence
of H∂1 with h′ : A × I −! B, we obtain the correspondence
0σ 0σ
ju H H′ sB 0
ph h′
of ju with ph · h′ , which shows that
λ = (Ωp)∗ µ · n∗ (λ, µ) or n∗ (λ, µ) = [(Ωp)∗ µ]−1 · λ.
Thus the map n in (6) is the same as that in (5).
(i,0,0)
The map f −−−−! E ×B B I ×B (∗) is a weak equivalence by M5 since it may
(sE i,id)
be factored F −−−−−! E I ×E F = E I ×B (∗) −! E ×B B I ×B (∗) where the
43
(dE ,pI )
second map is a trivial fibration (base extension of E I −−−
0
−−! E ×B B I and
pr
where the first map is a section of the trivial fibration E I ×E F −−!
2
F (base
extension of dE
1 .) We shall show that the diagram in Ho C
ΩB
(0,j,0)
∂ (7)
(i,0,0)
F E ×B B I ×B (∗)
commutes. Let λ ∈ [A, ΩB] be represented by k : A −! B I and let

H : A×I −! B I be a correspondence of k with h. Then ∂∗ α = 0·α is represented
by i−1 h′ ∂1 : A −! F where h′ is the dotted arrow in
0
A E
h′ p
∂0
h
A×I B
So (i, 0, 0)∗ ∂∗ λ is represented by
(h′ ∂1 ,0,0)
A −−−−−−! E ×B B I ×B (∗),
(0,k,0)
(0, j, 0)∗ λ is represented by A −−−−! E ×B B I ×B (∗), and
(h′ , H, 0) : A×I −! E ×B B I ×B (∗)
is a left homotopy between these maps, showing that the triangle (7) commutes in
Ho C . As pr1 ◦(i, 0, 0) = i we see that idΩB , (i, 0, 0), and idE give as isomorphism
of (5) with the fibration sequence (6), and so by definition (5) is a fibration
sequence.
44
Proposition 1.3.4. Let (4) be a fibration sequence in Ho C , let ∂ : ΩB −! F

be defined as in Propostion 1.3.3 and let A be any object of Ho C . Then the
sequence
(Ωq ∂)∗ (Ωq i)∗ (Ωq p)∗

... [A, Ωq+1 B] [A, Ωq F ] [A, Ωq E] ...
(Ωp)∗ (∂∗ ) i∗ p∗
... [A, ΩE] [A, ΩB] [A, F ] [A, E] [A, B]
is exact in the following sense:
(i) (p∗ )−1 {0} = Im i∗
(ii) i∗ ∂∗ = 0 and i∗ α1 = i∗ α2 ⇐⇒ α2 = α1 · λ for some λ ∈ [A, ΩB]
(iii) ∂∗ (Ωi)∗ = 0 and ∂∗ λ1 = ∂∗ λ2 ⇐⇒ λ2 = (Ωp)∗ µ · λ1 for some µ ∈ [A, ΩE]
(iv) The sequence of group homomorphisms from [A, ΩE] to the left is exact in
the usual sense.
The dual proposition for cofibration sequences is
Proposition 1.3.4’. Let
u v n
X
A −! X −! C C −! C ∧ A
be a cofibration sequence in Ho C and let ∂ : C −! A be (idC +0) ◦ n. If B is

P
any object in Ho C , then the sequence
v)∗ ( u)∗
P P
( P P ∂∗ v∗ u∗
[ X, B] [ A, B] [C, B] [X, B] [A, B]
45
is exact in the sense that (i) - (iv) hold with i∗ , p∗ , ∂∗ replaced by v ∗ , u∗ , ∂ ∗ and
where the · in (ii) refers to the right action n∗ : [C, B] × [ A, B] −! [C, B].
P
Proof if Prop. 1.3.4. We may assume (4) is the sequence constructed from the
fibration p.
(i) Clearly pi = 0. If p∗ α = 0 represennt α by u : A −! E, let h : A × I −! B

be such that h∂0 = pu, h∂ = 0. By the covering homotopy theorem (dual
of Corrolary of Lemma 1.1.2) we may cover h by k : A × I −! E with
∂0 k = u. Then if β is represented by i−1 k∂1 we have i∗ β = α.
(ii) With the notation of Prop. 1.3.2, we have that h′ is a homotopy from iu
which represents i∗ α to h′ ∂1 which represents i∗ (α·λ). Hence i∗ (α·λ) = i∗ α
and in particular
i∗ ∂∗ λ = i∗ (0 · λ) = i∗ 0 = 0,
so i∗ ∂∗ = 0. Conversely given α1 α2 with i∗ α1 = i∗ α2 , represent α1 by

ui , i = 1, 2, let h : A × I −! E be such that h∂0 = iu1 , h∂1 = iu2 whence
if λ is the class of ph, α1 · λ = α2 by Prop 1.3.2.
(iii) follows from (ii) and Proposition 1.3.3
(iv) follows by repeated use of Proposition 3.
Proposition 1.3.5. The class of fibration sequences in Ho C has the following

properties:
(i) Any map f : X −! Y may be embedded in a fibration sequence
f
F −! X −! Y, F × ΩY −! F.
46
(ii) Given a diagram of solid arrows
i p m
F E B F × ΩB F
γ β α γ×Ωα γ
i′ p′ m′
F′ E′ B′ F ′ × ΩB ′ F′
(8)
where the rows are fibration sequences, the dotted arrow γ exists.
(iii) In any diagram (8) where the rows are fibration sequences, if α and β are
isomorphisms so is γ.
(iv) Proposition 1.3.3.
Remark 2. Proposition 1.3.4 gives the analogues for fibration sequences of

all non-trivial axioms for the triangles in a triangulated category (see [Ver] or
[Har66]) except the octahedral axiom. The analogue of that axiom holds also,
but as far as the author knows, it’s not worth the trouble required to write it
down.
Proof. (i) Any map in Ho C is isomorphic to a fibration of objects in Ccf .
(iii) If A is any object in Ho C , then Prop. 1.3.4 gives a diagram
[A, ΩE] [A, ΩB] [A, F ] [A, E] [A, B]
S S Y∗ S S
[A, ΩE ′ ] [A, ΩB ′ ] [A, F ′ ] [A, E ′ ] [A, B ′ ]
where the rows are “exact” in the sense that (i)-(iii) of Prop. 1.3.4 hold.
However this is enough to conclude by the usual 5-lemma argument that
γ∗ : [A, F ] −! [A, F ′ ] is a bijection for all A and hence γ is an isomorphism.
47
(ii) We may suppose by replacing the diagram (8) by an isomorphic diagram if

necessary that the rows are constructed in the standard way from fibrations
u
p and p′ in Cf . Let B
e −! B be a trivial fibration with B
e cofibrant and let
v
e −! e be a trivial fibration with E
e cofibrant. By M4
E ExB B
e −! E is a trivial fibration and pr2 : ExB B
pr1 : ExB B e −! B
e is a fibration
so we obtain a diagram
i pr2 v
Fe E B
e
e e
ε pr1 v u
F i
E p B
in C , where pr1 v and u are weak equivalences. It follows easily from the
calculation given in Prop. 1.3.2, that
m
Fe × ΩB
e e
Fe
m
F × ΩB F
commutes. Hence by (iii) the sequence ∼ is isomorphic to first row of (8)

and so we may suppose that the rows of (8) are not only constructed in the
standard way form fibrations p and p′ but that E and B are in Ccf . Then by
Theorem 1.1 α and β are represented by maps u and v in C with p′ v ∼ up.
As E is cofibrant, we may by the corollary of Lemma 1.1.2, modify v, so
that p′ v = up. Then we may take γ : F −! F ′ in (8) to be the map in C
induced by v. The first part of (8) commutes clearly and the second square
may be shown to commute in Ho C by use of Proposition 1.3.2. This proves
(ii)
48
The dual proposition for cofibration sequences is left to the reader.
The following proposition will be used in the definition of the Toda bracket.
Proposition 1.3.6. Let
u v ∂′ n
A X C ΣA C C ∨ ΣA
f
β γ δ
∂ i p m
ΩB F E B F × ΩB F
(9)
be a solid arrow diagram in Ho C where the first row except ∂ is a cofibration ′
sequence, and where the second row except for ∂ is a fibration sequence. We
suppose that ∂ ′ = (idC +0) ◦ n and ∂ = m ◦ (0, idΩB ) as in Proposition 1.3.4 and
1.3.4’. Suppose that f u = 0 and pf = 0. Then dotted arrow α, β, γ, δ exist and
the set of possibilities for α formas a left Ωp∗ [A, ΩE] - right u∗ [X, ΩB] double
coset in [A, ΩB] and the set of possibilities for δ forms a left (Σu)∗ [ΣX, B] -
right p∗ [ΣA, E] double coset in [ΣA, B]. Furthermore under the identification
[A, ΩB] = [ΣA, B] the first coset is the inverse of the second.
Proof. By Prop. 1.3.4

pf = 0 ⇒ ∃β : X −! F
with f = iβ. Similarly iβu = 0 ⇒ ∃α with ∂α = βu. Hence α, β exist. Suppose

that α′ , β ′ are other maps. By the exact sequence of Prop. 1.3.4 β ′ = β · λ for
some λ ∈ [X, ΩB]. More precisely β ′ = m ◦ (β, λ) hence
∂α′ = β ′ u = m ◦ (β, λ)u = m(βu, λu) = m(∂α, λu) = ∂α · (λu) = (0 · α) · λu = 0(α · λu) = ∂(α · λu).
By exactness
α′ = (Ωp)∗ µ · α · λu = (Ωp)∗ µ · α · u∗ (λ)
49
and so α′ lies in the double coset Ωp∗ [A, ΩE] · α · u∗ [X, ΩB]. As µ and λ may
be arbitrary we see that any element of this double coset may be an α′ . Dual
assertions hold for γ and δ and so the first statement of the proposition is proved.
To prove the second assertion we must construct α, β, γ, δ so that α corre-

sponds to δ −1 . We may assume that u is a cofibration of cofibrant objects, that p
is a fibration of fibrant objects and that the top and bottom rows of (1) are con-
structed as above. In this case Theorem 1.1 shows that the map f in Ho C may
be represented by a map in C which we shall denote again by f. Now pf ∼ 0 and
as X is cofibrant and u is a fibration we may by the corrollary to Lemma 1.1.2 lift
this homotopy to E and so assume that pf=0. (We may not, however, simultane-
ously assume that f u = 0.) Let h : A × I −! E be such that h∂0 = f u, h∂1 = 0
and consider the following diagram
u in1 0+q+0
A X X ∨A A × I ∨A (∗) ΣA
i−1 f
f
f +h+0 phq −1 (10)
p
F i
E B
where q : A × I −! ΣA is the cokernel of A ∨ A −! A × I and where we extend

to epimorphisms the convention for morphisms introduced at the beginning of
this section so that phq −1 is the unique map such that (phq −1 )q = ph. Now the
top line of (10) is isomorphic in Ho C to the top line of the first part of (9) –
see the proof of Proposition 1.3.3 especially the homotopy commutativity of (7)
for the dual considerations. Consequently by means of this isomorphism we may
define β in (9) to be represented by i−1 f, γ by f + h + 0, and δ by phq −1 . But
50
we also have the diagram
fu
A E
h p , h∂1 = 0
∂0
ph
A×I B
which by Prop. 1.3.2 shows that βu · δ = 0 since i−1 f u represents βu in (1).

Hence βu = 0 · δ −1 = ∂(δ −1 ) and we may take α in (1) to be δ −1 .
u f p
Definition 1.3.2. Let A −! X −! E −! B be three maps in Ho C such that
f u = pf = 0. Form a solid arrow diagram
u v ∂ n
A X C ΣA C C ∨ ΣA
f
γ δ
p
E B
(11)
by choosing by Prop. 1.3.5(i) for the first row a cofibration sequence containing
u, and then fill in the dotted arrows as in Prop. 1.3.6. The set of possibilities for
δ is as in Prop. 1.3.6 a left (Σu)∗ [ΣX, B] -right p∗ [ΣA, E] double coset in [ΣA, B]
which is called Toda bracket of u, f, and p, and is denoted ⟨u, f, p⟩.
Remark 3. 1. The Toda bracket is independent of the choice of the top row
of (3) by Prop. 1.3.5(ii) and (iii).
2. The Toda bracket ⟨u, f, p⟩ may also be computed by choosing a solid arrow
51
diagram
u
A X
f
α β
∂ i p m
ΩB F E B F × ΩB F
(12)
where the bottom row comes from a fibration sequence, and filling in the
dotted arrows. By Proposition 1.3.6 we have
(Σu)∗ [ΣY, B] · α−1 , p∗ [ΣA, B] ⊆ [ΣA, B].
1.4 Equivalences of homotopy theories
We begin with some general categorical considerations.
Definition 1.4.1. Let γ : A −! A′ and F : A −! B be two functors. By the

left-derived functor of F with respect to γ we mean a functor Lγ F : A′ −! B
with a natural transformation ε : Lγ F ◦ γ −! F having the following universal
property: Given any G : A′ −! B and natural transformation ζ : G ◦ γ −! F
there is a unique natural transformation Θ : G −! Lγ F such that
G◦γ
ζ
Θ∗γ F (1)
ε
Lγ F ◦ γ
commutes.
Remark. 1. Lγ F is the functor from A′ to B such that Lγ F ◦ γ is closest

to F from the left. Similarly we may define the right-derived functor
52
Section 1.4: Equivalences of homotopy theories
of F with respect to γ to be “the” functor Rγ F : A′ −! B with a natural

transformation η : F −! Rγ F ◦ γ which is closest to F from the right.
2. The terminology left-derived functor comes from Verdier’s treatment of

homological algebra[Ver]. In that case A is the category K(A), where A is
an abelian category, γ is the localization K(A) −! D(A), F : K(A) −! B
is a cohomological functor from K(A) to an abelian category B and Lγ F ,
Rγ F are what Verdier calls the left and right derived functors of F .
3. We shall be concerned only with the case where A is a model category C

and γ is the localization functor γ : C −! Ho C. In this case we will write
just LF .
4. If C is a model category and F : C −! B is a functor then it is clear

that ε : LF ◦ γ −! F is an isomorphism if and only if F carries weak
equivalences in C into isomorphisms in B. In this case we may assume that
LF is induced by F in the sense that LF is the unique functor Ho C −! B
with LF ◦ γ = F . Moreover RF = LF .
Proposition 1.4.1. Let F : C −! B be a functor where C is a model cate-

gory. Suppose that F carries weak equivalences in CC into isomorphisms in B.
Then LF : Ho C −! B exists. Furthermore ε(X) : LF (X) −! F (X) is an
isomorphism if X is cofibrant.
Proof. Let X −! Q(X), f −! Q(f ), pX : Q(X) −! (X) be as in the proof

of theorem 1.1, so that Q induces a well-defined functor Q : C −! πCC . By
Lemma 1.1.8(ii), X −! F QX, f −! F Q(f ) is a functor F Q : C −! B which
induces a functor LF : Ho C −! B since Q(f ) is a weak equivalence if f is. Let
ε : LF ◦ γ −! F be the natural transformation given by
ε(X) = F (pX ) : F QX −! F X. To show that ε has the universal property of
definition 1.4.1, let ζ : G ◦ γ −! F where G : Ho C −! B. Define
53
Θ(X) : G(X) −! LF (X) to be the composition
G(γ(pX ))−1 ζ
G(X) −−−−−−−−! GQX −
! F QX = LF (X).
It is clear that Θ is a natural transformation G ◦ γ −! LF ◦ γ, and since every

map is Ho CC is a finite composition of maps γ(f ) or γ(s)−1 , Θ is a natural
transformation Θ : G −! LF . The diagram
G(γ(pX )−1 ) ζ ∼
GX GQX F QX LF (X)
idGX
G(pX ) F (pX )
ε
ζ
GX FX
shows that ε(Θ ∗ γ) = ζ. The uniqueness of Θ : G −! LF is clear since it is
determined by on Ho CC = Ho C and so ε has the required universal property.
Finally if X is cofibrant LF X = F QX = F X and ε(X) = idF (X) .
Definition 1.4.2. Let F : C −! C′ be a functor where C and C′ are model

categories. By the total left-derived functor of F we mean the functor
LF : Ho C −! Ho C′ given by LF = Lγ (γ ′ ◦ F ) where γ : C −! Ho C and
γ ′ : C′ −! Ho C′ are the localization functors.
Remark. The diagram
f
C C′
γ γ′ (2)
LF
Ho C Ho C′
does not commute, but rather there is a natural transformation

ε : LF ◦ γ −! γ ′ ◦ F such that the pair (LF, ε) comes as close to making (1)
commutative as possible.
54
Corollary. If F carries weak equivalence in Cc into weak equivalences C′ , then

LF : Ho C −! Ho C′ exists and ε(X) : L(X) −! F (X) is an isomorphism in
Ho C′ for X cofibrant.
Proposition 1.4.2. Let C and C′ be pointed model categories with suspension

functors Σ and Σ′ on Ho C and Ho C′ , respectively. Let F : C −! C′ be a functor
which is right exact (i.e. compatible with finite inductive limits), which carries
cofibrations in C into cofibrations in C′ , and which carries weak equivalences in
Cc into weak equivalences in C′ . Then LF is compatible with finite direct sums,
there is a canonical isomorphism of functors LF ◦ Σ ≃ Σ′ ◦ LF , and with respect
to this isomorphism LF carries cofibration sequences in Ho C into cofibration
sequences in Ho C′ .
Proof. LF exists by proposition 1.4.1 and we may assume that LF (A) = F (A) if
A is cofibrant. If A1 and A2 are in Cc then A1 ∨ A2 , the direct sum of A1 and A2
in C, is also the direct sum of A1 and A2 in Ho C. By assumption F (Cc ) ⊂ C′c
and so
LF (A1 ∨ A2 ) = F (A1 ∨ A2 ) = F (A1 ) ∨ F (A2 ) = LF (A1 ) ∨ LF (A2 )
where the last v means direct sum in Ho C′ . This proves the first assertion about
F.
Next observe that if A is cofibrant, then for a given object A × I we have that
F (∂0 )+F (∂1 ) F (σ)

F (A) ∨ F (A) −−−−−−−−! F (A × I) −−−! F (A)
is a factorization of ∇F (A) into the cofibration F (∂0 ) + F (∂1 ) = F (∂0 + ∂1 )

followed by the weak equivalence F (σ). Hence F (A × I) = F (A) × I and since
F is compatible with cofibre products F (ΣA) = ΣF (A). As F (A) is cofibrant
Σ(F (A)) represents Σ(F (A)) in Ho C and so the second assertion is proved.
55
i×I
Finally note that if i : A −! B is a cofibration in Cc and A × I −−! B × I
is a compatible choice in the dual sense that pI : E I −! B I was a compatible
choice in §1.3, then F (A × I) −! F (B × I) is also a compatible choice for
F A × I −! F B × I. It follows that F carries the diagram in Cc
I q in ξ
A −! B −! C C −−!
1
C ∨ B ×I ∨ C −− C ∨ A
B B
where ξ is a weak equivalence into a similar diagram with A replaced by F A, etc.

This proves the last assertion about LF .
Theorem 1.3. Let C and C′ be model categories and let
C C′
R
be a pair of adjoint functors, L being the left and R the right adjoint functor.
Suppose that L preserves cofibrations and that L carries weak equivalences in
Cc into weak equivalences in C′ . Also suppose that R preserves fibrations and
that R carries weak equivalences in C′f into weak equivalences in C. Then the
functors
L(L)
Ho C R(R)
Ho C′
are canonically adjoint.

