Lenny Taelman - Categories and Modules

Modules and Categories∗
Lenny Taelman
∗Preliminary version, August 17, 2023

Contents
Foreword 7
Prerequisites 7
Other sources 7
Acknowledgements 7
Chapter 1. Modules over a ring 9

1. Left and right modules 9
2. First examples 10
3. Homomorphisms, submodules and quotient modules 12
4. Products, direct sums and free modules 13
Exercises 17
Chapter 2. Exact sequences 21

1. Exact sequences 21
2. The Five Lemma and the Snake Lemma 22
3. Split short exact sequences 24
Exercises 26
Chapter 3. Finitely generated modules over a PID 31

1. Introduction 31
2. Review of principal ideal domains 31
3. Free modules of finite rank over a PID 32
4. Structure of finitely generated modules over a PID 33
5. Application to Jordan normal form 35
Exercises 38
Chapter 4. Categories 41
1. Definition 41
2. Big examples 42
3. Small examples 42
4. Final and cofinal objects 45
3
4 CONTENTS
Exercises 46
Chapter 5. Functors 49
1. Definition of a functor 49
2. Many examples 49
3. Contravariant functors 52
4. Functors with multiple arguments 53
Exercises 55
Chapter 6. Morphisms of functors 57

1. Morphisms of functors 57
2. Equivalences of categories 58
Exercises 62
Chapter 7. Tensor product 65

1. Tensor product of a right and a left module 65
2. Tensor products and bimodules 68
3. Tensor product as a functor 70
4. The adjunction 72
Exercises 74
Chapter 8. Adjoint functors 77

1. Adjoint pairs of functors 77
2. Many examples 78
3. Yoneda and uniqueness of adjoints 81
Exercises 84
Chapter 9. Limits and colimits 87

1. Product and coproduct 87
2. Pullback and pushout 90
3. Limits and colimits 93
4. Yoneda and limits and colimits of sets 96
5. Adjoint functors and limits 98
Exercises 100
Chapter 10. Chain complexes 105

1. Chain complexes and their homology modules 105
2. The long exact sequence 106
3. The homotopy category 107
Exercises 110
CONTENTS 5
Chapter 11. Free resolutions 113

1. Definition and existence 113
2. The free resolution functor 115
Exercises 119
Chapter 12. The Ext functors 121
1. The functors Extn 121
2. The long exact sequence 123
3. Ext1 and extensions 124
Exercises 127
Bibliography 131
Foreword
These are course notes for a one-semester third-year course on cat-

egories and modules taught at the University of Amsterdam.
Prerequisites
Linear algebra, groups, elementary point-set topology, and the ba-
sics of rings and fields. Several exercises and examples in the parts
about categories refer to basic concepts in algebraic topology, Galois
theory, or representation theory. Although in principle these can be
skipped, developing a rich vocabulary of natural examples is probably
the most important aspect of becoming acquainted with categories.
Other sources
Excellent alternative sources for much of the material in this course
are the concise but clear and well-written Atiyah & MacDonald [1], the
more lengthy Lang [2], the course notes by Moerdijk [3] and (in Dutch)
the Algebra 2 course notes by Stevenhagen [4]. A good introduction
in the language of categories and functors containing much more than
these notes is [5].
Acknowledgements
Many thanks to Wessel Bindt, David de Boer, Jeroen Dekker, Koen
van Duin, Eline Filius, Pepijn Hoefgeest, Tycho van Hoof, Juultje Kok,
Christopher Spelt, Tijn de Vos, for their many corrections and sugges-
tions to the first drafts of these course notes.
7
CHAPTER 1
Modules over a ring
1. Left and right modules

Let R be a ring. Recall that this means R is a set equipped with
an addition (s, t) 7→ s + t, multiplication (s, t) 7→ st, and distinguished
elements 0 ∈ R and 1 ∈ R satisfying
(R1) (R, +, 0) is an abelian group,
(R2) for all r, s, t ∈ R we have (rs)t = r(st),
(R3) for all r, s, t ∈ R we have r(s+t) = rs+rt and (r+s)t = rt+st,
(R4) for all r ∈ R we have 1r = r1 = r.
Definition 1.1. A left module over a ring R is an abelian group M
equipped with an operation
R × M → M, (r, x) 7→ rx
satisfying for all r, s ∈ R and x, y ∈ M the following identities:
(M1) r(x + y) = rx + ry,
(M2) (r + s)x = rx + sx,
(M3) (rs)x = r(sx),
(M4) 1x = x.
One also says that ‘R acts on M ’, so that axiom (M3) for example
expresses that acting by rs is the same as first acting by s, and then
by r.
We use the same symbol 0 to denote the elements 0 ∈ M and 0 ∈ R.
This should not lead to confusion, see Exercise 1.1.
A right module over R is defined similarly: the action is written on
the right: M × R → M, (x, r) 7→ xr, and must satisfy
(M1’) (x + y)r = xr + yr,
(M2’) x(r + s) = xr + xs,
(M3’) x(rs) = (xr)s,
(M4’) x1 = x.
9
10 1. MODULES OVER A RING
If R is commutative, then the difference between a left and a right

module is purely a matter of notation, but over a non-commutative ring
the axioms (M3) and (M3’) give genuinely different conditions.
We will mostly work with left modules, and simply call them mod-
ules over R or R-modules.
There is another way to describe (left) modules, using the endomor-
phism ring of an abelian group. Let A be an abelian group. Denote by
End(A) the set of group homomorphisms A → A. This forms a ring
with addition and multiplication of f, g ∈ End(A) defined by pointwise
addition
f + g : A → A, a 7→ f (a) + g(a)
and composition
f g : A → A, a 7→ f (g(a)).
The zero element of this ring is the constant map 0 : a 7→ 0, and the
unit element is the identity map idA : a 7→ a.
Lemma 1.2. Let M be an R-module. Then the map
R → End(M ), r 7→ (x 7→ rx)
is a ring homomorphism. Conversely, let M be an abelian group and
ϕ : R → End(M ) be a ring homomorphism. Then the operation
R × M → M, (r, x) 7→ rx := ϕ(r)(x)
gives M the structure of an R-module.
Proof. See Exercise 1.3. □
In other words, a (left) R-module is the same as an abelian group
M together with a ring homomorphism R → End(M ).
2. First examples
Example 1.3. Let R be a ring. Then the trivial group {0} is an R-
module with r0 := 0 for all r ∈ R. We denote this module by 0, and
call it the zero module.
Example 1.4. Let R be a ring and n ≥ 0. Then M := Rn is an
R-module with addition
(x1 , . . . , xn ) + (y1 , . . . , yn ) := (x1 + y1 , . . . , xn + yn )
and R-action
r · (x1 , . . . , xn ) := (rx1 , . . . , rxn ).
2. FIRST EXAMPLES 11
For n = 0 we obtain the zero module R0 = 0 and for n = 1 we obtain

the R-module R.
Example 1.5. For every abelian group A there is a unique ring homo-
morphism Z → End(A). It follows that a Z-module is the same as an
abelian group. For r ∈ Z and x ∈ A we have
(
x + · · · + x (r terms) r≥0
r·x=
−(x + · · · + x) (−r terms) r ≤ 0
Example 1.6. Let K be a field. Then a K-module is the same as a
K-vector space.
Example 1.7. Let K be a field and n ≥ 0. Let Matn (K) be the ring
of n by n matrices over K. Then K n is a left Matn (K)-module via
Matn (K) × K n 7→ K n , (A, v) 7→ A · v,
where we interpret vectors v ∈ K n as column matrices.
Example 1.8. Let R be a ring and I ⊂ R an ideal. Then I is an
R-module.
Example 1.9. Let K be a field. Let V be a K-vector space, and
α : V → V be a K-linear endomorphism of V . Then there is a unique
ring homomorphism
ρ : K[X] → End(V )
such that
(1) ρ(λ)(v) = λv for all λ ∈ K and v ∈ V,
(2) ρ(X) = α.
This homomorphism is given by
X X
(1) ρ: λi X i 7→ v 7→ λi αi (v) ,
where αi denotes the iterated composition α ◦ · · · ◦ α. In particular, V

obtains the structure of a K[X]-module.
Conversely, given a K[X]-module V , the restriction of the action of
K[X] to K makes V into a K-vector space, and the map
α : V → V, v 7→ X · v
is K-linear. We conclude that a K[X]-module is the same as a K-vector
space equipped with an endomorphism (namely the action of X).
We will see in Chapter 3 that the ‘Jordan normal form’ of complex

square matrices (C-linear endomorphisms of Cn ), is really a theorem
about the structure of C[X]-modules, and that it is most naturally
explained in terms of ideals in C[X].
Example 1.10 (The group algebra and representations). Let K be
a field and G a group. Let K[G] be the group algebra of G over K.
Elements of K[G] are formal expressions
X
ag g (ag ∈ K)
g∈G
with ag = 0 for all but finitely many g (this is automatic if G is a

finite group). Addition is defined in the obvious way. Multiplication is
defined by extending the multiplication in G. We have
  !
X X X
 ag g  · bh h = ct t
g∈G h∈G t∈G
with X
ct = ag bh .
gh=t
A K[G]-module is the same as a K-vector space V , together with a
group homomorphism
G → GLK (V ) = EndK (V )× .
In other words, a K[G]-module is a K-linear representation of G.
3. Homomorphisms, submodules and quotient modules

Definition 1.11. Let M and N be R-modules. An R-module homo-
morphism from M to N is a map f : M → N such that for all r ∈ R
and x, y ∈ M we have
f (x + y) = f (x) + f (y)
and
f (rx) = rf (x).
We also say that the map f : M → N is R-linear. The set of R-module
homomorphisms from M to N is denoted HomR (M, N ). An isomor-
phism of R-modules is a bijective R-module homomorphism. Two R-
modules are called isomorphic if there exists an isomorphism between
them.
4. PRODUCTS, DIRECT SUMS AND FREE MODULES 13
A submodule of an R-module M is a subgroup N ⊂ M such that for

all r ∈ R and x ∈ N we have rx ∈ N . A submodule of an R-module is
itself an R-module. If N ⊂ M is a submodule, then the abelian group
M/N has the structure of an R-module, via
r(x + N ) := rx + N.
We call M/N the quotient module.
To a module homomorphism f : M → N are associated three im-
portant modules. The kernel
ker f := {x ∈ M | f (x) = 0} ⊂ M,
which is a submodule of M , the image
im f := f (M ) ⊂ N,
which is a submodule of N , and the cokernel
coker f := N/(im f ),
which is a quotient module of N . A homomorphism f is injective if
and only if ker f is trivial, and it is surjective if and only if coker f is
trivial.
As with groups or vector spaces, we have the natural isomorphism
∼
M/(ker f ) → im f, x̄ 7→ f (x).
4. Products, direct sums and free modules

Let R be a ring and let M and N be R-modules. The cartesian
product M × N is naturally an R-module with (x1 , x2 ) + (y1 , y2 ) =
(x1 + y1 , x2 + y2 ) and r(x1 , x2 ) = (rx1 , rx2 ). We call this R-module the
product or direct product of the R-modules M and N . More generally,
Q(Mi )i∈I is a family of R-modules indexed by a set I, then the product
if
i∈I Mi is naturally an R-module with
(xi )i∈I + (yi )i∈I = (xi + yi )i∈I , r(xi )i∈I = (rxi )i∈I ,
Q
for all (xi )i∈I , (yi )i∈I in i∈I Mi and r ∈ R. The empty product gives
the zero module.
The direct L sum of a collection (Mi )i∈I of R-modules
Q indexed by a
set I, denoted i∈I Mi is the R-submodule of i∈I Mi defined as
M n Y o
Mi := (xi )i∈I ∈ Mi | {i ∈ I : xi ̸= 0} is finite .
i∈I i∈I
Note that this is indeed a submodule: if (xi )i∈I and (yi )i∈I have only
finitely many non-zero terms, then so do (xi + yi )i∈I and (rxi )i∈I .
One sometimes phrases the condition on (xi )i∈I as ‘xi is zero for
all but finitely many L i’. Of course,
Q if I is finite then the condition is
vacuous and we have i∈I Mi = i∈I Mi .
We have natural ‘inclusion’ maps
(
M x i=j
ι j : Mj → Mi , x 7→ ιj (x), ιj (x)i =
i∈I
0 i ̸= j
and ‘projection’ maps
Y
πj : Mi → Mj , (xi )i∈I 7→ xj .
i∈I
If we have Mi = M for all i, then we write M I :=

Q
i∈I M and
M (I) := ⊕i∈I M , so that we have
M I = {(xi )i∈I | xi ∈ M }
and
M (I) = (xi )i∈I | xi ∈ M, and xi = 0 for all but finitely many i .

If I is a finite set of cardinality n, then we have M I = M (I) ∼

= M n.
Let (xi )i∈I be a family of elements of an R-module M . Then we
call the intersection of all submodules N ⊂ M that contain all xi the
submodule generated by (xi )i∈I . We denote it by ⟨xi ⟩i∈I . It is the
smallest submodule of M containing all the xi . It consists of all finite
R-linear combinations of the xi .
We say that M is finitely generated if there exists a finite family
(xi )i∈I with M = ⟨xi ⟩.
Example 1.12. Let R be a ring and I a set. Consider the module
n o
R(I) = (ri )i∈I ∈ RI | ri = 0 for all but finitely many i ∈ I .
For an index i ∈ I we denote by ei ∈ R(I) the element ei := ιi (1). One

may think of ei as the ‘standard basis’ element
ei = (. . . , 0, 0, 1, 0, 0, . . .)
with a 1 at position i. Clearly, every element of R(I) is a finite R-linear
combination of ei ’s, so the family (ei )i∈I generates R(I) . Note that if I
is infinite, then the ei do not generate the direct product RI .
4. PRODUCTS, DIRECT SUMS AND FREE MODULES 15
Proposition 1.13. Let R be a ring, M an R-module, and (xi )i∈I a

family of elements of M . Then there exists a unique R-linear map
φ : R(I) → M with φ(ei ) = xi for every i ∈ I.
Moreover, φ is surjective if and only if M is generated by (xi )i∈I .
This generalises the basic fact from linear algebra that giving a
linear map Rn → V is the same as giving the images of the standard
basis vectors.
Proof of Proposition 1.13. The map φ is given by
X
φ : R(I) → M, (ri )i∈I 7→ ri xi
i∈I
(note that in the sum only finitely many terms are non-zero). This map
is surjective if and only if every element of M can be written as a finite
R-linear combination of elements xi . □
Definition 1.14. Let M be an R-module and (xi )i∈I a family of el-
ements of M . Let φ : R(I) → M be the unique R-linear map with
φ(ei ) = xi for all i ∈ I. We say that (xi )i∈I is a basis of M if φ is an
isomorphism. We say that an R-module M is free if it has a basis. If
it has a basis of cardinality n, then we say that M is free of rank n.
In particular, M is free of rank n if and only if M ∼
= Rn .
In contrast with the case of vector spaces (modules over a field), a
finitely generated R-module need not have a basis. For example: for
m > 1 the Z-module Z/mZ does not have a basis, and hence is not
free.
Proposition 1.15. Let M be an R-module and (xi )i∈I a family of
elements of M . Then (xi )i∈I is a basis of M if and only if for every
x ∈ M there is a unique family (ri )i∈I of elements in R with
(1) ri = P
0 for all but finitely many i, and
(2) x = i∈I ri xi .
Note that the condition in (1) guarantees that only finitely many
terms in the sum in (2) are non-zero.
Proof of Proposition 1.15. This is a direct translation of the
definition: existence of (ri ) is equivalent with x being in the image of
φ : R(I) → M , and uniqueness is equivalent with φ : R(I) → M being
injective. □
Example 1.16. Let U be an open subset of Rn . Then the space of 1-

forms Ω1 (U ) on U forms a module over the ring C ∞ (U ) of C ∞ functions
on U . This module is free of rank n, with basis dx1 , . . . , dxn .
Proposition 1.17. Let R be a commutative ring with 0 ̸= 1. If Rn
and Rm are isomorphic R-modules, then n = m.
In other words: an R-module over a non-zero commutative ring
which is free of finite rank has a well-defined rank. We already know
this if R is a field (any two bases of a vector space have the same
cardinality), and the proof of the proposition will be by reduction to
the case of a field.
Proof of Proposition 1.17. Let M = Rm , N = Rn and let
φ: M → N
be an isomorphism. Let I ⊂ R be a maximal ideal (which exists in
every non-zero commutative ring). Then φ induces an isomorphism
φ̄ : M/IM → N/IN
of R/I-modules (see also Exercise 1.11). We have M/IM = (R/I)m
and N/IN = (R/I)n . But R/I is a field, so using the fact that a vector
space has a well-defined dimension we find
m = dimR/I M/IM = dimR/I N/IN = n,
which is what we had to prove. □
EXERCISES 17
Exercises
Exercise 1.1. Let M be an R-module. Show that for every r ∈ R and
x ∈ M the following identities in M hold:
(1) r0 = 0,
(2) 0x = 0,
(3) (−r)x = r(−x) = −(rx).
Exercise 1.2. Let R = {0} be the zero ring. Show that every R-module
is the zero module.
Exercise 1.3. Prove Lemma 1.2.
Exercise 1.4. Let R = (R, 0, 1, +, ·) be a ring. Consider the opposite
ring Rop = (R, 0, 1, +, ·op ) where multiplication is defined by
r ·op s := s · r.
Show that a right module over R is the same as an abelian group M
equipped with a ring homomorphism Rop → End(M ).
Exercise 1.5. Let M be an R-module, and let x1 , . . . , xn be elements
of M . Verify that the map
Rn → M, (r1 , . . . , rn ) 7→ r1 x1 + · · · + rn xn
is a homomorphism of R-modules.
Exercise 1.6. Let R be an integral domain and I ⊂ R a non-zero
principal ideal. Show that I, as an R-module, is isomorphic to the
R-module R.
Exercise 1.7. Let R be a ring, and let M and N be R-modules. Show
that point-wise addition makes HomR (M, N ) into an abelian group.
Show that if R is commutative, then HomR (M, N ) has a natural struc-
ture of R-module.
Exercise 1.8. Let R be a ring. Show HomR (R, M ) ∼ = M as abelian
groups (and if R is commutative, as R-modules).
Exercise 1.9. Verify that the map (1) in Example 1.9 is indeed a ring
homomorphism.
Exercise 1.10. Let K be a field and n a positive integer. Let V be a
K-vector space, and let α1 , α2 , . . . , αn be pairwise commuting linear
endomorphisms of V . Show that there is a unique ring homomorphism
ρ : K[X1 , . . . , Xn ] → End(V )
such that for all λ ∈ K and v ∈ V we have ρ(λ)(v) = λv and for all i
we have ρ(Xi )(v) = αi (v).
Convince yourself that a K[X1 , . . . , Xn ]-module is the same thing
as a vector space together with n pairwise commuting linear endomor-
phisms.
Exercise 1.11. Let R be a ring and I ⊂ R a (two-sided) ideal. Let M
be an R-module. Show that
X
IM := { ri xi | ri ∈ I, xi ∈ M }
is a sub-R-module of M , and show that M/IM is an R/I-module.
Show that if M is free of rank n as R-module, then M/IM is free of
rank n as R/I-module.
Exercise 1.12. Let R be a ring and let M be a left R-module. Show
that
AnnR (M ) := {r ∈ R | rx = 0 for all x ∈ M }
is a (two-sided) ideal in R.
Exercise 1.13. Show that Q is not a finitely generated Z-module.
Exercise 1.14. Let K be a field and I a countably infinite set. Show
that the K-module K I does not have a countable generating set.
Exercise 1.15. Show that Q is not a free Z-module.
Exercise 1.16. Prove the following generalisation of Proposition 1.13:
Let R be a ring and let (Mi )i∈I be a collection of R-modules indexed by
a set I. Let N be an R-module, and let (fi : Mi → N )i be a collection
of R-linear maps. Then there exists a unique R-linear map
M
f: Mi → N
i∈I
such that for all i ∈ I and x ∈ Mi we have f (ιi (x)) = fi (x).
Exercise 1.17. Let R be a ring, (Mi )i∈I a family of R-modules, and
N an R-module. Show that there are isomorphisms
∼
M Y
HomR ( Mi , N ) −→ HomR (Mi , N )
i∈I i∈I
and
∼
Y Y
HomR (N, Mi ) −→ HomR (N, Mi )
i∈I i∈I
of abelian groups.
EXERCISES 19
Exercise 1.18 (⋆). Let K be a field. Let R be the set of ∞ by ∞

matrices  
a11 a12 · · ·
 a21 a22 · · · 
.. ..
 
