M4P63

M4P63 Algebra IV
Lectured by Dr John Britnell

Typed by David Kurniadi Angdinata
Spring 2020
Syllabus
1
M4P63 Algebra IV Contents
Contents
1 Modules over a ring 3
1.1 Modules over rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Homomorphisms and submodules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Direct products and direct sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Exact sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Projective and injective modules 6

2.1 Projective modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Free modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Injective modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3 Hom and tensor products 10

3.1 Hom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 The snake lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3 Tensor products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.4 Flat modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4 Modules over a PID 18

4.1 Free and projective modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2 Injective and divisible modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3 Flat and torsion-free modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.4 Modules over PIDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5 Projective and injective resolutions 23

5.1 Existence of projective resolutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.2 Existence of injective resolutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.3 Uniqueness of projective resolutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5.4 Uniqueness of injective resolutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
6 Chain and cochain complexes 29

6.1 Chain complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
6.2 Homology groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
6.3 The long exact sequence in homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
7 Derived functors 31
7.1 Covariant and contravariant functors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
7.2 Left derived functors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.3 The long exact sequence of left derived functors . . . . . . . . . . . . . . . . . . . . . . . . . . 34
7.4 General derived functors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
8 Tor and Ext 36

8.1 Balancing theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
8.2 Tor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
8.3 Ext . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2
M4P63 Algebra IV 1 Modules over a ring
1 Modules over a ring

Lecture 1
Let R be an associative ring with unity, that is an abelian group written additively with a multiplication Friday
which is associative but not necessarily commutative, with an identity 1 and distributive laws a (b + c) = 10/01/20
ab + ac and (a + b) c = ac + bc. Then
R∗ = {r ∈ R | ∃s ∈ R, rs = 1 = sr}
is the unit group of R. If R∗ = R \ {0} then R is a division ring, or a skew field. In the case that R is
commutative, R is a field.
Example.
• Fields C, R, Q, and Fq , the field with q = pa elements with p a prime and a ≥ 1.
• Skew fields H = {a + bi + cj + dk | a, b, c, d ∈ R} where i2 = j 2 = k 2 = ijk = −1.
• Other rings are polynomial rings k [x] for k a field, more generally k [x1 , . . . , xp ], and Matn k, the n × n
matrices with entries from k, a field.
1.1 Modules over rings

Definition 1.1. Let R be a ring. A left R-module is an abelian group M , written additively, together
with a function ∗ : R × M → M satisfying
r∗(m1 + m2 ) = r∗m1 +r∗m2 , (r1 + r2 )∗m = r1 ∗m+r2 ∗m, (r1 r2 )∗m = r1 ∗(r2 ∗ m) , 1∗m = m.
We write rm for r ∗ m.
Example.
• R is itself a left R-module, with ∗ as ring multiplication. More generally, let I be a left ideal of R, so I
is an additive subgroup, and rI ≤ I for all r ∈ R. Then I is an R-module with ∗ as ring multiplication.
• Let k be a field. Then any vector space over k is a k-module, and vice versa.
• Any abelian group is a Z-module, with ∗ defined by na = a + · · · + a for n ∈ Z+ and a ∈ A, and
(−n) a = − (na).
• Let k be a field. Let k n be column vectors. Then k n is a left Matn k-module, with ∗ as the usual
matrix-vector multiplication.
∗ ∗
• Let M ∈ Matn k.
Then we can define a left k [x]-module structure on k by letting x act as M on k .
So x2 + 3x − 2 ∗ v = M 2 v + 3M v − 2v.
• Let G be a group. Any representation of G over the field k is a left module for k [G], the group
algebra, a vector space over k with elements of G as a basis, with multiplication derived from that of
G.
Definition 1.2. A right R-module is defined similarly, with the R-multiplication on the right, so M an
abelian group under +, and a map M × R → M satisfying
(m1 + m2 )∗r = m1 ∗r+m2 ∗r, m∗(r1 + r2 ) = m∗r1 +m∗r2 , m∗(r1 r2 ) = (m ∗ r1 )∗r2 , m∗1 = m.
Left and right modules are not quite the same. If we amend this definition by putting the ring multiplication
on the left, the third axiom becomes (r1 r2 ) m = r2 (r1 m). But in a left module, we have (r1 r2 ) m = r1 (r2 m).
Definition 1.3. Let R be a ring. The opposite ring Rop is R with a redefined multiplication r∗Rop s = s∗R r.
It is easy to see that a left R-module is the same as a right Rop -module and vice versa. If R is commutative
then R = Rop .
Exercise. Show that Matn k ∼ = Matn k op .
Except where otherwise stated, R-modules are assumed to be left R-modules.
3
1.2 Homomorphisms and submodules

Definition 1.4. Let M1 and M2 be R-modules. A map f : M1 → M2 is an R-module homomorphism if
• f is a group homomorphism, with respect to the + operation, and
• f (rm) = rf (m), for r ∈ R and m ∈ M .
If f is bijective, then it is an R-module isomorphism.
Definition 1.5. An additive subgroup L ≤ M is a submodule if rL ≤ L for r ∈ R. In this case we
automatically get an R-module structure on the quotient M/L with multiplication given by r (m + L) =
rm + L.
Theorem 1.6 (First isomorphism theorem). Let f : M1 → M2 be an R-module homomorphism. Then
Im f ≤ M2 , Ker f ≤ M1 , Im f ∼
= M/ Ker f.
The other isomorphism theorems have R-module versions too.
1.3 Direct products and direct sums

Lecture 2
Let S be a set. We have a collection of R-modules (Ms )S indexed by S. Monday
Definition 1.7. The direct product is 13/01/20
Y
Ms = {(ms )S | ms ∈ Ms } ,
s∈S
with coordinate-wise addition and R-multiplication, so

(ms )S + (ns )S = (ms + ns )S , r (ms )S = (rms )S .
If Ms = M for all s ∈ S, then we write M S for s∈S Ms .
Q
Definition 1.8. The direct sum is

M Y
Ms = {(ms )S | all but finitely many coordinates ms are zero} ≤ Ms .
s∈S s∈S
If S is finite then the direct product and the direct sum are equal.
L
Example. Let M = Z2 , as a Z-module, and let S = N. Then
Q s∈N Z2 is a countable Z-module but
N
s∈N Z2 = Z2 is uncountable.
When |S| = 2, generally we write M1 ⊕ M2 for the direct sum or product. There are natural injective maps
ιA : A −→ A⊕B ιB : B −→ A⊕B
, ,
a 7−→ (a, 0) b 7−→ (0, b)
and surjective maps
πA : A ⊕ B −→ A πB : A ⊕ B −→ B
, .
(a, b) 7−→ a (a, b) 7−→ b
1.4 Exact sequences

Definition 1.9. Suppose we have a sequence of R-modules
fn−1 fn fn+1
. . . −−−→ Mn −→ Mn+1 −−−→ . . . ,
with maps fn : Mn → Mn+1 . Say the sequence is exact at Mn if
Im fn−1 = Ker fn .
The sequence is exact if it is exact everywhere. A short exact sequence is an exact sequence
α β
0→A−
→B−
→ C → 0.
Note. α is injective and β is surjective.
4
The first isomorphism theorem implies that B/ Im α ∼

= C, where Im α ∼
= A. An easy case is
B∼= A ⊕ C,
with Im α = Im ιA = A ⊕ 0 and Im β = Im πβ = C. We say that the short exact sequence splits in this case.
Example. A non-split short exact sequence of Z-modules, or abelian groups, is
0 → Z/2Z → Z/4Z → Z/2Z → 0.
Proposition 1.10. A short exact sequence
α β
0→A−
→B−
→C→0
is split if and only if there exists an R-module homomorphism σ : C → B such that β ◦ σ = idC .
Such a σ is called a section of β.
Proof.
=⇒ Suppose that the short exact sequence is split. So assume B = A ⊕ C, with α = ιA and β = πC . Now
ιC is a section for β.
∼
⇐= For the converse, suppose that σ is a section for β. We want f : A ⊕ C −
→ B such that f ◦ ιA = α and
β ◦ f = πC , so
A⊕C
ιA πC
0 A f C 0 .
α β
B
Define
f : A × C −→ B
.
(a, c) 7−→ α (a) + σ (c)
Need to check the following.
– f is an R-module homomorphism. 1
– f is injective. Suppose f (a, c) = 0. Then α (a) + σ (c) = 0. Now α (a) ∈ Im α = Ker β, so
β (α (a) + σ (c)) = β (σ (c)) = c. Since α (a) + σ (c) = 0, we have c = 0. Hence α (a) = 0, and so
a = 0 since α is injective. We have shown that f is injective.
– f is surjective. Let b ∈ B. Let c = β (b). We have (β ◦ σ) (c) = c = β (b), so b − σ (c) ∈ Ker β =
Im α. So there exists a ∈ A with α (a) = b − σ (c). Then b = α (a) + σ (c) = f (a, c).
– f ◦ ιA = α and β ◦ f = πC . Immediate from the construction of f .
Proposition 1.11. The short exact sequence

α β
0→A−
→B−
→C→0
is split if and only if there exists ρ : B → A such that ρ ◦ α = idA .
Such a ρ is a retraction of α.
Proof.
=⇒ Once again, if the short exact sequence is split then the existence of ρ is clear.
∼
⇐= Suppose that ρ is a retraction for α. We define f : B −
→ A ⊕ C such that f ◦ α = ιA and πC ◦ f = β.
Do this by
g : B −→ A ⊕ C
.
b 7−→ (ρ (a) , β (c))
1 Exercise
5
M4P63 Algebra IV 2 Projective and injective modules
2 Projective and injective modules

2.1 Projective modules
Lecture 3
Definition 2.1. An R-module M is projective if any surjective map β : B → M has a section. In other Tuesday
words, any short exact sequence 14/01/20
0→A→B→M →0
splits.
Example. The R-module R is projective. Let
β
0→A→B−
→R→0
be a short exact sequence. Since β is surjective, there exists b ∈ B such that β (b) = 1. Now for all r ∈ R,
β (rb) = r. Now define
σ : R −→ B
.
r 7−→ rb
Then σ is a section for β.
Proposition 2.2. An R-module M is projective if and only if whenever β : B → C is surjective, and
f : M → C, there exists g : M → B such that f = β ◦ g, so
M
g .
f
0 A B β
C 0
Such a g is called a lift of f .

Proof.
⇐= Suppose that whenever β : B → C is surjective and f : M → C then there exists g : M → B with
f = β ◦ g. Suppose β : B → M is a surjective map. Define f : M → M to be idM . Then there exists
g : M → B such that f = β ◦ g, so idM = β ◦ g. So g is a section for β, and so M is projective.
=⇒ For the converse, suppose β : B → C is surjective, and f : M → C. We construct a module X to
complete a commuting square

X M
δ f .
B β
C
Let X be the submodule of B ⊕ M defined by
X = {(b, m) | β (b) = f (m)} .
The maps δ and are just πB and πM respectively, in their restrictions to X. It is clear that X ≤ B⊕M ,
and that the square above commutes. Now suppose that M is projective. Since β is surjective, we see
that for all m ∈ M there exists b ∈ B with β (b) = f (m). It follows that : X → M is surjective. So
has a section σ : M → X. Define g = δ ◦ σ : M → B, so

X M
σ
δ f .
g
B β
C
Since β ◦ δ = f ◦ , we have
(β ◦ g) (m) = (β ◦ δ ◦ σ) (m) = (f ◦ ◦ σ) (m) = (f ◦ idM ) (m) = f (m) , m ∈ M.
So β ◦ g = f as required.
6
Such an X is the pullback of β and f , and there is a short exact sequence

0 → A → X → M → 0.
2.2 Free modules

Definition 2.3. An R-module M is free if M is a direct sum of copies of R, so
M
M= R.
s∈S
A basis for a module M is a set T of elements such that every element m ∈ M has a unique expression as
m
X
m= ri ti , ri ∈ R, ti ∈ T.
i=1
L
If M = s∈S R, then M has a basis consisting of elements with exactly one coordinate L one, and the rest
zero. On the other hand, if M has a basis T then it is straightforward to show that M ∼
= t∈T R.
Proposition 2.4. Let F be a free R-module with basis T . Let M be some R-module, and let ψ : T → M be
a set map. Then ψ extends uniquely to an R-module homomorphism ψ : F → M .
P
Proof. Each element of F has a unique expression as i ri ti for ri ∈ R and ti ∈ T . Now define
ψ : F −→ M
P P .
i ti
ri 7−→ i ri ψ (ti )
It is easy to check that this respects + and R-multiplication.