Suppose in addition for X in Cc and Y in Cf that a map LX −! Y is a weak
equivalence if and only if the associated map X −! RY is a weak equivalence.
Then the adjunction morphisms id −! L(L) ◦ R(R) and R(R) ◦ L(L) −! id are
isomorphisms so the categories Ho C and Ho C′ are equivalent. Furthermore if C
and C′ are pointed then these equivalences L(L) and R(R) are compatible with
the suspension and loop functors and the fibration and cofibration sequences in
56
Ho C and Ho C′ .
Proof. For simplicity we write L instead of L(L) and we use Grothendieck’s

notation u♭ : X −! RY (resp. v ♯ : LX −! Y ) to denote the map corresponding
to u : LX −! Y (resp v : X −! RY ). If X is in Cc and Y is in Cf , then we
saw in the proof of proposition 1.4.1 that L(X × I) = LX × I. Hence to any left
homotopy h : X × I −! RY between f and g there corresponds the homotopy
H ♭ : LX × I −! Y between f ♭ and g ♭ and so [X, RY ] = [LX, Y ]. Hence if
X 7! Q(X) etc. is as in the proof of theorem 1.1 and Y 7! R′ (Y ), f 7! R′ (f ),
iY : Y −! R′ (Y ) is the functor-up-to-homotopy of theorem 1.1 for the category
C′ we have the isomorphisms:
HomHo C′ (LX, Y ) ≃ [LQX, R′ Y ] ≃ [QX, RR′ Y ] ≃ HomHo C (X, RY ), (3)
where the first and last isomorphisms come from the construction of L and R
given above in proposition 1.4.1. The isomorphisms (3) are clearly functorial as
(X, Y ) runs over C0 ×C1 , and hence as every map in Ho C is a finite composition
of maps of the form γ(f ) or γ(s)−1 , (3) is functorial as (X, Y ) runs over
(Ho C)0 × (Ho C′ ) proving that L and R are adjoint.
Suppose now that for X in Cc and Y in C′f , f : X −! RY is a weak
(iLX )♭
equivalence iff f ♯ : LX −! Y is a weak equivalence so X −−−−! RR′ (LX) is a
weak equivalence. But by propostion 1.4.1, RR′ LX = RLX and by examining
(3) we see that γ((iLX )♭ ) : X −! RR′ (LX) is the adjunction map X −! RLX.
∼ ∼
Hence X −! RLX for all X in Ho Cc and hence in Ho C. Similarly LR −! id
which proves the second assertion of the theorem.
If C and C′ are pointed we have by proposition 1.4.2 and its dual LΣ ≃ Σ′ L
and ΩR ≃ RΩ′ ,. Hence
RΣ′ ≃ RΣ′ LR ≃ RLΣR ≃ ΣR
57
and similarly L preserves loop functors. Also by proposition 1.4.2 L preserves

cofibration sequences and R preserves fibration sequences. Suppose that
i p n
ε = {F −! E −! B, ΩB × F −! F }
is a fibration sequence in Ho C. Then we may embed the map LE −! LB in a

fibration sequence ε′ of Ho C′ by proposition 1.3.5 (i) and the image Rε′ of the
sequence under R is a fibration seuqnece which is isomorphic to ε by Proposition
1.3.5, (ii) and (iii). Hence ε′ ≃ Lε and L preserves fibration sequences. Similarly
R preserves cofibration sequences.
Examples. 1. Let A be an abelian category with enough projectives and

injectives and let C and C′ be the two model categories which have Cb (A)
as underlying category described in Remark 3 following theorem 1.1. Then
the identity functor gives a pair of adjoint functor
C C′
satisfying the conditions of the theorem. The theorem implies that cofi-
bration and fibration sequences constructed from both categories coincide
which is clear since they coincide with Verdier’s triangles.
2. Let C′ = (spaces) C = (ss sets) as in examples A. and C. and let L be the

geometric realization functor, and R the singular complex functor. Then
theorem 1.3 applies because of [Mil57] and so the cofibration sequences in
the homotopy categories of ss sets of spaces coincide. This is not entirely
trivial since the singular functor does not commute with the operation of
taking the cofibre of a map.
Remark. We recall our vague definition of the homotopy theory associated to a
58
Section 1.5: Closed model categories
model category, namely the category Ho C with all extra structure which comes
by performing constructions in C. In §1.2 and §1.3 we gave the most important
examples of that extra structure and Theorem 1.3 gives a criterion which shows
when the homotopy theories coming from different model categories coincide, at
least when only the structure of §1.2 and §1.3 is concerned. There are other kinds
of structure, e.g. higher order ([Ver],[Spa63]) operations, which ate not included
in theorem 1.3, and it seems reasonable to conjecture that this extra structure is
preserved under the conditions of theorem 1.3.
1.5 Closed model categories
We will say that a map i : A −! B has the left lifting property with
respect to a class S of maps in a category C if the dotted arrow exists in any
diagram of the form
A X
i f (1)
B Y
where f is in the class S. Similarly f has the right lifting property with
respect to S if the dotted arrow exists in any diagram of the form (1) where i is
in S.
Definition 1.5.1. A model category C is said to be closed if it satisfies the

axiom
M6 Any two of the following classes of maps in C – the fibrations, cofibrations,

and weak equivalences – determine the third by the following rules:
(a) A map is a fibration ⇐⇒ it has the right lifting property with respect
to the maps which are both cofibrations and weak equivalences
59
(b) A map is a cofibration ⇐⇒ it has the left lifting property with respect
to the maps which are both fibrations and weak equivalences.
(c) A map f is a weak equivalence ⇐⇒ f = uv where v has the left

lifting property with respect to the class of fibrations and u has the
right lifting property with respect to the class of cofibrations.
Remarks. 1. It is clear that M6 implies M1, M3, and M4. Hence a closed
model category may be defined using axioms M0, M2, M5, and M6.
2. Examples A., B. and C. of §1.1 are all closed model categories (see proposi-
tion 1.5.2 below). Model categories which are not closed may be constructed
by reducing the class of cofibrations but keeping M2, M3 and M4 valid. For
example, take example B., §1.1, where A is the category of left R modules,
R a ring, and define cofibrations to be injective maps f in C+ (A) such that
Coker f is a complex of free R modules.
In the following C is a fixed model category and we retain the notations of

the previous sections.
Lemma 1.5.1. Let p : X −! Y be a fibration Ccf . The following are equivalent.
(i) p has the right lifting property with respect to the cofibrations.
(ii) p is the dual of a strong deformation retract map in the following prcise
sense: there is a map t : Y −! X with pt = idY and there is a homotopy
h : X × I −! X from tp to idX with ph = pσ.
(iii) γ(p) is an isomorphism.
Proof.
60
(i) =⇒ (ii) One lifts successively in
tp+idX
∅ X X ∨X X
t h
P ∂0 +∂1 f
pσ
Y Y X ×I Y
idY
(ii) =⇒ (i) Let pI : X I −! γ I be a compatible choice of path objects for X

and Y as in the beginning of §1.2 and let Q be a lifting in
sX
X XI
∂1 (dX I X
0 ,p ,d1 )
Q
(h,sY ,pσ,σ)
X ×I X ×Y Y I ×Y X
Then k = Q∂0 : X −! X I is a right homotopy from tp to idX with

pI k = sY p. Given the first diagram
kα
A α
X A XI
ϕ H
i p i (dX I
0 ,p )
β (tβ,sY β)
B Y B X ×γ γ I
the dotted arrow ϕ may be constructed by choosing a dotted arrow H in

the second and setting ϕ = dX
1 H.
(ii) =⇒ (iii) t is a homotopy inverse for p, hence p is a homotopy equivalence

and γ(p) is an isomorphism.
(iii) =⇒ (ii) By Theorem 1.1 γ(p) an isomorphism =⇒ p is a homotopy

equivalence and there is a map t : Y −! X with pt ∼ idY and tp ∼ idX .
61
By the covering homotopy theorem we may assume that pt = idY . Let

q : X × I −! X be a left homotopy from tp to idX . Then the composite
homotopy q −1 · tpq : X × I ′ −! X from idX to tp covers the composite
homotopy (pq)−1 · pq : X × I ′ −! Y from p to p. However proposition
1.2.1 implies that (pq)−1 · (pq) is left homotopic to pσ : X × I −! Y , that
is, there exists H : X × J −! Y with Hj0 = pσ and Hj1 = (pq)−1 · pq
where X × J, j0 , j1 , τ are as in (1) with A replaced by X. By a covering
homotopy argument which takes the form
(pq)−1 ·(pq)
X ×I X
j1 f
K
X ×J H
Y
we obtain a left homotopy Kj0 : X×I −! Y from idX to tp with pKj0 = pσ

whose inverse is the desired homotopy h.
Definition. A map f : X −! Y is said to be a retract of a map f ′ : X ′ −! Y ′

if there is a diagram
X′
f′
X idX
X
f Y′ f
idY
Y Y
Proposition 1.5.1. Let C be a closed model category and let f be a map in C.

Then γ(f ) is an isomorphism iff f is a weak equivalence.
62
Proof. The direction ⇐ is the basic property of γ, so we suppose that γ(f ) is an

isomorphism. By M5 and M2 we reduce to the case where f is a fibration Ccf
whence the result follows from the above lemma and M6(c).
Proposition 1.5.2. Let C be a model category. Then C is closed if and only

if each of the classes of fibrations, cofibrations, and weak equivalences has the
property that any retract of a member of the class is again a member.
Proof.
⇐= Let p : X −! Y have the lifting property (1) whenever i is a trivial cofi-

i u
bration. By M2 we may factor p into X −! Z −! Y where i is a trivial
cofibration and u is a fibration. By the property of p there is a dotted arrow
s in
id
X X
s p
i
u
Z Y
It follows that p is a) retract of the fibration u and hence that p is a fibration.

This proves a since M1 gives the ⇒ direction of M6 a), and the proof of
b) is similar. Suppose that f = uv as in c. Then by the above argument
u is a retract of a trivial cofibration and hence by assumption is a weak
equivalence. Similarly v is a weak equivalence and so f is also. This proves
c since the implication ⇒ is contained in M2, M5, and M1. So C is closed.
=⇒ It is immediate that a retract of a map with a lifting propety of the kind in

M6 a) b) c) again has that lifting property. Thus the classes of fibrations
and cofibrations, are closed under retracts. Let γ : C −! Ho C be the
canonical localization functor and suppose that f is a retract of a weak
63
equivalence. Then γ(f ) is a retract of an isomorphism and hence is an

isomorphism so f is a weak equivalence by proposition 1.5.1.
64
2. Examples of simplicial homotopy the-
ories
2.0 Introduction
The first four sections of Chapter II give some examples of model categories.
In §2.3 it is shown how the categories of topological spaces, simplicial groups, and
simplicial sets form model categories, and in §2.4 this result is extended to the
category sA of simplicial objects over a category A, where A is a category closed
under finite limits having sufficiently many projective objects and satisfying one
of the following additional assumptions:
(i) A has sufficiently many cogroup objects,
(ii) A is closed under arbitrary inductive limits and has a set of small projective
generators.
The proofs for topological spaces, simplicial groups, and sA when A satisfies (ii)
are similar and fairly simple, since every object in the model category is fibrant.
For simplicial sets we were unable to find a really elementary proof; the argument
given, which we think is the simplest, uses the classification theory of minimal
fibrations [BGM59]. It is possible to give another argument using the functor
Ex∞ of Kan [Kan57a]and a variant of this argument is used for sA in case (ii).
65
Chapter 2: Examples of simplicial homotopy theories
All of these categories are what we call simplicial categories, i.e. categories
C endowed with a simplicial set “function complex” HomC (X, Y ) for each pair
of objects X and Y satisfying suitable conditions. In §2.1 we define simpli-
cial categories and the generalized path and cylinder functors X, K 7! X ⊗ K,
Y, K 7! Y K , K a simplicial set , by the formulas
HomS (K HomsA (X, Y )) = HomsA (X ⊗ K, Y ) = HomsA (X, Y K )
where S is the category of simplicial sets. In §2.2 we define closed simpli-

cial model category which is a category having the structures of a simplicial
category and a closed model category compatibly related. All the examples of
Chapter 2 are closed simplicial model categories; moreover, for these model cat-
egories there are canonically adjoint path and cylinder functors, so much of the
work of the first chapter simplifies considerably (see [Kan57b]). However, there
are certain categories of differential graded algebras that do not seem to have
natural simplicial structures but which are model categories, which is the main
reason for the generality in Chapter 1.
In §2.5 we show under suitable assumptions how homology and cohomology

for model categories may be defined using abelian group objects and the abelian-
ization functor. In particular, we define cohomology groups of an object X with
values in an abelian group object A of a model category C. When C is the cat-
egory of simplicial objects in a category A and X and A are constant simplicial
objects, we show that these cohomology groups are equivalent to those obtained
from suitable cotriples and Grothendieck sheaves. We also indicate how this co-
homology gives a cohomology theory for arbitrary universal algebras coinciding
up to a dimension shift with usual cohomology in the case of groups, and Lie
algebras and associative algebras over a field.
In §2.6 we show that the category of simplicial modules over a simplicial ring
66
Section 2.1: Simplicial categories
forms a model category and use this to derive several Kunneth spectral sequences
which will be used in later applications.
The present framework for homotopical algebra is not the most general that
can be imagined. We have restricted ourselves to categories A closed under finite
limits and having sufficiently many projective objects. The sheaf cohomology
of Grothendieck is defined much more generally and Artin–Mazur [AM67] have
shown in the case of the etale topology for preschemes that it gives rise to an
analogue of ordinary homotopy theory using pro-objects in a homotopy category.
It would also be nice to weaken the hypothesis that finite limits exist on a model
category so the category of 2-connected pointed topological spaces would become
a model category. Finally further generalization might eliminate the following
inadequacy of this theory, that although derived functors may be defined for any
category A with finite limits and enough projectives, the category sA does not
form a model category without additional assumptions.
2.1 Simplicial categories
Simp will denote the category of (semi-) simplicial sets (see [GZ67]).
Definition 2.1.1. A simplicial category is a category C endowed with the

following structure:
(i) a functor X, Y 7! HomC (X, Y ) from Cop × C to Simp,
(ii) maps in Simp
HomC (X, Y ) × HomC (Y, Z) −! HomC (X, Z)
f, g 7! g ◦ f
called composition defined for each triple X, Y, Z of objects of C,
67
(iii) an isomorphism
∼
HomC (X, Y ) −! HomC (X, Y )0
u 7! u
e
of functors from Cop × C to Set.
This structure is subject to the following two conditions:
(1) If f ∈ HomC (X, Y )n , g ∈ HomC (Y, Z)n and h ∈ HomC (Z, W )n , then
(h ◦ g) ◦ f = h ◦ (g ◦ f ).
(2) If u ∈ HomC (X, Y ) and f ∈ HomC (Y, Z), then f ◦sn0 u

e = HomC (u, Z)n (f ).
Also sn0 U
e ◦ g = HomC (W, u)n (g) if g ∈ HomC (W, X)n .
Definition 2.1.2. Let C1 and C2 be simplicial categories. By a simplicial

functor F : C1 −! C2 we mean a functor F from C1 to C2 together with maps
HomC1 (X, Y ) −! HomC2 (F X, F Y ), denoted f 7! F (f ), such that
F (f ◦ g) = F (f ) ◦ F (g) and F (e
u) = Fg
(u).
Example. If X and Y are simplicial sets, let HomSimp (X, Y ) or simply Hom(X, Y )
be the “function complex” simplicial set of maps from X to Y . There is a canon-
ical “evaluation map”
ev : X × Hom(X, Y ) −! Y (1)
giving rise to isomorphisms
#
Hom(K, Hom(X, Y )) −! Hom(X × K, Y ) (2)
∼
for all K ∈ Ob Simp, where #(u) = ev ◦(idx ×u). The map
ev × id ev
X × Hom(X, Y ) × Hom(Y, Z) −−−−! Y × Hom(Y, Z) −! Z
68
thereby determines a composition map (ii), while taking K = ∆(0), the final
object of Simp, in (2) yields an isomorphism (iii). It is easily seen that Simp is
a simplicial category.
If X is a fixed object of Simp then the functor Y 7! Hom(X, Y ) is a simpli-
cial functor hX , where hX : Hom(Y, Z) −! Hom(Hom(X, Y ), Hom(X, Z)) is
given by #(hX ) = composition.
In the following C denotes a simplicial category. When convenient we will

identify HomC (X, Y ) with HomC (X, Y )0 and drop the “∼” notation. Also we
will often write Hom(X, Y ) instead of HomC (X, Y ).
Definition 2.1.3. Let X ∈ Ob Simp. By X ⊗ K we shall denote an object of

C with a distinguished map α : K −! HomC (X, X ⊗ K) such that
∼
ϕ : HomC (X ⊗ K, Y ) −! HomSimp (K, HomC (X, Y )) (3)
for all Y ∈ Ob C, where #(ϕ) is the map
α×id ◦
K ×Hom(X ⊗K, Y ) −−−! Hom(X, X ⊗K)×Hom(X ⊗K, Y ) −! Hom(X, Y ).
By X K we denote an object of C with a map β : K −! HomC (X K , X) such

that
∼
ψ : HomC (Y, X K ) −! HomSimp (K, HomC (Y, X)) (4)
for all Y ∈ Ob C, where #(ψ) is the composite
(pr ,β pr ) ◦
K × Hom(Y, X K ) −−−2−−−−
! Hom(Y, X K ) × Hom(X K , X) −! Hom(Y, X).
1
Example. If C = Simp, then X × K together with the map