. .
with aij ∈ K, and with the property that every column contains only
finitely many non-zero elements. Verify that R (with the usual rule
for matrix multiplication) is a ring. Show that the left R-modules R
and R ⊕ R are isomorphic. Conclude that the condition that R is
commutative cannot be dropped from Proposition 1.17.
CHAPTER 2
Exact sequences
Exact sequences form a useful and extensively used notational tool

in algebra. They allow to replace tedious and verbose arguments in-
volving kernels and quotients by quick and intuitive ‘diagram chases’.
1. Exact sequences
If f : M1 → M2 and g : M2 → M3 are R-module homomorphisms,
then we say that the sequence
f g
M1 −→ M2 −→ M3
is exact if and only if the image of f is the kernel of g, as submodules
of M2 . For example: the sequence
0 −→ M −→ N
is exact if and only if M → N is injective, and the sequence
M −→ N −→ 0
is exact if and only if M → N is surjective.
A general sequence
fi−1 fi
· · · −→ Mi−1 −→ Mi −→ Mi+1 −→ · · ·
is called exact if for every i we have ker fi = im fi−1 as submodules of
Mi .
An exact sequence of the form
f g
0 −→ M1 −→ M2 −→ M3 −→ 0
is called a short exact sequence. Note that f induces an isomorphism
M1 ∼= ker g
and g induces an isomorphism
coker f ∼
= M3 .
21
22 2. EXACT SEQUENCES
We will often interpret the injective map f as the inclusion of a sub-

module M1 into M2 , and M3 as the quotient of M2 by the submodule
M1 , so that we can think of any short exact sequence as a sequence of
the type
0 −→ M1 −→ M2 −→ M2 /M1 −→ 0.
2. The Five Lemma and the Snake Lemma

The Five Lemma and Snake Lemma are powerful and often-used
lemmas about modules that are hard to state (and even harder to prove)
without the language of commutative diagrams and exact sequences.
The proofs are classic examples of ‘diagram chasing’. Such arguments
are often fairly easy to verify by tracing elements around the diagram
on the blackboard or a piece of paper, but are sometimes headache-
provokingly resistant to being rendered or read in prose.
Theorem 2.1 (Five Lemma). Let R be a ring. Consider a commutative
diagram of R-modules
M1 M2 M3 M4 M5
f1 f2 f3 f4 f5
N1 N2 N3 N4 N5
with exact rows. If f1 , f2 , f4 , f5 are isomorphisms, then so is f3 .
In fact, the proof will show that it suffices to assume that f1 is
surjective, f5 is injective, and f2 and f4 are isomorphisms.
Proof. The proof consists of two parts, one showing that f3 is
injective, the other that it is surjective. Both parts require only part of
the hypotheses in the theorem.
Claim. If f1 is surjective, and f2 and f4 are injective, then f3 is
injective.
Indeed, assume f3 (x) = 0 for some x ∈ M3 . We need to show that
x = 0. Let x′ ∈ M4 be the image of x. By the commutativity of the
diagram, f4 (x′ ) = 0. But f4 was injective, hence x′ = 0. It follows that
x ∈ M3 is the image of some element y ∈ M2 . By commutativity, f2 (y)
maps to zero in N3 , hence f2 (y) is the image of some element z ∈ N1 .
By the assumption on f1 , there is a z̃ ∈ M1 with f1 (z̃) = z.
Consider the image y ′ of z̃ in M2 . By commutativity, we have
f2 (y ′ ) = f2 (y), but since f2 is injective, this implies y ′ = y. We see that
2. THE FIVE LEMMA AND THE SNAKE LEMMA 23
x ∈ M3 is the image of some element z̃ in M1 , and hence by exactness

we conclude x = 0.
Claim. If f5 is injective, and f2 and f4 are surjective, then f3 is
surjective. The proof of this second claim is left to the reader, see
Exercise 2.3.
The theorem follows immediately from the above two claims. □
Theorem 2.2 (Snake Lemma). Let R be a ring. Let
α β
0 M1 M2 M3 0
f1 f2 f3
α′ β′
0 N1 N2 N3 0
be a commutative diagram in which both horizontal rows are short exact
sequences. Then there is a commutative diagram
0 ker f1 ker f2 ker f3
α β
0 M1 M2 M3 0
f1 f2 f3
α′ β′
0 N1 N2 N3 0
coker f1 coker f2 coker f3 0

of R-modules in which the maps ker fi → Mi and Ni → coker fi are the
natural inclusions and projections, and in which the sequence
0 → ker f1 → ker f2 → ker f3 → coker f1 → coker f2 → coker f3 → 0
is exact.
Sketch of proof. We only give the most interesting part of the
proof: the construction of the ‘snake’ map
d : ker f3 → coker f1 .
Let x ∈ ker f3 ⊂ M3 . Since the map β : M2 → M3 is surjective,
there is a y ∈ M2 with β(y) = x. By the commutativity of the right
square, we have
β ′ (f2 (y)) = f3 (β(y)) = f3 (x) = 0.
So f2 (y) ∈ ker β ′ = im α′ , hence there is a z ∈ N1 with α′ (z) = f2 (y).

Note that z is unique, as the map α′ is injective. We define d(x) as the
element z̄ ∈ coker f1 = N1 /f1 (M1 ).
We must check that this is well-defined, since our construction de-
pended on the choice of y ∈ M2 with β(y) = x. Let y ′ ∈ M2 be
another element with β(y ′ ) = x, leading to a z ′ ∈ N1 as above. Since
β(y ′ − y) = x − x = 0 there is a unique δ ∈ M1 with y ′ − y = α(δ).
Now the commutativity of the diagram shows z ′ − z = f1 (δ) in N1 , and
hence z̄ ′ = z̄ in N1 /f1 (M1 ), as we had to show. □
3. Split short exact sequences

Let M and N be R-modules. Then their direct sum fits into a short
exact sequence
i p
0 −→ M −→ M ⊕ N −→ N −→ 0
where the maps are the natural inclusion i : x 7→ (x, 0), and projection
p : (x, y) 7→ y. It is often convenient to be able to recognize if a given
short exact sequence is of the above special form.
Theorem 2.3 (Splitting lemma). Let
f g
0 −→ M1 −→ M2 −→ M3 −→ 0
be an exact sequence of R-modules. Then the following are equivalent:
(1) there is a homomorphism h : M2 → M1 such that hf = idM1
(2) there is a homomorphism s : M3 → M2 such that gs = idM3
∼
(3) there is an isomorphism φ : M2 → M1 ⊕ M3 such that the
diagram
f g
0 M1 M2 M3 0
id φ id
i p
0 M1 M1 ⊕ M3 M3 0
commutes (where i(x) = (x, 0) and p(x, y) = y.)
If these conditions hold, we say that the sequence is split or split
exact. The map h is called a retraction of f , and the map s a section
of g. See Exercise 2.12 for examples of short exact sequences that are
not split.
3. SPLIT SHORT EXACT SEQUENCES 25
Proof of Theorem 2.3. We first show that (3) implies (2). In-
deed, the map
M3 → M2 , y 7→ φ−1 (0, y)
is a section of g.
Next, we show that (2) implies (1), so assume that s : M3 → M2 is
a section. We will construct a retraction h : M2 → M1 . Let x ∈ M2 .
Consider the element
y := x − s(g(x)) ∈ M2 .
Then, since gs = id we have
g(y) = g(x) − g(s(g(x))) = g(x) − g(x) = 0.
So y ∈ ker g = im f , and since f is injective, there is a unique z ∈ M1
with f (z) = y. Define h(x) := z. One checks that h is indeed a
retraction.
Finally, to show that (1) implies (3), assume that h : M2 → M1 is
a retraction. Then consider the map
φ : M2 → M1 ⊕ M3 , x 7→ (h(x), g(x)).
Note that this map is an R-module homomorphism. We verify that
it makes the diagram commute. We start with the left square. Take
an x ∈ M1 in the left-top corner of this square. Going down and
then right, it gets mapped to i(id(x)) = (x, 0) ∈ M1 ⊕ M3 . Following
the other path, we end up with φ(f (x)) = (h(f (x)), g(f (x))). But
now, h(f (x)) = x because h is a retraction, and g(f (x)) = 0 because
im f = ker g by the hypothesis that the sequence is exact. We conclude
that φ(f (x)) = (x, 0) and that the left square indeed commutes. For the
right-hand square, take x ∈ M2 . Then one path yields id(g(x)) = g(x),
and following the other path, we obtain p(φ(x)) = p(h(x), g(x)) =
g(x). These agree, so we conclude that the diagram indeed commutes.
Finally, by Exercise 2.5 we see that the map φ is automatically an
isomorphism, which shows that (3) indeed follows from (1). □
Exercises
Exercise 2.1. Show that
0 −→ M −→ 0
is exact if and only if M is the zero module.
Exercise 2.2. Let f : M → N be an R-module homomorphism. Show
that there is an exact sequence
f
0 −→ ker f −→ M −→ N −→ coker f −→ 0
of R-modules.
Exercise 2.3. Complete the proof of the Five Lemma (Theorem 2.1):
show that if f2 and f4 are surjective, and if f5 is injective, then f3 is
surjective.
Exercise 2.4. Let
M1 M2 M3 0
f1 f2
N1 N2 N3 0
be a commutative diagram of R-modules with exact rows. Show that
there exists a unique R-linear map f3 : M3 → N3 making the resulting
diagram commute.
Exercise 2.5. Consider a commutative diagram of R-modules
α
0 M E N 0
idM f idN
0 M E′ N 0
in which both rows are short exact sequences. Deduce from the snake
or five lemma that f must be an isomorphism. Show that f is an iso-
morphism without using the snake or five lemma. (Hint for surjectivity:
given y ∈ E ′ choose an x ∈ E with same image as y in N . Show that
there is an z ∈ M with f (α(z) + x) = y.)
Exercise 2.6. Give an example of a diagram as in Theorem 2.2, for
which the ‘snake map’ d : ker f3 → coker f1 is non-zero.
EXERCISES 27
Exercise 2.7. Let R be a ring and let

M1 M2
α1 α2
N1 N2
be a commutative diagram of R-modules, in which the two horizontal
maps are injective. Show that there exists an R-module E and an exact
sequence
0 −→ ker α1 −→ ker α2 −→ E −→ coker α1 −→ coker α2
of R-modules.
Exercise 2.8. Let R be a ring, let
(2) 0 −→ M1 −→ M2 −→ M3
be an exact sequence of R-modules, and let N be an R-module. Show
that there is an exact sequence of abelian groups
0 −→ HomR (N, M1 ) −→ HomR (N, M2 ) −→ HomR (N, M3 ).
Give an example to show that the exactness of 0 → M1 → M2 →
M3 → 0 need not imply that the map HomR (N, M2 ) → HomR (N, M3 )
is surjective.
Exercise 2.9. Let R be a ring, let
M1 −→ M2 −→ M3 −→ 0
be an exact sequence of R-modules, and let N be an R-module. Show
that there is an exact sequence of abelian groups
0 −→ HomR (M3 , N ) −→ HomR (M2 , N ) −→ HomR (M1 , N ).
Give an example to show that the exactness of 0 → M1 → M2 →
M3 → 0 need not imply that the map HomR (M2 , N ) → HomR (M1 , N )
is surjective.
Exercise 2.10. Let I and J be left ideals in a ring R. Show that there
are exact sequences
0→I ∩J →I ⊕J →I +J →0
and
0 → R/(I ∩ J) → R/I ⊕ R/J → R/(I + J) → 0
of R-modules.
Exercise 2.11. Let R be a commutative ring and let

0 −→ M −→ E −→ N −→ 0
be a short exact sequence of R-modules. Let I := AnnR M and J :=
AnnR N (see Exercise 1.12). Show that
IJ ⊂ AnnR E ⊂ I ∩ J
as ideals in R.
Exercise 2.12. Show that the short exact sequences of Z-modules
2
0 −→ Z −→ Z −→ Z/2Z −→ 0
and
2
0 −→ Z/2Z −→ Z/4Z −→ Z/2Z −→ 0
are not split. (Here the ‘2’ above the arrows are shorthand for the maps
x 7→ 2x and x̄ 7→ 2x.)
Exercise 2.13. Let K be a field. Show that every short exact sequence
of K-modules is split exact.
Exercise 2.14. Let G be a finite group, and K a field of characteristic
zero. Maschke’s theorem asserts that for every representation V of G
over K, and for every G-stable subspace W ⊂ V there exists a G-
stable complement U ⊂ V . Show that every short exact sequence of
K[G]-modules is split.
Exercise 2.15 (⋆). Let G be the cyclic group of 2 elements. Consider
the group ring R := F2 [G]. Give an example of a non-split short ex-
act sequence of R-modules. Show that not every representation of the
group G over the field F2 is isomorphic to a direct sum of irreducible
representations.
Exercise 2.16. Let K be a field. Consider the subring

a b
R := | a, b, c ∈ K
0 c
of the ring Mat(2, K) of two-by-two matrices over K. Let M be the
module of column vectors xy on which R acts by the usual matrix

multiplication:
a b x ax + by
=
0 c y cy
EXERCISES 29
Show that N := { x0 | x ∈ K} ⊂ M is a sub-R-module, and that the

short exact sequence of R-modules
0 −→ N −→ M −→ M/N −→ 0
does not split.
Exercise 2.17. Let R be a ring and let M and N be R-modules. Show
that any short exact sequence of the form
0 −→ M −→ N −→ Rn −→ 0
is split.
Exercise 2.18. Let R be a ring and let I ⊂ R be a left ideal. Show
that a short exact sequence of left R-modules of the form
π
0 −→ M −→ N −→ R/I −→ 0
splits if and only if there exists an x ∈ N with π(x) = 1 + I and rx = 0
for all r ∈ I.
Exercise 2.19. Let
0 −→ M1 −→ M2 −→ M3 −→ 0
be a split short exact sequence of R-modules, and let N be an R-module.
Show that the induced sequences
0 −→ HomR (N, M1 ) −→ HomR (N, M2 ) −→ HomR (N, M3 ) −→ 0
and
0 −→ HomR (M3 , N ) −→ HomR (M2 , N ) −→ HomR (M1 , N ) −→ 0
are exact.
0 −→ M1 −→ M2 −→ M3 −→ 0
be a short exact sequence of R-modules. Show that if M1 is free of rank
n1 and M3 is free of rank n3 , then M2 is free of rank n1 + n3 .
CHAPTER 3
Finitely generated modules over a PID
1. Introduction
The classification of finite abelian groups states that for every finite
abelian group A there are prime numbers pi (not necessarily distinct)
and exponents ei ≥ 1 such that
A∼
= (Z/pe11 Z) × · · · × (Z/penn Z).
The existence of Jordan normal forms states that for every square
matrix P over C there exist complex numbers λi (not necessarily dis-
tinct) and integers ei ≥ 1 so that P is conjugate to a block diagonal
matrix with blocks  
λi 0 · · · 0 0
 1 λi · · · 0 0 
 
 .. .. .. .. 
. . . .
 
 0 0 · · · λi 0 
0 0 · · · 1 λi
of size ei .
In this chapter, we will see that these two theorems are just two
instances of one and the same theorem about finitely generated modules
over a principal ideal domain. In the first case, the PID will be Z, in
the second case it will be C[X].
2. Review of principal ideal domains

Let R be a PID (principal ideal domain). Recall that this means
that R is an integral domain (a nonzero commutative ring without
zero divisors), and that for every ideal I ⊂ R there is an r ∈ R with
I = (r) = Rr. For our purposes, the most important examples are
R = Z and R = K[X] with K a field. Other important examples are
the ring of power series K[[X]], the ring of p-adic integers Zp , and the
31
32 3. FINITELY GENERATED MODULES OVER A PID
ring of Gaussian integers Z[i]. Also a field K is a PID, but of a rather

trivial kind.
PID’s are unique factorization domains. This means that every
non-zero element r in a principal ideal domain R can be written as
r = upe11 · · · penn
with u ∈ R× , with the pi ∈ R irreducible, and with ei non-negative
integers. Moreover, such factorization is unique up to multiplying u
and the pi ’s by units, and up to permuting the factors.
Using this prime factorization, we can define greatest common divi-
sors gcd(r, s) of two non-zero elements of R. They are uniquely deter-
mined up to units (for example both 2 and −2 are a gcd of 4 and 6 in
Z). Similarly, we can define gcds of any sequence r1 , . . . , rn of elements
of R which is not identically zero (ignoring the zeroes in the sequence).
An element d ∈ R is a gcd of r1 , . . . , rn if and only if d generates
the ideal (r1 , . . . , rn ) of R. In particular, there are a1 , . . . , an in R with
d = a1 r1 + · · · + an rn . For example, if r and s are coprime (have no
common prime factor), then there are a and b in R with ar + bs = 1.
3. Free modules of finite rank over a PID

Proposition 3.1. Let R be a PID and M ⊂ Rn a submodule. Then
M∼= Rk for some k ≤ n.
Hence over a PID a submodule of a free module of finite rank is
itself free of finite rank. The condition that R be a PID cannot be
dropped from the proposition, see Exercise 3.3.
Proof of Proposition 3.1. We use induction on n. For n = 0

we have M = Rn = 0, and M is indeed free of rank 0. Assume that the
proposition has been shown to hold for submodules of Rn−1 . Consider
the projection
π : Rn → R, (r1 , . . . , rn ) → rn
with kernel Rn−1 × {0}. Then we have a short exact sequence
0 −→ M ∩ ker π −→ M −→ π(M ) −→ 0.
By the induction hypothesis, we have that the submodule M ∩ ker π of
Rn−1 × {0} is free of rank k ≤ n − 1. If π(M ) = 0, then we are done.
If π(M ) ̸= 0, then π(M ) is an ideal in R, hence a principal ideal, hence
4. STRUCTURE OF FINITELY GENERATED MODULES OVER A PID 33
π(M ) ∼
= R as R-module. By Exercise 2.17 any short exact sequence of
R-modules of the form
0 −→ N −→ M −→ R −→ 0
splits, and we find M ∼
= R ⊕ (M ∩ ker π) ∼
= Rk+1 with k + 1 ≤ n. □
Corollary 3.2. Let R be a PID and M a finitely generated R-module.
Then there exists an exact sequence
0 −→ F1 −→ F2 −→ M −→ 0
with F1 and F2 free R-modules of finite rank.
Proof. Since M is finitely generated, by Proposition 1.13 there is
a surjection F2 → M with F2 a free module of finite rank. The kernel
F1 ⊂ F2 is also free of finite rank, thanks to Proposition 3.1. □
4. Structure of finitely generated modules over a PID

The main theorem of this chapter is the following structure theorem.
Theorem 3.3. Let R be a principal ideal domain and M a finitely
generated R-module. Then there exists an integer n and non-zero ideals
I1 , . . . , Ik of R such that
M∼ = Rn ⊕ R/I1 ⊕ · · · ⊕ R/Ik
as R-modules.
For the proof we need the notion of content of an element of a
module. Let R be a commutative ring and let M be an R-module. An
element x ∈ M determines a map
(3) HomR (M, R) → R, f 7→ f (x).
This map is R-linear, hence the image is an ideal. We denote it by
cM (x), and call it the content of x.
Lemma 3.4. If M is free of finite rank, and x ∈ M is non-zero, then
cM (x) is a non-zero ideal in R.
Proof. Without loss of generality, we may assume M = Rn and
̸ 0.
x = (x1 , . . . , xn ). Since x is non-zero, there exists an i with xi =
Consider the map
pri : Rn → R, (y1 , . . . , yn ) 7→ yi .
We have pri ∈ HomR (M, R) and pri (x) ̸= 0, hence cM (x) ̸= 0. □
Note that the coordinates xi of x ∈ M ∼ = Rn depend on the choice

of basis of M , but that the content ideal cM (x) is independent of such
choice!
Lemma 3.5. Let R be a principal ideal domain, M a free R-module
of finite rank, and x ∈ M a non-zero element. Then there is a surjec-
tive R-linear map f : M → R such that f (x) ∈ R generates the ideal
cM (x) ⊂ R.
Proof. Since R is a principal ideal domain, there exists an f ∈
HomR (M, R) such that f (x) generates the ideal cM (x). Moreover, by
Lemma 3.4 we have f (x) ̸= 0.
Consider the image of f : M → R. This is an ideal I ⊂ R, and it
suffices to show that I = R. As R is a PID, the ideal I is generated by
some element r ∈ R. For every y ∈ M we have f (y) ∈ I and (since R
is an integral domain) there exists a unique g(y) ∈ R such that
f (y) = g(y) · r.
This defines an R-linear map g : M → R. Now consider the element
g(x) ∈ R. By definition of cM (x) we have g(x) ∈ cM (x). Since f (x)
generates cM (x) there is an s ∈ R with g(x) = sf (x). But we also have
f (x) = rg(x), hence rs = 1, hence I = R and f is surjective. □
Proof of Theorem 3.3. By Corollary 3.2 every finitely gener-

ated R-module M sits in a short exact sequence
0 −→ F1 −→ F2 −→ M −→ 0
with the Fi free R-modules of finite rank. We will interpret F1 as a
submodule of F2 .
The proof goes by induction on the rank of F1 . If F1 is zero, then
M∼ = F2 ∼
= Rn for some n, and the theorem holds.
Otherwise, let x ∈ F1 be a non-zero element whose content I :=
cF2 (x) with respect to F2 is maximal (amongst all the ideals in R of the
form cF2 (x) with x ∈ F1 ).
By Lemma 3.5 there is a surjective map f : F2 → R such that f (x)
generates I.
Claim. f (F1 ) = I ⊂ R. Indeed, since f (x) generates I we certainly
have f (F1 ) ⊃ I. Conversely, for z ∈ F1 , let d be a gcd of f (z) and
f (x). Then there exists r, s ∈ R with rf (z) + sf (x) = d, and hence
f (rz + sx) = d. By the maximality of the content of x we have that d
5. APPLICATION TO JORDAN NORMAL FORM 35
must equal f (x) up to a unit, hence f (z) must be divisible by f (x) and
hence f (z) ∈ I.
Denote the kernel of f : F2 → R by F2′ (note that by Proposition
3.1, this is also a free module). We have a short exact sequence
f
0 −→ F2′ −→ F2 −→ R −→ 0.
Similarly, let F1′ be the kernel of the restriction f : F1 → R. We find a
short exact ‘sub-sequence’
f
0 −→ F1′ −→ F1 −→ I −→ 0.
Since f (x) generates I = cF2 (x) there is a y ∈ F2 with f (x)y = x. Now
the first sequence splits by the section R → F2 , 1 7→ y, and this section
restricts to a section I → F1 in the second short exact sequence.
We find M ∼ = M ′ ⊕ R/I with M ′ given by the short exact sequence
0 −→ F1′ −→ F2′ −→ M ′ −→ 0.
Since F1′ has lower rank, the induction hypothesis guarantees
′
M′ ∼= Rn ⊕ R/I1 ⊕ · · · ⊕ R/Ik ,
which finishes the proof. □
5. Application to Jordan normal form

Theorem 3.3 has the following corollary.
Corollary 3.6. Let R be a PID and M a finitely generated R-module.
Then there exists an integer n, irreducible elements p1 , . . . , pk of R and
positive integers e1 , . . . , ek such that
M∼ e
= Rn ⊕ R/pe1 R ⊕ · · · ⊕ R/p k R
1 k
as R-modules.
Proof. By Theorem 3.3 it suffices to show that for every non-zero
ideal I the R-module R/I can be written in the desired form. Let x be
a generator of I, and consider its prime factorization
x = upe11 · · · pekk
with pi pairwise non-associated primes. Then by the Chinese Remain-
der Theorem, we have
R/I ∼ e
= R/pe1 R ⊕ · · · ⊕ R/p k R,
1 k
as we had to show. □
We now consider two special cases of this corollary. The first one is
a structure theorem for finitely generated and finite abelian groups.
Theorem 3.7 (Classification of finitely generated abelian groups). Let
A be a finitely generated abelian group. Then there exists an integer n,
prime numbers p1 , . . . , pk , and positive integers e1 , . . . , ek such that
A∼ e
= Zn × Z/pe1 Z × · · · × Z/p k Z.
1 k
If A is a finite abelian group then
A∼ e
= Z/pe1 Z × · · · × Z/p k Z.
1 k
Proof. Apply Corollary 3.6 to the case R = Z. □
The second special case is a structure theorem for endomorphisms of
finite-dimensional vector spaces over C (or over an algebraically closed
field).
Theorem 3.8. Let K be an algebraically closed field. Let V be a finite-
dimensional vector space over K. Let α : V → V be an endomorphism.
Then there exist λ1 , . . . , λk in K, positive integers e1 , . . . , ek , and a
decomposition
V = V1 ⊕ · · · ⊕ Vk
such that α(Vi ) ⊂ Vi , and such that each Vi has a basis on which α is
expressed as the standard Jordan matrix
 
λi 0 · · · 0 0
 1 λi · · · 0 0 
 
 .. .. .. .. 
. . . .
 