Proposition 2.5. A module M is projective if and only if there exists N such that M ⊕ N is free, so
projective modules are direct summands of free modules.
Proof.
=⇒ Suppose M is projective. Let F be the free module with basis {bm | m ∈ M }. Now the map bm 7→ m
extends to an R-module homomorphism F → M , which is clearly surjective. Then if K = Ker ψ, we
have a short exact sequence
ψ
0→K→F − → M → 0.
∼ K ⊕M .
Since M is projective, there is a section σ for ψ, and so the short exact sequence splits, and F = Lecture 4
Friday
⇐= Suppose that M ⊕ N = F , a free module with basis T . Suppose β : B → C is surjective, and that 17/01/20
f : M → C. Note that f ◦ πM : F → C. For each t ∈ T , let bt ∈ B be such that β (bt ) = (f ◦ πM ) (t).
The set map
T −→ B
t 7−→ bt
extends to a homomorphism gb : F → B. Now define g : M → B by g = P gb ◦ ιM . We need to show
f = β ◦ g. Take m ∈ M . P Then ιM (m) = (m, 0) ∈ F can be written as i ri ti , where ti ∈ T and
ri ∈ R. Applying πM , m = i ri mti . Then
!
X X
g ◦ ιM ) (m) = gb
g (m) = (b ri ti = ri bti .
i i
So ! !
X X X X
(β ◦ g) (m) = β ri bti = ri β (bti ) = ri f (mti ) = f ri mti = f (m) .
i i i i
Hence β ◦ g = f . So M is projective.
7
2.3 Injective modules

Definition 2.6. Let M be an R-module. Then M is injective if whenever α : M → B is an injective map,
it has a retraction ρ : B → M , so ρ ◦ α = idM . Equivalently, every short exact sequence
0→M →B→C→0
splits.
Example. Let k be a field. Then k-modules are vector spaces. Every k-module is injective. Suppose M
and N are k-vector spaces and α : M → N is a injective map. Then Im α is a submodule, or subspace, of
N . Take a basis for Im α, and extend to a basis for N . The basis vectors not in Im α form a basis for a
complementary subspace U , so N = Im α ⊕U . Now πIm α is surjective, and α : M → Im α is an isomorphism.
This gives a retraction N → M .
If R is a general ring, the module R need not be injective.
Example. Let R = Z. Then R-modules are abelian groups. There exists an injective α : Z → Q. But Z is
not a quotient of Q, 2 so no retraction exists for α.
Proposition 2.7. An R-module M is injective if and only if whenever α : A → B is injective, and f : A →
M , there exists g : B → M such that f = g ◦ α.
Proof.
⇐= Suppose that whenever α : A → B is injective, and f : A → M , there exists g : B → M such that
f = g ◦ α. Suppose that α : M → B is injective. We have a map M → M , namely idM . There exists
g : B → M such that idM = g ◦ α. So g is a retraction for α, and so M is injective.
=⇒ For the converse, suppose α : A → B is injective, and M is an injective module, with f : A → M . We
define a module Y completing a square
α
A B
f δ ,
M Y
with ◦ f = δ ◦ α. Let Y be a quotient of B ⊕ M , by the kernel
K = {(α (a) , −f (a)) | a ∈ A} .
Let γ : B ⊕M → (B ⊕ M ) /K be the canonical quotient map. Then we define δ = γ ◦ιB and = γ ◦ιM .
By construction, we have
( ◦ f ) (a) = (γ ◦ ιM ◦ f ) (a) = γ (0, f (a)) = (0, f (a)) + K
= (α (a) , 0) + K = γ (α (a) , 0) = (γ ◦ ιB ◦ α) (a) = (δ ◦ α) (a) .
Hence ◦ f = δ ◦ α. Claim that is injective. Suppose (m) = 0. Then ιM (m) ∈ K, so (0, m) =
(α (a) , −f (a)) for some a ∈ A. But α (a) = 0 implies that a = 0, and so m = −f (0) = 0. Since M is
injective, has a retraction ρ : Y → M . Define g : B → M by g = ρ ◦ δ, so
α
A B
g ,
f δ
ρ
M Y

We know that ( ◦ f ) (a) = (δ ◦ α) (a) for all a ∈ A. So

f (a) = (idM ◦f ) (a) = (ρ ◦ ◦ f ) (a) = (ρ ◦ δ ◦ α) (a) = (g ◦ α) (a) ,
so f = g ◦ α as required.
2 Exercise
8
We know that projectives are direct summands of free modules. We might hope for a dual version of this
for injective modules. But there is no straightforward way of doing this. Lecture 5
Proposition 2.8 (Baer’s criterion for injectivity). Let M be an R-module. Then M is injective if and only Monday
if every R-module map f : I → M , where I is a left ideal of R, has the form f (x) = xm for some m ∈ M . 20/01/20
Equivalently, every map I → M extends to a map R → M .
Why are these two conditions equivalent? If f (x) = xm for x ∈ I, then we can extend f to R by f (r) = rm.
Conversely, suppose that f : I → M extends to f + : R → M . Let m = f + (1). Then for all r ∈ R,
f + (r) = rm, and so f (x) = xm for x ∈ I. The proof requires Zorn’s lemma.
Lemma 2.9 (Zorn’s lemma). Let X be a non-empty set, partially ordered by ≤. If every chain, or totally
ordered subset, in X has an upper bound in X, then X has a maximal element.
Proof.
⇐= Suppose α : A → B, where α is injective. Suppose f : A → M . We want to show there exists
g : B → M such that f = g ◦ α. We have Im α ≤ B. Define
X = {(L, h) | Im α ≤ L ≤ B, h : L → M, f = h ◦ α} .
Note that X 6= ∅ since Im α, f ◦ α−1 is in it. Define ≤ on X by (L1 , h1 ) ≤ (L2 , h2 ) S

if L1 ≤ L2 and
h2 extends h1 , so h2 |L1 = h1 . Suppose {(Ls , hs ) | s ∈ S} is a chain in X. Set L = s∈S Ls . Then
Im α ≤ L ≤ B. Define
h : L −→ M
, l ∈ Ls .
l 7−→ hs (l)
This does not depend on the choice of s. Then (L, h) is an upper bound for the chain {(Ls , hs ) | s ∈ S}.
Hence X has a maximal element, (L0 , h0 ). We want to show that L0 = B. Then we may set g = h0 .
Suppose that L0 6= B. Let b ∈ B \ L0 . Note that Rb ≤ B. Consider
L0 + Rb = {l + rb | l ∈ L0 , r ∈ R} ≤ B.
We would like to extend h0 to h+ +
0 by specifying an image for h0 (b). The problem is that Rb ∩ L0 may
not be {0}, and if rb ∈ L0 then we require rh0 (b) = h0 (rb), otherwise h+
+
0 will not be well-defined.
Note that I = {r ∈ R | rb ∈ L0 } is a left ideal for R. Suppose that M has the condition from Baer’s
criterion, so every map I → M has the form x 7→ xm for some m ∈ M . Note that {xb | x ∈ I} is a
submodule of L0 . Define
δ : I −→ M
.
x 7−→ h0 (xb)
This is an R-module homomorphism. So δ (x) = xm for somem ∈ M . Hence h0 (xb) = xm for all
x ∈ I. So we can safely define h+ +
∈ X, and (L0 , h0 ) < L0 + Rb, h+

0 (b) = m. Now L0 + Rb, h0 0 ,
which contradicts the maximality of (L0 , h0 ). Hence L0 = B, and we are done.
3
=⇒ The converse is left as an exercise.
Example.
• Suppose R is a field. Then the only ideals of R are zero and R. Any map 0 → M , for M an R-module,
can be extended to the zero map R → M . Hence any R-module is injective.
• Let Z be a module for itself. The ideals of Z are kZ for k ∈ Z. Define
f : kZ −→ Z
.
km 7−→ m
If k 6= 0, ±1, then f (k) = 1, and so f (x) 6= xm for m ∈ Z, since one is not divisible by k in Z. So
Baer’s criterion fails, and Z is not injective. We already knew that Z → Q has no retraction.
• Q is injective as a Z-module. Suppose we have a map f : kZ → Q. Let q = f (k). Then f (kt) = qt =
(q/k) kt. So f (x) = x (q/k) for all x, so Q satisfies Baer’s criterion.
3 Exercise
9
M4P63 Algebra IV 3 Hom and tensor products
3 Hom and tensor products

3.1 Hom
Lecture 6
Let A and B be two R-modules. Tuesday
21/01/20
Definition 3.1. Define
HomR (A, B) = {R-module homomorphisms A → B} .
We can define a natural addition on HomR (A, B) by defining f1 + f2 by
(f1 + f2 ) (a) = f1 (a) + f2 (b) , f1 , f2 ∈ HomR (A, B) .
This gives HomR (A, B) the structure of an abelian group. Why does HomR (A, B) not carry an R-module
structure in general? The only obvious candidate for rf is
(rf ) (a) = rf (a) = f (ra) , r ∈ R, f ∈ HomR (A, B) .
Now suppose s ∈ R. We have (rf ) (sa) = rf (sa) = rsf (a). But for rf to be a homomorphism, we would
need (rf ) (sa) = s (rf ) (a) = srf (a). If R is non-commutative, then rs may not be sr, and so rf is not
an R-module homomorphism in general. Clearly, however, if R is commutative then rf is an R-module
homomorphism, and HomR (A, B) has an R-module structure. The following are observations.
Proposition 3.2. Suppose A, A1 , A2 , B, B1 , B2 , M are R-modules, and α : A → B.
• HomR (A1 ⊕ A2 , B) ∼
= HomR (A1 , B) ⊕ HomR (A2 , B).
• HomR (A, B1 ⊕ B2 ) ∼
= HomR (A, B1 ) ⊕ HomR (A, B2 ).
• Then we can define
α∗ : HomR (M, A) −→ HomR (M, B)
, f : M → A.
f 7−→ α◦f
• We can also define

α∗ : HomR (B, M ) −→ HomR (A, M )
, g : B → M.
g 7−→ g◦α
Thus Hom is a bifunctor between the category of R-modules and the category of abelian groups, additive in
both arguments, covariant in the second argument and contravariant in the first argument.
• Bi means Hom takes two arguments.
• Functor means that homomorphisms between R-modules turn into abelian group homomorphisms.
• Covariant means the homomorphism goes in the same direction.
• Contravariant means the direction gets reversed.
• Additive in both arguments means Hom respects direct sums.
Proposition 3.3. Suppose α : A → B is surjective. Then α∗ : HomR (B, M ) → HomR (A, M ) is injective.
Proof. Suppose f1 , f2 : B → M are such that α∗ (f1 ) = α∗ (f2 ). Then f1 ◦ α = f2 ◦ α, so (f1 ◦ α) (a) =
(f2 ◦ α) (a) for all a ∈ A. Let b ∈ B. Then b = α (a) for some a, since α is surjective, so f1 (b) = (f1 ◦ α) (a) =
(f2 ◦ α) (a) = f2 (b), so f1 = f2 .
Proposition 3.4. Suppose α : A → B is injective. Then α∗ : HomR (M, A) → HomR (M, B) is injective.
Proof. Suppose f1 , f2 : M → A, and α∗ (f1 ) = α∗ (f2 ). Then α ◦ f1 = α ◦ f2 , so (α ◦ f1 ) (m) = (α ◦ f2 ) (m)
for all m ∈ M . But α is injective, so this implies f1 (m) = f2 (m) for all m ∈ M .
10
Proposition 3.5. Suppose

α β
0→A−
→B−
→C→0
is a short exact sequence of R-modules. Then we have an exact sequence
β∗ α∗
0 → HomR (C, M ) −→ HomR (B, M ) −−→ HomR (A, M ) .
Proof. This is exact at HomR (C, M ), since β ∗ is injective. Claim that the sequence is also exact at
HomR (B, M ), so it is an exact sequence. It is not necessarily a short exact sequence since α∗ is not
generally surjective. Let g : B → M . We have
g ∈ Ker α∗ ⇐⇒ α∗ (g) = 0 ⇐⇒ g ◦ α = 0 ⇐⇒ g (α (A)) = 0 ⇐⇒ Im α ≤ Ker g ⇐⇒ Ker β ≤ Ker g,
Then g ∈ Ker α∗ if and only if for all b1 , b2 ∈ B, β (b1 ) = β (b2 ) implies that g (b1 ) = g (b2 ), which is if and
only if the map defined by
f : C −→ M
, β (b) = c
c 7−→ g (b)
is well-defined, since β is surjective, and f is an R-module homomorphism. Thus
g ∈ Ker α∗ ⇐⇒ ∃f ∈ HomR (C, M ) , β ∗ (f ) = g ⇐⇒ g ∈ Im β ∗ .
Hence Ker α∗ = Im β ∗ . So the sequence is exact at HomR (B, M ).