α : K −! Hom(X, X × K) such that #(α) = idX×K is an object X ⊗ K.
Hom(K, X) with the map β : K −! Hom(Hom(K, X), X) such that #β = the
69
composite
(pr ,pr ) ev
Hom(K, X) × K −−−2−−−
! K × Hom(K, X) −! X
1
is an object X K .
Proposition 2.1.1. If X ∈ Ob C and K, L ∈ Ob Simp, then there are canonical

isomorphisms
X ⊗ (X × L) ≃ (X ⊗ K) ⊗ L (X K )L ≃ X K×L (5)
when all the objects are defined.
Proof.
Hom(X ⊗ (K × L), Y ) ≃ Hom(K × L, Hom(X, Y ))
≃ Hom(L, Hom(K, Hom(X, Y )))
≃ Hom(L, Hom(X ⊗ K, Y ))
≃ Hom((X ⊗ K) ⊗ L, Y ).
This yields the first isomorphism; the second is proved similarly.
Remarks.
1. The degree-0 part of (3) yields the formula
HomC (X ⊗ K, Y ) ≃ HomSimp (K, HomC (X, Y )). (6)
The difference between (6) and (3) is roughly the first isomorphism of (5)
as on sees by analyzing the proof of (5). In practice (see Prop. 2.1.2 below)
one defines an operation X ⊗ K satisfying (6) and (5) and then proves (3)
70
by inverting the proof of (5).
2. The objects X ⊗ K and X K have the following interpretation whose de-

tails we leave to the reader. The functor Y 7! Hom(X, Y ) is a simplicial
functor hX from C to Simp in a natural way. Call a simplicial func-
tor F : C −! Simp representable if it is isomorphic to hX for some
X ∈ Ob C. (Yoneda’s lemma holds: HomF (hX , F ) ≃ F (X) where F is the
simplicial category of simplicial functors from C to Simp.) Then X ⊗ K
represents the simplicial functor Y 7! Hom(K, Hom(X, Y )).
Let π0 (K) be the set of components of the simplicial set K so that we have
adjoint functors
HomSimp (K, K(S, 0)) ≃ HomSet (π0 (K), S) (7)
where if S is a set K(S, 0) denotes the constant simplicial set which is S in each
dimension and has all simplicial operators = idS . If x, y ∈ K0 we say that x is
strictly homotopic to y if there is a z in K1 with d1 z = x and d0 z = y and that
x is homotopic to y if x and y are equivalent with respect to the equivalence
relation generated by the relation “is strictly homotopic to”. π0 (K) is the quotient
of K0 by the relation “is homotopic to” and hence
∼
π0 (K × L) −! π0 (K) × π0 (L). (8)
Let J denote a generalized unit interval, that is, a simplicial set which is
a string of copies of ∆(1) joined end to end. Let {0} ⊂ J and {1} ⊂ J be the
subcomplexes generated by the first and last vertices of J. A typical J may be
pictured
0. 1.
71
and it is clear that two simplices x and y of K are homotopic if there exists a
generalized unit interval J and a map u : J −! K with u(0) = x and u(1) = y.
Definition 2.1.4. Let X, Y be two objects of C and f , g two maps from X

to Y . We say that f is strictly homotopic (resp. homotopic) to g if this is
the case when f and g are regarded as 0-simplices of Hom(X, Y ). By a strict
homotopy (resp. homotopy) from f to g we mean an element h ∈ Hom(X, Y )1
with d1 h = f and d0 h = g (resp. a map u : J −! Hom(X, Y ) with u(0) = f and
u(1) = g). Let π0 (X, Y ) = π0 Hom(X, Y ) be the homotopy classes of maps from
X to Y . We define the category π0 C to be the category with the same objects
as C, with Homπ0 C (X, Y ) = π0 (X, Y ), and with composition induced from the
composition in C (this is legitimate by (8)).
When objects X ⊗ K and X K exist in C, then a homotopy from f to g is

the same as a map H : X ⊗ J −! Y with Hi0 = f and Hi1 = g. Here J is a
generalized unit interval and ie : X −! X ⊗ J denotes the map induced by the
0-simplex e of J where e = 0 or 1. The homotopy may also be identified with a
map H ′ : Y −! Y J with j0 H ′ = f and j1 H ′ = g where je : X J −! X is induced
by e ∈ J0 . The reader will note that we have changed notation from ∂, d of Ch. 1
to i, j. This is because d0 corresponds to i1 . However we will retain the notation
s : X −! X J and σ : X ⊗ J −! X to denote the constant homotopy of idX .
These are the maps induced by the unique map J −! ∆(0).
Let A be a category and let sA be the category of simplicial objects over
A, that is, contravariant functors ∆ −! A, where ∆ is the category having for
objects the ordered sets [n] = {0, 1, . . . , n} for each integer n ≥ 0, and where a
map ϕ : [p] −! [q] in ∆ is a (weakly) monotone map. If X is an object of sA,
we write Xn instead of X[n] and ϕ∗X (or simply ϕ∗ ) for X(ϕ) when ϕ is a map in
∆. If X, Y are objects of sA and if K is a simplicial set, then ([θ], [β]) a map
f : X × K −! Y is defined to be a collection of maps f (σ) : Xq −! Yq , one for
72
each q ≥ 0 and σ ∈ Kq , such that ϕ∗Y f (σ) = f (ϕ∗K σ)ϕ∗X for any map ϕ in ∆.
X × K is not to be understood as an object of sA and f is not a morphism in a
category. Letting Map(X × K, Y ) be the set of maps f : X × K −! Y we obtain
a functor
(sA)op × Simpop × (sA) −! Set
and hence a functor X, Y 7! HomsA (X, Y ) from (sA)op × (sA) to Simp given
by
HomsA (X, Y )n = Map(X × ∆(n), Y )
with simplicial operator ϕ∗ = Map(X × ϕ,

e Y ). Here ∆(n) is the “standard n-
simplex” simplicial set, which is the functor ∆op −! Set represented by [n], and
for any simplicial set K and σ ∈ Kn we let σ
e : ∆(n) −! K be the unique map
in Simp with σ
e(id[n] ) = σ.
If X, Y, Z ∈ Ob sA and K is a simplicial set, then we map define the composite
g ◦ f of two maps f : X × K −! Y and g : Y × K −! Z by (g ◦ f )(σ) = g(σ)f (σ).
This yields a composition operations as in (ii) of Def. 2.1.1, and (iii) comes from
the fact that ∆(0)q consists of exactly one element for each q. It is clear that
sA thereby becomes a simplicial category. Also if the functor F : A −! B is
extended degree-wise to sF : sA −! sB, then sF is a simplicial functor where if
f : X × K −! Y we let
(sF )(f ) : F (X) × K −! F (Y ), [(sF )(f )](σ) = F (f (σ)).
Recall that a simplicial set is said to be finite if it has only finitely many
non-degenerate simplices. A finite simplicial set is always a simplicial finite set,
i.e. a simplicial object over the category of finite sets, but not conversely.
Proposition 2.1.2. Let A be a category and let X be a simplicial object over

A. If A is closed under (finite) direct sums, then X ⊗ K exists in sA for every
73
simplicial (finite) set K. If A is closed under (finite) projective limits, then X K

exists for every (finite) simplicial set K.
Proof. Let
_ X
(X⊠K)n = Xn with ϕ∗X⊠K = in(ϕ∗K σ) ϕ∗X .
σ∈Kn σ
Here Xi denotes the direct sum of an indexed family {Xi ; i ∈ I} of objects of

W
i∈I
A, ini : Xi −! Xi is the injection of the ith component, and fi : Xi −! Y

W P W
is the unique map with ( fi ) inj = fj for all j ∈ I if {fi : Xi −! Y, i ∈ I} is a

P
family of maps in A. These direct sums exist by the assumptions on A and K.

Let ξ : X × K −! X⊠K be given by ξ(σ) = inσ . Finally let
ev′ : X × HomsA (X, Y ) −! Y, ev′ (fn ) = fn (id[n] ) : Xn −! Yn .
Then there are isomorphisms
#′ ξ∗
HomSimp (K, HomsA (X, Y )) −! Map(X × K, Y ) − HomsA (X⊠K, Y )
∼ ∼
where #′ is induced by ev′ and ξ ∗ by ξ. Letting # = (ξ ∗ )−1 ◦ (#′ ) it is clear that

# is functorial as X, Y run over sA and K varies over the category of simplicial
(finite) sets. If L is another simplicial (finite) set, then there is a canonical
isomorphism
∼
θ : X⊠(K × L) −! (X⊠K)⊠L
given by
X
θn = inτ inσ .
(σ,τ )∈(K×L)n
Now if α : K −! Hom(X, X⊠K) is given by #(α) = idX⊠K , then X⊠K

with α is an object X ⊗ K. In effect letting ϕ be the map (3) determined by α,
74
we have the diagram
ϕ∗
Hom(L, Hom(X⊠K, Y )) Hom(L, Hom(K, Hom(X, Y )))
# ∼
# ∼ Hom(K × L, Hom(X, Y ))
# ∼
∗
θ
Hom((X⊠K)⊠L, Y ) ∼ Hom(X⊠(K × L), Y )
which may be shown to be commutative by a straightforward analysis of the

definitions. Taking L = ∆(n) for each n we see that ϕ is an isomorphism and
hence the first part of the proposition is proved.
Let A′ be the category of functors Aop −! Set and let X 7! hX be the

canonical fully faithful functor (this forces us to leave the haven of our universe).
Denoting the degree-wise extension of h by h : sA −! sA′ , one sees that
MapsA (X, Y ) ≃ MapsA (hX × K, hY ),
so
HomsA (X, Y ) ≃ HomsA′ (hX, hY ) (9)
and h is a “fully faithful” simplicial functor. Now if F ∈ Ob sA′ , then F K exists

and is given by F K (A) = F (A)K for all A ∈ Ob A, where we have identified
sA′ with the category of functors Aop −! Simp in the natural way. One sees
immediately from (9) that X K exists if and only if (hX)K is isomorphic to hZ
for some Z ∈ Ob sA, or equivalently if [(hX)K ]n is a representable functor for
each n. There is a cokernel diagram in Simp
_ _
∆(qj ) ∆(pi ) K × ∆(n)
j∈J i∈I
75
where if K is finite so is K × ∆(n) and hence I and J are finite sets. But the
functor
A 7! [(hX)K ]n (A) = HomSimp (K × ∆(n), hX(A))

nY Y o
= Ker hXpi (A) ⇒ hXqj (A)
I J
nY Y o
= h Ker hXpi ⇒ hXqj (A)
I J
is representable by the assumptions made on A.
Corollary. If F : A −! B commutes with (finite) direct sums (resp. projective

∼
limits), then F (X) ⊗ K −! F (X ⊗ K) for all X ∈ Ob sA and simplicial (finite)
∼
sets K (resp. F (X K ) −! F (X)K for all X and (finite) simplicial sets K).
This is immediate from the formulas for X ⊗ K and X K obtained in the proof
of Prop. 2.1.2.
Remark. The corollary implies that if G is a simplicial group then the underlying
simplicial set of GK is (underlying simplicial set of G)K , and similarly for any
other algebraic species.
2.2 Closed simplicial model categories

•
∆(n) for n ≥ 0 (resp. V (n, k) for 0 ≤ k ≤ n > 0) denotes the simplicial subset
of ∆(n) which is the union of the images of the faces ∂i : ∆(n − 1) −! ∆(n) for
˙ = ∅ the initial object in S. In the
0 ≤ i ≤ n (resp. 0 ≤ i ≤ n, i ̸= k). ∆(0)
following, RLP (resp. LLP) stands for right (resp. left) lifting property ( §1.5).
Proposition 2.2.1. The following are equivalent for a map f in S.

•
(i) f has the RLP with respect to ∆(n) ,−! ∆(n) for all n
76
Section 2.2: Closed simplicial model categories
(ii) f has the RLP with respect to any injective (i.e. injective in each degree)
map of simplicial sets.
This follows immediately from the skeletal decomposition of an injective map

(see [GZ67, Ch. II, 3.8]). The following is proved in [GZ67, Ch. IV, §2.1].
{e} ⊂ ∆(1) denotes the subcomplex consisting of the degeneracies of the vertex
e, where e = 0, 1.
Proposition 2.2.2. The following are equivalent for a map f in S.
(i) f has the RLP with respect to V (n, k) ,−! for 0 ≤ k ≤ n > 0
(ii) f has the RLP with respect to
•
∆(n) × ∆(1) ∪ ∆(n) × {e} ,−! ∆(n) × ∆(1)
for n ≥ 0 and e = 0, 1.
(iii) f has the RLP with respect to
L × ∆(1) ∪ K × {e} ,−! K × ∆(1)
for all injective maps L ,−! K in S and e = 0, 1.
Definition 2.2.1. A map of simplicial sets will be called a trivial fibration

(resp. fibration) if it satisfies the equivalent conditions in Proposition 2.2.1
(resp. Proposition 2.2.2).
Thus a fibration is a fiber map in the sense of Kan. It is easy to see that a
trivial fibration is a fibration whose fibers are contractible.
Definition 2.2.2. By a closed simplicial model category we mean a closed

model category C which is also a simplicial category satisfying the following two
conditions:
77
SM0 If X ∈ Ob C, then the objects X ⊗ K and X K exist for any finite simplicial
set K.
SM7 If i : A −! B is a cofibration and p : X −! Y is a fibration, then
(i∗ ,p∗ )
Hom(B, X) −−−−! Hom(A, X) × Hom(B, Y ) (1)
Hom(A,Y )
Convention. It will be convenient to use the notation Hom(i, p) for the target
of the map (1).
Proposition 2.2.3. Suppose that C is a simplicial category satisfying M0 and

SM0 with four distinguished classes of maps–fibrations, cofibrations, trivial fi-
brations, and trivial cofibration–such that the first and fourth (resp. second and
third) determine each other by lifting properties as in M6(a) and (b). (This holds
in particular if C is a closed simplicial model category). Then SM7 is equivalent
separately to each of the following:
SM7(a) If X −! Y is a fibration (resp. trivial fibration), then
X ∆(n) −! X ∆(n) × Y ∆(n)

•
Y ∆(n)
is a fibration (resp. trivial fibration) and X ∆(1) −! X {e} × Y {e} Y ∆(1) is a

trivial fibration for e = 0, 1.
SM7(b) If A −! B is a cofibration (resp. trivial cofibration), then
•
A ⊗ ∆(n) ∨ B ⊗ ∆(n) −! B ⊗ ∆(n)
•
A⊗∆(n)
is a cofibration (resp. trivial cofibration, and
A ⊗ ∆(1) ∨ B ⊗ {e} −! B ⊗ ∆(1)

A⊗{e}
78
is a trivial cofibration for e = 0, 1).
Proof. To show that X K −! X L ×Y L Y K is a fibration where L −! K is a

map of simplicial sets, it suffices to show that it has the RLP with respect to any
trivial cofibration A −! B. By the definition of the object X K this is equivalent
to showing that
Hom(B, X) −! Hom(A, X) × Hom(B, Y )

Hom(A,Y )
has the RLP with respect to L −! K. Manipulating in this way one proves the
proposition.
Remark. It is clear that SM7(a) holds for the fibrations and trivial fibrations in
S.
For the rest of this section C denotes a closed simplicial model category. We
shall be concerned with relating the simplicial homotopy structure of C with the
s.s s
left and right homotopy structure of Ch. 1. Let f ∼ g (resp. f ∼ g) mean f
is strictly (simplicially) homotopic (resp. (simplicially) homotopic) to g. The
following is the covering homotopy extension theorem for simplicial homotopies.
It should be noted how much stronger it is when than the Cor. of Lemma 1.1.2
and Lemma 1.1.7.
Proposition 2.2.4. Let i : A −! B be a cofibration and let p : X −! Y be a

fibration. Let h : A ⊗ J −! X and h : B ⊗ J −! Y be simplicial homotopies
compatible with i and p in the sense that pk = h(i ⊗ idJ ).
(1) If θ : B −! X satisfies pθ = hj0 , θi = ki0 , then there is a homotopy

H : B ⊗ J −! X with Hi0 = θ, pH = h, and H(i ⊗ idJ ) = k.
(2) If either i or p is trivial and if θe : B −! X satisfies pθe = hie , θi = kie ,

e = 0, 1, then there is a homotopy H : B ⊗ J −! X with Hie = θe , e = 0, 1,
79
pH = h, and H(i ⊗ idJ ) = k.
Proof. This follows immediately from SM7 by an induction on the length of J.
Corollary. Let i : A −! B be a cofibration of fibrant objects. Then i is trivial

iff i is a strong deformation retract map (i.e. there exists r : B −! A,
h : B ⊗ ∆(1) −! B with ri = idA , h0 = idB , h1 = ir, h(i ⊗ ∆(1)) = iσ)). Dually
if p : X −! Y is a fibration of cofibrant objects, then p is trivial iff there are
maps s : Y −! X, h : X ⊗ ∆(1) −! X with ps = idY , h0 = idX , h1 = sp,
ph = σ(p ⊗ ∆(1)).
Proof. ( =⇒ ) r and h may be obtained by lifting successively in

id A si
A A A BI
r h (j0 ,j1 )
i
(idB ,ir)
B e B B×B
( ⇐= ) is clear from Proposition 2.2.4
Proposition 2.2.5. (1) If f, g : X ⇒ Y are two maps in C, then
s l
f ∼ g =⇒ f ∼ g
r
and f ∼ g. If X cofibrant and Y is fibrant, then the strict simplicial, left,
and right homotopy relations on Hom(X, Y ) coincide and are equivalence
relations.
(2) The conclusions of Theorem 1.1,§1.1 remain valid if πCC , πCf , and πCcf
are replaced by π0 (Cc ), π0 (Cf ), and π0 (Ccf ), respectively.
Proof. (2) The inclusion {0} ⊂ J has the LLP with respect to fibrations in S,
hence if X is cofibrant one finds, as in the proof of Prop. 2.2.3(b), that
i0 : X −! X ⊗ J is a trivial cofibration. By M5 the map σ : X ⊗ J −! X
80
i +i
is a weak equivalence. also by Prop 2.2.3(b) X ∨ X −−
0
−!
1
X ⊗ J is a
cofibration and so X ⊗ J is a cylinder object for J. It follows as in the
s
proof of Lemma 1.1.8 that if f, g : X ⇒ Y are two maps in Cc and f ∼ g,
then γc (f ) = γc (g) and hence γc induces γ c : π0 Cc −! Ho Cc . Similarly
one shows that γ, γ f as in Theorem 1.1, exist with π replaced by π0 . Next
note that the “quasi-” functors X 7! Q(X) and X 7! R(X) of the proof of
this theorem yield functors Q : π0 C −! π0 Cc , R : π0 C −! π0 Cf in virtue
of Prop. 2.2.4 (2) above. The rest of the proof of Theorem 1.1 goes through
without change so (2) follows.
(1) The quasi-inverse of γ : π0 Ccf −! Ho C constructed in the proof of Theo-

rem 1.1 is induced by RQ : C −! π0 Ccf . But we have just seen that
s s
f ∼ g =⇒ RQ(f ) ∼ RQ(g)
and therefore we conclude that
s
f ∼ g =⇒ γ(f ) = γ(g).
Now if J is a generalized unit interval, there is a canonical homotopy

h : J × J −! J with h(idJ ×e
0) = idJ and h(idJ ×e
I) = idJ where
e : ∆(0) −! J is the map with e
e e(id[0] ) = e, e = 0, 1 and σ is the unique
map J −! ∆(0). This homotopy in a representative case may be pictured
where the arrows denote the direction of each 1 simplex of J × J and where
a simplex of J × J labelled as s0 a goes to s0 a in J under h. Consequently if
X is any object of C, σ : X ⊗ J −! X is a simplicial homotopy equivalence
and therefore γ(σ) is an isomorphism. By 1.5.1, σ is a weak equivalence
s l s
and therefore f ∼ g =⇒ f ∼ g. Similarly X −! X J is a weak equivalence
s r
for all X in C so f ∼ g =⇒ f ∼ g; thus this first part of (1) is proved. The
81
1 s0 1 s0 1 1
s0 b s0 b
b s0 1
s1 b s0 b
h 0 a b 1
s1 b s0 b
a s0 1
s1 b s1 b
0 a b 1
last assertion follows from Lemma 1.2.1 which shows when X is cofibrant
and Y is fibrant the cylinder object X ⊗ ∆(1) (see proof of (2) above) may
be used to represent any left homotopy from f to g and from Lemma 1.1.4.
Remark. Propostion 2.2.5 shows that the simplicial homotopy relation of

Hom(X, Y ) is finer than either left or right homotopy, but when X is cofibrant and
Y is fibrant the three relations coincide. One may compare the constructions of
§1.2 and 1.3 with the correspending well-known simplicial constructions and show
that the resulting structure on Ho C is the same. This the fundamental groupoid
of the Kan complex Hom(X, Y ) coincides with the one constructed in §1.2, and
if E −! B is a fibration in Cf where C is pointed, then the long exact sequence
of homotopy groups arising from the fibration Hom(A, E) −! Hom(A, B) (SM7
when A ∈ Ob Cc ) coincides with that of §1.3.
Proposition 2.2.6. If C is a closed simplicial model category, then in a natural

way so are the dual Cop and the category C/X of the objects of C over a fixed
82
Section 2.3: Topological spaces, simplicial sets, and simplicial
groups
object X.
Proof. The assertion about Cop is trivial. If A and B are two objects of C/X,
we let HomC/X (A, B) be the subcomplex of HomC (A, B) consisting of elements
fn of dimension n with (sn0 ) ◦ f = sn0 u, where u : A −! X and v : B −! X are
the structural maps. With the induced composition C/X becomes a simplicial
category closed under finite limits. If K is a finite simplicial set, then the object
u σ(u⊗id)
(A −! X)⊗K in C/X is the map A⊗K −−−−−! X, where σ : X⊗K −! X is the
map corresponding to the map K −! Hom(X, X) sending all elements of K to
u
degeneracies of idX . The objects (A −! X)K in C/X is the map pr2 : AK ×X K X,
whose source is the fiber product of uK and the map s : X −! X K corresponding
to σ. Thus C/X satisfies SM0.
A map in C/X will be called a fibration, cofibration or weak equivalence if

it is so in C. Axioms M2 and M5 are clear if i : A −! A′ and p : B ′ −! B
are maps in C/X, then the map HomC/X (A′ , B ′ ) −! HomC/X (i, p) is the base
extension by the structural map ∆(0) −! HomC (A′ , X) of the map
HomC (A′ , B ′ ) −! HomC (i, p). Hence SM7 holds, hence also M1. To obtain M6
argue as follows: Supposing a map f in C/X has the LLP with respect to the
fibrations in C/X, factor f = pi where i is a trivial cofibration and p is a fibration
C/X; Then f is a retract of i hence is a trivial cofibration in C and hence in
C/X. The other cases of M6 are similar.
2.3 Topological spaces, simplicial sets, and simplicial groups
Let Top be the category of topological spaces and continuous maps. If X and
Y are spaces, define the function complex Hom(X, Y ) by
Hom(X, Y )n = Hom(X × |∆(n)|, Y )
83
with natural simplical operations, where | | denotes geometric realisation. If

f ∈ Hom(X, Y )n and g ∈ Hom(Y, Z), let g ◦ f be the composite map
id ×∆ f ×id g
X × |∆(n)| −−−−! X × |∆(n)| × |∆(n)| −−−! Y × |∆(n)| −! Z.
Top thereby becomes a simplical category where X ⊗ K = X × |K| and

X k = the function space X |K| .
A map f : X ! Y in Top will be called a fibration if it is a fiber map in the
sense of Serre and a weak equivalence if it is a weak homotopy equivalence (i.e.
∼
πq (X, x) −! πq (Y, f x) for all x ∈ X and q ≥ 0). Finally a map will be called a
cofibration if it has the LLP with respect to all trivial fibrations.
Theorem 2.1. With these definitions the category Top of topological spaces is
a closed simplical model category.
Let Sing : Top ! Simp be the singular complex functor so that
HomSimp (K, Sing X) = HomTop (|K|, X) (1)
(Actually Sing and | | are adjoint simplical functors which means that Hom can
be replaced by Hom in (1).)
Lemma 2.3.1. The following are equivalent for a map f in Top.
(i) f is a fibration.
(ii) Sing f is a fibration in Simp.
(iii) f has the RLP with respect to |V (n, k)| ,−! |∆(n)| for 0 ≤ k ≤ n > 0.
Proof. (ii) and (iii) are equivalent by (1), and (i) and (iii) are equivalent since
|V (n, k)| ,−! |∆(n)| is isomorphic in Top to I n−1 × 0 ,−! I n .
Lemma 2.3.2. The following are equivalent for a map f in Top.
84
groups
(i) f is a trivial fibration.
(ii) Sing f is a trivial fibration in Simp.