 0 0 · · · λi 0 
0 0 · · · 1 λi
of size ei .
Proof. As in Example 1.9, we turn the vector space V into a
K[X]-module, with X acting via the endomorphism α. Since V is finite-
dimensional, it is finitely generated as a K-module, hence a fortiori also
as a K[X]-module.
Since K is algebraically closed, the irreducible elements of K[X] are
(up to units) the linear polynomials X − λ with λ ∈ K. By Corollary
3.6, we have
K[X] K[X]
V ∼
= K[X]n ⊕ ⊕ ··· ⊕ .
(X − λ1 )e1 K[X] (X − λk )ek K[X]
5. APPLICATION TO JORDAN NORMAL FORM 37
Note that K[X] is infinite-dimensional as a K-vector space, so we nec-

essarily must have n = 0.
Without loss of generality, we may assume
V = K[X]/(X − λ)e K[X]
for some λ ∈ K and e > 0. Consider the elements
vj := (X̄ − λ)j ∈ V j ∈ {0, . . . , e − 1}.
Note that the vj form a K-basis of V . The action of α on this basis is
given by
α(vj ) = X · (X̄ − λ)j = (X̄ − λ)j+1 + λ(X̄ − λ)j
hence (
vj+1 + λvj j < e − 1
α(vj ) =
λvj j = e − 1.
and we see that the matrix of α with respect to this basis is the standard
Jordan block of eigenvalue λ and size e. □
Exercises
Exercise 3.1. Let R be a commutative ring and m ⊂ R a maximal
ideal. Let M be an R-module. Show that M/mM is a vector space
over R/m. Show that if M is generated by x1 , . . . , xn then M/mM has
dimension at most n over R/m.
Exercise 3.2. Let R be a commutative ring and M an R-module. An
element x ∈ M is called a torsion element if there exists a non-zero
r ∈ R with rx = 0.
Assume that R is an integral domain. Show that the torsion ele-
ments of an R-module M form a submodule of M .
Give an example to show that the condition that R is an integral
domain cannot be dropped.
Exercise 3.3. Let R be an integral domain. Show that the following
are equivalent:
(1) every submodule of a free R-module of finite rank is free of
finite rank,
(2) R is a principal ideal domain.
Exercise 3.4. Let R be a PID and let M be an R-module which can
be generated by n elements. Show that every submodule N ⊂ M can
be generated by n elements. Show that the condition that R is a PID
cannot be dropped.
Exercise 3.5. Let R be a PID and let M be a finitely generated R-
module such that for all r ∈ R and x ∈ M we have that rx = 0 implies
r = 0 or x = 0. Show that M is free of finite rank.
0 −→ M −→E−→N −→ 0
be a short exact sequence of R-modules. Assume that M can be gener-
ated by m elements, and that N can be generated by n elements. Show
that E can be generated by m + n elements.
Exercise 3.7. Let R be a PID, and let p1 and p2 be irreducible elements
with (p1 ) ̸= (p2 ). Let e1 , e2 be non-negative integers. Show that the
only R-module homomorphism
R/pe11 R → R/pe22 R
is the zero homomorphism.
EXERCISES 39
Exercise 3.8. Let R be a PID and let p ∈ R be irreducible. Let

e1 , e2 be non-negative integers. Show that there is an isomorphism of
R-modules
HomR (R/pe1 R, R/pe2 R) ∼
= R/pe R
with e = min(e1 , e2 ).
Exercise 3.9. Describe all Z[i]-modules with at most 10 elements, up
to isomorphism.
Exercise 3.10. Let R be a PID and let p ∈ R be an irreducible element.
Let E be an R-module contained in a short exact sequence
f g
0 −→ R/pR −→ E −→ R/pR −→ 0.
Show that either E ∼= R/pR ⊕ R/pR or E ∼
= R/p2 R, that both options
occur, and that R/pR ⊕ R/pR ≠∼ R/p R.
2
Exercise 3.11 (⋆). Let

0 −→ A1 −→ A2 −→ A3 −→ · · · −→ An −→ 0
be an exact sequence of finite abelian groups. Show that the equality
i
Y
|Ai |(−1) = 1
i
holds.
Exercise 3.12. Let K be a field and let f1 , . . . , fn ∈ K[X] be monic
polynomials. Consider the K[X]-module
V := K[X]/(f1 ) ⊕ · · · ⊕ K[X]/(fn ).
Show that the characteristic polynomial
Q of the endomorphism v 7→ X ·v
of the K-vector space V equals i fi .
Exercise 3.13. Let K be a field and let V be a finite-dimensional
vector space over K. Let α be an endomorphism of V . Assume that
the characteristic polynomial of α is irreducible. Show that there is no
proper non-zero subspace W ⊂ V with α(W ) ⊂ W .
Exercise 3.14. Let K be a field and let V be a finite-dimensional
vector space over K. Let α be an endomorphism of V . Assume that
the characteristic polynomial of α is separable. Show that there are
subspaces W1 , . . . , Wn of V , such that the following hold:
(1) V = W1 ⊕ W2 ⊕ · · · ⊕ Wn ,
(2) α(Wi ) ⊂ Wi for every i,

(3) the characteristic polynomial of α|Wi is irreducible for every i.
Give an example to show that the condition that the characteristic
polynomial is separable cannot be dropped.
Exercise 3.15 (⋆⋆). Let R be a PID. An R-module M is called torsion
if for every x ∈ M there is a non-zero r ∈ R with rx = 0. For a finitely
generated torsion R-module
M∼ = R/I1 ⊕ · · · ⊕ R/Ik
denote by F (M ) the ideal I1 I2 · · · Ik . Show that F (M ) is independent
of the chosen decomposition. Show that if
0 −→ M1 −→ M2 −→ M3 −→ 0
is a short exact sequence of R-modules, and if M1 and M3 are finitely
generated torsion R-modules, then M2 is a finitely generated torsion
R-module and F (M2 ) = F (M1 )F (M3 ).
CHAPTER 4
Categories
1. Definition
Definition 4.1. A category C consists of the data of
(1) a class of objects ob C,
(2) for every X, Y ∈ ob C a class HomC (X, Y )
(3) for every X ∈ ob C an element idX ∈ HomC (X, X)
(4) for every X, Y, Z ∈ ob C a map
HomC (Y, Z) × HomC (X, Y ) → HomC (X, Z), (g, f ) 7→ gf,
called composition, subject to the conditions

(C1) for every X, Y, Z, T ∈ ob C and for every f ∈ HomC (X, Y ), g ∈
HomC (Y, Z) and h ∈ HomC (Z, T ) the identity h(gf ) = (hg)f
holds
(C2) for every X, Y ∈ C and for every f ∈ HomC (X, Y ) the identi-
ties f idX = f and idY f = f hold.
We often write f : X → Y instead of f ∈ HomC (X, Y ), and think

of f as being a ‘map from X to Y ’. However, the objects X and Y
need not be sets, and expressions such as ‘x ∈ X’ or ‘f (x) ∈ Y ’ can be
completely meaningless.
Remark 4.2. The use of the word ‘class’ is to avoid set-theoretical

problems. We want to consider examples such as the category of all
sets, but must be careful to avoid paradoxes. The class of all sets does
not form a set itself, otherwise one could consider the subset of all the
sets that are not contained in itself, leading to Russell’s paradox.
This subtlety in the definition of a category is mostly harmless, and
in almost all applications of categories one can safely pretend the class
of objects form a set. In fact, one can often restrict the objects to a
suitable chosen sub-set of the given class without losing much.
41
42 4. CATEGORIES
2. Big examples
The notion of a category is modeled on the properties of the col-
lection of all objects of a certain kind (sets, rings, spaces) together
with the collection of all structure-preserving maps (functions, ring ho-
momorphisms, continuous maps) between them. The most important
examples are of this kind.
Example 4.3 (The category of sets). The category Set with ob Set
the class of all sets, HomSet (X, Y ) the set of all maps from X to Y ,
idX the identity map and the usual composition gf := g ◦ f forms a
category.
Example 4.4 (The category of topological spaces). The category Top
with ob Top the class of all topological spaces, HomTop (X, Y ) the set of
continuous maps from X to Y and the usual identity and composition
form a category. (Note that the composition of two continuous maps is
continuous!).
Example 4.5 (The categories of left and right R-modules). If R is a
ring, we denote by R Mod the category whose objects are the left R-
modules, and whose morphisms are the R-module homomorphisms. So
for M, N ∈ ob R Mod we have
HomR Mod (M, N ) := HomR (M, N ).
Similarly, we denote by ModR the category of right R-modules.
In the same style as the above examples, we have the category Ring
of rings and ring homomorphisms, the category CRing of commuta-
tive rings and ring homomorphisms, the category Grp of groups and
group homomorphisms, the category Ab of abelian groups and group
homomorphisms, etcetera. Note that in all these examples the objects
of the categories are sets equipped with some extra structure, and the
morphisms are functions that are compatible with the structure. This
is the case for many, but certainly not all, commonly used categories.
3. Small examples
A category is also a mathematical object in its own right, and one
can write down explicit examples by specifying the objects and maps,
in the same way one can specify say a ring by giving its elements, the
addition, and multiplication.
3. SMALL EXAMPLES 43
Example 4.6 (One arrow). Consider the category C consisting of pre-

cisely two objects, X and Y , and with precisely three maps: idX , idY ,
and a map f : X → Y . We can render this category in a picture:
f
idX idY
X Y
Example 4.7 (A group as a category). Let G be a group. Con-

sider the category BG with one object ⋆ (so ob BG = {⋆}), and with
HomBG (⋆, ⋆) := G, id⋆ := 1 ∈ G, and where composition is defined by
multiplication in G. Note that axiom (C1) follows from the associa-
tivity of the group operation, and axiom (C2) from the axiom for the
neutral element 1 ∈ G.
Example 4.8 (Discrete category). If S is a set, then S defines a cate-
gory C with ob C := S and with
(
{id} if x = y
HomC (x, y) :=
∅ ̸ y
if x =
for all x, y ∈ S.
Example 4.9 (Pre-ordered set). A pre-ordered set is a set S equipped
with a relation ≤ that is reflexive (x ≤ x) and transitive (x ≤ y and
y ≤ z implies x ≤ z). A pre-ordered set defines a category C with
ob C := S and with
(
{⋆} if x ≤ y
HomC (x, y) :=
∅ if x ̸≤ y
where {⋆} denotes any singleton.
Example 4.10 (The category of matrices). Let R be a ring. Then we
can form a category C with ob C := {0, 1, 2, . . .}, with
HomC (n, m) := Matm,n (R) := {m × n matrices over R},
and with composition being matrix multiplication
Matm,ℓ × Matℓ,k → Matm,k , (B, A) 7→ BA.
The identity element idn : n → n is the n × n identity matrix. Axiom
(C1) corresponds to the associativity of matrix multiplication.
44 4. CATEGORIES
Definition 4.11. A category is called locally small if for every X and

Y in ob C the class HomC (X, Y ) is a set. It is called small if moreover
the class of objects ob C is itself a set.
Examples 4.12. All the examples in sections 2 and 3 are locally small.
The examples in 3 are in fact small. However, the examples in 2 are
not small, essentially because the class of all sets is not a set.
Finally, we discuss two ways of constructing new categories out
of old ones. The first is the opposite category, which is obtained by
formally reversing all the arrows in a given category:
Definition 4.13. The opposite or dual of a category C is the category
C op with ob C op := ob C and with
HomC op (X, Y ) := HomC (Y, X).
Composition is done ‘the other way around’:
HomC op (Y, Z) × HomC op (X, Y ) → HomC op (X, Z), (g, f ) 7→ f g
where f : Y → X and g : Z → Y and f g : Z → X are maps in C.
Definition 4.14. If C and D are categories, then the product category
C × D is the category whose objects are pairs (X, Y ) with X ∈ ob C and
Y ∈ ob D, and whose morphisms are pairs of morphisms:
HomC×D ((X, Y ), (X ′ , Y ′ )) := HomC (X, X ′ ) × HomD (Y, Y ′ ).
In general it makes no sense to ask if a morphism f : X → Y in
a category C is bijective, injective, or surjective. The objects X and
Y are just elements of some class ob C, and the morphism f is just an
element of some class HomC (X, Y ). It makes no sense to talk about
elements of X or Y . However, one can define what it means for f to
be an isomorphism.
Definition 4.15. Let f : X → Y be a morphism in a category C. We
say that f is an isomorphism if there exists a morphism g : Y → X
such that f g = idY and gf = idX .
Example 4.16. An isomorphism in Set is a bijection. An isomorphism
in Grp is a group isomorphism. An isomorphism in Top is a homeo-
morphism (note that this is stronger than being a bijective continuous
map!). An isomorphism in the category of matrices (Example 4.10) is
an invertible square matrix.
4. FINAL AND COFINAL OBJECTS 45
Definition 4.17. We say that objects X and Y in a category are

isomorphic if there exists an isomorphism f : X → Y .
4. Final and cofinal objects

Some categories have special objects, called final and cofinal (also
known as initial) objects.
Definition 4.18. Let C be a category. An object X ∈ ob C is called
final in C if for every Y ∈ ob C there is a unique morphism Y → X.
Definition 4.19. Let C be a category. An object X ∈ ob C is called
initial or cofinal in C if for every Y ∈ ob C there is a unique morphism
X →Y.
Examples 4.20. In Top the empty space ∅ is a cofinal object, and a
one-point space {⋆} is final. In R Mod the zero module 0 is both cofinal
and final. In Ring the ring Z is cofinal and the zero ring 0 is final. In
Grp the trivial group {1} is both cofinal and final.
Proposition 4.21. If it exists, a final object in a category C is unique
up to unique isomorphism.
Proof. Assume X1 and X2 are final. We need to show that there
exists a unique isomorphism f : X1 → X2 .
Since X2 is final, there exists a map f : X1 → X2 . Since X1 is final,
there exists a map g : X2 → X1 . Consider the composite f g : X2 → X2 .
Since X2 is final, there is a unique map X2 → X2 , so f g = idX2 .
Similarly gf = idX1 . So we see that f is an isomorphism X1 → X2 .
Now assume that there are two isomorphisms f1 , f2 : X1 → X2 .
Then again, because X2 is final, we must have f1 = f2 , which shows
that f is unique. □
Like so many statements about categories, this proposition has a
co-proposition:
Proposition 4.22. If it exists, a cofinal object in a category C is unique
up to unique isomorphism.
Proof. Reverse the arrows in the proof of Proposition 4.21. Al-
ternatively, apply Proposition 4.21 to the opposite category C op . □
46 4. CATEGORIES
Exercises
Exercise 4.1 (Automorphism group of an object). Let C be a (locally
small) category and X an object in C. Show that
AutC (X) := {f : X → X | f is an isomorphism}
forms a group under composition.
Exercise 4.2. Let C be a (locally small) category and X and Y objects
in C. Assume that X and Y are isomorphic. Show that AutC (X) and
AutC (Y ) are isomorphic groups.
Exercise 4.3. Let C be a category. Show that the relation ‘X and Y
are isomorphic’ forms an equivalence relation on ob C.
Exercise 4.4. Let f : X → Y be morphisms in a category C. Show
that if f has both a left and a right inverse, then these must agree. In
particular, f has at most one two-sided inverse.
Exercise 4.5. Let C be a category. Define ob C × := ob C and
HomC × (X, Y ) := {f ∈ HomC (X, Y ) | f is an isomorphism}.
Show that C × (with composition and identity maps inherited from C)
is a category.
Exercise 4.6. Verify that Examples 4.6 and 4.8 are special cases of
Example 4.9.
Exercise 4.7. Show that the category of fields has neither final nor
cofinal object. Show that the category of fields of a given fixed charac-
teristic does have a cofinal object.
Exercise 4.8 (The category of G-sets). Let G be a group. A G-set is
a set S together with a left action of G on S, that is, a map G × S →
S, (g, s) 7→ gs satisfying 1s = s and g(hs) = (gh)s for all g, h ∈ G
and s ∈ S. A morphism of G-sets is a map f : S → T satisfying
f (gs) = gf (s) for every s ∈ S and g ∈ G. The category of G-sets is
denoted G Set.
What are the final and cofinal objects of G Set?
Exercise 4.9 (Homotopy category). Let hTop be the homotopy cat-
egory of topological spaces. Its objects are topological spaces, and its
morphisms are homotopy classes of continuous maps:
HomhTop (X, Y ) := {f : X → Y | f continuous}/ ∼
EXERCISES 47
(Note that composition is compatible with homotopy, so that composi-

tion in hTop is well-defined). Show that two topological spaces X and
Y are homotopy-equivalent if and only if X and Y are isomorphic in
the category hTop.
Exercise 4.10. Let f : X → Y be a morphism in C. Show that f is an
isomorphism if and only if for all objects T in C the map
HomC (T, X) → HomC (T, Y ), h 7→ f h
is a bijection. (Hint, for the hard direction: use T = Y to find a
g : Y → X with f g = idY , and use T = X to deduce that gf = idX .)
Exercise 4.11. Formulate and prove the co-Exercise of Exercise 4.10.
CHAPTER 5
Functors
If in order to study groups, modules or topological spaces, one

should study the homomorphisms or continuous maps between them,
then to study categories one should study ‘morphisms of categories’.
Such morphisms of categories are called functors.
1. Definition of a functor
Definition 5.1. Let C and D be categories. A functor F from C to D,
consists of the data of
(1) for every object X in C an object F (X) in D,
(2) for every morphism f : X → Y in C a morphism F (f ) : F (X) →
F (Y ) in D
subject to the conditions
(F1) for every X in C we have F (idX ) = idF (X)
(F2) for every f : X → Y and g : Y → Z in C we have F (gf ) =
F (g)F (f ).
We will write F : C → D to denote that F is a functor from C to D.
Note that if F : C → D and G : D → E are functors, then the com-
posite GF : C → E defined by (GF )(X) := G(F (X)) and (GF )(f ) :=
G(F (f )) is also a functor.
To avoid overloading notation, we will often write F X and F f
instead of F (X) and F (f ).
2. Many examples
Example 5.2 (Identity). For every category C there is an identity
functor idC : C → C with idC (X) = X and idC (f ) = f .
Example 5.3 (Forgetful functors). Let R be a ring. Then we have a
functor
F : R Mod → Ab
49
50 5. FUNCTORS
defined by F (M ) := M (as an abelian group) and F (f ) := f . An

R-module M is an abelian group equipped with extra structure, and
this functor ‘forgets’ the extra structure. Other examples of forgetful
functors are the obvious functors Top → Set (forgetting the topology),
Ab → Set (forgetting the addition), and Ring → Ab (forgetting the
multiplication).
Example 5.4 (Hom functor). Let C be a category in which the Hom
classes are sets, and let X be an object of C. Then we define a functor
F : C → Set as follows. For an object Y in C we define
F (Y ) := HomC (X, Y ),
and for a morphism f : Y1 → Y2 in C we define
F (f ) : HomC (X, Y1 ) → HomC (X, Y2 ), g 7→ f g.
One easily checks that F is a functor. We will denote it by HomC (X, −).
Example 5.5 (Abelianization of a group). If G is a group then we
denote by [G, G] its commutator subgroup. This is the subgroup gen-
erated by the elements sts−1 t−1 with s, t ∈ G. It is a normal subgroup,
and the quotient group
Gab := G/[G, G]
is abelian. It is called the abelianization of G. If f : G → H is a group
homomorphism, then f ([G, G]) ⊂ [H, H], hence f induces a group ho-
momorphism
f ab : Gab → H ab .
Together, these constructions define a functor (−)ab : Grp → Ab.
Example 5.6 (Free module). Let R be a ring. For every set I we have
the free module
R(I) := (xi )i∈I ∈ RI | xi = 0 for all but finitely many i ,

see Example 1.12. If f : I → J is a map of sets, then we have an induced

map R(I) → R(J) determined by requiring that f maps the standard
basis vector ei of R(I) to the standard basis vector e′f (i) of R(J) . This
construction defines a functor Set → R Mod.
Example 5.7 (Units in a ring). If f : R → S is a ring homomorphism,
then f restricts to a group homomorphism R× → S × , and we obtain a
functor Ring → Grp.
2. MANY EXAMPLES 51
Example 5.8 (The fundamental group of a pointed space). Let Top⋆

be the category of pointed topological spaces. Objects in Top⋆ are pairs
(X, x) with X a topological space and x ∈ X. A morphism from (X, x)
to (Y, y) is a continuous map f : X → Y such that f (x) = y.
Then the fundamental group defines a functor
π1 : Top⋆ → Grp.
On the level of objects it is simply defined by mapping a pair (X, x) to

the fundamental group π1 (X, x). On the level of morphisms it is defined
as follows. Let (X, x) and (Y, y) be pointed spaces, and let f : X → Y
be a continuous map such that f (x) = y. Then we defined π1 (f ) by
π1 (f ) : π1 (X, x) → π1 (Y, y), [γ] 7→ [f ◦ γ],
where [γ] denotes the class of a loop γ : [0, 1] 7→ X based at x.
Note that the definition of fundamental group requires a base point.