Lecture 7
Example. These examples show that α : A → B is injective does not imply α ∗ : HomR (B, M ) → Friday
HomR (A, M ) is surjective. 24/01/20
• The inclusion α : Z → Q is a Z-module homomorphism. Let M = Z. Then we get α∗ : HomZ (Q, Z) →

HomZ (Z, Z). Then α is injective, but α∗ is not surjective. Why is this? In fact HomZ (Q, Z) = 0.
Suppose
f : Q −→ Z
.
1 7−→ k 6= 0
Suppose p - k. Then there is no possible image for 1/p ∈ Q, since we would require pf (1/p) = f (1) = k.
But HomZ (Z, Z) ∼= Z, so α∗ is not surjective.
• Let α : kZ → Z be the inclusion, so α is injective and not surjective. Let M = Z. So we get
α∗ : HomZ (Z, Z) → HomZ (kZ, Z). Suppose that g ∈ Im α∗ . Then g = f ◦ α, where f : Z → Z. Then
g (k) = f (k) = kf (1), so Im g ≤ kZ. But there exists g ∈ HomZ (kZ, Z) such that g (k) = 1. So this
g∈ / Im α∗ , so α∗ is not surjective.
Proposition 3.6. Let
α β
0→A−
→B−
→C→0
be exact. Then
∗ α β∗
0 → HomR (M, A) −−→ HomR (M, B) −→ HomR (M, C)
is exact.
Proof. We already know that α injective implies that α∗ is injective, so the sequence is exact at HomR (M, A).
We show that Ker β∗ = Im α∗ . Suppose g ∈ HomR (M, B). Then
g ∈ Ker β∗ ⇐⇒ (β ◦ g) (M ) = 0 ⇐⇒ Im g ≤ Ker β ⇐⇒ Im g ≤ Im α.
Note there exists α−1 : Im α → A. If Im g ≤ Im α, then α−1 ◦ g : M → A. If f = α−1 ◦ g, then α ◦ f = g, so

g ∈ Im α∗ . Conversely, if g ∈ Im α∗ , then g = α ◦ f for some f ∈ HomR (M, A) and so Im g ≤ Im α. So
g ∈ Ker β∗ ⇐⇒ Im g ≤ Im α ⇐⇒ g ∈ Im α∗ .
Hence Ker β∗ = Im α∗ . So the sequence is exact at HomR (M, B).
11
Example. These examples show that β : B → C is surjective does not imply β∗ : HomR (M, B) →
HomR (M, C) is surjective.
• Let P
β : q∈Q Z −→ Q
.
eq 7−→ q
P
In general β : m∈M R → M defined by mapping the basis vector Pem to m, is a surjective homo-
morphism, so β is surjective. Let M = Q. So we get β∗ : HomZ Q, q∈Q Z → HomZ (Q, Q). Claim
P P
that HomZ Q, q∈Q Z is trivial. Suppose f : Q → q∈Q Z is not zero. Suppose f (q0 ) 6= 0. Then
Pt
there exist q1 , . . . , qt ∈ Q and a1 , . . . , at ∈ Z such that f (q0 ) = i=1 ai eqi . Now the projection of
P ∼
q∈Q Z onto Zeq1 is a non-trivial Z-module homomorphism. But Zeq1 = Z, and so no non-trivial map
Q → Zeq1 exists. But HomZ (Q, Q) is not trivial, so β∗ is not surjective.
• Let
0 → Z2 → Z4 → Z2 → 0
be a short exact sequence of Z-modules. Then we have
α∗ β∗
0 HomZ (Z2 , Z2 ) HomZ (Z2 , Z4 ) HomZ (Z2 , Z2 )
.
∼
∼
=
=
Z2 Z2 Z2
But there is no short exact sequence of abelian groups
0 → Z2 → Z2 → Z2 → 0,
and so β∗ cannot be surjective.
Proposition 3.7. Let M be an R-module. Then M is injective if and only if for every injective map
α : A → B, we get α∗ : HomR (B, M ) → HomR (A, M ) is surjective.
Proof. M is injective if and only if for all injective α : A → B, for all f ∈ HomR (A, M ), there exists
g ∈ HomR (B, M ) such that f = g ◦ α, so f = α∗ (g). This is if and only if for all injective α : A → B,
f ∈ Im α∗ for all f ∈ HomR (A, M ), which is if and only if α∗ is surjective.
Proposition 3.8. Let M be an R-module. Then M is projective if and only if whenever β : B → C is
surjective, the map β∗ : HomR (M, B) → HomR (M, C) is surjective.
Proof. M is projective if and only if whenever β : B → C is surjective, and f ∈ HomR (M, C), there exists
g ∈ HomR (M, B) such that f = β ◦ g. This is if and only if whenever β : B → C is surjective, and
f ∈ HomR (M, C), then f ∈ Im β∗ , which is if and only if β∗ is surjective.
3.2 The snake lemma

Lecture 8
Let α : A → B be an R-module homomorphism. The cokernel of α is B/ Im α, written Coker α. The Monday
sequence 27/01/20
α
0 → Ker α → A −
→ B → Coker α → 0
is exact.
Lemma 3.9 (The snake lemma). Suppose we have a commutative diagram
α β
A B C 0
f g h ,
0 X φ
Y ψ
Z
where the rows are exact. Then we obtain an exact sequence

α β δ φ ψ
Ker f −
→ Ker g −
→ Ker h −
→ Coker f −
→ Coker g −
→ Coker h.
12
Proof.
• The maps α : Ker f → Ker g and β : Ker g → Ker h are obtained simply by restricting α and β
respectively. Observe that if a ∈ Ker f then f (a) = 0, so (φ ◦ f ) (a) = 0. But φ ◦ f = g ◦ α, and so
(g ◦ α) (a) = 0, so α (a) ∈ Ker g, which is what we wanted.
• The maps φ : Coker f → Coker g and ψ : Coker g → Coker h are induced from φ and ψ by
φ (x + Im f ) = φ (x) + Im g, ψ (y + Im g) = ψ (g) + Im h.
Check that these maps make sense. Suppose x1 + Im f = x2 + Im f . Then x1 − x2 ∈ Im f , so there
exists a ∈ A such that f (a) = x1 − x2 . Now
φ (x1 ) − φ (x2 ) = φ (x1 − x2 ) = (φ ◦ f ) (a) = (g ◦ α) (a) ∈ Im g.
So φ (x1 ) + Im g = φ (x2 ) + Im g. So φ is well-defined, and ψ is shown to be well-defined by a similar
argument.
• How is the connecting homomorphism δ defined? Since β is surjective, for all c ∈ C, there
exists b ∈ B with β (b) = c. Suppose c ∈ Ker h. Then (h ◦ β) (b) = 0, so (ψ ◦ g) (b) = 0. Hence
g (b) ∈ Ker ψ = Im φ. Define
δ (c) = x + Im f, φ (x) = g (b) , β (b) = c.
Check this is well-defined. Suppose b1 , b2 , x1 , x2 are such that φ (x1 ) = g (b1 ) and φ (x2 ) = g (b2 ), and
β (b1 ) = β (b2 ) = c. We have b1 − b2 ∈ Ker β = Im α. So b1 − b2 = α (a) for some a ∈ A. Then
(φ ◦ f ) (a) = (g ◦ α) (a) = g (b1 − b2 ) = g (b1 ) − g (b2 ) = φ (x1 ) − φ (x2 ) = φ (x1 − x2 ) .
But φ is injective, and so f (a) = x1 − x2 , and so x1 + Im f = x2 + Im f . So δ is well-defined.
Exactness of the sequence is an exercise, on problem sheet.
3.3 Tensor products

Definition 3.10. Let M be a left R-module, and let L be a right R-module. The tensor product L ⊗R M
is an abelian group generated as an abelian group by a set of pure tensors
{l ⊗ m | l ∈ L, m ∈ M } ,
subject to the relations
l1 ⊗ m + l2 ⊗ m = (l1 + l2 ) ⊗ m, l1 , l2 ∈ L, m ∈ M,
l ⊗ m1 + l ⊗ m2 = l ⊗ (m1 + m2 ) , l ∈ L, m1 , m2 ∈ M,
(lr) ⊗ m = l ⊗ (rm) , l ∈ L, m ∈ M, r ∈ R.
The following are observations.
• In general, not every element of L ⊗R M is a pure tensor. A general element of L ⊗R M is a Z-linear
combination of pure tensors.
• If R is commutative, L can be a left module, since left and right modules are the same. Also, in this
case, L ⊗R M has an R-module structure, by r (l ⊗ m) = rl ⊗ m.
• Suppose that S is a set of generators for L, as an abelian group, and T is a set of generators for M , as
an abelian group. Then a smaller generating set for L ⊗R M is {s ⊗ t | s ∈ S, t ∈ T }. This is because
if
Xp Xq
l= ai si , m= bj tj , si ∈ S, ti ∈ T, ai , bi ∈ Z,
i=1 i=1
then, from the relations,
p X
X q
l⊗m= ai bj (si ⊗ tj ) .
i=1 j=1
13
Example. Tensor products can be counter intuitive, such as Z2 ⊗Z Z3 = 0. Why? Observe that for x ∈ Z2 ,
x3 = 3x = x. So for all x ∈ Z2 and y ∈ Z3 ,
x ⊗ y = x3 ⊗ y = x ⊗ 3y = x ⊗ 0 = x ⊗ y − x ⊗ y = 0.
Lecture 9
Tuesday
Theorem 3.11 (Universal property of tensor products). Let A be a right R-module and B a left R-module.
28/01/20
Let C be an abelian group. Let f : A × B → C be a map, not necessarily a homomorphism, which is Z-linear
in both arguments, so
f (a1 + a2 , b) = f (a1 , b) + f (a2 , b) , a1 , a2 ∈ A, b ∈ B,
f (a, b1 + b2 ) = f (a, b1 ) + f (a, b2 ) , a ∈ A, b1 , b2 ∈ B,

and such that
f (ar, b) = f (a, rb) , a ∈ A, b ∈ B, r ∈ R.
Then there is a unique homomorphism
g : A ⊗R B −→ C
.
a ⊗ b 7−→ f (a, b)
Proof. In formal group theoretic terms, the tensor product A ⊗R B is a quotient F/K, where F is the free
abelian group on the set of pure tensors a ⊗ b, and K is the subgroup of F generated by elements of the form
(a1 + a2 ) ⊗ b − a1 ⊗ b − a2 ⊗ b, a ⊗ (b1 + b2 ) − a ⊗ b1 − a ⊗ b2 , ar ⊗ b − a ⊗ rb.
The universal property of free abelian groups states that if F is free abelian on a set S, then any set map
S → C, for C an abelian group, extends uniquely to a homomorphism F → C. In the situation under
discussion, we have a map
g 0 : {a ⊗ b | a ∈ A, b ∈ B} → C.
So g 0 extends uniquely to a homomorphism F → C. The conditions stipulated on f guarantee that g 0 (K) = 0.
So g 0 induces a map g : F/K → C, which is what we want, since F/K = A ⊗R B. This establishes the
existence of g. Since the images of the pure tensors under g are specified, it is clear that g is unique.
Corollary 3.12.
1. Let M be a left R-module. Then R ⊗R M ∼
= M , via the map
f : M −→ R ⊗R M
.
m 7−→ 1⊗m
2. Let M be a right R-module. Then M ⊗R R ∼

= M.
Proof.
1. It is clear that f is a homomorphism of abelian groups. Now r ⊗ m = 1 ⊗ rm, so R ⊗R M is generated
by {1 ⊗ m | m ∈ M }, so f is surjective. For injectivity of f , we need the universal property. Define a
bilinear map
R × M −→ M
.
(r, m) 7−→ rm
This induces a homomorphism
g : R ⊗R M −→ M
.
r⊗m 7−→ rm
It is easy to check that g is an inverse for f , so f is bijective.
2. By the same argument as 1.
14
Corollary 3.13. Let A and B be right R-modules, and let C be a left R-module.
1. (A ⊕ B) ⊗R C ∼
= (A ⊗R C) ⊕ (B ⊗R C), via the map
f : (A ⊕ B) ⊗R C −→ (A ⊗R C) ⊕ (B ⊗R C)
.
(a, b) ⊗ c 7−→ (a ⊗ c, b ⊗ c)
2. A ⊗R (B ⊕ C) ∼
= (A ⊗R B) ⊕ (A ⊗R C).
Proof.
1. Take a bilinear map, that is Z-bilinear in both arguments, and respecting R-multiplication,
A ⊕ B × C −→ (A ⊗R C) ⊕ (B ⊗R C)
.
((a, b) , c) 7−→ (a ⊗ c, b ⊗ c)
This induces a homomorphism f : (A ⊕ B) ⊗R C → (A ⊗R C) ⊕ (B ⊗R C) with the description as
given above. Now take the bilinear map given by
A × C −→ (A ⊕ B) ⊗R C
.
(a, c) 7−→ (a, 0) ⊗ c
This induces a homomorphism g1 : A ⊗R C → (A ⊕ B) ⊗R C. Similarly, we get a homomorphism
g2 : B ⊗R C → (A ⊕ B) ⊗R C. Now define
g = g1 ⊕ g2 : (A ⊗R C) ⊕ (B ⊗R C) −→ (A ⊕ B) ⊗R C
.
(x, y) 7−→ g1 (x) + g2 (y)
It is easy to check that f and g are mutually inverse, so both isomorphisms.
2. Similarly.
Corollary 3.14. Let A be an abelian group. Then