•
(iii) f has the RLP with respect to ∆(qD ) ,−! |∆(n)| for n ≥ 0.
•
Proof. (ii) and (ii) are equivalent by (1). As ∆(qD ) ,−! |∆(n)| is isomorphic in
Top to S n−1 ⊂ Dn (where S −1 = ϕ if n = 0), the equivalence of (i) and (iii)
becomes a standard obstruction theory argument which we omit.
Corollary. In Top every object is fibrant and the fibrations and trivial fibrations
satisfy SM7(a).
Proof. Since Sing(X |K| ) = (Sing X)K , SM7(a) for Simp implies SM7(a) for Top.
Lemma 2.3.3. Any map f may be factored f = pi where i is a cofibration and

p is a trivial fibration.
Proof. Letting f : X ! Y we construct a diagram

j0 j1
X Z0 Z1 ···
p0
f p1
Y
as follows. Let Z −1 = X and p−1 = f , and having obtained Z n−1 , consider the
set D of all diagrams D of the form
•
αD
|∆(qD )| z n−1
pn−1
βD
|∆(qD )| Y
and define jn : Z n−1 ! Z n by a co–cartesian diagram
85
W • W
|∆(qD )| |∆(qD )|
D∈D D∈D
ΣαD in2
jn =in1
z n−1 zn
β
Define pn : Z n ! Y by pn jn = pn−1 , pn in2 = σD , let Z = lim Z n , p = lim pn
−! −!
and i = lim jn ◦ . . . ◦ j0 . By Lemma 2.3.1 jn has the LLP with respect to trivial
−! •
fibrations, hence i does too and so i is a cofibration. Now as ∆(qD ) is compact
•
any map α : ∆(qD ) ! Z factors through Z m for m sufficiently large. In effect
the well–known argument works because all the points of Z − i(X) are closed.
•
Hence given α : ∆(qD ) ! Z, β : |∆(n)| ! Y with pα = the restriction of β,
there is an m with Im α ⊂ Z m , and hence by the construction of Z m+1 a map
•
γ : |∆(n)| ! Z m+1 ⊂ Z such that pα = β and α = the restriction of γ to ∆(qD ).
By Lemma 2.3.1, p is a trivial fibration.
Remark. The argument used to prove Lemma 2.3.3 relied primarily on the fact
• •
that Hom(∆(qD ), lim Z m ) = lim Hom(∆(qD ), Z m ) and may be used to prove
−! −!
factorization whenever the fibrations (or trivial fibrations) are characterized by
the RLP with respect to a set of maps {Ai ! Bi } where each Ai is “sequentially
small” in the sense that Hom(Ai , •) commutes with sequential inductive limits.
We will have further occasions to use this argument and will refer to it as the
small object argument.
Lemma 2.3.4. The following are equivalent for a map i : A ! B.
(i) i is a trivial cofibration.
(ii) i has the LLP with respect to the fibrations.
(iii) i is a cofibration and a strong deformation retract map.
86
groups
Proof. (iii) =⇒ (i) since a strong deformatino retract map is a homotopy equiv-
alence and hence a weak homotopy equivalence.
(ii) =⇒ (iii). Any trivial fibration is a fibration so i is a cofibration. The
retract and strong deformation may be constructed by lifting in
∼ si
A B A BI
r h
i i (j0 ,j1 )
(ir,idB )
B e B B×B
which is possible since A ! e and B I ! B × B are fibrations by the corollary to

Lemma 2.3.2.
(iii) =⇒ (ii). A lifting in the first diagram, where p is a fibration, may be
constructed by lifting in the second
α sα
A X A XI
u p H (j0 ,pI )
i i
β
B Y B X ×Y Y I
(αr ,h)
and setting u = j1 H. Here r and h are the retract and strong deformation for i
and lifting in the second diagram is possible because (j0 , pI ) is a trivial fibration
by the corollary of Lemma 2.3.2.
(i) =⇒ (iii). Consider the following factorization of i
A ×B B I
j
A p
i
B
87
which is the dual of the mapping cylinder construction. j is a strong deformation

map, hence a weak equivalence, and p is a fibration. But i is a weak equivalence
and so p is a trivial fibration. As i is a cofibration there is a section u of p with
ui = j. Hence i is a retract of j and so i is a strong deformation retract map.
Proof of Theorem 2.1. Axioms M0, SM0, and M5 are clear. Axiom M6 follows
immediately from definitions and lemmas 2.3.1, 2.3.2, 2.3.4. M6 and the corollary
to Lemma 2.3.2 yield SM7. Lemma 2.3.3 gives one case of M2; to obtain the other,
j p
take f : X ! Y and factor it X −! X ×Y Y i −! Y where p is a fibration and j
is a weak equivalence. Then factor j = qi by Lemma 2.3.3 where i is a cofibration
and q is a trivial fibration. By M5 i is a trivial cofibration hence f = (qp)i is the
desired factorization. This proves M2 and hence the theorem.
Let SimpGrp be the category of simplicial groups endowed with its natural
simplicial structure (see §2.1). Then G ⊗ K and GK exists if G ∈ Ob SimpGrp
and K is a simplicial set. In fact (G ⊗ K)q = σ∈Kq Gq with natural simplicial
W
operations and GK is the function complex HomSimp (K, G) with its natural
group structure. Define the normalization of SimpGrp by
\
Nq (G) = ker(di : Gq −! Gq−1 ) (= G0 if q = 0)
i>0
d : Nq (G) −! Nq−1 (G) induced by d0 . (= 0 if q = 0)
and the (Moore) homotopy groups of G by
ker(d : Nq G −! Nq−1 G)
πq (G) = .
Im(d : Nq+1 G −! Nq G)
Then πq (G) is abelian for q ≥ 1 and π0 (G) is the set of components of G as a

simplical set.
88
groups
A map in SimpGrp will be called a weak equivalence if it induces isomor-

phisms for the functor π• . A map will be called a fibration if it is a fibration as a
map of simplicial sets and a cofibration if it has the LLP with respect to trivial
fibrations.
Theorem 2.2. With these definitions the category SimpGrp of simplicial groups
is a closed simplicial model category.
The proof will be exactly the same as for topological spaces once we get the
corollary of Lemma 2.3.2 for SimpGrp and the homotopy axiom for the functor
π• .
Proposition 2.3.1. The following are equivalent for a map f : G −! H of

simplicial groups.
(i) f is a fibration in Simp (hence in SimpGrp).
(ii) Nq f : Nq G −! Nq H is surjective for q > 0.

(f,ε)
(iii) G −−−! H ×K(π0 H,0) K(π0 G, 0) is surjective (in each dimension).
Here if A is a group we let K(A, 0) be the constant simplicial group which is A

in each degree and which has all φ∗ = idA . It is readily verified that G 7! π0 (G)
is adjoint to A 7! K(A, 0), that is
HomSimpGrp (G, K(A, 0)) = HomGrp (π0 (G), A)
and ε : G −! K(π0 G, 0) is the adjunction map. The above proposition is

essentially an elaboration of the following well-known fact which we shall assume.
Corollary Moore. A simplical group is a Kan complexes.
We shall also need the following fact which may be proved in exactly the same
way as [DP61].
89
Lemma 2.3.5. f : G −! H is surjective (resp. injective) iff N f : N G −! N H

is surjective (resp. injective).
Proof of proposition 2.3.1. (i) =⇒ (ii) since (ii) is equivalent to lifting in any
diagram of the form
0
V (n, 0) G
∆(n) H
where 0 denotes the map sending all simplicies to the identity elements of SimpGrp.
(ii) =⇒ (iii). By Lemma 2.3.5 it suffices to show that N (f, ε) is surjective.
As N is left exact and Nj K(A, 0) = {1} for j > 0 and A if j = 0, we find that
Nj H ×K(π0 H,0) K(π0 G, 0) = Nj H for j > 0, and hence Nj (f, ε) is surjective for

j > 0. It remains to show that G0 −! H0 ×π0 H π0 G is surjective which follows

immediately by diagram chasing in the diagram
d
N1 G G0 π0 G 1
d
N1 H H0 π0 H 1.
(iii) =⇒ (i). First suppose f : G −! H is surjective. Given

u : V (n, k) −! G covering v : ∆(n) −! H we may extend u to a map
u′ : ∆(n) −! G by the corollary. We may solve the lifting problem for u and v
iff we may solve it for 0 : V (n, k) −! G and v · (f u′ )−1 : ∆(n) −! H. Hence we
reduce to the case u = 0. As f is surjective when there is a map w : ∆(n) −! G
with f w = v. Then w|V (n,k) maps V (n, k) to ker f and by the corollary there is
a z : ∆(n) −! ker f with z|V (n,k) = w|V (n,k) . Then w · z −1 : ∆(n) −! G satisfies
(w · z −1 )|V (n,k) = 0 = u and f ◦ (w · z −1 ) = f ◦ w = v, thus providing the desired
90
groups
lifting. Hence any surjective map of simplicial groups is a fibration.

Returning to the general case we consider the diagram
(f,ε) pr1
G H ×K(π0 H,0) K(π0 G, 0) H
K(π0 f,0)
K(π0 G, 0) K(π0 H, 0)
where the square is cartesian. K(π0 f, 0) is clearly a fibration hence so is pr1 , and
(f, ε) being surjective is a fibration. Hence f = pr1 (f, ε) is a fibration.
Corollary. f is surjective iff f is a fibration and π0 (f ) is surjective.
Proposition 2.3.2. The following are equivalent for a map f in SimpGrp.
(i) f is a trivial fibration in Simp.
(ii) f is a trivial fibration in SimpGrp.
(iii) f is surjective and π• (ker f ) = 0.
Proof. (ii) ⇐⇒ (iii). First of all the above corollary shows that f is surjective in
case (ii). Letting K be the kernel of f we have the exact sequence of non-abelian
group complexes
1 −! N (K) −! N (G) −! N (H) −! 1
where exactness at N (K) and N (G) is because N is left exact and exactness at
N (H) comes from lemma 2.3.5. From this one gets by the usual diagram chasing
a long exact sequence
· · · −! π1 (G) −! π1 (H) −! π0 (K) −! π0 (G) −! π0 (H) −! 1
91
which shows that π• (K) = 0 iff π• (f ) is an isomorphism.

(i) =⇒ (iii). First of all a trivial fibration is surjective in dimension 0 since it
•
has the RLP with respect to ∆(0) ⊂ ∆(0); hence by the Corollary of Proposition
2.3.1 f is surjective. Next if α ∈ πq (ker f ) we represent α by x ∈ Kq with dj x = 0
•
for 0 ≤ j ≤ q and define u : ∆(q + 1) −! ker f by sending all faces to the identity
element of ker f except the 0-th which goes to x. Lifting in
•
u
∆(q + 1) G
0
∆(q + 1) H
we obtain y ∈ Nq+1 (Ker f ) with dy = x showing that α = 0.

(ii) + (iii) =⇒ (i). Given u : ∆(n) ! G covering v : ∆(n) ! H we may
lift if n = 0 since f is surjective. If n > 0, then as f is a fibration we may
find w : ∆(n) ! G with w|V (n,0) = u|V (n,0) and f w = u. Lifting for u and v
is equivalent to lifting for u · w−1 and 0 so we reduce to the case v = 0 and
•
u|V (n, 0) = 0. Then u applied to the 0-th face of ∆(n) is an element x of
(Ker f )n−1 with all faces the identity element. As π• (Ker f ) = 0, there is a
z ∈ Nn (Ker f ) with dz = x. Then z̃ : ∆(n) ! G satisfies z̃| • = u and f z̃ = 0,
∆(n)
hence z̃ is the desired lifting.
Corollary. Every object of SimpGrp is fibrant and the fibrations and trivial
fibrations of SimpGrp satisfy SM7(a)
Lemma 2.3.6. If f, g : G ⇒ H are homotopic maps in SimpGrp, then
π• (f ) = π• (g) : π• (G) −! π• (H).
Proof. We may assume that f is strictly homotopic to g. Let h : G × ∆(1) −! H
92
groups
be a homotopy with hi0 = f, hi1 = g. Then h = {hσ } where σ is a simplex of

∆(1), hσ : Gq −! Hq is a group homomorphism and q is the degree of σ. σ may
be identified with the sequence (σ0, . . . , σq), which is a sequence (0 . . . 0, 1 . . . 1).
Let h1 : Gq −! Hq be hσ where σ has i + 1 zeroes and q − 1 ones. Then h−1 = f
and hq = g in degree q. If α ∈ πq G, represent α by x ∈ Gq with dj x = 1, 0 ≤ j ≤ q,
and set
q
z1 = (h0 s0 x) · (h1 s1 x)−1 . . . (hq sq x)(−1)
q
z2 = (f s0 x) · (f s1 x)−1 . . . (f sq x)(−1) .
Then z1 z2−1 ∈ Nq+1 H and d(z1 z2−1 ) = gx · (f x)−1 showing that πq (f )α = πq (g)α
Proof of Theorem 2.2. : We first note that Lemma 2.3.4 holds in SimpGrp. In
effect (iii) =⇒ (i) because a homotopy equivalence is a weak equivalence by
Lemma 2.3.6 and the rest of the proof used only the definition of cofibration and
the corollary to Lemma 2.3.2 which for SimpGrp is replaced by the corollary
to Prop. 2. The factorization axiom Lemma 2.3.2 may be proved by the small
object argument since trivial fibrations are characterized by the RLP with respect
• •
to F ∆(n) −! F ∆(n) (F = free group functor), and since F ∆(n) is small. The
rest of the proof follows that of Theorem 2.1.
Let the category of Simp of simplicial sets be considered as a simplicial cat-

egory as in §2.1. Define fibrations and trivial fibrations as in §2.2 and call a map
a cofibration (resp. trivial cofibration) if it has the LLP with respect to the class
of trivial fibrations (resp. fibrations). Finally define a weak equivalence in Simp
to be a map f which may be factored f = pi where i is a trivial cofibration and
p is a trivial fibration.
93
Theorem 2.3. With these definitions the category Simp of simplicial sets is a
closed simplicial model category.
Proof. First note that “trivial” has its customary meaning in the sense that a
map is a trivial cofibration (resp. fibration) iff it is a cofibration (resp. fibration)
and a weak equivalence. Indeed the direction ( =⇒ ) is clear. If f : A −! B is a
cofibration and
Z
1
A p (2)
f
B
is a factorization of f , where i is a trivial cofibration and p is a trivial fibration,

then there exists a section s of p with sf = i. Hence f is a retract of i and so f
is a trivial cofibration. Fibrations are handled similarly.
The factorization axiom M2 may be proved by the small object argument
•
using Prop 2.2.1(i) and 2.2.2(i) and the fact that ∆(n) and V (n, k) are small.
This actually proves that any map f may be factored f = pi where p is a trivial
fibration (resp. fibration) and where i is a sequential composition of cobase
•
extensions of direct sums of the maps ∆(n) −! ∆(n) (resp. V (n, k) −! ∆(n)).
In particular i is injective (resp. an “anodyne extension” in the terminology of
Gabriel-Zisman). If f is already a cofibration (resp. trivial cofibration), then as
above (see (2)) f is a retract of i, hence is injective (resp. an “anodyne extension”).
The converse is also true (2.1.1 and [GZ67, p. 3.1]). Hence:
Proposition 2.3.3. In Simp the cofibrations are the injective maps and the
trivial cofibrations are the anodyne extensions. Any object of Simp is cofibrant.
All of the axioms except M5 are now clear. M0, SM0 are trivial and M6
is true by the way things have been defined. M2 follows from the small object
94
groups
argument, and as the fibrations and trivial fibrations of Simp satisfies SM7(a),
M6 implies that SM7 holds.
The fibrant objects of Simp are the Kan complexes. If E is a Kan complex
and A is a simplicial set, then by SM7 Hom(A, E) is a Kan complex so “is strictly
homotopic to” is an equivalence relation on Hom(A, E). Let
[A, E] = π0 Hom(A, E) denote the equivalence classes. Then M5 follows imme-
diately from:
Proposition 2.3.4. A map f : X −! Y in Simp is a weak equivalence if and

only if for all Kan complexes E, [f, E] : [Y, E] −! [X, E] is bijective.
Proof.
( =⇒ ) If f is a trivial cofibration then this follows from the covering homotopy

extension theorem (Prop 2.2.4) which depends only on SM7. If f is a trivial
fibration then as every simplicial set is cofibrant one sees by the dual of the
argument used to prove (ii) =⇒ (iii) in Lemma 2.3.3 that f is the dual of a
strong deformation retract map. In particular f is a homotopy equivalence
so [f, E] is bijective. If f is a weak equivalence then f is the composition
of a trivial cofibration and a trivial fibration so [f, E] is bijective.
( ⇐= ) Factoring f = pi where i is a cofibration and p is a trivial fibration we have