Although for a path-connected space X and points x, y ∈ X the funda-
mental groups π1 (X, x) and π1 (X, y) are isomorphic, the isomorphism
is not unique, as it depends on the choice of a path.
Example 5.9 (Solutions to polynomial equations). Let f1 , . . . , fm ∈

Z[X1 , . . . , Xn ]. Then we can consider solutions to the system of poly-
nomial equations f1 = f2 = · · · = fm = 0 in arbitrary commutative
rings. Varying the commutative ring, one obtains a functor
F : CRing → Set
defined on the level of objects by
R 7→ {x = (x1 , . . . , xn ) ∈ Rn | f1 (x) = f2 (x) = · · · = fm (x) = 0}.
and on the level of morphisms by

φ : R → S 7→ (x1 , . . . , xn ) 7→ (φ(x1 ), . . . , φ(xn )) .
Indeed, if φ : R → S is a ring homomorphism, and if x ∈ Rn is a

solution to the system of equations, then also (φ(x1 ), . . . , φ(xn )) ∈ S n
is a solution to the system of equations.
52 5. FUNCTORS
Example 5.10 (Commutative diagrams as functors). Let C be the

small category
x f y
g
h
z
with f = hg (and where we omitted idx , idy and idz from the picture).
Let D be any category. Then a functor F : C → D consists of objects
X := F (x), Y := F (y), Z := F (z) together with morphisms F (f ),
F (g) and F (h) between them, such that F (f ) = F (h)F (g). In other
words, a functor F : C → D is the same as a triangular commutative
diagram
X Y
Z
in the category D. With a similar construction, a commutative diagram
in D of any shape can be thought of as a functor from some small
category C to D.
3. Contravariant functors
Definition 5.11. Let C and D be categories. A contravariant functor
from C to D consists of the data of
(1) for every object X in C an object F (X) in D,
(2) for every morphism f : X → Y in C a morphism F (f ) : F (Y ) →
F (X) in D
subject to the conditions
(F1) for every X in C we have F (idX ) = idF (X)
(F2’) for every f : X → Y and g : Y → Z in C we have F (gf ) =
F (f )F (g).
To stress the difference, one sometimes calls an ordinary functor a
covariant functor.
Remark 5.12. The only difference with the notion of a functor is that
F reverses the order of composition. In other words, a contravariant
functor F from C to D is the same as a functor F : C op → D.
4. FUNCTORS WITH MULTIPLE ARGUMENTS 53
To avoid clashing notation, we will reserve the notation F : C → D

for a functor from C to D, and will write F : C op → D for a contravariant
functor from C to D.
Example 5.13 (Contravariant Hom functor). Let C be a category and
X an object of C. If f : Y1 → Y2 is a morphism in C, then we have an
induced map of sets
HomC (Y2 , X) → HomC (Y1 , X), g 7→ gf,
and varying Y we obtain a contravariant functor from C to Set given
by
HomC (−, X) : C op → Set, Y 7→ HomC (Y, X).
This is a contravariant variation on Example 5.4.
Example 5.14. Another (related) example of a contravariant functor
is the ‘dual vector space’
(−)∨ : Vecop
K → VecK
which maps a K-vector space V to its dual V ∨ := HomK (V, K) and a
linear map f : V → W to the induced map
f ∨ : W ∨ → V ∨ , φ 7→ φ ◦ f.
Example 5.15 (Ring of functions on a space). Yet another (related)
example is the functor
C : Topop → CRing, X 7→ C(X) = {φ : X → R | f continuous},
mapping a topological space X to the ring of continuous R-valued func-
tions on X (with point-wise addition and multiplication). On the level
of maps it is defined as follows: if f : X → Y is continuous, then we
have an induced map
C(f ) : C(Y ) → C(X), φ 7→ φ ◦ f,
which is clearly a ring homomorphism.
4. Functors with multiple arguments

It is sometimes useful to consider functors with multiple arguments,
living in various categories. This can easily be formalized using the
notion of product category (see Definition 4.14). For example, a functor
F : C1 × C 2 → D
54 5. FUNCTORS
assigns to any pair of objects (X1 , X2 ) with Xi ∈ ob Ci an object

F (X1 , X2 ) in D, and to any pair of morphisms (f1 , f2 ) with fi : Xi → Yi
in Ci a morphism F (f1 , f2 ) : F (X1 , X2 ) → F (Y1 , Y2 ) in D.
This can be combined with the notion of a contravariant functor.
Example 5.16. A typical example is the functor
Hom(−, −) : C op × C → Set,
which can be defined for any locally small category C. If f : X ′ → X
and g : Y → Y ′ are morphisms in C (take note of the directions of the
arrows), then we have an induced morphism
Hom(f, g) : Hom(X, Y ) → HomR (X ′ , Y ′ ),
which is given by α 7→ gαf . One says that Hom(−, −) is contravariant
in the first, and covariant in the second argument.
EXERCISES 55
Exercises
Exercise 5.1. Let C and D be categories, let F : C → D be a functor.
Let f be an isomorphism in C. Show that F (f ) is an isomorphism in D.
Give an example where F (f ) is an isomorphism but f is not.
Exercise 5.2. Verify the claims in Example 5.5.
Exercise 5.3. For a non-negative integer n we denote by GLn (R) the
group of invertible n by n matrices with entries in R. In other words,
the group of units in the ring Matn (R). For example GL1 (R) = R× .
Show that GLn defines a functor from the category of commutative
rings CRing to the category of groups Grp.
Exercise 5.4 (Center is not a functor. . . ). Show that there exist mor-
phisms f : S2 → S3 and g : S3 → S2 in Grp with the property that
gf = idS2 . Deduce that there is no functor F : Grp → Ab such that
for every group G we have that F G is isomorphic to the center of G.
Exercise 5.5 (. . . but it is when we restrict to isomorphisms). Let
Grp× be the category whose objects are groups, and whose morphisms
are isomorphisms of groups (see also Exercise 4.5). Show that there is
a functor Z : Grp× → Ab× that maps a group G to its center.
CHAPTER 6
Morphisms of functors
1. Morphisms of functors
Definition 6.1. Let C and D be categories, and let F : C → D and
G : C → D be functors. A morphism or natural transformation η from
F to G consists of the data of
(1) for every object X in C a morphism ηX : F X → GX in D
subject to the condition
(N1) for every morphism f : X → Y in C the square
Ff
FX FY
ηX ηY
Gf
GX GY
in D commutes.
An isomorphism from F to G is a morphism of functors η such that ηX
is an isomorphism in D for every X in C.
We will write η : F → G to denote that η is a morphism from the
functor F to the functor G.
Example 6.2 (double dual). Let K be a field and V a K-vector-space.
Then we have a natural map
ηV : V → V ∨∨ = HomK (HomK (V, K), K), v 7→ (φ 7→ φ(v)) .
The word ‘natural’ is usually used in an informal sense, meaning ‘not
depending on the choice of a basis’. But it also has a precise math-
ematical meaning, namely that the collection of maps (ηV )V with V
running over all the vector spaces forms a morphism
η : idVecK → (−)∨∨
from the functor id : VecK → VecK to the functor (−)∨∨ : VecK →
VecK .
57
58 6. MORPHISMS OF FUNCTORS
Example 6.3. Consider the forgetful functor F : Grp → Set. Let n

be an integer. Then we have a morphism of functors η : F → F defined
by
ηG : G → G, g 7→ g n .
Note that g 7→ g n is a morphism of sets, but in general not a morphism
of groups.
Remark 6.4. Note that the world of categories has three layers:
(0) categories
(1) functors between categories
(2) morphisms (natural transformations) between functors
A similar picture arises in topology, where one distinguishes
(0) topological spaces
(1) continuous maps between spaces
(2) homotopies between continuous maps
There is more to this than just an analogy and modern category theory
and algebraic topology are heavily intertwined.
2. Equivalences of categories
Definition 6.5. An functor F : C → D is called an equivalence or
an equivalence of categories if there exists a functor G : D → C and
isomorphisms of functors
∼ ∼
ϵ : F G → idD , η : GF → idC .
A functor G with this property is called a quasi-inverse of F . If there
exists an equivalence from C to D then C and D are called equivalent
categories.
Remark 6.6. Note that this is formally very similar to the notion of
a homotopy equivalence in topology. See also Remark 6.4.
Equivalent categories tend to be ‘indistinguishable’ from the point
of view of category theory. See Exercise 6.13 for an example. It is
however often difficult to decide if a functor F is an equivalence from
the definition, since it can be hard to construct a quasi-inverse functor.
We end this chapter with a powerful criterion for testing if a functor F
is an equivalence.
Definition 6.7. Let F : C → D be a functor. We say that F is
2. EQUIVALENCES OF CATEGORIES 59
(1) full if for every X, Y in C the map

HomC (X, Y ) → HomD (F X, F Y ), f 7→ F f
is surjective;
(2) faithful if for every X, Y in C the map
HomC (X, Y ) → HomD (F X, F Y ), f 7→ F f
is injective;
(3) essentially surjective if for every object Z in D there is an X
in C such that F X and Z are isomorphic in D.
A functor which is full and faithful is often called fully faithful.
Theorem 6.8. A functor F : C → D is an equivalence of categories if
and only if it is full, faithful and essentially surjective.
We start with two lemmas.
Lemma 6.9. Let H : C → C be a functor, and assume that H is iso-
morphic to the functor idC . Then for all objects X and Y of C the
map
HomC (X, Y ) → HomC (HX, HY ), f 7→ Hf
is a bijection.
Proof. By assumption, there exists an isomorphism φ : idC → H.
By definition, this means that for all f : X → Y in C the square
φX
X HX
f Hf
φY
Y HY.
commutes. Note that φX and φY are isomorphisms (since φ is an
isomorphism). In particular, we have Hf = φY ◦ f ◦ φ−1
X . One can now
check directly that the map
HomC (HX, HY ) → HomC (X, Y ), g 7→ φ−1
Y ◦ g ◦ φX
is a two-sided inverse to the map f 7→ Hf . □
Lemma 6.10. Let
α β γ
S −→ T −→ U −→ V
be functions. Assume that the compositions βα and γβ are bijections.
Then α, β, and γ are bijections.
Proof. Since βα is surjective, β must be surjective. Since γβ is

injective β must also be injective. So β is a bijection. But then α and
γ must be bijections too. □
Proof of Theorem 6.8. Assume F is an equivalence of cate-

gories, with quasi-inverse G and isomorphisms
ϵ : F G → idD , η : GF → idC .
Then for every Y in D the object X := GY satisfies F X = F GY ∼ =Y
(via ϵY ), hence F is essentially surjective.
To see that F is full and faithful, let X and Y be objects of C and
consider the composition
HomC (X, Y ) → HomD (F X, F Y ) → HomC (GF X, GF Y ).
By Lemma 6.9 this composition is a bijection (since GF ∼
= idC ). Simi-
larly, the composition
HomD (F X, F Y ) → HomC (GF X, GF Y ) → HomD (F GF X, F GF Y ).
is a bijection (since F G ∼
= idC ). But now applying Lemma 6.10 we
conclude that
HomC (X, Y ) → HomD (F X, F Y )
is a bijection for all X and Y , and hence that F is full and faithful.
Conversely, assume that F is full, faithful and essentially surjective.
Essential surjectivity means that for every X in D there exists an object
GX in C and an isomorphism αX : F (GX) → X. Using a suitable form
of axiom of choice, we choose such a pair (GX, αX ) for every X in D.
We want to make the construction X 7→ GX into a functor (which
will be a quasi-inverse to F ). For this, we need to define for every
f : X → Y a map Gf : GX → GY . Note that the map f induces a
map
αY−1 f αX : F (GX) → F (GY ).
But since the functor F is full and faithful the map
HomC (GX, GY ) → HomD (F (GX), F (GY )), g 7→ F g
is a bijection. Hence there exists a unique g : GX → GY with F g =
αY−1 f αX , and we define Gf := g. One verifies that this G defines indeed
satisfies the axioms for a functor from D to C.
2. EQUIVALENCES OF CATEGORIES 61
Finally, to see that G is a quasi-inverse to F , note that the αX define

an isomorphism of functors α : F G → idD . In the other direction, given
an object X in C we define
βX : GF X → X
to be the pre-image of αF X under the bijection
HomC (GF X, X) → HomD (F GF X, F X).
Then one verifies that the βX define an isomorphism of functors GF →
idC , and we conclude that F is an equivalence of categories. □
Example 6.11. Let K be a field, let FVecK be the category of finite-
dimensional K-vector spaces. Let C be the category of matrices over
K, see Example 4.10. Consider the functor
F : C → FVecK
that maps an object n of C to the vector space K n and a matrix
A ∈ HomC (n, m) = Matm,n (K)
to the corresponding linear map K n → K m . We claim that F is an
equivalence of categories.
Indeed, using Theorem 6.8 it suffices to observe that the functor
is essentially surjective, since every finite-dimensional vector space is
isomorphic to K n for some n, and that the functor is full and faithful,
since for every m and n the map
Matm,n (K) → HomK (K n , K m )
is a bijection.
Note that C and FVecK are not isomorphic as categories. The
category FVecK is much bigger than C since for every n there are
infinitely many n-dimensional vector spaces (all isomorphic, but not
equal ). In fact, C is a small category, whereas FVecK is not.
Exercises
Exercise 6.1. Let η : idAb → idAb be a morphism of functors. Show
that there is an integer n such that for every A ∈ ob Ab and for every
x ∈ A the identity ηA (x) = nx holds.
Exercise 6.2. Show that taking determinants defines a morphism
det : GLn → GL1
between functors from CRing to Grp. (See Exercise 5.3).
Exercise 6.3. Let G and H be groups, and BG and BH the corre-
sponding one-object categories (see Example 4.7). Show that a functor
F : BG → BH is the same as a group homomorphism f : G → H, and
a morphism of functors η : F1 → F2 is the same as an element h ∈ H
such that hf1 (g)h−1 = f2 (g) for all g ∈ G.
Exercise 6.4. Let R be a ring. Recall that the center of a ring is the
subring
Z(R) = {z ∈ R | zr = rz for all r ∈ R}.
Denote by C the category of left R-modules.
(1) Let z ∈ Z(R). Show that ηz,M : M → M, x 7→ zx defines a
morphism of functors ηz : idC → idC .
(2) Let η : idC → idC be a morphism of functors. Show that there
is a z ∈ Z(R) with η = ηz .
Exercise 6.5. Show that there are precisely two morphisms of functors
idGrp → idGrp .
Exercise 6.6. Consider categories and functors as in the following
diagram:
F0
G
C D E
F1
Let η : F0 → F1 be a morphism of functors. Construct a morphism of
functors GF0 → GF1 .
Exercise 6.7 (Equivalence is an equivalence relation). Let F : C → D
and G : D → E be equivalences of categories. Show that GF : C → E is
an equivalence of categories.
Exercise 6.8. Let F : C → D be a full and faithful functor. Let X and
Y be objects in C.
EXERCISES 63
(1) Let f : X → Y be a morphism. Show that f is an isomorphism

if and only if F f is an isomorphism.
(2) Show that X and Y are isomorphic if and only if F X and F Y
are isomorphic.
Exercise 6.9. For a category C, we denote by [C] the class of isomor-

phism classes of objects in C. Let F : C → D be a functor. Show that
F induces a map [F ] : [C] → [D]. Show that if F is fully faithful, then
[F ] is injective, and if F is essentially surjective, then [F ] is surjective.
Exercise 6.10. Let C be a non-empty locally small category in which

all objects are isomorphic and in which every morphism is an isomor-
phism. Show that there is a group G and an equivalence of categories
BG → C.
Exercise 6.11 (Fundamental groupoid). Let X be a topological space.

Let Π1 (X) be the category with
(1) objects: ob Π1 (X) = X
(2) morphisms: HomΠ1 (X) (x, y) the set of homotopy classes of
paths from x to y
(3) composition: composition of paths
Verify that this indeed defines a category. It is called the fundamental
groupoid of X.
Assume that X is path connected, and let x ∈ X. Show that Π1 (X)
is equivalent with the category Bπ1 (X, x). (See Example 4.7).
Exercise 6.12. Give for every n ∈ N a category Cn so that Cn has

exactly n objects, and all the Cn are equivalent.
Exercise 6.13. Let F : C → D be an equivalence of categories. Show

that C has a final object if and only if D has a final object.
Exercise 6.14. Let R and S be rings. Show that the categories

R Mod × S Mod and R×S Mod are equivalent.
Exercise 6.15 (Morita equivalence (⋆)). Let R be a ring and n a

positive integer. If M is an R-module, then we can consider elements
of M n as length n column matrices with entries in M . In this way, we
have an action
   
x1 x1
 x2   x2 
Matn (R) × M n → M n , (A,  .. ) 7→ A ·  .. 
   
 .   . 
xn xn
This makes Mn into a left Matn (R)-module. Verify that this defines a
functor
R Mod → Matn (R) Mod, M 7→ M n .
Show that this functor is an equivalence of categories.
Exercise 6.16 (Abelianized fundamental group without base point
(⋆)). Let C be the category of path connected topological spaces. Let
P be the functor from Top⋆ to C that maps a pair (X, x) the the path
component of x ∈ X. Show that there is a functor F : C → Ab and an
isomorphism between the functors
π1ab : Top⋆ → Ab, (X, x) 7→ π1 (X, x)ab
and F ◦ P . Bonus question: show that there is no functor F : C → Grp
and an isomorphism between F ◦ P and π1 .
CHAPTER 7
Tensor product
1. Tensor product of a right and a left module

Definition 7.1. Let R be a ring, M a right R-module, and N a left
R-module. Let A be an abelian group. A map
f: M ×N →A
is said to be R-bilinear if for all x, x1 , x2 ∈ M , y, y1 , y2 ∈ N and r ∈ R

the following hold:
(1) f (x1 + x2 , y) = f (x1 , y) + f (x2 , y)
(2) f (x, y1 + y2 ) = f (x, y1 ) + f (x, y2 )
(3) f (xr, y) = f (x, ry)
Remark 7.2. These conditions imply that moreover

(4) f (x, 0) = 0, and
(5) f (0, y) = 0
hold for all x ∈ M and y ∈ N . See Exercise 7.1.
Note that if f : M × N → A is bilinear, and h : A → B is a homo-

morphism of abelian groups, then the composition hf : M × N → B
is also a bilinear map. The following theorem states that there is a
‘universal’ bilinear map, from which all others can be obtained by a
unique composition with a homomorphism of abelian groups.
Theorem 7.3. Let R be a ring, M a right R-module and N a left

R-module. Then there exists an abelian group T and an R-bilinear map
g: M × N → T
65
66 7. TENSOR PRODUCT
such that for every abelian group A and every R-bilinear map f : M ×
N → A there is a unique group homomorphism h : T → A with f = hg:
g
M ×N T
f
h
A
Moreover, the pair (T, g) is unique up to unique isomorphism in the
following sense: if both (T1 , g1 ) and (T2 , g2 ) satisfy the above property,
then there is a unique isomorphism h : T1 → T2 such that g2 = hg1 .
Definition 7.4. We will call the abelian group T (unique up to unique
isomorphism) the tensor product of M and N , and denote it by M ⊗R
N := T . For x ∈ M and y ∈ N we denote the image of (x, y) in M ⊗R N
by x ⊗ y := g(x, y).
Proof of Theorem 7.3. Uniqueness. This is purely formal: ev-
erything defined by a universal property is unique up to unique iso-
morphism, by an argument that is basically the same as the proof of
Proposition 4.21: Assume (T1 , g1 ) and (T2 , g2 ) both satisfy the required
property. Since g1 : M × N → T1 is bilinear there is a unique map
h1 : T2 → T1 with g1 = h1 g2 . Reversing the roles of T1 and T2 , we get
a unique map h2 : T1 → T2 . Moreover, both the compositions h1 h2 and
h2 h1 must be the identity, so we conclude that h1 is an isomorphism.
Existence. This part is certainly not formal! The proof is a bit
messy, but in a way very natural: we just construct an abelian group
T (and a map g) with all the desired properties built-in.
Let F := Z(M ×N ) be the free Z-module on the set M × N . Given
an element (x, y) ∈ M × N , we denote by e(x,y) ∈ Z(M ×N ) the corre-
sponding basis vector, see 1.12. There is a canonical map of sets
M × N → F, (x, y) 7→ e(x,y) .
This map has no reason to be bilinear. We will force it to become
bilinear by dividing out the necessary relations. Let G ⊂ F be the
subgroup generated by the elements
e(x1 +x2 ,y) − e(x1 ,y) − e(x2 ,y) ,
e(x,y1 +y2 ) − e(x,y1 ) − e(x,y2 ) ,
e(xr,y) − e(x,ry) ,
1. TENSOR PRODUCT OF A RIGHT AND A LEFT MODULE 67
for all x1 , x2 , x ∈ M and y1 , y2 , y ∈ N and r ∈ R. Let T be the quotient

group F/G, and consider the composition
M ×N F
g
T
Then the map g is bilinear by construction.
We now show that (T, g) is a tensor product. Let f : M × N → A
be a billinear map. Then there is a unique homomorphism f ′ : F → A,
which sends the basis vector e(x,y) to f (x, y), see Proposition 1.13. Since
f is bilinear, we have that f ′ vanishes on all the generators of G, and
therefore that f ′ (G) = 0. Hence f ′ induces a homomorphism h : T → A
with f = hg. To see that a map h with this property is unique, note
that T is generated by the images of the elements (x, y) ∈ M × N , and
that h must send the image of (x, y) to f (x, y). □
Remark 7.5. In practice it is often easier not to use the actual con-
struction of the tensor product in the proof of Theorem 7.3, but only
the defining universal property in the statement of Theorem 7.3, to-
gether with the fact that the the tensor product is generated by the
elements of the form x ⊗ y with x ∈ M and y ∈ N .
Remark 7.6. Elements of M ⊗R N are finite sums of elements of the
form x ⊗ y, but these are not independent. In fact, the map
M × N → M ⊗R N, (x, y) 7→ x ⊗ y
is R-bilinear (by definition of the tensor product), so that for all x, x1 , x2 ∈
M , y, y1 , y2 ∈ N and r ∈ R the identities
(x1 + x2 ) ⊗ y = (x1 ⊗ y) + (x2 ⊗ y)
x ⊗ (y1 + y2 ) = (x ⊗ y1 ) + (x ⊗ y2 )
(xr) ⊗ y = x ⊗ (ry)
x⊗0=0
0⊗y =0
hold in M ⊗R N .
We end this section with a few examples of tensor products.
Example 7.7. Let M be a right R-module, then we claim M ⊗R {0} ∼ =

{0}. Indeed, the tensor product is generated as an abelian group by
the elements x ⊗ 0, but these all are equal to 0 in M ⊗R {0}.
Example 7.8. We claim that for any left R-module M there is a unique
isomorphism
∼
f : R ⊗R M → M
satisfying
r ⊗ x 7→ rx.
Indeed, the map R × M → M, (r, x) 7→ rx is R-bilinear, and hence
induces an R-linear homomorphism f : R ⊗R M → M with f (r ⊗ x) =
rx. Conversely, the map M → R ⊗R M given by x 7→ 1 ⊗ x is R-linear,
and is a two-sided inverse to the map f .
Of course, in the same way one can produce an isomorhpism
∼
M ⊗R R → M, x ⊗ r 7→ xr
for every right R-module M .
Example 7.9. The tensor product of two non-zero modules can be
zero. For example, we have
(Z/2Z) ⊗Z (Z/3Z) = 0.
Indeed, the tensor product is generated by elements of the form x ⊗ y
with x ∈ Z/2Z and y ∈ Z/3Z. But since 3x = x for all x ∈ Z/2Z we
find
x ⊗ y = (3x) ⊗ y = x ⊗ (3y) = x ⊗ 0 = 0.
Similarly we have (Z/nZ) ⊗Z (Z/mZ) = 0 whenever n and m are co-
prime. See also Exercise 7.3.
2. Tensor products and bimodules

In many cases, the tensor product of two modules is not just an
abelian group, but itself again a module.
Definition 7.10. Let R and S be rings. An (R, S)-bimodule is an
abelian group M equipped with operations
R × M → M, (r, x) 7→ rx
and
M × S → M, (x, s) 7→ xs
such that
2. TENSOR PRODUCTS AND BIMODULES 69
(B1) the first operation makes M into a left R-module