1. Zn ⊗Z A ∼
= A/nA, and
2. A ⊗Z Zn ∼
= A/nA.
Proof.
1. Define a map by
f : A −→ Zn ⊗Z A
.
a 7−→ 1⊗a
Suppose a0 ∈ A such that a0 = na for some a. Then f (a0 ) = 1 ⊗ a0 = 1 ⊗ na = n ⊗ a = 0 so
nA ≤ Ker f . So f induces a map
f : A/nA → Zn ⊗Z A.
Notice that the pure tensor k ⊗ a is equal to 1 ⊗ ka, so Zn ⊗Z A is generated by {1 ⊗ a | a ∈ A}. So f
is surjective. For injectivity, use the universal property. We have a bilinear map
g : Zn × A −→ A/nA
.
(k, a) 7−→ ka + nA
This is well-defined and bilinear. So extends to a homomorphism
g : Zn ⊗Z A → A/nA.
It is easy to check that g ◦ f = idA/nA , so f is injective.
2. Similarly.
15
Lecture 10
Proposition 3.15. Let α : A → B be a homomorphism of right R-modules. Let M be a left R-module. Friday
There is a unique abelian group homomorphism 31/01/20
α0 : A ⊗R M −→ B ⊗R M
, a ∈ A, m ∈ M.
a⊗m 7−→ α (a) ⊗ m
Proof. The set map defined by
f : A × M −→ B ⊗R M
(a, m) 7−→ α (a) ⊗ m
is linear in both arguments, and we have
f (ar, m) = α (ar) ⊗ m = α (a) r ⊗ m = α (a) ⊗ rm = f (a, rm) .
Now by the universal property of tensor products, f gives rise to a unique homomorphism α0 : A ⊗R M →
B ⊗R M with the properties claimed.
Proposition 3.16. Suppose α : A → B is surjective. Then α0 : A ⊗R M → B ⊗R M is surjective.
Proof. Since α is surjective, every pure tensor b ⊗ m ∈ B ⊗R M is equal to α (a) ⊗ m for some a ∈ A. So
b ⊗ m = α0 (a ⊗ m) ∈ Im α0 . Since B ⊗R M is generated by its pure tensors, α0 is surjective.
An observation is that it is not true that A → B is injective implies A ⊗R M → B ⊗R M is injective.
Example. Let
α : Z2 −→ Z4
,
1 7−→ 2
which is injective. Consider
α0 : Z2 ⊗Z Z2 ∼
= Z2 −→ Z4 ⊗Z Z2
.
1⊗1 7−→ 2⊗1=1⊗2=0
So α0 is the zero map, which is not injective.
α β
0→A−
→B−
→C→0
be a short exact sequence of right R-modules. Then the sequence
α0 β0
A ⊗R M −→ B ⊗R M −→ C ⊗R M → 0
is exact.
Proof. Since β 0 is surjective, the sequence is exact at C ⊗R M . We show it is exact at B ⊗R M . Since β
is surjective, for every c ∈ C, there exists f (c) ∈ B such that β (f (c)) = c. Here f is a set map C → B,
which is not uniquely defined in general. Suppose that β (b) = c. Then b − f (c) ∈ Ker β = Im α, so
f (c) + Im α = b + Im α. Define a set map by
g : C × M −→ (B ⊗R M ) / Im α0
.
(c, m) 7−→ f (c) ⊗ m + Im α0
Note that if β (b) = c, then b⊗m−f (c)⊗m = α (a)⊗m ∈ Im α0 for some a ∈ A. We can check that g is linear
in both arguments. For example, for the first argument, we have g (c1 + c2 , m) = f (c1 + c2 ) ⊗ m + Im α0 .
Now β (f (c1 + c2 )) = c1 + c2 = β (f (c1 )) + β (f (c2 )) = β (f (c1 ) + f (c2 )) so
g (c1 + c2 , m) = (f (c1 ) + f (c2 )) ⊗ m + Im α0 = f (c1 ) ⊗ m + f (c2 ) ⊗ m + Im α0 = g (c1 , m) + g (c2 , m) .
Also, we have g (cr, m) = f (cr) ⊗ m + Im α0 . But β (f (cr)) = cr = β (f (c) r), so f (cr) ⊗ m + Im α0 =
f (c) r ⊗ m + Im α0 . So
g (cr, m) = f (c) r ⊗ m + Im α0 = f (c) ⊗ rm + Im α0 = g (c, rm) .
16
By the universal property, there is a unique homomorphism

ψ : C ⊗R M −→ (B ⊗R M ) / Im α0
.
c⊗m 7−→ f (c) ⊗ m + Im α0
Next observe that (β 0 ◦ α0 ) (a ⊗ m) = (β ◦ α) (a) ⊗ m = 0, since Im α = Ker β. Since A ⊗R M is generated
by pure tensors, we have β 0 ◦ α0 = 0. So Im α0 ≤ Ker β 0 . Hence β 0 induces a map
φ : (B ⊗R M ) / Im α0 → C ⊗R M.
It is easy to check that φ and ψ are mutually inverse, and so both are isomorphisms. In particular φ is
injective, and so Im α0 = Ker β 0 as required.
3.4 Flat modules

Definition 3.18. A left R-module M is flat if A → B is injective implies that A ⊗R M → B ⊗R M is
injective.
If M is flat then any short exact sequence of right R-modules
0→A→B→C→0
corresponds to a short exact sequence of abelian groups
0 → A ⊗R M → B ⊗R M → C ⊗R M → 0.
Proposition 3.19. Every projective module is flat.
This follows from two lemmas.
Lemma 3.20. P ⊕ Q is flat if and only if P and Q are both flat.
Proof. Recall there is a canonical isomorphism
A ⊗R (P ⊕ Q) ∼
= (A ⊗R P ) ⊕ (A ⊗R Q) .
Suppose α : A → B is injective. Then α0 : A ⊗R (P ⊕ Q) → B ⊗R (P ⊕ Q) corresponds to
α0 : (A ⊗R P ) ⊕ (A ⊗R Q) −→ (B ⊗R P ) ⊕ (B ⊗R Q)
(a ⊗ p, 0) 7−→ (α (a) ⊗ p, 0) .
(0, a ⊗ q) 7−→ (0, α (a) ⊗ q)
It is clear from this that α0 is injective if and only if A ⊗R P → B ⊗R P and A ⊗R Q → B ⊗R Q are injective,
and Lemma 3.20 follows immediately.
Lemma 3.21. Every free R-module is flat.
Lecture 11
Lecture 11 is a problems class. Monday
03/02/20
Proof. We know (A ⊕ B) ⊗R C ∼
= (A ⊗R C) ⊕ (B ⊗R C). Similarly,
! Lecture 12
Tuesday
As ⊗ R C ∼
M M
= (As ⊗R C) .
04/02/20
s∈S s∈S
L
3.20 generalises, so s∈S As is flat if and only if all of the As is flat for s ∈ S. Let F be free.
So LemmaL
Then F = s∈S R, and so F is flat if and only if R is flat. But for any R-module in A, we have A ⊗R R ∼
= A,
so
α
A B
,
∼
∼
=
A ⊗R R 0
B ⊗ R R
α
and it is easy to check that R is flat.
Proof of Proposition 3.19. Lemma 3.20 and Lemma 3.21 imply Proposition 3.19, since a projective module
is a direct summand of a free module.
17
M4P63 Algebra IV 4 Modules over a PID
4 Modules over a PID

There exist flat modules which are not projective. We will show that Q as a module for Z is flat, and it is
easy to see it is not projective. To do this we will study the case of modules over a PID. Recall that R is
an integral domain if R is commutative and rs = 0 implies that r = 0 or s = 0 for r, s ∈ R. An integral
domain is a PID if every ideal is hai = {ra | r ∈ R} for some a ∈ R.
Example. The ring Z is an example of a PID.
4.1 Free and projective modules

Proposition 4.1. Let R be a PID. Then every projective R-module is free. Equivalently, every summand
of a free module is free.
In fact we will show that any submodule of a free module is free. Moreover, if F1 ≤ F2 , where F1 and F2
are free, and if B1 and B2 are bases for F1 and F2 respectively, then |B1 | ≤ |B2 |. In particular, if M ≤ Rn ,
then M ∼ = Rm for some m ≤ n. For this, we will need the well-ordering theorem.
Theorem 4.2 (Well-ordering theorem). Let X be a set. There exists a well-order ≤ on X, that is a total
order such that every non-empty subset of X has a least element.
Corollary 4.3 (Transfinite induction). Let X be a non-empty set well-ordered by ≤. Let x0 be the least
element of X. Let S ⊆ X. If x0 ∈ S, and s < t implies s ∈ S implies that t ∈ S, then S = X.
L
Proof. Let F = s∈S R. Let ≤ be a well-order on S. For s ∈ S, let πs be the projection map F → R onto
the s-coordinate. Let es be the element of F with one in coordinate s, and zero elsewhere. Suppose U ≤ F
is an R-submodule of F . Define Rt to be the submodule of F generated by {es | s ≤ t}, so
Rt = sp {es | s ≤ t} .
So if t1 ≤ t2 then Rt1 ≤ Rt2 . Let
Ut = U ∩ Rt .
So t1 < t2 implies that Ut1 ≤ Ut2 . Consider πs (Us ). This is an ideal of R. Hence there exists as ∈ R such
that πs (Us ) = has i, since R is a PID. For each s, let us ∈ Us be such that πs (us ) = as . In cases where
as = 0, assume us = 0. Let
B = {us | s ∈ S, us 6= 0} .
• Claim that B generates U . We will actually prove that Bt = {us | s ≤ t} generates Ut , using transfinite
induction. If s0 is the least element of S, it is easy to see that Bs0 = {us0 } generates Us0 . Suppose
Bt generates Ut for all t < t0 . Let u ∈ Ut0 . Then πt0 (u) = rat0 . Hence πt0 (u − rut0 ) = 0. So
u − rut0 has
Pzero in the t0 -coordinate, so u − rut0 ∈ sp {es | s < t0 }. Clearly u − rut0 ∈ U . We have
q
u − rut0 = i=1 ri esi , where si < t0 , and s1 < · · · < sq . Then
u − rut0 ∈ U ∩ Rsq = Usq = sp Bsq ,

by the inductive hypothesis. Hence u ∈ sp Bsq ∪ {ut0 } ⊆ sp Bt0 . Hence Bt0 generates Ut0 , as
required.
• Next we show the linear
Pk independence of B. Suppose we have a linear combination of elements of B
equal to zero. Say i=1 ri usi = 0. Assume s1 < · · · < sk . We have
k
! k
X X
πs k ri usi = ri πsk (usi ) .
i=1 i=1
Now usi ∈ Usi ⊆ Rsi , and so πsk (usi ) = 0 if si < sk . Hence rk πsk (usk ) = 0, so rk ask = 0. But
ask 6= 0, and R is an integral domain. So rk = 0. It follows easily that ri = 0 for all i, so B is linearly
independent.
We have shown that B is a basis for U . Hence U is free. Since the elements of B are indexed by a subset of
S, we have |B| ≤ |S|.
Lecture 13
Lecture 13 is a problems class. Friday
07/02/20
18
4.2 Injective and divisible modules

Lecture 14
Definition 4.4. Let R be an integral domain, and M an R-module. Let m ∈ M . Say that m is infinitely Monday
divisible if for all r ∈ R \ {0} there exists l ∈ M such that rl = m. 10/02/20
Proposition 4.5. The divisible elements of M form a submodule D (M ).
Proof. Easy.
Definition 4.6. If D (M ) = M , then M is divisible.
Proposition 4.7. Let R be an integral domain. Then if an R-module M is injective then it is divisible.
Proof. Recall that for an integral domain R, and a ∈ R \ {0}, the map
f : R −→ hai
r 7−→ ra
is an isomorphism. Suppose M is an injective R-module. Let
g : R−→ M
.
17−→ m
Then g ◦ f −1 is a homomorphism hai → M , and g ◦ f −1 (a) = g (1) =

−1 −1
m. Now by Baer’s criterion, there
is a map h : R → M extending g ◦ f . Now ah (1) = h (a) = g ◦ f (a) = m. Hence there exists l ∈ M
such that al = m. So m is a divisible element, and so M is divisible.
Proposition 4.8. Let R be a PID. If M is a divisible R-module then M is injective.
So divisible equals injective when R is a PID.
Proof. We use Baer’s criterion. Let I be an ideal of R, and f : I → M an R-module homomorphism. Since
R is a PID, I = hai for some a ∈ R. Suppose f (a) = m. If a = 0 there is nothing to prove, since the zero
map R → M extends f . So assume a 6= 0. Since m is divisible, there exists l ∈ M with al = m. Now the
map given by
R −→ M
1 7−→ l
extends f . So Baer’s criterion is satisfied, and so M is injective.
4.3 Flat and torsion-free modules

Definition 4.9. Let R be an integral domain. Let M be an R-module. Say that m ∈ M is a torsion
element if there exists r ∈ R \ {0} such that rm = 0.
Proposition 4.10. The torsion elements of M form a submodule T (M ).
Proof. Easy, using the fact that integral domains are commutative.
Definition 4.11. If T (M ) = 0, then M is torsion-free. If T (M ) = M , then M is a torsion module.
Proposition 4.12. Let R be an integral domain. Let M be a flat R-module. Then M is torsion-free.
Proof. Let a ∈ R \ {0}. Then
f : R −→ R
1 7−→ a
is an injective R-module homomorphism. Suppose that M is flat. Then the map
g : R ⊗R M −→ R ⊗R M
r⊗m 7−→ ra ⊗ m = r ⊗ am
is injective. But R ⊗R M is canonically isomorphic to M , under which the map g corresponds to m 7→ am.
Since g is injective, we have am 6= 0 for m 6= 0. Hence m is not a torsion element, if m 6= 0, and so M is
torsion-free.
19
We now build up to the following.