[p, E] bijective by the above so we reduce to the case where f is a cofibration.
In this case f is a trivial cofibration by the following two lemmas.
Lemma 2.3.7. If i is a cofibration and [i, e] is bijective for all Kan complexes
E, then i has the LLP with respect to all fibrations of Kan complexes.
Lemma 2.3.8. If a cofibration i has the LLP with respect to all fibrations of
Kan complexes, then it has the LLP with respect to all fibrations and so is a
trivial cofibration.
95
Proof of Lemma 2.3.7. We begin by showing that if p : X −! Y is a fibration

of Kan complexes, then p is a trivial fibration if and only if p is a homotopy
equivalence. The direction =⇒ has been proved above. To prove ⇐= let s be
a homotopy inverse for p. By lifting the homotopy from ps to idY we may assume
that ps = idY . Then idX and sp are homotopic and as X is a Kan complex we
may choose h : X × ∆(1) −! X with hi0 = sp and
hi1 = idY . Now
Hom(X, p) : Hom(X, X) −! Hom(X, Y )
is a fibration and the l-simplicies h and sph define a map a : V (2, 0) −! Hom(X, X)
which covers the map β : ∆(2) −! Hom(X, X) given by the 2-simplex s1 (ph).
Hence there is a map γ : ∆(2) −! Hom(X, X) which covers the map β and
restricts to α; the 0-th face of γ(id) is a homotopy k : X × ∆(1) −! X from idX
to sp which is fiber-wise, i.e. pk = σ(p × ∆(1)). This shows that p : X −! Y
is a fibration and the dual of a strong deformation retract and hence is a trivial
fibration.
Now let i : A −! B and E be as in the statement of Lemma 2.3.7 and apply

this fact to the fibration
Hom(i, E) : Hom(B, E) −! Hom(A, E).
If K is any simplicial set, then [K, Hom(B, E)] −! [K, Hom(A, E)] may be
identified with [B, Hom(K, E)] −! [A, Hom(K, E)] which is bijective since
Hom(K, E) is a Kan complex and the assumption on i. Hence Hom(i, E) is
a trivial fibration.
Let p : X −! Y be a fibration in Simp where Y and hence X is a Kan
96
groups
complex and consider the diagram
(i∗ ,p∗ ) pr1

Hom(B, X) Hom(A, X) ×Hom(A,Y ) Hom(B, Y ) Hom(A, X)
pr1 Hom(A,p)
Hom(i,Y )
Hom(B, Y ) Hom(A, Y )
where the square is cartesian. We have just shown that Hom(i, Y ) is a trivial
fibration and hence so is pr1 . Thus pr1 and
pr1 (i∗ , p∗ ) = 1∗ = Hom(i, X)
are trivial fibrations, hence homotopy equivalences, and so (i∗ , p∗ ) is a fibration

(SM7) it is a trivial fibration hence surjective in dimension zero and so i has the
LLP with respect to p.
Proof of Lemma 2.3.8. If p : X −! Y is an arbitrary fibration in Simp, then

by [BGM59] there is a minimal fibration q : Z −! Y such that Z is a strong
deformation retract of X over Y (i.e. the homotopies are fiber-wise). As i is a
cofibration SM7 implies that i has the LLP with respect to p iff i has the LLP
with respect to q. But q is induced from a fibration of Kan complexes. To see
this we may suppose Y is connected and let F be the fiber of q over a 0-simplex
of Y . Then by [BGM59] there is a cartesian square
X W (Aut F ) ×Aut F F
q r
Y W (Aut F )
where r is a fibration and W (Aut F ) is a Kan complex. As i has the LLP
97
with respect to r it does so also for q, and hence i is a trivial cofibration. This
completes the proof of Lemma 2.3.8 and hence also of Theorem 2.3.
Combining Prop 2.3.2 with Prop 1.5.1 we obtain
Corollary. The anodyne extensions are precisely the injective maps in Simp
which become isomorphisms in the homotopy category.
Remark. We have presented what we consider to be the next elementary proof

of Theorem 2.3. The problem is to characterize the weak equivalences in some
way so that M5 becomes clear. We now present a list of different characterizations
of the weak equivalences. Some of these may be used to give alternative proofs
of M5 and will be useful later.
Proposition 2.3.5. The following assertions are equivalent for a map f : X −!

Y of simplicial sets:
(i) f is a weak equivalence (isomorphism in homotopy category).

∼
(ii) [Y, E] −
! [X, E] for all Kan complexes E.
(iii) |X| −! |Y | is a homotopy equivalence in T .
(iv) Ex∞ X −! Ex∞ Y is a homotopy equivalence in Simp.

∼ ∼
(v) H 0 (Y, S) −
! H 0 (X, S) for any set S, H 1 (Y, G) −
! H 1 (X, G) for any group
∼
G, and H q (Y, L) −
! H q (X, f ∗ L) for any local coefficient system L of abelian
groups on Y and q ≥ 0.
∼ ∼
(vi) π0 X −
! π0 Y , π1 (X, x) −
! π1 (Y, f x) for any x ∈ X0 , and
∼
H q (Y, L) −
! H q (X, f ∗ L) where L, q are as in (v).
Proof. (i) ⇐⇒ (ii) is Proposition 2.3.4. (ii) ⇐⇒ (iii) ⇐⇒ (iv) are proved
in [Kan57b]. Here X −! Ex∞ X is the functorial embedding of X into a Kan
complex constructed by Kan.
98
groups
(v) ⇐⇒ (vi). Here
H 0 (X, S) = Hom(π0 X, S), H 1 (X, G) = [X, W (G)],
and π1 (X, x) is the fundamental group of X at x calulated by the method of the

maximal tree. The first assertion of (v) and (vi) are equivalent and we may assume
X and Y are connected. Let x ∈ X0 . Then [X, W (G)] = HomGrp (π1 (X, x), G)G
where G acts on a homomorphism Ψ by (g · Ψ)(λ) = gΨ(λ)g −1 . In other words,
[X, W (G)] is the set of homomorphisms from π1 (X, x) to G in the category of
groups up to inner automorphisms, so the second condition of (v) means that
π1 (X, x) −! π1 (Y, f x) is an isomorphism in this category. But this is clearly the
same as π1 (X, x) −! π1 (Y, f x) being an isomorphism of groups, and so we see
that the second conditions of (v) and (vi) are equivalent. Thus (v) and (vi) are
equivalent.
(iii) =⇒ (vi). As π0 |X| = π0 X we may assume that X and Y are connected.

∼
As π1 (|X|, x) = π1 (X, x) we conclude that π1 (X, x) −
! π1 (Y, f x) for all x ∈ X0 .
∼
Let x0 be a fixed 0-simplex of X, let y0 = f x0 and let π = π1 (X, x0 ) −
! π1 (Y, y0 ).
Let p : (X,
e xe0 ) −! (X, x) (resp. q : (Ye , ye0 ) −! (Y, y0 )) be the universal coverings
and fe : X
e −! Ye be the unique map covering f with fexe0 = ye0 .
If L is a local coefficient system on Y , then there is a morphism of Cartan-

Leray spectral sequences
E2pq = H p (π, H q (Ye , q ∗ L)) H p+q (Y, L)
E2pq = H p (π, H q (X,

e p∗ f L)) H p+q (X, f ∗ L)
As |X|
e and |Ye | are the universal coverings of X and Y , (iii) =⇒ |fe| is a homotopy
99
equivalence. As H ∗ (|X|, e A) for any abelian group A we see that the

e A) = H ∗ (X,
map on the E2 is an isomorphism and so (vi) is proved.

(vi) =⇒ (iii). We may assume X and Y are connected and we let X,
e Ye , π,
etc., be as above. By a theorem of Whitehead it suffices to prove that

∼
πq (|X|, x0 ) −
! πq (|Y |, y0 ) for all q. For q = 1, this comes from
π1 (|X|, x0 ) = π1 (X, x0 ) and the similar assertion for Y . For q > 1 it suffices
to prove |fe| is a homotopy equivalence or equivalently, since |X|
e and |Ye | are 1-
∼
connected, that H ∗ (Ye , A) −
! H ∗ (X,
e A) for any abelian group A. But the Leray
spectral sequences for p and q degenerate giving a diagram
H n (Y, q∗ , A) ∼ H n (Ye , A)
f∗ e∗
f
∼
H n (X, p∗ , A) H n (X,
e A)
where p∗ A, q∗ A are the local coefficient systems of the cohomology of the fiber,
and where f ∗ is the map on cohomology coming from f ∗ (q∗ A) = p∗ A. By (vi)
f ∗ is an isomorphism and so we are finished.
2.4 sA as a model category
Let A be a category closed under finite limits. A map f : X −! Y is said to

be an effective epimorphism if for any object T the diagram of sets
f∗ pr∗
1
Hom(Y, T ) Hom(X, T ) Hom(X ×Y X, T )
pr∗
2
is exact. We shall say that an object P of A is projective if

Hom(P, X) −! Hom(P, Y ) is surjective whenever X −! Y is an effective epi-
morphism and that A has sufficiently many projectives if for any object X
100
Section 2.4: sA as a model category
there is a projective P and an effective epimorphism P −! X. If A is closed

under inductive limits, we call an object X small if Hom(X, −) commutes with
filtered inductive limits, and call a class U of objects of A a class of generators
if for every object X there is an effective epimorphism Q −! X, where Q is a
direct sum of copies of members of U.
Theorem 2.4. Let A be a category closed under finite limits and having suffi-
ciently many projectives. Let sA be the simplicial category of simplicial objects
over A. Define a map f in sA to be a fibration (resp. weak equivalence) if
Hom(P, f ) is a fibration (resp. weak equivalence) in Simp for each projective
object P of A, and a cofibration if f has the LLP with respect to the class of
trivial fibrations. Then sA is a closed simplicial model category if A satisfies one
of the following extra conditions:
(∗) Every object of sA is fibrant.
(∗∗) A is closed under inductive limits and has a set of small projective genera-
tors.
Here, and in the following, objects of A will be identified with constant sim-
plicial objects. For the rest of this section A will denote a category closed under
finite limits and having sufficiently many projectives. We will not use conditions
(∗) and (∗∗) until we absolutely have to. We first make some remarks about the
theorem.
Proposition 2.4.1. Suppose that every object X of A is a quotient of a cogroup

object C (i.e. there exists an effective epimorphism C −! X). Then A satisfies
(∗).
Proof. Given X ∈ Ob sA and a projective object P of A, choose an effective

epimorphism C −! P where C is a cogroup object. Then P is a retract of
101
C, so Hom(P, X) is a retract of Hom(C, X) which is a group complex. By

Moore, Hom(C, X) is a Kan complex hence so is Hom(P, X), and hence X is
fibrant.
Remarks. 1. By a theorem of Lawvere [Law63] a category closed under in-

ductive limits and having a single small projective generator U is equivalent
to the category of universal algebras with a specified set of finitary opera-
tions and identities in such a way that U corresponds to the free algebra on
one generator. Hence the theorem applies when A is the category of rings,
monoids, etc. One may show that effective epimorphism = set-theoretically
surjective map in this case.
2. The category of profinite groups satisfies (∗) but not (∗∗). The free profinite
group generated by a profinite set is both projective and a cogroup object
in this category and every object is a quotient of such an object.
The rest of this section contains the proof of Theorem 2.4.
Proposition 2.4.2. Let A be a category closed under finite limits and having
sufficiently many projectives. Then X −! Y is effective epimorphism ⇐⇒
Hom(P, X) −! Hom(P, Y ) is surjective for every projective object P .
Proof. ( =⇒ ) is by definition. For ( ⇐= ) we first establish three properties of

effective epimorphisms which hold without assuming A has enough projectives.
It is clear that f : X −! Y is an effective epimorphism iff for any object T and
map α : X −! T there is a unique β : Y −! T with βf = α provided α satisfies
the necessary condition that αu = αv whenever u, v : S ⇒ X are two maps such
that f u = f v.
(1) If f : X −! Y has a section s : Y −! X with f s = idy , then f is an

effective epimorphism.
102
In effect, given α : X −! T satisfying the necessary condition let

β = αs : Y −! T . As sf, idX : X ⇒ X are two maps with f (sf ) = f (idX ) we
have βf = αsf = α. β is clearly unique.
f g
(2) If X −! Y −! Z are maps, where gf is an effective epimorphism and f is
an epimorphism, then g is an effective epimorphism.
Given α : Y −! T with αu = αv whenever u, v : S −! Y and gu = gv, it

follows that αf : X −! T has the property that αf u = αf v, whenever u, v : S −!
Y and gf u = gf v. As gf is an effective epimorphism, there is a unique map
β : Z −! T with βgf = αf . As f is an epimorphism βg = α.
f g
(3) If X −! Y −! Z are maps, where g is an effective epimorphism and f has
a section s, then gf is an effective epimorphism.
In effect given α : X −! T satisfying the necessary conditions that it factors

through gf , it in particular satisfies the necessary conditions for factoring through
f . By (1). there is a unique β with βf = α given by β = αs. Suppose u, v : S ⇒ Y
are such that gu = gv. Then gf su = gf sv so αsu = αsv or βu = βv . Hence
since g is an effective epimorphism there is a unique γ with γg = β and hence a
unique γ with γgf = α. Thus gf is an effective epimorphism.
Now suppose that f : X −! Y has Hom(P, X) −! Hom(P, Y ) surjective for

all projective objects P . Choose an effective epimorphism u : P −! X with P
projective. As f u has the same property as f we are reduced by (2). to the case
where X is projective. Choose an effective epimorphism v : Q −! Y with Q
projective. As X is projective there is a map α : X −! Q with vα = f and by
the property of f there is a map β : Q −! X with f β = v. The maps α and β
103
yield sections of the maps pr1 and pr2 in
pr2
X ×Y Q Q
pr1 v
X f
Y
By (1). and (3). v pr2 = f pr1 is an effective epimorphism and so by (2). f is

an effective epimorphism.
Corollary. The class of effective epimorphisms in A is closed under composition

and base change and it contains all isomorphisms. If gf is an effective epimor-
phism so is g.
In particular, the effective epimorphisms are universally effective.
Proposition 2.4.3. Any map f may be factored f = pi where i is a cofibration

and where p is a trivial fibration.
Proof. Given f : X −! Y construct a diagram
j0 j1 j2
X Z0 Z1 ...
p0
f p1
as follows. Let Z −1 = X, p−1 = f and having obtained pn−1 : Z n−1 −! Y ,

choose a projective object Pn of A and a map (α, β) so that
∆(n)
(α,β)+(pn−1 ,(Z n−1 )in ) •
Pn ∨ (Z n−1 )∆(n) Y ∆(n) × •

∆(n)
(Z n−1 )∆(n) (1)
Y
•
is an effective epimorphism in dimension 0, where in : ∆(n) −! ∆(n) is the
104
canonical inclusion. Now define the map jn by a cocartesian diagram
•
Pn ⊗in
Pn ⊗ ∆(n) Pn ⊗ ∆(n)
(2)
β in2
Z n−1 in1 =jn

Zn
and let pn : Z n −! Y be the unique map with pn jn = pn−1 and pn in2 = α.

•
As in : ∆(n) −! ∆(n) is an isomorphism in dimensions < n so is jn , hence
i p
lim Z n = Z exists and we may define map X −! Z −! Y by i = lim Jn . . . J0 ,
−!
p = lim pn . It is clear that Pn ⊗ in in (2) is a cofibration, hence each jn and hence
i is a cofibration. To see that p is a trivial fibration it suffices to show that
(P ∆(n) , Z in ) : Z ∆(n) −! Y ∆(n) × • (Z n−1 )∆(n)

Y ∆(n)
is an effective epimorphism in dimension 0. Consider the diagram
Pn ∨ (Z n−1 )∆(n) Y ∆(n) × •

∆(n)
(Z n−1 )∆(n)
Y
∆(n) •
β+jn ∆(n)
(id,kn−1 )
(Z n )∆(n) ∆(n)
Z ∆(n) Y ∆(n) × • Z ∆(n)
kn (p∆(n) ,Z in ) Y ∆(n)
where the top map is the effective epimorphism (1), and where kq = lim jn . . . jq+1 .
•
∆(n)
kq+1 is an isomorphism in dimension < n, hence (id, kn−1 ) is an isomorphism
in dimension zero. By the corollary of Prop. 2.4.2, (p∆(n) , Z in ) is an effective
epimorphism in dimension 0.
Proof of Theorem 2.4. (∗) This is exactly the same as the proof in §2.3 for Top
and SimpGrp, so we present an outline only. If f : A −! B is a map, then as
105
i p
A and B are fibrant, A −! A ×B B I −! B is a factorization of f into a strong
deformation retract map followed by a fibration. The homotopy equivalence i in
sA is carried by Hom(P, −) into a homotopy equivalence in Simp; hence i is
a weak equivalence in sA. If f has the LLP with respect to fibrations, f is a
cofibration and a retract of i; hence f is a trivial cofibration. Conversely, if f is a
trivial cofibration, M5 implies p is a trivial fibration so f is a retract of i; hence
f is a strong deformation retract map, so by SM7(a), f has the LLP with respect
to the fibrations. With this we have M6, hence SM7. Finally, M2 results from
Prop. 2.4.3 for the cofibration–trivial fibration case and for the other case one
uses this case to write i = qj, j cofibration, q trivial fibration, whence f = (pq)j
is a factorization where j is a trivial cofibration and pq is a fibration.
(∗∗) Let U be a set of small projective generators for A. Then the retract
argument used in the proof of Prop. 2.4.1 shows that a map f in sA is a fibration
or weak equivalence iff Hom(P, f ) is so is Simp for all P ∈ U. In particular,
the fibrations are characterized by the RLP with respect to the set of maps
P ⊗ V (n, k) −! P ⊗ ∆(n) for each P ∈ U and 0 ≤ k ≤ n > 0. However
P ⊗ V (n, k) is small in sA since P is small in A, hence the small object argument
implies that any map f may be factored f = pi where p is a fibration and i has
the LLP with respect to all fibrations. We must show that i is a weak equivalence.
For this purpose, we shall use Kan’s Ex∞ functor [Kan57a]. We recall that
(Ex K)n is the projective limit in the category of sets of a finite diagram involving
Kn , Kn−1 and the face operators of K. AS A is closed under finite limits, we
may define Ex : sA −! sA by the formula
Hom(A, Ex X) = Ex Hom(A, X) (3)
for all A ∈ Ob A, X ∈ Ob sA. The natural map K −! Ex K in Simp extends

to a map X −! Ex X, and hence we may define Ex∞ (X) = lim Exn (X) and a
−!
106
map εX : X −! Ex∞ (X). If P ∈ U, then as P is small
Hom(P, Ex∞ X) = Ex∞ Hom(P, X).
Therefore Ex∞ X is fibrant and εX : X −! Ex∞ X is a weak equivalence.