(B2) the second operation makes M into a right S-module
(B3) for all r ∈ R, s ∈ S and x ∈ M the identity r(xs) = (rx)s
holds in M .
A map f : M → N is called a morphism of (R, S)-bimodules if it is both
a morphism of left R-modules and a morphism of right S-modules. We
denote the category of (R, S)-bimodules by R ModS .
Because of axiom (B3), we can simply write rxs for r(xs) = (rx)s,
and we will frequently describe the structure of an (R, S)-bimodule by
the map
R × M × S → M, (r, x, s) 7→ rxs,
simultaneously encoding the left and right module structures.
Example 7.11. An abelian group A has a unique structure of (Z, Z)-
bimodule, which is given by nam := (nm)a for all n, m ∈ Z and a ∈ A.
It follows that a (Z, Z)-bimodule is the same thing as an abelian group.
Similarly, an (R, Z)-bimodule is the same as a left R-module, and a
(Z, R)-bimodule is the same as a right R-module.
Example 7.12. R be a commutative ring and M an R-module. Then
M also is an (R, R)-bimodule by setting
rxs := rsx
for all r, s ∈ R and x ∈ M .
Example 7.13. Let R be a ring and n a non-negative integer. Then
Rn becomes an (R, R)-bimodule with
r(x1 , . . . , xn )s := (rx1 s, . . . , rxn s).
Example 7.14. Let R be a commutative ring and n and m non-
negative integers. Then the additive group M = Matm,n (R) of n by m
matrices is a (Matn (R), Matm (R))-bimodule where for A ∈ Matn (R),
B ∈ Matm (R) and X ∈ M the element AXB of M is defined by matrix
multiplication.
Example 7.15. Let M a left S-module and N be a left R-module.
Consider the group Hom(M, N ) of homomorphisms of abelian groups.
This group carries a natural structure of (R, S)-bimodule with

rf s := M → N, x 7→ rf (sx)
for all r ∈ R, f ∈ Hom(M, N ) and s ∈ S.
Proposition 7.16. Let R, S, and T be rings. Let M be an (R, S)-

bimodule, and let N be an (S, T )-bimodule. Then the tensor product
M ⊗S N has a unique structure of an (R, T )-bimodule satisfying
r(x ⊗ y)t = (rx) ⊗ (yt)
for all r ∈ R, x ∈ M , y ∈ N and t ∈ T .
If R is a commutative ring then any R-module is canonically an
(R, R)-bimodule, and we find that the tensor product of two R-modules
over R is naturally an R-module. When dealing with commutative
rings, there is no essential distinction between left and right modules,
and we will usually treat the tensor product as an operation that pro-
duces a left R-module out of two left R-modules.
3. Tensor product as a functor

Let R be a ring and let f : M1 → M2 be a morphism of right R-
modules, and let g : N1 → N2 be a morphism of left R-modules. Then
the map
M1 × N1 → M2 ⊗R N2 , (x, y) 7→ f (x) ⊗ g(y)
is R-bilinear. Hence, by the universal property of the tensor product,
there exists a unique homomorphism of abelian groups
f ⊗ g : M1 ⊗R N1 → M2 ⊗R N2
that maps an element x ⊗ y to f (x) ⊗ g(y). This upgrades the tensor
product from being just a construction on pairs of modules to a functor
− ⊗R − : ModR × R Mod → Ab.
Similarly, if R, S and T are rings, then we have a functor
− ⊗S − : R ModS × S ModT → R ModT .
Proposition 7.17. Let R be a ring and let N be a left R-module. If
f g
(4) M1 −→ M2 −→ M3 −→ 0
is an exact sequence of right R-modules, then the induced sequence
f ⊗id g⊗id
(5) M1 ⊗R N −→ M2 ⊗R N −→ M3 ⊗R N −→ 0
of abelian groups is exact. Similarly, if
N1 −→ N2 −→ N3 −→ 0
3. TENSOR PRODUCT AS A FUNCTOR 71
is an exact sequence of left R-modules, and M a right R-module, then

the induced sequence
M ⊗R N1 −→ M ⊗R N2 −→ M ⊗R N3 −→ 0
is exact.
The functor − ⊗R N in general does not preserve short exact se-
quences: even if
0 → M1 → M 2 → M3 → 0
is a short exact sequence, then we only obtain a partial exact sequence
M1 ⊗R N → M2 ⊗R N → M3 ⊗R N → 0.
See Exercise 7.15.
Proof of Proposition 7.17. For all x ∈ M1 and y ∈ N we have
gf (x) ⊗ y = 0, hence im(f ⊗ id) ⊂ ker(g ⊗ id). In particular, we have
an induced map
M2 ⊗R N
(6) Φ: −→ M3 ⊗R N.
im(f ⊗ id)
To show that the sequence (5) is exact, it suffices to show that the
above map is an isomorphism. We will verify this by constructing an
inverse map.
For every x ∈ M3 choose an x′ ∈ M2 with g(x′ ) = x (note that g is
surjective). We define a map
M2 ⊗R N
M3 × N → , (x, y) 7→ x′ ⊗ y.
im(f ⊗ id)
This is well-defined, since if x′′ is another element with g(x′′ ) = x, then
(x′′ ⊗ y) − (x′ ⊗ y) = (x′′ − x′ ) ⊗ y ∈ im(f ⊗ id),
using the exactness of the original sequence (4). Moreover, the map is
bilinear, so it induces a homomorphism
M2 ⊗R N
M3 ⊗R N →
im(f ⊗ id)
and one verifies that this is a two-sided inverse to the map Φ. □
Example 7.18. Proposition 7.17 can be a powerful tool in computing
tensor products. As an example, let us use it to compute M ⊗Z (Z/nZ)
for any Z-module M . We have an exact sequence
n
Z −→ Z −→ Z/nZ −→ 0
of Z-modules, which induces an exact sequence

n⊗id
Z ⊗Z M −→ Z ⊗Z M −→ Z/nZ ⊗Z M −→ 0.
We have Z⊗Z M ∼ = M (see Example 7.8), and the above exact sequence
is isomorphic with the exact sequence
n
M −→ M −→ Z/nZ ⊗Z M −→ 0,
from which we conclude that Z/nZ ⊗Z M is isomorphic with M/nM .
4. The adjunction
Let R and S be rings. If N is an (R, S)-bimodule, and P a right
S-module, then HomS (N, P ) is naturally a right R-module, with the
action of R defined by:
f r : N → P, x 7→ f (rx).
See Exercise 7.9.
Theorem 7.19 (Tensor-Hom adjunction). Let R and S be rings. Let
M be a right R-module, N an (R, S)-bimodule, and P be a right S-
module. Then the map of abelian groups
HomS (M ⊗R N, P ) → HomR (M, HomS (N, P ))
given by
f 7→ x 7→ (y 7→ f (x ⊗ y))
is an isomorphism.
Remark 7.20. The HomR and HomS in the theorem denote the set of
homomorphisms in the categories of right R- and S-modules.
Proof of Theorem 7.19. Given an R-linear map f : M → HomS (N, P )
we obtain a map
M × N → P, (x, y) 7→ f (x)(y)
which is R-bilinear, hence it defines a homomorphism
f ′ : M ⊗R N → P
with the property that it maps x ⊗ y to f (x)(y). This map is S-linear.
This construction defines a homomorphism
HomR (M, HomS (N, P )) → HomS (M ⊗R N, P ), f 7→ f ′ .
We leave it to the reader to verify that this is a two-sided inverse to the
map in the theorem. □
4. THE ADJUNCTION 73
Remark 7.21. There is completely analogous theorem about tensoring

on the left with a fixed (S, R)-bimodule N . It gives an isomorphism
HomS (N ⊗R M, P ) → HomR (M, HomS (N, P ))
where M is a left R-module and P a left S-module. In this version,
HomS and HomR denote the sets of homomorphisms in the categories
of left S- and R-modules.
Exercises
Exercise 7.1. Prove that if f : M ×N → A is R-bilinear, then f (x, 0) =
0 for all x ∈ M and f (0, y) = 0 for all y ∈ N .
Exercise 7.2. Let R be a ring, let M be a left R-module and let n be
a non-negative integer.
(1) Show that the map Rn × M → M n given by
((r1 , . . . , rn ), x) 7→ (r1 x, . . . , rn x)
is R-bilinear.
(2) Show that the above map is universal, and conclude that Rn ⊗R
M∼ = M n.
Exercise 7.3. Let n and m be positive integers with greatest common
divisor d. Show that
(1) (Z/nZ) ⊗Z (Z/mZ) ∼= Z/dZ;
(2) Q ⊗Z (Z/mZ) ∼= 0;
(3) Q ⊗Z Q ∼= Q.
Exercise 7.4. Let K be a field and R := Matn (K) the ring of n by n
matrices. Let M = K n be the right R-module of length n row vectors,
and N = K n the left R-module of length n column vectors. Show that
the abelian group M ⊗R N is isomorphic to K.
Exercise 7.5. Let R be a commutative ring and f, g ∈ R. Show that
R/(f ) ⊗R R/(g) ∼
= R/(f, g)
as R-modules.
Exercise 7.6. Let R be a ring, let I ⊂ R be a (two-sided) ideal and
let M be a left R-module. Show that (R/I) ⊗R M is isomorphic to
M/IM .
Exercise 7.7. Let R be a commutative ring. Show that the functors
R Mod × R Mod → R Mod given by (M, N ) 7→ M ⊗R N and (M, N ) 7→
(N ⊗R M ) are isomorphic.
Exercise 7.8. Verify that the bimodule in Example 7.15 indeed satisfies
the bimodule axioms.
Exercise 7.9. Let R and S be rings. Let M be a left R-module and
N an (R, S)-bimodule. Show that HomR (M, N ) is naturally a right
S-module and that HomR (N, M ) is naturally a left S-module.
EXERCISES 75
Exercise 7.10. Let R be a ring and let M1 , M2 be right R-modules

and N a left R-module. Show that there is a unique isomorphism
(M1 ⊕ M2 ) ⊗R N → (M1 ⊗R N ) ⊕ (M2 ⊗R N )
such that
(x, y) ⊗ z 7→ (x ⊗ z, y ⊗ z)
for all x ∈ M1 , y ∈ M2 and z ∈ N .
Exercise 7.11. Let R and S be rings, let M1 be a right R-module,
M2 an (R, S)-bimodule, and M3 a left S-module. Show that there is a
unique isomorphism
(M1 ⊗R M2 ) ⊗S M3 → M1 ⊗R (M2 ⊗S M3 )
such that
(x ⊗ y) ⊗ z 7→ x ⊗ (y ⊗ z)
for all x ∈ M1 , y ∈ M2 and z ∈ M3 .
Exercise 7.12. Let R be an integral domain with fraction field K. Let
M be an R-module. Show that every element of K ⊗R M is of the form
λ ⊗ x with λ ∈ K and x ∈ M .
Exercise 7.13. Let R be a ring, M a right and N a left R-module.
Show that there is a surjective morphism of abelian groups
M ⊗Z N → M ⊗R N
which maps x ⊗ y to x ⊗ y.
Exercise 7.14. Let R, S, T be rings. Let M be an (R, S)-bimodule and
N an (S, T )-bimodule. Let P be an (R, T )-bimodule and let f : M ×
N → P be an S-bilinear map. Let h : M ⊗S N → P be the unique
additive map such that h(x ⊗ y) = f (x, y) for all x ∈ M , y ∈ N .
(1) show that h is a homomorphism of (R, T )-bimodules if and
only if the bilinear map f satisfies f (rx, y) = rf (x, y) and
f (x, yt) = f (x, y)t for all x ∈ M , y ∈ N , r ∈ R and t ∈ T .
(2) conclude that the map M ×N → M ⊗S N is universal amongst
bilinear maps f : M × N → P to (R, T )-bimodules satisfying
the identities in (1).
Exercise 7.15. Consider the short exact sequence of Z-modules
2
0 −→ Z −→ Z −→ Z/2Z −→ 0.
Show that the sequence obtained by applying the functor − ⊗Z Z/2Z

is not exact.
Exercise 7.16. Let R be a ring and let 0 → M1 → M2 → M3 → 0
be a short exact sequence of right R-modules. Let N be a free left
R-module. Show that the induced sequence
0 → M1 ⊗R N → M2 ⊗R N → M3 ⊗R N → 0
is a short exact sequence.
Exercise 7.17. Let R be a ring and let 0 → M1 → M2 → M3 → 0
be a split short exact sequence of right R-modules. Let N be a left
R-module. Show that the induced sequence
0 → M1 ⊗R N → M2 ⊗R N → M3 ⊗R N → 0
is a short exact sequence.
Exercise 7.18. Let R be a principal ideal domain and let
M∼ = Rm ⊕ R/pe1 R ⊕ · · · ⊕ R/pen R
1 n
be a finitely generated R-module (with the notation of Corollary 3.6).
(1) Let K be the fraction field of R. Compute dimK (K ⊗R M ).
(2) Let p ∈ R be irreducible. Compute dimR/pR (R/pR ⊗R M ).
CHAPTER 8
Adjoint functors
1. Adjoint pairs of functors

Definition 8.1. Let F : C → D and G : D → C be functors between
locally small categories. An adjunction between F and G is an isomor-
phism
∼
α : HomD (F (−), −) −→ HomC (−, G(−))
of functors C op × D → Set (see Example 5.16). If such an adjunction
exists, we say F is left adjoint to G, and G is right adjoint to F .
In other words, an adjunction from F to G consists of the data of
a bijection
∼
(7) αX,Y : HomD (F X, Y ) −→ HomC (X, GY ),
for every X in C and Y in D, such that for every f : X1 → X2 the
square
αX1 ,Y
HomD (F X1 , Y ) HomC (X1 , GY )
−◦F f −◦f
αX2 ,Y
HomD (F X2 , Y ) HomC (X2 , GY )
commutes, and for every g : Y1 → Y2 the square
αX,Y1
HomD (F X, Y1 ) HomC (X, GY1 )
g◦− Gg◦−
αX,Y2
HomD (F X, Y2 ) HomC (X, GY2 )
commutes.
Remark 8.2. The terminology comes from an analogy with linear al-
gebra: if V and W are vector spaces equipped with inner products,
77
78 8. ADJOINT FUNCTORS
then linear maps f : V → W and g : W → V are called adjoint if we

have
⟨f (v), w⟩W = ⟨v, g(w)⟩V
for all v ∈ V and w ∈ W .
Assume that F : C → D is a left adjoint of G : D → C, with an
adjunction α. Taking Y = F X in (7), we obtain a bijection
∼
αX,F X : HomD (F X, F X) −→ HomC (X, GF X).
The image of idF X under this map gives a map
ηX : X → GF X
in C. Using the fact that α is a morphism of functors, one shows that
the ηX form a morphism of functors
η : idC → GF.
Similarly, taking X = GY in (7) we obtain a morphism of functors
ϵ : F G → idD .
The morphisms of functors η and ϵ are called the unit and co-unit of
the adjunction between F and G.
2. Many examples
The main reason that adjunctions between functors are interesting,
is that they are ubiquitous: they arise surprisingly often in multiple
branches of mathematics. Here is a short list of examples.
Example 8.3 (Cartesian product and set of maps). Fix a set A. Then
for all sets X and Y we we have a canonical bijection
∼
αX,Y : Hom(X × A, Y ) → Hom(X, Hom(A, Y ))
given by mapping a function f : X × A → Y to the function
X → Hom(A, Y ), x 7→ (a 7→ f (x, a)) .
An inverse is given by mapping a function g : X → Hom(A, Y ) to
X × A → Y, (x, a) 7→ g(x)(a).
It is easy to check that α defines an adjunction, making the functor
Set → Set, X 7→ X × A
2. MANY EXAMPLES 79
into a left adjoint to the functor

Set → Set, Y 7→ Hom(A, Y ).
The unit η : id → Hom(A, − × A) of this adjunction is given by
ηX : X → Hom(A, X × A), x 7→ (a 7→ (x, a))
and the co-unit ϵ : Hom(A, −) × A → id is given by
ϵX : Hom(A, X) × A → X, (f, a) 7→ f (a).
Example 8.4 (Tensor product and Hom). This is a variation on the
previous example. Let R and S be rings, and let A be an (R, S)-
bimodule. Then the functor
ModR → ModS , M 7→ M ⊗R A
is left adjoint to the functor
ModS → ModR , N 7→ HomS (A, N ),
which comes down to the functorial isomorphism
∼
αM,N : HomS (M ⊗R A, N ) −→ HomR (M, HomS (A, N ))
of Theorem 7.19.
Example 8.5 (Free module and forgetful functor). Let R be a ring.
Let M be an R-module and let R(I) be the free R-module on a set I
(see Example 5.6). Then we have a canonical map
αI,M : HomR Mod (R(I) , M ) −→ HomSet (I, M ),
given by restricting a module homomorphism φ : R(I) → M to the
standard basis {ei : i ∈ I}. This map is a bijection, since a module
homomorphism R(I) → M is uniquely determined by the images of the
basis vectors ei , and conversely, given a map of sets f : I → M we
obtain an R-module homomorphism
X X
R(I) → M, ri ei 7→ ri f (i).
i∈I i∈I
This is just a reformulation of the familiar fact from linear algebra: to
give a linear map from V to W is the same as to give the images of the
vectors in a basis of V .
If we denote by
G : R Mod → Set, M 7→ M
the forgetful functor (see Example 5.3) and by
F : Set → R Mod, I 7→ R(I)

the free module functor, then α defines a bijection
∼
αI,M : HomR Mod (F I, M ) −→ HomSet (I, GM ),
and one can verify directly that this defines an adjunction, making the
free module functor F into a left adjoint to the forgetful functor G.
The unit of this adjunction is the morphism η : id → GF given by the
function
ηI : I → R(I) , i 7→ ei ,
for every set I.
Example 8.6 (Discrete topology, forgetful functor, trivial topology).
Any function from a discrete topological space is automatically contin-
uous. Likewise, any function to a trivial topological space is automati-
cally continuous. That is, we have
HomTop (Xdisc , Y ) = HomSet (X, Y )
and
HomSet (X, Y ) = HomTop (X, Ytriv ),
and we see that the discrete topology functor
Set → Top, X 7→ Xdisc
is left adjoint to the forgetful functor Top → Set, and that the trivial
topology functor
Set → Top, Y 7→ Ytriv
is right adjoint to the forgetful functor.
Example 8.7 (Frobenius reciprocity). Let k be a field, let G be a
group and let H ⊂ G be a subgroup. Then Frobenius reciprocity gives
for every k-linear representation V of H and W of G a canonical iso-
morphism
∼
Homk[G] (IndG G
H V, W ) → Homk[H] (V, ResH W ),
which makes the functor IndG G

H into a left adjoint to ResH .
3. YONEDA AND UNIQUENESS OF ADJOINTS 81
3. Yoneda and uniqueness of adjoints

Let C be a locally small category. If X is an object in C, then we
have a functor
hX := HomC (−, X) : C op → Set,
see also 5.13. Now the functors from C op to Set form themselves the
objects of a category Fun(C op , Set), in which the morphisms are the
morphisms of functors. We obtain a functor
h : C → Fun(C op , Set), X 7→ hX
On the level of morphisms it is given by sending a map f : X → Y to
the natural transformation hf : hX → hY given by
hf,T : HomC (T, X) → HomC (T, Y ), g 7→ f g
for every T in C.
Theorem 8.8 (Yoneda’s Lemma). The functor
C → Fun(C op , Set), X 7→ hX
is fully faithful.
Proof. In other words, we need to show that for all pairs of objects
X, Y in C the map
(8) HomC (X, Y ) → HomFun(C op ,Set) (hX , hY ), f 7→ hf
is a bijection.
We show this by constructing an inverse bijection. Let φ : hX →
hY be a morphism of functors. Then for every T we have a map
φT : hX (T ) → hY (T ), and in particular, taking T = X, we have a
map
φX : HomC (X, X) → HomC (X, Y )
and the image of idX defines an element φX (idX ) in HomC (X, Y ). We
obtain a map
(9) HomFun(C op ,Set) (hX , hY ) → HomC (X, Y ), φ 7→ φX (idX ).
Using the definition of hf , we see that for a morphism f : X → Y
we have
hf,X (idX ) = f ◦ idX = f,
and hence that the composition of (8) followed by (9) is the identity on
HomC (X, Y ).
To see that the other composition is the identity, let φ : hX → hY

be a morphism of functors. Under (9) it is mapped to φX (idX ), which
in turn under (8) is mapped to hφX (idX ) . To see that the morphisms
of functors φ and hφX (idX ) coincide, it suffice to verify that for all T in
C the maps hφX (idX ),T and φT from hX (T ) = Hom(T, X) to hY (T ) =
Hom(T, Y ) coincide.
So let g ∈ Hom(T, X). We have
hφX (idX ),T (g) = φX (idX ) ◦ g.
Since φ is a morphism of functors, the square
φX
HomC (X, X) HomC (X, Y )
−◦g −◦g
φT
HomC (T, X) HomC (T, Y )
commutes. Tracing the element idX under the two paths from HomC (X, X)
to HomC (T, Y ) we find
φX (idX ) ◦ g = φT (g)
and hence
hφX (idX ),T (g) = φT (g),
as we had to show. □
Remark 8.9. There is something quite striking in the proof. The
inverse bijection φ 7→ φX (idX ) proceeds by first removing from the
morphism of functors φ = (φT )T all components except for the one
at T = X, and then restricting the remaining function φX to just the
element idX of Hom(X, X). Yet, despite this apparent massive loss of
information, φ 7→ φX (idX ) is a bijection, so that φ can be completely
recovered from φX (idX ).
Corollary 8.10. If hX and hY are isomorphic functors, then X and
Y are isomorphic objects in C.
Proof. See Exercise 6.8. □
Corollary 8.11 (Uniqueness of right adjoints). If both G1 : D → C and
G2 : D → C are right adjoints to a functor F : C → D, then G1 and G2
are isomorphic functors.
3. YONEDA AND UNIQUENESS OF ADJOINTS 83
Proof. Choose adjunctions between F and G1 and between F and