Proposition 4.13. Let R be a PID. If M is a torsion-free R-module then M is flat.
The following is the strategy. We want to prove that whenever α : A → B is injective, so is α0 : A ⊗R M →
B ⊗R M , where M is torsion-free.
1. Prove this in the case that B is free, and A is a submodule of B, and α is the inclusion map, by
• first reducing the problem to the case that A and B are finitely generated, so B ∼
= Rn , and
• then using induction on the rank n of B.
2. Show the general case follows from 1.
Lemma 4.14. Let R be a PID, let I = hai be an ideal of R, and let M be a torsion-free R-module. Then
g : I ⊗R M → R ⊗R M is injective.
Proof. The homomorphism given by
R −→ I
r 7−→ ra
gives a map f : R ⊗R M → I ⊗R M . Now g ◦ f is a map
R ⊗R M −→ R ⊗R M
.
r 7−→ ra
Now f is surjective, and g ◦ f is injective, since R is an integral domain. But this implies that g is injective,
as required.
Pt
Lemma 4.15. Let A be a right R-module. Let M be a left R-module. Suppose i=1 (ai ⊗ mi ) = 0 in A⊗R M .
Pt
There exists a finitely generated submodule A0 ≤ A such that ai ∈ A0 for all i, and i=1 (ai ⊗ mi ) = 0 in
A0 ⊗ R M .
Proof. Recall that
A ⊗R M = Fab (A × M ) /K,
Pt
where K is generated by certain relators. If i=1 (ai ⊗ mi ) = 0 in A ⊗R M , then in Fab (A × M ), we have
Pt
i=1 (ai ⊗ mi ) ∈ K. So there exist relators s1 , . . . , sq , or their negations, such that
t
X q
X
(ai ⊗ mi ) = si .
i=1 i=1
Only finitely many elements of A are involved in the relators s1 , . . . , sq . Let A0 be generated by these
Pt Pq
together with a1 , . . . , at . Then certainly ai ∈ A0 for all i. And i=1 (ai ⊗ mi ) = i=1 si in Fab (A0 × M )
Pt
so i=1 (ai ⊗ mi ) = 0 in A0 ⊗R M . Clearly A0 is finitely generated.
Lecture 15
L
Lemma 4.16. Let F = F (S) = s∈S R. Let U be a finitely generated submodule of F . Then there exists Tuesday
a finite T ⊆ S such that U ≤ F (T ), and for any M , the map F (T ) ⊗R M → F (S) ⊗R M is injective. 11/02/20
Proof. Let u1 , . . . , uq be generators for U . Every ui is an R-linear combination of elements of S. Since each
of these linear combinations mentions only finitely many elements of S, there is a finite subset T ⊆ S such
that every ui is an R-linear combination of elements of T . So U ≤ F (T ). We have
F (S) = F (T ) ⊕ F (S \ T ) ,
and so
F (S) ⊗R M ∼
= (F (T ) ⊗R M ) ⊕ (F (S \ T ) ⊗R M ) .
It follows that the natural map F (T ) ⊗R M → F (S) ⊗R M is injective.
20
Lemma 4.15 and Lemma 4.16 tell us that if F is free and U ≤ F , and if M is an R-module, if U ⊗R M →
F ⊗R M is not injective, then there exists a finitely generated U0 < U and a finite rank free submodule
F0 < F such that U0 ⊗R M → F0 ⊗R M is not injective.
Lemma 4.17. Let R be a PID. Let F be free, and U ≤ F . Let M be torsion-free. Then U ⊗R M → F ⊗R M
is injective.
Proof. We assume that F = Rn . We do this by induction on n.

Base case. Let n = 1. So F is R, and U is an ideal of R. By Lemma 4.14, U ⊗R M → F ⊗R M
is injective in this case.
Inductive hypothesis. U ≤ F = Rn−1 implies that U ⊗R M → F ⊗R M is injective.
Inductive step. Assume U ≤ F = Rn . Write Rn = R ⊕ Rn−1 . So we have a short exact sequence
0 → R → Rn → Rn−1 → 0.
We also have a short exact sequence
0 → U1 → U → πRn−1 (U ) → 0,

where U1 = U ∩ R ⊕ 0n−1 . Identifying R with R ⊕ 0n−1 , we get a commuting
diagram
0 U1 U πRn−1 (U ) 0
,
0 R Rn Rn−1 0
where the vertical maps are inclusions, and the rows are exact. Tensoring everything
with M , we get a new commuting diagram
U1 ⊗R M U ⊗R M πRn−1 (U ) ⊗R M 0
f g h .
0 R ⊗R M R n ⊗R M Rn−1 ⊗R M 0
The initial zero in the bottom row comes from the fact that
0 → R → Rn → Rn−1 → 0
is split, since Rn = R ⊕ Rn−1 , and so
Rn ⊗R M ∼
= (R ⊗R M ) ⊕ Rn−1 ⊗R M .

Now f is injective by Lemma 4.14, and h is injective by the inductive hypothesis.

The snake lemma tells us that the sequence
Ker f → Ker g → Ker h
is exact at Ker g. So
0 → Ker g → 0
is exact, and so Ker g = 0. So g is injective, and this completes the induction.
21
Proof of Proposition 4.13. Prove that if α : A → B is injective, and M is torsion-free, over a PID R, then
α0 : A ⊗R M → B ⊗R M is injective. There exists a free module F such that B is quotient of F . So there is
a short exact sequence
δ
0→K→F − → B → 0.
Now A ∼= αA = Im α. Let FA be the δ-preimage of αA. Then K < FA , and we have another short exact
sequence
0 → K → FA → αA → 0.
We have a commuting diagram
FA αA 0
0 K .
F B 0
Tensoring with M ,
γ
β
FA ⊗R M αA ⊗R M 0
K ⊗R M f g
δ
F ⊗R M B ⊗R M 0
is commuting, and exact along rows. Let u ∈ Ker g ≤ αA ⊗R M ∼ = A ⊗R M . Since γ is surjective, there
is w ∈ FA ⊗R M with γ (w) = u. So (g ◦ γ) (w) = 0. So ( ◦ f ) (w) = 0. So f (w) ∈ Ker = Im δ, so
f (w) = δ (k) for k ∈ K ⊗R M . Since f is injective, by Lemma 4.17, we get w = β (k) ∈ Im β. So w ∈ Ker γ,
so u = 0. Hence g is injective, as required.
We have shown that if R is a PID, and if M is torsion-free, then M is flat.
4.4 Modules over PIDs

Lecture 16
For an R-module M Friday
14/02/20
free =⇒ projective =⇒ flat =⇒ torsion-free, injective =⇒ divisible.
Over a PID
free ⇐⇒ projective =⇒ flat ⇐⇒ torsion-free, injective ⇐⇒ divisible.
Do we have projective if and only if flat, over a general ring, or over a PID? The answer is no.
Example. The Z-module
L Q is torsion-free, so flat. Is Q projective? Is Q free, since Z is a PID? Consider a
free Z-module F = s∈S Z. Let s0 ∈ S. Then let
(
1 s = s0
x = (xs )s∈S = ∈ F.
0 otherwise
It is clear there are no y ∈ F such that 2y = x. So x is not a divisible element of F . Indeed, D (F ) = {0}.
But D (Q) = Q. Hence Q ∼ 6 F . So Q is an example of a flat module which is not projective.
=
22
M4P63 Algebra IV 5 Projective and injective resolutions
5 Projective and injective resolutions

Definition 5.1. Let M be an R-module. A resolution, or left resolution, for M is a sequence of R-
modules A0 , A1 , A2 , . . . , with homomorphisms d : Ai+1 → Ai , and also a homomorphism A0 → M , such
that
d d d
... −
→ A2 −
→ A1 −→ A0 → M → 0
is an exact sequence, where d is the differential. If all of the modules Ai have a property P, we call this a
P-resolution.
So we can talk about free resolutions, projective resolutions, flat resolutions. We do not use the term injective
resolution in this context.
Definition 5.2. A right resolution, or coresolution, for M is a sequence of R-modules A0 , A1 , A2 , . . . ,
with homomorphisms d : Ai → Ai+1 , and M → A0 , such that
d d d
0 → M → A0 −
→ A1 −
→ A2 −
→ ...
is exact. If the modules Ai have a property P, we can refer to a right P-resolution.
An injective resolution always means a right injective resolution.
5.1 Existence of projective resolutions

Proposition 5.3. Let M be an R-module. Then M has free, projective, and flat resolutions.
Proof. Since free implies projective implies flat, it is enough to show that free resolutions exist. Use the fact
that for any module L, there exist a free module F and K ≤ F such that L ∼ = F/K. So we get a short exact
sequence
0 → K → F → L → 0.
It follows that we can find F0 , F1 , F2 , . . . , and K0 ≤ F0 , K1 ≤ F1 , K2 ≤ F2 , . . . such that
0 → K0 → F0 → M → 0, 0 → K1 → F1 → K0 → 0, 0 → K2 → F2 → K1 → 0, ...
are all exact. Since Ki ≤ Fi , we may consider the maps Fi+1 → Ki as maps Fi+1 → Fi with image Ki . But
Ki is the kernel of the map Fi → Ki−1 , so the sequence
· · · → F2 → F1 → F0 → M → 0
is exact, and a free resolution for M .
5.2 Existence of injective resolutions

Injective coresolutions exist too, but the proof is more intricate. It involves making use of properties of the
abelian group Q/Z.
Proposition 5.4. Let A be an abelian group, and let a ∈ A \ {0}. There is a homomorphism f : A → Q/Z
such that f (a) 6= 0.
Proof. Start by defining f0 : hai → Q/Z. If a has finite order t, then f0 : a 7→ 1/t + Z. If a has infinite order,
then f0 : a 7→ 21 + Z. We will use Zorn’s lemma. Let X be the set
{(B, f ) | B ≤ A, a ∈ B, f : B → Q/Z, f extends f0 } .
Then X is non-empty, since (hai , f0 ) ∈ X. Define a partial order ≤ on X by (B1 , f1 ) ≤ (B2 , f2 ) if B1 ≤
B2 and f2 extends f . Let {(Bs , fs ) | s ∈ S} be a chain in X, Swhere S is a suitable indexing set. Then
{Bs | s ∈ S} is a chain of subgroups of A. So the union B = s∈S Bs is a subgroup of A, containing a.
Define
f : B −→ Q/Z
, b ∈ Bs .
b 7−→ fs (b)
23
This is well-defined since if b ∈ Bt then fs (b) = ft (b). Now (B, f ) is an upper bound for {Bs | s ∈ S} in
X. So by Zorn’s lemma, X has a maximal element, which we will call (B, f ). We show that B = A. Since
f (a) = f0 (a), this will complete the proof. Suppose x ∈ A \ B. Then let I < Z be defined by
I = {k | kx ∈ B} .
Since Z is a PID, we have I = nZ for some n. We have hB, xi ≤ A, and hB, xi ∼ = B ⊕ hxi / hnx − b0 i, where
b0 = nx in A. Define
φ : B ⊕ hxi −→ Q/Z
kf (b0 ) ,
(b, kx) 7−→ f (b) +
n
so sending x to f (b0 ) /n. We see that φ (nx − b0 ) = 0, so φ induces a map B ⊕ hxi / hnx − b0 i → Q/Z, and
hence a map f 0 : hB, xi → Q/Z. But f 0 (a) = f0 (a), so (hB, xi , f ) is an element of X greater than (B, f ),
contradicting maximality of (B, f ). Hence B = A as required.
Lecture 17
Proposition 5.5. For every abelian group A, there is an injective abelian group I such that A is isomorphic Monday
to a subgroup of I. 17/02/20
Q
Proof. We know that Q/Z is injective, as a Z-module, since it is divisible, and Z is a PID. So s∈S Q/Z is
also injective. Take S = A \ {0}. Then define, for each s ∈ S, fs : A → Q/Z such that fs (s) 6= 0. Define
Q
f : A −→ s∈S Q/Z .
a 7−→ (fs (a))s∈S
Now if s ∈ A \ {0}, then fs (s) 6= 0, so f (s) 6= 0. So f is injective. It is easy to check that f is a
homomorphism.
Proposition 5.6. Let M be a right R-module, and let A be an abelian group. Then HomZ (M, A) is a left
R-module, with the R-action defined by (rf ) (m) = f (mr).
Proof. This is clearer if we write the map f on the right instead of the left. Then the proposition becomes
(m) (rf ) = (mr) f , and it is easy to see this works.
Proposition 5.7. Let M be a left R-module, and A an abelian group. Then HomZ (R, A) is a left R-module,
and there is a natural isomorphism
∼ HomZ (M, A) .
HomR (M, HomZ (R, A)) =
Proof. Write H = HomZ (R, A). Define
Φ : HomR (M, H) −→ HomZ (M, A)
, m ∈ M, 1 ∈ R.
f 7−→ (m 7→ f (m) (1))
Check the following.
• Φ (f ) is a homomorphism, since
Φ (f ) (m1 + m2 ) = f (m1 + m2 ) (1)
= (f (m1 ) + f (m2 )) (1)
= f (m1 ) (1) + f (m2 ) (1) definition of + in HomZ (R, A)
= Φ (f ) (m1 ) + Φ (f ) (m2 ) .
• Φ is a homomorphism, since
Φ (f1 + f2 ) (m) = (f1 + f2 ) (m) (1)
= (f1 (m) + f2 (m)) (1) definition of + in HomZ (M, A)
= f1 (m) (1) + f2 (m) (1)
= Φ (f1 ) (m) + Φ (f2 ) (m)
= (Φ (f1 ) + Φ (f2 )) (m) definition of + in HomZ (M, A) ,
so since m was arbitrary, Φ (f1 + f2 ) = Φ (f1 ) + Φ (f2 ).
24
Now define
Ψ : HomZ (M, A) −→ HomR (M, H)
, m ∈ M, r ∈ R.
p 7−→ (m 7→ (r 7→ p (rm)))
Check the following.
• Ψ (p) (m) is a homomorphism, since
Ψ (p) (m) (r1 + r2 ) = p ((r1 + r2 ) m) = p (r1 m + r2 m)

= p (r1 m) + p (r2 m) = Ψ (p) (m) (r1 ) + Ψ (p) (m) (r1 ) .
• Ψ (p) is an R-module homomorphism, since
Ψ (p) (m1 + m2 ) (r) = p (r (m1 + m2 )) = p (rm1 + rm2 ) = p (rm1 ) + p (rm2 )

= Ψ (p) (m1 ) (r) + Ψ (p) (m2 ) (r) = (Ψ (p) (m1 ) + Ψ (p) (m2 )) (r) ,
so Ψ (p) (m1 + m2 ) = Ψ (p) (m1 ) + Ψ (p) (m2 ), and for h ∈ H, we have (sh) (r) = h (rs), by definition
of the R-module structure on H, so
sΨ (p) (m) (r) = Ψ (p) (m) (rs) = p (rsm) = Ψ (p) (sm) (r) ,
so sΨ (p) (m) = Ψ (p) (sm).