Now suppose that i : A −! B has the LLP with respect to fibrations. Then
we may lift successively in
sεB i
A
εA
Ex∞ A A (Ex∞ B)∆(1)
i u i H (j0 ,j1 )
B e B (Ex∞ B) × (Ex∞ B)
(εB ,(Ex∞ i)u)
obtaining the formulas ui = εA , (Ex∞ i)u ∼ εb , εB i = (Ex∞ i)εA . Let P ∈ U

and apply the functor γ ◦ Hom(P, −) where γ is the canonical localization map
Simp −! Ho Simp. It follows that γ Hom(P, i) is an isomorphism hence (Prop.
1.5.1) Hom(P, i) is a weak equivalence. Thus i is a weak equivalence and we
have proved that a map with the LLP with respect to the fibrations is a trivial
cofibration. Conversely if f is a trivial cofibration we may factor f = pi where p
is a fibration and i has the LLP with respect to the fibrations; by what we have
just shown i is a weak equivalence, hence p is trivial, so f is a retract of i and
hence has the LLP with respect to the fibrations. This proves half of M6 and
M2; the other is similar using Prop. 2.4.3. M6 implies SM7 and M5 is clear, so
the theorem is proved.
Remarks. 1. Some extra conditions on A like (∗) or (∗∗) is necessary since

the category of simplicial finite sets fails to satisfy M2. In effect there are
simplicial finite sets with infinite homotopy groups.
i p
2. If the map ∅ −! X in sA is factored ∅ −! Z −! X where p is a trivial
107
fibration and i is a cofibration, then it is easily seen by using Prop. 2.2.4

that this factorization is unique up to simplicial homotopy over X. Now
for Z −! X to be a trivial fibration is the analogue of Z being a resolution
of X, while for Z to be cofibrant is the analogue of Z being a complex
of projective objects. Hence Prop. 2.4.3 asserts for sA the existence of
projective resolutions and so one may define derived functors for A even
when A does not satisfy (∗) or (∗∗).
•
3. It is worthwhile noting that (X ∆(n) )0 = (coskn−1 X)n where coskq is the

q-th coskeleton functor of Verdier [Ver]. Consequently a trivial fibration
X −! A where A is an object of A is the same as hypercovering of A for
the Grothendieck topology whose covering families consist of single maps
{v −! u} which are effective epimorphisms. We will discuss this in the
next section.
4. When A is a category of universal algebras (see Remark 1 after Prop. 2.4.1),

then the Pn in the proof of Prop. 2.4.3 may be chosen to be free algebras,
i
and so the map X −! Z is free in the following sense: there are subsets
Cq ⊂ Zq for each q such that
(i) η ∗ Cp ⊂ Cq whenever η : [q] −! [p] is a surjective monotone map,
(ii) fq + gq : Xq ∨ F Cq −! Zq is an isomorphism for all q, where F Cq is

the free algebra generated by Cq and gq : F Cq −! Zq is the unique
algebra map which is the identity on Cq .
i
Conversely, one may show [Kan57b, Thm. 6.1] that any free map X −! Z
may be factored
X −! Z 0 −! Z 1 −! . . . −! Z
where there are co-cartesian squares (2) with Pn free and hence any free
map is a cofibration. Furthermore given a cofibration f we may factor it
108
Section 2.5: Homology and cohomology
f = pi where i is free and p is a trivial fibration: then f is a retract of i

hence a map is a cofibration iff it is a retract of a free map.
5. If A is an abelian category with sufficiently many projective objects, then

Theorem 2.4 endows sA with the structure of a closed model category. On
the other hand by Dold–Puppe [DP61] the normalization functor
N : sA −! Ch A, the category of chain complexes in A is an equivalence
of categories, and moreover the simplicial homotopy relation on maps in
sA corresponds to the chain homotopy relation on maps in Ch A. The cor-
responding closed model category structure on Ch A may be described as
follows: Weak equivalences are maps inducing isomorphisms on homology
groups (since H(N X) = πX) and fibrations are maps which are epimor-
phisms in positive degrees (straightforward generalization of Prop. 2.3.1 to
abelian categories). Finally cofibrations are monomorphisms whose cock-
erels are dimension-wise projective. In effect what is called the fundamental
theorem of homological algebra amounts essentially to the following: (i) any
monomorphism with dimension-wise projective cokernel has the LLP with
respect to trivial fibrations and (ii) any map f may be factored f = pi
where p is a trivial fibration and i is a monomorphism with dimension-wise
projective cokernel.
As the class of monomorphisms with dimension-wise projective cokernels is

closed under retracts, it is seen to be the class of cofibrations by a retract
argument.
2.5 Homology and cohomology
If homotopical algebra is thought of as “non-linear” or “non-additive” homo-

logical algebra, then it is natural to ask what is the “linearization” or “abelion-
aization” of this non-linear situation. This leads to a uniform description of
109
homology and cohomology for model categories and in the case of sA the result-
ing cohomology agrees with the cohomology constructed using suitable cotriples
and Grothendieck topologies.
Let C be a model category and let Cab be the category of abelian group
objects in C. We assume that the abelionaization Xab of any object X of C
exists so that there are adjoint functors
ab
C Cab (1)
i
where i is the faith inclusion functor. We also assume that Cab is a model category
in such a way that these adjoint functors satisfy the conditions of the first part
of 1.3, so that there are adjoint functors
L ab
Ho C HoCab (2)
Ri
[X, Ri(A)] = [L ab(X), A]
Finally we shall assume that Ho Cab satisfies the following two condition:
A. The Adjunction map θ : A ∼

= ΩΣA is an isomorphism for all objects A.
B. If
i j δ
A′ −! A −! A′′ −! ΣA′
is a cofibration sequence, then
−i·θ −1 j δ
ΩΣA′ −−−−! A −! A′′ −! ΣA
is a fibration sequence (Note that as Ho C−ab is additive the action

m
F × ΩB −! F is determined by ∂ = m(0, id) : ∂B −! F via the rule
m(α, λ) = α + ∂λ if α : T −! F and λ : T −! ΩB).
110
These conditions hold for example if Cab = sA, where A is an abelian category
with enough projectives and if Cab is the model category of simplicial modules
over a simplicial ring (see following section.)
We define the cohomology groups of an object X of Ho C with coefficients

an object A of Ho Cab to be
q
HM (X, A) = [L ab(X), Ωq+N ΣN A]
where N is an integer ≥ 0 with q + N ≥ 0. By (A) it does not matter what N

we choose. Suppose now that C is pointed. Then
q
HM (ΣX, A) = [L ab(ΣX), Ωq+N ΣN A] = [ΣL ab(X), Ωq+N ΣN A]
q+1
= [L ab(X), Ωq+N +1 ΣN A] = HM (X, A)
Using this and the fact that L ab preserves cofibration sequence, we find that if
X −! Y −! C, etc. is a cofibration sequence, then there is a long exact sequence
q q q δ
q+1
· · · −! HM (C, A) −! HM (Y, A) −! HM (X, A) −! HM (C, A) −! · · ·
From (B) it follows that if
A′ −! A −! A′ −! ΣA′
is a cofibration sequence in Ho Cab then there is a long exact sequence
q q q δq+1
· · · −! HM (X, A′ ) −! HM (X, A) −! HM (X, A′′ ) −! HM (X, A′ ) −! · · ·
It is reasonable to call an object of Ho C of the form Ri(A) a generalized
111
Eilenberg-Maclane object and to call L ab(X) the homology of X. In effect
0
HM (X, A) = [L ab(X), A]
is a universal coefficient theorem while
0
HM (X, A) = [X, Ri(A)]
is representability theorem for cohomology.
Examples. 1. C = Simp so that Simpab = s(Ab) the category of simpli-

cial abelian groups and Xab = ZX,the free abelian group functor applied
dimension-wise to X. The assumption on Simp and Sab hold and as every
object of Simp is cofibrant L ab(X) = Xab . Hence
•
HM (X, K(R, 0)) = H • (X, R),
the usual cohomology of X with values in the abelian group R. Also
π• (Xab ) = H• (X, Z)
which partially justifies calling Xab the homology of X.
2. Let C = SimpGrp so that SimpGrpab = s(Ab) and Gab = G/[G, G].
Then L ab(G) = Gab if G is a free simplicial group and so by a result of

[Kan58b] (See also (16))



 0 q<0


q
HM (G, K(R, 0)) =



H q+1 (W G, R) q ≥ 0

112
where W G is the “classifying space” simplicial set of G. Also
πq (L ab(G)) = Hq+1 (W G, Z).
These formulas are seen to hold for any simplicial group G since to calculate
L ab(G) we may replace G by a free simplicial group.
We now show how these model cohomology groups compare with other kinds
of cohomology. In the following A denotes a category closed under finite projec-
tive limits, X is an object of A, and A is an abelian group object in A/X. We
consider four definitions of cohomology of X with values in A.
(1) Suppose that the effective epimorphisms of A are universal effective epi-
morphisms (which is the case if A has sufficiently many projectives– Cor.
to Prop. 2). We define a Grothendieck topology on A ([Art62]) by defining
a covering of an object Y to be a family consisting of a single map U −! Y
which is an effective epimorphism. The induced topology on A/X is coarser
than the canonical topology so the representable functor hA is a sheaf of
abelian groups; hence sheaf cohomology groups, which we shall denote by
q
•
HGT (X, A), are defined. Thus HGT (X, A) = H q (I ′ (X)) where I ′ is an
injective resolution of hA in the category of abelian sheaves on X.
(2) Suppose that there are adjoint functors
F
A B
S
HomA (F B, Y ) = HomB (B, SY ) (3)
Such that (i) F SY −! Y is an effective epimorphisms for all Y ∈ Ob A, (ii)

F B is projective for all B ∈ Ob B. these adjoint functors define a cotriple
113
(see [BB66]) and hence cohomology groups Hcot

∗
(X, A) defined by
∗
Hcot (X, A) = H ∗ [hA (C• (X))]
where C• (X) is the simplicial object of A/X with Cq (X) = (F S)q+1 (X)
with face and degeneracy operators coming from the adjunction maps
id −! SF, F S −! id.
(3) Suppose that A is closed under finite limits and has sufficiently many pro-
jective objects. Regarding X as a constant simplicial object there exists
by Prop 2.4.3 a trivial fibration P• −! X, Where P• is cofibrant, which is
unique up to homotopy over X. The group H q [hA (P• )] is therefore inde-
pendent of the choice of P• and we denote it by Rq hA (X).
(4) Suppose that A satisfies the conditions of theorem 2.4, §2.4 that the abelian-
ization functor ab : A/X −! (A/X)ab exists, and that (A/X)ab is an
abelian category. Then the model category C = s(A/X) satisfies the
assumption made at the beginning of this section and hence cohomology
groups HM
•
(X, A) are defined, where X and A are identified with constant
simplicial objects.
q q
Theorem 2.5. When each of groups HM (X, A), Hcot (X, A), and RqhA (X) is
defined, it is canonically isomorphic with the Grothendieck sheaf cohomology
q
group HGT (X, A).
q
Proof. We begin by showing that HM (X, A) = RqhA (X). Let F be the abelian
category (A/X)ab . F has enough projectives, namely those of the form Pab where
P is a projective object A/X. Hence sF and Ch(F) are model categories (see
Remark 5 at the end of §2.4) and N : sF −! Ch(F) is an equivalence of model
categories. The loop and suspension functors on Ho(Ch F) are given very simply
114
by functors Ω and Σ on Ch F defined by the formulas


Xq−1

q>0 dΣX = −ΣdX
(ΣX)q =
0 q=0


Xq+1

q>0 dΩX = −ΩdX
(ΩX)q =
Ker{d : X1 −! X0 } q = 0

Let A[q] be the chain complex in F which is A in dimension q and 0 elsewhere

(A[q] if q < 0). As
N K(A, 0) = A[0], N Ωq+N ΣN K(A, 0) = A[q],
hence
q
HM (X, A) = [L ab(X), Ωq+N ΣN K(A, 0)] = π0 ((P• )ab , Ωq+N ΣN K(A, 0))
= π(N (P• )ab , A[q]) = H q HomF (N (P• )ab , A)
= H q (HomF ((P• )ab , A)) = H q hA (P• ) = Rq hA (X).
To finish the theorem we need some results about Grothendieck sheaves

([Art62], [Ver]). Let T denote a Grothendieck topology whose underlying cate-
gory is closed under finite projective limits and has sufficiently many projectives,
and where a covering of an object Y in T is a family U = (Z −! Y ) consisting
of a single effective epimorphism. Eventually we will let T be A/X. A presheaf
of sets (resp. abelian groups) is a functor T0 −! Set(resp Ab) and a presheaf of
sets (resp. abelian groups) is a presheaf F such that for any effective epimorphism
115
Z −! Y the diagram
F (Y ) −! F (Z) ⇒ F (ZXY Z)
is exact.
Letting Pr and Sh (resp. Prab and Shab) denote the categories of presheaves
and sheaves of sets (resp. abelian groups) we have the diagram
a
Sh Pr
i
Z j Z j (4)
a
Shab Prab
i
Here i and j are inclusion functors which are right adjoint functors and the other
functors are left adjoint functors. The square of left (resp. right) adjoint functors
commutes up to canonical isomorphisms.
We recall the construction of a, the associated sheaf functor. If F ∈ Ob Pr

(resp. Ob Prab), then the 0−th (resp. q−th) Čech cohmology presheaf of F is
defined by
Ȟ◦ (F )(Y ) = lim

!
−
Ȟ ◦ (U, F )
U
(resp. Ȟq (F )(Y ) = lim

!
−
Ȟ q (U, F ))
U
where the limit is taken over the category coverings U = (U −! Y ) of Y and

where
Ȟ ◦ ((U −! Y ), F ) = Ker{F (U ) ⇒ F (U ×Y U )}
(resp. Ȟ q ((U −! Y ), F ) = the q−th cohomology of the cosimplicial abelian
116
group
!
F (U ) ⇒ F (U ×Y U ) !
! F (U ×Y U ×Y U ) . . . )
Then aF = Ȟ◦ Ȟ◦ (F ). Given Y , choose an effective epimorphism P −! Y with

P projective; it follows that (P −! Y ) is cofinal in the category of coverings of
Y and hence
Ȟ ◦ (F )(Y ) = Ker{F (P ) ⇒ F (P ×Y P )}
In particular Ȟ◦ (F )(P ) = F (P ) if P is projective, and hence
(aF )(P ) = F (P ) (5)
If Y is arbitrary choose effective epimorphism P0 −! Y , Pl −! P0 ×Y P0 , whence
a(F )(Y ) = Ker{(aF )(P0 ) ⇒ (aF )(Pl )}
= Ker{F (P0 ) ⇒ F (Pl )}
It follows that for F ∈ Ob Prab, aF = 0 if and only if F (P ) = 0 for all projective

u v
P . Now if F ′ −! F −! F ′′ are maps in Shab with vu = 0, then this sequence
is exact iff aH = 0 where H = Ker v/ Im u in the category Prab. Hence we have
proved
Lemma 2.5.1. A sequence F ′ −! F −! F ′′ of abelian sheaves is exact iff

F ′ (P ) −! F (P ) −! F ′′ (P ) is exact for all projective objects P .
Let Z(S) denote the free abelian group generated by a set S. Then the
abelianization functor Z for presheaves is given by (ZF )(Y ) = Z(F (Y )) for all Y
hence combining (5) and the commutativity of (5) we obtain
117
Lemma 2.5.2. If F is a sheaf of sets, then its abelianization ZF is such that
ZF (P ) = Z(F (P ))
for all projectives P .
Let Hq : Shab −! Prab be the q−th cohomology presheaf functors. Then

Hq q ≥ 0 are the right derived functors of i : Simp −! P and
H• (F )(Y ) = H • (Y, F ) is the cohomology of F over Y . We define a weak equiv-
alence in sT to be a map Z• −! Y• such that for any projective object P ,
Hom(P, Y• ) −! Hom(P, Z• ) is a weak equivalence in Simp. This agrees with
the definition in §2.4.
Proposition 2.5.1. The following are equivalent for a sheaf of abelian groups:
(i) Hq (F ) = 0 q > 0.
(ii) Ȟ q ((U −! Y ), F ) = 0 q > 0 for all effective epimorphisms U −! Y .
(iii) For any weak equivalence Z• −! Y• in sT
∼
H ∗ (F (Y• )) −! H ∗ (F (Z• ))
A sheaf satisfying the equivalent conditions of Prop. 1 will be called flask.

By (i) any injective sheaf is flask.
Proof. (i) =⇒ (iii). Let hY : T0 −! Set be the functor repesented by Y ; then

hY is a sheaf. Let ZY = ZhY so that
HomShab (ZY , F ) = F (Y )
118
Let I • be an injective resolution of F in Shab so that H q (I • (Y )) = H q (Y, F ) = 0

for all Y . Then

H p (F (Y• )) q = 0

Hhp Hvq Hom(ZY , I • ) = H p H q (Y• , F ) =
0 q>0

Hvp Hhq Hom(ZY• , I • ) = H p Hom(Hq (ZY• ), I • ) = Extp (Hq (ZY• ), F )
and so we obtain a spectral sequence
E2pq = Extp (Hq (ZY• ), F ) =⇒ H p+q (F (Y• ))
and a similar spectral sequence for Z• . Hence we are reduced to showing that
∼
H∗ (ZZ• ) −! H∗ (ZY• ). By Lemmas 2.5.1 and 2.5.2 we are reduced to showing
that Z Hom(P, Z• ) −! Z Hom(P, Y• ) is a weak equivalence of simplicial abelian
groups for each projective P• But this is clear since Hom(P, Z• ) −! Hom(P, Y• )
is a weak equivalence and since π∗ (ZK• ), the homology of a simplicial set K• , is
a weak homotopy invariant.
(ii) =⇒ (i). There is a Carten-Leray spectral sequence

E2pq = Ȟp (Hq F ) =⇒ Hp+q F [Art62, 3.5, Ch.I.]. By assumption E2p0 = Ȟp F = 0
for p > 0 hence by induction on n one sees that Hn F = 0.
(iii) =⇒ (ii). Ȟ q ((U −! Y ), F ) = H q (F (Z• )) where Z• is the object of sT

with
Zq = U ×Y . . . ×Y U .
| {z }
q+1 times
Regarding Y as a constant simplicial object, Z• −! Y is a weak equivalence. In

effect if P is projective Hom(P, U ) −! Hom(P, Y ) is surjective; denoting this by
119
S −! T we have that Hom(P, Z• ) −! Hom(P, Y• ) is the map
... S×S×S S×S S

T T T
... T T T
which is a homotopy equivalence by the cone construction.
Lemma 2.5.3. With the notations of (2.5.2) C• (X) −! X is a weak equivalence.
Proof. Let P be projective, as F SP −! P is an effective epimorphism it follows

that P is a retract of F SF . It suffices to show therefore that
HomA (F B, C• (X)) −! HomA (F B, X)
or
HomB (B, SC• (X)) −! HomB (B, SX)
is a weak equivalence of simplicial sets. However SC• (X) −! SX is a homotopy

equivalence by the “cone construction”.
We can now finish the proof of the theorem. Let T = A/X and let I • be a
flask resolution of the sheaf hA and let P• −! X be a weak equivalence where
each Pq is projective. For the double complex I • (P• ) we have
 
H p (I • (X)) q = 0
 H p (X, A)

GT
Hhp Hvq (I • (P• )) = =
0 q>0 0 q>0
 
120
by Prop. 2.5.1 and

 
H p (hA (P• )) q = 0
 Rp ha (X) q = 0

Hvp Hhq (I • (P• )) = =
0 q>0 0 q>0
 
by Lemma 2.5.1. Thus the two spectral sequences of a double complex degen-
p p
erate giving HGT (X, A) ≃ Hcot (X, A) by Lemma 2.5.3 and condition (ii) on the
functors (3).
Examples. Let A = Grp and let G be a group. Then any abelian group object
pr2
in A/G is of the form M ×T G −−! G where M is a G module and M ×T G
denotes the semi-direct product of M and G. Hence (A/G)ab is the abelian cat-
egory of G modules. Moreover if X −! G is a group over G, then
HomA/G (X, M ×T G) = Der(X, M ), the derivative of X with values in M re-
garded as an X module via the map X −! G. For each group X over G, let
C q (X, M ) = HomSet (X q , M ) be the group of q cochain of X with values in M
and let δ : C q (X, M ) −! C q+1 (X, M ) be the usual coboundary operator. Then
δ δ
0 −! Der(·, M ) −! C ′ (·, M ) −! C 2 (·, M ) −! . . .
is a flask resolution of the sheaf hM ×T G on A/G. In effect any weak equivalence

of simplicial groups is a homotopy equivalence of sets and the functor C q (X, M )
depend only on the underlying set of X; hence C q (·, M ) is flask by Prop. 2.5.1(iii).
On the other hand the sequence is exact by Lemma 2.5.1 and the fact that
cohomology of a free group vanishes in dimension ≥ 2. Thus we find that