G2 . Then we obtain isomorphisms
∼ ∼
HomC (X, G1 Y ) ←− HomD (F X, Y ) −→ HomC (X, G2 Y ),
functorial in X and Y . Composing these, we find for all X in C and Y
in D an isomorphism
HomC (X, G1 Y ) → HomC (X, G2 Y ).
Functoriality in X implies that for every Y in D we find an isomorphism
∼
HomC (−, G1 Y ) −→ HomC (−, G2 Y )
of functors C op → Set, which by Yoneda’s Lemma and Exercise 6.8
comes from a unique isomorphism
∼
γY : G1 Y −→ G2 Y
in C. Functoriality in Y implies that the collection (γY )Y defines an
isomorphism of functors
∼
γ : G1 −→ G2
which finishes the proof. □
There is (of course) a dual of Yoneda’s lemma. Given an object X
in C, consider the functor
hX := HomC (X, −) : C → Set.
We have a functor
C op → Fun(C, Set), X 7→ hX
On the level of morphisms it is given by sending a map f : X → Y to
the natural transformation hf : hY → hX given by
hfT : HomC (Y, T ) → HomC (X, T ), g 7→ gf
for every T in C.
Theorem 8.12 (co-Yoneda’s Lemma). The functor
C op → Fun(C, Set), X 7→ hX
is fully faithful. □
Corollary 8.13 (Uniqueness of left adjoints). If both F1 : C → D and
F2 : C → D are left adjoints to a functor G : D → C, then F1 and F2
are isomorphic functors. □
Exercises
Exercise 8.1. Verify that the unit η and the co-unit ϵ of an adjunction
are indeed morphisms of functors.
Exercise 8.2. Let F : C → D be an equivalence with quasi-inverse
G : D → C. Show that F is both left and right adjoint to G.
Exercise 8.3. Show that the abelianization functor G 7→ Gab (see
Example 5.5) is a left adjoint to the inclusion functor Ab → Grp.
What are the unit and co-unit of this adjunction?
Exercise 8.4. Let R be the category with ob R = R and
(
{⋆} x ≤ y
HomR (x, y) =
∅ x>y
for all x, y ∈ R (see also Example 4.9). Let Z be the full subcategory
with ob Z = Z and let F : Z → R be the inclusion functor. Does this
functor have a left adjoint? And a right adjoint?
Exercise 8.5. Assume F : C1 → C2 is left adjoint to G : C2 → C1 and
F ′ : C2 → C3 is left adjoint to G′ : C3 → C2 . Show that F ′ F is left
adjoint to GG′ .
Exercise 8.6. Let {⋆} be the ‘one-point category’ consisting of a unique
object ⋆ and a unique morphism id⋆ . Let C be an arbitrary category.
When does the (unique) functor C → {⋆} have a left adjoint? And a
right adjoint?
Exercise 8.7. For a set I denote by Z[Xi | i ∈ I] the polynomial ring
in variables (Xi ) indexed by I. Elements of Z[Xi | i ∈ I] are finite
Z-linear combinations of monomials in finitely many of the variables.
Verify that I 7→ Z[Xi | i ∈ I] defines a functor Set → CRing which is
left adjoint to the forgetful functor CRing → Set.
Exercise 8.8. Let F : C → D be left adjoint to G : D → C. Let X be
a cofinal object in C, show that F X is cofinal in D. Similarly, if Y is
final in D, show that GY is final in C. (Compare with Exercise 6.13).
Exercise 8.9. Show that the forgetful functor Top⋆ → Top has a left
adjoint but not a right adjoint.
Exercise 8.10. Look up the definition of Stone-Čech compactification,
and verify that it gives a left adjoint to the inclusion functor from the
category of compact Hausdorff spaces to Top.
EXERCISES 85
Exercise 8.11. Let G be a group and let R be a ring. Show that

restriction defines a bijection between the sets of
(1) ring homomorphisms f : Z[G] → R
(2) group homomorphisms G → R× .
Interpret this bijection as an adjunction between functors between the
categories of groups and rings.
Exercise 8.12 (Triangle identities (⋆)). Let F : C → D be a left adjoint
to G : D → C, with unit η : idC → GF and co-unit ϵ : F G → idD . Show
that the diagrams
F ηX ηGY
FX F GF X GY GF GY
ϵF X GϵY
id id
FX GY
commute for every X in C and Y in D. Conversely, assume that F : C →
D and G : D → C are functors, and that η : idC → GF and ϵ : F G → idD
are morphisms of functors for which the above triangles commute. Show
that F and G form an adjoint pair of functors.
Exercise 8.13. A functor F : C op → Set is called representable if there
exists an object X in C with hX ∼= F . We say that F is represented by
X. Show that C has a final object if and only if the constant functor
C op → Set, T 7→ {⋆} is representable.
Exercise 8.14. Let M be a right R-module and N a left R-module.
Describe a functor F : Abop → Set which is represented by the abelian
group M ⊗R N . Use this to verify that if R is commutative, then
M ⊗R N ∼ = N ⊗R M .
Exercise 8.15. A functor F : C → Set is called co-representable if there
exists an object X in C with hX ∼
= F . Let f1 , . . . , fm ∈ Z[X1 , . . . , Xn ].
Show that the functor
CRing → Set, R 7→ {x ∈ Rn | f1 (x) = · · · = fm (x) = 0}
of Example 5.9 is co-representable.
Exercise 8.16 (⋆). Show that the functor
GLn : CRing → Set, R 7→ GLn (R)
of Exercise 5.3 is co-representable. (Hint: first show that the functor
GL1 : R 7→ R× is isomorphic to hR1 with R1 = Z[X, Y ]/(XY − 1).) Let
Rn be the commutative ring such that GLn ∼= hRn . By the co-Yoneda

lemma there is a unique ring homomorphism R1 → Rn inducing the
natural transformation det : GLn → GL1 . Describe this ring homomor-
phism explicitly.
Exercise 8.17 (⋆). For topological spaces A and Y define C(A, Y ) to
be the set of continuous maps from A to Y . For every compact K ⊂ A
and open U ⊂ Y let CK,U ⊂ C(A, Y ) be the subset consisting of those
f : A → Y with f (K) ⊂ U . We give C(A, Y ) the topology generated
by the subsets CK,U . (It is known as the compact-open topology).
Show that if A is compact and Hausdorff, then the functor
Top → Top : X 7→ X × A
(giving X × A the product topology) is left adjoint to the functor
Top → Top : Y 7→ C(A, Y ).
CHAPTER 9
Limits and colimits
1. Product and coproduct

Definition 9.1. Let C be a category, let X and Y be objects in C. A
product of X and Y consists of the data of
(1) an object P
(2) morphisms πX : P → X and πY : P → Y
such that for all objects T in C, and for all f : T → X and g : T → Y

there is a unique map h : T → P making the following diagram commute
f X
πX
h
T P
πY
g
Y
A product need not exist, but if it does, it is unique up to unique

isomorphism. In particular, we will refer to any product of X and Y as
the product of X and Y . We will usually denote it by X × Y , omitting
the morphisms πX and πY . The proof of uniqueness is formal, and
essentially identical to the proofs of Proposition 4.21 and the uniqueness
part of Theorem 7.3. We repeat the argument for the last time.
Proposition 9.2. Let (P, πX , πY ) and (P ′ , πX

′ , π ′ ) be products of X
Y
and Y . Then there exists a unique isomorphism h : P → P ′ such that
87
88 9. LIMITS AND COLIMITS
the diagram
X
′
πX
πX
h
P P′
πY ′
πY
Y
commutes.
Proof of Proposition 9.2. Since P ′ is a product, there exists
a unique h : P → P ′ making the diagram commute. Likewise, since
P is a product, there exists a unique h′ : P ′ → P making the diagram
commute. By the same reasoning, there are unique maps i and i′ making
the diagrams
X X
′
πX ′
πX
πX πX
i i′
P P P′ P′
πY πY ′
πY ′
πY
Y Y
commute. Combining both parts, we see that both i = idP and i = h′ h
make the first diagram commute, and that both i′ = idP ′ and i′ = hh′
make the second diagram commute. By unicity we have h′ h = idP and
hh′ = idP ′ , hence h is an isomorphism. □
Examples 9.3. In Set the product is the cartesian product, together
with the usual projections. In Top the product is the cartesian product
equipped the product topology, together with the usual projections.
The product of two rings R and S in Ring is the product ring R × S.
The product of two modules in R Mod is the cartesian product M × N .
The dual notion of a product is a coproduct (sometimes called sum).
Definition 9.4. Let C be a category, let X and Y be objects in C. A
coproduct of X and Y consists of the data of
(1) an object S
(2) morphisms ιX : X → S and ιY : Y → S
1. PRODUCT AND COPRODUCT 89
such that for all objects T in C, and for all f : X → T and g : Y → T

there is a unique map h : S → T making the diagram
X f
ιX
h
S T
ιY
g
Y
commute.
By the same argument as before, a coproduct, if it exists, is unique
up to isomorphism. We talk about the coproduct, and denote it with
X ⨿Y.
Example 9.5. Let X and Y be sets. Then the coproduct of X and Y in
Set is the disjoint union X ⨿ Y , together with the canonical inclusions
ιX : X → X ⨿ Y and ιY : Y → X ⨿ Y .
Similarly, the coproduct in Top of topological spaces X and Y is
the disjoint union X ⨿ Y , with the natural topology (in which X and
Y are both open and closed).
Example 9.6. Let R be a ring. Then the coproduct of R-modules M
and N is the direct sum M ⊕ N , and hence in this case the product
and the coproduct coincide (although the former is considered together
with the projections, and the latter with the inclusions).
More generally one can define products and coproducts of arbitrary
(possibly infinite) families of objects.
Definition
Q 9.7. The product of a familyQ (Xi )i∈I of objects in C is an
object i∈I Xi together with maps πn : i∈I Xi → Xn such that for
every T and forQevery collection of morphisms fi : T → Xi there is a
unique h : T → i∈I Xi such that fi = πi h for all i ∈ I.
Definition i∈I of objects in C is
` 9.8. The coproduct of a family (Xi )`
an object i∈I Xi together with maps ιn : Xn → i∈I Xi such that for
` for every collection of morphisms fi : Xi → T there is a
every T and
unique h : i∈I Xi → T such that fi = hιi for all i ∈ I.
Again, products and coproducts of arbitrary families need not exist,
but if they do they are unique up to unique isomorphism.
Example 9.9 (Product and coproduct of modules). Let R be a ring

and (Mi )i∈I a collection of R-modules. Then the product of (Mi ) is the
R-module Y
Mi = {(xi )i∈I | xi ∈ Mi }
i∈I
Q
together with the projection maps πn : i∈I Mi → Mn . The universal
property of products gives a natural (i.e. functorial) bijection
∼
Y Y
HomR (N, Mi ) −→ HomR (N, Mi ).
i∈I i∈I
L
The coproduct of (Mi )i isL the direct sum i∈I Mi together with the
inclusion maps ιn : Mn → i∈I Mi given by
(
x i=n
ιn (x)i =
0 i ̸= n
for all x ∈ Mn and i ∈ I. Indeed, the natural bijection
∼
M Y
HomR ( Mi , N ) −→ HomR (Mi , N )
i∈I i∈I
collection of morphisms (fi : Mi →

of Exercise 1.17 shows that for every L
N )i there is a unique morphism h : i∈I Mi → N with hιi = fi .
2. Pullback and pushout

Definition 9.10. Let C be a category and let f : X → Z and g : Y → Z
be morphisms in C. The pullback or fibered product of f and g consists
of
(1) an object P in C
(2) morphisms πX : P → X and πY : P → Y
such that f πX = gπY as maps P → Z, and such that for every object
T and for every pair of maps s : T → X, t : T → Y with f s = gt there
is a unique morphism h : T → P making the diagram
s X
πX
f
h
T P Z
g
πY
t Y
2. PULLBACK AND PUSHOUT 91
commute.
If it exists, the pullback is unique up to unique isomorphism. It is
usually denoted by P = X ×Z Y , but care should be taken since the
pullback does depend on the maps f and g. Alternatively, one says that
the square
πY
P Y
πX g
f
X Z
is cartesian if (P, πX , πY ) is the fibered product of f and g.
Example 9.11. In the category of sets, the fiber product of any pair
of maps f : X → Z and g : Y → Z exists. It is given by
X ×Z Y = {(x, y) ∈ X × Y | f (x) = g(y)},
together with the projection maps. Indeed, if s : T → X and t : T → Y
satisfy f s = gt, then the map
h : T → X ×Z Y, u 7→ (s(u), t(u)).
is the unique map making the diagram of the definition commute.
In the case that X = {⋆} and f the map ⋆ 7→ z, then we find
{⋆} ×Z Y = g −1 (z),
the fiber of Y over z.
In the case that X, Y are subsets of Z and f and g are the respective
inclusions we find X ×Z Y = X ∩ Y .
Example 9.12. Similarly, in Top we have
X ×Z Y = {(x, y) ∈ X × Y | f (x) = g(y)}
with the induced topology from the product topology on X × Y .
Example 9.13. Similarly, if f1 : M1 → N and f2 : M2 → N are R-
module homomorphisms, then
{(x1 , x2 ) ∈ M1 × M2 | f1 (x1 ) = f2 (x2 )}
is a sub-R-module of M1 × M2 , and one verifies that it is the pullback
of f1 and f2 .
Example 9.14. In particular, the kernel of a morphism f : M → N is

the pull-back of f and the map 0 → N . In other words, the square
ker f M
f
0 N
is cartesian in R Mod.
The dual notion of pullback/fibered product is pushout/fibered co-

product. The definition is obtained by reversing all the arrows.
Definition 9.15. Let C be a category and let f : Z → X and g : Z → Y

be morphisms in C. The pushout or fiber coproduct or fiber sum of f
and g consists of
(1) an object S in C
(2) morphisms ιX : X → S and ιY : Y → S
such that ιX f = ιY g as maps Z → S, and such that for every object T
and for every pair of maps s : X → T , t : Y → T with sf = tg there is
a unique morphism h : S → T making the diagram
X s
f
ιX
h
Z S T
ιY
g t
Y
commute.
If it exists, the pushout is unique up to unique isomorphism.
Example 9.16. Pushouts exist in Set, and can be constructed as fol-

lows. Let f : Z → X and g : Z → Y be functions. Consider the equiva-
lence relation ∼ on the disjoint union X ⨿ Y generated by f (z) ∼ g(z).
Let S be the quotient (X ⨿Y )/ ∼. Then we claim that S is the pushout
3. LIMITS AND COLIMITS 93
of f and g. Indeed, assume that we have a commutative diagram
X s
f
ιX
h
Z S T
ιY
g t
Y
Then there is only one possibility for the map h:

(
[x] 7→ s(x) x ∈ X
h : S → T,
[y] 7→ t(y) y ∈ Y
This is indeed well-defined, since for any z ∈ Z we have s(f (z)) =

t(g(z)).
Example 9.17 (gluing). The same construction as in Set defines a

pushout in Top, it suffices to put the quotient topology on X ⨿ Y / ∼.
This construction is particularly useful in topology when both f : Z →
X and g : Z → Y are injective. In this case it constructs a space S by
‘gluing’ X and Y along their common subspace Z.
Conversely, if S is a topological space and X and Y subspaces with
S = X ∪ Y , then S is the pushout of the inclusion maps X ∩ Y → X
and X ∩ Y → Y .
Example 9.18. The cokernel of a morphism f : M → N is the pushout

of f and the map M → 0. See also Exercise 9.2.
3. Limits and colimits

The above constructions are examples of two dual general classes
of categorical constructions called limits and colimits.
Let I be a small category (see Definition 4.11), let C be a category,
and let X : I → C be a functor. We will often denote the image of an
object i ∈ I by Xi .
It is useful to think of the functor X : I → C informally as a diagram

in C, indexed by I. For example, if I is the three-object category
1
φ
2
ψ
3
then a functor X : I → C can be thought of as a diagram
X1
X(φ)
X2
X(ψ)
X3
of objects and morphisms in C. See also Example 5.10.
Definition 9.19. The limit of a functor X : I → C consists of
(1) an object limI X in C
(2) for every object i in I a morphism πi : limI X → Xi
such that
(1) for every φ : i → j in I we have πj = X(φ) ◦ πi
(2) for every T in C and for every collection of morphisms ti : T →
Xi satisfying tj = X(φ) ◦ ti for all φ : i → j, there is a unique
h : T → limI X with ti = πi h for all i.
Of course, the limit of I → C, if it exists, is unique up to unique
isomorphism. The limit is sometimes written limi Xi or limI Xi , but
one should be careful to remember that it depends on the full functor
X, and not just on the objects (Xi )i∈ob I .
Examples 9.20. If I theQ discrete category on a set I (see ??), then
limI X is the product i∈I Xi (if it exists). If I is empty, then the
limit of the unique functor ∅ → C is the final object of C (if it exists).
Taking for I the category
1
2,
3
3. LIMITS AND COLIMITS 95
recovers the notion of fibered product.
There is no surprise in the definition of colimit:
Definition 9.21. The colimit of a functor X : I → C consists of

(1) an object colimI X in C,
(2) for every object i in I a morphism ιi : Xi → colimI X
such that
(1) for every φ : i → j in I we have ιj ◦ X(φ) = ιi ,
(2) for every T in C and for every collection of morphisms ti : Xi →
T satisfying tj ◦ X(φ) = ti for all φ : i → j, there is a unique
h : colimI X → T with ti = hιi for all i.
If it exists, it is unique up to unique isomorphism.
Examples 9.22.`If I the discrete category on a set I, then colimI X

is the coproduct i∈I Xi (if it exists). If I is empty, then the colimit of
the unique functor ∅ → C is the cofinal object of C (if it exists). Taking
for I the category
1
3
recovers the notion of pushout.
We end with an example of an infinite diagram in Set and its limit

and colimit.
Example 9.23. Let I be the category with ob I = N and such that for
all i < j we have Hom(i, j) = ∅, and for all i ≥ j we have Hom(i, j) =
{⋆}. In a picture:
0 ←− 1 ←− 2 ←− 3 ←− · · · .
Let S0 ⊃ S1 ⊃ S2 ⊃ · · · be a decreasing chain of sets. Then this defines
a functor
S : I → Set, i 7→ Si
which for i ≥ j maps the unique map i → j to the inclusion Si ,→ Sj .
To give a collection of maps ti : T → Si such that for all i ≥ j the

diagram
T
ti
tj
Sj Si
T
commutes, is the same as toTgive a map t : T → i Si . From this it
follows easily that limi Si = i Si .
For the colimit, note that a collection of maps ti : Si → T such that
for all i ≥ j the diagram
Sj Si
ti
tj
T
commutes is completely determined by t0 : S0 → T (since ti must be
the restriction of t0 to the subset Si ⊂ S0 ), and one easily verifies that
colimi Si = S0 .
4. Yoneda and limits and colimits of sets

The category Set has all limits and colimits.
Proposition 9.24. Let I be a small category and let X : I → Set, i 7→
Xi be a functor. Then limI Xi exists, and is given by
Y
(xi )i ∈ Xi | X(φ)(xi ) = xj for all φ : i → j
i∈ob I
together with the projection maps to the sets Xi .
Proof.Q The proof is a straightforward verification that the set L :=
{(xi )i ∈ i Xi | · · · } described in the proposition, with the projection
maps
πn : L → Xn , (xi )i 7→ xn
satisfies the definition of a limit. The first property holds by construc-
tion: for every x ∈ L and φ : i → j in I we have
πj (x) = xj = Xφ (xi ) = (Xφ ◦ πi )(x),
hence πj = Xφ ◦ πi . For the second property, assume that (T, (ti : T →
Xi )i ) satisfies tj = Xφ ◦ ti for every φ : i → j. Then the map
h : T → L, x 7→ (ti (x))i
4. YONEDA AND LIMITS AND COLIMITS OF SETS 97
Q
is well-defined (that is, h(x) lands in L ⊂ i Xi ), and clearly is the
unique map such that ti = πi h for every i. □
Proposition 9.25. Let I be a small category and `let X : I → Set, i 7→
Xi be a functor. Consider on the disjoint union i∈ob I Xi the equiv-
alence relation ∼ generated by xi ∼ X(φ)(xi ) for all φ : i → j and all
xi ∈ Xi . Then colimI Xi exists and is given by
a
colimI X = Xi ∼
i∈ob I
together with the compositions
a
ιi : XI ,−→ Xi −↠ colimI X.
i∈ob I
Proof. This is shown by a direct verification, somewhat similar to

the proof of Proposition 9.24. See also Exercise 9.17. □
Using the explicit descriptions of limits of sets, we can now rephrase
the universal property for limits and colimits in an arbitrary category:
Theorem 9.26. Let I be a small category and let X : I → C be a
functor. Let L be an object of C. Then limI X exists and is isomorphic
to L if and only if there is an isomorphism
∼
HomC (−, L) → limI HomC (−, Xi ),
of functors C op → Set.
Note that by Yoneda (Theorem 8.8) the above theorem completely
characterizes limI X.
Proof of Theorem 9.26. Let T be an object in C. By the ex-
plicit description of limit of sets in Proposition 9.24 we have

lim HomC (T, Xi ) = (ti : T → Xi )i | ∀φ : i → j, X(φ) ◦ ti = tj .
i∈I
By the universal property of the limit in C the map
∼
HomC (T, limi Xi ) −→ (ti : T → Xi )i | ∀φ : i → j, X(φ) ◦ ti = tj
given by h 7→ (πi ◦ h)i∈I is a bijection. The bijection is functorial in T ,
and hence defines an isomorphism of functors. Conversely, if there is a
functorial bijection
∼
HomC (T, L) −→ (ti : T → Xi )i | ∀φ : i → j, X(φ) ◦ ti = tj ,
then L satisfies the universal property of the limit. □
Theorem 9.27. Let I be a small category and let X : I → C be a func-

tor. Let C be an object of C. Then colimI X exists and is isomorphic
to C if and only if there is an isomorphism
HomC (C, −) = limI op HomC (Xi , −)
of functors C → Set. □
Note that by co-Yoneda (Theorem 8.12) the above theorem com-
pletely characterizes colimI X.
5. Adjoint functors and limits

Theorem 9.28 (Left adjoints commute with colimits). Let F : C → D
be a left adjoint to G : D → C. Let X : I → C be a functor, and suppose
that colimI X exists in C. Then colimI (F X) exists and
colimI (F X) ∼
= F (colimI X)
in D.
Informally, we say that ‘F commutes with colimits’.
Proof of Theorem 9.28. Using Theorem 9.27 and the definition
of adjoint functors we find for every T in D a chain of isomorphisms
HomD (F (colimI Xi ), T ) ∼
= HomC (colimI Xi , GT )
∼
= limI op HomC (Xi , GT )
∼
= limI op HomD (F Xi , T ),
functorial in T , and hence an isomorphism of functors
HomD (F (colimI Xi ), −) ∼
= limI op HomD (F Xi , −)
which by Theorem 9.27 shows that F (colimI Xi ) is the colimit of the
diagram I → D, i 7→ F Xi . □
The co-theorem states that right adjoints commute with limits:
Theorem 9.29 (Right adjoints commute with limits). Let F : C → D
be a left adjoint to G : D → C. Let X : I → D be a functor, and suppose
that limI X exists in D. Then limI (GX) exists and
limI (GX) ∼
= G(limI X)
in C. □
5. ADJOINT FUNCTORS AND LIMITS 99
Example 9.30. Let R be a ring and let A be an (S, R)-bimodule. By

Theorem 7.19 the functor
R Mod → S Mod, M 7→ A ⊗R M
is left adjoint (to the functor HomS (A, −)), and hence by Theorem 9.28
it commutes with colimits. In particular, it commutes with coproducts:
A ⊗R (⊕i∈I Mi ) = ⊕i∈I (A ⊗R Mi )
and with cokernels:
coker(id ⊗ f : A ⊗R M → A ⊗R N ) = A ⊗R coker(f : M → N ).
(Note that the latter gives a one-line proof of Proposition 7.17!).
There is a priori no reason for the functor to commute with limits,
and indeed in general A ⊗R − does not respect kernels (see Exercise
7.15) or (infinite) products (see Exercise 9.9).
Example 9.31. The forgetful functor R Mod → Set is right adjoint to
the free module functor (see Example 8.5) and hence by Theorem 9.29
it commutes with limits. For example, this implies that the underlying
set of a product of modules is the product of the underlying sets of the
modules.
On the other side of the adjunction, we see that the free mod-
ule functor I 7→ R(I) must commute with colimits. For example, this
contains the (trivial) statement that a basis for the direct sum of free
modules is given by the disjoint union of their bases, that is
R(I) ⊕ R(J) ∼
= R(I⨿J)
as R-modules.
Example 9.32. We have seen in Example 8.6 that the forgetful functor
Top → Set is both right adjoint (to the discrete topology functor) and
left adjoint (to the trivial topology functor). It therefore commutes with
both limits and colimits, and hence any limit or colimit of topological
spaces can be constructed by putting a suitable topology on the limit
or colimit of the underlying sets.
Exercises
Exercise 9.1. Let X and Y be topological spaces. Show that the
cartesian product X × Y with the product topology is the product of
X and Y in the category Top.
Exercise 9.2. Let f0 : M → N0 and f1 : M → N1 be morphisms in
R Mod. Show that their fibered coproduct exists. (Hint: construct the
fibered coproduct as a quotient module of N0 ⊕ N1 ).
Exercise 9.3. Does the category of pointed topological spaces Top⋆
have products and/or coproducts? And if so, what are they? Does
the forgetful functor Top⋆ → Top commute with products and/or
coproducts?
Exercise 9.4. Let C be a category and let X and Y be objects in C.
Find a category P in which the products of X and Y are precisely the
final objects. Conclude that Proposition 9.2 can be deduced directly
from Proposition 4.21.
Exercise 9.5. Show that the pushout of the inclusion map {0, 1} →
[0, 1] and the map {0, 1} → {⋆} in Top is the circle.
Exercise 9.6. Let C be a category. Consider the diagonal functor
∆: C → C × C
defined by X 7→ (X, X) and f 7→ (f, f ). When does ∆ have a left
adjoint? And a right adjoint?
Exercise 9.7 (Coproduct of commutative rings). Let R and S be com-
mutative rings.
(1) Show that there is a unique ring structure on the Z-module
R ⊗Z S such that
(r1 ⊗ s1 )(r2 ⊗ s2 ) = r1 r2 ⊗ s1 s2 .
(2) Show that R → R ⊗Z S, r 7→ r ⊗ 1 and S → R ⊗Z S, s 7→ 1 ⊗ s
are ring homomorphisms.
(3) Show that R ⊗Z S is the coproduct of R and S in CRing.
f
M P
g φ
ψ
Q N
EXERCISES 101
be a commutative square of R-modules.