• Ψ is a homomorphism, since
Ψ (p1 + p2 ) (m) (r) = (p1 + p2 ) (rm) = p1 (rm) + p2 (rm)

= Ψ (p1 ) (m) (r) + Ψ (p2 ) (m) (r) = (Ψ (p1 ) + Ψ (p2 )) (m) (r) ,
so Ψ (p1 + p2 ) = Ψ (p1 ) + Ψ (p2 ).

4
Then Ψ ◦ Φ = idHomR (M,H) and Φ ◦ Ψ = idHomZ (M,A) . Hence Φ and Ψ are isomorphisms.
We are interested in the case A = Q/Z. Write S = HomZ (M, Q/Z). Lecture 18
Tuesday
Proposition 5.8. S is injective as a left R-module.
18/02/20
Proof. Let M and N be R-modules, and α : M → N an injective homomorphism. By identifying M with
Im α, we may assume that M ≤ N , and α is the inclusion map. Since Q/Z is injective as an abelian group,
any Z-module homomorphism M → S extends to a homomorphism N → S. Define
Θ : HomZ (N, Q/Z) −→ HomZ (M, Q/Z)
,
f 7−→ f |M
the restriction to M . We see that Θ is surjective. Similarly, we can define

Θ0 : HomR (N, S) −→ HomR (M, S)
.
f 7−→ f |M
Then Θ0 is an abelian group homomorphism. But we know there is a naturally defined isomorphism between
HomR (M, S) and HomZ (M, Q/Z). So we get
Θ
HomZ (N, Q/Z) HomZ (M, Q/Z)
Ψ ∼ ∼ Ψ .
HomR (N, S) HomR (M, S)

Θ0
It is easy to see that this diagram commutes. It follows that Θ0 is surjective. So any R-module homomorphism
M → S extends to a homomorphism N → S. Hence S is injective.
4 Exercise
25
Proposition 5.9. Let M be a left R-module, and m ∈ M \ {0}. Then there exists f : M → S such that
f (m) 6= 0.
Proof. We know there is an abelian group homomorphism g : M → Q/Z such that g (m) 6= 0. Now
Ψ (g) ∈ HomR (M, S), and Ψ (g) (m) (1) = g (m) 6= 0 for 1 ∈ R, so Ψ (g) (m) is not the zero map.
Proposition 5.10. Let M be a left R-module. There exists an injective R-module I such that M is iso-
morphic to a submodule of I. Equivalently, there exists an injection M → I.
Q
Proof. Same as abelian groups. Let T = M \ {0}. Then I = t∈T S is injective. Let ft be a homomorphism
M → S such that ft (t) 6= 0. Then
f : M −→ I
m 7−→ (ft (m))t∈T
is injective, and a homomorphism.
Proposition 5.11. Every R-module admits an injective resolution.
Thus there exist injective I0 , I1 , I2 , . . . such that
0 → M → I0 → I1 → I2 → . . .
is exact.
Proof. Let M be an R-module. Then M injects into some injective module I0 . Let C0 = I0 / Im (M → I0 ).
Then C0 injects into some injective I1 . This induces a map I0 → I1 whose kernel is Im (M → I0 ). Further
terms in the sequence are constructed in an identical manner.
5.3 Uniqueness of projective resolutions

Proposition 5.12. Let M and N be R-modules, and φ : M → N . Let (Pi ) be a projective resolution for
M , and (Qi ) a projective resolution for N .
1. There exist R-module homomorphisms fi : Pi → Qi such that the diagram
d2 d1 d0 p
... P2 P1 P0 M 0
f2 f1 f0 φ
... Q2 Q1 Q0 q N 0
d02 d01 d00
commutes.
2. Let gi : Pi → Qi be such that the diagram
d2 d1 d0 p
... P2 P1 P0 M 0
g2 f2 g1 f1 g0 f0 φ
... Q2 Q1 Q0 q N 0
d02 d01 d00
commutes. Then there exist homomorphisms si : Pi → Qi+1 such that

(
si−1 ◦ di−1 + d0i ◦ si i > 0
gi − fi = ,
d00 ◦ s0 i=0
so
d2 d1 d0 p
... P2 P1 P0 M 0
s1 s0
s2 φ .
... Q2 Q1 Q0 q N 0
d02 d01 d00
26
Proof.
1. The map q : Q0 → N is surjective. There is a map p : P0 → N , given by composing P0 → M with φ.
Since P0 is projective there exists f0 : P0 → Q0 such that p = q ◦ f0 . Suppose the maps f0 , . . . , ft−1
have been constructed, so
dt dt−1 dt−2
... Pt Pt−1 Pt−2 ...
ft ft−1 ft−2 .
... Qt Qt−1 Qt−2 ...
d0t d0t−1 d0t−2
Observe that d0t−2 ◦ ft−1 ◦ dt−1 = ft−2 ◦ dt−2 ◦ dt−1 , since the existing squares of the diagram commute.
But dt−2 ◦ dt−1 = 0. So d0t−2 ◦ ft−1 ◦ dt−1 = 0, so Im (ft−1 ◦ dt−1 ) ≤ Ker d0t−2 = Im d0t−1 . Now the map
d0t−1 : Qt → Im d0t−1 is obviously surjective, and Pt is projective. So there is a map ft : Pt → Qt such
that ft−1 ◦ dt−1 = d0t−1 ◦ ft . Now inductively, maps fi exist for all i.
2. We want si such that gi − fi = d0i ◦ si + si−1 ◦ di−1 . Let hi = gi − fi . We see that the diagram
d2 d1 d0 p
... P2 P1 P0 M 0
h2 h1 h0 0
... Q2 Q1 Q0 q N 0
d02 d01 d00
commutes, since we want hi ◦ di = d0i ◦ hi+1 , but we have

hi ◦ di = gi ◦ di − fi ◦ di = d0i ◦ gi+1 − d0i ◦ fi+1 = d0i ◦ h0i+1 ,
so we are fine.
Base case. Let x ∈ P0 . Then (q ◦ h0 ) (x) = (0 ◦ p) (x) = 0 so Im h0 ≤ Ker q = Im d00 . We have a
surjective map d00 : Q1 → Im d00 , and a map h0 : P0 → Im d00 . Since P0 is projective,
there exists s0 : P0 → Q1 such that h0 = d00 ◦ s0 .
Inductive step. Suppose we have maps s0 , . . . , st−1 , with si : Pi → Qi+1 , and hi = d0i ◦ si + si−1 ◦ di−1
for i = 1, . . . , t − 1, so
dt+1 dt dt−1 dt−2 dt−3
... Pt+1 Pt Pt−1 Pt−2 ...
st st−1 st−2
ht+1 ht ht−1 ht−2 .
... Qt+1 Qt Qt−1 Qt−2 ...
d0t+1 d0t d0t−1 d0t−2 d0t−3
Look at ht − st−1 ◦ dt−1 . We want to show that the image of this map is contained in
Im d0t = Ker d0t−1 . So check
d0t−1 ◦ (ht − st−1 ◦ dt−1 ) = d0t−1 ◦ ht − d0t−1 ◦ st−1 ◦ dt−1
= ht−1 ◦ dt−1 − (ht−1 − st−2 ◦ dt−2 ) ◦ dt−1
= ht−1 ◦ dt−1 − ht−1 ◦ dt−1 + st−2 ◦ dt−2 ◦ dt−1 .
Now dt−2 ◦ dt−1 = 0, so we have d0t−1 ◦ (ht − st−1 ◦ dt−1 ) = 0. So ht − st−1 ◦ dt−1 ∈
Ker d0t−1 . Now we have the situation
Pt
st
ht −st−1 ◦dt−1 ,
Qt+1 Im d0t
d0t
and since Pt is projective, there exists st such that d0t ◦ st = ht − st−1 ◦ dt−1 , so
ht = d0t ◦ st + st−1 ◦ dt−1 as required.
27
5.4 Uniqueness of injective resolutions

The following is the equivalent result for injectives.
Proposition 5.13. Let M and N be R-modules, and φ : M → N a homomorphism. Let (It ) be an injective
resolution for M , and (Jt ) another injective resolution for N . Then
• there exist maps fi : Ii → Ji such that the diagram
i d0 d1 d2
0 M I0 I1 I2 ...
φ f0 f1 f2
0 N J0 J1 J2 ...
j d00 d01 d02
commutes, and
• if (gi ) is another set of maps gi : Ii → Ji such that the diagram
i d0 d1 d2
0 M I0 I1 I2 ...
ψ f0 g0 f1 g1 f2 g2
0 N J0 J1 J2 ...
j d00 d01 d02
commutes, then there exist maps si : Ii+1 → Ji such that

(
si ◦ di + d0i−1 ◦ si−1 i>0
gi − fi = ,
s0 ◦ d0 i=0
so
i d0 d1 d2
0 M I0 I1 I2 ...
s0 s1 s2
ψ .
0 N J0 J1 J2 ...
j d00 d01 d02
Proof. Very similar to Proposition 5.12.

Lecture 19
Lecture 19 is a problems class. Friday
21/02/20
28
M4P63 Algebra IV 6 Chain and cochain complexes
6 Chain and cochain complexes

6.1 Chain complexes
Lecture 20
Definition 6.1. A chain complex is a series A∗ = (Ai ), with maps dA
i = di = d : Ai+1 → Ai such that Monday
2
d = 0, that is di+1 ◦ di = 0, or Im di+1 ≤ Ker di . 24/02/20
Definition 6.2. A cochain complex is a series A∗ = Ai with maps dA
i i+1
i = di = d : A → A such that
2
d = 0, or Im di ≤ Ker di+1 .
Let A∗ and B∗ be chain complexes. Let f = (fi ) be a family of R-module homomorphisms fi : Ai → Bi .
Say that f is a map of chain complexes if f ◦ d = d ◦ f , that is fi ◦ dA B
i = di ◦ fi+1 . So
dn+1 dn dn−1 dn−2

... An+1 An An−1 ...
fn+1 fn fn−1
... dn+1
Bn+1 dn
Bn dn−1
An−1 dn−2
...
commutes. Say that f has property P if all fi have property P, where P is injective, surjective, etc. A
sequence
f g
A∗ −→ B∗ −→ C∗
is exact at B∗ if
fn gn
An −→ Bn −→ Cn
is exact at Bn for all n. A sequence of chain complexes is exact if it is exact everywhere. An exact
sequence
0 → A∗ → B ∗ → C ∗ → 0
is a short exact sequence of chain complexes.
6.2 Homology groups

Definition 6.3. Let A∗ be a chain complex. The n-th homology group of A∗ is Ker dn−1 / Im dn . We
write Hn (A∗ ). Also write H∗ (A∗ ) = (Hn (A∗ )).
Definition 6.4. Let A∗ be a cochain complex. The n-th cohomology group of A∗ is Ker dn / Im dn−1 .
We write Hn (A∗ ), and H∗ (A∗ ) = (Hn (A∗ )).
Example. Let Ai = Z3 for all i, and let d (a, b, c) = (0, 0, a). Certainly d2 = 0, so this is a chain complex.
Then
Ker d = {(0, b, c)} = 0 ⊕ Z2 , Im d = {(0, 0, a)} = 02 ⊕ Z.
Now
Ker dn−1 / Im dn = (0, b, 0) + 02 ⊕ Z .

Proposition 6.5. A map of chain complexes f : A∗ → B∗ induces a map on the homology,

f∗ : H∗ (A∗ ) → H∗ (B∗ ) ,
given by
f∗i Hi (A∗ ) −→ Hi (B∗ )
:
.
x + Im dAi 7−→ fi (x) + Im dB i

Proof. Let x ∈ Ker dA A B B
i−1 . Then fi−1 ◦ di−1 (x) = 0, so di−1 ◦ fi (x) = 0. Hence fi (x) ∈ Ker di−1 . So fi
certainly induces a map fi : Ker dA → Ker dB B A
i−1 / Im di . So there exists y ∈ Ai+1 with di (y) = x. Now
A
B
i−1 B A
fi (x) = fi ◦ di (y) = di ◦ fi+1 (y) ∈ Im di , so fi (x) = 0. Hence Im di ≤ Ker fi so fi induces a map
Ker dA A B B
i−1 / Im di = Hi (A∗ ) → Ker di−1 / Im di = Hi (B∗ ) .
29
M4P63 Algebra IV 6 Chain and cochain complexes
Let A∗ and B∗ be chain complexes, and let f and g be maps between them. We say that f and g are equal
up to homotopy if there exist maps si : Ai → Bi+1 such that
gi − fi = si−1 ◦ dA B
i−1 + di ◦ si .
Proposition 6.6. If f, g : A∗ → B∗ are equal up to homotopy, then f∗ = g∗ , so f and g induce the same
map on homology.
5
Proof. Exercise.
6.3 The long exact sequence in homology

f g
0 → A∗ −
→ B∗ −
→ C∗ → 0
be a short exact sequence. This induces a long exact sequence
· · · → Hn+1 (A∗ ) → Hn+1 (B∗ ) → Hn+1 (C∗ ) → Hn (A∗ ) → Hn (B∗ ) → Hn (C∗ ) → . . . .
Proof. We have a commuting diagram with exact rows

fn+1 gn+1
0 An+1 Bn+1 Cn+1 0
dA
n dB
n dC
n
.
0 An fn
Bn gn Cn 0
Notice Im dn ≤ Ker dn−1 , so we can change this to

fn+1 gn+1
0 An+1 Bn+1 Cn+1 0
dA
n dB
n dC
n
.
0 Ker dA
n−1 fn
Ker dB
n−1 gn Ker dC
n−1
Now Im dn+1 ≤ Ker dn , so the maps An+1 → Ker dn+1 induce maps An+1 / Im dn+1 → Ker dn−1 . So we get
a diagram
fn+1 gn+1
An+1 / Im dA
n+1 Bn+1 / Im dB
n+1 Cn+1 / Im dC
n+1 0
dA
n dB
n dC
n
.
0 Ker dA
n−1 fn
Ker dB
n−1 gn Ker dC
n−1
We are now in the position to apply the snake lemma, so
Ker dA B C A B C
n → Ker dn → Ker dn → Coker dn → Coker dn → Coker dn
is an exact sequence. Then