H q+1 (G, M ) q ≥ 1

q q q
HGT (G, M ) = Hcot (G, M ) = HH (G, M ) =
Der(G, M )

q=0
121
where H • (G, M ) is the ordinary group cohomology.
Remarks. (i) The preceding example generalizes immediately to cover the

usual cohomology of Lie algebras and associative algebras over a field ([BB66]).
Moreover one is lead to the following general picture for cohomology of any
kind of universal algebras. Letting A be a category of universal algebras
and X ∈ Ob A, then an X-module is an abelian group object A in A/X,
and the cohomology of X with values in A may be defined to be either
•
HM •
(X, A), HGT (X, A), or Hcot
•
(X, A) where the cotriple is for example the
“underlying set” and “free algebra” functors A ⇆ Set. A cochain com-
plex for computing this cohomology is just a flask resolution of the sheaf
hA on A/X.
(ii) The isomorphism HGT

•
(X, A) = H • (hA (P• )) is a special case of a gen-
eral theorem of Verdier that the Grothendieck sheaf cohomology group
H q (X, F ) may be computed as lim!
− H • (U, F ) where U runs over the
U
category of hypercoverings of X for the topology. In effect P• −! X is
cofinal in this category of hypercoverings. See [GV72] especially, exposé V,
appendice.
2.6 Modules over a simplicial ring
In this section we show how the category MR of left simplicial modules over a
simplicial ring forms closed simplicial model category. MR occurs as the category
(sA/X)ab where X is a non-constant simplicial object in sA and hence is worth
studying in virtue of §2.5. We also derive Kunneth spectral sequences which are
useful in applications. Some applications to simplicial groups are given.
In this section a ring is always associative with unit, not necessarily commu-
tative, and left or right modules are always unitary.
122
Section 2.6: Modules over a simplicial ring
Let R be a simplicial ring. By a left simplicial R-module we mean a

simplicial abelian group M together with a map R × M −! M of simplicial sets
which for each q makes Mq into a left Rq -module. The left simplicial R-modules
form an abelian category MR where a sequence is exact iff it is exact in each
dimension. The category of right simplicial R-modules is the category MRop
where Rop is the simplicial ring which is the simplicial abelian group R with the
multiplication opposed to that of R.
If X, Y ∈ Ob MR , let HomR (X, Y )n = HomMR (X ⊗Z Z∆(n), Y ) with the

simplicial operator ϕ• induced by ϕ
e in the obvious way. Here ZK denotes the
simplicial abelian group obtained by applying the free abelian group functor
dimension-wise to the simplicial set K and ⊗ denotes dimension-wise tensor prod-
uct. There is a bilinear map
HomR (X, Y ) ⊗ HomR (Y, Z) −! HomR (X, Z) (1)
defined by letting g ◦ f for f : X ⊗ Z∆(n) −! Y and g : Y ⊗ Z∆(n) −! Z be

the map
id ⊗∆ f ⊗id
X ⊗ Z∆(n) −−−−! X ⊗ Z∆(n) ⊗ Z∆(n) −−−! Y ⊗ Z∆(n) !
− Z.
It is clear that MR is a simplicial category with HomMR (X, Y ) is equal to the

underlying simplicial set of set of HomR (X, Y ) and with composition induced
by (1). If K is a simplicial set, let X ⊗Z ZK and HomS (K, Y ) be considered as
simplicial R-modules in the natural way. Then there are canonical isomorphisms
HomS (K, HomR (X, Y )) = HomMR (X ⊗Z ZK, Y ) = HomMR (X, HomS (K, X))
X ⊗Z Z(K × L) = (X ⊗Z ZK) ⊗Z ZL (2)
HomS (K × L, Y ) = HomS (L, HomS (K, Y ))
123
which may be used in the proof of Prop. 2.1.2 to show that X ⊗Z ZK is an object
X ⊗ K and that HomS (K, Y ) is an object Y K in the simplicial category MR .
We will use the notation X ⊗ ZK instead of X ⊗ K in the following.
Define a map in MR to be a fibration (resp. weak equivalence) if it is so as

a map in S, and call a map a cofibration if it has the LLP with respect to the
trivial fibrations. The proof that MR is a closed simplicial model category follows
that for SimpGrp (§2.3) and sA in the case (∗) (§2.4); in effect every object
is fibrant and factorization axiom may be proved by the small object argument.
The following descriptions hold: A map f : X −! Y in MR is a fibration if
(f, ε) : X −! Y ×K(π0 Y,0) K(π0 X, 0)
∼
is surjective, a weak equivalence if π• f : π• X −! π• Y, and a trivial fibration if f
is a surjective weak equivalence. f is a cofibration iff it is a retract of a free map,
and a trivial cofibration iff f is a cofibration and a strong deformation retract
map. Here f : X −! Y is said to be free if there are subsets Cq ⊂ Yq for each q
∼
such that C• is stable under the degeneracy operators of Y and Xq ⊕Rq Cq −! Yq
for each q.
If A is a ring, MA is the category of left A modules, and R is the constant

simplicial ring obtained from A, then MR = s(MA ) and the above structure of
a closed simplicial model category on MR is the same as that defined in §2.4.
Moreover if Ch(MA ) denotes the category of chain complexes in MA , then the
normalization functor N : MR −! Ch(MA ) is an equivalence of closed model
categories. Here Ch(MA ) is defined to be a closed model category by a slight
modification of example B.. The following fact is of course clear for Ch(MA ).
Proposition 2.6.1. Let Ω and Σ be the loop and suspension functors in the
124
category Ho(MR ). Then
∼
θ : M −! ΩΣM
∼
ΣΩM −! M ⇐⇒ π0 M = 0
where the maps are adjunction morphisms. Furthermore if
i j δ
A′ −! A −! A′′ −! ΣA′
is a cofibration sequence in Ho(MR ), then
−iθ −1 j δ
ΩΣA′ −−−−! A −! A′′ −! ΣA′
is a fibration sequence.
Proof. For any simplicial left R module X there are canonical exact sequences in
MR
0 −! X −! CX −! ΣX −! 0 (3)
0 −! ΩX −! ∧X −! X −! K(π0 X, 0) −! 0 (4)
which in more detail are the maps
•
i
X −!
1
X ⊗ Z∆(1)/X ⊗ Z{0} −! X ⊗ Z∆(1)/X ⊗ Z∆(1)
(pr ,pr ) j1 pr ε
0 ×X X ∆(1) ×X 0 −−−1−−−
! 0 ×X X ∆(1) −−−!
2 2
X −! K(π0 X, 0)
Here K(π0 X, 0) is the simplicial R module which is the constant simplicial abelian
group of π0 X with R module structure determined via ε : R −! K(π0 R, 0)
and the natural π0 R action on π0 X, and ε : X −! K(π0 X, 0) is the canonical
125
augmentation. The exactness of (3) is clear dimension-wise and (4) is exact

for simplicial groups hence also for MR , since X ∆(1) is calculated in MR as in
SimpGrp. The canonical homotopy h : ∆(1) × ∆(1) −! ∆(1) with hi0 = i0 σ
and hi1 = id induces a homotopy H : CX ⊗ Z∆(1) −! CX with Hi0 = 0 and
Hi1 = id and a homotopy K : ∧X ⊗ Z∆(1) −! ∧X with Ki0 = 0 and Ki1 = id.
Hence π(CX) = π(∧X) = 0.
The functor Ω on MR defined by (4) actually becomes the functor Ω in
Ho(MR ), since every X in MR is fibrant and so X ∆(1) is a path object for
X. Similarly one sees that ΣX represents the suspension of X in Ho(MR ) pro-
vided X is cofibrant. However if Y −! X is a trivial fibration with Y cofibrant
we obtain a map into (3) of the corresponding sequence for Y , so by the homo-
topy long exact sequence and the 5 lemma ΣY −! ΣX is a weak equivalence.
Therefore ΣX represents the suspension of X in Ho(MR ) for all X.
If π0 X = 0, then the diagram
0 ΩX CΩX ΣΩX 0
1 u v (5)
0 ΩX ∧X X 0
where u and v are induced by the contracting homotopy K of ∧X described

above, and the five lemma show that v is a weak equivalence. However v is
the adjunction map for the adjoint functors Σ and Ω in MR and hence also
in Ho(MR ), so the direction ⇐= of the second assertion of the proposition is
proved. The direction =⇒ results from the formula π0 (ΣX) = 0 which follows
since (ΣX)0 = 0. The first assertion of the proposition may be proved by a
diagram similar to (5). For the last assertion of the proposition we may assume
that i : A −! A′ is a cofibration of cofibrant objects, that A′′ is the cone on i,
that j is the embedding of A as the base of this cone, and finally that δ is the
126
cokernel of j. As MR is abelian δ is a fibration with fiber j : A −! A′′ and there

is a diagram
j
A′ i
A A′′ δ
ΣA′
−θ 1 1 1
j
ΩΣA′ ∂
A A′′ δ
ΣA′
where ∂ is the boundary operator of the fibration sequence associated to δ. For the
commutativity of the first square see proof of Prop. 1.3.6. As θ is an isomorphism
we find that ∂ = −iθ−1 and so the proposition is proved.
Kunneth spectral sequences. If X and Y are simplicial abelian groups

and if x ∈ Xp , y ∈ Yq , the the element x ∧ y ∈ (X ⊗ Y )p+q is defined by the
formula
X
x∧y = ε(µ, ν)sν x ⊗ sµ y (6)
(µ,ν)
where (µ, ν) runs over all (p, q) shuffles, i.e., permutations (µ1 , . . . , µp , ν1 , . . . , νq )
of {0, . . . , p + q + 1} such that µ1 < · · · < µp and ν1 < · · · < νq , where ε(µ, ν) is
the sign of the permutation, and where
sµ y = sµp · · · sµ1 y, sν x = sνq · · · sν1 x.
The following properties of the operation ∧ are well known.
(1) x ∈ N X, y ∈ N Y =⇒ x ⊗ y ∈ N (X ⊗ Y )
(2) d(x ∧ y) = dx ∧ y + (−1)p x ∧ dy where p = degree x and d = (−1)i di .

P
(3) x ∧ (y ∧ z) = (x ∧ y) ∧ z
127
∼
(4) If τ : X ⊗ Y −! Y ⊗ X is the isomorphism τ (x ⊗ y) = y ⊗ x, then
τ (x ∧ y) = (−1)pq τ (y ∧ x)
if p = degree x, q = degree y.
If R is a simplicial ring, then these properties show that ∧ induces on N R the

structure of a differential graded ring which is anti-commutative if R is commuta-
tive. In fact N R is even strictly anti-commutative (x2 = 0 if degree x is odd) when
R is commutative as one sees directly from (6). Consequently π• R = H• (N R)
is a graded ring which is strictly anti-commutative if R is commutative. If X is
a left (resp. right) simplicial R module then by virtue of ⊗, N X is a left (resp.
right) differential graded N R module, and so π• X is a left (resp. right) graded
π• R module.
By a projective resolution of a left simplicial R module X we mean a
trivial fibration u : P −! X in MR such that P is cofibrant. By Prop. 2.2.4 u is
unique up to homotopy over X, and moreover if we choose projective resolutions
pY : Q(Y ) −! Y for each Y ∈ Ob MR and a map Q(f ) for each map f : Y −! Y ′
such that PY ′ Q(f ) = f pY , then we obtain a functor π0 (MR ) −! π0 (MR,c ) right
adjoint to the inclusion functor. Hence projective resolution is up to homotopy
a homotopy preserving functor of X.
If X is a right simplicial R module and Y is a left simplicial R module, and if
u v
P −! X and Q −! Y are projective resolutions of X and Y in MRop and MR
respectively, then the abelian group P ⊗R Q is independent up to homotopy over
L
X ⊗R Y of the choices of u and v. We denote P ⊗R Q by X ⊗R Y and call it the
derived tensor product of X and Y since in the terminology of §1.4, it is the
total left derived functor of ⊗R : MRop × MR −! MZ .
Theorem 2.6. Let R be a simplicial ring and let X (resp. Y ) be a left (resp.
128
right) simplicial R module. Then there are canonical first quadrant spectral
sequences
(a) Epq
2
− πp (TorR
q (X, Y )) =⇒ πp+q (X ⊗R Y )
(b) Epq
2
= TorπR
p (πM, πN )q =⇒ πp+q (X ⊗R Y )
(c) Epq
2
= πp (πq X ⊗R Y ) =⇒ πp+q (X ⊗R Y )
(d) Epq
2
= πp (X ⊗R πq Y ) =⇒ πp+q (X ⊗R Y )
which are functorial in R, X, Y .
In (a) TorR
q (X, Y ) denotes the simplicial abelian group obtained by apply-
ing the functor Tor− πR

q (−, −) to R, X, Y dimension-wise. In (b) Torp (πM, πN )q
denotes the homogeneous submodule of degree in q in TorπR

p (πM, πN ) which is
naturally graded since the ring πR and the modules πM, πN are graded. In (c)
πq X is an abbreviation for the constant simplicial abelian group K(πq X, 0) which
becomes a right R module via the augmentation R −! K(π0 R, 0) and the action
πq X ⊗ π0 R −! πq X induced by ∧. Similarly for πq Y in (d).
Proof. (a) Construct recursively an exact sequence in MR
· · · −! P1 −! P0 −! X −! 0 (7)
by letting X0 = X, Pq −! Xq be the projective resolution of Xq , and

Xq+1 = ker(Pq −! Xq ). Then πPq = 0 for q > 0 so Pq −! 0 is a weak
equivalence of cofibrant objects and hence a homotopy equivalence. Hence there
is a map h : Pq ⊗Z Z∆(1) −! Pq with h(id ⊗i0 ) = id and h(id ⊗i1 ) = 0. Thus
h⊗id
(Pq ⊗R Y ) ⊗Z Z∆(1) −−−! Pq ⊗R Y
129
is a contracting homotopy of Pq ⊗R Y and so π(Pq ⊗R Y ) = 0. Think of the

simplicial operators in (7) as being vertical and consider the double complex
N•V (P• ⊗R Y ) obtained by applying the normalization to the functor to the
simplicial structure. Then Hph Hqv = 0 for q > 0 and = πp (P0 ⊗R Y ) if q = 0. As
the cofibrant simplicial R module Pq is a direct summand of a free simplicial R
module, (Pq )n is projective over Rn for each n, and so in simplicial dimension n
(7) is a projective resolution of the Rn module Xn . Hence
Hqh Nnv (P• ⊗R Y ) = Nnv Hqh (P• ⊗R Y ) = Nn TorR

q (Xn , Yn ),
n
where we have used that N is an exact functor from the simplicial abelian groups
to chain complexes.
Thus we obtain the spectral sequence
2
Epq = πp (TorR
q (X, Y )) =⇒ πp+q (P0 ⊗R Y ) (8)
having the edge homomorphism πn (P0 ⊗R Y ) −! πn (X ⊗R Y ) induced by the

map P0 −! X. By repeating this procedure with Y instead of X we obtain a
spectral sequence
2
Epq = πp (TorR
q (X, Y )) =⇒ πp+q (X ⊗R Q0 ) (9)
where Q0 −! Y is a projective resolution of Y , whose edge homomorphism

πn (X ⊗R Q0 ) −! πn (X ⊗R Y ) is induced by v. Substituting P0 for X in (9), it
degenerates showing that P0 ⊗R Q0 −! P0 ⊗R Y is a weak equivalence and hence
L
that π(X ⊗R Y ) = π(P0 ⊗R Y ). Substituting this into (8) we obtain the spectral
sequence (a) and the following fact which will be used later.
L
Corollary. The edge homomorphism π(X ⊗R Y ) −! π(X ⊗R Y ) of spectral
130
L
sequence (a) is induced by the canonical map X ⊗R Y −! X ⊗R Y . This map is
a weak equivalence if TorR
q (Xn , Yn ) = 0 for q > 0, n ≥ 0.
n
To prove (a) is functorial let R, X, Y −! R′ , X ′ , Y ′ be a map and suppose

that a sequence (7)’ corresponding to (7) has been constructed. As a map of
simplicial sets, the maps Pq′ −! Xq′ are trivial fibrations as maps in MRop . Hence
we may construct a map θ from (7) to (7)’ covering the given map X −! X ′ by
inductively defining θq : Pq −! Pq′ by lifting in
0 Pq′
θq
(10)
θq−1
Pq Xq Xq′
We then obtain a map of the spectral sequence (8) into the corresponding one (8)’
which is independent of the choice of θ because its E 2 term is clearly independent
and the map P0 −! P0′ covering X −! X ′ is unique upto homotopy. Conse-
quently there is a canonical map from spectral sequence (a) to the corresponding
one (a)’ and this proves the functoriality of (a) as well as its independence of the
choices made for its construction.
(b) We need two lemmas.
Lemma 2.6.1. Suppose that P is a cofibrant right simplicial R module such

that π• P is a free graded π• R module. Then for any left simplicial R module Y
the map
π• P ⊗π• R π• Y −! π• (P ⊗R Y )
induced by ∧ is an isomorphism.
Lemma 2.6.2. Suppose P is as in Lemma 2.6.1 and let f : X −! Y be a
131
fibration in MRop such that π• f : π• X −! π• Y is surjective. Then given any

map u : P −! Y there is a v : P −! X with f v = u.
•
Proof. Let RS n = Coker(R ⊗ Z∆(n)) n ≥ 1 considered as a right simplicial R
module in the obvious way, let tn ∈ (RS n )n be the residue class of the element
1 ⊗ id[n] , and let un be the element of πn (RS n ) represented by tn . We claim that
A. π(RS n ) is a free right graded πR module generated by un .
B. The map π(RS n )⊗πR πY −! π(RS n ⊗R Y ) induced by ∧ is an isomorphism.
Indeed there is an exact sequence of right simplicial R modules
i j
0 −! RS n−1 −! RDn −! RS n −! 0 (11)
where RDn = Coker(R ⊗ ZV (n, 0) −! R ⊗ Z∆(n)), where i is induced by

∂˜0 : ∆(n − 1) −! ∆(n) and j is the canonical surjection. Moreover 0 −! RDn
is a trivial cofibration because it is a cobase extension of the map
R ⊗ ZV (n, 0) −! R ⊗ Z∆(n), which is a trivial cofibration by SM7 since R is
cofibrant. Hence RDn is contractible and the long exact sequence in homotopy
yields an isomorphism

πq−1 (RS n−1 ) q ≥ 1

∂
πq (RS n ) −! .
∼
0 q = 0.