(1) Show that M (with the maps f and g) is the pullback of P →
N and Q → N if and only if the sequence
(f,g) φ−ψ
0 −→ M −→ P ⊕ Q −→ N
is exact.
(2) Show that N (with the maps φ and ψ) is the pushout of M →
P and M → Q if and only if the sequence
(f,g) φ−ψ
M −→ P ⊕ Q −→ N −→ 0
is exact.
Exercise 9.9. Show that the element
1 ⊗ (1, 1, . . .)
Q
of the module Q ⊗Z n>0 Z/nZ is non-zero. Conclude that the func-
tor Q ⊗Z − from Ab to Ab does not commute with infinite products.
Exercise 9.10. Let X, Y : I → C be functors, and let η : X → Y be
a morphism of functors. Assume that limI X and limI Y exist. Show
that α induces a morphism limI X → limI Y in C. Formulate and prove
the analogous statement for colimits.
Exercise 9.11. Show that the lim and colim in Example 9.23 coincide
with those described by Propositions 9.24 and 9.25.
Exercise 9.12. Let S0 ⊂ S1 ⊂ S2 · · · be an infinite sequence of in-
clusions of sets. Show that the union ∪i Si is the colimit of a suitably
chosen diagram I → Set.
Exercise 9.13. Let K be a field. Let I be the category with ob I = N
and (
{⋆} j ≤ i
HomI (i, j) =
∅ j>i
Consider the diagram
R : I → CRing, i 7→ Ri := K[X]/(X i ),
(where for j ≤ i the map K[X]/(X i ) → K[X]/(X j ) is the quotient
map). Show that limI Ri exists in CRing, and is isomorphic to the
power series ring K[[X]].
Exercise 9.14. Let G be a group and let BG the category of Example

4.7. Let F : BG → Set be a functor.
(1) Show that F (⋆) is a set X equipped with an action of G.
(2) Show that lim F is the set of fixed points of the action.
(3) Show that colim F is the set of orbits of the action.
Exercise 9.15. Let X be a topological space, and let (Ui )i∈I be an
open cover. Let I be the category with ob I = I and
(
{⋆} Ui ⊂ Uj
Hom(i, j) =
∅ otherwise
Assume that for all i, j there is a k ∈ I such that Ui ∩ Uj = Uk . Show
that colimI Ui = X in Top.
Exercise 9.16. Let Z and R be the categories of Exercise 8.4. Let I
be the category with ob I = N and
(
{⋆} i ≤ j
HomI (i, j) =
∅ i>j
Verify that a functor I → R is the same as an increasing sequence of
real numbers
x0 ≤ x1 ≤ x2 ≤ · · ·
When does this functor have a limit? And a colimit? Verify directly if
they are preserved by the left and right adjoints of the inclusion functor
Z → R.
Exercise 9.17. Prove Proposition 9.25.
Exercise 9.18. Show that all limits and colimits exist in the category
Top. (Hint: see Propositions 9.24 and 9.25).
Exercise 9.19 (Arbitrary limits of modules). Let R be a ring, I a
small category, and M : I → R Mod, i 7→ Mi a functor. Show that
there is an exact sequence
Y Y
0 −→ lim M −→ Mi −→ Mj
I
i f : i→j
of R-modules (in particular the limit exists). Here the first product
ranges over all objects i in I, and the second ranges over all triples
(i, j, f ) with i and j objects in I and f : i → j a morphism in I. Verify
EXERCISES 103
by hand that your exact sequence is correct in the special cases where
the limit is a product or a pullback.
Exercise 9.20 (Arbitrary colimits of modules). Formulate and prove
an analogous statement for colimits of modules.
Exercise 9.21. Let R and S be rings, let A be an (R, S)-module and
let
0 −→ M1 −→ M2 −→ M3
be an exact sequence of R-modules. Use Theorem 9.29 to prove that
the induced sequence
0 −→ HomR (A, M1 ) −→ HomR (A, M2 ) −→ HomR (A, M3 )
is exact in S Mod. (See Exercise 2.8 for a more direct approach).
Exercise 9.22. Consider the functor F : Grp → Ab, G 7→ Gab . (See
Example 5.5 and Exercise 8.3).
(1) Let f : G1 → H and g : G2 → H be group homomorphisms.
Show that the pullback of f and g exists, and is isomorphic to
{(s, t) ∈ G × G | f (s) = g(t)}.
(2) Let f : Z/3Z → S3 be an injective homomorphism and let
g : {1} → S3 be the trivial homomorphism. Compute the pull-
back in Grp of f and g, as well as the pullback in Ab of F (f )
and F (g).
(3) Conclude that F does not have a left adjoint.
(4) Show that F does commute with finite products.
CHAPTER 10
Chain complexes
1. Chain complexes and their homology modules

Definition 10.1. Let R be a ring. A chain complex of R-modules
consists of
(1) for every i ∈ Z an R-module Mi ,
(2) for every i ∈ Z an R-linear map di : Mi → Mi−1
such that for every i the identity di ◦ di+1 = 0 holds.
We depict a chain complex as a diagram
2d 1 0 d d
· · · −→ M2 −→ M1 −→ M0 −→ M−1 −→ · · ·
and often denote the chain complex by M• .
Definition 10.2. A morphism of chain complexes M• → M•′ consists
of an R-module homomorphism fi : Mi → Mi′ for every i, such that the
resulting diagram
d2 d1
··· M2 M1 M0 ···
f2 f1 f0
d′2 d′1
··· M2′ M1′ M0′ ···
commutes. The resulting category of chain complexes of left R-modules

is denoted R Ch. The similarly-defined category of chain complexes of
right R-modules is denoted ChR .
The condition di ◦ di+1 = 0 in a chain complex
di+1 i d
· · · −→ Mi+1 −→ Mi −→ Mi−1 −→ · · ·
implies that im di+1 ⊂ ker di inside Mi .
105
106 10. CHAIN COMPLEXES
Definition 10.3. Let M• be a chain complex of R-modules and i an

integer. The i-th homology module of M• is the R-module
ker di
Hi (M• ) := .
im di+1
If f : M• → M•′ is a morphism of chain complexes, then fi : Mi →
Mi′ induces a morphism
Hi (f ) : Hi (M• ) → Hi (M•′ )
(see Exercise 10.1). We obtain for every i a functor
Hi : R Ch → R Mod.
These modules measure the failure of the sequence M• to be exact,
in the following sense.
Lemma 10.4. A chain complex
2 d 1 d
0 d
· · · −→ M2 −→ M1 −→ M0 −→ M−1 −→ · · ·
is exact if and only if Hi (M• ) = 0 for all i ∈ Z. □
2. The long exact sequence

Let M• , N• and P• be chain complexes of R-modules. We say that
a sequence of morphisms
α β
0 −→ M• −→ N• −→ P• −→ 0
is a short exact sequence in R Ch if it is termwise exact. In other words,
if in the commutative diagram
0 0 0
··· Mi+1 Mi Mi−1 ···

αi+1 αi βi−1
··· Ni+1 Ni Ni−1 ···

βi+1 βi βi−1
··· Pi+1 Pi Pi−1 ···
0 0 0
3. THE HOMOTOPY CATEGORY 107
all the columns are short exact sequences of R-modules.
Theorem 10.5. Let

α β
0 −→ M• −→ N• −→ P• −→ 0
be a short exact sequence of chain complexes of R-modules. Then there

exists an exact sequence
Hi+1 (β)
··· Hi+1 (N• ) Hi+1 (P• )
Hi (α) Hi (β)
Hi (M• ) Hi (N• ) Hi (P• )
Hi−1 (α)
Hi−1 (M• ) Hi−1 (N• ) ···
of R-modules.
The resulting exact sequence of homology modules is called the long

exact sequence of homology associated to the short exact sequence of
chain complexes.
Proof of Theorem 10.5. We omit the proof, which is a tedious

diagram chase in the style of the proof of the Snake Lemma (Theorem
2.2). In fact, the Snake Lemma is a special case of the theorem, obtained
by assuming the Mi , Ni and Pi vanish for i ̸∈ {0, 1}. □
3. The homotopy category

Definition 10.6. Let R be a ring and let f : M• → M•′ and g : M• →
M•′ be morphisms of chain complexes of R-modules. A homotopy from
′
f to g consists of a collection of R-linear maps hi : Mi → Mi+1 indexed
by i ∈ Z, such that for every i ∈ Z the identity
gi − fi = d′i+1 hi + hi−1 di
holds in HomR (Mi , Mi′ ). We say that f and g are homotopic, and write
f ∼ g, if there exists a homotopy from f to g.
It is convenient to keep track of these maps in a diagram:

d2 d1
··· M2 M1 M0 ···
f2 g2 h1 f1 g1 h0 f0 g0
··· M2′ M1′ M0′ ···

d′2 d′1
but note that in the definition of homotopy it is not required that the
hi ’s commute with the horizontal maps d in any way.
The equation defining homotopy can be remembered as
g − f = dh + hd,
omitting the indices which can be reinserted in only one meaningful
way.
Proposition 10.7. Let R be a ring and M• , M•′ chain complex of R-
modules. Then homotopy is an equivalence relation on the set HomR Ch (M• , M•′ ).
□
Proposition 10.8. Let R be a ring and f, g : M• → M•′ homotopic
morphisms of chain complexes of R-modules. Then
(1) for any morphism s : M•′ → N• in R Ch, the compositions sf
and sg are homotopic;
(2) for any morphism t : N• → M• in R Ch, the compositions f t
and gt are homotopic.
Proof. Let (hi )i be a homotopy from f to g. In the first case, one
verifies that (si+1 hi )i is a homotopy from sf to sg, and in the second
case, that (hi ti )i is a homotopy from f t to gt. □
Definition 10.9. The homotopy category of chain complexes of R-
modules, denoted R Ho, is the category with
(1) ob R Ho := ob R Ch
(2) HomR Ho (M• , N• ) := HomR Ch (M• , N• )/ ∼
where composition and identity maps are inherited from composition
and identity maps in R Ch.
Proposition 10.8 guarantees that composition in R Ho is well-defined.
Proposition 10.10. Let f, g : M• → M•′ be homotopic maps in ChR ,
and let i be an integer. Then Hi (f ) = Hi (g) as maps Hi (M• ) → Hi (M•′ ).
3. THE HOMOTOPY CATEGORY 109
Proof. By definition, an element of Hi (M• ) is a coset

x̄ := x + im(di+1 )
for some x ∈ ker(di ). We have
Hi (f )(x̄) − Hi (g)(x̄) = f (x) − g(x) + im(d′i+1 )
= d′i+1 (hi (x)) + hi−1 (di (x)) + im(d′i+1 ).
Since d′i+1 (hi (x)) ∈ im(d′i+1 ), the first term vanishes in Hi (M•′ ). Since
x ∈ ker(di ), also the second term vanishes, and Hi (f ) = Hi (g). □
A consequence of Proposition 10.10 is that the functors Hi on R Ch
induce functors
Hi : R Ho → R Mod, M• 7→ Hi (M• )
on the homotopy category of chain complexes.
Remark 10.11. As the terminology suggests, there is a close rela-
tionship with the notions of homotopic maps and homology groups in
algebraic topology. Homology groups of topological spaces are usually
defined in terms of the functor
C : Top → Z Ch, X 7→ C• (X)
that maps a space X to the chain complex of singular chains. One then
defines the i-th homology group of X by
Hi (X, Z) := Hi (C• (X))
and obtains functors Hi : Top → Ab.
If f, g : X → Y are homotopic continuous maps, then their induced
maps C• (X) → C• (Y ) are homotopic maps of chain complexes, and
hence their induced maps Hi (X, Z) → Hi (Y, Z) coincide.
Remark 10.12. Chain complexes form a mathematical context in
which three layers play a role: objects (chain complexes), morphisms
(morphisms of chain complexes), and maps between morphisms (ho-
motopies). See Remark 6.4 for two other such contexts: topological
spaces (spaces, continuous maps, homotopies), and categories (cate-
gories, functors, morphisms of functors).
Exercises
Exercise 10.1 (Functoriality of homology). Let f : M• → M•′ be a
morphism of chain complexes of R-modules.
(1) Show fi (ker di ) ⊂ ker d′i ;
(2) Show fi (im di+1 ) ⊂ im d′i+1 ;
(3) Conclude that fi induces an R-linear map Hi (M ) → Hi (M ′ ).
Exercise 10.2. Let
. . . → M 2 → M 1 → M0 → N → 0
be an exact sequence of R-modules. Show that the complex

M • = . . . → M 2 → M1 → M0 → 0 → · · ·
satisfies H0 (M• ) ∼
= N and Hn (M• ) = 0 for all n ̸= 0.
Exercise 10.3. Let R be a non-zero ring. Show that the two chain
complexes
··· 0 0 0 0 ···
id
··· 0 R R 0 ···
are isomorphic in R Ho.
Exercise 10.4. Let n > 1 and let f be the morphism in Z Ch given by
the diagram
n
··· 0 Z Z 0 ···
π
··· 0 0 Z/nZ 0 ···

with π the canonical map. Show that Hi (f ) is an isomorphism for all
i, but that f is not an isomorphism in Z Ho.
Exercise 10.5. Let R and S be rings. A functor F : R Mod → S Mod
is called additive if for all R-modules M , N the map
F : HomR (M, N ) → HomS (F M, F N )
is a homomorphism of abelian groups. Let A be an (R, S)-bimodule.
Show that the functor
R Mod → S Mod, M 7→ HomR (A, M )
is additive.
EXERCISES 111
Exercise 10.6. Show that an additive functor F : R Mod → S Mod

induces functors R Ch → S Ch and R Ho → S Ho.
Exercise 10.7. Let R and S be rings and A an (R, S)-bimodule. For
a chain complex M• in R Ch define a chain complex M•′ in ChS by
(1) Mi′ := HomR (M−i , A)
(2) di : Mi′ → Mi−1
′ the map induced from d1−i : M1−i → M−i
′
Verify that M• is a chain complex of right S-modules, and that the
operation M• 7→ M•′ defines functors R Chop → ChS and R Hoop →
HoS .
Exercise 10.8 (⋆). Let G be a group. Consider the abelian groups
n
Cn (G) := Z(G ) . To ease the notation, denote the basis vector e(g1 ,...,gn )
by [g1 , . . . , gn ] ∈ Cn (G). Consider the morphisms
dn : Cn (G) → Cn−1 (G)
defined on the basis of Cn (G) by
[g1 , . . . , gn ] 7→[g2 , g3 , . . . , gn ]
n−1
X
+ (−1)i [g1 , . . . , gi−1 , gi gi+1 , gi+2 , . . . , gn ]
i=1
+ (−1)n [g1 , . . . , gn−1 ].
Set Cn (G) = 0 for n < 0.
(1) Show that C• (G) is a chain complex of abelian groups.
The homology groups Hn (C• (G)) are called the homology groups of G,
and are denoted Hn (G, Z).
(2) Show that H0 (G, Z) ∼= Z;
(3) Show that H1 (G, Z) ∼= Gab .
Q
Exercise 10.9 (⋆). Let K be a field. Let i∈Z VecK be the category
whose objects are sequences (Vi )i∈Z of K-vector spaces, and whose mor-
phisms are sequences (fi )i∈Z of K-linear maps. Show that the functor
Y
HoK → VecK , V• 7→ (Hi (V• ))i
i∈Z
is an equivalence of categories.
CHAPTER 11
Free resolutions
1. Definition and existence

Definition 11.1. Let R be a ring and let M be an R-module. A free
resolution of M is an exact sequence of R-modules
d2 1 d π
· · · −→ F2 −→ F1 −→ F0 −→ M −→ 0
in which the modules Fi are free.
Example 11.2. If M itself is free, then the exact sequence
id
· · · −→ 0 −→ 0 −→ M −→ M −→ 0
is a free resolution (with F0 = M and Fi = 0 for i ̸= 0). We will usually
suppress leading zeroes from the notation, and simply write
0 −→ M −→ M −→ 0
for the above resolution.
Example 11.3. Let R be an integral domain. If I ⊂ R is a non-zero
principal ideal, then I is free of rank 1 as an R-module (see Exercise
1.6), hence the exact sequence
0 −→ I −→ R −→ R/I −→ 0
is a free resolution of the R-module R/I.
Example 11.4. If R is a principal ideal domain, and M a finitely
generated R-module, then we have seen in Corollary 3.2 that M has a
free resolution of the form
0 −→ F1 −→ F0 −→ M −→ 0
with F0 and F1 free R-modules of finite rank.
113
114 11. FREE RESOLUTIONS
Example 11.5. Let K be a field, and let R = K[X, Y ]. Then the

sequence
2 d 1 d π
0 −→ R −→ R ⊕ R −→ R −→ K[X, Y ]/(X, Y ) −→ 0
with d2 (1) = (Y, −X), d1 (1, 0) = X and d1 (0, 1) = Y , and π the
quotient map is a free resolution of the R-module K[X, Y ]/(X, Y ).
Every R-module has a free resolution, although in general the res-
olution need not be of finite length, and the free modules occurring in
it need not be of finite rank.
Proposition 11.6. Let R be a ring. Then every R-module M has a
free resolution.
Proof. Choose a generating set I of M , and let F0 = R(I) be the
free R-module with basis I. Then we have a natural surjective map
π : F0 → M , and hence an exact sequence
π
F0 −→ M −→ 0.
Now choose a generating set I1 of the R-module K0 := ker π ⊂ F0 and
let F1 = R(I0 ) . The natural map F1 → K0 ⊂ F0 extends the above to
an exact sequence
d1 π
F1 −→ F0 −→ M −→ 0.
Repeating this argument with K1 := ker d1 and so forth, we obtain an
exact sequence
d
2 1 d π
· · · −→ F2 −→ F1 −→ F0 −→ M −→ 0,
with the Fi free, as we had to show. □
Remark 11.7. From a free resolution
d
2 1 d π
· · · −→ F2 −→ F1 −→ F0 −→ M −→ 0
of M we obtain a chain complex F• of the form
d2 1 d
· · · −→ F2 −→ F1 −→ F0 −→ 0 −→ 0 −→ · · ·
by setting set Fi = 0 for i < 0. This chain complex satisfies
(
M i=0
Hi (F• ) ∼
=
0 i ̸= 0
Conversely, given a pair (F• , α) consisting of
(1) a chain complex F•
2. THE FREE RESOLUTION FUNCTOR 115
(2) an isomorphism α : H0 (F• ) → M

with Fi free for all i and zero for i < 0, and with Hi (F• ) = 0 for all
i ̸= 0, the sequence
π
· · · −→ F2 −→ F1 −→ F0 −→ M −→ 0,
where π is induced by α, is a free resolution of M .
It will often be convenient to think of a free resolution as a pair
(F• , α) consisting of a chain complex F• and an isomorphism α as above.
2. The free resolution functor

Theorem 11.8. Let M and M ′ be R-modules. Let F• and F•′ be free
resolutions of M and M ′ respectively. Let φ : M → M ′ be a morphism
of R-modules. Then there exists a morphism f : F• → F•′ such that
H0 (f ) = φ. Moreover, f is unique up to homotopy.
Proof. The existence of f amounts to the existence of R-linear
maps fi making the diagram (with exact rows)
d2 d1 π
··· F2 F1 F0 M 0
f2 f1 f0 φ
d′2 d′1 π′
··· F2′ F1′ F0′ M′ 0
commute. We will construct such a diagram inductively.