A A A A
Ker dA
n = Ker dn / Im dn+1 = Hn+1 (A∗ ) , Coker dA
n = Ker dn−1 / Im dn = Hn (A∗ ) .
Similarly for B∗ and C∗ . So we have an exact sequence
Hn+1 (A∗ ) → Hn+1 (B∗ ) → Hn+1 (C∗ ) → Hn (A∗ ) → Hn (B∗ ) → Hn (C∗ ) .
Since consecutive values of i give a sequence overlapping in three terms we can glue them together, to give
the long exact sequence in the proposition.
5 Exercise
30
M4P63 Algebra IV 7 Derived functors
7 Derived functors
7.1 Covariant and contravariant functors
Lecture 21
The following are two variations. Tuesday
25/02/20
Definition 7.1. A covariant functor F from the category of left or right R-modules to the category
of abelian groups is a map from R-modules to abelian groups such that if φ : M → N is an R-module
homomorphism then there exists an abelian group homomorphism
F (φ) : F (M ) → F (N ) ,
which respects identity maps, so F (idM ) = idF (M ) , and respects composition, so
F (φ1 ◦ φ2 ) = F (φ1 ) ◦ F (φ2 ) .
The map F on homomorphisms is additive if F (φ1 + φ2 ) = F (φ1 ) + F (φ2 ). If
0→A→B→C→0
be a short exact sequence, then F is right exact if
F (A) → F (B) → F (C) → 0
is exact, left exact if
0 → F (A) → F (B) → F (C)
is exact. Then F is exact if both left and right exact.
Definition 7.2. A contravariant functor F from the category of left or right R-modules to the category
of abelian groups is a map from R-modules to abelian groups such that if φ : M → N is an R-module
homomorphism then there exists an abelian group homomorphism
F (φ) : F (N ) → F (M ) ,
which respects identity maps, so F (idM ) = idF (M ) , and respects composition, so
F (φ1 ◦ φ2 ) = F (φ2 ) ◦ F (φ1 ) .
Similarly, if
0→A→B→C→0
be a short exact sequence, then F is right exact if
F (C) → F (B) → F (A) → 0
is exact, and left exact if
0 → F (C) → F (B) → F (A)
is exact.
Example. Some functors we have seen. Fix a left R-module M .
• F (A) = HomR (M, A), where
F (φ) : F (A) = HomR (M, A) −→ F (B) = HomR (M, B)
, φ : A → B,
f 7−→ φ◦f
is covariant, left exact, and exact if and only if M is projective.
• F (A) = HomR (A, M ), where
F (φ) : F (B) = HomR (B, M ) −→ F (A) = HomR (A, M )
, φ : A → B,
f 7−→ f ◦φ
is contravariant, left exact, and exact if and only if M is injective.
• For a right R-module A, F (A) = A ⊗R M is covariant, right exact, and exact if and only if M is flat.
31
7.2 Left derived functors

Let F be the functor F (A) = A ⊗R M , where M is a fixed R-module. Let P∗ → A be a projective resolution
for A. So P∗ = (Pi )i≥0 for projective Pi , and
d
2 1 0 d d φ
. . . −→ P2 −→ P1 −→ P0 −
→A→0
is exact. Consider the sequence

· · · → P2 → P1 → P0 → 0.
This is no longer exact, but it is a chain complex. And if we apply F , we get a chain complex F (P∗ ),
· · · → F (P2 ) → F (P1 ) → F (P0 ) → 0.
Define left derived functors

Ln F (A) = Hn (F (P∗ )) .
Theorem 7.3.
1. Ln F (A) does not depend on the choice of resolution P∗ .
2. Ln F is an additive functor from right R-modules to abelian groups.
3. L0 F (A) = F (A).
Lecture 22
Friday
Proof.
28/02/20
1. Let P∗ → A and Q∗ → A be projective resolutions. Then there exist maps of chain complexes
f : P∗ → Q∗ and g : Q∗ → P∗ . So g ◦ f : P∗ → P∗ , so
d2 d1 d0
... P2 P1 P0 A 0
g2 ◦f2 g1 ◦f1 g0 ◦f0 id ,
... d2
P2 d1
P1 d0
P0 A 0
and g ◦ f is equal to id up to homotopy. Apply F to everything. Since F is right exact,

F (d2 ) F (d1 ) F (d0 )
... F (P2 ) F (P1 ) F (P0 ) F (A) 0
F (g2 )◦F (f2 ) F (g1 )◦F (f1 ) F (g0 )◦F (f0 ) id .
... F (P2 ) F (P1 ) F (P0 ) F (A) 0

F (d2 ) F (d1 ) F (d0 )
The diagram remains commutative, since F preserves composition. Now
gi ◦ fi − id = si−1 ◦ di−1 + di ◦ si ,
for suitable maps si . Then
F (gi ) ◦ F (fi ) − id = F (si−1 ) ◦ F (di−1 ) + F (di ) ◦ F (si ) .
So F (gi )◦F (fi ) is id up to homotopy. Hence F (g)◦F (f ) induces the identity on homology H∗ (F (P∗ )).
Also F (f ) ◦ F (g) induces the identity on H∗ (F (Q∗ )). Now we have
F (fi ) : Hi (F (P∗ )) → Hi (F (Q∗ )) , F (gi ) : Hi (F (Q∗ )) → Hi (F (P∗ )) ,
and F (fi ) ◦ F (gi ) = id and F (gi ) ◦ F (fi ) = id, so F (fi ) and F (gi ) are isomorphisms. So
Hn (F (P∗ )) ∼
= Hn (F (Q∗ )) ,
as required. This argument tells us nothing about H0 (F (P∗ )).
32
2. Let φ : A → B. Let P∗ → A and Q∗ → B be projective resolutions. Then there exists f : P∗ → Q∗

such that
P∗ A 0
f φ
Q∗ B 0
commutes. Then F is covariant and right exact. So
F (P∗ ) F (A) 0
F (f ) F (φ)
F (Q∗ ) F (B) 0
is commutative, where F (f ) = (F (fi )). If g : P∗ → Q∗ is such that
P∗ A 0
g φ
Q∗ B 0
commutes, then f and g, the induced maps on homology, are equal. So there exists a map F (fi ) :
Li F (A) → Li F (B), and is independent of the choice of f . So we can write Ln F (φ) = F (fi ). Then
Ln F preserves identity and compositions and is additive, since F is an additive functor. 6
3. We have a short exact sequence
⊂ φ
0 → Im d0 −
→ P0 −
→ A → 0.
Since F is right exact, we get an exact sequence
F (Im d0 ) → F (P0 ) → F (A) → 0.
Now d0 : P1 → Im d0 is surjective, and F preserves surjectivity. So F (d0 ) : F (P1 ) → F (Im d0 ) is

surjective. So
F (P1 ) → F (P0 ) → F (A) → 0
is exact. So, setting P−1 = 0, we get L0 F (P∗ ) = F (P0 ) / Im F (d0 ) = F (A).
6 Exercise
33
7.3 The long exact sequence of left derived functors

Proposition 7.4 (Horseshoe lemma). Suppose
0→A→B→C→0
is an exact sequence of R-modules. Suppose P∗ → A and R∗ → C are projective resolutions. Define

Qi = Pi ⊕ Ri . Then there exist maps Qi+1 → Qi and Q0 → B such that Q∗ → B is a projective resolution,
and such that
0 0 0 0
... P2 P1 P0 A 0
ι ι ι
... Q2 Q1 Q0 B 0
π π π
... R2 R1 R0 C 0
0 0 0 0
commutes, where if x ∈ Pi and y ∈ Ri then ι (x) = (x, 0) and π (x, y) = y.
Note. Qi is a direct sum of projectives, so is itself projective.
Proof. We have the setup
0 0
φ
P0 A 0
ι f
Q0 B .
π g
R0 ψ
C 0
0 0
Since B → C is surjective, and R0 is projective, there exists h : R0 → B such that g ◦ h = ψ. Now define
χ : Q0 −→ B
, x ∈ P0 , y ∈ R0 .
(x, y) 7−→ (f ◦ φ) (x) + h (y)
This construction guarantees that the squares are commutative. It is easy to see that χ is surjective, so
0 0
φ
P0 A 0
ι f
χ
Q0 B 0 .
h g
π
R0 ψ
C 0
0 0
Now we have a short exact sequence

ι π
0 → Ker φ →
− Ker χ −
→ Ker ψ → 0,
by the snake lemma. So now we can iterate, replacing A, B, C with these kernels, to construct a map
Q1 → Q0 , and so on.
34
Proposition 7.5. Let F be an additive functor, and let A and B be R-modules. There is a canonical
isomorphism F (A) ⊕ F (B) → F (A ⊕ B).
Proof. Let M = A ⊕ B. Consider functions
p1 : M −→ M p2 : M −→ M
, .
(a, b) 7−→ (a, 0) (a, b) 7−→ (0, b)
Then p2i = pi , p1 ◦ p2 = p2 ◦ p1 = 0, and p1 + p2 = idM . If q1 and q2 are maps on a module M satisfying
these relations, then M = q1 (M ) ⊕ q2 (M ).
Lecture 23
Proposition 7.6. Let Monday
0→A→B→C→0 02/03/20
be a short exact sequence of right R-modules. This gives rise to a long exact sequence
· · · → Ln F (A) → Ln F (B) → Ln F (C) → · · · → L0 F (A) → L0 F (B) → L0 F (C) → 0.
Proof. Let P∗ → A be a projective resolution and R∗ → C be a projective resolution. By the horseshoe
lemma, there exists a projective resolution Q∗ → B such that
0 → P∗ → Q∗ → R∗ → 0
is a split short exact sequence of chain complexes, that is Qi = Pi ⊕ Ri . Since Qi = Pi ⊕ Ri , and since F is
an additive functor, we have F (Qi ) = F (Pi ) ⊕ F (Ri ). So
0 → F (P∗ ) → F (Q∗ ) → F (R∗ ) → 0
is a short exact sequence. Therefore we get a long exact sequence on homology,
· · · → Hn (F (P∗ )) → Hn (F (Q∗ )) → Hn (F (R∗ )) → . . . .
Since Hn (F (P∗ )) = Ln F (A) this gives the long sequence that we need. Since L0 F (A) = F (A), and F is
right exact, the sequence terminates
L0 F (A) → L0 F (B) → L0 F (C) → 0,
as required.
7.4 General derived functors

Proposition 7.7.
• Let F be any covariant, right exact, additive functor from left or right R-modules to abelian groups.
Then the left derived functors Ln F can be defined in just the same way as we did for the case F (A) =
A ⊗R M . All of the results we have proved follow in the more general case, by the same arguments.
• If F is a covariant, left exact, additive functor from R-modules to abelian groups, then we can define
right derived functors Ri F in a similar manner. Instead of working with a projective resolution,
we use an injective resolution,
0 → A → I0 → I1 → I2 → . . . .
By similar arguments, we show that Ri F (A) is independent of the choice of injective resolution. All
of the results we proved for left derived functors have natural analogies for right derived functors. The
argument requires a version of the horseshoe lemma for injective resolutions, which is exactly what one
might expect.
• We can even construct derived functors for contravariant functors. If F is contravariant and right
exact, so
0 → A → I0 → I1 → I2 → . . .
is exact implies that
· · · → F (I2 ) → F (I1 ) → F (I0 ) → F (A) → 0
is exact, we get a left derived functor, which is defined using an injective resolution. If F is contrav-
ariant and left exact, we get a right derived functor, which is defined using a projective resolution.
35
M4P63 Algebra IV 8 Tor and Ext
8 Tor and Ext

8.1 Balancing theorems
Definition 8.1. Let F be the functor F (A) = A ⊗R B. Then F is covariant, left exact, and additive. So
Ln F exists. Define
TorR
i (A, B) = Li F (A) .
Fact. Let F 0 be the functor F 0 (B) = A ⊗R B. Then F 0 is covariant, right exact, and additive. So Ln F 0
exists. We have
Li F 0 (B) ∼
= Li F (A) = TorR
i (A, B) .
Definition 8.2. Let F be the functor F (B) = HomR (A, B). Then F is covariant, left exact, and additive,
so Rn F exists. Define
ExtiR (A, B) = Ri F (B) .
Fact. Let F 0 be the functor F 0 (A) = HomR (A, B). Then F 0 is contravariant, left exact, and additive, so
Rn F 0 exists. We have
Ri F 0 (A) ∼
= Ri F (B) = ExtiR (A, B) .
The two facts above are the balancing theorems for Tor and Ext. Their proof is beyond the scope of the
course.
8.2 Tor
The following is an observation. Suppose A is projective. Then a projective resolution for A is
id
0 → · · · → 0 → A −→ A → 0.
So Li F (A) = 0 for i ≥ 1, and L0 F (A) = F (A), for F possessing left derived functors. Similarly, if A is
injective, then an injective resolution is
id
0 → A −→ A → 0 → · · · → 0,
so Ri F (A) = 0 for i ≥ 1, and R0 F (A) = F (A), for F possessing right derived functors. In fact the property
TorR
i (A, B) = 0 for all i ≥ 0 characterises flat modules, so either A or B is flat.
Proposition 8.3. Let F (A) = A ⊗R B. Then Li F (A) = TorR

i (A, B) = 0 for all i ≥ 1 and for all A if and
only if B is flat.
Similarly if F 0 (B) = A ⊗R B, then Li F 0 (B) = 0 for all i ≥ 1 and for all B if and only if A is flat. Lecture 24
Tuesday
Proof. 03/03/20
⇐= If B is flat then F (A) = A ⊗R B is exact, so
0→L→M →N →0
is exact implies that