such that ∂un = un−1 . By property 2 of ∧ ∂ is an isomorphism

∼
π(RS n ) −! Σπ(RS n−1 ) of right graded πR modules, where if M is a right graded
module over a graded ring S, we defined ΣM to be the right graded R module
with (ΣM )k = Mk−1 and (Σm)s = Σ(ms); here if m ∈ Mk−1 , Σm denotes m as
an delement of (ΣM )k . A then follows by induction on n. To obtain B note that
(11) splits in each dimension so it remains exact after tensoring with Y over R.
132
The resulting long exact homotopy sequence yields the bottom isomorphism in
the square
∂⊗id
π(RS n ) ⊗πR πY Σπ(RS n−1 ) ⊗πR πY
∂
π(RS n ⊗R Y ) ∼ Σπ(RS n−1 ⊗R πY )
where the vertical arrows come from ∧ and the diagram commutes by property
2 of ∧. Induction on n then proves B.
If P is as in Lemma 2.6.1 choose elements x1 ∈ Pn1 , i ∈ I whose represen-
tatives in πP form a free basis over πR and let Ψ : ⊕ RS ni −! P the map of
right simplicial R modules sending tn1 to x1 . By the assumption on P and A
Ψ is a weak equivalence hence a homotopy equivalence since both are cofibrant.
Lemma 2.6.1 then reduces to the case P = RS n in which case it follows from B.
To prove Lemma 2.6.2 we reduce by the covering homotopy theorem to the
case P = RS n , and we must show that Zn f : Zn X −! Zn Y is surjective where
Zn denotes the group of n cycles in the normalization. As f is a fibration Nj f is
surjective j > 0 and as πf is surjective one sees easily that Zf is surjective.
To obtain (b) construct an exact sequence
−! P1 −! P0 −! X −! 0 (12)
of right simplicial R modules by setting X0 = X, Xq+1 = Ker(uq : Pq −! Xq )

where uq is surjective, πuq is surjective, and πPq is a free graded right πR
module. uq may be obtained by choosing generators {α1 } for πXq over πR,
letting v : ni
−! Xq be a map sending tn1 onto a representative for
L
i RS
α1 and then factoring v = uq i where uq is a fibration and i is a trivial fibra-
133
tion. If Q −! Y is a projective resolution of Y , consider the double complex

N•V (P• ⊗R Q) where v refers to the (vertical) simplicial structure. By Lemma
2.6.1 π(Pq ⊗R Q) = πPq ⊗πR πQ and by the construction of (12), π(P• ) is a free
πR resolution of πX. Thus
Hph Hqv (N• (P• ⊗R Q)) = Hph (πP• ⊗πR πQ)q = TorπR
p (πX, πQ)q .
On the other hand, Q is projective over R in each dimension, hence
Hpv Hqh (N• (P• ⊗R Q) = Hpv N• Hqh (P• ⊗R Q) = 0
∼ L
if q > 0 and πp (X ⊗R Q) if q = 0. As π(X ⊗R Q) ! π(X ⊗R Y ) by the
above corollary and π(Q) = π(Y ) we obtain spectral sequence (b) as well as its
independence of (12) may be proved in exactly the same was as for (a), except
the lifting analogous to (10) is constructed via Lemma 2.6.2.
(c) These are derived by the Serre-Postnikov method. In effect we have (see
Prop 2.6.1 (4)) canonical exact sequences
^
0 −! ΩX −! X −! X −! π0 X −! 0 (13)
in MRop , where X is contractible and where π0 X is short for the right simplicial
V
R module which is the constant simplicial abelian group K(π0 , X, 0), and whose
R module comes via the augmentation ε : R −! K(π0 R, 0) from the map
π0 X ⊗ π0 R −! π0 X induced by ∧. From the long exact homotopy sequence we
have
∂
πq (X) −! πq−1 (ΩX) q > 0 (14)
∼
where ∂(α · ρ) = (∂α)ρ if ρ ∈ πR. Hence substituting Ωk X into we obtain exact
134
sequences
^
0 −! Ωk+1 X −! Ωk X −! Ωk X −! πk X −! 0 k ≥ 0 (15)
where πk X stands for the right simplicial R module as described in the theorem.
Letting Q −! Y be a projective resolution of Y , ⊗R Q is exact and Ωk X ⊗R Q
V
is contractible, hence from (15) we obtain exact sequences
−! πn−1 (Ωk+1 X⊗R Q) −! πn (Ωk X⊗R Q) −! πn (πk X⊗R Q) −! πn−2 (Ωk+1 X⊗R Q) −! . . . .
L
for k ≥ 0. By the corollary − ⊗R Q may be replaced by − ⊗R Y and so we obtain
L L
an exact couple (Dpq
2 2
, Epq ) with Epq
2
= πp (πq X ⊗R Y ) and Dpq
2
= πp (Ωq X ⊗R Y )
and hence the spectral sequence (c). It is clearly canonical and functorial since
the only only choice made was that of Q which is unique and functorial up to
homotopy. Spectral sequence (d) is proved similarly. There is no sign trouble
from the fact that ∂ : πq Y −! πq−1 (ΩY ) satisfies ∂(ρα) = (−1)k ρ · ∂α if ρ ∈ πk R
because only k = 0 occurs when we consider πk Y as a left simplicial R module.
Theorem 2.6 is now proved.
Applications to simplicial groups. Let G be a simplicial group. If M

L
is a simplicial G module we call H• (G, M ) = π(Z ⊗ZG M ) the homology of
G with coefficients in M . Here Z is short for K(Z, 0) with trivial G action.
To calculate the homology we choose a projective resolutive of Z as a right ZG
module, e.g. ZW G where W G −! W G is the universal principal G bundle,
whence H(G, M ) = π(ZW G ⊗ZG M ). If M is an abelian group on which π0 G
acts and we consider M as a constant simplicial G module, then it follows that
H• (G, M ) is the homology of the simplicial set W G with values in the local
coefficient system defined by M . In particular when G is a constant simplicial
group and M is a G module H(G, M ) in the above sense coincides with the
135
ordinary group homology of G with values in M .
If F is a free group, then



Z q=0


TorZF
q (Z, Z) = Fab q=1 .




0 q >= 2.
hence if G is a simplicial group which is free in each dimension spectral sequence

(a) degenerates giving

n=0

Z
Hn (G, Z) = (16)
πn−1 (Gab ) n > 0.

which is a formula due to [Kan57a] when G is a free simplicial group.
Let f : G −! H be a weak equivalence of simplicial groups. Then f is a weak

equivalence in Simpf and as every object of Simp is cofibrant f is a homotopy
equivalence in Simp. Thus ZG −! ZH is a homotopy equivalence of simplicial
∼
abelian groups and so πZG ! πZH. From spectral sequence (b) we deduce that
∼
H• (G, Z) ! H• (H, Z) which shows that homology is a weak homotopy invariant.
Suppose now that

1 −! K −! G −! H −! 1
is an exact sequence of simplicial groups and that M is a simplicial G module.

Let P −! M be a projective resolution of M as on left ZG module. Then
∼ ∼
Z ⊗ZG M ! Z⊗ZG ! Z ⊗ZH (ZH ⊗ZG P ).
and
∼
πq (ZH ⊗ZG P ) ! πq (Z ⊗ZK P ) = Hq (K, M ).
136
Substituting R = ZH, X = Z, Y = Z ⊗ZK P in spectral sequence (d) we obtain a
spectral sequence
2
Epq = Hp (H, Hq (K, M )) =⇒ Hp+q (G, M ) (17)
which generalizes the Hoschild-Serre spectral sequence for group homology and
the Serre spectral sequence for the fibration W K −! W G −! W H.
Spectral sequence (a) with R = ZG, X = Z, Y = Z has the edge homomor-
phism
Hn (G, Z) −! πn−1 (Gab ) n > 0.
which is an isomorphism for n = 1 in general and for all n if G is free. So we

obtain Poincare’s theorem
H1 (G, Z) = (π0 G)ab .
Now by the method of [Ser53] it is possible to start from this fact and the spec-
tral sequence (17) and prove directly the Hurewicz and Whitehead theorems for
simplicial groups. We leave the details to the reader.
137
138
Bibliography
[AM67] M. Artin and B. Mazur. “Homotopy of Varieties in the Etale Topology”.

In: Proceedings of a Conference on Local Fields. Ed. by T. A. Springer.
Berlin, Heidelberg: Springer, 1967, pp. 1–15. isbn: 978-3-642-87942-5. doi:
10.1007/978-3-642-87942-5_1.
[Art62] Michael Artin. Grothendieck Topologies: Notes on a Seminar. Google-Books-
ID: wERAAQAAIAAJ. Harvard University, Department of Mathematics,
1962. 146 pp. url: https://www.math.nagoya-u.ac.jp/~larsh/teaching
/S2013_AG/grothendiecktopologies.pdf.
[BB66] Michael Barr and Jon Beck. “Acyclic Models and Triples”. In: Proceedings
of the Conference on Categorical Algebra. Ed. by S. Eilenberg et al. Berlin,
Heidelberg: Springer, 1966, pp. 336–343. isbn: 978-3-642-99902-4. doi: 10
.1007/978-3-642-99902-4_16. url: https://www.math.mcgill.ca/barr
/papers/amt.pdf.
[BGM59] M. G. Barratt, V. K. A. M. Gugenheim, and J. C. Moore. “On Semisim-
plicial Fibre-Bundles”. In: American Journal of Mathematics 81.3 (1959).
Publisher: Johns Hopkins University Press, pp. 639–657. issn: 0002-9327.
doi: 10.2307/2372920. url: https://www.jstor.org/stable/2372920
(visited on 12/04/2022).
[Dol58] Albrecht Dold. “Homology of Symmetric Products and Other Functors of
Complexes”. In: Annals of Mathematics 68.1 (1958). Publisher: Annals of
Mathematics, pp. 54–80. issn: 0003-486X. doi: 10 . 2307 / 1970043. url:
https://www.jstor.org/stable/1970043 (visited on 12/04/2022).
[DP61] Albrecht Dold and Dieter Puppe. “Homologie nicht-additiver Funktoren.
Anwendungen”. In: Annales de l’institut Fourier 11 (1961). Publisher: As-
sociation des Annales de l’Institut Fourier, pp. 201–312. issn: 0373-0956.
url: https://eudml.org/doc/73776 (visited on 12/04/2022).
[Ger65] Hillel H. Gershenson. “Higher composition products.” In: Journal of Math-
ematics of Kyoto University 5.1 (Jan. 1965). Publisher: Duke University
Press, pp. 1–37. issn: 0023-608X. doi: 10 . 1215 / kjm / 1250524554. url:
https://projecteuclid.org/journals/journal-of-mathematics-of-ky
oto-university/volume-5/issue-1/Higher-composition-products/10
.1215/kjm/1250524554.full (visited on 12/04/2022).
[GV72] A. Grothendieck and J. L. Verdier. Théorie des Topos et Cohomologie Etale
des Schémas. Vol. 269. Lecture Notes in Mathematics. Berlin, Heidelberg:
Springer, 1972. isbn: 978-3-540-05896-0 978-3-540-37549-4. doi: 10.1007
/ BFb0081551. url: http : / / link . springer . com / 10 . 1007 / BFb0081551
(visited on 12/04/2022).
139
[GZ67] Peter Gabriel and Michel Zisman. Calculus of Fractions and Homotopy
Theory. Berlin, Heidelberg: Springer, 1967. isbn: 978-3-642-85846-8 978-3-
642-85844-4. doi: 10.1007/978- 3- 642- 85844- 4. url: https://people
.math.rochester.edu/faculty/doug/otherpapers/GZ.pdf (visited on
12/04/2022).
[Har66] Robin Hartshorne. “Residues and duality”. In: (1966). Publisher: Springer(Berlin
[u.a.]) url: https://eudml.org/doc/203789 (visited on 12/04/2022).
[Kan57a] Daniel M. Kan. “On c.s.s. Categories”. In: Boletín de la Sociedad Matemática
Mexicana (1957), pp. 82–94.
[Kan57b] Daniel M. Kan. “On c.s.s. Complexes”. In: American Journal of Mathe-
matics 79.3 (1957). Publisher: Johns Hopkins University Press, pp. 449–
476. issn: 0002-9327. doi: 10 . 2307 / 2372558. url: https : / / people . ma
th.rochester.edu/faculty/doug/otherpapers/kan57.pdf (visited on
12/04/2022).
[Kan58a] Daniel M. Kan. “On Homotopy Theory and C.S.S. Groups”. In: Annals of
Mathematics 68.1 (1958). Publisher: Annals of Mathematics, pp. 38–53.
issn: 0003-486X. doi: 10.2307/1970042. url: https://www.jstor.org/st
able/1970042 (visited on 12/04/2022).
[Kan58b] Daniel M. Kan. “The Hurewicz Theorem”. In: Proceedings of the Interna-
tional Symposium on ALgebraic Topology and its applications. Symposium
Internacional de Topología Algebraica. OCLC: 560705. México: Universi-
dad Nacional Autónoma de México y la UNESCO, 1958, pp. 225–231. url:
https://aareyanmanzoor.github.io/assets/articles/Kan- Hurewicz-
Theorem.pdf.
[Law63] F. William Lawvere. “Functorial semantics of algebraic theories*”. In: Pro-
ceedings of the National Academy of Sciences 50.5 (Nov. 1963). Publisher:
Proceedings of the National Academy of Sciences. Reprint: http://tac.mta.ca/tac/reprints/
pp. 869–872. doi: 10.1073/pnas.50.5.869. url: https://www.pnas.org
/doi/10.1073/pnas.50.5.869 (visited on 12/04/2022).
[Mil57] John Milnor. “The Geometric Realization of a Semi-Simplicial Complex”.
In: Annals of Mathematics 65.2 (1957). Publisher: Annals of Mathematics,
pp. 357–362. issn: 0003-486X. doi: 10.2307/1969967. url: https://stat
ic.aminer.org/pdf/PDF/000/225/050/geometric_realization_of_simp
licial_complexes.pdf (visited on 12/04/2022).
[Ser53] Jean-Pierre Serre. “Groupes D’Homotopie Et Classes De Groupes Abeliens”.
In: Annals of Mathematics 58.2 (1953). Publisher: Annals of Mathematics,
pp. 258–294. issn: 0003-486X. doi: 10.2307/1969789. url: https://peopl
e.math.rochester.edu/faculty/doug/otherpapers/Serre-Classes.pdf
(visited on 12/04/2022).
[Spa63] E. Spanier. “Higher Order Operations”. In: Transactions of the American
Mathematical Society 109.3 (1963). Publisher: American Mathematical So-
ciety, pp. 509–539. issn: 0002-9947. doi: 10.2307/1993858. url: https:
//www.jstor.org/stable/1993858 (visited on 12/04/2022).
[Ver] Jean-Louis Verdier. “Des catégories dérivées des catégories abéliennes”. In:
(), p. 270. url: http://www.numdam.org/item/?id=AST_1996__239__R1_0.
140
Index
cartesian, 16 left homotopy, 18

closed model category, 13, 59 left lifting property, 59
closed simplicial model category, 66, 77 left-derived functor, 52
co-cartesian, 16 — total, 54
cochain complex, 122 left-homotopic, 17
codiagional, 17 localization, 23
cofibrant, 12, 15
cofibration sequence, 42
cofibre, 35 model category, 11
cofibre product, 16 — closed, 59
composition, 32, 67 module, 122
constant homotopy, 72 Moore homotopy groups, 88
correspondence, 29
cylinder object, 12, 18
null-object, 35
derived tensor product, 128
diagonal, 16 path object, 18
pointed category, 35
fibrant, 12, 15 pointed model category, 35
— co, 12, 15 presheaf, 115
fibration, 77 projective resolution, 128
— trivial, 15
— trivial co, 15
fibration sequence, 42 representable, 71
fibre, 35 right homotopy, 18
finite simplicial set, 73 right-derived functor, 52
flask, 118 right-homotopic, 17
generalized unit interval, 71 simplicial categories, 66

simplicial functor, 68
homotopy, 12, 71 simplicial set
— category, 12, 23 — finite, 73
— constant, 72 strictly homotopic, 71, 72
— left, 18, 29
— right, 18, 29
— strict, 71, 72 Toda bracket, 51
total left-derived functor, 54
inverse, 32 trivial fibration, 77
141
Daniel Quillen, 1940–2011, Fields Medalist, transformed many
aspects of algebra, geometry, and topology. Especially in a
succession of remarkable papers during the ten-year period of
1967–1977, Quillen created astonishing mathematics which
continues to inspire current research in many fields. Quillen’s
mathematical exposition serves as the ultimate model of clarity.
Despite his brilliance, those who knew Quillen were regularly
impressed by his generosity and modesty
(Sources: Memorium article, edited by Eric

Friedlander and Daniel Grayson)

Quillen Homotopical-Algebra

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Quillen Homotopical-Algebra

Uploaded by

Copyright:

Available Formats

Daniel Quillen

Revised and modernized edition by

This book is the product of a community effort. It was typeset by TEXromancers:

More books in this series:

• Noel J. Hicks, Notes on Differential Geometry

• J.C. Lagarias and A.M. Odlyzko, Effective Versions of the Chebotarev

• Daniel Quillen, Homotopical Algebra

Revised and modernized edition by

Original edition: 1965

1 Axiomatic homotopy theory 11

2 Examples of simplicial homotopy theories 65

Homotopical algebra or non-linear homological algebra is the generalization

All diagrams are redrawn in tikz or tikzcd.

Chapter 1 is an attempt to define what is meant by a “homotopy theory” in a

M1 Given a commutative solid arrow diagram

where i is a cofibration, p is a fibration, and either i or p is also a weak

M2 Any map f may be factored f = pi and f = p′ i′ where i, i′ are cofibrations,

Call an object cofibrant if the map ∅ ! X is a cofibration (hence in the

where ∨ denotes direct sum, f + g is the map with components f and g,

Given a model category C, the homotopy category Ho C is obtained from

two definitions coincide.

1.1 The axioms

All diagrams are assumed to be commutative unless stated otherwise.

Definition 1.1.1. By a model category we mean a category together with

M0 C is closed under finite projective and inductive limits.

M1 Given a solid arrow diagram

where i is a cofibration, p is a fibration, and where either i or p is a weak

M2 Any map f may be factored f = pi where i is a cofibration and weak

Examples. A. Let C be the category of topological spaces and continuous

B. Let A be an abelian category with sufficiently many projectives and let

C. Let C be the category of semi-simplicial sets and let fibrations in C be

there is a unique map A −! B ×Y X denotes (α, β)Y or simply (α, β) such

and the codiagional map

of f . We write ∆X (resp. ∇Y ) if Y = e (resp. X = ∅).

Definition 1.1.3. Let f, g : A ⇒ B be maps. We say that f is left-homotopic

where σ is a weak equivalence. Dually we say that f is right-homotopic to g

where s is a weak equivalence.

homotopic =⇒ right homotopic =⇒ left homotopic (5)

Definition 1.1.4. By cylinder object for an object A we mean an object A × I

such that ∂0 + ∂1 is a cofibration and σ is a weak equivalence. Dually, a path

where s is a weak equivalence and (d0 , d1 ) is a fibration. By a left homotopy

from f to g is a diagram (4) where B

Remark. 1. A × I is not the product of A and an object I nor is it a functor

2. Since the dual of a model category is again a model category in an evident

where ∂0′ + ∂1′ is a trivial cofibration and ρ is a trivial fibration. By M5

σ ′ = σρ : A′ −! A is a weak equivalences so A′ with ∂0′ , ∂1′ , and σ ′ is a cylinder

Lemma 1.1.2. Let A be a cofibrant object and let A × I be a cylinder object

Proof. in1 : A −! A ∨ A is a commutativ by the coase change assertion in M3,

Corollary (Covering Homotopy theorem). Let A be cofibrant and let

Proof. By M1, H exists in

The dual assertion is the homotopy extension theorem.

Lemma 1.1.3. Let A be cofibrant and let A × I and A × I ′ be two cylinder

∂1 A = ∂0′ A, defined precisely to be the object A

is also a cylinder object A × I ′′ for A with

∂0′′ = in1 ∂0 , ∂1′′ = in2 ∂1′ , σ ′′ in1 = σ, σ ′′ in2 = σ ′ .

Proof. The relation is reflexive since if f = g we may take A

Lemma 1.1.5. Let A be cofibrant and let f, g ∈ Hom(A, B). Then

Proof. (i) By Lemma 1.1.1 there is a left homotopy h : A × I −! B from f to

and k = K∂1 : A −! B I is the desired right homotopy from f to g.

and k gives the desired homotopy from f to g.

(iii) Let k be as in (ii) and let C I be a path object for C. By M1 it is possible

and kϕ : A −! C I is a right homotopy from uf to ug.

Lemma 1.1.6. If A is cofibrant, then composition in C induces a map