Let S0 ⊂ F0 be a basis of the free module F0 . For every s ∈ S0 ,
choose an s′ ∈ F0′ such that π ′ (s′ ) = φπ(s). Such s′ exists by the
surjectivity of π ′ . Now, since F0 is free with basis S0 , there exists a
unique R-linear map f0 : F0 → F0′ that maps every s ∈ S0 to its chosen
counterpart s′ ∈ F0′ . By construction the right-hand square commutes.
Next, let S1 ⊂ F1 be a basis of F1 . For every s ∈ S1 , we have
πd1 (s) = 0 by the exactness of the top-row, hence π ′ f0 d1 (s) = 0 by the
commutativity of the right-hand square, and hence f0 d1 (s) ∈ ker π ′ .
We conclude that there exists an element s′ ∈ F1′ with d′1 (s′ ) = f0 d1 (s).
Choosing such an s′ for every s ∈ S1 yields an R-linear map f1 : F1 → F1′
as above. Repeating the argument, we construct maps fi as required.
For the uniqueness assertion in the theorem, assume that we have
chain complex homomorphisms f : F• → F•′ and g : F• → F•′ with
H0 (f ) = H0 (g) = φ. Let δ := g − f . We need to show that there
are hi as in the diagram below

d2 d1 π
··· F2 F1 F0 M 0
δ2 h1 δ1 h0 δ0 0
··· F2′ F1′ F0′ M′ 0

d′2 d′1 π′
satisfying
δ0 = d′1 h0
δi = d′i+1 hi + hi−1 di (i ≥ 1)
As in the proof of the first part of the theorem, we construct these hi

inductively.
For the base step, choose a basis S0 of F0 and for every s ∈ S0
choose an s′ ∈ F1′ with d′1 s′ = δ0 s. Such s′ exists, since π ′ δ0 s = 0, and
the bottom row in the diagram is exact. Since F0 is free with basis S0 ,
there is a (unique) morphism h0 : F0 → F1′ with s → s′ for every s ∈ S0 ,
and we have δ0 = d′1 h0 by construction.
Next, let S1 be a basis of F1 , and choose for every s ∈ S1 an s′ ∈ F2′
with d′2 s′ = δ1 s − h0 d1 s. Such s′ exists, since
d′1 (δ1 s − h0 d1 s) = d′1 δ1 s − d′1 h0 d1 s = δ0 d1 s − δ0 d1 s = 0.
As before, the collection of s′ defines a map h1 : F1 → F2′ , and re-

peating the argument gives a collection of maps hi defining the desired
homotopy between f and g. □
Corollary 11.9. If F• and F•′ are free resolutions of M , then F• and

F•′ are isomorphic in R Ho.
Proof. The proof is ‘abstract nonsense’ and quite similar to the

argument that showed that final objects are unique up to unique iso-
morphism (see Proposition 4.21).
Take M ′ := M and apply Theorem 11.8 to id : M → M ′ and
id : M ′ → M to obtain morphisms f : F• → F•′ and g : F•′ → F• . Then
apply Theorem 11.8 again to idM and idM ′ to show that gf is homotopic
to idF• and f g is homotopic to idF•′ . This gives equalities f g = idF•
and gf = idF•′ in R Ho, which shows that f and g are mutually inverse
isomorphisms in R Ho. □
2. THE FREE RESOLUTION FUNCTOR 117
Example 11.10. The zero module M = {0} has the zero resolution,
but also the non-trivial free resolution
id π
0 −→ R −→ R −→ M −→ 0,
hence the corresponding complexes
··· 0 0 0 0 ···
··· 0 R R 0 ···
are isomorphic in R Ho. See also Exercise 10.3.
We can now summarise this section into one powerful theorem.
Theorem 11.11. There exists a functor
F : R Mod → R Ho, M 7→ F• (M )
and an isomorphism of functors
∼
α : H0 ◦ F −→ idR Mod
such that for every R-module M , the complex F• (M ) together with the
isomorphism αM forms a free resolution of M .
The proof goes directly against our basic principle that ‘construc-
tions depending on choices do not give rise to functors’.
Proof of Theorem 11.11. Using Proposition 11.6, choose for ev-
ery R-module M a free resolution
M π
· · · −→ F2 (M ) −→ F1 (M ) −→ F0 (M ) −→ M −→ 0.
This defines for every R-module M an object F• (M ) ∈ R Ho, and an
∼
isomorphism αM : H0 (F• (M )) → M (induced by πM ).
Now for every φ : M → N in R Mod, Theorem 11.8 gives a unique
morphism F• (φ) : F• (M ) → F• (N ) such that the square of R-modules
αM
H0 (F• (M )) M
H0 (F• (φ)) φ
αN
H0 (F• (N )) N
commutes. This provides the necessary data for a functor F• , and
immediately shows that α is an isomorphism of functors, provided that
the data defining F• indeed forms a functor.
For this, we need to check that F• respects identity and composi-

tion. But this follows quite formally from the uniqueness statement in
Theorem 11.8. Given an R-module M , both idF• (M ) and F• (id) induce
the identity on H0 (F• (M )) = M , so they must be homotopic and hence
they define the same morphism in R Ho. Similarly, given f : M → N
and g : N → P then both
F• (g)F• (f ) : F• (M ) → F• (N )
and
F• (gf ) : F• (M ) → F• (N )
induce the map gf : M → P on H0 , so they must be homotopic and
hence define the same morphism in R Ho. □
EXERCISES 119
Exercises
Exercise 11.1. Consider the ring R = Z[X]. Give a free resolution of
the R-module Z[X]/(X, 2).
Exercise 11.2. Let K be a field and consider the subring R = K[X 2 , X 3 ]
of the polynomial ring K[X]. Let M be the R-module R/(X 2 , X 3 ).
Find a free resolution of M .
Exercise 11.3. Let R be a commutative ring.
(1) Assume that r ∈ R is not a zero divisor in R. Show that R/rR
has a free resolution of the form
0 −→ R −→ R −→ R/(r) −→ 0.
(2) Let r, s ∈ R. Assume that r is not a zero divisor in R, and
that s̄ is not a zero divisor in R/rR. Show that R/(s, r) has a
free resolution of the form
0 −→ R −→ R2 −→ R −→ R/(r, s) −→ 0.
(3, ⋆) Try to formulate and prove an analogous statement for mod-
ules of the form R/(r, s, t), etcetera.
Exercise 11.4. Let n be a positive integer, and consider the ring R :=
Z[X]/(X n − 1). Let M be the quotient module R/(X − 1), π : R → M
the quotient map. Show that
β α β α π
· · · −→ R −→ R −→ R −→ R −→ M −→ 0,
with α(r) = (X − 1)r and β(r) = (X n−1 + · · · + X + 1)r, is a free
resolution of the R-module M .
Exercise 11.5. Let R = Z/4Z and let M be the R-module Z/2Z.
Find a free resolution of M .
Exercise 11.6. Let K be a field, n > 1 and let R be the matrix
ring Matn (K). Let M = K n be the left R-module of column vectors.
Show that M does not have a finite free resolution consisting of finitely
generated free R-modules.
Exercise 11.7. Let 0 → M1 → M2 → M3 → 0 be a short exact
sequence of R-modules. Show that there exist free R-modules F1 , F2 ,
F3 , and a commutative diagram

0 F1 F2 F3 0
0 M1 M2 M3 0
with exact rows and surjective vertical maps.
Exercise 11.8 (Free resolution of a short exact sequence). Let 0 →
M1 → M2 → M3 → 0 be a short exact sequence of R-modules. Show
that there exist free resolutions
· · · −→ Fi,2 −→ Fi,1 −→ Fi,0 −→ Mi −→ 0
for i = 1, 2, 3, and a short exact sequence
0 −→ F1,• −→ F2,• −→ F3,• −→ 0
of chain complexes compatible with the exact sequence 0 → M1 →
M2 → M3 → 0.
Exercise 11.9 (Uniqueness of free resolution functor). Let F and G
∼ ∼
be functors R Mod → R Ho. Let α : H0 ◦ F → id and β : H0 ◦ G → id
be isomorphisms. Assume that for every M the pairs (F (M ), α) and
(G(M ), β) are free resolutions of M . Show that the functors F and G
are isomorphic.
Exercise 11.10. Let M• be a chain complex of R-modules with Mi = 0
for all i ̸= 0, 1. Show that there exists a chain complex F• and a
morphism α : F• → M• such that
(1) Hi (α) is an isomorphism for all i, and
(2) Fi is free for all i.
Exercise 11.11 (⋆). Let M• be a chain complex of R-modules with
Mi = 0 for all i < 0. Show that there exists an F• and α as in Exercise
11.10.
CHAPTER 12
The Ext functors
1. The functors Extn

Let R be a ring. If M• is the complex of R-modules given by
· · · −→ Mi+1 −→ Mi −→ Mi−1 −→ · · ·
and if N is an R-module, then the induced sequence
· · · −→ HomR (Mi−1 , N ) −→ HomR (Mi , N ) −→ HomR (Mi+1 , N ) −→ · · ·
forms a complex of abelian groups, with the group HomR (M−i , N ) in
degree i. This determines a functor
op
R Ch × R Mod → Z Ch
and it induces a functor
op
R Ho × R Mod → Z Ho.
on the homotopy categories, by Exercise 10.7.
Definition 12.1. Let n be an integer. We define the functor
ExtnR (−, −) : R Modop × R Mod → Ab, (M, N ) 7→ ExtnR (M, N )
as the composition of the following functors:
(1) the free resolution functor of Theorem 11.11 (applied to the
first coordinate)
F• (−) × id : R Modop × R Mod −→ R Hoop × R Mod
(2) the functor induced by HomR as above
op
R Ho × R Mod −→ Z Ho
(3) the homology functor
H−n : Z Ho → Ab,
which is well-defined by Proposition 10.10.
121
122 12. THE Ext FUNCTORS
In other words, if
· · · −→ F1 −→ F0 −→ M −→ 0
is a free resolution of M , then the group ExtnR (M, N ) is defined as the
quotient group

ker Hom R (Fn , N ) → Hom R (F n+1 , N )
ExtnR (M, N ) = ,
im HomR (Fn−1 , N ) → HomR (Fn , N )
where the maps between the Hom groups are induced from the maps
in the free resolution, and where we set Fi = 0 for i < 0, as before.
Remark 12.2. A priori the functor ExtnR (−, −) depends on the choices
of free resolutions involved in F• (−), but different choices give rise to
isomorphic functors ExtnR (−, −).
Example 12.3. Let m be a positive integer. We compute the groups
ExtnZ (Z/mZ, Z) using the definition. As a first step we need to find a
free resolution of Z/mZ. The obvious choice
m
0 −→ Z −→ Z −→ Z/mZ −→ 0
leads to the complex
m
F• = · · · −→ 0 −→ Z −→ Z −→ 0 −→ 0 −→ · · ·
with F1 = F0 = Z. Note that Hom(Z, Z) = Z, so that applying
the contravariant additive functor HomZ (−, Z) to this complex gives a
complex of the form

H• = · · · −→ 0 −→ 0 −→ Z −→ Z −→ 0 −→ · · ·
with H0 = H−1 = Z. One checks that the map Z → Z is multiplication
by m. We find Hi (H• ) = 0 for all i ̸= −1, and H−1 (H• ) = Z/mZ.
From this we conclude
(
Z/mZ n = 1
ExtZ (Z/mZ, Z) ∼
n
=
0 n ̸= 1
By construction, the groups Extn (M, N ) are zero for n < 0.

Proposition 12.4. The functors Ext0R (−, −) and HomR (−, −) are iso-
morphic.
2. THE LONG EXACT SEQUENCE 123
Proof. Let
· · · −→ F1 −→ F0 −→ M −→ 0
be a free resolution of an R-module M . Then by Exercise 2.9 we get
an exact sequence
0 −→ Hom(M, N ) −→ Hom(F0 , N ) −→ Hom(F1 , N ).
Using the definition of Ext0 we find
Ext0 (M, N ) = H0 (Hom(F• , N )) = ker(Hom(F0 , N ) −→ Hom(F1 , N )),
which gives an isomorphism Ext0 (M, N ) = Hom(M, N ), functorial in
M. □
We will see that the module Ext1 (M, N ) is in bijection with iso-
morphism classes of short exact sequences
0 −→ N −→ E −→ M −→ 0
of R-modules. There is also an interpretation to the modules Extn (M, N )
with n > 1 in terms of exact sequences
0 −→ N −→ E1 −→ · · · −→ En −→ M −→ 0,
but the statement is more delicate.
2. The long exact sequence

Theorem 12.5. Let N be an R-module, and let
0 −→ M1 −→ M2 −→ M3 −→ 0
be a short exact sequence of R-modules. Then there is a natural exact
sequence
0 Hom(M3 , N ) Hom(M2 , N ) Hom(M1 , N )
Ext1 (M3 , N ) Ext1 (M2 , N ) Ext1 (M1 , N )
Ext2 (M3 , N ) Ext2 (M2 , N ) ···
of abelian groups.
Proof. Choose free resolutions of the Mi as in Exercise 11.8, so

that we have a short exact sequence of complexes
0 −→ F1,• −→ F2,• −→ F3,• −→ 0.
Denote by Hi,• the complex obtained by applying Hom(−, N ) to Fi,• ,
so we have
Hi,j = Hom(Fi,−j , N ).
For every i the exact sequence
0 −→ F1,i −→ F2,i −→ F3,i −→ 0
is split exact because F3,i is a free R-module (see Exercise 2.17), and
hence also the induced sequence
0 −→ H3,−i −→ H2,−i −→ H1,−i −→ 0
is exact (see Exercise 2.19). Applying Theorem 10.5 to the short exact
sequence of complexes of abelian groups
0 −→ H3,• −→ H2,• −→ H1,• −→ 0
we obtain a long exact sequence
··· Hi+1 (H2,• ) Hi+1 (H1,• )
Hi (H3,• ) Hi (H2,• ) Hi (H1,• )
Hi−1 (H3,• ) Hi−1 (H2,• ) ···
which, taking into account the vanishing of Exti for i < 0 and the fact
that Ext0 = Hom, gives precisely the exact sequence of the theorem.
One verifies that this sequence does not depend on the choice of free
resolutions. □
3. Ext1 and extensions

Definition 12.6. Let R be a ring and let M and N be R-modules. An
extension of M by N is a short exact sequence
0 −→ N −→ E −→ M −→ 0
3. Ext1 AND EXTENSIONS 125
of R-modules. Two such extensions are called equivalent if there exists

a commutative diagram
0 N E M 0
id id
0 N E′ M 0
of R-modules. We define the set extR (M, N ) to be the set of equivalence
classes of extensions of M by N .
Note that the morphism E → E ′ in the above diagram is auto-
matically an isomorphism, see Exercise 2.5. Warning: there can be
non-equivalent extensions with E ∼
= E ′ , see Exercise 12.11.
We will now construct a map
θ : extR (M, N ) → Ext1R (M, N )
and show that it is a bijection. To define the map, consider an element
e ∈ extR (M, N ), represented by a short exact sequence
0 −→ N −→ E −→ M −→ 0.
By Theorem 12.5 this induces a long exact sequence, of which a part
reads
· · · −→ HomR (N, N ) −→ Ext1R (M, N ) −→ · · ·
We define θ(e) ∈ Ext1R (M, N ) to be the image of idN under the above
map.
Theorem 12.7. The map
θ : extR (M, N ) → Ext1R (M, N )
is a bijection.
Proof. We construct map
ψ : Ext1R (M, N ) → extR (M, N )
but omit the tedious verification that it is a two-sided inverse.
Choose a free module F and a surjection F → M . Let K be the
kernel. Then we have a short exact sequence
γ
0 −→ K −→ F −→ M −→ 0.
The induced long exact sequence of Theorem 12.5 contains
δ
HomR (F, N ) −→ HomR (K, N ) −→ Ext1R (M, N ) −→ Ext1R (F, N ).
Since F is a free module, it has a ‘trivial’ one-term free resolution, and

one sees that ExtiR (F, N ) = 0 for all i > 0. It follows that the map δ is
surjective.
Let e ∈ Ext1R (M, N ). Choose an R-linear map f : K → N with
δ(f ) = e. In the commutative diagram
id
0 K K 0
(f,γ) γ
ι1 π2
0 N N ⊕F F 0
both rows are exact, and all the vertical maps are injective (since γ is
injective). It follows that there is an induced short exact sequence of
cokernels
0 −→ N −→ E −→ M −→ 0
with E = coker(K → N ⊕ F ). We define ψ(e) ∈ extR (M, N ) to be
this extension. One verifies that this is well-defined: if also f ′ : K → N
satisfies δ(f ) = e, leading to an extension E ′ , then f ′ = f + h for some
linear map h : F → N , and one shows that the isomorphism
F ⊕ N → F ⊕ N, (x, y) 7→ (x, y + hx)
induces an isomorphism E → E ′ of extensions. □
EXERCISES 127
Exercises
Exercise 12.1. Let R be a ring, F a free R-module, and N an R-
module. Show that ExtiR (F, N ) = 0 for all i ̸= 0.
Exercise 12.2. Let K be a field. Let M and N be K-modules. Show
that ExtiK (M, N ) = 0 for all i ̸= 0.
Exercise 12.3. Let n and m be positive integers. Compute for all i
the Z-modules
(1) ExtiZ (Z, Z/mZ);
(2) ExtiZ (Z/nZ, Z);
(3) ExtiZ (Z/nZ, Z/mZ).
Exercise 12.4. Let K be a field. Consider the ring R = K[X 2 , X 3 ] ⊂
K[X]. Let I = (X 2 , X 3 ). Show that Ext1R (I, R/I) is non-zero, and
conclude that I is not a principal ideal. Show that also Ext2R (R/I, R/I)
is non-zero.
Exercise 12.5. Let K be a field. Consider the ring R = K[X, Y ] and
the R-module M = K[X, Y ]/(X, Y ). Compute Ext2R (M, R). Conclude
that M does not have a free resolution of the form 0 → F1 → F0 →
M → 0.
Exercise 12.6. Let R be a principal ideal domain. Show that for every
i ≥ 2, for every finitely generated R-module M , and for every R-module
N we have ExtiR (M, N ) = 0.
Exercise 12.7. Let R be a ring, N an R-module and n ≥ 1. As-
sume that ExtnR (M, N ) = 0 for all R-modules M . Show that also
Extn+1
R (M, N ) = 0 for all R-modules M . (Hint: consider a short exact
sequence of the form 0 → K → F → M → 0 with F a free module).
Exercise 12.8. Let
0 −→ N −→ E −→ M −→ 0
be a short exact sequence of R-modules. Let φ : N → N ′ be a morphism
of R-modules, and let E ′ be the pushout of N → E and φ : N → N ′ .
Show that there is a short exact sequence
0 −→ N ′ −→ E ′ −→ M −→ 0
of R-modules. (Hint, see Exercise 9.8).
Exercise 12.9. Let

0 −→ N −→ E −→ M −→ 0
be a short exact sequence of R-modules. Let φ : M ′ → M be a mor-
phism of R-modules, and let E ′ be the pullback of E → M and
φ : M ′ → M . Show that there is a short exact sequence
0 −→ N −→ E ′ −→ M ′ −→ 0
of R-modules. (Hint, see Exercise 9.8).
Exercise 12.10. Let R be a ring. Show directly (without using the
relation with Ext1 ) that
op
R Mod × R Mod → Set, (M, N ) 7→ extR (M, N )
is a functor. (In particular: explain what the functor does on the level
of morphisms).
Exercise 12.11. Describe explicitly the n elements of the set extZ (Z/nZ, Z)
for:
(1) n a prime number;
(2) n = pq with p and q distinct primes;
(3) n = 4.
Exercise 12.12. For i = 1, 2 let
i α βi
0 −→ N −→ Ei −→ M −→ 0
be a short exact sequence of R-modules. Define an R-module E as the
quotient

ker β1 − β2 : E1 ⊕ E2 → M
E := .
im (α1 , α2 ) : N → E1 ⊕ E2
Show that there is a natural short exact sequence
0 −→ N −→ E −→ M −→ 0
of R-modules.
Exercise 12.13. Let K be a field and let λ1 , λ2 ∈ K. Let Mi :=
K[X]/(X − λi ). Show that Ext1K[X] (M2 , M1 ) = 0 if λ1 =
̸ λ2 . Compute
1
ExtK[X] (M2 , M1 ) when λ2 = λ1 .
EXERCISES 129
Exercise 12.14. Let K be a field, λ1 , λ2 , µ ∈ K. Consider the matrix

λ1 µ
A :=
0 λ2
and let E be the K[X]-module given by E = K 2 on which X acts by
A. Show that E sits in a short exact sequence of K[X]-modules
0 −→ M1 −→ E −→ M2 −→ 0
where Mi = K with X acting as λi . Show that this sequence splits if
and only if λ2 ̸= λ1 or µ = 0. Relate this to the previous exercise.
Exercise 12.15. Let n be a positive integer, and consider the ring
R := Z[X]/(X n − 1). Let M be the quotient module R/(X − 1).
Compute the R-modules ExtiR (M, M ) and ExtiR (M, R). (Hint: use the
free resolution from Exercise 11.4).
Bibliography
[1] Atiyah, M.F. & MacDonald, I.G. Introduction To Commutative Algebra,

Westview Press, 1969.
[2] Lang, S. — Algebra, Graduate Texts in Mathematics 211, Springer, 2002.
[3] Moerdijk, I. — Notes on Homological Algebra, Lecture notes, 2008.
[4] Stevenhagen, P. — Algebra 2, http://websites.math.leidenuniv.nl/
algebra/algebra2.pdf.
[5] Leinster, T. — Basic Category Theory, Cambridge Studies in Advanced
Mathematics, Cambridge University Press, 2014.
131

Lenny Taelman - Categories and Modules

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Lenny Taelman - Categories and Modules

Uploaded by

Document Information

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

Lenny Taelman - Categories and Modules

Uploaded by

Copyright:

Available Formats

Modules and Categories∗

∗Preliminary version, August 17, 2023

Chapter 1. Modules over a ring 9

Chapter 2. Exact sequences 21

Chapter 3. Finitely generated modules over a PID 31

Chapter 6. Morphisms of functors 57

Chapter 7. Tensor product 65

Chapter 8. Adjoint functors 77

Chapter 9. Limits and colimits 87

Chapter 10. Chain complexes 105

Chapter 11. Free resolutions 113

These are course notes for a one-semester third-year course on cat-

Modules over a ring

1. Left and right modules

If R is commutative, then the difference between a left and a right

For n = 0 we obtain the zero module R0 = 0 and for n = 1 we obtain

where αi denotes the iterated composition α ◦ · · · ◦ α. In particular, V

We will see in Chapter 3 that the ‘Jordan normal form’ of complex

with ag = 0 for all but finitely many g (this is automatic if G is a

3. Homomorphisms, submodules and quotient modules

A submodule of an R-module M is a subgroup N ⊂ M such that for

4. Products, direct sums and free modules

If we have Mi = M for all i, then we write M I :=

If I is a finite set of cardinality n, then we have M I = M (I) ∼

For an index i ∈ I we denote by ei ∈ R(I) the element ei := ιi (1). One

Proposition 1.13. Let R be a ring, M an R-module, and (xi )i∈I a

Example 1.16. Let U be an open subset of Rn . Then the space of 1-

Exercise 1.18 (⋆). Let K be a field. Let R be the set of ∞ by ∞

Exact sequences form a useful and extensively used notational tool

We will often interpret the injective map f as the inclusion of a sub-

2. The Five Lemma and the Snake Lemma

x ∈ M3 is the image of some element z̃ in M1 , and hence by exactness

coker f1 coker f2 coker f3 0

So f2 (y) ∈ ker β ′ = im α′ , hence there is a z ∈ N1 with α′ (z) = f2 (y).

3. Split short exact sequences

Exercise 2.7. Let R be a ring and let

Exercise 2.11. Let R be a commutative ring and let

Show that N := { x0 | x ∈ K} ⊂ M is a sub-R-module, and that the

Finitely generated modules over a PID

2. Review of principal ideal domains

ring of Gaussian integers Z[i]. Also a field K is a PID, but of a rather

3. Free modules of finite rank over a PID

Proof of Proposition 3.1. We use induction on n. For n = 0

4. Structure of finitely generated modules over a PID

Note that the coordinates xi of x ∈ M ∼ = Rn depend on the choice

Proof of Theorem 3.3. By Corollary 3.2 every finitely gener-

5. Application to Jordan normal form

Note that K[X] is infinite-dimensional as a K-vector space, so we nec-

Exercise 3.8. Let R be a PID and let p ∈ R be irreducible. Let

Exercise 3.11 (⋆). Let

(2) α(Wi ) ⊂ Wi for every i,

HomC (Y, Z) × HomC (X, Y ) → HomC (X, Z), (g, f ) 7→ gf,

called composition, subject to the conditions

We often write f : X → Y instead of f ∈ HomC (X, Y ), and think

Remark 4.2. The use of the word ‘class’ is to avoid set-theoretical

Example 4.6 (One arrow). Consider the category C consisting of pre-

Example 4.7 (A group as a category). Let G be a group. Con-

Definition 4.11. A category is called locally small if for every X and