0 → F (L) → F (M ) → F (N ) → 0
is exact, or F maps kernels to kernels and cokernels to cokernels. Let P∗ → A be a projective resolution.
Then
· · · → P2 → P1 → P 0 → 0
is exact everywhere except P0 . So
· · · → F (P2 ) → F (P1 ) → F (P0 ) → 0
is exact everywhere except F (P0 ). So Ln F (P∗ ) = 0 for n ≥ 1. But Ln F (P∗ ) = TorR

n (A, B).
36
=⇒ Conversely, suppose TorR

i (A, B) = 0 for all A. Let
0→L→M →N →0
be exact. This gives a long exact sequence of homology groups
... L1 F (L) L1 F (M ) L1 F (N ) L ⊗R B M ⊗R B N ⊗R B 0
.
∼
=
TorR
1 (N, B)
Since TorR
1 (N, B) = 0, we get a short exact sequence
0 → L ⊗R B → M ⊗R B → N ⊗R B → 0.
So F (A) = A ⊗R B is left exact, and so B is flat.
Proposition 8.4. Let A and B be abelian groups. Then TorZn (A, B) = 0 for n > 1.
Proof. A is a quotient of some free module K, say
f
K−
→ A → 0.
Now Ker f ≤ K, and since Z is a PID, Ker f is free, since it is a submodule of a free module. So
· · · → 0 → Ker f → K → A → 0
is a projective resolution for A. Since all of the modules above P1 in the resolution are zero, clearly Hn (P∗ ) =
0 for n > 1.
Fact. TorZ1 (A, Q/Z) = T (A) = {a ∈ A | a has finite order}. Proof omitted.
8.3 Ext
Proposition 8.5. Let A and B be abelian groups. Then ExtnZ (A, B) = 0 for n > 1.
Proof. Problem sheet question.
More generally, Ext1R (A, C) tells us about extensions of C by A, that is B such that
0 → A → B → C → 0.
Let B1 and B2 be two extensions of C by A. Write B1 ∼ B2 if there exists a map of extensions f : B1 → B2
such that
α
B1 β
1 1
0 A f C 0
α2 β2
B2
commutes.
Proposition 8.6. Any such f is an isomorphism.
Proof.
• f is surjective. Suppose y ∈ B2 . Then β2 (y) ∈ C, and β1 is surjective, so β2 (y) = β1 (x) for some
x ∈ B1 . Now f (x)−y ∈ Ker β2 = Im α2 , so f (x)−y = α2 (a) for some a ∈ A. So f (x)−y = (f ◦ α1 ) (a),
and so y = f (x) − (f ◦ α1 ) (a) = f (x − α1 (a)).
• f is injective. Suppose f (x) = f (y) for x, y ∈ B1 . Then f (x − y) = 0, so (β2 ◦ f ) (x − y) = 0,
so β1 (x − y) = 0. So x − y ∈ Ker β1 = Im α1 , so x − y = α1 (a) for some a ∈ A. Now α2 (a) =
(f ◦ α1 ) (a) = f (x − y) = 0. But α2 is injective, so a = 0, so x − y = 0.
37
Hence the relation ∼ is an equivalence relation. Write EC (A) for the set of ∼-equivalence classes. We will
put an abelian group structure on EC (A). Let B1 and B2 be extensions, so
1α β1 2 α β2
0 → A −→ B1 −→ C → 0, 0 → A −→ B2 −→ C → 0.
Define maps α∗ and β ∗ by
α∗ : A −→ B1 ⊕ B2 β∗ : B1 ⊕ B2 −→ C
, .
a 7−→ (α1 (a) , −α2 (a)) (b1 , b2 ) 7−→ β1 (b1 ) − β2 (b2 )
Now β ∗ ◦ α∗ = 0. So
α∗ β∗
0 → A −−→ B1 ⊕ B2 −→ C → 0
is a chain complex. Define H = H (B1 , B2 ), the Baer sum of [B1 ] and [B2 ], to be the homology group at
B1 ⊕ B2 , that is H = Ker β ∗ / Im α∗ . More explicitly,
H = {(b1 , b2 ) ∈ B1 ⊕ B2 | β1 (b1 ) = β2 (b2 )} / {(α1 (a) , −α2 (a)) | a ∈ A} .
Clearly H is an R-module. Now define maps
α : A −→ H β : H −→ C
, .
a 7−→ (α1 (a) , 0) + Im α∗ (b1 , b2 ) + Im α∗ 7−→ β1 (b1 )
Lecture 25
∗
Note. (b1 , b2 ) ∈ Ker β , so β1 (b1 ) = β2 (b2 ). Also (α1 (a) , 0) = (0, α2 (a))+(α1 (a) , −α2 (a)), so (α1 (a) , 0)+ Friday
Im α∗ = (0, α2 (a)) + Im α∗ . 06/03/20
Proposition 8.7.
α β
0→A−
→H−
→C→0
is a short exact sequence.
Proof.
• First check that β is well-defined. If (b1 , b2 ) ∈ (b01 , b02 ) + Im α∗ then (b1 , b2 ) = (b01 , b02 ) + (α1 (a) , −α2 (a))
for some a ∈ A. So
β ((b1 , b2 ) + Im α∗ ) = β1 (b1 ) = β1 (b1 − α1 (a)) = β ((b01 , b02 ) + Im α∗ ) ,
since β1 ◦ α1 = 0.
• Next check α is injective. Suppose α (a) = (0, 0) + Im α∗ . Then (α1 (a) , 0) = α∗ (a0 ) for some a0 . So
(α1 (a) , 0) = (α1 (a0 ) , −α2 (a0 )). Since α1 and α2 are injective, a = a0 = 0.
• Next, show β is surjective. Take c ∈ C. Then c = β1 (b1 ) for some b1 ∈ B1 . Since β2 is surjective, there
exists b2 ∈ B2 with β2 (b2 ) = β1 (b1 ) = c. Now (b1 , b2 ) ∈ Ker β ∗ , and β ((b1 , b2 ) + Im α∗ ) = β1 (b1 ) = c.
• Finally, show that
0→A→H→C→0
is exact, that is Ker β = Im α. It is clear that Im α ≤ Ker β, since β1 ◦ α1 = 0. For the reverse
containment, let (b1 , b2 ) + Im α∗ ∈ Ker β. So (b1 , b2 ) ∈ Ker β ∗ , so β1 (b1 ) = β2 (b2 ). And β1 (b1 ) = 0,
so β2 (b2 ) = 0 as well. But Ker βi = Im αi for i = 1, 2, so there exist a1 , a2 ∈ A with α1 (a1 ) = b1 and
α2 (a2 ) = b2 . Now
(b1 , b2 ) = (α1 (a1 ) , α2 (a2 )) = (α1 (a1 + a2 ) , 0) + (−α1 (a2 ) , α2 (a2 ))

∈ (α1 (a1 + a2 ) , 0) + Im α∗ = α (a1 + a2 ) ∈ Im α.
We have shown that H is an extension of C by A.
38
Proposition 8.8. If B1 ∼ B10 and B2 ∼ B20 then H (B1 , B2 ) ∼ H (B10 , B20 ), where B ∼ B 0 if there exists a
map of extensions f : B → B 0 such that
B
0 A f C 0
B0
commutes.
Proof. Suppose f1 : B1 → B10 and f2 : B2 → B20 are maps of extensions. Then there exists a map of chain
complexes
B1 ⊕ B2
0 A (f1 ,f2 ) C 0 .
B10 ⊕ B20
This induces a map on homology, f : H (B1 , B2 ) → H (B10 , B20 ). It is easy to check
H (B1 , B2 )
0 A f C 0
H (B10 , B20 )
commutes, so H (B1 , B2 ) ∼ H (B10 , B20 ).

Write [B] for the equivalence class of B. If H = H (B1 , B2 ), write [H] = [B1 ] + [B2 ], or [H] = [B1 ] +B [B2 ].
Proposition 8.9. + gives an abelian group operation on the set EC (A) of equivalence classes of extensions.
Proof.
• Check + is commutative. This follows easily from the facts that
α (a) = (α1 (0) , 0) + Im α∗ = (0, α2 (a)) + Im α∗ , β ((b1 , b2 ) + Im α∗ ) = β1 (b1 ) = β2 (b2 ) .
7
• Associativity is an exercise.
• The identity is [A ⊕ C], the split extension. Let
α β α0 β0
0→A−
→A⊕C −
→ C → 0, 0 → A −→ B −→ C → 0.
There is a map π : A ⊕ C → A such that π ◦ α = idA . Consider a map

f : H (B, A ⊕ C) −→ B
, β 0 (b1 ) = β (b2 ) , b2 = (a, c) ∈ A ⊕ C.
(b1 , b2 ) + Im α∗ 7−→ b1 + α0 (a)
It is easy to check this gives a map of extensions.
• Inverses. Suppose
α β
0→A−
→B−
→ C → 0.
Then the inverse of [B] is given by the extension
α −β
0→A−
→ B −−→ C → 0.
7 Exercise
39

M4P63

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

M4P63

Uploaded by

Copyright:

Available Formats

M4P63 Algebra IV

Lectured by Dr John Britnell

2 Projective and injective modules 6

3 Hom and tensor products 10

4 Modules over a PID 18

5 Projective and injective resolutions 23

6 Chain and cochain complexes 29

8 Tor and Ext 36

1 Modules over a ring

1.1 Modules over rings

1.2 Homomorphisms and submodules

1.3 Direct products and direct sums

with coordinate-wise addition and R-multiplication, so

Definition 1.8. The direct sum is

1.4 Exact sequences

The first isomorphism theorem implies that B/ Im α ∼

Proposition 1.11. The short exact sequence

2 Projective and injective modules

Such a g is called a lift of f .

Such an X is the pullback of β and f , and there is a short exact sequence

2.2 Free modules

It is easy to check that this respects + and R-multiplication.

2.3 Injective modules

We know that ( ◦ f ) (a) = (δ ◦ α) (a) for all a ∈ A. So

3 Hom and tensor products

• We can also define

Proposition 3.5. Suppose

g ∈ Ker α∗ ⇐⇒ α∗ (g) = 0 ⇐⇒ g ◦ α = 0 ⇐⇒ g (α (A)) = 0 ⇐⇒ Im α ≤ Ker g ⇐⇒ Ker β ≤ Ker g,

g ∈ Ker α∗ ⇐⇒ ∃f ∈ HomR (C, M ) , β ∗ (f ) = g ⇐⇒ g ∈ Im β ∗ .

Hence Ker α∗ = Im β ∗ . So the sequence is exact at HomR (B, M ).

• The inclusion α : Z → Q is a Z-module homomorphism. Let M = Z. Then we get α∗ : HomZ (Q, Z) →

Note there exists α−1 : Im α → A. If Im g ≤ Im α, then α−1 ◦ g : M → A. If f = α−1 ◦ g, then α ◦ f = g, so

Hence Ker β∗ = Im α∗ . So the sequence is exact at HomR (M, B).

3.2 The snake lemma

where the rows are exact. Then we obtain an exact sequence

3.3 Tensor products

f (a1 + a2 , b) = f (a1 , b) + f (a2 , b) , a1 , a2 ∈ A, b ∈ B,

f (a, b1 + b2 ) = f (a, b1 ) + f (a, b2 ) , a ∈ A, b1 , b2 ∈ B,

(a1 + a2 ) ⊗ b − a1 ⊗ b − a2 ⊗ b, a ⊗ (b1 + b2 ) − a ⊗ b1 − a ⊗ b2 , ar ⊗ b − a ⊗ rb.

2. Let M be a right R-module. Then M ⊗R R ∼

Corollary 3.14. Let A be an abelian group. Then

By the universal property, there is a unique homomorphism

3.4 Flat modules

4 Modules over a PID

4.1 Free and projective modules

4.2 Injective and divisible modules

4.3 Flat and torsion-free modules

We now build up to the following.

Proof. We assume that F = Rn . We do this by induction on n.

Inductive step. Assume U ≤ F = Rn . Write Rn = R ⊕ Rn−1 . So we have a short exact sequence

We also have a short exact sequence

is split, since Rn = R ⊕ Rn−1 , and so

Now f is injective by Lemma 4.14, and h is injective by the inductive hypothesis.

Ker f → Ker g → Ker h

We have shown that if R is a PID, and if M is torsion-free, then M is flat.

4.4 Modules over PIDs

free ⇐⇒ projective =⇒ flat ⇐⇒ torsion-free, injective ⇐⇒ divisible.

5 Projective and injective resolutions

5.1 Existence of projective resolutions

5.2 Existence of injective resolutions

Ψ (p) (m) (r1 + r2 ) = p ((r1 + r2 ) m) = p (r1 m + r2 m)

• Ψ (p) is an R-module homomorphism, since

Ψ (p) (m1 + m2 ) (r) = p (r (m1 + m2 )) = p (rm1 + rm2 ) = p (rm1 ) + p (rm2 )

We know that ( ◦ f ) (a) = (δ ◦ α) (a) for all a ∈ A. So